An Elementary Textbook for Students of

Mathematics, Engineering, and the Sciences

..
\ I ....

\ I
,
'

,.• .." "

.. ".
"
• "
'\

,• ---
.
. .
i

I \
• ,I

•
.-

I \

Morris Tenenbaum
Harry Pollard
ORDINARY DIFFERENTIAL
EQUATIONS
An Elementary Textbook for
Students of Mathematics, Engineering,
and the Sciences

Morris Tenenbaum
CorneU University

Harry Pollard
Purdue University

DOVER PUBLICATIONS, INC., NEW YORK

 Copyright © 1963 by Morris renenbaum and Harry Pollard
All rights reserved under Pan American and International Copyright Con-
ventions.

This Dover edition, first published in 1985, is an unabridged and corrected

republication of the work first published by Harper & Row, Publishers, Inc.,
New York, in 1963.
Manufactured in the United States of America
Dover Publications, Inc., 31 East 2nd Street, Mineola, N. Y. 11501

Library ojCongress Cataloging in Publication Data

Tenenbaum, Morris.
Ordinary differential equations.
Reprint. Originally published: New York : Harper & Row, 1963.
Bibliography: p.
Includes index.
1. Differential equations. I. Pollard, Harry, 1919- II. Title.
QA372. T4 1985 515.3'5 85-12983
ISBN 0-486-64940-7
 Contents

PREFACE FOR THE TEACHER XV

PREFACE FOR THE STUDENT xvii

1. BASIC CONCEPTS I

Lesson I. How Differential Equations Originate. I

Lesson 2. The Meaning of the Terms Set and Function. Im-

plicit Functions. Elementary Functions. 5
A. The Meaning of the Term Set. 5 B. The Meaning of the Term
Function oj One Independent Variable. 6 C. Function of Two Inde-
pendent Variables. 11 D. Implicit Function. 14- E. The Elemen-
tary Functions. 11

Lesson 3. The Differential Equation. 20

A. Definition of an Ordinary Differential Equation. Order of a Differ-
ential Equation. 20 B. Solution of a Differential Equation. Explicit
Solution. 21 C. Implicit Solution of a Differential Equation. 24-

Lesson 4. The General Solution of a Differential Equation. 28

A. Multiplicity of Solutions of a Differential Equation. 28 B. Method
of Finding a Differential Equation if Its n-Parameter Family of Solu-
tions Is Known. 31 C. General Solution. Particular Solution. Initial
Conditions. 33

Lesson 5. Direction Field. 38

A. Construction of a Direction Field. The Isoclines of a Direction
Field. 38 B. The Ordinary and Singular Points of the First Order
Equation (5.11). 41

2. SPECIAL TYPES OF DIFFERENTIAL EQUATIONS

OF THE FIRST ORDER 46

Lesson 6. Meaning of the Differential of a Function.

Separable Differential Equations. 47
A. Differential of a Function of One Independent Variable. 4-1 B. Dif-
ferential of a Function of Two Independent Variables. 50 C. Differ-
ential Equations with Separable Variables. 51
v
vi CONTENTS

Lesson 7. First Order Differential Equation with

Homogeneous Coefficients. 57
A. Definition of a Homogeneous Function. 57 B. Solution of a Dif-
ferential Equation in Which the Coefficients of dx and dy Are Each
Homogeneous Functions of the Same Order. 58

Lesson 8. Differential Equations with Linear Coefficients. 62

A. A Review of Some Plane Analytic Geometry. 62 B. Solution of
a Differential Equation in Which the Coefficients of dx and dy are
Linear, Nonhomogeneous, and When Equated to Zero Represent Non-
parallel Lines. 63 C. A Second Method of Solving the Differential
Equation (8.2) with Nonhomogeneous Coefficients. 66 D. Solution
of a Differential Equation in Which the Coefficients of dx and dy Define
Parallel or Coincident Lines. 67

Lesson 9. Exact Differential Equations. 70

A. Definition of an Exact Differential and of an Exact Differential
Equation. 72 B. Necessary and Sufficient Condition for Exactness
and Method of Solving an Exact Differential Equation. 73

Lesson 10. Recognizable Exact Differential Equations.

Integrating Factors. 80
A. Recognizable Exact Differential Equations. 80 B. Integrating
Factors. 82 C. Finding an Integrating Factor. 84-

Lesson II. The Linear J)ifferentiaI Equation of the First Order.

Bernoulli Equation. 91
A. Definition of a Linear Differential Equation of the First Order. 91
B. Method of Solution of a Linear Differential Equation of the First
Order. 92 C. Determination of the Integrating Factor ef P(z)dz. 94-
D. Bernoulli Equation. 95

Lesson 12. Miscellaneous Methods of Solving a First Order

Differential Equation. 99
A. Equations Permitting a Choice of Method. 99 B. Solution by
Substitution and Other Means. 101

3. PROBLEMS LEADING TO DIFFERENTIAL EQUATIONS

OF THE FIRST ORDER 107

Lesson 13. Geometric Problems. 107

Lesson 14. Trajectories. 115

A. Isogonal Trajectories. 115 B. Orthogonal Trajectories. 117
C. Orthogonal Trajectory Formula in Polar Coordinates. 118
 CONTENTS vii

Lesson 15. Dilution and Accretion Problems. Interest Prob-

lems. Temperature Problems. Decomposition and
Growth Problems. Second Order Processes. 122
A. Dilution and Accretion Problems. 122 B. Interest Problems. 126
C. Temperature Problems. 129 D. Decomposition and Growth
Problems. 131 E. Second Order Processes. 134

Lesson 16. Motion of a Particle Along a Straight Line-

Vertical, Horizontal, Inclined. 138
A. Vertical Motion. 139 B. Horizontal Motion. 160 C. Inclined
Motion. 164

Lesson 17. Pursuit Curves. Relative Pursuit Curves. 168

A. Pursuit Curves. 168 B. Relative Pursuit Curve. 177

Lesson 17M. Miscellaneous Types of Problems Leading to

Equations of the First Order 183
A. Flow of Water Through an Orifice. 183 B. First Order Linear
Electric Circuit. 184 C. Steady State Flow of Heat. 185 D. Pres-
sure-Atmospheric and Oceanic. 186 E. Rope or Chain Around a
Cylinder. 188 F. Motion of a Complex System. 189 G. Variable
Mass. Rocket Motion. 191 II. Rotation of the Liquid in a Cylin-
der. 193

4. LINEAR DIFFERENTIAL EQUATIONS OF ORDER

GREATER THAN ONE 196

Lesson 18. Complex Numbers and Complex Functions. 197

A. Complex Numbers. 197 B. Algebra of Complex Numbers. 200
C. Exponential, Trigonometric, and Hyperbolic Functions of Complex
Numbers. 201

Lesson 19. Linear Independence of Functions. The Linear

Differen tial Equation of Order n. 205
A. Linear Independence of Functions. 205 B. The Linear Differ-
ential Equation of Order n. 207

Lesson 20. Solution of the Homogeneous Linear Differential

Equation of Order n with Constant Coefficients. 211
A. General Form of Its Solutions. 211 B. Roots of the Characteristic
Equation (20.14) Real and Distinct. 213 C. Roots of Characteristic
Equation (20.14) Real but Some Multiple. 214 D. Some or All Roots
of the Characteristic Equation (20.14) Imaginary. 217
viii CONTENTS

Lesson 21. Solution of the Nonhomogeneous Linear Differential

Equation of Order n with Constant Coefficients. 221
A. Solution by the Method of Undetermined Coefficients. 221 B. So-
lution by the Use of Complex Variables. 230

Lesson 22. Solution of the Nonhomogeneous Linear Differential

Equa tion by the Method of Varia tion of Parameters. 233
A. Introductory Remarks. 233 B. The Method of Variation of
Parameters. 233

Lesson 23. Solution of ,the Linear Differential Equation with

Nonconstant Coefficients. Reduction of Order
Method. 241
A. Introductory Remarks. 241 B. Solution of the Linear Differential
Equation with Nonconstant. Coefficients by t.he Reduction of Order
Method. 242

5. OPERATORS AND LAPLACE TRANSFORMS 250

Lesson 24. Differential and Polynomial Operators. 251

A. Definition of an Operator. Linear Property of Polynomial Opera-
tors. 251 B. Algebraic Properties of Polynomial Operators. 255
C. Exponential Shift Theorem for Polynomial Operators. 260 D. S0-
lution of a Linear Differential Equation with Constant Coefficients by
Means of Polynomial Operators. 262

Lesson 25. Inverse Operators. 268

A. Meaning of an Inverse Operator. 269 B. Solution of (25.1) by
Means of Inverse Operators. 272

Lesson 26. Solution of a Linear Differential Equation by

Means of the Partial Fraction Expansion of Inverse
Opera tors. 283
A. Partial Fraction Expansion Theorem. 283 B. First Method of
Solving a Linear Equation by Means of the Partial Fraction Expansion
of Inverse Operators. 288 C. A Second Method of Solving a Linear
Equation by Means of the Partial Fraction Expansion of Inverse
Operators. 290

Lesson 27. The Laplace Transform. Gamma Function. 292

A. Improper Integral. Definition of a Laplace Transform. 292 B. Prop-
erties of the Laplace Transform. 295 C. Solution of a Linear Equa-
tion with Constant Coefficients by Means of a Laplace Transform. 296
D. Construction of a Table of Laplace Transforms. 302 E. The
Gamma Function. 306
 CONTENTS ix

6. PROBLEMS LEADING TO LINEAR DIFFERENTIAL

EQUATIONS OF ORDER TWO 313

Lesson 28. Undamped Motion. 313

A. Free Undamped Motion. (Simple Harmonic Motion.) 313 B. Defi-
nitions in Connection with Simple Harmonic Motion. 317 C. Exam-
ples of Particles Executing Simple Harmonic Motion. Harmonic Oscil-
lators. 323 D. Forced Undamped Motion. 338

Lesson 29. Damped Motion. 347

A. Free Damped Motion. (Damped Harmonic Motion.) 347 B. Forced
Motion with Damping. 359

Lesson 30. Electric Circuits. Analog Computation. 369

A. Simple Electric Circuit. 369 B. Analog Computation. 375

Lesson 30M. Miscellaneous Types of Problems Leading to

Linear Equations of the Second Order 380
A. Problems Involving a Centrifugal Force. 380 B. Rolling Bodies.
381 C. Twisting Bodies. 383 D. Bending of Beams. 383

7. SYSTEMS OF DIFFERENTIAL EQUATIONS.

LINEARIZATION OF FIRST ORDER SYSTEMS 393

Lesson 31. Solution of a System of Differential Equations. 393

A. Meaning of a Solution of a System of Differential Equations. 393
B. Definition and Solution of a System of First Order Equations. 394-
C. Definition and Solution of a System of Linear First Order Equa-
tions. 396 D. Solution of a System of Linear Equations with Con-
stant Coefficients by the Use of Operators. Nondegenerate Case. 398
E. An Equivalent Triangular System. 405 F. Degenerate Case.
II (D)g2(D) - gl (D)/2(D) == o. 413 G. Systems of Three Linear
Equations. 415 H. Solution of a System of Linear Differential Equa-
tions with Constant Coefficients by Means of Laplace Transforms. 418

Lesson 32. Linearization of First Order'Systems. 424

8. PROBLEMS GIVING RISE TO SYSTEMS OF EQUATIONS.

SPECIAL TYPES OF SECOND ORDER LINEAR AND NON-
LINEAR EQUATIONS SOLVABLE BY REDUCING TO SYSTEMS 440

Lesson 33. Mechanical, Biological, Electrical Problems Giving

Rise to Systems of Equations. 440
A. A Mechanical Problem-Coupled Springs. 440 B. A Biological
Problem. 447 C. An Electrical Problem. More Complex Circuits. 451
X CONTENTS

Lesson 34. Plane Motions Giving Rise to Systems of Equations. 459

A. Derivation of Velocity and Acceleration Formulas. 459 B. The
Plane Motion of a Projectile. 463 C. Definition of a Central Force.
Properties of the Motion of a Particle Subject to a Central Force. 470
D. Definitions of Force Field, Potential, Conservative Field. Conser-
vation of Energy in a Conservative Field. 473 E. Path of a Particle
in Motion Subject to a Central Force Whose Magnitude Is Proportional
to Its Distance from a Fixed Point O. 476 F. Path of a Particle in
Motion Subject to a Central Force Whose Magnitude Is Inversely Pro-
portional to the Square of Its Distance from a Fixed Point O. 481
G. Planetary Motion. 491 H. Kepler's (1571-1630) Laws of Plane-
tary Motion. Proof of Newton's Inverse Square Law. 492

Lesson 35. Special Types of Second Order Linear and Nonlinear

Differential Equations Solvable by Reduction to a
System of Two First Order Equations. 500
A. Solution of a Second Order Nonlinear Differential Equation in
Which y' and the Independent Variable x Are Absent. 500 B. Solu-
tion of a Second Order Nonlinear Differential Equation in Which the
Dependent Variable y Is Absent. 502 C. Solution of a Second Order
Nonlinear Equation in Which the Independent Variable x Is Absent. 503

Lesson 36. Problems Giving Rise to Special Types of Second

Order Nonlinear Equations. 506
A. The Suspension Cable. 506 B. A Special Central Force Prob-
lem. 521 C. A Pursuit froblem Leading to a Second Order Nonlinear
Differential Equation. 523 D. Geometric Problems. 528

9. SERIES METHODS 531

Lesson 37. Power Series Solutions of Linear Differential

Equations. 531
A. Review of Taylor Series and Related Matters. 531 B. Solution
of Linear Differential Equations by Series Methods. 537

Lesson 38. Series Solution of y' == f(x,y). 548

Lesson 39. Series Solution of a Nonlinear Differential Equation

of Order Greater Than One and of a System of First
Order Differential Equations. 555
A. Series Solution of a System of First Order Differential Equations. 555
B. Series Solution of a System of Linear First Order Equations. 559
C. Series Solution of a Nonlinear Differential Equation of Order Greater
Than One. 562
 CONTENTS xi

Lesson 40. Ordinary Points and Singularities of a Linear

Differential Equation. Method of Frobenius. 570
A. Ordinary Points and Singularities of a Linear Differential Equa-
tion. 510 B. Solution of a Homogeneous Linear Differential Equation
About a Regular Singularity. Method of Frobenius. 512

Lesson 41. The Legendre Differential Equation. Legendre

Functions. Legendre Polynomials Pk(X). Properties
of Legendre Polynomials Pk(X). 591
A. The Legendre Differential Equation. 591 B. Comments on the
Solution (41.18) of the Legendre Equation (41.1). Legendre Functions.
Legendre Polynomials PIc(X). 593 C. Properties of Legendre Poly-
nomials PIc(X). 598

Lesson 42. The Bessel Differential Equation. Bessel Function

of the First Kind Jk(X). Differential Equations
Leading to a Bessel Equation. Properties of Jk(X). 609
A. The Bessel Differential Equation. 609 B. Bessel Functions of the
First Kind JIc(X). 611 C. Differential Equations Which Lead to a
Bessel Equation. 615 D. Properties of Bessel Functions of the First
Kind J Ic(X). 619

Lesson 43. The Laguerre Differential Equation. Laguerre

Polynomials Lk(X). Properties of Lk(X). 624
A. The Laguerre Differential Equation and Its Solution. 624- B. The
Laguerre Polynomial LIc(x). 625 C. Some Properties of Laguerre
Polynomials LIc(x). 627

10. NUMERICAL METHODS 631

Lesson 44. Starting Method. Polygonal Approximation. 632

Lesson 45. An Improvement of the Polygonal Starting Method. 641

Lesson 46. Starting Method-Taylor Series. 645

A. Numerical Solution of y' = J(x,y) by Direct Substitution in a Taylor
Series. 61,.6 B. Numerical Solution of y' = J(x,y) by the "Creeping
Up" Process. 61,.6

Lesson 47. Starting Method-Runge-Kutta Formulas. 653

Lesson 48. Finite Differences. Interpolation. 659

A. Finite Differences. 659 B. Polynomial Interpolation. 661
xii CONTENTS

Lesson 49. Newton's Interpolation Formulas.

A. Newton's (Forward) Interpolation Formula. 669 B. Newton's
(Backward) Interpolation Formula. 668 C. The Error in Polyno-
mial Interpolation. 670

Lesson 50. Approximation Formulas Including Simpson's and

Weddle's Rule. 672

Lesson 51. Milne's Method of Finding an Approximate

Numerical Solution of y' = f(x,y). 684

Lesson 52. General Comments. Selecting h. Reducing h.

Summary and an Example. 690
A. Comment on Errors. 690 B. Choosing the Size of h. 691 C. Re-
ducing and Increasing h. 692 D. Summary and an Illustrative Exam-
pIe. 69J,.

Lesson 53. Numerical Methods Applied to a System of Two

First Order Equations. 702

Lesson 54. Numerical Solution of a Second Order Differential

Equation. 707

Lesson 55. Perturbation Method. First Order Equation. 713

Lesson 56. Perturbation Method. Second Order Equation. 715

11. EXISTENCE AND UNIQUENESS THEOREM FOR THE FIRST

ORDER DIFFERENTIAL EQUATION y' = f(x,y). PICARD'S
METHOD. ENVELOPES. CLAIRAUT EQUATION. 719

Lesson 57. Picard's Method of Successive Approximations. 720

Lesson 58. An Existence and Uniqueness Theorem for the First

Order Differential Equation y' = f(x,y) Satisfying
y(xo) = Yo- 728
A. Convergence and Uniform Convergence of a Sequence of Functions.
Definition of a Continuous Function. 728 B. Lipschitz Condition.
Theorems from Analysis. 731 C. Proof of the Existence and Unique-
ness Theorem for the First Order Differential Equation y' = f(x,y). 133

Lesson 59. The Ordinary and Singular Points of a First Order

Differential Equation y' == f(x,y). 744
 CONTENTS xiii

Lesson 60. Envelopes. 747

A. Envelopes of a Family of Curves. 748 B. Envelopes of a 1-Param-
eter Family of Solutions. 754-

Lesson 61. The Clairaut Equation. 757

12. EXISTENCE AND UNIQUENESS THEOREMS FOR A SYSTEM

OF FIRST ORDER DIFFERENTIAL EQUATIONS AND FOR
LINEAR AND NONLINEAR DIFFERENTIAL EQUATIONS OF
ORDER GREATER THAN ONE. WRONSKIANS. 763

Lesson 62. An Existence and Uniqueness Theorem for a System

of n First Order Differential Equations and for a
Nonlinear Differential Equation of Order Greater
Than One. 763
A. The Existence and Uniqueness Theorem for a System of n First
Order Differential Equations. 763 B. Existence and Uniqueness The-
orem for a Nonlinear Differential Equation of Order n. 765 C. Exist-
ence and Uniqueness Theorem for a System of n Linear First Order
Equations. 768

Lesson 63. Determinants. Wronskians. 770

A. A Brief Introduction to the Theory of Determinants. 770
B. Wronskians. 774

Lesson 64. Theorems About Wronskians and the Linear

Independence of a Set of Solutions of a
Homogeneous Linear Differential Equation. 778

Lesson 65. Existence and Uniqueness Theorem for the Linear

Differential Equation of Order n. 783

Bibliography 791

Index 793
 Preface for the Teacher

IN WRITING THIS BOOK, it has been our aim to make it readable for the
student, to include topics of increasing importance (such as transfonns,
numerical analysis, the perturbation concept) and to avoid the errors
traditionally transmitted in an elementary text. In this last connection,
we have abandoned the use of the tenninology "general solution" of a
differential equation unless the solution is in fact general, i.e., unless the
solution actually contains every solution of the differential equation. We
have also avoided the term "singular solution." We have exercised great
care in defining function, differentials and solutions; in particular we have
tried to make it clear that functions have domains.
On the other hand, this accuracy has been secondary to our main pur-
pose: to teach the student how to use differential equations. We hope
and believe that we have not overlooked any of the major applications
which can be made comprehensible at this elementary level. You will
find in this text an extensive list of worked examples and homework
problems with answers.
We acknowledge our indebtedness to the publishers for their coopera-
tion and willingness to let us use new pedagogical devices and to Prof.
C. A. Hutchinson for his thorough editing.
M.T.
H.P.
Ithaca, New York
West Lafayette, Indiana

xv
 Preface for the Student

THIS BOOK HAS BEEN WRITTEN primarily for you, the student. We have
tried to make it easy to read and easy to follow.
We do not wish to imply, however, that you will be able to read this
text as if it were a novel. If you wish to derive any benefit from it, you
must study each page slowly and carefully. You must have pencil and
plenty of paper beside you so that you yourself can reproduce each step
and equation in an argument. When we say '''verify a statement," "make
a substitution," "add two equations," "multiply two factors," etc., you
yourself must actually perform these operations. If you carry out the
explicit and detailed instructions we have given you, we can almost
guarantee that you will, with relative ease, reach the conclusion.
One final suggestion-as you come across formulas, record them and
their equation numbers on a separate sheet of paper for easy reference.
You may also find it advantageous to do the same for Definitions and
Theorems.
M.T.
H.P.
Ithaca, New York
West Lafayette, Indiana

xvii
 Chapter I

Basic Concepts

LESSON 1. How Differential Equations Originate.

We live in a world of interrelated changing entities. The position of

the earth changes with time, the velocity of a falling body changes with
distance, the bending of a beam changes with the weight of the load
placed on it, the area of a circle changes with the size of the radius, the
path of a projectile changes with the velocity and angle at which it is fired.
In the language of mathematics, changing entities are called variables
and the rate of change of one variable with respect to another a deriva-
tive. Equations which express a relationship among these variables and
their derivatives are called differential equations. In both the natural
and social sciences many of the problems with which they are concerned
give rise to such differential equations. But what we are interested in
knowing is not how the variables and their derivatives are related but
only how the variables themselves are related. For example, from certain
facts about the variable position of a particle and its rate of change with
respect to time, we wish to determine how the position of the particle is
related to the time so that we can know where the particle was, is, or will
be at any time t. Differential equations thus originate whenever a uni-
versal law is expressed by means of variables and their derivatives. A
course in differential equations is then concerned with the problem of
detennining a relationship among the variables from the infonnation
given to us about themselves and their derivatives.
We shall use an actual historical event to illustrate how a differential
equation arose, how a relationship was then established between the two
variables involved, and finally hpw from the relationship, the answer to a
very interesting problem was determined. In the year 1940, a group of
boys was hiking in the vicinity of a town in France named Lascaux.
They suddenly became aware that their dog had disappeared. In the
ensuing search he was found in a deep hole from which he was unable to
climb out. When one of the boys lowered himself into the hole to help
extricate the dog, he made a startling discovery. The hole was once a
1
2 BASIC CONCEPTS Chapter I

part of the roof of an ancient cave that had become covered with brush.
On the walls of the cave there were marvellous paintings of stags, wild
horses, cattle, and of a fierce-looking black beast which resembled our
bull. * This accidental discovery, as you may guess, created a sensation.
In addition to the wall paintings and other articles of archaeological
interest, there were also found the charcoal remains of a fire. The problem
we wish to solve is the following: determine from the charcoal remains
how long ago the cave dwellers lived.
It is well known that charcoal is burnt wood and that with time cer-
tain changes take place in all dead organic matter. It is also known that
all living organisms contain two isotopes of carbon, namely C 12 and C 14 •
The first element is stable; the second is radioactive. Furthermore the
ratio of the amounts of each present in any macroscopic piece of living
organism remains constant. However from the moment the organism
dies, the C 14 is lost because of radiation, is no longer replaced. Hence
the amount of the unstable C 14 present in a dead organism, as well as its
ratio to the stable C 12 , changes with time. The changing entities in this
problem are therefore the element C 14 and time. If the law which tells
us how one of these changing entities is related to the other cannot be
expressed without involving their derivative, then a differential equation
will result.
Let t represent the elapsed time since the tree from which the charcoal
came, died, and let x represent the amount of C 14 present in the dead
tree at any time t. Then the instantaneous rate at which the element C 14
decomposes is expressed in mathematical symbols as
dx
(1.1) dt ·
We now make the assumption that this rate of decomposition of C 14
varies as the first power of x (remember x is the amount of C 14 present
at any time t). Then the equation which expresses this assumption is
dx
(1.11) - = -kx
dt '
where k > 0 is a proportionality constant, and the negative sign is used
to indicate that x, the quantity of C 14 present, is decreasing. Equation
(1.11) is a differential equation. It states that the instantaneous rate of
decomposition of C 14 is k times the amount of C 14 present at a moment of
time. For example, if k = 0.01 and t is measured in years, then when
x = 200 units at a moment in time, (1.11) tells us that the rate of decom-
position of C 14 at that moment is 1/100 of 200 or at the rate of 2 units per
·You can see some of these pictures in Primitive Art by Erwin O. Christensen,
Viking Press, 1955, and in The Picture History oj Painting by H. W. and D. J. Jansen,
Harry N. Abrams, 1957.
Lesson 1 How DIFFERENTIAL EQUATIONS ORIGINATE 3

year. If, at another moment of time, x = 50 units, then (1.11) tells us that
the rate of decomposition of C 14 at that moment is 1/100 of 50 or at the
rate of i unit per year.
Our next task is to try to determine from (1.11) a law that will express
the relationship between the variable x (which, remember, is the amount
of C 14 present at any time t) and the time t. To do this, we multiply
(1.11) by dt/x and obtain
(1.12) -dx = -kdt.
x
Integration of (1.12) gives

(1.13) log x = -kt + c,

where c is an arbitrary constant. By the definition of the logarithm, we
can write (1.13) as
(1.14)

where we have replaced the constant eC by a new constant A.

Although (1.14) is an equation which expresses the relationship between
the variable x and the variable t, it will not. give us the answer we seek
until we know the values of A and k. For this purpose, we fall back on
other available information which as yet we have not used. Since time is
being measured from the moment the tree died, i.e., t = 0 at death, we
learn from (1.14) by substituting t = 0 in it, that x = A. Hence we now
know, since x is the amount of C 14 present at any time t, that A units of
C 14 were present when the tree, from which the charcoal came, died.
From the chemist we learn that approximately 99.876 percent* of C 14
present at death will remain in dead wood after 10 years and that the
assumption made after (1.1) is correct. Mathematically this means that
when t = 10, x = 0.99876A. Substituting these values of x and t in
(1.14), we obtain
(1.15) 0.99876A = Ae- 10k , 0.99876 = e- 10k •
We can now find the value of k in either or' two ways. There are tables
which tell us for what value of -10k, e- 10k = 0.99876. Division of this
value by -10 will then give us the value of k. Or if we take the natural
logarithm of both sides of (1.15) there results

(1.2) log 0.99876 = -10k.

*There is some difference among chemists in regard to this figure. The one used
above is based on a half-life of C14 of 5600 years, i.e., half of C14 present at death will
decompose in 5600 years. It is an approximate average of 5100 years, the lowest half-
life figure, and 6200 years, the largest half-life figure.
" BASIC CONCEPTS Chapter 1

From a table of natural logarithms, we find

(1.21) -0.00124 = -10k, k = 0.000124

approximately. Equation (1.14) now becomes

(1.22) x = Ae-O.000124',

where A is the amount of C 14 present at the moment the tree died.

Equation (1.22) expresses the relationship between the variable quan-
tity x and the variable time t. We are therefore at last in a position to
answer the original question,: How long ago did the cave dwellers live?
By a chemical analysis of the charcoal, the chemist was able to determine
the ratio of the amounts of C 14 to C 12 present at the time of the dis-
covery of the cave. A comparison of this ratio with the fixed ratio of
these two carbons in living trees disclosed that 85.5 percent of the amount
of C 14 present at death had decomposed. Hence 0.145A units of C 14
I

remained. Substituting this value for x in (1.22), we obtain

(1.23) 0.145A = Ae-O.000124t
0.145 = e-O.000124t
log 0.145 = -0.OOO124t
-1.9310 = -0.OOO124t
t = 15573.
Hence the cave dwellers lived approximately 15,500 years ago.
Comment 1.3. Differential equation (1.11) originated from the as-
sumption that the rate of decomposition of C 14 varied as the first power
of the amount of C 14 present at any time t. The resulting relationship
between the variables was then verified by independent experiment.
Assumptions of this kind are continually being made by scientists. From
the assumption a differential equation originates. From the differential
equation a relationship between variables is determined, usually in the
form of an equation. From the equation certain predictions can be made.
Experiments must then be devised to test these predictions. If the pre-
dictions are validated, we accept the equation as expressing a true law.
It has happened in the history of science, because experiments performed
were not sensitive enough, that laws which were considered as valid for
many years were found to be invalid when new and more refined experi-
ments were devised. A classical example is the laws of Newton. These
were accepted as valid for a few hundred years. As long as the experi-
ments concerned bodies which were macroscopic and speeds which were
reasonable, the laws were valid. If the bodies were of the size of atoms or
the speeds near that of light, then new assumptions had to be made, new
Lesson 2A THE MEANING OF THE TERM Set 5

equations born, new predictions foretold, and new experiments devised to

test the validity of these predictions.
Comment 1.4. The method we have described for determining the
age of an organic archaeological remain is known as the carbon-14 test. *

EXERCISE I
1. The radium in a piece of lead decomposes at a rate which is proporiional to
the amount present. If 10 percent of the radium decomposes in 200 years,
\vhat percent of the original amount of radium will be present in a piece of
lead after 1000 years?
2. Assume that the half life of the radium in a piece of lead is 1600 years. How
much radium will be lost in 100 years?
3. The following item appeared in a ne,wspaper. "The expedition used the
carbon-14 test to measure the amount of radioactivity still present in the
organic material found in the ruins, thereby determining that a town existed
there as long ago as 7000 B.C." Using the half-life figure of C14 as given in
the text, determine the approximate percentage of C14 still present in the
organic material at the time of the discovery.

ANSWERS I
1. 59.05 percent. 2. 4.2 percent. 3. Between 32 percent and 33 percent.

LESSON 2. The Meaning of the Terms Set and Function.

Implicit Functions. Elementary Functions.

Before we can hope to solve problems in differential equations, we must

first learn certain rules, methods and laws which must be observed. In
the lessons that follow, we shall therefore concentrate on explaining the
meaning of certain terms which we shall use and on devising methods by
which certain types of differential equations can be solved. We shall
then apply these methods to solving a wide variety of problems of which
the one in Lesson 1 was an example.
We begin our study of differential equations by clarifying for you two
of the basic notions underlying the calculus and ones ,vhich we shall use
repeatedly. These are the notions of set and Junction.

LESSON 2A. The Meaning of the Term Set. Each of you is familiar
with the word collection. Some of you in fact may have or may have had
collections-such as collections of stamps, of sea shells, of coins, of butter-
flies. In mathematics we call a collection of objects a set, and the indi-

*Dr. Willard F. Libby was awarded the 1960 Nobel Physics Prize for developing this
method of ascertaining the age of ancient objects. His C14 half-life figure is 5600 years,
the same as the one we used. According to Dr. Libby, the measurable age span by
this test is from 1000 to 30,000 years.
6 BASIC CONCEPTS Chapter 1

vidual members of the set elements. A set therefore may be described

by specifying what property an object must have in order to belong to it
or by giving a list of the elements of the set.
Examples of Sets. 1. The collection of positive integers less than 10
is a set. Its elements are 1, 2, 3, 4, 5, 6, 7, 8, 9.
2. The collection of individuals whose surnames are Smith is a set.
3. The collection of all negative integers is a set. Its elements are
· · · , -4, -3, -2, -1.
Since to each point on a line, there corresponds one and only one rea]
number, called the coordinate of the point, and to each real number
there corresponds one and orlly one point on the line, we frequently refer
to a point on a line by its corresponding number and vice versa.
Definition 2.1. The set oj all numbers between any two points on a
line is called an interval and is usually denoted by the letter I.
If the two points on a line are designated by a and b, then the notation

(2.11) I: a < x <b

will mean the set of all real numbers x (or of all real values of x) which lie
between the points a and b, but not including a and b. For convenience,
we shall frequently omit the I and write only

(2.111) a < x < b

to represent this set of numbers. Similarly,

(2.12) I: - 00 < x < 00 will mean the set of all real values of x.
I: a < x < b will mean the set of all real values of x
between a and b, including the two end points.
I: a < x < b will mean the set of all real values of x
between a and b, including a but not b.
I: -1 < x < 3, x = 10, will mean the set of all num-
bers between -1 and 3 plus the number 10.
I: x > 0 will mean the set of all positive real values of x
plus zero.
I: x = a will mean the set consisting of the single num-
ber a.

LESSON 2B. The Meaning of the Term Function of One Inde-

pendent Variable. If two variables are connected in some way so that
the value of one is uniquely determined when a value is given to the other,
we say that one is a function of the other. (This concept will be given
a more precise meaning in Definitions 2.3 and 2.31 below.)
Lesson 2B THE MEANING OF THE TERM Function 7

We shall show by examples below that the manner in which the rela-
tionship between the variables is expressed is unimportant. It may be
by an equation, of the kind with which you are familiar, or by other means.
It is only important for the definition of a function that there be this
unambiguous relationship between the variables so that, when a value is
given to one, a corresponding value to the other is thereby uniquely
detennined.
Example 2.2. Let l be the length of the side of a square and A its
area. It is then customary to say that the area A depends on the length
I, so that l is given an independent status and A a dependent one. How-
ever, there is no valid reason why l could not be considered as being
dependent on A. The decision as to which variable in a problem is to be
considered as dependent and which independent lies entirely within the
discretion of the individual. The choice will usually be detennined by
convenience. It is customary to write, whenever it is possible to do so,
first the dependent variable, then an equals sign, then the independent
variable in a manner which expresses mathematically the relationship be-
tween the two variables.
If in this example, therefore, we express the relationship between our
two variables A and l by writing
(a)

we thereby give to A a dependent status and to l an independent one.

Equation (a) now defines A as a function of l since for each l, it detennines
A uniquely. The relationship between the two variables, expressed mathe-
matically by equation (a), is, however, not rigidly correct. It says that
for each value of the length l, A, the area, is the square of l. But what if
we let l = -3? The square of -3 is 9; yet no area exists if the side of a
square has length less than zero. Hence we must place a restriction on l
and say that (a) defines the area A as a function of the length l only for
a 8et of positive values of l and for l = o. We must therefore write
(b) A = l2 , l > o.
Example 2.21. The relationship between two variables x and y is the
following. If x is between 0 and 1, y is to equal 2. If x is between 2 and
3, y is to equal VX. The equations which express the relationship between
the two variables are, with the end points of the interval included,
(a) y = 2, 0 < x < 1,
y = v'X, 2 < x < 3.
These two equations now define y as a function of x. For each value of x
in the specified intervals, a value of y is detennined uniquely. The graph
8 BASIC CONCEPTS Chapter 1

of this function is shown in Fig. 2.211. Note that these equations do not
define y as a function of x for values of x outside the two stated intervals.

1 -

(0,0) 1 2 3 x
Figure 2.211

The reason is obvious. We have not been told what this relationship is.
For values of x therefore, equal to say i or -lor 4, etc., we say that y
is undefined or that the function is undefined.
Example 2.22. In Fig. 2.221, we have shown the temperature T of a
body, recorded by an automatic device, in a period of 24 consecutive
hours. The horizon tal axis represents the time in hours; the vertical axis
the temperature T at any time t > o. Even though we cannot express
the relationship between the variables t and T by an equation, there can
be no doubt that a precise, unique relationship between the two variables
T

3 6 12 15 18 t

Figure 2.221

exists. For each value of the time t, the graph will give a unique value of
the temperature T. Hence the graph in this case defines Tas a function of t.
The main feature we wished to emphasize in the above differing exam-
ples was that for the definition of a function it was not essential to be able
to set up the relationship between the two variables by a single equation.
(Most of the functions that you have encountered thus far were of this
type.) As we mentioned at the outset, what is essential for the definition
Lesson 2B THE MEANING OF THE TERM Functwn 9

of a function is that the relationship between the two variables be specific

and unambiguous so that for each value taken on by an independent
variable on a specified set, there should correspond one and only one value
of a dependent variable. As you can verify, all the examples we gave
above had this one common important property. We incorporate all the
essential features of a function in the following definition.
Definition 2.3. If to each value of an independent variable x on a
set E (the set must be specified) there corresponds one and only one real
value of the dependent variable y, we say that the dependent variable y
is a function* of the independent variable x on the set E.
It is customary to call the specified set E of values of the independent
variable, the domain of definition of the independent variable and to
call the set of resulting values of the dependent variable, the range of the
dependent variable or the range of the function. Using this terminology,
we may define a function alternately, as follows.
Definition 2.31. A function is a correspondence between a domain
set D and a set R that assigns to each element of D a unique element of R.
Comment 2.32. A function is thus equivalent to a rule which tells us
how to determine the unique element y of the range which is to be assigned
to an element x of the domain. When we say therefore that the formula
(a) y = Vx - 1, x> 1,
defines y as a function of x, we mean that the formula has given us a rule
by which, for each value of the independent variable x in its domain D:
x 1, we can determine the unique assigned value of y in the range R.
(Here R > 0.) The rule is as follows: for each x in D, subtract one and
take the positive square root of the result. For example, to the element
x = 5 of D, the function or rule defined by (a) assigns the unique value
2 of R. Conversely, we can first give the rule, and then express the rule,
if possible, by a formula.
Comment 2.33. We may at times for convenience refer to an equa-
tion or formula as if it were a functioil. For example, we may frequently
refer to the equation
(b)
as a function. What we actually mean is that the equation y = x 2 defines
a function; that is y = x 2 gives us a rule by which to each x in D we can

*In advanced mathematics, the function is called a real function. Since we shall for
the most part be considering only real functions, we shall omit the word real whenever
a real function is meant.
10 BASIC CONCEPTS Chapter 1

assign a yin R. Here the rule is: for each x in D, the number yin R assigned
to it is obtained by squaring x.
In view of Definition 2.3, such frequently encountered formulas as

y = x
2
,

Y = vI - X2,

x 2 + 5x
Y=---
x-3
are meaningless because they do not specify the domain, i.e., the set of
values of x, for which the formulas apply. In practice, however, we inter-
pret such formulas to define functions for all values of x for which they make
sense. The first equation, therefore, defines a function for all values of x,
the second for values of x in the interval -1 S x S 1, and the last for
all values of x except x = 3.
Because of the above comment, the function defined by

y = VI - x2

is not the same as the function defined by

y = VI - X2, 0 < X < 1.

(The first is defined for all x in -1 < x S I; the second only for x in
o < x < 1.)
It is customary and convenient to record the fact that the dependent
variable y is a function of the independent variable x (these are the letters
most commonly used for the dependent and independent variable respec-
tively), by means of the symbolic expression
(2.34) y = f(x).

I t is read as tty equals f of x" or tty is a function of x. "

By Definition 2.3 and (2.34), we could therefore write for the tempera-
ture Example 2.22,
(2.35) T = f(t).

We then say T is a function of t and refer to the graph itself as the defini-
tion or rule which tells us which T to assign to each value of t.
Comment 2.36. We shall at times write an expression in x, say
(a) f(x) = x2 + eX, -00 < x < 00

and refer to f(x) as a function. What we mean is that J(x) , here (x 2 + eX),
gives a rule by which for each x, we can assign a unique value to J(x).
Lesson 2C FUNCTION OF Two INDEPENDENT VARIABLES 11

Definition 2.4. If I(x) is a function of x defined on a set E, then the

symbolf(a), for any a in E, means the unique value assigned to I(x) ob-
tained by substituting a for x.
Example 2.5. If
(a) I(x) = x2 + 2x + 1, 0 < x < 1,
find 1(0),/(1), I(!), 1(2), 1(-1).
Solution. By Definition 2.4 we have
(b) 1(0) = 0 2 + 2·0 + 1 = 1,
1(1) = 12 +2·1+1 = 4,
I(!) = (!)2 + 2 · ! + 1 = 2i,
1(2) is undefined since 2 is not in our set E: 0 < x < 1,
1(-1) is undefined since -1 is not in our set E.

If several functions appear in a single context so that the use of the same
letter for each would be confusing, it is permissible to replace I by other
letters. Those most frequently used are g, h, G, H, F, etc. Similarly, we
may use other letters in place of x and y. Those usually used are the ones
at the end of the alphabet, namely u, v, W, Z, 8, t.

LESSON 2C. Function of Two Independent Variables. In Lesson

2B, we defined a function of one independent variable. In an analogous
manner, we define a function of two independent variables x and y as
follows.
Definition 2.6. If to each element (x,y) of a set E in the plane (the
set must be specified) there corresponds one and only one real value of z,
then z is said to be a function of x and y for the set E. In this event, x,y
are called independent variables and z a dependent variable.
As in the case of one independent variable, the set E is sometimes
called the domain of definition of the function and the set of resulting
values of z, the range of the function. '
In view of Definition 2.6, a formula such as

z = YVI - x2

is meaningless since it does not specify the domain of definition. Here

again as in the case of one variable, we interpret such formulas to define
functions for all values of x and y for which they make sense. In this
example therefore the elements of the domain D are the points (x,y) in
the plane where -1 < x < 1, - 00 < y < 00. The domain, therefore,
for which the formula defines z as a function of x and y consists of all points
12 BASIC CONCEPTS Chapter 1

in the plane between and including the lines x = 1 and x = -1. It is

the shaded area in Fig. 2.61.
z

%=-1

Figure 2.61

Example 2.62. Determine the domain D for which each of the fol-
lowing formulas define z as a function of x and y .
.i

I. z=
VI- y2 .
VI - x 2
2. z = x + y.
3. z = Vx 2 + y2 - 25.
4. z = V-(x 2 + y2).
5. Z.= V-(x 2 + y2 + 1).

Solutions. 1. The elements of the domain D for which the formula

defines z as a function of x and y, consists of those points (x,y) where
-1 < x < 1, .,-1 < y < I. The domain D is the shaded square shown
z z

%=-1

y y

(a) (b)
Figure 2.63

in Fig. 2.63(a). It is bounded by the lines x = ±1 and y = ±1. It

includes the lines y = ±I but not the lines x = ±1.
2. The domain is the entire plane.
Lesson 2C FUNCTION OF Two INDEPENDENT VARIABLES 13

3. The domain consists of those points (x,y) for which x 2 y2 25. +

It is the shaded area in Fig. 2.63(b), i.e., it is the area outside the circle
x2 +y2 = 25 plus the points on its circumference.
4. The domain consists of the single point (0,0).
5. This formula does not define z as a function of x and y. There does
not exist a domain which will determine a value of z.
It is evident from the above examples, that a two-dimensional domain
may cover the whole plane or part of the plane; it may cover the whole
plane with the exception of a hole in its interior; its boundaries may be
circular or straight lines, or it may consist of only a finite number of points.
In short, in contrast to a one-dimensional domain, a two-dimensional
domain may assume a great variety of shapes and figures.
It is customary and convenient to record the fact that z is a function
of x and y by means of the symbolic expression
(2.64) z = f(x,y).
It is read as liZ equals f of x,y" or liZ is a function of x,y. "
Definition 2.65. If f(x,y) is a function of two independent variables
x,y, defined over a domain D, then the symbol f(x,a) for any element
(x,a) in D, means the function of x obtained by replacing y by a.
Example 2.66. If
(a) f(x,y) = X2 +xy2+5y+3, -00 <x < 00, -00 < y < 00,
find f(x,2), f(x,a), f[x,g(x)].
Solution. By Definition 2.65
(b) f(x,2) = x 2 + 4x + 10 + 3 = x 2 + 4x + 13, -00 < x < 00,
f(x,a) = x 2 +a 2x+5a+3, -00 < x < 00.
And, if g(x) is defined for all x,

f[x,g(x)] = x 2 + x[g(X)]2 + 5g(x) + 3, - 00 < x < 00.

Example 2.67. If

(a) f(x,y) = x + y, -1 < x <1, 0 < y < 2,

find f(x, i) , f(x,3).

Solution. By Definition 2.65
(b) f(x,i) = x + i, -1 < x < 1.

But f(x,3) is undefined since the domain of y is the interval °< y < 2.
A special type of set, called a region, is defined as follows.
14 BASIC CONCEPTS Chapter 1

Definition 2.68. A set in the plane is called a region if it satisfies

the following two conditions:
1. Each point of the set is the center of a circle whose entire interior con-
sists of points of the set.
2. Every two points of the set can be joined by a curve which consists
entirely of points of the set.
In Example 2.62, the domain defined in 2 is a region. If the boundary
points are excluded from each set defined in 1 and 3, then each resulting
domain is also a region. Each point of each set satisfies requirement 1,
and every two points of eacp set satisfies requirement 2. On the other
hand, the set consisting of the points on a line is not a region. The set
satisfies requirement 2 but not 1. Also the set consisting of isolated points
is not a region-the points in the set do not satisfy either of the two re-
quirements.
Definition 2.69. A region is said to be bounded if there is a circle
which will enclose it.
Comment 2.691. In a manner analogous to Definition 2.6, we can
define a function of three or more independent variables.

LESSON 2D. Implicit Function. Consider a relationship between

two variables x,y given by the formula
(2.7) x2 + y2 -= o.
25
Does it define a function? If x > 5 or x < -5, then the formula will
not determine a value of y. For example, if x = 7, there is no value of y
which will make the left side of (2.7) equal to zero. (Why?) However, if
x lies between -5 and 5 inclusive, then there is a value of y which will
make the left side of (2.7) equal to zero. To find it, we solve (2.7) for y
and obtain
(2.71) y = ±v'25 - X2, -5 S x S 5.
When the relation between x and y is written in this form, however, we
see that the formula does not define y uniquely for a value of x. Hence,
by our Definition 2.3, it does not define a function. We can correct this
defect by specifying which value of y is to be chosen. For example, we
can choose anyone of the following three formulas to determine y.

(2.72) y = v'25 - X2, -5 < x < 5.

(2.73) y = -v'25 - X2, -5 < x < 5.
(2.74) y = '\1'25 - X2, -5 < x 0;
= -v'25 - X2, o< X < 5.
Lesson 2D IMPLICIT FUNCTION 15

By Definition 2.3, each of these formulas now defines a function. It gives

a rule which assigns a unique y to each x on the specified interval.
N ow consider the formula

(2.75)

which also connects two variables x and y, and ask of it the same ques-
tion. Does it define a function? If it does, then there must be values of x
for which it will determine uniquely values of y. It should be evident to
you that there are no values of y for any x. (Write the equation as
x 2 + y2 = -1.) Hence this equation does not define a function by our
Definition 2.3.
As a final example, we consider the formula

(2.76)

and again ask the question. Does it define a function? And if it does, for
what values of x will it determine uniquely a value of y? The answer to
both questions, unlike the answer to the previous formula (2.7), is not
easy to give. For unlike it, (2.76) cannot be solved easily for y in terms
of x. Hence we must resort to other means. The graph of equation (2.76)
is shown in Fig. 2.77.

Figure 2.77

From the graph, we see that for x < 0 and x > 22/3, Y is uniquely
determined. Hence formula (2.76) does define y as a function of x in
these two intervals, but not in the interval 0 < x < 22/3. It is possible,
however, to make formula (2.76) define y as a function of x for all x if we
16 BASIC CONCEPTS Chapter 1

choose one of the three possible values of y for each x in the interval
o < x < 2 2/ 3 and choose one of the two possible values of y for x = 22/3.
With these restrictions, we then would be able to assert that (2.76) defines
y as a function of x for all x.
Whenever a relationship which exists between two variables x and y is
expressed in the form (2.7) or (2.76), we write it symbolically as

(2.8) f(x,y) = O.

I t is read as "f of x,y equals zero, " or as "a function of x,y equals zero. "
If the relation which defines y as a function of x is expressed in the
form f(x,y) = 0, it is customary to call y an implicit function of x. When
we say therefore that y is an implicit function of x, we mean, as the name
suggests, that the functional relationship between the two variables is not
explicitly visible as when we write y = f(x), but that it nevertheless im-
plicitly exists. That is, there is a function, let us call it g(x), which is
implicitly defined by the relationf(x,y) = 0 and which determines uniquely
a value of y for each x on a set E. Hence:
Definition 2.81. The relation

(2.82) f(x,y) = 0

defines y as an implicit function of x on an interval I: a < x < b, if

there exists a function g(x) defined on I such that

(2.83) ffx,g(x)] = 0

for every x in I.
Example 2.84. Show that

(a) f(x,y) = x 2 + y2 - 25 = 0

defines y as an implicit function of x on the interval I: -5 < x < 5.

Solution. Choose for g(x) anyone of the functions (2.72), (2.73), or
(2.74). If, for example, we choose (2.72), then g(x) = V25 - x 2. It is
defined on I, and by (a) and Definition 2.65,

(b) f[x,g(x)] = x 2 + [V25 - X2]2 - 25 = O.

Hence Definition 2.81 is satisfied.
Example 2.85. Show that
(a) J(x,y) = x 3 + y3 - 3xy = 0

defines y as an implicit function of x for all x.

LeSSOD 2E THE ELEMENTARY FuNCTIONS 17

Solution. Here, as pointed out earlier, it is not easy to solve for y in

terms of x, so that it is not easy to find the required function g(x). How-
ever, if we select from Fig. 2.77 anyone of the graphs shown in Fig. 2.86

Figure 2.86

to represent the function g(x), then f[x,g(x)] = 0 for every x. Hence

Definition 2.81 is satisfied. (Note that g(x) ,may be selected in infinitely
many more ways.)

LESSON 2E. The Elementary Functions. In addition to the terms

function and implicit function, we shall refer at times to a special class of
functions called the elementary lunctions. These are the constants
and the following functions of a variable x:
1. Powers of x: x, x 2 , x 3 , etc.
2. Roots of x: Vx, etc.
3. Exponentials: e •
Z

4. Logarithms: log x.
5. Trigonometric functions: sin x, cos x, tan x, etc.
6. Inverse trigonometric functions: Arc cos x, etc.
7. All functions obtained by replacing x any number of times by any of
the other functions 1 to 6. Examples are: log sin x, sin (sin x), e8in Z,
2
eGZ , etc.
8. All functions obtained by adding, subtracting, multiplying, and divid-
ing any of the above seven types a finite number of times. Examples
e8in ez2+2z+1
+ vax -
Z

are: 2x - log x + x
,(Arc cos X)2
3x
log (log 4x).

In the calculus course, you learned how to differentiate elementary

functions and how to integrate the resulting derivatives. If you have for-
gotten how, it would be an excellent idea at this point to open your cal-
culus book and review this material.
18 BASIC CONCEPTS Chapter 1

EXERCISE 2
1. Describe, in words, each of the following sets:
(a) x < O. (b) x <0. (c) a < x S b. (d) - 00 < x < 5, x == 7.
(e) -3 < x < -2, x > o. (I) V2 < x < 11". (g) 211" S X < 311".
2. Define the area A of a circle as a function of its radius r. Which is the de-
pendent variable and which is the independent variable? Draw a rough
graph which will sho,v how A depends on T, when T is given values between
o and 5.
3. Under certain circumstances the pressure p of a gas and its volume V are
related by the formula p V 3 / 2 = 1. Express each variable as a function of the
other. '
4. Explain the difference between the function
y = vi, 2 < x S 3,
and the function
y = ¥X, x > o.
5. Let
F (x) = 2 if x < 0,
= 7 if 0 < x < 1,
= x3 - 2x if x > 4.
Find (a) F(-l), (b) F(O), (c) F(0.7), (d) F(4), (e) F(8), (f) F(2).
6. If 2
X - 2x+ 1
g(x) = l' x 1,
x-
find (a) g(2), (b) g(-5),· (c) g(l), (d) g(u), (e) g(x 2 ), (f) g(x - 1).
7. Why is the function defined in 6 not the same as the function defined by
g(x) = x-I?
8. Determine the domain D for \vhich each of the following formulas defines
z as a function of x and y.

(a) z =
VI - x2
. (b) z = x - + 2.y
VI - y2
(c) Z = vx2 + (y - 1)2. (d) z = Vx2 + y2 - 9.
(e) z = V -(x + y)2.
I-x
(f) z = V-(x 2 + y2 + 3). (g) z = ·
I-y
9. Which of the domains in problem 8 are regions?
10. If !(x,y) is a function of two independent variables x,y defined over a domain
D, then the symbol !(b,y) for any element (b,y) in D, means the function of
y obtained by replacing x by b; see Definition 2.65. Let
!(x,y) = x 2 + 2xy + log (xy), xy > o.
Find:
(a) !(x,I), !(x,b), !(x,O), !(x,x 2 ). (b) !(I,y), I(a,y), I(O,y), J(x 2 ,y).
(c) !(1,1), !(2,-1), !(a,b), !(u,v).
Lesson 2-Exercise 19

II. Draw the graphs of three different functions defined by

x3 + y3 - 3xy = o.
See Fig. 2.77. Is there one which is continuous for all x?
12. The following is a standard type of exercise in the calculus.
If x 3 + y3 - 3xy = 0 then 3x 2
,
+ 3y2 dx
dy - 3x dy -
dx
3y = o. Therefore
2
dy = y - x, 2 x
dx y2-x y
Explain by the use of Fig. 2.77 what this means geometrically.
13. Explain why the procedure followed in problem 12, applied to the relation
x2 + +
y2 1 = 0 and yielding the result
dy x
dx = - y
is meaningless.
14. Find the function g(x) that is implicitly defined by the relation

v'x2 - y2 + Arc cos y 0

= ' y 0,
15. Explain why
v' x2 - y2 + Arc sin y = 0

does not define y as an implicit function of x:

16. Can you apply the method of implicit differentiation as taught in the calculus
to the function of problem 14, of pr.oblem 15?
17. Define a function of three independent variables Xl, X2, X3; of n independent
variables Xl • • • , x n • Hint. See Definition 2.6.

ANSWERS 2
I. (a) The set of all negative values of x. tb) The set (a) plus zero.
(c) The set of values of X between a and b, including b but not a. (d) The
set of all values of X less than five, plus the number 7. (e) The set of all
values of X between -3 and -2, plus all positive values of x.
2. A = 'l'r2, r o.
3. p = V -3/2, V > 0; V = P -2/3, P > O.
4. The first function is defined only for all X between and including two and
three; the second function is defined for all X greater than zero.
5. (a) 2. (b) 7. (c) 7. (d) Undefined. (e) 496. (f) Undefined.
6. (a) 1. (b) -6. (c) l\1eaningless. 2
(d) (u - 2u 1)/(u - 1). +
u 1. (c) (x4 - 2x2 + 1)/(x - 1), x
2 2 1. (f) (x - 4x + 4)/
2
(x - 2), x 2.
7. Function defined in 6 is meaningless when x = 1.
8. (a) -1 S x <1, -1 < y < 1. (b) Entire plane. (c) Entire plane.
(d) Area outside circle x 2 + y2 = 9 plus points on the circumference of the
circle. (e) Line x + y = o. (f) Nonexistent. (g) y 1.
9. (b), (c), (g). Also (a) and (d) if their boundary points are excluded.
10. (a) x 2 + +
2x + +
log x; x 2 2bx log (bx); undefined; x 2 2x 3 +
log x 3 • +
(b) 1 + +
2y log y; a + +
2 2ay log ay; undefined; x4 + 2x y 2 log (x 2y). +
(c) 3; undefined; a2 + +2ab log (a b) ; u 2 2uv + log (uv). +
20 BASIC CONCEPTS Chapter 1

13. x 2 + +
y2 1 = 0 does not define a function.
14. y = g(x) = x.
15. The first term requires that I x I > I y I ; the second term that Ix I <: I Y I.
16. No, both examples.

LESSON 3. The Differential Equation.

LESSON 3A. Definition of an Ordinary Differential Equation.

Order of a Differential Equation. In the calculus, you studied various
methods by which you could differentiate the elementary functions. For
example, the successive derivatives of y = log x are
, 1 -1 2
(a) y =-, y" = -2 , y'" - -, etc.
x x x3
And if z = x 3 - 3xy +
2y2, its partial derivatives with respect to x and
with respect to yare respectively
oz
(b) ox = 3x -
2
3y,
oz =
- -3x + 4y,
oy

Equations such as (a) and (b) which involve variables and their deriva-
tives are called differential equations. The first involves only one inde-
pendent variable x; the second two independent variables x and y. Equa-
tions of the type (a) are called ordinary differential equations; of the type
(b) partial differential equations. Hence,
Definition 3.1. Let f(x) define a function of x on an interval I:
a < x < b. By an ordinary differential equation we mean an equa-
tion involving x, the function f(x) and one or more of its derivatives.
Note. It is the usual custom in writing differential equations to replace
f(x) by y. Hence the differential equation d Z) + x(f(X)]2 = 0 is usually

written as: + xy2 = 0; the differential equation D%2[j(X)] + xDJ(x) =

eX as Dx2y + xDxY = eX or as y" + xy' = eX.
Examples of ordinary differential equations are:
dy
(3.11) dx +y = o.
(3.12) y' = eX.
2 1
(3.13) d y =
dx 2 1 - x2
(3.14) f'(x) = J"(x).
(3.15) xy' = 2y.
Lesson 3B EXPLICIT SOLUTION OF A DIFFERENTIAL EQUATION 21

(3.16) y"+ (3y,)3 + 2x = 7.

(3.17) (y"') 2 + (y") 4 + y' = x.
(3.18) xy(4) + 2y" + (xy,)5 = x 3.

Note. Since only ordinary differential equations will be considered in

this text, we shall hereafter omit the word ordinary.
Definition 3.2. The order of a differential equation is the order
of the highest derivative involved in the equation.
For the differential equations listed above, verify that (3.11), (3.12),
and (3.15) are of the first order; (3.13), (3.14), and (3.16) are of the second
order; (3.17) is of the third order; (3.18) is of the fourth order.
A WORD OF CAUTION. You might be tempted to assert, if you were not
careful, that
y" - y" +
y' - y = 0

is a second order differential equation because of the presence of y". How-

ever, y" is not really involved in the equation since it is removable. Hence
the equation is of order 1.

LESSON 3B. Solution of a Differential Equation. Explicit Solu-

tion. Consider the algebraic equation

(3.3) x2 - 2x - 3 = O.
When we say x = 3 is a solution of (3.3), we mean that x = 3 satisfies it,
i.e., if x is replaced by 3 in (3.3), the equality will hold. Similarly, when
we say the function f(x) defined by
(3.31) y = f(x) = log x + x, x > 0,
is a solution of
(3.32) x 2 y" + 2xy' + y = log x + 3x + 1, x > 0,

we mean that (3.31) satisfies (3.32), i.e., if in (3.32) we substitute the

function f(x) = log x +
x for y, and the first and second derivatives of
the function for y' and y", respectively, the equality will hold. [Be sure
to verify the assertion that (3.31) does in fact satisfy (3.32).]
We want you to note two things. First in accordance with Definition
2.3, we specified in (3.31) the values of x for which the function is defined.
But even if we had not, the interval x > 0 would have been tacitly
assumed log x is undefined for x < O. Second, we also specified in
(3.32) the interval for which the differential equation makes sense. Since
it too contains the term log x, it too is meaningless when x < o.
22 BASIC CONCEPTS Chapter 1

Definition 3.4. Let y = I(x) define y as a function of x on an interval

I: a < x < b. We say that the function I(x) is an explicit solution or
simply a solution of an ordinary differential equation involving x, I(x),
and its derivatives, if it satisfies the equation for every x in I, i.e., if we
replace y by I(x) , y' by f'(x), y" by f"(x), · · • , y(n) by I(n)(x), the differen-
tial equation reduces to an identity in x. In mathematical symbols the
definition says: the function I(x) is a solution of the differential equation
(3.41) F(x,y,y', · · · ,y(n») = 0,
if
(3.42) F[x,f(x),f'(x), · · • ,f(n)(x)] = 0
for every x in I.
Comment 3.43. We shall frequently use the expression, "solve a dif-
ferential equation," or "find a solution of a differential equation." Both
are to be interpreted to mean, find a function which is a solution of the
differential equation in accordance with Definition 3.4. Analogously when
we refer to a certain equation as the solution of a differential equation, we
mean that the function defined by the equation is the solution. If the equa-
tion does not define a function, then it is not a solution of any differential
equation, even though by following a formal procedure, you can show that
the equation satisfies the differential equation. For example, the equation
y = V-(1 + x 2 ) does not define a function. To say, therefore, that it is
a solution of the differential equation x + yy' = 0 is meaningless even
though the fonnal substitution in it of y = V- (1 x 2 ) and y' = +
-x/V-(1 + x 2 ) yields an identity. (Verify it.)
Example 3.S. Verify that the function defined by
(a) y = x 2, -00 < X < 00,

is a solution of the differential equation

(b) (y")3 + (y')2 - Y - 3x 2 - 8 = O.
Solution. By (a), the function I(x) = x 2 • Therefore f' (x) = 2x,
I"(x) = 2. Substituting these values in (b) for y, y', y", we obtain
(c) 8 + 4x 2 - x2 - 3x 2 - 8 = o.
Since the left side of (c) is zero, (a), by Definition 3.4, is an explicit solu-
tion or simply a solution of (p). Note that (b) is also defined for all x.
Remark. I t is the usual practice, when testing whether the function
defined by the relation y = I(x) on an interval I is a solution of a given
differential equation, to substitute in the given equation the values of y
and its derivatives. In the previous Example 3.5, therefore, if we had
followed this practice, we would have substituted in (b): y = x 2 , y' = 2x,
y" = 2. If an identity resulted, we would then say that (a) is a solution
Lesson 3D EXPLICIT SOLUTION OF A DIFFERENTIAL EQUATION 23

of (b). We shall hereafter, for convenience also follow this practice, but
you should always remember that it is the function/(x) and its derivatives
which must be substituted in the given differential equation for y and its
corresponding derivatives. And if y = f(x) does not define a function
then y or f(x) cannot be the solution of any differential equation.
Example 3.51. Verify that the function defined by
(a) y = log x + c, x > 0
is a solution of
1
(b) y' = -.
x

Solution. Note first that (b) is also defined for all x > O. By (a),
y' = Ilx. Substituting this value of y' in (b) gives an identity. Hence
(a) is a solution of (b) for all x > O.
Example 3.52. Verify that the function defined by
(a) y = tan x - x, X (2n + 1) 2'
'Tr
n = 0, ±I, ±2, · · · ,
is a solution of
(b)

Solution. Here y = tan x - x, y' = sec 2 x - I = tan 2 x. Substi-

tution of these values in (b) for y and y' gives the identity

(c) tan 2 x = (x + tan x - X)2 = tan 2 x.

Hence (a) is a solution of (b) in each of the intervals specified in (a).

Comment 3.521. Note by (b) that the differential equation is defined
for all x. Its solution, however, as given in (a), is not defined for all x.
Hence, the interval, for which the function defined in (a) may be a solu-
tion of (b), is the smaller set of intervals given in (a).
Comment 3.53. It is also pos- y
sible for a function to be defined
over an interval and be the solution
of a differential equation in only
part of this interval. For example
y = Ixl is defined for all x. Its graph
is shown in Fig. 3.54. It has no de- X
rivative when x = O. It satisfies
the differential equation y' = 1 in Figure 3.54
the interval x > 0, and the differ-
ential equation y' = -1 in the interval x < O. But it does not satisfy
any differential equation in an interval which includes the point x = O.
24 BASIC CONCEPTS Chapter 1

LESSON se. Implicit Solution of a Differential Equation. To

test whether an implicit function defined by the relation J(x,y) = 0 is a
solution of a given differential equation, involves a much more compli-
cated procedure than the testing of one explicitly expressed by y = J(x).
The trouble arises because it is usually not easy or possible to solve the
equation f(x,y) = 0 for y in terms of x in order to obtain the needed
function g(x) demanded by Definition 2.81. However, whenever it can be
shown that an implicit function does satisfy a given differential equation
on an interval I: a < x < b, then the relation f(x,y) = 0 is called (by
an unfortunate usage)* an implicit solution of the differential equation.
Definition 3.6. A relation f(x,y) = 0 will be called an implicit 801u-
tion of the differential equation
(3.61) F(x,y,y', • · • , y(n» = 0
on an interval I: a < x < b, if
1. it defines y as an implicit function of x on I, i.e., if there exists a func-
tion g(x) defined on I such that f[x,g(x)] = 0 for every x in I, and if
2. g(x) satisfies (3.61), i.e., if
(3.62) F[x,g(x),g'(x), · · · , g(n)(x)] = 0
for every x in I.
Example 3.63. Test whether
(a) f(x,y)' x2 + y2 - 25 = 0
is an implicit solution of the differential equation
(b) F(x,y,y') = yy' +x = 0
on the interval I: -5 < x < 5.
Solution. We have already shown that (a) defines y as an implicit
function of x on I, if we choose for g(x) anyone of the functions (2.72),
(2.73), (2.74). If we choose (2.72), then
(c) g(x) = V25 - X2, g'(x) = - x ; -5 < x < 5.
v'25 - x2

Substituting in (b), g(x) for y, g'(x) for y', there results

(d) F[x,g(x), g'(x)] = v'25 - x2(- X ) +x = o.

------ v'25 - x 2
• Actually J(x,y) = 0 is an equation and an equation is never a solution of a differential
equation. Only a Junction can be a solution. What we really mean when we say J(x,y) = 0
is a solution of a differential equation is that the Junction g(x) defined by the relation
J(x,y) = 0 is the solution. See Definition 3.6, also Comment 3.43.
Lesson 3C IMPLICIT SOLUTION OF A DIFFERENTIAL EQUATION 25

Since the left side of (d) is zero, the equation is an identity in x. There-
fore, both requirements of Definition 3.6 are satisfied, and (a) is therefore
an implicit solution of (b) on I.
Example 3.64. Test whether

(a) f(x,y) = x 3 + y3 - 3xy = 0, -00 < x < 00,

is an implicit solution of

(b) F(x,y,y') = (y2 - x)y' - y + x2 = 0, -00 < x < 00.

Solution. Unlike the previous example, it is not easy to solve for y

to find the required function g(x). However, we have already shown that
(a) defines y as an implicit function of x, if we choose for g(x) anyone of
the curves in Fig. 2.86. (Be sure to refer to the graphs shown in this
figure.) If we choose the first, then g'(x) does not exist when x = 22/3.
Hence, (a) cannot be a solution of (b) for all x, but we shall show that
(a) is a solution of (b) in any interval which excludes this point x = 2 2/ 3
Since we do not have an explicit expression for g(x), we cannot substitute
in (b), g(x) for y, g'(x) for y' to determine whether F[x,g(x), g'(x)] = o.
What we do is to differentiate (a) implicitly to obtain

(c) 3x 2 + 3y2y' - 3xy' - 3y = 0, (y2 - x)y' - y + x 2 = o.

Since (c) now agrees with (b), we know that the slope of the function
g(x) implicitly defined by (a) and explicitly defined by the graph, satisfies
(b) at every point x in any interval excluding x = 22/3. Hence both re-
quirements of Definition 3.6 are satisfied in any interval which does not
include the point x = 22/3. Therefore the function g(x) defined by the
first graph in Fig. 2.86 is an implicit solution of (a) in any interval not
containing the point x = 22/3.
If we choose for g(x) the second curve in Fig. 2.86, then g'(x) does not
exist when x = 0 and x = 22/3. For this g(x), (a) will be a solution of
(b) in any interval which excludes these two points.
If we choose for g(x) the third curve in Fig. 2.86, then g'(x) does not
exist when x = O. For this g(x), (a) will be a solution of (b) in any inter-
val which excludes this point.
Comment 3.65. The example above demonstrates the possibility of
an implicit function being defined over an interval and being the solution
of a differential equation in only part of the interval.
Comment 3.651. The standard procedure in calculus texts to prove
that (a) is a solution of (b) is the following. Differentiate (a) implicitly.
If it yields (b), then (a) is said to be an implicit solution of (b). If you
operate blindly in this manner, then you are likely to assert that x 2 +
26 BASIC CONCEPTS Chapter 1

y2 = 0 is an implicit solution of x + yy' = 0, since differentiation of the

+
first gives the second. But x 2 y2 = 0 does not define y implicitly as a
function of x on an interval. Only the point (0,0) satisfies this formula.
+ +
To assert, therefore, that x 2 y2 = 0 is an implicit solution of x yy' = 0
because it satisfies the differential equation is meaningless.
Example 3.66. Test whether
(a) xy2 - e-'ll - 1 = 0

is an implicit solution of the differential equation

(b) (xy2 + 2xy - l)y' + y2 = o.
Solution. If we worked blindly and used the method of implicit dif-
ferentiation as taught in the calculus, then from (a), we would obtain by
differentiation
(c) 2xyy' + y2 + e-Yy' = 0, (2xy + e-Y)y' + y2 = o.
Although (c) is not identical with (b), it can be made so if we replace e--'ll
in the second equation of (c) by its value xy2 - 1 as determined from
(a). Working blindly then, we would assert that (a) is an implicit solu-
tion of (b). But is it ? Well, let us see. By Definition 3.6, we must first
show that (a) defines y as an implicit function of x on an interval. If we
write (a) as
(d) y.= ± + e-I/ ,
x
y

x = 2.07 --------------____
x

Y= - 2 . 2 2 . - - - - - - - -

Figure 3.67

we see, since e-Y is always positive, that y is defined only for x > o.
Hence the interval for which (a) may be a solution of (b) must exclude
Lesson 3-Exercise 27

values of x S O. Here again &s in the previous example, we cannot easily

solve for y explicitly in terms of x, so that we must resort to a graph to
determine g(x). I t is given in Fig. 3.67. From the graph, we see that
there are three choices of g(x). If g(x) is the upper branch, then (a) is an
implicit solution of (b) for all x > o. And if g(x) is either of the two lower
branches, one above the line y = -2.22, the other below this line, then
(a) is an implicit solution of (b) for I: x > 2.07 approximately.

EXERCISE 3
1. Determine the order of the following differential equations.
(a) dy + (xy - cos x) dx = o. (b) y" + xy" + 2y(y')3 + xy = O.

(c) - (y",)4 +X = o. (d) efl '" + xy" + y = o.

2. Prove that the functions in the right-hand column below are solutions of the
differential equations in the left-hand columns. (Be sure to state the common
interval for which solution and differential equation make sense.)
(a) y' +y =0 y = e-Z •
(b) y' = eZ y = eZ •
2
d y
(c) - 2 =
dx
1
--;:::====
VI - x2
y = x Arc sin x + VI - x2 •

(d) I'(x) = I"(x) y = eZ + 2.

(e) xy' = 2y y = x2•
(f) (1 +2
x )y' = xy Y = v''--1-+-x-2 •
dr .
(g ) cos 8 d8 - 2r Sln 8 = 0 r = a sec 2 8.
(h) y" - y = 0 y = ae z +
be-z •
(i) I' (x) = !/(x) lex) = 2ez / 3 •
2
(j) xy' +y = y2 y = x+2·
(k) x + yy' = 0 y = V16 - x 2 •
3. Show that the differential equation

+ Iyl + 1 = 0
has no solutions.
4. Determine whether, the equations on the right define implicit functions of x.
For those ,vhich do, determine whether they are implicit solutions of the
differential equations on the left.
(a) y2 - 1 - (2y + xy)y' = 0 Y -
2
1 = (x + 2) 2.
(b) eZ - fI +e fl - z
dy = 0 e2J1 + e2z = 1.
dx
(c) dy =
dx
Y
x
X
2
+ y2 + 1 = O.
28 BASIC CONCEPTS Chapter 1

ANSWERS 3
1. (a) 1. (b) 2. (c) 3. (d) 3.
2. (a) - co <x < co. (b) - aJ < X < aJ. (c) -1 < x < 1.
(d) -aJ <x< co. (e) x o. (f) - a J < x < co.
... 3...
(g) 8 ± 2' ± 2"' ± .... (h) - co < X < co.
(i) -aJ < x < co. (j) x 0, -2. (k) -4 < x < 4.
4. (a) Yes, if one makes y single valued. Implicit solution.
(b) Yes. Implicit solution, x o.
(c) Function undefined.
,
LESSON 4. The General Solution of a Differential Equation.

LESSON 4A. Multiplicity of Solutions of a Differential Equation.

We assume at the outset that you have understood clearly the material of
the previous lesson so that when we say "solve a differential equation" or
"find a solution of a differential equation," or "the solution of a differen-
tial equation is," you will know what is meant (see Comment 3.43). Or
if we omit intervals for which a function or a differential equation is
defined, we expect that you will be able to fill in this omission yourself.
When you studied the theory of integration in the calculus, you solved
some simple differential equations of the form y' = I(x). For example,
you learned that, if
(4.1)

then its solution, obtained by a simple integration, is

(4.11 )
where C can take on any numerical value. And if
(4.12)

then its solution, obtained by integrating (4.12) twice, is

(4.13)

where now Cl and C2 can take on arbitrary values. Finally, if

(4.14)

then its solution, obtained by integrating (4.14) three times, is

(4.15)

where Cl, C2, C3 can take on any numerical values.

Lesson 4A MULTIPLICITY OF SOLUTIONS OF A DIFFERENTIAL EQUATION 29

Two conclusions seem to stem from these examples. First, if a differen-

tial equation has a solution, it has infinitely many solutions (remember the
c's can have infinitely many values). Second, if the differential equation
is of the first order, its solution contains one arbitrary constant; if of the
second order, its solution contains two arbitrary constants; if of the nth
order, its solution contains n arbitrary constants. That both conjectures
are in fact false can be seen from the following examples.
Example 4.2. The first order differential equation
(a)
also the second order differential equation
(b) (y")2 + y2 = 0,

each has only the one solution y = 0.

Example 4.21. The first order differential equation
(a) Iy'l + 1 = 0,
also the second order differential equation
(b) ly"l + 1 = 0,
has no solution.
Example 4.22. The first order differential equation
(a) xy' = 1
has no solution if the interval [ is -1 < x < 1. Formally one can solve
(a) to obtain
(b) y = log Ixl + c,
but this function is discontinuous at x = 0. By Definition 3.4, a solution
must satisfy the differential equation for every x in [.
Remark. If x < 0, then by (b) above
(c) y = log (-x) + Ct, x < 0,
is a valid solution of (a). And if x > 0, then by (b)
(d) y = log x + C2, X > 0,
is a valid solution of (a). The line x = 0, therefore, divides the plane into
two regions; in one (c) is valid, in the other (d) is valid. There is no
solution, however, if the region includes the line x = 0.
30 BASIC CONCEPTS Chapter 1

Example 4.23. The first order differential equation

(a) (y' - y)(y' - 2y) = 0

has the solution

(b)

which has two arbitrary constants instead of the usual one.

These examples should warn you not to jump immediately to the con-
clusion that every differential equation has a solution, or if it does have a
solution that this solution will contain arbitrary constants equal in num-
ber to the order of the differential equation. It should comfort you to
know, however, that there are large classes of differential equations for
which the above conjectures are true, and that these classes include most
of the equations which you are likely to encounter. For these classes only,
then, we can assert: the solution of a differential equation of order n containB
n arbitrary constants Cl, C2, · · · , Cn.
It is customary to call a solution which contains n constants Ct, C2,
• • • , Cn an n-parameter family of solutions, and to refer to the con-
stants Cl to Cn as parameters. In this new notation, we would say (4.11)
is a 1-parameter family of solutions of (4.1); (4.13) is a 2-parameter family
of solutions of (4.12), etc.
Definition 4.3. The functions defined by
(4.31)

of the n + 1 variables, x, Cl, C2,

· · · , Cn will be called an n-parameter
family of solutions of the nth order differential equation

(4.32) F(x,y,y', · · · , yen») = 0,

if for each choice of a set of values Cl, C2, · · · , Cn, the resulting function
f(x) defined by (4.31) (it will now define a function of x alone) satisfies
(4.32), i.e., if
(4.33) F(x,f,f', • · • ,f(n») = o.
For the classes of differential equations we shall consider, we can now
assert: a differential equation of the nth order has an n-parameter family of
solutions.
Example 4.34. Show that the functions defined by
(a)

of the three variables x, Cl, C2, are a 2-parameter family of solutions of the
Lesson 4B FINDING EQUATION FROM FAMILY OF SOLUTIONS 31

second order differential equation

(b) F(x,y,y' ,y") = y" - 3y' + 2y - 4x = o.

Solution. Let a,b be any two values of Cl,C2 respectively. Then, by (a),

(c) y = I(x) --: 2x + 3 + ae% + be 2 %.

[Note that (c) now defines a function only of x.] The first and second
derivatives of (c) are

(d) y' = I'(x) = 2 + ae% + 2be %, 2

SUbstituting in (b) the values of I, I' and I", as found in (c) and (d), for
y, y', y", we obtain

(e) F(x,I,I',I") = ae% + 4be 2%- 6 - 3ae% - 6be 2%

+ + +
4x 6 2ae% +
2be 2%- 4x = o.
You can verify that the left side of (e) reduces to zero. Hence by Defini-
tion 4.3, (a) is a 2-parameter family of solutions of (b).

LESSON 4B. Method of Finding a Differential Equation if Its

n-Parameter Family of Solutions Is Known. We shall now show
you how to find the differential equation when its n-parameter family of
solutions is known. You must bear in mind that although the family
will contain the requisite number of n arbitrary constants, the nth order
differential equation whose solution it is, contains no such constants. In
solving problems of this type, therefore, these constants must be elimi-
nated. Unfortunately a standard method of eliminating these constants
is not always the easiest to use. There are frequently simpler methods
which cannot be standardized and which will depend on your own in-
genuity.
Example 4.4. Find a differential equation whose 1-parameter family
of solutions is

(a) y = C cos x + x.
Solution. In view of what we have already said, we assume that
since (a) contains one constant, it is the solution of a first order differential
equation. Differentiating (a), we obtain

(b) y' = -c sin x + 1.

This differential equation cannot be the one we seek since it contains the
parameter c. To eliminate it, we multiply (a) by sin x, (b) by cos x and
32 BASIC CONCEPTS Chapter 1

add the equations. There results

(c) y sin x + y' cos X = x sin x + cos x,
(y' - 1) cos x + (y - x) sin x = 0,

y) tan x + 1, x
'Tr 3'Tr
y' = (x - ± 2 ' ± 2 ' ···,

which is the required differential equation. [We could also have solved
(b) for C and substituted its value in (a).] Note that the interval for
which (a) is a solution of (c) must exclude certain points even though the
function (a) is defined for these points.
,
Example 4.5. Find a differential equation whose 2-parameter family
of solutions is
(a)
Solution. Since (a) contains two parameters, we assume it is the
solution of a second order differential equation. We therefore differentiate
(a) twice, and obtain
(b) YI = Cle
X
- C2 e
-x
,

(c) y" = cle

x
+ C2e- x
•

Because of the presence of the constants Cl and C2 in (c), it cannot be the

differential equation we seek. A number of choices are available for elimi-
nating Cl and C2. We could,. for example, solve (a) and (b) simultaneously
for Cl and C2 and then substitute these values in (c). This method is a
standard one which is always available to you, provided you know how
to solve the pair of equations. An easier method is to observe that the
right side of (c) is the same as the right side of (a). lIenee, by equating
their left sides, we have
(d) y" - y = 0,
which is the required differential equation.
Example 4.51. Find a differential equation whose 2-parameter family
of solutions is
(a)
Solution. Since (a) contains two constants, we assume it is the solu-
tion of a second order differential equation. Hence we differentiate (a)
twice, and obtain
(b) y' = Cl cos X - C2 + 2x,
sin x
(c) y" = -Cl sin x - C2 cos X + 2
Lesson 4C GENERAL AND PARTICULAR SOLUTIONS 33

Here again you could use the standard method of finding Cl and C2 by
solving (a) and (b) simultaneously, and then substituting these values
in (c). [Or you could solve (b) and (c) simultaneously for Cl and C2 and
substitute these values in (a)]. An easier method is to observe from (c) that

(d) Cl sin x + C2 cos X = 2 - y"

Substitution of (d) in (a) gives

(e) y = 2 - y" + x2 or y" = x2 - Y + 2,

which is the required differential equation.
Example 4.52. Find a differential equation whose I-parameter family
of solutions represents a family of circles with centers at the origin.
Solution. Here the family of solutions is not given to us in the form
of a mathematical equation. However, the family of circles with center
at the origin is

(a)

Since (a) has only I-parameter T, we assume, it is the solution of a first

order differential equation. Hence we differentiate (a) once, and obtain
(b) x + yy' = 0,

which is the required differential equation. Note that in this example the
parameter T was eliminated in differentiating (a), and we were thus able
to obtain the required differential equation immediately.

LESSON 4C. General Solution. Particular Solution. Initial Con-

ditions. An n-parameter family of solutions of an nth order differential
equation has been called traditionally a "general" solution of the differen-
tial equation. And the function which results when we give a definite set
of values to the constants Cl, C2, • • • , Cn in the family has been called a
"particular solution" of the differential equation.
Traditionally then, for example, y = ceZ which is a I-par&meter family
of solutions of y' - y = 0, would be called its general solution. And if
we let C = -2, then y = -2eZ would be called a particular solution of
the equation. It is evident that an infinite number of particular solutions
can be obtained from a general solution: one for each value of c.
A general solution, if it is to be worthy of its name, should contain all
solutions of the differential equation, Le., it should be possible to obtain
every particular solution by giving proper values to the constants Ct, C2,
• • • ,Cn • Unfortunately, there are differential equations which have solu-
tions not obtainable from the n-parameter family no matter what values
34 BASIC CONCEPTS Chapter 1

are given to the constants. For example, the first order differential equation
(4.6) Y = xy' + (y')2
has for a solution the I-parameter family
(4.61)

Traditionally, this solution, since it contains the required one parameter,

would be called the general solution of (4.6). However, it is not the gen-
eral solution in the real meaning of this term since it does not include
every particular solution. The function
x2
(4.62) y= - -
4

is also a solution of (4.6). (Verify it.) And you cannot obtain this func-
tion from (4.61) no matter what value you assign to c. [(4.61) is a first
degree equation; (4.62) is a second degree equation.]
Unusual solutions of the type (4.62), i.e., those which cannot be ob-
tained from an n-parameter family or the so-called general solution, have
traditionally been called "singular solutions." We shall show below by
examples that the use of these terms-general solution and singular solu-
tion-in their traditional meanings is undesirable. Rather than being
helpful in the study of differential equations, their use leads only to
confusion.
Consider for example the. first order differential equation

(4.63)

Its solution is
1
(4.64) y = (x + C)2·
(Verify it.) But (4.63) has another solution
(4.65) y = 0,

which cannot be obtained from (4.64) by assigning any value to c. By the

traditional definition, therefore, y = 0 would be called a singular solution
of (4.63). However, we can also write the solution of (4.63) as
C2 •
(4.651) y = (Cx + 1)2·
[Now verify that (4.651) is a solution of (4.63).] In this form, y = 0 is
not a singular solution at all. It can be obtained from (4.651) by setting
C = O. Hence use of the traditional definitions for general solution and
Lesson 4C GENERAL AND PARTICULAR SOLUTIONS 35

singular solution in this example leads us to the uncomfortable contradic-

tion that a solution can be both singular and nonsingular, depending on
the choice of representation of the 1-parameter family.
Here is another example. The first order differential equation
(4.652) (y' - y)(y' - 2y) = 0
has the following two distinct 1-parameter family of solutions
(4.653) z
y = cle ,

(4.654) y = C2 e2z •

[Verify that each of these families satisfies (4.652).]

If we call (4.653) the general solution of (4.652), as it should be called
traditionally since it contains the required one parameter, then the entire
family of functions (4.654) is, in the traditional sense, singular solutions.
They cannot be obtained from (4.653) by giving any values whatever to
Ct. If we call (4.654) the general solution of (4.652), as well we may in
the traditional sense, since it too contains the requisite one parameter,
then all the functions (4.653) are singular solutions. Hence, use of the
traditional definitions for general solution and singular solution again leads
us, in this example, to the uncomfortable contradiction that a family of
solu tions can be both general and singular.
In this text, therefore, we shall not call an n-parameter family of solu-
tions a general solution, unless we can prove that it actually contains
every particular solution without exception. If we cannot, we shall use
the term n-parameter family of solutions. In such cases, we shall make
no attempt to assert that we have obtained all possible solutions, but shall
claim only to having found an n-parameter family. Every solution of the
given differential equation, in which no arbitrary constants are present,
whether obtained from the family by giving values to the arbitrary con-
stants in it or by any other means, will be called, in this text, a particular
solution of a differential equation. In our meaning of the term, therefore,
(4.62) is a particular solution of (4.6), not a singular solution.
Definition 4.66. A solution of a differential equation will be called a
particular solution if it satisfies the equation and does not contain arbi-
trary constants.
Definition 4.7. An n-parameter family of solutions of a differential
equation will be called a general solution if it contains every particular
solution of the equation.
Since there is an infinite number of ways of choosing the n arbitrary
constants CI, C2, • • • , Cn in an n-parameter family, one may well wonder
how they are determined. What we usually want is the one solution of
the infinitely many that will satisfy certain conditions. For instance, we
36 BASIC CONCEPTS Chapter 1

may observe in an experiment, that at time t = 0 (i.e., at the start of

an experiment) a body is 10 feet from an origin and is moving with a
velocity of 20 ft/sec. The constants then must be so chosen that when
t = 0, the solution will give the value 10 feet for its position and 20 ft/sec
for its velocity. For example, assume the motion of the body is given by
the 2-parameter family

(a)

where x is the distance of the particle from an origin at time t. Its velocity,
obtained by differentiating (a),
,
is

(b) v = 32t + Cl.

Hence we must choose the constants Cl and C2 so that WRen t = 0, x = 10,
and v = 20. Substituting these values of t, x, and v in (a) and (b), we find
C2 = 10, Cl = 20. The particular solution, therefore, which satisfies the
given conditions of this problem is

(c) x = 16t 2 + 20t + 10.

Definition 4.71. The n conditions which enable us to determine the
values of the arbitrary constants Cl, C2, • • • , Cn in an n-parameter family,
if given in tenns of one value of the independent variable, are called
initial conditions.
In the example above, the given conditions were initial ones. Both the
value of the function and of its derivative were given in terms of the one
value t = o.
Comment 4.72. Normally the number of initial conditions must
equal the order of the differential equation. There are, as usual, excep-
tional cases where this requirement can be modified. For our classes of
differential equations, however, this statement will be a true one.
Example 4.8. Find a I-parameter family of solutions of the differen-
tial equation
(a)

and the particular solution for which y(2) = O. [This notation, y(2) = 0,
is a shorthand way of stating the initial conditions. Here these are x = 2,
y = O. It means that the point (2,0) must lie on or satisfy the particular
solution.]
Solution. If y :F -1, we may divide (a) by (y + 1)2 and obtain
(b) f (y 1)2 dy = f ch, y -1.
Lesson 4-Exercise 37

Perfonning the indicated integrations gives

1
(c) Y + 1 + log Iy + 11 = x + c, y -1,

which is the required I-parameter family. To find the particular solution

for which x = 2, Y = 0, we substitute these values in (c) and obtain
(d) 1= 2 +c or c = -1.

Substituting (d) in (c), there results the required particular solution,

(e) y +1 1 + log Iy + 11 = x-I, y -1.

NOTE. The function defined by y = -1 which we had to discard to

obtain (c) is also a solution of (a). (Verify it.) Hence,
(f)

is also a particular solution of (a). It is a particular solution which cannot

be obtained from the family (c) by assigning any value to the constant c.

EXERCISE"
In problems 1-3, show that each of the functions on the left is a 2-param-
eter family of solutions of the differential equation on its right.
3
1. y = CI + c2e- + 3x ' Y" + y' - x 2 - 2x = o.
z

2. y = cle- 2z + c2e- + 2e z
y" + 3y' + 2y - 12e = o.
z
,
z

3. y = CIl: + C2X- + Ix log x,

1 x 2 y" + xy' - y - x = o.
In problems 4 and 5, show that each of the functions on the left is
a 3-parameter family of solutions of the differential equation on its right.
4. y = e" (Cl + C2X + ca x2 + x:), y'" - 3y" + 3y' - y - eZ = O.

5. y -
_
CI + C2e + C3e-z +
Z
12 + 9 cos 2x 520
- 7 sin 2X) 2z
e ,
y'" - y' - e2z sin 2 x = o.
In each of problems 6-17, find a differential equation whose solution is
the given n-parameter family.
6. y == cx + c3• 12. y = CleC2Z •

7. x - cy + c = o.
2 2
13. y == x
3
+ -.
C
x
8. y = Cl cos 3x + C2 sin 3x. 14. y = cle2z + c2e- 2z•
9. r = 8 tan (8 + c). IS. (y - c)2 = cx.
10. y = cx + 3c 2 - 4c. 16. r = a(l - cos 8).
y = VCIX 2 + C2. log y = CIX + C2.
2
11. 17.
38 BASIC CONCEPTS Chapter 1

Find a differential equation whose solution is

18. A family of circles of fixed radii and centers on the z axis.
19. A family of circles of variable radii, centers on the z axis and passing through
the origin.
20. A family of circles with centers at (h,k) and of fixed radius.
21. A family of circles with centers in the xy-plane and of variable radii. Hin'.
Write the equation of the family as z2 +
y2 - 2CIZ - 2C2Y +
2ca == o.
22. A family of parabolas with vertices at the origin and foci on the z axis.
23. A family of parabolas with foci at the origin and vertices on the z axis.
24. A family of parabolas with foci and vertices on the z axis.
25. A family of parabolas with axes parallel to the z axis and with a fixed dis-
tance a/2 between the vertex and focus of each parabola.
26. A family of equilateral hyperbolas whose asymptotes are the coordinate axes.
27. A family of straight lines whose y intercept is a function of its slope.
28. A family of straight lines that are tangents to the parabola y2 == 2z.
29. A family of straight lines that are tangents to the circle z2 +
y2 == c2 ,
where c is a constant.
30. Find a I-parameter family of solutions of the differential equation dy == y tb;
and the particular solution for which y(3) == 1.

ANSWERS"
6. y == xy' + (y') 3 • 19. 2xyy' + x 2 - y2 == .0.
7. x 2(y')2 - 2xyy' + 4x2 == O. 20. [1 + (y')2]3 == a2(y") 2.
8. y" + 9y == o. 21. y"'[1 + (y')2] == 3y'(y") 2.
9. 8r' == 82 + + r.
r2 22. 2xy' == y.
10. y == (x - 4)y' + 3(y')2. 23. y(y')2 + 2xy' - y == o.
II. xyy" + x(y')2 - yy' = o. 24. yy" + (y')2 == o.
12. yy" = (y')2. 25. ay" + (y')3 == o.
13. y'x = 4x 3 - y. 26. xy' + y = o.
14. y" = 4y. 27. y = xy' + fey').
IS. 4x(y') 2 + 2xy' - y = o. 28. 2x(y')2 - 2yy' + 1 == o.

16. (1 - cos 8) :; = T sin 8. 29. 11 == zy' ± cV (y')2 + 1.

17. xyy" - yy' - x(y')2 == O. 30. y == y ==
18. (yy')2 + y2 = a2•

LESSON S. Direction Field.

LESSON SA. Construction of a Direction Field. The Isoclines of

a Direction Field. Before beginning a formal presentation of techniques
which are available for solving certain types of differential equations, we
wish to emphasize the geometric significance of a solution of a first order
differential equation. In many practical problems, a rough geometrical
approximation to a solution, such as those we shall describe below and in
later lessons, may be all that is needed. Let
(5.1) y = f(x) or f(x,y) = 0
define a function of x, whose derivative y' exists on an interval I: a < x < b.
Lesson SA CONSTRUCTION OF A DIRECTION FIELD. ISOCLINES 39

Then y' will give the slope of the graph of this function at each point whose
x coordinate is in I, i.e., y' will give the direction of the tangent to the
curve at each of these points. When, therefore, we are asked to find a
I-parameter family of solutions of

(5.11) y' = F (x,y), a < x < b,

we are in effect being asked the following. Find a family of curves, every
member of which has at each of its points a slope given by (5.11).
Definition 5.12. If y = f(x) or f(x,y) = 0 defines y as a function of
x which satisfies (5.11) on an interval I, then the graph of this function
is called an integral curve, i.e., it is the graph of a function which is a
solution of (5.11).
Therefore even if we cannot find an elementary function which is a
solution of (5.11), we can by (5.11) draw a small line element at any point
(x,y), for which x is in I, to represent the slope of an integral curve. And
if this line is short enough, the curve itself over that length will resemble
the line. For example, let us assume that y' by (5.11) has the value 2 at
the point (4,3). This means that at (4,3), the slope of an integral curve is
2. Hence we can draw a short line at this point with slope 2. In a similar
manner we can draw, theoretically, such short lines over all that part of
the plane for which (5.11) is valid.
These lines are called line elements or sometimes lineal ele-
ments. The totality of such lines has been given various descriptive
names. We shall use the term direction field. * Any curve which has at
each of its points one of these line elements as a tangent will satisfy (5.11),
and will therefore be the graph of a particular solution.
Example 5.2. Construct a direction field for the differential equation

(a) y' = x + y.
Solution. Table 5.21 gives the values of y' for the integer coordi-
nates from -5 to 5. In Fig. 5.22, we have drawn the line elements for
these values of y' and also one integral curve. It is the graph of the par-
ticular solution
(b) y=e%-x-1
of (a)
The construction of line elements is unquestionably a tedious job.
Further, if a sufficient number of them is not constructed in close prox-
imity, it may be difficult or impossible to choose the correct line element

·Other names are slope field, lineal element diagram.

40 BASIC CONCEPTS Chapter 1

Table 5.21

-5 -4 -3 -2 -1 0 1 2 3 4 5
-5 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0
-4 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1
-3 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2
-2 -7 -6 -5 -4 -3 -2 -1 0 1 2 3
-1 -6 -5 -4 -3 -2 -1 0 1 2 3 4
0 -5 -4 -3 -2 -1 0 1 2 3 4 5
1 -4 --3 -2 -.-1 0 1 2 3 4 5 6
2 -3 -2 -1 0 1 2 3 4 5 6 7
3 -2 -1 0 1 2 3 4 5 6 7 8
4 -1 0 1 2 3 4 5 6 7 8 9
5 0 1 2 3 4 5 6 7 8 9 10

for the particular integral curve we wish to find. If such doubt exists in a
certain neighborhood, it then becomes necessary to construct additional
line elements in this area until the doubt is resolved. Fortunately there

Figure 5.22

exist certain aids which can facilitate the construction of line elements.
One of these is to make use of the isoclines of a direction field. We
shall explain its meaning below.
Lesson 5B ORDINARY AND SINGULAR POINTS OF y' = F(z,y) 41

In (a), let y' equal any value, say 3. Then (a) becomes

(c) x +y = 3,

which in effect says that the slope y' has the value 3 at each point where
the integral curve crosses this line. [Look at the table of values given in
5.21. At all points which satisfy (c) and are therefore points on this line,
as for example (5,-2), (0,3), (1,2), etc., y' = 3.] Hence we can quickly
draw a great many lineal elements on the line (c). All we need do is to
construct at any point on it a line element with slope 3. This line there-
fore has been called appropriately an isocline of the direction field. For
each different value of y', we obtain a different isocline. All the straight
lines drawn in Fig. 5.22 are isoclines. In general, therefore, if

(5.23) y' = F(x,y),

then each curve for which

(5.24) F(x,y) = k,

where k is any number, will be an isocline pf the direction field deter-

mined by (5.23). Every integral curve will cross the isocline with a slope k.

Remark. For our illustration, we chose an F(x,y) which, when set

equal to k, could be solved explicitly for y. We were therefore able to find
the isoclines of the direction field without much trouble. We should
warn you however, that in many practical cases, (5.24) may be more
difficult to solve than the given differential equation itself. In such cases,
we must resort to other means to find a solution.
An integral curve which has been drawn by means of a direction field
may be looked upon as if it were formed by a particle moving in such a
way that it is tangent to each of its line elements. Therefore the path of
this particle (which remember is an integral curve) is sometimes referred
to as a streamline of the field moving in the , direction of the field. Every
student of physics has witnessed the formation of a direction field when
he has gently tapped a glass, covered with iron filings, which had been
placed over a bar magnet. Each iron filing assumes the direction of a
line element, and the imaginary curve which has the proper line elements
as tangents is a streamline.

LESSON SH. The Ordinary and Singular Points of the First Order
Equation (5.11). In the example of the previous lesson, each point
(x,y) in the plane determined one and only one lineal element. Now con-
sider the following example.
42 BASIC CONCEPTS Chapter 1

Example 5.3. Construct a direction field for the differential equation

.
, _ 2(y - 1) 0
(a) y - , X r •
X
,
Solution. (See Fig. 5.31.) We observe from (a) that we can con-
struct line elements at every point of the plane excepting at those points
whose x coordinate is zero. If therefore we were attempting to find a
particular integral curve of (a) by means of a direction field construction,
we could do so as long as we did not cross the x axis. For example, if we

B
y

Figure 5.31

began at a point in the second quadrant of the plane and followed a

streamline, we would be stopped at the point (0,1) since by (a), y' is
meaningless there. Even if we had concluded that an arbitrary assign-
ment of the value zero to y' at this- point would seem reasonable and give
continuity to the integral curve made by the streamline, so that (a) would
now read
(b) y' = 2(y - 1), x ¢ 0
x
= 0, x = 0, y = 1,
we would be at a loss to know which streamline to follow after crossing
(0,1). If you will draw sufficient lineal elements in the neighborhood of
Lesson 5B ORDINARY AND SINGULAR POINTS OF y' = F(x,y) 43

(0,1), it will soon become evident to you, that with the new definition of
y' as given in (b), an infinite number of integral curves lies on the point
(0,1) with slope zero. Hence after crossing this point, one could follow
any first, third, or fourth quadrant streamline, or even another second
quadrant st.reamline.
To overcome this difficulty, we could specify two sets of initial condi-
tions in place of the usual one. For example, we could require our solu-
tion to lie on the point (-3,2), and after crossing (0,1) to go through the
point (2,-1). These two initial conditions would then fix a particular
integral curve. This annoying difficulty arises because of the necessity of
excluding x = 0 from the interval of definition. Actually, there are two
distinct solutions of (a), namely

(c) y = CIX
2
+ 1, x < 0,
y = C2X2 + 1, x > o.
[Be sure to verify that each function defined in (c) is a solution of (a).]
Since the point (0,1) satisfies both equations in (c) and since we have
agreed to define the slope y' as equal to zero at this point, we can write
(c) as

(d) y = CIX
2
+ 1, x < 0,
y = C2X2 + 1, x > o.
In this form the solutions (d) include every particular solution of the
given differential equation with the agreement that y' equals zero when
x = 0, y = 1.
In the form of (d), we can now make the further observation
that with the exception of (0,1), no integral curve lies on any point in the
plane whose x coordinate is zero. For example the point (0,3) does not
satisfy either equation in (d) no matter what values you assign to Cl
and C2.
We characterize the difference between points like (0,1), (0,3), and
(2,3) in the following definitions.
Definition 5.4. An ordinary point of the first order differential equa-
tion (5.11) is a point in the plane which lies on one and only one of its
integral curves.
Definition 5.41. A singular point of the first order differential equa-
tion (5.11) is a point in the plane which meets the following two require-
ments:
1. It is not an ordinary point, i.e., it does not lie on any integral curve or
it lies on more than one integral curve of (5.11).
44 BASIC CONCEPTS Chapter 1

2. If a circle of arbitrarily small radius is drawn about the point (Le., the
radius may be as small as one wishes), there is at least one ordinary
point in its interior. (We describe this condition by saying the singular
point is a limit of ordinary points.)
In the above Example 5.3, every circle, no matter how small, drawn
about any point on the y axis, say about the point (0,3), contains not
only one ordinary point but also infinitely many such points.
Remark. Requirement 2 is needed to exclude extraneous points. For
example, if y' = VI - X2, then only points whose x coordinates lie
between -1 and 1 need be cQnsidered. If, therefore, we defined a singular
point by requirement 1 alone, then a point like (3,7) would be singular.
This point, however, is extraneous to the problem.
This example has served three purposes:
1. It has shown you what an ordinary point and a singular point are.
2. It has shown you the need for specifying intervals for which a differ-
ential equation and its solution have meaning. You cannot work auto-
matically and blindly and write as a solution of (a)
(e) y = cx 2 + 1, - 00 < x < 00.

If you did that, and even if you defined y' = 0 at (0,1), you could get
from (e) only the parabolic curves as solutions. One of these is shown in
Fig. 5.31. It is marked A, and was obtained from (e) by setting c = 1
[equivalent to setting Cl = 1 and C2 = 1 in (d)]. If in (d), we set Cl = 2
and C2 = -2, we get the integral curve marked B in the graph. And if
in (d) we set Cl = 1 and C2 = -2, we get the curve which is marked A
in the second quadrant and B in the fourth quadrant.
3. It shows once more that not every first order differential equation has
a I-parameter family of solutions for its general solution. The differen-
tial equation in this example requires two I-parameter families to
include all possible solutions.

EXERCISE 5
1. Construct a direction field for the differential equation
y' = 2x.
Draw an integral curve.
2. Construct a direction field for the differential equation
y' = 2y.
x
Where are its singular points? How many parameters are required to include
all possible solutions? Draw the integral curve that goes through the points
(-1,2) and (2,-1).
Lesson 5-Exercise

3. Construct a direction field for the differential equation

x-y
- x+
y' - - - y.
Draw an integral curve that goes through the point (1,1).
4. Find the isoclines of the direction field and indicate the slope at each point
where an integral curve crosses an isocline, (a) for problem 2, (b) for problem 3.
5. Describe the isoclines of the direction field, if y' = x 2 + 2y2.

ANSWERS 5
2. On line x = 0; two parameters.
4. (a) Isoclines are the family of straight lines through the origin: 2y = cx .
Slope of an integral curve at each point where it crosses an isocline is equal to c.
(b) Isoclines: x(l - c) = y(l + c), slope c.
5. Isoclines are the family of ellipses x 2 + 2y2 = c. At each point where an
integral curve crosses one of these ellipses, the slope of the integral curve is c.
 Chapter 2

Special Types of Differential Equations

of the First Order

Introductory Remarks. In this chapter we begin the study of formal

methods of solving special types of first order differential equations. A
few preliminary observations however should be instructive and helpful.
1. It is unfortunately true that only very special types of first order
differential equations possess solutions (remember a solution is a function)
which can be expressed in terms of the elementary functions mentioned
in Lesson 2E. Most first order differential equations, in fact, one could say
almost all, cannot be thus expressed.
2. There is no connection between the appearance of a differential
equation and the ease or difficulty of finding its solution in terms of ele-
mentary functions. The differential equation

does not look less complicated than

dy = x2
dx
+ y2 or dy
dx=
ez2
·

Yet the first has an elementary function for its solution; the other two
do not.
3. If the solution you have found can be expressed only in the implicit
formf(x,y) = 0, it will usually be of little practical value. An implicit solu-
tion is frequently such a complicated expression that it is almost impos-
sible to find the needed function g(x) which it implicitly defines, (see
Definition 2.81). And without a knowledge of the function g(x) or at
least a knowledge of what a rough graph of g(x) looks like, the solution
will not be of much use to you. While we shall show you, therefore, in
the lessons which follow, formal techniques for finding solutions of a first
order differential equation, keep in mind, if the solution is an implicit one,
46
Lesson 6A DIFFERENTIAL OF A FUNCTION 47

that other forms of solutions we shall describe later, such as geometric

solutions, series solutions, and numerical solutions, will be of far greater
practical importance to you.
4. If you start with an algebraic equation and follow a certain pro-
cedure to find a solution for a variable x, it is possible that the value thus
obtained is extraneous. For example, the usual procedure followed to
solve the equation v'x 2 + 4x - 3 = 1 - 2x is to square both sides and
then factor the resulting equation. If you do this you will obtain the solu-
tions x = 2 and x = i. However, both values are extraneous. Neither
solution satisfies the given equation. (Verify it.) Similarly, in showing
you a procedure that will lead you to a solution of a differential equation,
it is possible that the function thus obtained will be extraneous. Hence,
to be certain a function is a solution of a given differential equation, you
should always verify that it does in fact satisfy the given equation.
5. Finally, and we cannot emphasize this point too strongly, examples
generally found in textbooks are "textbook" examples. They are inserted
as illustrations in order to clarify the subject matter under discussion.
Hence they are carefully selected to yield "nice, relatively easy" solutions.
Actual practical problems are avoided since they may require, for in-
stance, the determination of the imaginary roots of a fourth degree equa-
tion, or the solving of a system of four or more equations in a corresponding
number of unknowns-burdensome and time-consuming problems to say the
least.

LESSON 6. Meaning of the Differential of a Function.

Separable Differential Equations.

LESSON 6A. Differential of a Function of One Independent

Variable. We assume in this Lesson 6A that all functions are differ-
entiable on an interval. Let y = f(x) define y as a function of x. Then
its derivative f'(x) will give the slope of the curve at any point P(x,y)
on it, i.e., it is the slope of the tangent line drawn to the curve at P.
It is evident from Fig. 6.12, that
(6.1) f'(x) = tan a = dy ·
ax
Hence,
(6.11) dy = f'(x) dx.

We call dy the differential of y, i.e., it is the differential of the function

defined by y = f(x). From (6.11) we note that the differential of y,
namely dy, is dependent on the abscissa x (remember as the point P
changes, f'(x) changes), and on the size of dx. We see, therefore, that
whereas y = f(x) defines y as a function of one independent variable x,
48 SPECIAL TYPES OF FIRsT ORDER EQUATIONS Chapter 2

(0,0) x
Fisure 6.12

the differential dy is a function of two independent variables x and flx.

We indicate this dependence of dy on x and L\x by writing it &8

Hence,
Definition 6.13. Let y = J(x) define y as a function of x on an inter-
val I. The differential of y, written as dy (or df) is defined by
(6.14) = J'(x) flx.

Note. We shall want to apply Definition 6.13 to the function defined

by y = x. Therefore, in order to distinguish between the Junction defined
by y = x and the variable x, we place the symbol A over the z so that
(6.15) y=f
will define the function that assigns to each value of the independent
variable x the same unique value to the dependent variable y.
Theorem 6.2. If
(6.21) y = f,
then
(6.22) (dy){x,l1x) = (dx)(x,L\x) = L\x.
Proof. Since
(a) y==f
defines the function that assigns to each value of the independent variable
x the same unique value to the dependent variable y, its graph is the
straight line whose slope is given by
(b) y' = f'(x) = 1.
Substituting f'{x) = 1 in (6.14), we obtain (6.22).
Le8son 6A DIFFERENTIAL OF A FuNCTION 49

Commen.t 6.23. If in (6.14) we replace Ilx by its value as given in

(6.22), it becomes

(6.24) (dy)(x,llx) =
In words (6.24) says that if y = f(x) defines y as a function of x, then the
differential of y is the product of the derivative of the function f and the
differential of the function defined by y = f. The relation (6.24) is the
correct one, but in tlie course of time, it became customary to write
(6.24) in the more familiar fonn
(6.25) dy = f'(x) dx, :; = f'(x),

Example 6.26. If
(a)

defines y as a function of x, find dy.

Solution. Heref(x) = x 2 • Thereforef'(x) = 2x. Hence by (6.24)

(b)

which is customarily written as

(c) dy = 2x dx.

The importance of the definition of the differential as given in 6.13 lies

in the following theorem.
Theorem 6.3. If y = f(x) defines y as a Junction of x and x = g(t),
y = f[g( t)] = F (t), define x and y as Junctions of t, then

(6.31) (dy) (t,llt) = J' (x) (dx) (t,llt).

Proof. Since x = get) defines x as a function of t, we have by (6.·24)

(a) (dx) (t,llt) = g' (t) dt (t,llt) ,

where x = t defines the function which assigns to each value of the inde-
pendent variable t, the same value to the dependent variable x. By
hypothesis y = J[g(t)] defines y as a function of t. Therefore by (6.24)
and the chain rule of differentiation,

(b) (dy)(t,llt) = J'[g(t)][g'(t) dt(t,dt)].

In (b) replace the last expression in brackets on the right, by its equal
left side of (a) and replace get) by its equal x. The result is (6.31).
so SPECIAL TYPES OF FIRST ORDER EQUATIONS Chapter 2

To summarize:
1. If y = f(x), then = f'(x)
2. If y = f(x) and x = get) so that y = f[g(t)], then (dy)(t,flt) =
f'(x)(dx)(t,dt), where dy and dx are differentials of y and x respectively.
The first is the differential of f[g(t)]; the second is the differential of get).
In both cases 1 and 2, we shall follow the usual custom and write

(6.32) dy = f'(x) dx or dy = f'(x).

dx
Comment 6.33. Jf y = f(x) and x = get), then y = f[g(t)] defines y
as a function of t. The independent variable is therefore t; the dependent
variables x and y. In general if y is a dependent variable, the increment
fly dy, see Fig. 6.12. It follows therefore that flx dx since here x is
also a dependent variable. Thus there is no justification in replacing an
increment dx by dx in "dy = f'(x) dx." However, if both dy and dx are
differentials as defined in 6.13, then as we proved in Theorem 6.3, "dy =
f'(x) dx" even when x is itself dependent on a third variable t.

LESSON 6B. Differential of a Function of Two Independent

Variables. Let z = f(x,y) define z as a function of the two independent
variables x and y. Then, following the analogy of the one independent
variable treatment, we define the differential of z as follows.
Definition 6.4. Let z = f(x,y) define z as a function of x and y.
The differential of z, written as dz or df, is defined by

(6.41) (dz)(x,y,t:.x,lly) = Ilx + Ily.

Note that whereas z is a function of two independent variables, the

differential of z is a function of four independent variables.
Theorem 6.42. If
(6.43) z = f(x,y) = x,
where x has the usual meaning, then
(6.44) (dZ)(x,y,dx,dy) = dX(X,y,dx,dy) = dx.
Proof. Here z = f(x,y) = x. Hence
af(x,y) == ax = 1 ax = 0.
ax ax ' ay
Substituting these values in (6.41), we obtain (6.44).
Similarly, it can be shown, if z = f(x,y)= '0, that
(6.45) (dY)(X,y,dx,dy) = dy,
Lesson 6C DIFFERENTIAL EQUATIONS WITH SEPARABLE VARIABLES 51

where y has the usual meaning. Substituting (6.44) and (6.45) in (6.41),
we obtain
(6.46)
(dz)(x,y,llx,lly) = (dX)(x,y,llx,lly) + (dY)(x,y,llx,lly).

This relation (6.46) is the correct one. However, in the course of time,
it became the custom to write (6.46) as

(6.47) dZ = af(x,y) d
ax x
+ af(x,y)
ay
d
y.

Keep in mind that dx and dy mean di and dy and are therefore differentials,
not increments. With this understanding of the meaning of dx and dy we
shall now state, but not prove, an important theorem analogous to Theo-
rem 6.3 in the case of one independent variable. The theorem asserts that
(6.47) is valid even when x and yare both dependent on other variables.
Theorem 6.5. If Z = f(x,y) defines z as a function of x and y, and
x = x(r,s, · · .), y = y(r,s, · · .), z = f[x(r,s, · · .), y(r,s, · · .)] = F(r,s, · • .)
define x, y, and z as functions of r, s, and a finite number of other variables
(indicated by the dots after s), then

(6.51) (dz)(r,s,"', Ilr,lls, · · .) = (dx)(r,s, · • · , Ilr,lls · · .)

+ af(x,y)
ay (dy) (r,s, . · · , dr,ds, · · .).

Here also we shall follow the usual custom and write (6.51) as

(6.52) dz = af(x,y) d
ax x
+ af(x,y)
ay
d
y.

Example _6.53. Fintl dz if

(a) z = f(x,y) = x 3 + 3x 2 y + y3 + 5.
Solution. Here

(b) = 3x 2 + 6xy, = 3x 2 + 3y2.

Hence by (6.52)

(c) dz = (3x 2 + 6xy) dx + (3x 2 + 3y2) dye

LESSON 6C. Differential Equations with Separable Variables.
The first order differential equations we shall study in this chapter will be
32 SPECIAL TYPES OF FIRST ORDER EQUATIONS Chapter 2.

those which can be written in the fonn

(6.6) Q(x,y) + P(x,y) = o.

Written in this form it is assumed that y is the dependent variable and x
is the independent variable. If we multiply (6.6) by dx, it becomes
(6.61) P(x,y) dx + Q(x,y) dy = o.
Written in this form, either x or y may be considered as being the de-
pendent variable. In both cases, however, dy and dx are differentials and
not increments.
Although (6.6) and (6.61) are not the most general equations of the
first order, they are sufficiently inclusive to cover most of the applications
which you will meet. Examples of such equations are

(a) ;;; = 2xy +e S

,

(b) y' == log x + y,

(c) (x - 2y) dx + (x + 2y + 1) dy = 0,
(d) eX cos y dx + x sin y dy = o.

If it is possible to rewrite (6.6) or (6.61) in the form

(6.62) f(x) dx + g(y) dy = 0,
so that the coefficient of dx is a function of x alone and the coefficient of
ay is a function of y alone, then the variables are called separable. And
after they have been put in the form (6.62), they are said to be separated.
A I-parameter family of solutions of (6.62) is then

(6.63) f I(x) dx + f g(y) dy = 0,

where C is an arbitrary constant.

Example 6.64. Find a I-parameter family of solutions of
(a) 2x dx - 9y2 dy = o.
Solution. A comparison of (a) with (6.62) shows that the variables
are separated. Hence, by (6.63), its solution is
(b)

Example 6.65. Find a I-parameter family of solutions of

(a) VI - x 2 dx + V5 + y dy = 0, -1 x 1, y >
Lesson 6C DIFFERENTIAL EQUATIONS WITH SEPARABLE VARIABLES 53

Solution. A comparison of (a) with (6.62) shows that the variables

are separated. Hence its solution by (6.63) is
(b) !xVI - x 2+iArcsinx+i(5+y)3/2=C, -1 < x <1, y> -5.
Comment 6.651. Because of the presence of the inverse sine, (b) im-
a function. By our definition of a func-
tion it must be single-valued, i.e., each value of x should determine one
and only one value of y. For this reason we have written the inverse sine
with a capital A to indicate that we mean only its principal values, namely
those values which lie between - 7r/2 and 7r/2.
Example 6.66. Find a I-parameter family of solutions of
(a) XVI - y dx - VI - x 2 dy = 0;
also a particular solution not obtainable from the family.
Solution. We note first that (a) makes sense only if y < 1 and
-1 x S 1. Further if y ¢ 1, x ¢ ±I, we can divide (a) by
vI - y VI - x 2 and obtain
xdx dy
(b) = 0, -1 < x < 1, y < 1.
VI - x2 VI - y
'l his equation is now of the form (6.62). A I-parameter family of solu-
tions by (6.63) is
(c) VI - x2 - 2VI - y = C, -1 < x < 1, y < 1.
The function y = 1, which we had to exclude to obtain (c) also satisfies
(a) for values of x between -1 and 1. (Be sure to verify it.) It is a par-
ticular solution of (a) that cannot be obtained from the family (c).
Remark. In Fig. 6.67 we have y
indicated the set in the plane for y=l
which the solution (c) is valid. It (-1,1) __
is bounded on the top by the line
y = 1, and on the sides by the lines
x = 1 and x = -1. By (c), when x
x = ± 1 and y < 1, a unique value
of the constant C and therefore a
unique particular solution of (a) is
determined. However, if y < 1, Figure 6.67
dy / dx ---+ ± 00 as x ---+ ± 1. Hence
if y < 1, the lines x = ± 1 are tangents to the family of integral curves.
The corner points (1,1) and ( - 1,1) can be made part of the set. For these
two points, we obtain from (c) the solution
(d) vI - x 2 - 2Vl - y = o.
54 SPECIAL TYPES OF FIRST ORDER EQUATIONS Chapter 2

Squaring (d) and taking the derivative of the reSUlting function, we obtain
, x
(e) y = -.
2

By (a), y' is meaningless at the two corner points (1,1) and (-1,1). How-
ever, because of (e) we are led to define the slope of the integral curve at
these two corner points as i and -i. We see now that every particular
solution of (a) obtained from (c) lies in the region below the line y = 1.
The line y = 1 however, is, as we observed previously, also a particular
integral curve of (a) and is part of the set for which solutions of (a) are
valid. In Fig. 6.67 we have drawn the integral curve (d) and an integral
curve of (c) through (0,0).

Example 6.68. Find a I-parameter family of solutions of

(a) x cos y dx + vix + 1 sin y dy = 0;

also a particular solution not obtainable from the family.

Solution. We note first that (a) makes sense only if x > -1. Fur-

ther, if x ¢ -1, Y ± ; , ± 3 7r,

2
•• •, we can divide (a) by vx + 1 cos y
and obtain

(b) vxx+ 1 dx + sin yd =

cos y y
0
, x > -,1 y ¢
7r
± "2' ± 2 ±,
37r

The equation is now of the fonn (6.62). A I-parameter family, by is

2(x - 2) ,....--
(c) 3 +
y' x 1- log Icos yl = C,
7r 37r
X > -1, y ¢ ± "2' ± 2 ±, · · · ·
The functions
7r ± 37r ,
(d) Y = ± -, ±, ... ,
2 2

which we had to exclude to obtain (c), also satisfy (a). They are particular
solutions of (a) which cannot be obtained from the family (c).

Remark. In Fig. 6.69 we have indicated the set in the plane for
which the solutions (c) are valid. It is bounded on the left by the line
x = - 1, and excludes the lines y = ± ;. These last two lines, how-
ever, also are particular solutions of (a) not obtainable from (c), and are
therefore also part of the set for which solutions of (a) are valid. Each
line y = 37r/2, -31l/2, etc., is also a solution of (a).
Le8son 6-Exercise 55

...--.. . ----------Y=- 31r

2

. . ._. . . _---------Y= -- 311"

2
Figure 6.69

Example 6.7. Find a particUlar solution of

(a) xy2 dx + (1 - x) dy = 0,
for which y(2) = 1.
Solution. If y 0 and x 1, we can obtain from (a)
(b) 1 x
-x
dx + y-2 dy = 0, x F- 1, y F- 0,

which we can write as

(c) C x-I) dx + y-2 dy = 0, x F- 1, y F- 0.

By (6.63), a family of solutions of (&) is

1
(d) log 11 - xl +x+ y = C, x 1, y o.
To find the particular solution for which x = 2, y = 1, we substitute
these values in (d) and obtain 0 + 2 + 1 = C, or C = 3. Hence (d)
becomes
1
(e) log 11 - xl + x + - = 3, x 1, y ¢ O.
y

EXERCISE 6

Find a I-parameter family of solutions of each of the differential equa-

tions 1-16 listed below. Be careful to justify all steps used in obtaining a
solution and to indicate intervals for which the differential equation and
the solution are valid. Also try to discover particular solutions which are
not members of the family of solutions.
1. y' = y. 2. x dy - y dx = O.
dr . B
3. dB = -Sln •

dr
4. (y
2
+ 1) dx - (x
2
+ 1) dy = O. 5. dB cot B - r = 2.
56 SPECIAL TYPES OF FIRST ORDER EQUATIONS Chapter 2

6. yx 2 dy - y3 dx = 2x2 dYe 2
7. (y - 1) dz - (2y +
xy) dy == O.
8. x log x dy + VI + y2 dx = O. 9. tan y dx + cos y dy == O.
10. x cos y dx + x 2 sin y dy = a2 sin y dye
dr
II. d8 = r tan 8.
12. (x - 1) cos y dy = 2x sin y dx.
13. y' = y log y cot x.
14. x dy (1+ +y2) Arc tan y dx = o.
15. dy + x(y + 1) dx = o.
16. e1l2 (x2 + +
2x 1) dx (xy +
y) dy = O. +
Find a particular solution satisfying the initial condition, of each of the
following differential equations 17-21. The initial condition is indicated
alongside each equation.
dy
17. dx +y = 0, y(l) = 1.
18. sin x cos 2y dx + cos x sin 2y dy = 0, yeO) - T /2.
19. (1 - x) dy = x(y +
1) dx, y(O) = o.
20. y dy +
x dx = 3xy2 dx, y(2) = 1.
21. dy = e +1I dx, y(O) = O.
Z

22. Define the differential of a function of three independent variables; of n

independent variables.

ANSWERS 6
I. y = ceZ • 2. y = cx. 3. r = cos 8 + c.
4. Arc tan x = Arc tan y + c or x - y = 0(1 + xy), where 0 = tan c.
5. r = c sec 8 - 2, r -2, 8
T
2 + nT; r == -2.
6. (ex + 1)y2 - l)x, x F- 0,
y F- OJ Y = O.
7. x + 2 = cvR-
1, x -2, Y ±1; y = ::1:1.
8. log Ixl(y + Vy2 + 1) = c, x 0, x 1.
z 1
9. e + +
log (csc y - cot y) cos y = c, y +
nT; y = nT.
2
10. a - x
2
=
2
c cos y, x
2
a,
2
y
'If'
2 + nT; y = 2
'If'
+ nT.
II. r cos 8 = c, r
0, 8 2 nT; r == O.
T
+
12. sin y = (x - 1)2e 2z +c, x 1, y nT; y = nT.
13. y = ecsinz, x nT, y O.
14. y = tan (c/x), x o.
15. y = ce- z2/2 - 1, y -1; y = -1.
16. x2 + 2x = e- y2 c, x+ -1.
17. 1
y = e - • z

18. cos 2 x cos 2y = -l.

19. (y + 1)(1 - x) = e- Z •
20. 3y2 = 1 +
2e3z2-12.
21. eZ + e-Y = 2.
Lesson 7A DEFINITION OF A HOMOGENEOUS FuNCTION 57

LESSON 7. First Order Differential Equation with

Homogeneous Coefficients.

LESSON 7A. Definition of a Homogeneous Function.

Definition 7.1. Let z = f(x,y) define z as a function of x and y in a

region R. The function f(x,y) is said to be homogeneous of order n if
it can be written as
(7.11) f(x,y) = xng(u),
where u = y/x and g(u) is a function of u; or alternately if it can be
written as
(7.12) f(x,y) = ynh(u),
where u = x/y and h(u) is a function of u.

Example 7.13. Determine whether the function

(a) f( x,y) = x2 + y2 log xy, R: x > 0, y > 0,

is homogeneous. If it is, give its order.

Solution. We can write the right side of (a) as

(b) x2 (1 + y2 log
x2
y) .
X

If we now let u = y/x, it becomes

(c)
Hence, by Definition 7.1, (a) is a homogeneous function. Comparing
(c) with (7.11), we see that it is of order 2.
Or if we wished, we could have written (a) as

(d) y2 (:: + log = y2(U

2
- log u) = 2
y h(u),

where u = x/yo Hence by Definition 7.1, (a) is homogeneous of order 2.

Example 7.14. Determine whether the function

(a) f(x,y) = Vy sin G)

is homogeneous. If it is, give its order.
Solution. With u = x/y, we can write the right side of (a) as
(b) yl/2 sin u = yl/2h(u).

Hence, by Definition 7.1, (a) is homogeneous of order t.

58 SPECIAL TYPES OF FIRST ORDER EQUATIONS Chapter 2

Follow the procedure used in Examples 7.13 to 7.14 to check the accu-
racy of the answers given in the examples below.

Answers
1. eY /+ tan (y/x).
x
Homogeneous of order zero.
x + sin x cos y.
2
2. Nonhomogeneous.
3. v'x + y. Homogeneous of order !.
4. v'X2 3xy + 2y2. + Homogeneous of order 1.
5. X4 - 3x 3 y + 5y 2x 2 - 2y4. Homogeneous of order 4.

Comment 7.15. An alternate definition of a homogeneous function is

the following. A function f(x,y) is said to be homogeneous of order n if

(7.16) f(tx,ty) = tnf(x,y) ,

where t > 0 and n is a constant. By using this definition, (a) of Example

7.13 becomes
2 2 2
(a) f(tx,ty) = t x + t y2 log
2 2
= t (x + y2 log
= t 2f(x,y).

+
Hence the given function x 2 y 2 log (y/x) is homogeneous of order two.
By using this definition, (a) of Example 7.14 becomes

(b) f(tx,ty) = (ty) 1/2 sin ( -tx) = tl/2 ( y 1/2 sln-

• x)
ty y
= t 1 / 2f(x, y).

Hence the given function is of order t. As an exercise, use (7.16) to test

the of the answers given for the functions 1 to 5 after Example
7.14.

LESSON 7B. Solution of a Differential Equation in Which the

Coefficients of dx and dy Are Each Homogeneous Functions of
the Same Order.
Definition 7.2. The differential equation

(7.3) P(X,y) dx + Q(x,y) dy = 0,

where P(x,y) and Q(x,y) are each homogeneous functions of order n is

called a first order differential equation with homogeneous coeffi-
cients.
Lesson 7B EQUATIONS WITH HOMOGENEOUS COEFFICIENTS 59

We shall now prove that the substitution in (7.3) of

(7.31) y = ux, dy = udx + xdu

will always lead to a differential equation in x and u in which the variables
are separable and hence solvable for u by Lesson 6. The solution y will
then be obtainable by (7.31). The proof is incorporated in the following
theorem.
Theorem 7.32. If the coejJicients in (7.3) are each homogeneous func-
tions of order n, then the substitution in it of (7.31) will lead to an equation
in which the variables are separable.
Proof. By hypothesis P(x,y) and Q(x,y) are each homogeneous func-
tions of order n. Hence by Definition 7.1 with u = y/x, each can be
written as

(a)

Substituting in (7.3) the value of dy as given in (7.31) and the values of

P(x,y), Q(x,y) as given in (a), we obtain

(b)

which simplifies to

(c) [gl(U) + Ug2(U)] dx + Xg2(U) du = 0,

-dx
x
+ ()
gl U
g2(U)
+ ug2 ()
U
du = 0, x 0, gl(U) + ug2(U) 0,

an equation in which the variables x and u have been separated.

Prove as an exercise that the substitution in (7.3) of

(7.33) x = uy, dx = udy + ydu

will also lead to a separable equation in u and y.
Remark. If the differential equation (7.3) is written in the form

(7.4) dy _ P(x,y) - F(x )

dx - Q(x,y) - ,y ,

then the statement that P(x,y) and Q(x,y) are each homogeneous of order
n is equivalent to saying F(x,y) is homogeneous of order 0. For by (7.4)
and Definition 7.1

(7.41)
60 SPECIAL TYPES OF FIRST ORDER EQUATIONS Chapter 2

Example 7.5. Find a I-parameter family of solutions of

(a) (VX2 - y2 + y) dx - x dy = 0;
also any particular solution not obtainable from the family.

Solution. We observe first that (a) makes sense only if Iyl lxi, or
Iy/xl 1, x o. Second we note by Definition 7.1 that (a) is a differen-
tial equation with homogeneous coefficients of order one. We have a
choice therefore of either of the substitutions (7.31) or (7.33). Byexperi-
menting with both, you quickly will discover that the first is preferable.
By using this substitution in (a), we obtain

(b) (vx 2 - U 2X2 + ux) dx - x(u dx + x du) = 0,

lui ....: I I <1, x 0.

Since x 0, we can divide (b) by it to obtain after simplification

(c) ±VI - u 2 dx - x du = 0, x 0, lui = I I< 1,

where the + sign is to be used if x > 0; the - sign if x < O. * Further

if u ±1, we may divide (c) by vI - u 2 • Therefore (c) becomes
dx _ du 0
(d)
x
-±
VI - u2
,X"-, lui = < 1.

The variables are now separated. Hence, by (6.63), a I-parameter family

of solutions of (d) is

(e) log x = Arc sin u + c, lui < 1, x > 0,

-log (-x) = Arc sin 'U, + c, lui < 1, x < o.
Replacing u in (e) by its value as given in (7.31), we have

(f) log x = Arc sin y

x
+ c, IxYI < 1, x > 0,

-log (-x) = Arc sin Y

x
+ c, I YI
x
< 1, x < o.

In obtaining the solution (f), we had to exclude the values lui = Iy/xl =
1. This means we had to exclude the functions y = ± x. You can and
should verify that these two functions also satisfy (a). They are particular
solutions of (a) not obtainable from the family (f).

*For real x, vx 2 = x if x 0 and vx 2 = -x if x < o. For example, if x = 2,

v'22 = 2 and if x = -2, v( _2)2 = -( -2) = 2.
Lesson 7-Exercise 61
EXERCISE 7
I. Prove that the substitution in (7.3) of
x = uy, dx = u dy + y du, y 0,
leads to a separable equation.
Find a 1-parameter family of solutions of each of the following equa-
tions. Assume in each case that the coefficient of dy o.
2. 2xy dx + (x 2 + y2) dy = O.
3. (x + V y2 - xy) dy - y dx = O.
4. (x + y) dx - (x - y) dy = o.
5. xy' - y - x sin (y / x) = o.
6. (2x 2y + y3) dx + (xy2 - 2x 3) dy = O.
7. y dx + (xv y2 -- x 2 - xy) dy = o.
2 ___ r ; ; - -

8. '#.. cos '#.. dx - sin '#.. + cos '#..) dy = o.

x x y x x
9. y dx + x log'#.. dy - 2x dy = o.
x
10. 2ye dx + (y - 2xe
zl1l dy = O. z / 1I )

ll. (xe dx + xsin dy = o.

lll
Z Y sin -

Find a particular solution, satisfying the initial condition, of each of the

following differential equations.
12. (x 2 + y2) dx = 2xy dy, y(-I) = O.
13. (xe" lz + y) dx = x dy, y(l) = O.
14. y' - '#.. + csc '#.. = 0, y(l) = O.
x x
IS. (xy - y2) dx - x 2 dy = 0, y(l) = 1.

ANSWERS 7
2. 3x 2y +
y3 = c.
3. y > 0, x < Yj y = ce 2 -Vl-z/1I,
y = ce- 2..J1-zI1l, y < 0, x > y.
4. Arc tan (y/x) - I log (x 2 y2) == c. +
5. y = 2x Arc tan cx.
2
6. 2'
Y
X
+
log xy = c, x 0, y o.
7. y2 - cx = YVy2 - X2, or equivalently, c(y + Vy2 - x 2) = xy, y2 > x 2•
. y
8• y sln - = c.
x
9. y = c(1 log x/y). +
z
10. 2e / 1I +
log y = c.
II. log x
2
- e-
lI/z
(sin + = c.
12. y2 = x
2
+ x. 14. log x - cos'#.. + 1
x
= o.
13. log x + e- lIlz
= I. IS • X
(zI1l)-1
= e •
62 SPECIAL TYPES OF FIRST ORDER EQUATIONS Chapter 2

LESSON 8. Differential Equations with Linear Coefficients.

A Review of Some Plane Analytic Geometry. The

LESSON 8A.
first degree equation ax + by + c = °
represents a straight line. For this
reason it is called a linear equation. (N OTE. The presence of the con-
stant c in the equation prevents the function defined by it from being
homogeneous.) If the coefficients of x and y in one linear equation are
proportional to the x and y coefficients in another, the two lines they
represent are parallel. For example, the two lines
3x - 2y +7= 0,
6x - 4y + 3 = 0,
are parallel since 3 : -2 = 6: -4. If the three constants in one linear
equation are proportional to the three constants respectively in a second
linear equation, the two lines coincide, i.e., they are the same line. For
example, the two lines
2x + 3y + 1 = 0,
4x + 6y + 2 = 0,
are coincident. (Do you see why?)
Another concept of analytic geometry that we shall need for this lesson
is that of "translation of axes. " Let (x,y) be the coordinates of a point P
with respect to an origin (0,0) (Fig. 8.1), and let us translate the origin to

P(x,y) BE (f,y)
y

(h, k)
(0,0)
i
y
1 Y

h %
(0,0) X

Figure 8.1

a new position whose x and y distances from (0,0) are hand k respectively.
To distinguish the new origin from the old one, we call its coordinates
(0,0). The point P will then have two sets of coordinates, one with re-
spect to (0,0), which we designate by (l!,U), and the other with respect to
Lesson 8B EQUATIONS WITH LINEAR COEFFICIENTS 63

(0,0), which we have already designated as (x,y). This means that if a

point is measured from (0,0), its coordinates have no bars over them; if
it is measured from (0,0) its coordinates have bars over them. The ques-
tion we now ask and whose answer we seek is this. What is the relation-
ship between the two sets of coordinates (x,y) and (x,7i)?
If you will examine l"ig. 8.1 carefully, you will see that
(8.11) x = % + h, y = 7i + k.
Hence by (8.11),
(8.12) %=x-h, 7i = y - k.
These are the equations of translation. Their purpose, you may recall, is
to change more complicated second degree equations into simpler ones by
eliminating the first degree terms. We shall now demonstrate how a
translation of axes can help solve a differential equation with linear
coefficients.

LESSON 8B. Solution of a Differential Equation in Which the

Coefficients of dx and dy Are Linear, Nonhomogeneous, and When
Equated to Zero Represent Nonparallel Lines. Consider the differ-
ential equation

in which the coefficients of dx and dy are linear and when equated to zero
represent nonparallel lines. We assume also that both CI and C2 are not
°
zero. (If both CI = and C2 = 0, then (8.2) is a differential equation
with homogeneous coefficients which can be solved by the method of
Lesson 7.) Since the coefficients in (8.2) are assumed to define nonparallel
lines, the pair of equations
(8.21) alX + bly + CI = 0,
a2 x + b2y + C2 = 0,
formed with them, have a unique point of intersection and therefore a
unique solution for x and y. Let us call this point (h,k). If we now trans-
late the origin to (h,k), then by (8.11), (8.2) becomes, with respect to this
new origin (0,0),
(8.22) [al (% + h) + bl (7i + k) + CI] d%
+ + h) + b2(7i + k) + C2] d'U = 0,
which simplifies to
(8.23) [al% + bl'U + (alh + blk + CI)] d%
+ [a2% + b2'U + (a2h + b2k + C2)] d7i = 0.
M SPECIAL TYPES OF FIRST ORDER EQUATIONS Chapter 2

But (h,k) is the point of intersection of the two lines in (8.21) and there-
fore lies on both of them. Hence the term in the parentheses in each
bracket of (8.23) is zero. This equation therefore reduces to
(8.24)

which is now a homogeneous type solvable for % and '0 by the method of
Lesson 7. By (8.11) we can then find solutions in terms of x and y.
Note.
1. The left-hand members of the system (8.21) by which hand k are
determined are the coefficients in the given differential equation (8.2).
2. Equation (8.24) which is equivalent to (8.2) with respect to a new
origin translated to the point (h,k), can be easily obtained from (8.2).
Omit the constants Cl and C2 and place bars over x and y.
Example 8.25. Find a I-parameter family of solutions of
(a) (2x - y + 1) dx + (x + y) dy = 0.

Solution. The coefficient of dx is linear but nonhomogeneous, and

the two lines defined by the coefficients of dx and dy are nonparallel.
Hence the procedure outlined above applies. Solving simultaneously the
two equations determined by the coefficients of dx and dy, namely,
(b) 2x - y +1= 0,
x +y = 0,
we find that their point of intersection is (-1,1). Hence, in (8.11)
h = -1, k = 1. Translating the origin to the point (-1,1), the equa-
tions of translation are by (8.11) and (8.12)

(c) x = %- 1, y = '0 + 1, %= x + 1, '0 = y - 1·

By (8.24), (a) becomes with respect to this new origin (see Note 2 above)
(d) (2% - '0) d'f + (% + '0) d'O = 0.

To solve it, we apply the method of Lesson 7. Let

(e) '0 = U%, d'O = U d% + % du,

By substituting (e) in (d) and following the procedure outlined in Lesson
7, we obtain

(f) log 1%1 = Cl - 1_ Arc tan _ ":n - 21 log 12 + u 2 1, 0,

v2 v2
Lesson 8D EQUATIONS WITH LINEAR COEFFICIENTS 65

which is a I-parameter family of solutions of (d). In (f), we replace u by

its value in (e) and multiply by 2. There results

(g) log f2 (2 + = c -0 Arc tan J f ' f;;<! O.

Substituting the last two equations of (c) in (g) gives finally

(h)
Y
log 2 (3X: 9 + (3 ;- 9 =
2 2
C -y12 Arc tan 1) ,

I
x¢ - -.
3
Comment 8.26. The solution (h) above is an excellent example of
the point made at the beginning of this chapter in introductory remark
No.3. Here is a solution written in implicit form, which has little value
for practical purposes. To try to find the function g(x) implicitly defined
by this relation would be an extremely laborious if not a hopeless task.
In general a 1-parameter family of solutions of a differential equation
with linear coefficients or with homogeneous coefficients will usually be a
complicated expression of this kind. In these cases, more important for
practical purposes than an implicit solution is a knowledge of the approxi-
mate behavior of the integral curves. There are fortunately means avail-
able by which it is possible to determine the character of these integral
curves from the differential equation (8.S) itself, without the need to solve
it. Since differential equations with homogeneous or linear coefficients
arise in practical problems when trying to find an approximation to the
behavior of the motion of a particle whose velocities in the x and y direc-
tions are given by the two differential equations
dx
dt = f(x,y),

dy
dt = g(x,y),

we have deferred to Lesson 32 a further discussion of this important topic.

We shall show there, how it is possible to find an approximation of the
particle's motion by changing the two equations into one equation with
linear coefficients, and then showing how more useful information can be
obtained from the resulting differential equation itself than from its usual
complicated implicit solution. We shall thus be able to learn what the
solution (h) in the above example approximately looks like, not from this
solution, but from the given differential equation (a); see Example 32.44.
66 SPECIAL TYPES OF FIRST ORDER EQUATIONS Chapter 2

LESSON 8C. A Second Method of Solving the Differential Equa-

tion (8.2) with Nonhomogeneous Coefficients. In (8.2), let

(8.3) u = alX + bty + Ct,

v = a2X + b2 y + C2.

Therefore

(8.31) du = at dx + bt dy,
dv = a2 dx + b2 dye

Now solve (8.31) for dx and dye The substitution in (8.2) of (8.3) and these
values of dx and dy will also lead to a differential equation with homo-
geneous coefficients solvable by the method of Lesson 7.

Example 8.32. Find a I-parameter family of solutions of

(a) (2x - y + 1) dx + (x + y) dy = O.

Solution. As indicated in (8.3), we let

(b) u = 2x - y + 1, v = x + y.
Therefore

(c) du = 2dx - dy,

dv = dx + dy,
The solution of (c) for dx and dy is

(d) dx = du + dv, dy = _ du - 2 dv .
3 3

SUbstituting (b) and (d) in (a), we obtain

(e)

which simplifies to

(f) (u - v) du + (u + 2v) dv = O.

This equation is now of the type with homogeneous coefficients. Follow-

ing the method of Lesson 7, we let

(g) u = tv, du = t dv + v dt.

Lesson 8D COEFFICIENTS PARALLEL OR COINCIDENT LINES 67

Substituting these values in (f), we obtain

(h) (tv - v)(tdv + vdt) + (tv + 2v) dv = 0,

which reduces to

(i) vdv + t2t -+ 12 dt = 0, v ¢ o.

Its solution is
0) log Ivl + ! log (t
2
+ 2) - Arc tan = c, v;;<! 0,

log [V
2
(t
2
+ 2)] = c + v'2 Arc tan v;;<! O.

By (g) and (b),

(k) t = u = 2x - y + 1, x +y ¢ o.
v x +y
Substituting (k) in (j), we obtain
(1) log [(2x - y + 1)2 + 2(x + y)2] = C + v'2 Arc tan + 1,
y
2 (x + y)
x +y ¢ o.
LESSON SD. Solution of a Differential Equation in Which the
Coefficients of dx and dy Define Parallel or Coincident Lines.
If the lines defined by the coefficients of dx and dy in (8.2) are parallel, the
method of Lesson 8B will not work. Parallel lines do not have a point of
intersection and therefore (8.21) has no solution for x and y. In this case
we must resort to a different substitution. It is illustrated in the following
example.
Example 8.4. Find a I-parameter family of solutions of

(a) (2x + 3y - 1) dx + (4x + 6y + 2) dy = 0,

also any particular solution not obtainable from the family.

Solution. We observe that the lines defined by the coefficients of dx
and dy are parallel but not coincident lines. In all such cases, the substi-
tution of a new variable for the coefficient of dx or of dy will transform
the equation into one which is separable. We therefore let

(b) u = 2x + 3y - 1, du = 2dx + 3dy, dx = du - 3dy.

2
Then by (b)
(c) 2u +4 = 4x + 6y + 2.
68 TYPES OF FIRST ORDER EQUATIONS Chapter 2

Substituting (b) and (c) in (a), we obtain

(d) U(dU -; 3dY) + (2u + 4) dy = 0,

which simplifies to

(e) u du + (u + 8) dy = 0,

an equation Whose variables are separable. If u -8, (e) can be written

as
u
(f) u + 8 du + dy = 0, u ¢ -8.

Integration of (f) gives

(g) u - 8 log lu + 81 + y = c, u -8.

Finally, replace in (g) the value of u as given in (b), noting at the same
time that the exclusion of u = -8 implies the exclusion of the line
2x + +3y 7 = O. Hence (g) becomes

(h) 2x + 3y - 1 - 8 log 12x + 3y + 71 + y = c, 2x + 3y + 7 ¢ 0,

which is a I-parameter family of solutions of (a).

The function defined by

(i) 2x + 3y + 7 = 0,

which had to be excluded in obtaining (h) also satisfies (a). (Be sure to
verify it.) It is a particular solution not obtainable from the family til).
Example 8.41. Find a 1-parameter family of solutions of

(a) (2x + 3y + 2) dx + (4x + 6y + 4) dy = 0;

also any particular solution not obtainable from the family.

Solution. We observe that the coefficients in (a) define the same line.
If we exclude values of x and y for which

(b) 2x + 3y + 2 = 0,

we may divide (a) by it and obtain

(c) dx + 2dy = o.
Its solution is

(d) x + 2y = C,
Le8son 8-Exercise 69

which is valid for those values of x and y which do not lie on the line
2x + 3y + 2 = O. It is the required I-parameter family. However, the
function defined by
(e) 2x + 3y + 2 = 0
also satisfies (a). It is a particular solution not obtainable from the family.

EXERCISE 8
Find a l ...parameter family of solutions of each of the following equations.
1. (x + 2y - 4) dx - (2x - 4y) dy = O.
2. (ax + +2y 1) dx - (3x +2y - 1) dy = o.
3. (z + +
y +
1) dx (2x + +
2y 2) dy = o.
4. (x + y - 1) dx + (2x +
2y - 3) dy = O.
5. (x + y - 1) dx - (l: - Y - 1) dy O.
Ie:

6. (x + y) dx + +(2x 2y - 1) dy = O.
7. (7y - 3) dx +(2x + 1) dy = O.
8. (x + 2y) dx +
(ax + +
6y 3) dy = O.
9. (x + 2y) dx +
(y - 1) dy = O.
10. (3x - 2y +4) dx - (2x + 7y - 1) dy == O.

Find a particular solution, satisfying the initial condition, of each of

the following differential equations.
11. (x + y) dx + (3x + 3y - 4) dy = 0, y(l) == O.
12. (3x+ +2y 3) dx - (x + 2y - 1) dy = 0, y(-2) = 1.
13. (y + 7) dx + (2x +
y + 3) dy = 0, y(O) == 1.
14. (x + y + 2) - (x - y - 4) dy = 0, y(l) == O.

ANSWERS 8

1. log [4(y - 1)2 + (x - 2)2] - 2 Arc tan 2y - :

x-
= c.

2. log 115z + lOy - 11 + i(x - y) == c.

3. x + 2y = c.
+ +
4. x 2y log Ix + y - 21 = c.
5. (x - 1)2 + y2 = ce 2 Arc tan [y/(z-l».
6. x + 2y + log Ix + y - 1I == c.
7. 7 log 12x + 11 + 2 log 17y - 31 = c.
8. x + 3y - 3 log Ix + 2y + 31 = c.
9. [(x + 2)/(x + y + 1)] + log Ix + y + 11 = c.
10. 3x 2 - 4xy + 8x - 7y2 + 2y = c.
II. X + 3y + 2 log (2 - X - y) = 1.
12. (2x + 2y + 1)(3x - 2y + 9)4 = -1.
13. (y + 7)2(3x + y + 1) == 128.
x-I
14. log [(x - 1) + (y + 3) ] + 2 Arc tan y + 3 ==
2 2
210g 3.
70 SPECIAL TYPES OF FIRST ORDER EQUATIONS Chapter 2

LESSON 9. Exact Differential Equations.

Before beginning a study of this type of equation, we shall review those

concepts from the theory of integration which we shall need.

xo=a

Figure 9.1

1. Let f(x) be a function of x defined on an interval I: a < x S b.

Let I be divided into n subintervals and call dXk the width of the kth
subinterval, (Fig. 9.1). Then if
n
(9.11) lim L
k=l
f(Xk) dXk

exists as the number of subintervals increases in such a manner that the

largest subinterval approaches zero, we say that

(9.12)
r
k f(Xk) .:1Xk = Ja f(x) dx.
n
b

This limit is called the Riemann integral of f(x) over I. If this limit
does not exist, we say f(x) is not Riemann-integrable over I.
2. If f(x) is a continuous function of x on an interval I: a < x < b, and

(9.13) F(x) = 1'"

Zo
f(u) du,

then by the fundamental theorem of the calculus

(9.14) F'(x) = f(x), a < x < b,
or equivalently
(9.15) d
dX
1'"
Zo
f(u) du = f(x).

3. Let P(x,y) and Q(x,y) be functions of two independent variables,

x,y, both functions being defined on a common domain D. In Lesson 2C
Lesson 9 EXACT DIFFERENTIAL EQUATIONS 71

(we suggest your rereading this lesson), we showed that a two-dimensional

domain may assume various shapes. Hence if we wish to perform, for
example, the following integration, the first with respect to x (y constant),
and the second with respect to y (x constant),

(9.16) rJz o
P(x,y) dx -I- 11/ Q(x,y) dy,
Yo

we must be sure that (xo,Yo) is a point of D and that the rectangle deter-
mined by the line segments joining the points (xo,Yo), (x,Yo) and (xo,Yo),
(xo,Y) lies entirely in D. If the domain D is either of the regions shown
z z

Figure 9.17 Figure 9.18

in the shaded areas in Fig. 9.17 or 9.18, then you cannot integrate (9.16)
along the straight line from Xo to x or from Yo to Y because part of these
lines are not in D.
There is also another factor to be considered. If the domain D is the
region shown in Fig. 9.17, and the rectangle determined by the line seg-
ments joining the points (xo,Yo), (x,Yo) and (xo,Yo), (xo,Y) lies entirely in
D, then (9.16) can be integrated. If, however, the domain D is a region
with a hole in it such as in Fig. 9.18, then even if the rectangle were en-
tirely in D, (9.16) cannot be integrated because of this hole in its interior.
Therefore, we must in some way distinguish between these two types of
regions. The region in Fig. 9.17, i.e., the one which has no hole in it, is
called a simply connected region. Its formal definition is the following.
Definition 9.19. A region is called a simply connected region if
every simple closed curve lying entirely in the region encloses only points of
the region.
To summarize: We can perform the integrations called for in (9.16)
only if:
(a) The common domain of definition of the functions is a simply con-
nected region R.
72 SPECIAL TYPES OF FIRST ORDER EQUATIONS Chapter 2

(b) The point (xo,Yo) is in R.

(c) The rectangle determined by the lines joining the points (xo,Yo), (x,Yo)
and (xo,Yo), (xo,Y) lies entirely in R.

LESSON 9A. Definition of an Exact Differential and of an Exact

Differential Equation. We showed in Lesson 6B that if, for example,

(9.2) z = f(x,y) = 3x 2y + 5xy + y3 + 5,

then the differential of z [see (6.47)] is
(9.21)
dz = of(x,y) dx
ox
+ of(x,y)
oy
dy = (6xy + 5y) dx + (3x 2 + 5x + 3y2) dye
If therefore, we had started with the differential expression

(9.22) (6xy + 5y) dx + (3x 2 + 5x + 3y2) dy,

we would know that it was the total differential of the function f(x,y)
defined in (9.2). Differential expressions of the type (9.22); i.e., those
which are the total differentials of a function f(x,y), are called exact
differentials. Hence:
Definition 9.23. A differential expression

(9.24) P(x,y) dx + Q(x,y) dy

is called an exact differential if it is the total differential of a function
f(x,y), Le., if

(9.241)
o
P(x,y) = ox f(x,y) and Q(x,y)
a J(x,y).
= oy
Setting the differential expression (9.22) equal to zero, we obtain the
differential equation

(9.25) (6xy + 5y) dx + (3x 2 + 5x + 3y2) dy = 0,

whose solution, by (9.2), is

(9.26) f(x,y) = 3x 2y + 5xy + y3 = C,

valid for all values of x for which (9.26) defines y as an implicit function
of x and for which dy/dx exists. [If you have any doubt that (9.26) is an
implicit solution of (9.25) or that of(x,y) / ox is the dx coefficient in (9.25)
or that of(x,y) /oy is the dy coefficient in (9.25), verify these statements.]
A differential equation of the type (9.25), Le., one whose dx coefficient
is the partial derivative with respect to x of a function f(x,y) and whose dy
Lesson 9B SOLUTION OF AN EXACT EQUATION 73

coefficient is the partial derivative with respect to y of the same function,

is called an exact differential equation. Hence:
Definition 9.27. The differential equation

(9.28) P(x,y) dx + Q(x,y) dy = 0

is called exact if there exists a function f(x,y) such that its partial deriva-
tive with respect to x is P(x,y) and its partial derivative with respect to
y is Q(x,y). In symbolic notation, the definition says that (9.28) is an
exact differential equation if there exists a function f(x,y) such that

(9.29) aJ(x,y) = P(x y) = Q(x,y.).

ax ' ,
A I-parameter family of solutions of the exact differential equation (9.28)
is then
(9.291) f(x,y) = c.
In the example used above, we knew in advance that (9.25) was exact
and that the function defined in (9.26) was its solution, because we started
with this function f(x,y) and then set its total differential equal to zero to
obtain the differential equation (9.25). In general, however, if a first order
differential equation were selected at random, two questions would present
themselves. First, how would we know it was exact, and second, if it were
exact, how could we find the solution J(x,y) = c? The answer to both
questions is incorporated in the theorem and proof which follow.

LESSON 9B. Necessary and Sufficient Condition for Exactness

and Method of Solving an Exact Differential Equation.
Theorem 9.3. A necessary and sufficient condition that the differential
equation
(9.31) P(x,y) dx + Q(x,y) dy = 0

be exact is that
a a
(9.32) ay P(x,y) = ax Q(x,y),

where the functions defined by P(x,y) and Q(x,y), the partial derivatives in
(9.SS) and ap(x,y)/ax, CJQ(x,y)/ay exist and are continuous in a simply
connected region R.
NOTE. Although the theorem is valid as stated, the proof will be given
only for a rectangular domain contained entirely within the simply con-
nected region R.
74 SPECIAL 'I'YPES OF FIRST ORDER EQUATIONS Chapter 2

Proof of necessary condition, i.e., given (9.31) is exact, to prove (9.32).

Since (9.31) is exact, it follows from Definition 9.27 that there is a func-
tion f(x,y) such that
a
(9.33) a: f(x,y) = 'P(x,y), ay f(x,y) = Q(x,y).

Because of the assumptions about the functions P and Q stated after

(9.32) and by a theorem in analysis, we are permitted to assert that

1. (:x f(x,y) ) and :x (aay f(x,y) ) exist.

a a a a
2. ay ax f(x,y) = ax ay f(x,y),

i.e., the order in which we take the first and second partial derivatives of
f(x,y) is immaterial. Substituting (9.33) in 2, above, we obtain
a a
(9.34) ay P(x,y) = ax Q(x,y),
which is (9.32).

Proof of sufficient condition, i.e., given (9.32), to prove (9.31) is exact.

By Definition 9.27, the proof that (9.31) is exact is equivalent to proving
the existence of a function f(x,y) such that af/ax = P(x,y) and af/ay =
Q(x,y). [In the proof of this sufficient condition, we shall at the same
time discover the method of finding f(x,y).] Hence the function f(x,y) , if
it exists, must have the property that

(9.35) = P(x,y).

Therefore, with y constant, f(x,y), by (9.14) and (9.13), must be a function

such that
(9.36) f(x,y) = 1'"
%0
P(x,y) dx + R(y),
where Xo is a constant and R(y) stands for t1he arbitrary constant of inte-
gration. [Remember that in going from (9.26) to (9.35), R(y) and yare
constants.]
But this function f(x,y) must also have the property, by Definition 9.27,
that
a
(9.37) ay f(x,y) = Q(x,y).

Therefore, differentiating (9.36) with respect to y and setting the result

Lesson 9B SOLUTION OF AN EXACT EQUATION 75

equal to Q(x,y), we obtain in symbolic notation

(9.38) aa 1'"
y Zo
P(x,y) dx + R'(y) = Q(x,y).
By hypotheses P(x,y) is continuous. Hence, by a theorem in analysis, we
can, in (9.38), put the symbol ajay inside the integral sign. It will then
read

(9.381)
1'":
Zo vy
P(x,y) dx + R'(y) = Q(x,y).

By (9.32), we can write (9.381) as

(9.39)
1'":
Zo vX
Q(x,y) dx + R'(y) = Q(x,y).

Study the integral in (9.39) carefully. In words, it says: differentiate

Q(x,y) with respect to x, with y fixed, and then integrate this result with y
still fixed. The net result is to get back the function Q(x,y). [Try it for
the function Q(x,y) = x 2 y with y constant.] Hence (9.39) becomes

(9.4) Q(x,y)l:o + R'(y) = Q(x,y),

which simplifies to [remember Q(x,y) = Q(x,y) - Q(xo,y)]

(9.41) R'(y) = Q(xo,y).

Integration of (9.41) gives, by (9.14) and (9.13),

(9.42) R(y) = 11/ Q(xo, y) dy,

Yo

where Yo is a constant. Substituting (9.42) in (9.36), we obtain finally

(9.43) f(x,y) = 1'" P(x,y) dx + 11/ Q(xo,y) dy,

Zo Yo

where (xo,Yo) is a point in R, and the line segments joining the points
(xo,Yo), (x,Yo) and (xo,Yo), (xo,y) lie entirely in R.
This function f(x,y) we shall now show is the one we seek. Since the
second integral in (9.43) is a function of y, we have, by (9.13) and (9.14),

(9.44) = i: P(x,y) dx +0 = P(x,y).

76 TYPES OF FIRST ORDER EQUATIONS Chapter 2

And by following the steps from (9.38) to (9.4), we obtain

(9.441) !loa
uy Jzro P(x,y) dx = Q(x,y) - Q(xo,Y).

By (9.13) and (9.14)

(9.442) : /.11- Q(xo,Y) dy = Q(xo,Y).

y 110

Hence, by (9.43), (9.441) and (9.442), it follows that

a
(9.443) ay f(x,y) = Q(x,y).

We have thus not only proved the theorem, but have shown at the same
time how to find a I-parameter family of solutions of (9.31). It is, by
(9.291) and (9.43),

(9.45) f(x,y) =
Jzro P(x,y) dx + /."Yo Q(xo,Y) dy = c,
where (xo,Yo) is a point in R and the rectangle determined by the line seg-
ments joining the points (xo,Yo), (x,Yo) and (xo,Yo), (xo,Y) lies entirely in R.
Prove as an exercise that if in place of (9.35) we had started with
a
(9.46) ay f(x,y) = Q(x,y),

we would have obtained for the solution of (9.31)

(9.47) f(x,y) = /." Q(x,y) dy +

110 Jzro P(x,Yo) dx = c.

Remark. Both (9.45) and (9.47) will give a I-parameter family of

solutions of (9.31) . You may use whichever you find easiest in a particular
problem.
Example 9.5. Show that the following differential equation is exact
and find a I-parameter family of solutions.

(a) cos y dx - (x sin y - y2) dy = O.

Solution. Comparing (a) with (9.31) we see that P(x,y) = cos y and
Q(x,y) = -x sin y + y2. Therefore ap(x,y)/ay = -sin y and aQ(x,y)/ax =
-sin y. Since ap/ay = aQ/ax, the equation, by Theorem 9.3, is exact and
since P(x,y) and Q(x,y) are defined for all x,y, the region R is the whole
plane. Hence we may take Xo = 0 and Yo = o. With Xo = 0,
Lesson 9B SOLUTION OF AN EXACT EQUATION 77

Q(xo,Y) = Q(O,y) = y2. Thus (9.45) becomes

(b) foz cos y dx + foil y2 dy = c.

Integration of (b) gives (remember in the first integration y is a constant)
3
(c) = c,

which is the required I-parameter family.

Example 9.51. Show that the following differential equation is exact
and find a particular solution y(x) for which y(I) = o.

(a) (x - 2xy +e fl
) dx + (y - x2 + xe fl
) dy = o.
Solution. Comparing (a) with (9.31) we see that P(x,y) = x - 2xy
+ efl and Q(x,y) = y - x 2 + xe'll. Therefore ap/ay = -2x + e'll and
aQ/ax = -2x + efl • Since ap/ay = aQ/ax, the equation, by Theorem
9.3, is exact. And since the region R is the whole plane, we may take
Xo = 0 and Yo = O. With Xo = 0, Q(xo,Y) = y. Hence (9.45) becomes

(b) foz (x - 2xy + e") dx + foil Y dy = c.

Integration of (b) gives

x2 y2
(c) 2 - 2
x y + xe'll + -2 = c.

To obtain a particular solution for which x = 1, y = 0, we substitute

these values in (c) and find c = i. Hence the required particular solution is

(d)

Example 9.52. Show that the following equation is exact, and find a
I-parameter family of solutions.

(a) + xy2 sin 2x + y2 sin 2 x) dx + (2xy sin 2 x) dy = o.

(x 3

Solution. Here ap/ ay = 2xy sin 2x + 2y sin 2 x and

= 2y(2x sin x cos x + sin 2 x) = 2xy sin 2x + 2y sin 2 x.

Since ap/ay = aQ/ax, the equation by Theorem 9.3 is exact. In this case,
it will be found easier to use (9.47). Since the region R is the whole plane,
78 SPECIAL TYPES OF FIRST ORDER EQUATIONS Chapter 2

we may take Xo = 0 and Yo = 0 so that P(x,O) = x 3 • Hence (9.47)

becomes
(b) i" 2
2xy sin x dy + i'" 3
x dx = c.

Integration of (b) gives (remember that this time x is & constant in the
first integration)
X4
(c) xy2 sin 2 x + "4 = c.

which is the required 1-parameter family.

Remark. Formulas have the advantage of enabling one to obtain a
result relatively easily, but have the disadvantage of being easily forgot-
ten. We therefore outline a method by which you can solve any of the
above differential equations directly from Definition 9.27. In our opinion,
this method is the preferable one.
Example 9.6. Solve Example 9.5, using Definition 9.27.
Solution. The differential equation is
(a) cos y dx - (x sin y - y2) dy = o.
We have already proved (a) is exact. By Definition 9.27, therefore, there
exists a function f(x,y) such that

(b) aJ(x,y) = P(x,y) = cos y.

ax
Hence integrating (b) with respect to x, we obtain

(c) f(x,y) = f cos y dx + R(y) = x cos y + R(y).

Again by Definition 9.27, this function J(x,y) must also have the property
that

(d) = Q(x,y) = -x sin y + y2.

Hence differentiating the last expression in (c) partially with respect to y
and substituting this value in (d) we obtain
(e) -x sin y + R'(y) = -x sin y + y2,
which simplifies to

(f) R(y) = f y2 dy = 3. y3
Lesson 9-Exercise 79

Substituting (f) in (c) gives

y3
(g) f(x,y) = x cos y +3 ·
By (9.291), a I-parameter family of solutions of (a) is therefore
3
(h) xcosy + Y3 = c,
just as we found previously.
As an exercise, solve the other two Examples 9.51 and 9.52 directly from
the definition as we did above.

EXERCISE 9
1. Prove formula (9.47).
2. Solve Example 9.51 by the method of Example 9.6.
3. Solve Example 9.52 by the method of Example 9.6.
Show that each of the following differential equations 4-13 is exact and
find a I-parameter family of solutions using formula (9.45) or (9.47), and
also the method outlined in Example 9.6.
4. (3z 2y + 8xy2) dx + (x 3 + 8x 2y + 12y2) dy = O.

S. (2XYy+ 9 + (Y
dx x) dy = O.
6. 2zy dz + (z2 +
y2) dy = O.
7. siny + e- ) dx - (xe- lI -
II cos y dy) = o.
8. cos y dz - (x sin y - y2) dy = o.
9. (x - 2xy + ell ) dz + (y - x 2 + xe ll ) dy = O.
10. (x 2 - Z + y2) dz - (e ll - 2xy) dy = O.
11. (2x + y cos x) dx + (2y + sin x - sin y) dy = O.
2
12. xVz2 + y2 dx - x Y dy = O.
Y - v'x2 + y2
13. (4z3 - sin z + y3) dz - (y2 +1- 3xy2) dy = O.
14. If J(z,y) is the function defined in (9.43), prove (9.44).
Find a particular solution, satisfying the initial condition, of each of the
following differential equations.
15. + zy3 + 1) dz + 6) dy = 0, y(O)
- = 1.
16. sin z cos y dx + cos z sin y dy = 0, Y(1r/4) = 1r/4.
17. + 4x3 ) dz + - 3y2) dy = 0, y(l) = O.
ANSWERS 9
4. z3y+ 4z2y2 + 4y3 = c. 7. + ze- = c.
sin y II

5. z + - + log Iyl =
2 Z
y
c. 8. 3z cos y + y3 = c.
6. 3z 2y + y3 z - 2x y + y + 2ze = c.
2 2 2 II
= c. 9.
80 SPECIAL TYPES OF FIRST ORDER EQUATIONS Chapter 2

10. 2x 3 - 3x 2 + 6xy2 - 6e" = c. 11. x 2 + y sin x + y2 + cos y == c.

12. +
(x 2 y2)3/2+ y3 = c.
13. +
3x 4 + 3 cos x 3y3x - y3 - 3y = c.
15. xe z y3
+ e - 6y3 = -5.
Z
16. 2 cos x cos y = 1.
17. +
eZII2 z4 - y3 = 2.

LESSON 10. Recognizable Exact Differential Equations.

Integrating Factors.

LESSON lOA. Recognizable Exact Differential Equations. It is

sometimes possible to recognize the solution f(x,y) = c of an exact differ-
ential equation without the necessity of resorting to the methods of
Lesson 9b. For example, if the exact differential equation is
(10.1)

you might be able to recognize that its solution is

(10.11)

If you cannot, you must of course use the method of solution outlined in
the previous lesson.
We list below a number of exact differential equations and their solu-
tions. It will be profitable for you to verify some of them by taking the
total differential of the function on the right and seeing if it yields the
differential equation on the left.

Exact Differential Equation Solution

(10.2) y dx + x dy = 0 xy == c
(10.21) 2xy dx+ x 2 dy = 0 Z2y == c
(10.22) y2 dx+ 2xy dy = 0 xy2 = c
(10.23) 2xy2 dx+ 2x 2y dy = 0 x 2y2 == c
(10.24) +
3x 2y 3 dx 3x3y2 dy = 0 Z3 y 3 = c
(10:25) 3x 2y dx + x 3 dy = 0 z3y == c
(10.26) y cos x dx+ sin x dy = 0
.
y sin x == c
(10.27) sin y dx+ x cos y dy = 0 z sin y == c
(10.28) zlI
ye dx + xe ZlI dy = 0 eZlI == c
(10.29) dx + dy = 0 log (xy) = c
x Y
Y dx - x dy X
(10.3)
y2 = 0 - =C
Y
Y dx - x dy
(10.31) -
x2
= 0 -Y
X
= C

2xy dy - y 2 dx Y
2
(10.32)
x2
=0 -Z = C
Lesson lOA RECOGNIZABLE EXACT DIFFERENTIAL EQUATIONS 81

&act Differential Equation Solution

2 2
2zydz - z dy = 0 Z
(10.33) - = c
y2 y
Zdy-ydz=O y
(10.34) Arc tan - = c
Z2 + y2 z
2zdy-ydz=0 z+y
(10.35) log = c
Z2 - y2 z-y
2 3
3zydz-z dy =0 y
(10.36) - = C
Z6 z3
2
Y dz - 2zy dy = 0 z
(10.37) - = C
y4 y2

(10.38) _ Y dz + z dy = 0
-
1
= c
Z2 y 2 zy
3 2 3
(10.39)
Y dz - zy dy = 0 Y
-= C
Z4 3z3
(10.391) Z + 'II dy = 0
dx v'Z2 + y2 = C
v'Z2 + y2
(10.392) e3z dy + 3e 3z y dz = 0 e3z y = c
The exact differential equations on the left are sometimes referred to
as integrable combinations. We encourage you to add to this list
whenever you discover the solution of a new integrable combination.
If a differential equation is exact, but no integrable combination is
readily apparent to you, you can always fall back on the method of solu-
tion outlined in Lesson 9B. However, it is sometimes possible, if an equa-
tion is exact, to solve it more readily and easily by a judicious rearrange-
ment of terms so as to take advantage of any integrable combinations it
may possess.
Note. Hereafter ,vhen we use the word solve, in connection with first
order equations, we shall mean "find a I-parameter family of solutions of
the given differential equation."
Example 10.4. Solve

(a) xy : 1 dx + 2y y-; x dy = 0, y o.
Solution. By (9.31), P(x, y), Q(x, y) are the respective coefficients of
dx, dYe Therefore,
(b) ap = _ l. and aQ = _ 1.. .
ay y2 ax y2
Hence the equation is exact. A solution therefore can be obtained by
Theorem 9.3. However, if we rewrite the equation as

(c) x dx + yI dx + 2y dy -
x
y2 dy = 0, y 0,
82 SPECIAL TYPES OF FIRST ORDER EQUATIONS Chapter 2

which can be put in the form

(d) xdx 2d +ydx-xdy 0

=, y 7" 0,
+- y 2
Y Y
we observe that the last term of the left side is, by (10.30), d(x/y). Each
of the other terms can be integrated individually. Hence integration of
(d) yields
x2 X
(e) 2 + 2 log 'yl + y = c, y¢ 0,

which is the required solution.

Example 10.41. Solve
(a)
Solution. You can easily verify that (a) is exact. It therefore can
be solved by the method of Lesson 9B. However, if we rewrite the equa-
tion as
(b)

we observe that the sum of the first two tenns is by (10.392), d(e 3Z y),
and that the third term can be integrated individually. Hence integration
of (b) yields the solution
(c)
Example 10.42. Solve
(a) (x - 2xy + ell) dx + (y - x 2 + xell ) dy = O.
Solution. We have already shown (see Example 9.51) that this
equation is exact, and solved it by use of Lesson 9B. However, if we
rewri te it as
(b) x dx - (2xy dx + x 2 dy) + (ell dx + xell dy) + y dy = 0,

the second expression, by (10.21), is d(x 2 y) and the third expression is

recognizable as d(xe ll ). The solution of (b) is therefore
x2 2 ll
y2
(c) 2 - x y + xe + "2 = c.

just as we found previously.

LESSON lOB. Integrating Factors.

Definition 10.5. A multiplying factor which will convert an inexact

differential equation into an exact one is called an integrating lactor.
Lesson lOB INTEGRATING FACTORS 83

For example, the equation (y2 + y) dx - x dy = 0 is not exact. If,

however, we multiply it by y-2, the resulting equation

(1 +Y!) dx - -=-
y2 dy = 0, y 0,

is exact. (Verify it.) Hence by Definition 10.5, y-2 is an integrating

factor.
Remark. Theoretically an integrating factor exists for every differ-
ential equation of the form P(x,y) dx +
Q(x,y) dy = 0, but no general
rule is known to discover it. Methods have been devised for finding
integrating factors for certain special types of differential equations, but
the types are so special that the methods are of little practical value. It
is evident that if a standard method of finding integrating factors were
available then every first order equation of this form would be solvable
by this means. Unfortunately this is not the case. However, in the next
lesson we shall discuss a special, important, first order differential equation
for which an integrating factor is known and for which a standard method
for finding it exists.
In the meantime, we shall show by a few examples how an integrating
factor, if you are shrewd enough to discover one, can help you solve a
differential equation.
Example 10.51. Solve

(a) (y2 + y) dx - x dy = o.
Solution. We proved to you above that y-2 is an integrating factor
of (a). Hence multiplication of (a) by y-2 will convert it into the exact
differential equation

(b) (1 + Ddx - ;2 dy = 0, y;;<! O.

We can now solve (b) by means of Lesson 9B or lOA. If we rearrange the

terms to read
(c) dx + y dx y x dy = 0,

the second term by (10.3) is d(x/y). Hence by integration of (c) we

obtain the solution
(d) x + -xy = c·, y = c-x
x
,y o.
NOTE. The curve y = 0 also satisfies (a). It is a particular solution
not obtainable from the family (d).
84 SPECIAL TYPES OF FIRST ORDER EQUATIONS Chapter 2

Example 10.52. Solve

(a) y sec x dx + sin x dy = 0, x

'Tr
2 '
3'Tr
2'· · · ·
Solution. First verify that (a) is not exact. However, sec x is an
integrating factor. Multiplication of (a) by it will therefore yield the
exact equation
(b)
2
y sec x dx + tan x dy = 0, x
'Tr
2 '
3'Tr
2' · · · ·
[Now verify that (b) is exact.] The solution of (b) by means of Lesson 9B
or by recognizing that its left side is an integrable combination is

(c) y tan x = c or y = c cot x.

LESSON IOe. Finding an Integrating Factor. As noted in the
remark following Definition 10.5, a standard method of finding an inte-
grating factor is known only for certain very special types of differential
equations. We shall discuss some of these special types below.
We assume that

(10.6) P(x,y) dx + Q(x,y) dy = 0

is not an exact differential equation and that h is an integrating factor of

(10.6), where h is an unknown function which we wish to determine.
Hence, by Definition 10.5,

(10.61) hP(x,y) dx + hQ(x,y) dy = 0

is exact. It therefore follows, by Theorem 9.3, that

(10.62)
o 0
oy [hP(x,y)] = ox [hQ(x,y)].

We consider five possibilities.

1. h is a function only of x, i.e., h = h(x). In this case we obtain
from (10.62),
o 0 dh(x)
(10.63) hex) oy P(x,y) = hex) ox Q(x,y) + Q(x,y) dx '

which we can write as

dh(x) - :x
Q(x,y) ]
(10.64) hex) = Q(x,y) dx.

In the special case that the coefficient of dx in (10.64) also simplifies to a

Lesson IOC FINDING AN INTEGRATING FACTOR 85

function only of x, let us call it F(x), so that

aya P(x,y) - axa Q(x,y)

(10.65) F(x) = Q(x,y) ,

then, by (10.64) and (10.65), log [hex)] = f F(x) dx. Hence,

(10.66)
where we have omitted the constant of integration, is an integrating factor
of (10.6).
Example 10.661. First show that the differential equation

(a) (e Z - sin y) dx + cos y dy = 0

is not exact and then find an integrating factor.

Solution. Comparing (a) with (10.6), we see that

(b) P(x,y) = eZ - sin y, Q(x,y) = cos y.

Therefore
(c) ap(x,y) = -cos y aQ(x,y) = o.
ay , ax

Hence, by Theorem 9.3, (a) is not exact. By (b), (c), and (10.65),

(d) F(x) = -cos Y = -1.

cosy

Therefore, by (d) and (10.66),

(e) h(X ) = ef -dz = e-z
is an integrating factor of (a). (Verify it.)
2. h is afunction only of y, i.e., h == h(y). In this case we obtain
from (10.62),
a
(10.67) hey) ay P(x,y) + P(x,y) dh(y)
dy
a
= hey) ax Q(x,y),

which we can write as

dh(y) axa Q(x,y) - aya P(x,y)

(10.68) hey) = P(x,y) dye

If the coefficient of dy in (10.68) also simplifies to a function only of y-

86 SPECIAL TYPES OF FIRST ORDER EQUATIONS Chapter 2

let us call it G(y)-so that

a a
;- Q(x,y) - ;- P(x,y)
(10.69) G( ) _ ux uy
y - JD(x,y)

then, by (10.68) and (10.69), log [h(y)] = fG(y) dye Hence,

(10.7)

is an integrating factor of (10.6).

Example 10.701. First show that the differential equation
(a) xy dx + (1 + x 2 ) dy = 0

is not exact and then find an integrating factor.

Solution. Comparing (a) with (10.6), we see that

(b) P(x,y) = xy, Q(x,y) = 1 + x2•

Therefore
ap(x,y) aQ(x,y) _ 2
(c) = x, ax - x.
ay

Hence, by Theorem 9.3, (a) is not exact. By (b), (c), and (10.69),
2x - x 1
(d) G(y) = = -.
xy y
Therefore, by (d) and (10.7),
f!dU
(e) h(y) = e U = e10g U = Y

is an integrating factor of (a). (Verify it.)

3. h is a Junction of xy, i.e., h == h(u), where u = xy. In this
case we obtain from (10.62),
(10.71)
h(u) P(x,y) + P(x,y) [:y h(U)] = h(u) :x Q(x,y) + Q(x,y) [:x h(U)]'
.
Smce u = xy, ay
au = x. Therefore, ay h(u)
a
=
au =
h'(u) ay
d
x du h(u).

Similarly, :: = y and :x h(u) = h'(U) :: = y fu leu). Substituting in

Lesson IOC FINDING AN INTEGRATING FACTOR 87

(10.71) these values of k(u) and h(u) and simplifying the result,
we obtain
d[h(u)] aua P(x,y) - axa Q(x,y)
(10.72) h(u) = yQ(x,y) - xP(x,y) dUe

If the coefficient of du in (10.72) also simplifies to a function of u = xy-

let us call it F(u) == F(xy)-so that

aa P(x,y) - aa Q(x,y)
(10.73) F(u) = yyQ(X,y) - ,

then, by (10.72) and (10.73), log [h(u)] = fF(u) dUe Hence,

(10.74)
is an integrating factor of (10.6), where u = xy.
Example 10.741. First show that the differential equation
(a) (y3 + xy2 + y) dx + (x 3 + x 2y + x) dy = 0

is not exact and then find an integrating factor.

Solution. Comparing (a) with (10.6), we see that
(b) P(x,y) = y3 + xy2 + y,
Therefore

(c) ap(x,y) _ 3 2
ay - y
+ 2xy + 1, aQ(x,y) = 3x 2 + 2xy + 1.
ax
Hence, by Theorem 9.3, (a) is not exact. By (b), (c), and (10.73),

(d) F(u) -
3y2 +
2xy +1- 3x 2 - 2xy - 1
-
-3(x 2 _ y2)
- x 3y + x2y2 + xy - xy3 - x2y2 - xy - xy(x 2 - y2)
3
= --.
u
Therefore, by (d) and (10.74),
J-!du
(e) h(u) = e u = e- 3 logu = u- 3 = (xy)-3

is an integrating factor of (a). (Verify it.)

88 SPECIAL TYPES OF FIRST ORDER EQUATIONS Chapter 2

4. h is a function of x/y, i.e., h = h(u), where u = x/y. Here

u = x/y so that ou/oy = -x/y2 and ou/ox=l/y. Therefore
o ,ou xd
(10.75) oy h(u) = h (u) oy = - y2 du h(u),

o ou
;- h(u) = h'(u) ;-
{IX (IX
= -Y1 d
-d h(u).
u

Substituting (10.75) in (10.71) and simplifying the result, we obtain

2 [oP(x,y) _ oQ(X,y)]
d[h(u)] _ y ay ax du
(10.76)
h(u) - xP(x,y) + yQ(x,y) ·

If the coefficient of du in (10.76) also simplifies to a function of u = x/y-

let us call it G(u) = G(x/y)-so that

2 [oP(x,y) _ oQ(X,y)]
(10.77) G(u) = y ay ax
+ yQ(x,y)
I
xP(x,y)
then, by (10.76) and (10.77), log [h(u)] = f G(u) duo Hence,
(10.78) h(u) = efG(U) du,

is an integrating factor of (10.6), where u = x/Yo

Example 10.781. First show that the differential equation
(a) 3ydx - xdy = 0
is not exact and then find an integrating factor.
Solution. Comparing (a) with (10.6), we see that
(b) P(x,y) = 3y, Q(x,y) = -x.
Therefore
(c) oP(x,y) = 3 oQ(x,y) = -1
oy , ox '
Hence, by Theorem 9.3, (a) is not exact. By (b), (c), and (10.77),

(d) G(u) = + 1)
y2(3 = 2 y = 2.
3xy-xy xu
Therefore, by (d) and (10.78),

(e) )
h(u=e
f u
2 du
=e
log u 2
= u =y2
-
2 X
2

is an integrating factor of (a). (Verify it.)

Lesson IOC FINDING AN INTEGRATING FACTOR 89

5. h is a function of y/x, i.e., h = h(u), where u = y/x. In this

case, we leave it to you as an exercise-follow the method used in 4r-to
show that an integrating factor of (10.6) is

(10.79)
where u = y/x and
x 2 [oQ(X,y) _ oP(X,y)]
(10.8) K u _ ox oy
( ) - xP(x,y) yQ(x,y)+
Example 10.81. First show that the differential equation
(a) ydx - 3xdy = 0

is not exact and then find an integrating factor.

Solution. Comparing (a) with (10.6), we see that
(b) P(x,y) = y, Q(x,y) = -3x.
Therefore
(c) oP(x,y) = 1 oQ(x,y) = -3
oy , ox .

Hence, by Theorem 9.3, (a) is not exact. By (b), (c), and (10.8),

(d) K(u) = x2[-3 - 1] = 2x = 2.

xy - 3xy y u
Therefore, by (d) and (10.79),

(e) h(u) = e
f! du
U = e
10g
u
2
= U
2
= 2
y2
x
is an integrating factor of (a). (Verify it.)
If a differential equation can be put in the special form

(10.82)
where A, B, C, D are constants, then it be shown that an integrating
factor of (10.82) has the form xayb where a and b are suitably chosen
constants. We illustrate, by an example, the method of finding an inte-
grating factor of (10.82).
Example 10.83. First show that the differential equation
(a)

is not exact and then find an integrating factor.

90 SPECIAL TYPES OF FIRST ORDER EQUATIONS Chapter 2

Solution. Comparing (a) with (10.6), we see that

(b) P(x,y) = 2x2y4 + 3y, Q(x,y) = XSy3 - x.
Therefore

(c) oP(x,y) -_ 8x 2y 3 + 3,
oy
Hence, by Theorem 9.3, (a) is not exact. Since (a) has the form of (10.82),
an integrating factor will have the form xa,yb. Multiplying (a) by xa,y", we
have
(d) (2Xa,+2 y b+4 +
3xa yb+ 1) dx + (Xa,+3 y b+S - xa,+ ly") dy = o.

By Theorem 9.3, (d) will be exact if

(e) 2(b + 4)xa +2 yb+3 + (b + 1)3xa,y b = (a + 3)Xa,+2y b+3 - (a + l)xa y".
Multiplying (e) by 1/(xa,yb), we obtain
(f) (2b + 8)x 2 y3 + 3b + 3 = (a + 3)X 2y 3 - (a + 1).
Equation (f) will be an equality if we choose a and b so that
(g) 2b + 8 = a + 3, 3b +3 = -a - 1.
Solving (g) for a and b, we find
(h) a = I, b = -f·
Hence, x 7/5 y-9/ 5 is an integrating factor of (a). (Verify it.)

EXERCISE 10
Test each of the following equations 1-19 for exactness. If it is not exact,
try to find an integrating factor. (Integrating factors for nonexact equa-
tions are given in the answers.) After the equation is made exact, solve by
looking for integrable combinations. If you cannot find any, use method of
Lesson 9.
1. (2xy + x 2) dx + (x 2 + y2) dy = O.
2. (x 2 + y cos x) dx + (y3 + sin x) dy = o.
3. (x 2 + y2 + x) dx + xy dy = O.
4. (x - 2xy + efl ) dx + +
(y - x 2 xe fl ) dy = o.
5. (e z sin y + e-fl ) dx - (xe- fl - eZ cos y) dy = o.
6. (x 2 - y2 - y) dx - (x 2 - y2 - x) dy = O.
7. (x 4 y2 - y) dx + (x 2y 4 - x) dy = O.
8. y(2x + y3) dx - x(2x - y3) dy = O.
2 2
xy - 2Xy2) x - 2x y
9. ( Arc tan xy + + 1 x2y2 dx + + 1 x2y2 dy = O.
Leason IIA DEFINITION OF A FmST ORDER LINEAR EQUATION 91

10. eS(z + 1) dz + (ye" - zes) dy == O.

11. Z1I +
y
1 tk + 21/ -;
Y
2: d1/ _ O.
12. (y2 - 3zy - 2Z2) dz +
(zy - Z2) dy == O.
13. y(y +2z + 1) dz - z(2y +
z - 1) dy == O.
14. y(2z - y - 1) dz + z(2y - z - 1) dy == o.
15. (y2 + + +
I2z2y) dz (2zy 42;3) dy == O.
16. 3(y + +
z)2 dz + z(3y 2z) dy == O.
17. y dz - (y2+ + Z2 z) dy == O.
18. 2zy dz + + +
(Z2 y2 a) dy == O.
19. (2zy + + + + +
Z2 b) dx (y2 z2 a) dy == O.

ANSWERS 10
1. 3z2y + z3 + y3 == c.
2. 4z 3 + 3y' + I2y sin z == c.
3. 3z4 + 4z + 6z 2y2 == Cj integrating factor z.
3

4. z2 - 2z2y + y2 + 2xe" = c.
5. eS sin y + ze- II == c.
6. z - y + logv'z + y - logv'z - y == Cj integrating factor I/(z2 _ y2).
7. z4y + zy4 - czy == -3; integrating factor I/z2y2.
8. Z2 + zy3 == cy2; integrating factor l/y3.
9. z Arc tan zy - log (1 + z2y2) == c.
10. 2zes -" + y2 = Cj integrating factor e- II •
2z
11. z
2
+-y
+ 4 log Iyl == c.
12. z2y2 - 2z 3 y - z4 = Cj integrating factor 2z.
13. (y - z + 1)3 == cZyj integrating factor (zy) -4/3.
14. (z + +
y 1)3 == cZyj integrating factor z-ly-l(z + + 1)-1.
y
15. 4z3y + zy2 = c.
16. 6z 2y2 + 8z 3 y + 3z4 = Cj integrating factor z.
17. y + Arc tan Y.. = Cj integrating factor I/(z2 + y2).
z
18. y3 + 3z 2y + 3ay = c.
19. y3 + z3 + 3(z2y + ay + bz) = c.

LESSON 11. The Linear Differential Equation of the

First Order. Bernoulli Equation.

LESSON IIA. Definition of a Linear Differential Equation of the

First Order. The important differential equation which we are about
to discuss has many theoretical and practical applications. It is a special
type of first order differential equation in which both the dependent vari-
able and its derivative are of the first degree. An equation of this type is
called a linear differential equation of the first order. Hence:
92 SPECIAL TYPES OF FIRST ORDER EQUATIONS Chapter 2

Definition 11.1. A linear differential equation of the first order

is one which can be written as
dy
(11.11) dx + P(x)y = Q(x),

where P(x) and Q(x) are continuous functions of x over the intervals for
which solutions are sought. (Note that y and its derivative both have
exponent one.)
For this differential equation we shall prove in Lesson lIB that the
I-parameter family of solutions we shall obtain is actually a true general
solution as we defined the term in 4.7. Every particular solution of (11.11)
will be obtainable from this I-parameter family of solutions.

LESSON lIB. Method of Solution of a Linear Differential Equa-

tion of the First Order. As mentioned in Lesson lOB, an integrating
factor is known for this type of equation (11.11). It is
fp(x)dx
(11.12) e ,
where the constant of integration is taken to be zero.
The motivation and means by which this rather terrifying looking inte-
grating factor was obtained have been deferred to Lesson 11 C. In the
meantime let us verify that (11.12) is indeed an integrating factor for
(11.11). Multiplying (11.11) by (11.12) and changing the order in which
the tenns appear, we obtain
(11.13) [P(x)efp(X)dX y - Q(x)efp(X)dX] dx + Jp(x)dx dy = O.

The terms P(x)efp<x)dx and Q(x)ef<x)dx are functions of x. Hence the

partial derivative with respect to y of the coefficient of dx in (11.13) is

(11.14)
By (9.15) the partial derivative with respect to x of the coefficient of dy
in (11.13) is [remember ! e" = e" :: ; here u = f P(x) dx ]
(11.15)
Since the functions in (11.14) and (11.15) are the same, by Theorem 9.3,
(11.13) is exact. Hence by Definition 10.5, efp<x)dx is an integrating factor.
Let us now rewrite (11.13) in the form

(11.16)
Since we know (11.16) is exact, we can solve it either by the method of
Lesson 9B, or by trying to discover an integrable combination. A shrewd
Lesson lIB SOLUTION OF A FIRST ORDER LINEAR EQUATION 93

observer now discerns that the left side of (11.16) is indeed

(11.17) d(efp(X)dX y ).

[Be sure to verify this statement. Remember d(uy) = u dy + y duo Here

u = Jp(x)dx.] Hence (11.16) becomes

(11.18)
whose solution is
(11.19) efpC"'ld"'y = f e fpC"'ld"'Q (x) dx + c.
Proof that (11.19) is a true general solution of (11.11). The argu-
ment proceeds as follows. Since efp(x)dx ¢ 0, [for proof see Lesson 18,
(18.86)], (11.11) holds if and only if (11.13) holds; (11.13) holds if and only
if (11.16) holds; (11.16) holds if and only if (11.18) holds. Finally (11.18)
holds if and only if (11.19) holds. We have thus demonstrated that (11.11)
is true if and only if (11.19) is true. And since Jp(x)dx ¢ 0, we can in
(11.19) divide by this factor to obtain a I-parameter family of solutions
of (11.11) in the explicit form

(11.191)

Hence (11.11) holds if and only if (11.191) holds. The "if and only if"
clause is equivalent to saying that if (11.11) holds, then (11.191) holds,
and if (11.191) holds, then (11.11) holds. This means that if (11.11) is
true, then its solutions are (11.191), and if (11.191) is true (i.e., true for
each c), then (11.11) is satisfied.
Example 11.2. Find the general solution of
(a) y' - 2xy = ex2 •

Solution. By comparing (a) with (11.11), we see that the equation

is linear, P(x) = -2x and Q(x) = ex2 • By (11.12) an integrating factor
of (a) is therefore
eJ-2x dx = e_x .
2
(b)
Multiplication of (a) by e-x2 gives

(c)

an equation which now corresponds to (11.16). Hence by (11.17), the left

side of (c) should be (and is)
(d)
94 SPECIAL TYPES OF FIRST ORDER EQUATIONS Chapter 2

Replacing the left side of (c) by (d), and integrating the resulting expres-
sion, we obtain for the general solution of (a),

(e) e-Z y
2
= fdX = x + c,
which we can write as
(f)

Comment 11.21. After the integrating factor e-zl has been deter-
mined, we could by (11.19) go immediately from (a) to (e). For (11.19)
says, tty times the integrating factor = f
(integrating factor) Q(x) dx + c. "
Example 11.3. Find the particular solution of

(a) xdy
- + ay = sin x , x 0,
dx x2
for which Y('Ir/2) = 1.
Solution. Since x 0, we may divide (a) by it and obtain

(b) dy
-dx
+ -y=
a sin x
3'X,-.
0
X X

Comparing (b) with (11.11) we see that (b) is linear, P(x) = a/x and
Q(x) = sin x/x 3 • Hence by (11.12) an integrating factor is

ef x
Jdz 3
(c) = e310gz = e 10gz = x3•
.
(By taking the logarithm of both sides, you can show that e10KU = u.)
Multiplication of (b) by the integrating factor x 3 will give, by (11.19)
[see Comment (11.21)],

(d) X3 y = f sin x dx +c = -cos x + c.

To find the particular solution for which x = 'Ir/2, Y = 1, we substitute
these values in (d) and obtain
'lr3
(e) C =-.
8
Replacing (e) in (d), the required solution is
'lr3
(f) yx
3
+ cos X = 8".
LESSON IIC. Determination of the Integrating Factor Jp(%)d%.
We outline below a method by which the integrating factor for (11.11)
can be determined. First we rewrite (11.11) as
(11.4) [P(x)y - Q(x)] dx + dy = 0,
Lesson lID BERNOULLI EQUATION 95

and then multiply it by u(x). There results

(11.41) [u(x)P(x)y - u(x)Q(x)] dx + u(x) dy = o.

We now ask ourselves this question. What must u(x) look like if it is to
be an integrating factor for (11.11)? We know by Definition 10.5 that it
will be an integrating factor if it makes (11.41) exact. And by Theorem
9.3, we know that (11.41) will be exact if

(11.42)
a
ay [u(x)P(x)y - u(x)Q(x)] =
a
ax u(x).

Taking these derivatives [observe that u(x), P(x), and Q(x) are functions
only of x], we obtain
d
(11.43) u(x)P(x) = dx u(x),
which we can write as
(11.44) P(x) dx = du(x) .
u(x)

Integration of (11.44), with the constant of integration taken to be zero,

gives (remember f d: = log u )

(11.45) log u(x) = f P(x) dx,

which is equivalent to
(11.46)
Hence if u(x) has the value efp(x)dx, it will be an integrating factor for
(11.11).

LESSON lID. Bernoulli Equation. A special type of first order dif-

ferential equation, named for the Swiss mathematician James Bernoulli
(1654-1705), and solvable by the methods of this lesson is the following.

(11.5) : + P(x)y = Q(x)y",

If n = 1, (11.5) can be written as dy/dx = [Q(x) - P(x)]y, an equation
in which the variables are separable and therefore solvable by the method
of Lesson 6C. Hence we assume n 1. Note also that the presence of
y'" prevents the equation from being linear. If we multiply (11.5) by
(11.51)
we obtain
(11.52) (1 - n)y- : + (1 - n)P(x)(yl-"') = (1 - n)Q(x).
96 SPECIAL TYPES OF FIRST ORDER EQUATIONS Chapter 2

The first tenn in (11.52) is ! (yl-n). Hence (11.52) can be written as

(11.53) ! (yl-n) + (1 - n)P(x)(yl-n) = (1 - n)Q(x).

If we now think of yl-n as the dependent variable instead of the usual y

[or if you prefer you can replace yl-n by a new variable u so that (11.53)
becomes ;:: + (1 - n)P(x)u = (1 - n)Q(x)], then, by Definition 11.1,
(11.53) is linear in yl-n (or u). It can therefore be solved by the method
of Lesson lIB.
Example 11.54. Solve
(a) y '+ xy = -y3x , y o.
Solution. Comparing (a) with (11.5), we see that (a) is a Bernoulli
equation with n = -3. Hence, by (11.51), we must multiply (a) by 4 y 3.
There results
(b)

Because of the sentence after (11.52), we know that the first term of (b)
should be (and is)

(c)

Hence we can write (b) as

d
(d) dx [y 4] + 4x[y 4] = 4x,

an equation which is now linear in the variable y4. An integrating factor

for (d) is therefore, by (11.12),

(e)

Mter multiplying (d) by e2x2 , we can take advantage of (11.19) to write

immediately
(f)
Therefore,
(g)

is the required solution.

Lesson II-Exercise 97
EXERCISE 11
Find the general solution of each of the following.
1. xY' + Y = x 3•
2. y' + ay = b.
3. xy' +y = y2 log x.

4. : + 2yx = e-1/ Hint. Consider x as the dependent variable.

2
•

5. dB = (r +-1
dr e) tan B.

6 dy _ 2xy
• dx x2 +
1 - 1.
7. y' +
y = xy3.
3 dy
8. (1 - x ) dx - 2(1 + x)y = y
5/2
•
dr 2
9. tan B dB - r = tan B.

10. L + Ri = E sin kt. (This is the equation of a simple electric circuit

containing an inductor, a resistor, and an applied electromotive force. For
the meaning of these terms and for a Inore complete discussion of electric
circuits, see Lessons 30 and 33C.)
11. y' + 2y = 3e- 2z . 15. y' y cos x = i sin 2x. +
12. y' + 2y = !e- 2z . 16. xy' y = x sin x. +
13. y' + 2y = sin x. 17. xy' - y = x 2 sin x.
14. y' + y cos x = e2z . 18. xy' xy2 - Y = O. +
19. xy' - y(2y log x-I) = o.
20. x 2(x - l)y' - y2 - x(x - 2)y = O.
Find a particular solution of each of the differential equations 21-24.
21. y' - y = eZ , yeO) = 1.
2
22. y' + !x y x
= y(-l) = 1.
23. 2 cos x dy = (y sin x - y3) dx, yeO) = 1.
24. (x - sin y) dy +
tan y dx = 0, y(l) = 1r/6.
25. The differential equation
(11.6)
is called a Riccati equation. If Yl(X) is a particular solution of this equa-
tion, show that the substitution
1 , , 1,
(11.61) y = Yl + -,
u
y = Yl - -2 U
u '

will transform the equation into the first order linear equation
(11.62)
Hint. Since Yl is a particular solution of the given equation, Yl' -
fo(x) +fl(X)Yl +
f2(X) y1 2.
98 SPECIAL TYPES OF FIRST ORDER EQUATIONS Chapter 2

With the aid of problem 25 above, find the general solution of each of
the following Riccati equations.

26. y' = x
3
+ -x2 Y - 1 2
- Y,
x
Yl(X) = -x .
2

27. Y' = 2 tan x sec x - y2 sin x, Yl(X) = sec x.

, 1 Y 2
28. Y = x 2 - X - Y, Yl (x) =
1
x·
2
Y
29. Y' = 1 + - - -,
y
x x2
Yl(X) = x.

ANSWERS II
I. 4xy = x4 + c. 4. x = e-,,2(y + c).
2. Y = -b + ce • 5. 2r = c sec 8 - e--l(tan 8 + 1).
a
3. Y log x + Y + cxy = 1. 6. Y = (Arc tan z + c)(z2 + 1).
1 = ce2z
7. 2' + x + 2·
1
Y 2
-3/2 3 c(1 - x)
8. Y = - + x + x2) + + +
4(1 1 x x2 .
9. r = sin 8[log (sec (J + tan 8)] + c sin 8.
. _ -RtIL + E(R sin kt - kL cos kt) .
I o. ." - CC R2 + k2L2
II. Y = + ce- 2z .
12. y = ixe- 2z + ce- 2z•
13. Y = 1(2 sin x - cos x) + ce-2z•
14.y=e-lin z ( c + e f
2z+lin d )
x·
1 + ce -lin •
Z2
· X
IS• Y = sln -
20. Y = (x - l)c + 1·
sin x
16. Y = - - - cos x
x
+ -xc · 21. Y = + 1).
17. Y = x(c - cos x). 22. Y = 1.

18. y = x 2
2x
+ c' Y = o. 23. sec x = y2(tan z + 1).
19. 1 - 2y(1 + log x) = cxy. 24. 8x sin Y = 4 sin 2 y + 3.
26. u'+ (;+ 2X)U = i.
2
Y = -x 2 + 2x e
z2

ez2 + c
2
__ 1_+ 3cos z
27. u' - 2u tan x = sin x,
Y - cos x c - cos3 Z
, 3 1 2
28. u - - u = 1,
x Y = x+
,1 1 2x
29. u - - u = -2, y=z+cx2 -1·
:J; x
Lesson 12A EQUATIONS PERMrrl'ING A CHOICE OF METHOD 99

LESSON 12. Miscellaneous Methods of Solving a First Order

Differential Equation.

LESSON 12A. Equations Permitting a Choice of Method. If a

differential equation is selected at random, it may be solvable by more
than one method. The one selected will depend ultimately on your in-
genuity in detennining which will most readily lead to a solution. The
following examples will illustrate this point.
Example 12.1. Solve

(a) x dy - y dx = y2 dx.

Solution. As was shown in Example 10.51, multiplication by y-2,

Y ¢ 0, will make (a) exact. It will then be solvable by the method of
Lesson 9B, or by means of (10.30). Finally if we divide by x, (x ¢ 0),
the equation becomes
, y y2
(b) y - -
x
= -,
x
x ¢ 0,

which one recognizes as a Bernoulli equation. Hence it is solvable by

the method of Lesson lID. Of the three choices available to solve (a),
you will find that use of (10.30) is the easiest.
Example 12.11. Solve

(a)

Solution. This equation is of the type with homogeneous coefficients

and is therefore solvable by the method of Lesson 7B. It is also exact
and therefore solvable by the method of Lesson 9B. However, the easiest
method is to rewrite (a) as
(b) x 2 dy + 2xydx + y 2 dy = 0,

and then make use of (10.21).

Example 12.12. Solve
(a)

Solution. The equation is exact and is therefore solvable by the

method of Lesson 9B. If we divide (a) by e3z dx, we obtain

(b) + 3y = 2xe-3s,

an equation which is linear and hence solvable by the method of Lesson

100 SPECIAL TYPES OF FIRST ORDER EQUATIONS Chapter 2

liB. If, however, (a) is rewritten as

(c)

it can be solved most easily by making use of (10.392).

Example 12.13. Solve

(a)

Solution. The equation is of the type with homogeneous coefficients

and therefore is solvable by the method of Lesson 7B. However, the
presence of the combination x(x dy - y dx) leads one to try to make use
of (10.3) or (10.31). We therefore divide (a) by x, (x 0) and rewrite
it to read
(b) x dy - y dx = V x 2 + y2 dx, x;t6. o.
We then divide (b) by y2 to obtain

(c) x dy ;; y dx = vx 2 + y2 dx = ! + 1 dx, x;t6. 0, Y ;t6. O.

By (10.3) the left side of (c) is -d(x/y). Hence (c) can be written as
d(x/y) dx
(d) -;::==::::::::::::;== = - , x 0, y o.
+
v'1 (X/y)2 x
Integration of (d) now gives

(e) -log I; + VI + (x/y) 2 I= log Ixl + log lei, x;t6. 0, y ;t6. 0,

which can be written as

(f) y = ex, x 0, y o.
x+v'X 2 +y2

The solution of (a) is therefore

(g) y = ex(x + v'X2 + y2), X 0, y o.
Comment 12.14. The function y = 0 which we had to exclude to
obtain (g) also satisfies (a). Moreover every member of the family of
solutions (g) goes through the point (0,0). Observe from (a), however,
that dy/dx is undefined when x = O. The point (0,0) is one of those
singular points we discussed in Lesson 5B. Every member of the family
(g) lies on it, but no member goes through any other point on the line
x = O. This solution, therefore, actually should have been written with
Lesson 12B SOLUTION BY SUBSTITUTION AND OTHER MEANS 101

two parameters instead of one, namely

y = Ctx(x + VX2 + y2), X S 0
Y = C2X(X + vx 2 + y2), X >, o.
LESSON 12B. Solution by Substitution and Other Means. A
first order differential equation need not come under any of the headings
mentioned heretofore. This fact should not be too surprising. You your-
self could easily write a first order differential equation which would not
fit any of the types thus far discussed. It is possible in some cases to solve
a differential equation by means of a shrewd substitution, or by discovering
an integrating factor, or by some other ingenious method. We give below
a number of examples which do not, as they stand, lend themselves to
standardization. You should keep in mind that these examples have been
specially designed to yield a solution in terms of elementary functions.
It is easily conceivable, if a differential equation were selected at random,
that one could spend hours and days using every known method and device
at one's disposal and still fail to find an explicit or implicit solution in
terms of the elementary functions. More than likely no such solution exists.
Example 12.2. Solve
(a) (y - y2 - x 2) dx - x dy = O.

The equation as it stands cannot be solved by any of the methods out-

lined thus far. However, the presence of the combination y dx - x dy
leads one to try to make use of (10.3) or (10.31). Rearranging terms and
dividing by x 2 gives

(b) ydx xdy = [1 + (:YJdX, x o.

By (10.31), the left side of (b) is -d(y/x). Hence (b) can be written as

(c)
d(:) 2 = dx.
1 + (:)
Integration of (c) gives
(d) Arc tan y
x
= -(x + c) or y = -tan (x
x
+ c), x o.
Hence the solution of (a) is
(e) y = -x tan (x + c), x O.
NOTE.In regard to the line x = 0 and the point (0,0), we refer you to
our Comment 12.14.
102 SPECIAL TYPES OF FIRST ORDER EQUATIONS Chapter 2

Example 12.21. Solve

(a) (2 cos y) y' + sin y = x 2 csc y, Y o.

Solution. The equation as it stands cannot be solved by any of the
methods outlined thus far. If we multiply it by sin y, we obtain

(b) (2 sin y cos y) y' + sin 2 y = x 2, Y o.

The first tenn is equal to :x (sin 2 y). We therefore can write (b) 8S

(c)

an equation which is now linear in the variable sin 2 y. The integrating

factor by (11.12) is found to be eZ. Hence by (11.19)

(d) e" sin 2 y = f 2

x e" dx = e"(x
2
- 2x + 2) + c, y o.
The solution of (a) is therefore

(e) sin 2 y = (x 2 - 2x + 2) + ce-z , y o.

Example 12.22. Solve

(a) y' + 2x = 2(x 2 +y - 1)2/3.

Solution. The equation as it stands cannot be solved by any of the

methods outlined thus far. A powerful and useful method frequently used
by mathematicians to integrate functions is that of substitution. An
attempt is made to simplify the integrand by using a new variable to
represent a function of the given variable. In some cases this method of
substitution can also be used profitably to find solutions of differential
equations. For this example, we try the substitution

(b) u = x2 +y - 1,

and hope it will yield a differential equation in u which we can solve in

terms of elementary functions. Differentiating (b) we obtain

(c) du = 2x + dy. dy = du _ 2x.

dx dx' dx dx
We now substitute (b) and (c) in (a). There results

(d) du 2 2/3
dx = U ,
Lesson 12-Exercise 103

By integration of (d) we obtain

(e) 3U 1/ 3 = 2x + c, u ¢ o.
Replacing u by its value in (b) and cubing, we have

(f) x
2
+y - 1 = (2x + C)3, x
2
+y - 1 ¢ o.
The solution of (a) is .therefore

(g) y-
_ 1 -
x
2 + (2x +
27'
C)3 x2 +y - I ¢ O.

NOTE. The function y = I - x 2 which had to be excluded in obtain-

ing (g) also satisfies (a). It is a particular solution of (a) not obtainable
from the family (g).

EXERCISE 12

Solve each of the following differential equations.

I. 2ZY::: + (1 + X)y2 = e".

th + sin
2• cos Y dy · y == x.2 H,nt.
· th d (8m.)
y == COS Y dy
dx •
3. (x + 1) dy - (y + 1) dz == (x + 1)'V' y + 1 dz. Let u == y + 1.
4. e"(y' + 1) = e". Hint. ! e" .. eliy'o

S. y' sin y + sin x cos y = sin X. Hint. ! cos y = -(sin y)y'.

2 dy
6. (x - y) th == 4. Let u = x - y.
dy
7. z dz - Y == V x2 + y2.
8. (3x
2
+ +
2y 1) dy + (4x + 3y + 2) dx = O.
9. (x - y2) dy == 2xy dz.
10. 11 dz + +(1 dy == O. Let y ==
11. (x 2 y+ y2) dx + x3 dy == O.
12. (y 2eQI+ 4x3 ) dx + - 3y2) dy == O.
+
13. y' == (x 2 2y - 1)2/3 - x. Let u == x2 + 2y - 1.

14. x y .. x (1
2
+ ej'l.
MISCELLANEOUS PROBLEMS

In the following set of problems, classify each differential equation by

type before attempting to find a I-parameter family of solutions. Some
may fall into more than one type; some may not fall into any. In the
104 SPECIAL TYPES OF FIRST ORDER EQUATIONS Chapter 2

answers, you will find hints as to methods of solving. Always first try to
solve the equation before looking at these hints. Also keep in mind that
there may be other methods in addition to the ones we have given.
IS. (2y - xy log x) dx - 2x log x dy = o.
16. y' + ay = kebz •
17. y' = (x +
y)2.
18. y' + 3y
8x 3 +
2xy = o.
2
19. (xwx2 - y2 +
x)y' = y - x ..Jx2 - y2.
20. y' + ay = b sin kx.
21. xy' - y2 + 1 = o.
22. ( y 2 + . )
a Sln x dy
dx = cos x.
23. xy' = xe lllz + x + y.
24. y' + ain
y cos x = e- Z.
2S. xy' - y(log xy - 1) = O.
26. x 3y' - y2 - x 2y = o.
27. xy' + ay + bxn = 0, x > O.
28• xy' - x Sln. -Y - y = 0 .
x
29. +
(xy - x 2)y' y2 - 3xy - 2x2 == o.
30. (6xy + + + + +
x 2 3)y' 3y2 2xy 2x = o.
31. 2
x y'+ + + y2 xy x = o.
2
32. 2
(x - l)y'+ 2xy - cos x = O.
33. +
(x 2y - l)y' xy2 - 1 = O.
34. 2
(x - l)y'+ xy - 3xy2 = O.
3S. (x - l)y' - 2xy log y = o.
2
36. (x 2+ + + + +
y2 l)y' 2xy x2 3 = O.
37. + + +
y' cos x y (1 sin x) cos x = O.
38. (2xy + + +3
4x )y' y2 12x2y = o.
39. (x 2 - y)y'
2
+ x = O.
40. (x - y)y' - 4xy = O.
41. xyy' + + x2 y2 = O.
42. 2xyy' + 2
3x - y2 = O.
43. (2 xy 3 - x 4)y' + 2x 3 y - y4 = o.
44. (xy - 1)2xy' + +
(x 2y2 l)y = O.
4S. (x 2+ +y2)y' 2x(2x+ y) = o.
46. 3xy2y'+ y3 - 2x = O.
47. +
2 y 3y' xy2 - x 3 = O.
48. (2xy3 + + xy x 2)y' - xy + y2 == O.
49. (2y3 + y)y' - 2x 3 - X = o.
SO. y' - eZ - lI + eZ = o.

ANSWERS 12
1. 2y 2x = eZ + ce- Z • 4. y == log +
2. sin y = x 2 - 2x + + ce-
2 Z
• S. cos y == 1 +
3. Y'y + 1 = x + 1 + cY'x + 1.
x-y-2
6. y = log x - y + 2 + c.
7. y + Y' x 2 + y2 = cx , or equivalently,
2
y == C + v'x2 + y2.
8. 3xy + y2 + y + 2x2 + 2x = c.
Miscellaneou8 Problems-Answers 105

9. x 2 + +
y2 cy = o. 12. ezy2 + x4 - y3 = c.
10. e- 2z = 2y2(log Iyl - c). 13. 3(x 2 + 2y - 1)1/3 = 2x + c.
11. 3x 2 +y = cx3y. 14. xy(x + e + c) + 1 == o.
Z

ANSWERS FOR MISCELLANEOUS PROBLEMS

15. Separable. x + 2 log IYI - 2101 (log Ix/} = c.
16. Linear.
Y = a+be
bz+
k
ce,
-GZ
a +b 0

y = k xe bz + ce,
-GJ:
+ b = o. a
17. Let u = x + y. Resulting equation is separable. x + y = tan (x + c).
18. Bernoulli. y -2 = ce 2z2
- 4x - 22.
19. Let y = ux. Resulting equation can be made exact.
y = x sin [c - l(x 2 + y2»), x >0
y = x sin [c + l(x 2 + y2»), x < o.
20. Linear. y = a2 ! k 2 (a sin kx - k cos kx) + ce- Gz
•

21. Separable. y = (1 - cx 2)/(l + cx 2).

22. Let y be the independent variable. Bernoulli in sin x,· see problem 2 above.

•
Slnx = ce
Gil
- ( -+-+-
Y2
a
2y
a2
2)
a3
.
23. Homogeneous. (1 - cx)eY!z = cx.
24. Linear. y = (x +c)e-ain Z.
25. Let u = xy. Resulting equation is separable. xy = eCZ •
26. Bernoulli. Also l/x 2y2 is an integrating factor. x 2 - y = cxy.
27. Linear
-G b n
y=cx - x a -no
a+ n '
Y = cx -G - bx -Gl og x, a = -no
28. Homogeneous. csc - cot = cx.
x x
29. Integrating factor x. x 2y2 - 2x 3y - x4 = c.
30. Exact. 3xy2 ++ +
x 2y 3y x 2 = c.
31. +
Homogeneous. x = (y + x) (log Ixl c).
32. +
Linear. Also exact. (x 2 - l)y = sin x c.
33. 2 +
Exact. x y2 - 2(x y) = c.
34. +
Bernoulli. y-l = 3 cv''--lx-2---1-1.
35. Separable. y '= e (z2- ).
c 1

36. + +
Exact. x 3 y3 + +3(x 2 y y 3x) = c.
in
37. Linear. (1 + sin x)y = cos x (sin x - 2 log 1 + s2 x + c) ·
cos x
38. Integrable combinations. 4x 3y xy2 = c. +
39. Bernoulli, y independent variable. 2x2 = 2y - 1 + ce- 2y•
40. Let y = ux 2• Resulting equation is separable. (x 2 + y)2 = cy.
41. +
Homogeneous. x 2(x 2 2y2) = c.
106 SPECIAL TYPES OF FIRST ORDER EQUATIONS Chapter 2

42. Bernoulli. y2 = cx - 3x 2•
43. Integrating factor 1/x 2 y2. x3 + y3 = cxy.
44. Let u == xy. Resulting equation is separable. y2 ==
45. Homogeneous. Also exact. y3 + 4x 3 + 3x 2y == c.
46. +
Bernoulli. xy3 = x 2 c.
47. Homogeneous. Also x 2 + y2 is an integrating factor.
(x 2 + y2)2(2y2 - x 2) == c.
48. Integrating factor 1/xy2. y3 + y log /xy/ - x == cy.
49. Exact. Also separable. y4 + y2 = x4 + x 2 + c.
SO. Let u == e7l • Resulting equation is linear. elf == 1 + ce- r.
 Chapter 3

Problems Leading to Differential

Equations of the First Order

LESSON 13. Geometric Problems.

We are at last ready to study a wide variety of problems which lead to
differential equations of the first order. We consider first certain geo-
metric problems in which we seek the equation of a curve whose deriva-
tive y' has certain preassigned properties.
Example 13.1. Find the family of curves which has the property that
the segment of a tangent line drawn between a point of tangency and the
y axis is bisected by the x axis.

(0,0) x

Figure 13.11

Solution (Fig. 13.11). Let P(x,y) be a point on a curve of the required

family and let PQ be a segment of the tangent line drawn at this point.
By hypothesis, PQ is bisected by the x axis. Hence the coordinates of the
point Q are (O,-y). [The mid-point fonnula for two points (Xt,Yt) and

(X2,Y2) is (Xl X2, Yt Y2) .] The equation of the line PQ is, there-

107
108 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

fore, given by
2y , dy
(a) - = y =-.
x dx
By Lesson 6C, the solution of (a) is
(b) y = cx 2 , X F 0, y 0.
We have thus shown that if there exists a family of curves which meets
the requirements of our problem, it must satisfy (a) and hence must
satisfy (b).
Conversely, we must show' that if a family of curves satisfies (b), it
will meet the requirements of our problem. Let P(xo,Yo) be a point on a
curve of (b). Then by (b)
(c) Yo = cXo
2
, y' (xo.Uo) = 2cxo.
The equation of the line at P(xo,Yo) with slope 2cxo is
(d) y - Yo = 2cxo(x - xo)·
When x = 0, we obtain from (d)
(e)
which is the y coordinate of the intersection of the line (d) with the y
axis, (Q in Fig. 13.11). The coordinates of the mid-point of PQ are therefore

(f) Yo - CX 02 ) .

Replacing Yo in (f) by its value in (c), we obtain for the coordinates of

the mid-point of PQ
(g)

We have thus proved that the tangent

y
at any point P(xo,Yo) of a member of
the family (b), drawn to the y axis is
bisected by the x axis.
Example 13.2 (Fig. 13.21). Find
the family of curves with the property
that the area of the region bounded by
the x axis, the tangent line drawn at a
point P(x,y) of a curve of the family
0(0,0) Q(a,O) R(x,O) X and the projection of the tangent line
on the x axis has a constant value A.
Figure 13.21
Solution. At P(x,y) draw a tan-
gent to a curve of the required family. Call Q(a,O) the point of inter-
Lesson 13 GEOMETRIC PROBLEMS 109

section of this line with the x axis, R(x,O) the intersection of the projec-
tion of the tangent line with the x axis. The equation of the tangent line is

(a) y -
--"- - ,
y, x ¢ a.
x-a
Solving this equation for a, we obtain
y
(b) a=x--,
y'

which represents the distance OQ. Hence the distance QR is

(c) x_(x_lL)=1L.
y' y'

The region whose area is A, is therefore given by the formula

(d) A = !y (y)y' = :l..

2 2y'
Hence,
y2
(e) y' = - ,
2A
whose solution, by Lesson 6C, is
1 x 2A
(f) y = - 2A +c or y = 2Ac _ x' x ¢ 2Ac.

We have thus shown that if there exists a family of curves which meets
the requirements of our problems it must satisfy (a) and hence must
satisfy (f).
Conversely, we must show that if a family of curves satisfies (f), it will
meet the requirements of our problem. Let P(xo,Yo) be a point on a
curve of (f). Then by (f)
2A I 2A
(g) = (2Ac - xo) 2
Yo = 2Ac - xo' Y (%0,110)

The equation of the line at P(xo,Yo) with slope given in (g) is

2A
(h) y - Yo = (x - xo) (2Ac - XO)2·

When y = 0, we obtain from (h),

.) 2Axo- - yo(2Ac - XO)2
(1 x = 2A '

which is the x coordinate of the intersection of the line (h) with the x axis
110 PROBLEMS LEADING TO FiRST ORDER EQUATIONS Chapter 3

(Q in Fig. 13.21). The distance QR is therefore

2Axo - Yo(2Ae - XO)2 Yo(2Ae - xo)2
0) Xo - 2A = 2A •

Hence the area of the region A is

(k) !Yo XO)2].

Substituting in (k) the value of (2Ae - Xo) 2 as determined by the first
equation in (g), we obtain
y 4A2)
(1) !Yo ( '

which simplifies to the constant A. Hence our family (f) meets the re-
quirements of the problem.
E:mmple 13.3. Find the family of curves such that the angle from a
tangent to a normal at any point of a curve of the family is bisected by the ra-
dius vector at that point. (In problems involving radius vectors, it is usually
preferable to use polar coordinates.)
y
Solution (Fig. 13.31). Let P(r,8)
be the polar coordinates of a point on
a curve of the required family. Call fJ
the angle measured from the radius
vector r counterclockwise to the tan-
gent line at P. Then by a theorem in
the calculus
(0,0) (a) tan fJ =
aB
r- ·
Figure 13.31
ar
By hypothesis r bisects the angle be-
tween the normal and tangent lines. Hence fJ = 45° or -45°. In Fig.
13.31 it is -45°. Therefore (a) becomes (using fJ = 45°)
dB dr = dB,
(b) l=r ,
dr r
whose solution is

(c) log r = 8 + e' ,

We leave it to you as an exercise to solve (a) when fJ = -45° and to
prove the converse, i.e., that if a family of curves satisfies (c), then it
meets the requirements of our problem.
Lesson 13 GEOMETRIC PROBLEMS III

Example 13.4. Find the family of curves with the property that the
area of the region bounded by a curve of the family, the x axis, the lines
x = a, x = x is proportional to the length of the arc included between
these two vertical lines.
Solution (Fig. 13.41). The formula for the arc length of a curve be-
tween the points A ,B whose abscissas are x = a and x = x is

The area of the region bounded by

the lines, x = a, x = x, the x axis,
A
and the arc of the curve y = f(x)
between these lines, is y

(b) A = laO: f(x) dx.

By hypothesis A is proportional to 8;
o x=a d% %=X X
therefore, by (a) and (b), Figure 13.41

(c) laO: f(x) dx = k laO: + dx,

where k > 0 is a proportionality constant. Differentiation of (c) with

respect to x, and replacing f(x) by its equivalent y, give [see (9.15)]

(d) y = +
which simplifies to
ly2 - k 2 _ dy dx ± dy
(e) ±"\J. k 2 - dx' k = ...;y2 _ k2 ' Y k.

By integration of (e), we obtain

(f) = log (y ± v'y2 - k 2 ) - loge,

which can be written as

(g)

We leave it to you as an exercise to prove the converse, i.e., if a family

of curves satisfies (g), then it meets the requirements of our problem.
112 PROBLEMs LEADING TO FIRST ORDER EQUATIONS Chapter 3

EXERCISE 13
1. Let P(z,y) be a point on the curve y = f(x). At P draw a tangent and a
normal to the curve. The slope of the curve at P is therefore y'; the slope
of the normal is -1/y'. Prove (see Fig. 13.5) each of the following.
(a) z - y/y' is the x intercept of the tangent line.
(b) y - xy' is the y intercept of the tangent line.
(c) x +yy' is the x intercept of the normal line.
(d) y + x/y' is the y intercept of the normal line.
(e) /y/y'/ is the length AC of the projection on the x axis, of the segment of
the tangent AP. The length AC is called the subtangent.
(f) /yy'/ is the length CB of the projection on the x axis of the segment of
the normal BP. The length CB is called the subnormal.

(g) The length of the tangent segment AP = y +1 ·

(h) The length of the tangent segment DP = IxVI + (y')21.

(i) The length of the normal segment PB = IWI + (y')21.
(j) The length of the normal segment PE = x + (,)2 ·

Subnormal Iyy'l
(O,O) ........
(x+yy', O)

D (O,y - x.v')

Figure 13.5

In each of the following problems 2-20, use rectangular coordinates to

find the equation of the family of curves that satisfies the property described.
2. The slope of the tangent at each point of a curve is equal to the sum of the
coordinates of the point. Find the partiCUlar curve through the origin.
3. The sub tangent is a positive constant k for each point of the curve. Hint.
See I(e).
LeUOD 13-Exercise 113

4. The subnormal is a positive constant k for each point of the curve. Hint.
See l(f).
5. The subnormal is proportional to the square of the abscissa.
6. The segment of a normal line between the curve and the y axis is bisected
by the x axis. Find the particular curve through the point (4,2), with this
property.
7. The x intercept of a tangent line is equal to the ordinate. Hint. See l(a).
8. The length of a tangent segment from point of contact to the x intercept is
a constant. Hint. See 1(g).
9. Change the word tangent in 8 above to normal. Hint. See l(i).
10.· The area of the right triangle formed by a tangent line, the x axis and the
ordinate of the point of contact of tangent and curve, has constant area 8.
Hint. See l(e).
11. The area of the region bounded by a curve y = f(x), the x axis, and the lines
x = 2, x = x, is one-half the length of the arc included between these
vertical lines.
12. The slope of the curve is equal to the square of the abscissa of the point of
contact of tangent line and curve. Find the particular curve through the
point (-1,1).
13. The sUbtangent is equal to the sum of the coordinates of the point of contact
of tangent and curve. Hint. See l(e).
14. The length of a tangent segment from point of contact to the x intercept is
equal to the x intercept of the tangent line. Hint. See (lg) and l(a).
15. The normal and the line drawn to the origin from point of contact of normal
and curve form an isosceles triangle with the x axis.
16. Change the word normal in 15 above to tangent.
17. The point of contact of tangent and curve bisects the segment of the tan-
gent line between the coordinate axis.
18. Change the word tangent in 17 above to normal.
19. The length of arc between x = a and x = x is equal to x 2 /2.
20. The area of the region bounded by the curve y = I(x), the lines x = a,
x = x, and the x axis, is proportional to the difference of the ordinates.
21. Let y = I(x) define a curve that passes through the origin. Find a family of
curves with the property that the volume of the region bounded by y == I(x),
x = 0, x = x rotated about the x axis is equal to the volume of the region
bounded by y == I(x), y = 0, y = y rotated about the y axis.
22. Let P(r,B) be a point on the curve r = r(B). At P draw a tangent to the
curve. Prove (see Fig. 13.6) each of the following.

(a) = tr2, where A is the area of the region bounded by an arc of the
curve and two radii vectors.

(b) dBds = '(dr)2 + r , where

\j dB 2 .
S IS the length of arc of the curve between
two radii vectors.
dB
(c) tan (3 = r dr. See (a) of Example 13.3.

In each of the following problems 23-25, use polar coordinates to find the
equation of the family of curves that satisfies the property described.
23. The radius vector r and the tangent at P(r,B) intersect in a constant angle {3.
114 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

24. The radius vector r and the tangent at P(r,8) intersect in an angle which is
k times the polar angle 8.

r= r(8)

o
Figure 13.6

25. The area of the region bounded by an arc of a curve and the radii vectors to
the end points of the arc is equal to one-half the length of the arc. Hint. See
22(a) and (b).

ANSWERS 13

2. y = eZ - x - I .
3. ylc = ce z, (y/y') > 0; ylc = ce- z, (yjy') < O.
4. y2 = 2kx + c, yy' > 0; y2 = -2kx +
c, yy' < o.
5. 3y2 = 2kx 3 + C, yy' > 0; 3y2 = -2kx3 C, yy' <+ o.
6. x 2 + 2y2 = 24. 7. x y log y + + cy = O.

8. ± x+ c =

9. x = ± v'k2 - y2
v'k2 -

c.
y2 - k log

+
e+ y2) .
10. xy = cy - 16, y' > 0; xy = cy + 16, y' < O.
+
II. 8y = 4ce±2z c- 1 e=f2z. 12. 3y = x 3 + 4.
13. x = y log IcYI, (y/y') > 0; 2xy + y2 = C, (yjy') <: O.
14. x2+ y2 = cy. 17. xy = c.
15. 2
x - y2 = c. 18. x 2 - y2 = c.
16. xy = c.
19. y = ± vix 2 - 1 - 1 log (x + vix 2 - 1)] + c.
20. ylc = ceZ • 24. ric = c sin k8.
21. x - y = cxy. 25. r = 1; r = sec (8 + c), r 1.
23. r = ce'cot/j.
Lesson 14A ISOGONAL TRAJECTORIES 115

LESSON 14. Trajectories.

LESSON 14A. Isogonal Trajectories. When two curves intersect in

a plane, the angle between them is defined to be the angle made by their
respective tangents drawn at their point of intersection. Since these lines
determine two angles, it is customary to specify the particular one desired

o x
Figure 14.1

by stating from which tangent line we are to proceed in a counterclockwise

direction to reach the other. In Fig. 14.1, a is the positive angle from the
curve Cl with tangent line II to the curve C2 with tangent line l2; is the
positive angle from the curve C2 to the curve Ct. If we call ml the slope
of II and m2 the slope of l2, then by a formula in analytic geometry

(14.11)

Definition 14.12. A curve which cuts every member of a given

I-parameter family of curves in the same angle is called an isogonal
trajectory of the family.

If two I-parameter families have the property that every member of

one family cuts every member of the other family in the same angle,
then each family may be said to be a I-parameter family of isogonal
trajectories of the other, Le., the curves of either family are isogonal
traj ectories of the other.
An interesting problem is to find a family of isogonal trajectories that
makes a predetermined angle with a given I-parameter family of curves.
If we call Yl' the slope of a curve of a given I-parameter family, Y' the
slope of an isogonal trajectory of the family, and a their angle of inter-
116 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

section measured from the tangent line with slope y' to the tangent line
with slope Yl', then by (14.11)
Yl' - y'
(14.13) tana=I+' ,.
Y Yl
Example 14.14. Given the I-parameter family of parabolas

(a) Y = ax 2 ,

find a I-parameter family of isogonal trajectories of the family if the

angle of intersection a, measured from the required trajectory to the given
family, is 7r/4.
Solution. Differentiation of (a) gives

(b) y' = 2ax.

From (a) again, we obtain

(c)

Substitution of this value in (b) gives

(d) ,2y
Y=-' 0
x

which is the slope of the given family at any point (x,y), x o. It there-
fore corresponds to the Yl' in (14.13). Since by hypothesis a = 7r/4,
tan ex = 1. Hence (14.13) becomes
2y ,
--Y
1 = _x___ _ 2y - xy'
(e)
1 + 2y y' x 2yy' +
x
Simplification of (e) gives

(f) (x - 2y) dx + (x + 2y) dy = 0,

whose solution, by Lesson 7, is

(g) logV2y2 - xy + x2 + .
v7
tan 40 7x
x = c, x o.
Comment 14.15. Note that before we used (14.13), we eliminated
the parameter in (b) to obtain (d). This elimination is essential.

Comment 14.16. Because of the presence of the inverse tangent,

(g) implicitly defines a multiple-valued function. By our definition of a
Lesson 14B ORTHOGONAL TRAJECTORIES 117

function it must be single-valued, Le., each value of x should detennine

one and only one y. For this reason we have written the inverse tangent
with a capital A to indicate that we mean only its principal values, namely
those values which lie between - 7r/2 and 7r/2. By placing this restric-
tion on the inverse tangent, we have thus excluded infinitely many solu-
tions, for example solutions for which the arc tan lies between 7r/2 and
37r/2, between 37r/2 and 57r/2, etc.

LESSON 14B. Orthogonal Trajectories.

Definition 14.2. A curve which cuts every member of a given I-param-

eter family of curves in a 90° angle is called an orthogonal trajectory
of the family.
If two I-parameter families have the property that every member of
one family cuts every member of the other family in a right angle, then
each family may be said to be an orthogonal trajectory of the other.
Orthogonal trajectory problems are of special interest since they occur
in many physical fields.
Let Yt ' be the slope of a given family and let Y' be the slope of an or-
thogonal trajectory family. Then by a theorem in analytic geometry
, 1
(14.21) Yt'y' = -1, y = - Yt / ·

Example 14.22. Find the orthogonal trajectories of the I-parameter

family of curves
(a)

Solution (Fig. 14.23). Differentiation of (a) gives

(b)

From (a) again we obtain

(c)

Substituting this value in (b), we have

(d) y' = 5 Y, x 0,
x

which is the slope of the given family at any point (x,y), x O. It

corresponds therefore to the Yt ' in (14.21). Hence, by (14.21), the slope
y' of an orthogonal family is
,x 0
(e) y = - 5y' x 0, Y ¢ ·
118 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

Its solution, by Lesson 6C, is

(f) x 2 + 5y2 = k, x 0, y o.
which is a I-parameter family of ellipses.
y

c=o
x

Figure 14.23

Comment 14.24. Note that before we used (14.21), we eliminated

the parameter in (b) to obtain (d). This elimination is essential.

LESSON 14C. Orthogonal Trajectory Formula in Polar Coordi-

nates. Call P(r,O) (Fig. 14.3) the point of intersection in polar coordi-

o x
Figure 14.3

nates of two curves CI,C2 which are orthogonal trajectories of each other.
Call 4>1 and 4>2 the respective angle the tangent to each curve Cl and C2
Le8son 14C ORTHOGONAL TRAJECTORY FORMULA IN POLAR COORDINATES 119

makes with the radius vector r (measured from the radius vector counter-
clockwise to the tangent). Since the two tangents are orthogonal, it is
evident from the figure that
'Tr
4>1 = 4>2 + -.
2
Therefore
(14.31) tan 4>1 = tan (4)2 + '2Tr) = _ 1 ·
tan 4>2

As remarked previously in Example 13.3, in polar coordinates

(14.32)
Therefore (14.31) becomes
dr
(14.33) tan 4>1 = - --.
rd8

Comparing (14.32) with (14.33), we see that if two curves are orthogonal,
then T of one is the negative reciprocal of T of the other. Conversely,
if one of two curves satisfies (14.32) and the other satisfies (14.33), then
the curves are orthogonal. Hence to find an orthogonal family of a given
family, we proceed as follows. Calculate T of the given family. Replace

T by its negative reciprocal - d;o'

T The family of solutions of this
new resulting differential equation is orthogonal to the given family.
Example 14.34. Find, in polar form, the orthogonal trajectories of
the family of curves given by

(a) r = k sec 8.

Solution. Differentiation of (a) gives

dr
(b) d8 = k sec 8 tan 8.

From (a), again, we obtain

r
(c) k=-·
sec 8
Substituting this value in (b), we have
dr d8 1
(d) d8 = r tan 8, r-
dr = tan 8 ·
120 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

By (14.33), therefore, the differential equation of the orthogonal family is

dr 1 dr
(e) - --
r d8
= tan 8
or - -
r
= cot 8d8
'
whose solution is
(f) r sin 8 = c.
Comment 14.35. We remark once more that before we could use
(14.33), we had to eliminate the parameter in (b) to obtain (d).

EXERCISE 14

For each of the following family of curves, 1-4, find a I-parameter

family of isogonal trajectories of the family, where the angle of intersec-
tion, measured from the required trajectory to the given family, is the
angle shown alongside each problem.
1. y2 = 4kx, a = 45°.
2. x2 + y2 = k 2, a = 45°.
3. y = kx, tan a = 1.
4. y2 + 2xy - x 2 = k, a = 45°.
S. Given that the differential equation of a family of curves is M(x,y) dx +
N(x,y) dy = 0. Find the differential equation of a family of isogonal tra-
jectories which makes an angle a 'I' /2 with the given family, where a
is measured from trajectory family to given family.
6. Given that the differential equation of a family of curves is of the homo-
geneous type. Prove that the differential equation of a family of isogonal
trajectories which makes an angle a '1'/2 with the given family is also
homogeneous. Hint. You will find a proof in Case 3-2 of Lesson 32.
Find in rectangular form the orthogonal trajectories of each of the fol-
lowing family of eurves.
+
7. x 2 2xy - y2 = k. 13. x 2 y = k.
8. x 2+ (y - k) 2 = k 2 • 14. x 2 - y2 = k 2.
9. (x - k)2 + y2 = k 2. IS. y2 = kx 3 •
10. x - y = ke z • 16. e cos y = k.
Z

II. y2 = 4px. 17. sin y = ke z2 •

12. xy = kx - 1.

Find in rectangular form the orthogonal trajectories of each of the fol-

lowing family of curves.
18. A family of straight lines through the origin.
19. A family of circles with variable radii, centers on the x axis, and passing
through the origin.
20. A family of ellipses \vith centers at the origin and vertices at (± 1,0).
21. A family of equilateral hyperbolas whose asymptotes are the coordinate axes.
Leason 14-Exercise 121

22. A family of parabolas with vertices at the origin and foci on the z axis.
If the differential equation of a given family of curves remains unchanged
when y' is replaced by -l/y', the family is called self orthogonal, i.e., a
self-orthogonal family has the property that a curve orthogonal to a mem-
ber of the family also belongs to the family. Show that each of the fol-
lowing families is self orthogonal.
23. The family of parabolas that have a common focus and axis. Hint. See
answer.
24. The family of central conics that have common foci and axis. Hint. See
answer.
Find in polar form the orthogonal trajectories of each of the following
family of curves.
25. r = k cos 8. 30. r2 = k(r sin 8 - 1).
26. r8 = k. 31. r2 = k cos 28.
27. r == k(l+ sin 8). 32. r- 1 = sin 2 8 + k.
2B. r = sin 8 + k. 33. rn sin n8 = k.
29. r = k sin 28. 34. r = e"'.

ANSWERS 14
2 6 z+2y
1. log 12x + xy + y 2I + V7 Arc tan xV7 = c.

2. log Ic(z2 + y2) I - 2 Arc tan (y/z) = o.

3. log Ic(x 2 + y2)1 + 4 Arc tan (y/x) = o.
4. xy = c.
5. y' = (N tan a + M)/(M tan a - N).
7. x 2 - 2xy - y2 = c. IS. 3y2 +2x2 = c.
B. x2 + y2 = cx. 16. eX sin y = c.
9. x 2 + y2 = cy. 17. x = c cos 2 y.
10. ce 1l = y - x +2. lB. x2 + y2 = c.
II. 2x2 + y2 == c. 19. See problem 9.
12. x 3 + 3y = c. 20. x 2 = ce x2 +11 2 •
13. 2y2 - x 2 = c. 21. x2 - y2 = c.
14. xy = c. 22. See problem 11.
23. Equation of family is y2 = 4p(x
2 2
+ p).
24. Equation of family is x 2
a
+ a2 Y- c
2 = 1, where (±c,O) are the coordinates
of the foci.
25. r = c sin 8.
2
30. r = 2 sin 8 +
c cos 8.
26. cr = e' 12. 31. r2 = c sin 28.
27. r = c(l - sin 8). 32. tan 8 = ce 2r •
2B. ellr = c(sec 8 +
tan 8). 33. rn cos n8 = c.
29. r 4 = c cos 28. 34. r =
122 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

LESSON 15. Dilution and Accretion Problems. Interest

Problems. Temperature Problems.
Decomposition and Growth Problems.
Second Order Processes.

LESSON 15A. Dilution Accretion Problems.

In this type of problem, we seek a formula which will express the amount
of a substance in solution as a function of the time t, where this amount
is changing instantaneously with time.
Example 15.1. A tank contains 100 gallons of water. In error 300
pounds of salt are poured into the tank instead of 200 pounds. To correct
this condition, a stopper is removed from the bottom of the tank allowing
3 gallons of the brine to flow out each minute. At the same time 3 gallons
of fresh water per minute are pumped into the tank. If the mixture is
kept uniform by constant stirring, how long will it take for the brine to
contain the desired amount of salt?
Solution. The usual procedure in solving problems of this type is to
let a variable, say x, represent the number of pounds of salt in solution at
any time t. Then an equation is set up which will reflect the approximate
change in x in an arbitrarily small time interval flt. In this problem, 3
gallons of brine flow out each minute. Hence 3flt gallons of brine will
flow out in flt minutes. Since the 100 gallons of solution are thoroughly
mixed, and x represents the number of pounds of salt present in the solu-
tion at any instant of time t, we may assume that of the 3flt gallons of
brine flowing out, 3flt x will be the approximate loss of salt from the solu-
100
tion. (For example, if 250 pounds of salt are in the 100-gallon solution of
brine at time t, and if Ilt is sufficiently small, then 250 will be the
approximate loss of salt in time flt.) 1
If, therefore, flx represents the approximate loss of salt in time flt, then
3flt
flx - 100 x,

(the symbol means approximately equal to). The negative sign is

necessary to indicate that x is decreasing. Since no salt enters the solu-
tion, we are led to the differential equation
dx 3x dx
(a) dt = - 100' -0.03dt.
x
The solution of (a) is
(b) log x = -0.03t + c',
Lesson 15A DILUTION AND ACCRETION PROBLEMS 123

Inserting in (b) the initial conditions x = 300, t = 0, we obtain c = 300.

Hence (b) becomes
(c) x = 300e- 0 . 03t ,
an equation which gives the amount of salt in solution as a function of
the time t. And when x = 200, we obtain from (c)
(d) = e- 0 . 03f , log i = 0.03t,
from which we find
(e) t = 13.5 min,
Le., it will take 13.5 minutes for the amount of salt in the solution to be
reduced to 200 pounds.
Comment 15.11. A much shorter and more desirable method of
solving the above problem is to insert in the second equation of (a) the
initial and final conditions as limits of integration. We would thus obtain

(f) 1 x
200

dx = -0.03
%=300 t=o
dt.
Jt
Integration of (f) gives immediately
(g) log i = -0.03t, log i = 0.03t,
which is the same as (d) above.
Example 15.12. A tank contains 100 gallons of brine whose salt con-
centration is 3 pounds per gallon. Three gallons of brine whose salt concen-
tration is 2 pounds per gallon flow into the tank each minute, and at the
same time 3 gallons of the mixture flow out each minute. If the mixture is
kept uniform by constant stirring, find the salt content of the brine as a
function of the time t.
Solution. Let x represent the number of pounds of salt in solution at
any time t. By hypothesis 6 pounds of salt enter and 3 gallons of brine
leave the tank each minute. In time llt, therefore, 6llt pounds of salt
flow in and approximately C3llt) pounds flow out. Hence in time llt,
100
the approximate change of the salt content in the solution is
x llx
(a) llx 6llt - 1003llt, llt 6 - 0.03x.

We are thus led to the differential equation dx/dt = 6 - 0.03x, which

we write as
dx dx
(b) 6 - O.03x = dt, 0.03x - 6 - -dt.
124 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

As suggested in Comment 15.11, we integrate (b) and insert the initial

condition as a limit of integration. We thus obtain

f.'
(c)

whose solution is
1 %
dx
%==300 0.03x - 6
=
t==o
-dt
'

(d) x = 100(2 + e- 0 .
03t
).

NOTE. We also could have solved (b) by the method of Lesson lIB.
Example 15.13. Same problem as in Example 15.12, excepting that
3 gallons of fresh water flow into the tank instead of brine and 5 gallons
of the mixture flow out in place of 3.
Solution. Since liquid is flowing into the tank at the rate of 3 gallons
per minute and the mixture is flowing out at the rate of 5 gallons per
minute, there will be (100 - 2t) gallons of the brine in the tank at the
end of t minutes. If x represents the number of pounds of salt in solution
at time t, then the approximate change in the salt content of the brine in
a sufficiently small time interval llt is
x 5x
llx - 100 _ 2t (511t), 100 - 2t
We are thus led to the differential equation
dx 5x
(a)
dt = - 100 - 2t ' d_
5: = - 100 2t' 0 < t < 50.

Following the suggestion in comment 15.11, we integrate the second equa-

tion in (a) and insert the initial condition as a limit of integration. We
thus obtain
f.t
1
1 dx dt
%

(b)
5 %==300 X = t==o - 100 - 2t ·
Its solution is
(c)

EXERCISE I5A

It is assumed in the problems below that all mixtures are kept uniform
by constant stirring.
I. A taqk initially holds 100 gal of brine containing 30 lb of dissolved salt.
Fresh water flows into the tank at the rate of 3 gal/min and brine flows out
at the same rate. (a) Find the salt content of the brine at the end of 10 min.
(b) When will the salt content be 15 lb?
Lesson 15A-Exercise 125

2. Solve problem 1 if 2 gal/min of fresh water enter the tank instead of 3 gal/min.
3. Solve problem 1 if 4 gal/min of fresh water enter the tank instead of 3 gal/min.
4. A tank initially contains 200 gal of brine whose salt concentration is 3lb/gal.
Brine whose salt concentration is 2 lb/gal flows into the tank at the rate of
4 gal/min. The mixture flows out at the same rate. (a) Find the salt content
of the brine at the end of 20 min. (b) When will the salt concentration be
reduced to 2.5 lb/gal?
5. A tank initially contains 100 gal of brine whose salt concentration is i lb/gal.
Brine whose salt concentration is 2 lb/gal flows into the tank at the rate of
3 gal/min. The mixture flows out at the rate of 2 gal/min. Find the salt
content of the brine and its concentration at the end of 30 min. Hint. After
30 lnin, the tank contains 130 gal of brine.
6. A tank initially contains 100 gal of brine whose salt concentration is 0.6
lb/gal. Brine whose salt concentration is lib/gal flows into the tank at the
rate of 2 gal/min. The mixture flows out at the rate of 3 gal/min. Find the
salt content of the brine and its concentration at the end of 60 min. Hint.
After 60 min, the tank contains 40 gal of brine.
7. A tank initially contains 200 gal of fresh water. Brine whose salt concentra-
tion is 2 lb/gal flows into the tank at the rate of 2 gal/min. The mixture
flows out at the same rate.
(a) Find the salt content of the brine at the end of 100 min.
(b) At what time will the salt concentration reach lIb/gal?
(c) Could the salt content of the brine ever reach 400 lb?
8. A tank initially contains 100 gal of fresh water. Brine whose salt concentra-
tion is lIb/gal flows into the tank at the rate of 2 gal/min. The mixture
flows out at the rate of 1 gal/min. (a) Find the salt content of the brine and
its concentration at the end of 60 min.
9. A tank initially contains 200 gal of fresh water. It receives brine of an
unknown salt concentration at the rate of 2 gal/min. The mixture flows out
at the same rate. At the end of 120 min, 280 lb of salt are in the tank.
Find the salt concentration of the entering brine.
10. Two tanks, A and B, each contain 5000 gal of water. To each tank 150 gal
of a chemical should be added, but in error the entire 300 gal are poured into
the A tank. Pumps are set to work to circulate the liquid through the two

Figure 15.14"

tanks at the rate of 100 gal/min (Fig. 15.14). (a) How long will it take for
tank A to contain 200 gal of the chemical and tank B to contain 100 gal-
lons? (b) Is it theoretically possible for each tank to contain 150 gal?
11. The C02 content of the air in a 5000-cu-ft room is 0.3 percent. Fresh air
containing 0.1 percent C02 is pumped into the room at the rate of 1000
ft 3 /min. (a) Find the percentage of C02 in the room after 30 min. When
will the C02 content be 0.2 percent?
126 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

12. The C02 content of the air in a 7200-cu-ft room is 0.2 percent. What volume
of fresh air containing 0.05 percent C02 must be pumped into the room
each minute in order to reduce the C02 content to 0.1 percent in 15 min?

ANSWERS 15A
1. (a) x = 30e-O•3 = 22.2Ib. (b) e- O•03 ' = t, t = 23.1 min.
2. (a) x = 30(1 - 0.01t)3 = 30(0.9)3 = 21.87 lb. (b) t = 20.6 min.
3. (a) x = 30e- •O 286 = 22.5 lb. (b) t = 26.0 min.
4. (a) x = 200(2 + O 4
e- • ) = 534.1 lb. (b) t = 50 log 2 = 34.7 min.
5. 1711b, 1.32lb/gal.
6. 37.4Ib, 0.94 lb/gal.
7. (a) 252.8 lb. (b) 69.3 min. (c) Salt content approaches 400lb as e
increases without limit.
8. 981b, 0.61Ib/gal.
9. 2lb/gal.
10. (a) 27.5 min. (b) No. Tank B contains 150 gal of the chemical only if t
is infinite.
11. (a) 0.10 percent. (b) 3.47 min.
12. 3
527 ft /min.

LESSON 15H. Interest Problems. Let SA be invested at 6 percent

per annum. Then the principal P at the end of one year will be
(a) P = A(l + 0.06) if interest is compounded annually,

P -- A (1 + 0.2°6)2 1·f·Int eres t IS

· compounded semlannua
· 11y,

P -- A (1 + 0°406)4 1·f·In t eres t . compound

IS e d
quarter1y,

P = A (1 + °ir:y2 if interest is compounded monthly.

And, in general, the principal P at the end of one year will be

(b)

if the interest rate is r percent per annum compounded m times per year.
At the end of n years, it will be

(c)

If the number m of compoundings in one year, increases without limit,

then
Cd) P = A [(1 + = A [(1 +
Lesson 15H-Exercise 127

But lim
m-+CIO
(1 + !..)mlr
m
= e. Hence (d) becomes

(e) P = Aenr •
Finally, replacing n by t, we obtain
(15.2)
which gives the principal at the end of time t if $A are compounded
instantaneously or continuously at r percent per annum. The differential
equation of which (15.2) is the I-parameter family, is
dP
(15.21) (jj = rP.
(Verify it.)
Example 15.22. How long will it take for $1.00 to double itself if it is
compounded continuously at 4 percent annum.
Solution. By (15.21) with r = 0.04, we obtain
dP
(a) p = O.04dt.

Integrating (a) and inserting the initial and final conditions as limits of
integration, we have

(b)
1P==l
2 dpP = 0.04
t==o
dt, log 2 = 0.04t, t = 171 years,

approximately.
Remark 1. We could have solved this problem by using (.15.2) directly
with P = 2, A = 1, r = 0.04.
Remark 2. At 4 percent per annum, compounded semiannually, $1.00
will double itself in 17! years. Compounded continuously, as we saw
above, it doubles in 171 years. The continuous compounding of interest
property therefore is not as powerful as one might have believed.

EXERCISE 15H

It is assumed in the problems below that interest is compounded con-

tinuously unless otherwise stated.
I. In how many years will 31.00 double itself at 5 percent per annum?
2. At what interest rate will 31.00 double itself in 12 years?
3. How much will 31000.00 be worth at 41 percent interest after 10 years?
128 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

4. In a will, a man left a few million dollars, to be divided among several trusts.
The will provided that the money was to be deposited in savings institutions
and held for 500 years before being distributed to the designated legatees.
The will was contested by the government on the grounds that the monetary
wealth of the nation would be concentrated in these trusts. If the money
earned an average of 4 percent interest, approximately how much would
only 31,000,000 amount to at the end of the 500 years?
5. How much money would you need to deposit in a bank at 5 percent interest
in order to be able to withdraw 13600.00 per year for 20 years if you wish the
entire principal to be consumed at the end of this time: (a) if the money is
also being withdrawn continuously from the date of deposit as, for example,
withdrawing 33600/365 each day; (b) if the money is being withdrawn at
the rate of 3300.00 per month beginning with the first month after the de-
posit. Hint. After 1 month, money left in the bank equals AeO.05/12 - 300,
where A is the amount at the beginning of the month.
(c) Try to solve this problem assuming the more realistic situation of
interest being credited quarterly and 3900.00 withdrawn quarterly, begin-
ning with the first quarter after the deposit. NOTE. This problem no lopger
involves a differential equation. Hint. At the end of the first quarter, money
left in the bank equals A (1 + 0:5) - 900.
6. You plan to retire in 30 years. At the end of that time, you wish to have
$45,500.00, the approximate amount needed-see problem 5(b)-in order
to withdraw 3300.00 monthly for 20 years after retirement. (a) What amount
must you deposit monthly at 5 percent ? Hint. At the end of one month
P == A; at the end of two months P = AeO.05/12 +
A, where A is the
monthly deposit.
(b) What amount must you deposit semiannually if interest is credited
semiannually at 5 percent instead of continuously. NOTE. This problem
no longer involves a differential equation.

ANSWERS I5B
1. 13.86 years. (Compounded semiannually at 5 percent interest, 31.00 doubles
itself in 14 years.)
2. 5.78 percent approximately.
3. 31568.31.
4. $485,165,195,400,000.
5. (a) dP = (rP) dt - (3600) dt, A = $45,513.

(b) A = 300 C _1 = 145,418.

(c) A = 900 [1 - (1.0125)-80J = .45348

0.0125 ..".

6. (a) 345,500 = A [1
1 1.5 J,A = 354.57 monthly.
(1 025)60 - IJ
(b) 345,500 = A [ · 0.025 ,A = 3334.58 semiannually.
Lesson 15C TEMPERATURE PROBLEMS 129

LESSON 15C. Temperature Problems. It has been proved experi-

mentally that, under certain conditions, the rate of change of the tem-
perature of a body, immersed in a medium whose temperature (kept
constant) differs from it, is proportional to the difference in temperature
between it and the medium. In mathematical symbols, this statement is
written as
(15.3)

where k > 0 is a proportionality constant, T B is the temperature of the

body at any time t, and TM is the constant temperature of the medium.

Comment 15.31. In solving problems in which a proportionality

constant k is present, it is necessary to know another condition in addition
to the initial condition. In the temperature problem, for example, we
shall need to know, in addition to the initial condition, the temperature
of the body at some future time t. With these two sets of conditions, it
will then be possible to detennine the values of the proportionality con-
stant k and the arbitrary constant of integration c. And if we wish to
take advantage of Comment 15.11 we must use (15.3) twice, once to
find k, the second time to find the desired answer.

Example 15.32. A body whose temperature is 180° is immersed in a

liquid which is kept at a constant temperature of 60°. In one minute,
the temperature of the immersed body decreases to 120°. How long will
it take for the body's temperature to decrease to 90°?

Solution. Let T represent the temperature of the body at any time t.

Then by (15.3) with TM = 60,

dT dT
(a) - = -k(T - 60) T - 60 = -kdt,
dt '
where the negative sign is used to indicate a decreasing T. Writing (a)
twice as suggested in Comment 15.31, integrating both equations, and
inserting all given conditions, we obtain

(b)
i
120

T==180
dT
T - 60
= -k
fl 1==0
dt·
'
{DO
IT==180 T - 60
dT
-k t1
t==o
dt.

From the first integral equation, we obtain

(c) log 0.5 = -k, k = log 2.

With the proportionality constant k known, we find from the second inte-
130 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

gral equation
(d) log 0.25 = - (log 2)t, t = log 4 = 2 log 2 = 2
log 2 log 2 ·
Hence it will take 2 minutes for the body's temperature to decrease to 90°

EXERCISE 15C

In the problems below, assume that the rate of change of the tempera-
ture of a body obeys the law given in (15.3).
1. A body whose temperature is 100° is placed in a medium which is kept at a
constant temperature of 20°. In 10 min the temperature of the body falls
to 60°. (a) Find the temperature T of the body as a function of the time t.
(b) Find the temperature of the body after 40 min. (c) When will the body's
temperature be 50°?
2. The temperature of a body differs from that of a medium, whose temperature
is kept constant, by 40°. In 5 min, this difference is 20°. (a) What is the value
of k in (15.3)? (b) In how many minutes will the difference in temperature
be 10°?
3. A body whose temperature is 20° is placed in a medium which is kept at a
constant temperature of 60°. In 5 min the body's temperature has risen
to 30°. (a) Find the body's temperature after 20 min. (b) When will the
body's temperature be 40°?
4. The temperature in a room is 70°F. A thermometer which has been kept in
it is placed outside. In 5 min the thermometer reading is 60°F. Five minutes
later, it is 55°F. Find the outdoor temperature.
The specific heat of a substance is defined as the ratio of the quantity
of heat required to raise a unit weight of the substance 1° to the quantity
of heat required to raise the same unit weight of water 1°. For example, it
takes 1 calorie to change the temperature of 1 gram of water 1°C (or 1
British thermal unit to change the temperature of 1 lb of water 1°F). If,
therefore, it takes only to of a calorie to change the temperature of 1 gram
of a substance 1°C (or to of a British thermal unit to change 1 lb of the
substance 1°1 then the specific heat of the substance is to.
1
"),

In problems 5-7 below, assume that the only exchanges of heat occur
between the body and water.
5. A 50-lb iron ball is heated to 200°F and is then immediately plunged into a
vessel containing 100 lb of water whose temperature is 40°F. The specific
heat of iron is 0.11. (a) Find the temperature of the body as a function of
time. Hint. The quantity of heat lost by the iron body in time t is 50(0.11)
(200 - T B), where T B is its temperature at the end of time t. The quantity
of heat gained by the water-remember the specific heat of water is one-
is 100(1)(Tw - 40), where Tw is the temperature of the water at the end
of time t. Since the heat gained by the water is equal to the heat lost by the
ball, 50(0.11)(200 - T B ) = 100{Tw - 40). Solve for Tw and substitute
this value for TM in (15.3). Solve for T B • (b) Find the common temperature
approached by body and water as t --+ ex>.
LeUOD 15D DECOMPOSITION AND GROWTH PROBLEMS 131

6. The specific heat of tin is 0.05. A 10-Ib body of tin, whose temperature is
100°F, is plunged into a vessel containing 50 lb of water at 10°F. (a) Find
the temperature T of the tin as a function of time. (b) Find the common
temperature approached by tin and water as t --+ 00.
7. The temperature of a 100-lb body· whose specific heat is n
is 200°F. It is
plunged into a 40-lb liquid whose specific heat is 1 and whose temperature
is 50°. (a) Find the temperature T of the body as a function of time. (b) To
what temperature will the body eventually cool?

ANSWERS 15C

I. (a) T == 20(1 + 4e-O.06931'). (b) 25°. (c) 14.2 mm.

2. (a) k == -ilog (0.5) == 0.1386. (b) t == 10 min.
3. (a) 47.3°. (b) 12 min.
4. 50°F.
5. (a) TB .. (51 + 160e-1.055k'). (b) 48.3°F.

6. (a) T = (11 + 90e-1.0lkl). (b) 10.9°F.

7. (a) T == 100(1 + e-3k "2). (b) 100°F.

LESSON 15D. Decomposition and Growth Problems. These prob-

lems will also involve a proportionality constant and will therefore re-
quire an additional reading after an interval of time t. The method of
solution is essentially the same as that used to solve the problem discussed
in Lesson 1
Example 15.4. The number of bacteria in a yeast culture grows at a
rate which is proportional to the number present. If the population of a
colony of yeast bacteria doubles in one hour, find the number of bacteria
which will be present at the end of 3! hours.
Solution. Let x equal the number of bacteria present at any time t.
Then in mathematical symbols, the first sentence of the problem states
dx dx = k dt,
(a) dt = kx, x
where k is a proportionality constant. Writing (a) twice, integrating both
equations, and inserting all the given conditions, we obtain (omitting
percent signs)

(b)
1200 dx
- = k
11 dt·
i
z
dX=k
17/2 dt.
z-100 X t==o' z-lOO x t-O

From the first integral equation, we obtain

(c) k = log 2,
132 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

and from the second integral equation

(d) log (x/loo) =t log 2, 100 = 27/2

,x = 1131 ·

Hence 1131 percent or 11.31 times the initial number of bacteria will be
present at the end of 3! hours.
Example 15.41. The death rate of an ant colony is proportional to
the number present. If no births were to take place, the population at
the end of one week would be reduced by one-half. However because of
births, the rate of which is also proportional to the population present,
the ant population doubles in 2 weeks. Determine the birth rate of the
colony per week.

Solution. In this problem, we must determine two proportionality

constants, one for births which we call kl' the other for deaths which we
call k 2 • Using first the fact that the death rate is proportional to the
number present and that deaths without births would reduce the colony
in one week by one-half, we have
O.S -dx = -k2/1
(a)
l
%==1 1" t-O
dt
'

where x represents the population of the colony at any time t, and x = 1

stands for 100 percent. The solution of (a) is

(b) k2 = log 2.

Hence, the differential equation which takes into consideration both

births and deaths of the colony is

dx dx
(c) dt = k1x - (log 2)x, - = (k 1 - log 2) dt.
x

Integrating (c) and inserting the given conditions which reflect the net
change in the population, we obtain

(d)

The solution of (d) is

(e) log 2 = 2kl - log 4, kl = ! log 8 = 1.0397.

Hence the birth rate is 103.97 percent per week.

Lesson 15D-Exercise 133

EXERCISE 15D

In problems 1-10 below, assume that the decomposition of a substance is

proportional to the amount of the substance remaining and that the growth
of population is proportional to the number present. (A suggestion: review
Lesson 1.)
I. The population of a colony doubles in 50 days. In how many days will the
population triple?
2. Assume that the half life of the radium in a piece of lead is 1500 years. How
much radium will remain in the lead after 2500 years?
3. If 1.7 percent of a substance decomposes in 50 years, what percentage of the
substance will remain after 100 years? How many years will be required
for 10 percent to decompose?
4. The bacteria count in a culture is 100,000. In 2i hours, the number has in-
creased by 10 percent. (a) In how many hours will the count reach 200,000?
(b) What will the bacteria count be in 10 hours?
5. The populatioq of a country doubles in 50 years. Its present population is
20,000,000. (a) When will its population reach 30,000,000? (b) What will
its population be in 10 years?
6. Ten percent of a substance disintegrates in 100 years. What is its half life?
7. The bacteria count in a culture doubles in 3 hours. At the end of 15 hours,
the count is 1,000,000. How many bacteria were in the count initially?
8. By natural increase, a city, whose population is 40,000, will double in 50
years. There is a net addition of 400 persons per year because of people
leaving and moving into the city. Estimate its population in 10 years. Hint.
First find the natural growth proportionality factor.
9. Solve problem 8, if there is a net decrease in the population of 400 persons
per year.
10. A culture of bacteria whose population is No
will, by natural increase, double
in 4 log 2 days. If bacteria are extracted from the colony at the uniform
rate of R per day, find the number of bacteria present as a function of time.
Show that the population will increase if R < N 0/4, will remain stationary
if R = N 0/4, will decrease if R > N 0/4.
II. The rate of loss of the volume of a spherical substance, for example a moth
ball, due to evaporation, is proportional to its surface area. Express the
radius of the ball as a function of time.
12. The volume of a spherical raindrop increases as it falls because of the ad-
hesion to its surface of mist particles. Assume it retains its spherical shape
during its fall and that the rate of change of its volume with respect to the
distance y it has fallen, is proportional to the surface area at that distance.
Express the radius of the raindrop as a function of y.

ANSWERS 15D
I. 79 days. 5. (a) 29 years. (b) 22,970,000 approx.
2. 31 percent. 6. 658 years.
3. 96.6 percent; 307 years. 7. 31,250.
4. (a) 18.2 hours. (b) 146,400. 8. 50,240.

9. 41,660.[ (No ) t14]

10. x = 4 R + 4" - R e .
134 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

11. T = TO - kt, where TO is the initial radius and k is a positive proportionality

factor.
12. T = TO + ky where TO is the initial radius and k is a positive proportionality
factor.

LESSON 15E. Second Order Processes. A new substance C is some-

times formed from two given substances A and B by taking something
away from each; the growth of the new substance being jointly propor-
tional to the amount remaining of each of the original substances. Let
81 and 82 be the respective amounts of A and B present initially and let
x represent the number of units of the new substance C formed in time t.
If, for example, one unit of C is formed by combining 2 units of substance
A with three units of substance B, then when x units of the new substance
are present at time t
(81 - 2x) is the amount of A remaining at time t,
(82 - 3x) is the amount of B remaining at time t.
By the first sentence above, therefore, the differential equation which
represents the rate of change of C at any time t is given by
dx
dt = k(81 - 2X)(82 - 3x),

where k is a proportionality constant. In general, if one unit of C is formed

by combining m units of A and n units of B, then the differential equation
becomes
dx
(15.5) dt = k(81 - mx)(82 - nx),

where 81 and 82 are the respective number of units of A and B present

initially and x is the number of units of C present in time t.
A substance may also be dissolved in a solution, its rate of dissolution
being jointly proportional to:
1. The amount of the substance which is still undissolved.
2. The difference between the concentration of the substance in a satu-
rated solution and the actual concentration of the substance in the
solution. For example, if 10 gallons of water can hold a maximum of
30 pounds of salt, it is said to be saturated when it holds this amount
of salt. The concentration of salt in a saturated solution is then 3
pounds per gallon. When therefore the solution contains only 15
pounds of salt, the actual concentration of the salt in solution is 1.5
pounds per gallon or 50 percent of saturation.
Let x represent the amount of the substance undis80lved at any time t,
Xo the initial amount of the substance, and v the volume of the solution.
Leason 15E SECOND ORDER PROCESSES 135

Then at any time t,

(xo - x) is the amount of the substance dissolved in the solution,
xo -
- x.IS t he concentratIon
- - · 0 f t h e sub stance In
. th
e soiutlon.
·
v
If c represents the concentration of the substance in a saturated solution,
then the differential equation which expresses mathematically conditions
1 and 2 above is
(15.51) dx (
dt=kx c -
Xo -v x) ·
Problems which involve joint proportionality factors are known as
second order processes.
Example 15.52. A new substance C is to be formed by removing two
units from each of two substances whose initial quantities are 10 and 8 units
respectively. Assume that the rate at which the new substance is formed
is jolntiy proportional to the amount remaining of each of the original
substances. If x is the number of units of C formed at any time t and
x = 1 unit when t = 5 minutes, find x when t = 10 minutes.
Solution. In (15.5), SI = 10, S2 = 8, m = n = 2. Hence (15.5)
becomes
dx
(a) dt = k(10 - 2x)(8 - 2x) = 4k(5 - x)(4 - x).

Therefore

(b) 4kdt = (5 _ x)(4 _ x) = (1

(see Lesson 26) 4 _ x - 5 _ x dx. 1)
Writing (b) twice, integrating both equations and inserting all the given
conditions, we obtain

(c) 4k
1 11 ( 5
t==o
dt = %==0
1
4 - x
- 5 -
1
x
)dx;

4k to dt = Jt'
Jt-O %==0
( 1 _
4.- x
1 ) dx.
5 - x
From the first integral equation, we find

(d) k =
1 ( 5 -
20 log 4 - x
X)10 = 1 16
20 log 15 '

and from the second integral equation,

(e) 4
20
log 16) (10)
15
= log 5 -
4 - x
x_ log 45 '= log 45 (54 -- xx) .
136 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

Simplification of (e) gives

(f) (16)2 = 4(5 - x) , 5 - x 64

= -, 19x = 31, x = 1.63.
15 5 4 - x 4- x 45

Hence 1.63 units of the substance x are formed in 10 minutes.

Example 15.53. Six grams of sulfur are placed in a solution of 100 cc

of benzol which when saturated will hold 10 grams of sulfur. If 3 grams
of sulfur are in the solution in 50 minutes, how many grams will be in
the solution in 250 minutes?

Solution. Let x represent the number of grams of sulfur not yet dis-
solved at any time t. Then, at time t, (6 - x) is the amount of sulfur
dissolved and (6 - x)/100 is the concentration of sulfur in benzol. Here
c, the concentration of sulfur in a saturated solution of benzol, is given as
10/100 = 0.1, the initial amount Xo of the substance is given as 6 and
v = 100. Hence (15.51) becomes

(a) dx
dt
= k
x.
(0 1_ 6100
- x) = kx(4100+ x) .
Therefore

(b) dx _!(dX_ dX).

100 - x(x + 4) - 4 x x 4 +
Writing (b) twice, integrating both equations, and inserting all the given
conditions, we obtain

(c) -
k
25 1 50
t=o
dt=
13 (
%=6
1
--
x x
1
+
) 4
dx
'

k 1% (-1 - 1 ) dx
1
250
- dt=
25 t=o %=6 x x +4 .

From the first integral equation, we find

(d)
1
k = 2 log x +x 4 36 = 1( 3
2 log 7 - log 10
6) = 12 log 57'
and from the second integral equation,

(e)
1(1 5)
25 2 log 7 250 = log x +x 4 6 x

= log x 4- log = log C 4) ·

Lesson 15E-Exercise 137

Simplification of (e) gives

(f) = C 4)' x = 0.5 gram approximately,

which is the amount of sulfur not yet dissolved at the end of 250 minutes.
Therefore since 6 grams of sulfur were undissolved in the solution origi-
nally, 5.5 grams are in the solution at the end of 250 minutes.

EXERCISE 15E
In problems 1-8, assume all reactions are governed by formulas (15.5) or
(15.51), with the exception of problem 4 which is a modified version of
(15.51).
I. In (15.5) take 81 = 10, 82 = 10, m = 1, n = 1. If 5 units of C are formed
in 10 min, determine the number of C units formed in 50 min.
2. In (15.5) take 81 = 10, 82 = 8, m = 1, n = 1. If 1 unit of C is formed in 5
min, determine the number of C units formed in 10 min.
3. In (15.5) take m = 1, n = 1. (a) Solve for x as a function of time when
81 82 and when 81 = 82. (b) Show that as t -+ ex>, X -+ 81 if 82 81
and x -+ 82 if 82 S 81.
4. In a certain chemical reaction, substance A, initially weighing 12 Ib, is con-
verted into substance B. The rate at which B is formed is proportional to the
amount of A remaining. At the end of 2.5 min, 4 lb of B have been formed.
(a) How much of the B substance will be present after 6 min?
(b) How much time will be required to convert 60 percent of A1
Work this proBlem in two ways:
1. Letting x represent amount of A remaining at time t.
2. Letting x represent amount of B formed at time t.
Chemical reactions of this type are called first order processes.
5. A new substance C is to be formed from two given substances A and B by com-
bining one unit of A with two units of B. Initially A weighs 20 lb and B weighs
t
40 lb. (a) If 12 lb of C are formed in hr, express x as a function of time in
hours, where x is the number of units of C formed in time t. (b) What is the
maximum possible value of x?
6. A saturated solution of salt water will hold approximately 3 lb of salt per
gallon. A block of salt weighing 60 lb is placed into a vessel containing 100
gal of water. In 5 min, 20 lb of salt are dissolved.
(a) How much salt will be dissolved in 1 hr?
(b) When will 45 lb of salt be dissolved?
7. Five grams of a chemical A are placed in a solution of 100 cc of a liquid B
which, when saturated, will hold 10 g of A. If 2 g of A are in the solution
in 1 hr, how many grams of A will be in the solution in 2 hr?
8. Fifteen grams of a chemical A are placed into 50 cc of water, which when
saturated will hold 25 g of A. If 5 g of A are dissolved in 2 hr, how many
grams of A will be dissolved in 5 hr?
9. A substance containing 10 lb of moisture is placed in a sealed room, whose
volume is 2000 cu ft and which when saturated can hold 0.015 lb of moisture
per cubic foot. Initially the relative humidity of the air is 30 percent. If the
138 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

substance loses 4 lb of moisture in 1 hr, how much time is required for the sub-
stance to lose 80 percent of its moisture content? Assume the substance loses
moisture at a rate that is proportional to its moisture content and to the
difference between. the moisture content of saturated air and the moisture
content of the air.

ANSWERS 15E

1. 8! units. 2. 1.80 units.

k 2
3. x = 81 82[e ('1-'2) , - 1], x = 81 kt
81e
k('1-'2) t
- 82
1 + 81kt
4. (a) 7.47 lb. (b) 5.6 min.
5. (a) x = 180t/(2 9t). + (b) 20 lb.
6. (a) 59.2 lb. (b) 18.2 min.
7. 3.04. 8. 8.9 g. 9. dx/dt = kx[30 - (19 - x)]; 3.8 hr.

LESSON 16. Motion of a Particle Along a Straight Line-

Vertical, Horizontal, Inclined.

In this lesson we discuss a wide variety of problems involving the

motion of a particle along a straight line. In Lesson 34, we shall discuss
the motion of a particle moving in a plane.
By Newton's first law of motion, a body at rest will remain at rest, and
a body in motion will maintain its velocity, (i.e., its speed and direction),
unless acted upon by an outside force. By his second law, the rate of
change of the momentum of a body (momentum = mass X velocity) is
proportional to the resultant external force F acting upon it. In mathe-
matical symbols, the second law says

(a ) F = km dv
dt

where m is the mass of the body, t' its velocity, and k > 0 is a propor-
tionality constant whose value depends on the units used. If these are
foot for distance, pound for force, slug for mass (= 1/32 pound), second
for time, then k = 1 and (a) becomes
dv d 2s
(16.1) F = m dt = ma = m dt 2 '

where a is the rate of change in velocity, commonly called the acceleration

of the particle, and s is the distance the particle has moved from a fixed
point. A force of 1 lb therefore will give a mass of 1 slug an acceleration
of 1 ft/sec 2 • Remember that F, a, and v are vector quantities, i.e., they
not only have magnitude but also direction. (For a discussion of a vector
quantity, see Lesson 16C.) Hence it is always essential in a problem to
indicate the positive direction.
Lesson 16A VERTICAL MOTION 139

If we write
dv dv ds
(b)
dt· = ds dt
and recognize that v = ds/dt, then (b) becomes

(16.11) -dv.
dt
= dv
v-·
ds
Hence we can also write (16.1) as
dv
(16.111) F = mv-·
ds
Newton also gave us the law of attraction between bodies. If ml and
are the masses of two bodies whose centers of gravity are r distance apart,
the force of attraction between them is given by
(16.12)

where k > 0 is a proportionality constant.

LESSON 16A. Vertical Motion. Let, see Fig. 16.13,

M = mass of the earth, assumed to be a sphere,
m = mass of a body in the earth's gravitational field,
R = the radius of the earth,
y = the distance of the body above the earth's surface.

t+

Figure 16.13

By (16.12) the force of attraction between earth and body is (we assume
their masses are concentrated at their respective centers)
Mm
(16.14) F = -G (R + y)2 .

The proportionality constant G which we have used in place of k is called

the gravitational constant. The negative sign is necessary because the
140 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

resulting force acts downward toward the earth's center, and our positive
direction is upward. If the distance y of the body above the earth's sur-
face is small compared to the radius R of the earth, then the error in
writing (16.14) as
(16.15) F = _ GMm
R2
is also small. [R = 4000 miles approximately so that even if y is as high
as 1 mile above the earth, the difference between using (4000 X 5280)2
feet and (4001 X 5280)2 feet in the denominator is relatively negligible.]
By (16.1) with y replacing 8, we can write (16.15) as
GMm
(16.16) R2
Since G, M, and R are constants, we may replace GM/R 2 by a new con-
stant which we call g. We thus finally obtain for the differential equation
of motion of a falling body in the gravitational field of the earth,
dv
(16.17) m dt = -gm,

where v = dy/dt. The minus sign is necessary because we have taken the
upward direction as positive (see Fig. 16.13) and the force of the earth's
attraction is downward. From (16.17), we have
d2y
(16.18) dt 2 = -g.

The constant g is thus the acceleration of a body due to the earth's attrac-
tive force, commonly known as the force of gravity. Its value varies
slightly for different locations on the earth and for different heights. For
convenience we shall use the value 32 ft/sec 2 •
Integration of (16.18) gives the velocity equation

(16.19) v (= = -gt + CI.

And by integration of (16.19), we obtain the distance equation
2
gt
(16.2) y = - ""2 + CIt + C2·
Example 16.21. A ball is thrown upward from a building which is
64 feet above the ground, with a velocity of 48 ft/sec. Find:
1. How high the ball will rise.
2. How long it will take the ball to reach the ground.
3. The velocity of the ball when it touches the ground.
Lesson 16A VERTICAL MOTION 141

Solution (Fig. 16.22). By (16.2), with g = 32,

(a)
----,r__ y = 64
Differentiation of (a) gives

(b) v = -32t + Ct. f+

If the origin is taken at ground level, the
Ground
initial conditions are t = 0, y = 64,
v = 48. Inserting these values in (a) and
(b), we obtain Figure 16.22

(c) Cl == 48.
Hence (a) and (b) become respectively

(d) y = -16t 2 + 48t + 64, v = -32t + 48.

The ball will continue to rise until its velocity is zero. By (d), when
v = 0, t = 1.5 seconds, and when t == 1.5 seconds, y = 100 feet. Hence
the ball will rise 100 feet above the ground.
When the ball is at ground level, y = 0, and by (d) when y = 0,
t = 4 seconds. Hence the ball will reach the ground in 4 seconds. Its
velocity at that moment will then be, by the second equation in (d),

v = (-32)(4) + 48 = -80 ft/sec.

The negative sign indicates that the ball is moving in a downward direction.
Comment 16.23. In the above example, we ignored the very im-
portant factor of air resistance. In a real situation, this factor cannot be
thus ignored. Air resistance varies, among other things, with air density
and with the speed of the object. Furthermore, air density itself changes
with height and with time. It is different for different heights and may
be different from day to day. The factor of air resistance in a real problem
is thus a complicated one.
When, therefore, we assume in the examples which follow, a constant
atmosphere and an air resistance which is dependent only on the speed of
the object, we have simplified the practical problem enormously. And
when in addition we suppose that this simplified air resistance is propor-
tional to an integral power of the speed, we have simplified the problem
considerably further. There is no valid reason why air resistance may not
be proportional to the logarithm of the speed or to the square root of the
speed, etc.
In all cases, however, air resistance always acts in a direction to oppose
the motion.
142 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

Example 16.24. A body of mass m slugs is dropped from a height of

5000 feet. Find the velocity and the distance it will fall in time t. Assume
that the force of the air resistance is proportional to the first power of
the velocity, the proportionality constant being m/40.
Solution (Fig. 16.241). The force of the air resistance is given as
(m/40)v. The downward force due to the weight of the mass m is mg
pounds. Hence the differential equation of
---w'--y==o motion (16.17) must be modified to read,
with the poBitive direction dO'UJnward (remem-
ber mass X acceleration of a body = the net
l
+ forces acting upon it),
dv m
y=5,OOO Ground
(a) m dt = gm - 40 v.

Figure 16.241 Note that the force of gravity gm is now posi-

tive since it acts in the chosen positive direc-
tion. This equation can be solved by the method of Lesson 60 or liB.
Using the latter method, we write (a) as

dv 1
(b) dt + 40 v = g.

The integrating factor by (11.12) is et/40. The solution of (b) is therefore

(c)

Integration of (c) gives

(d)

If the origin is taken at the point where the body is dropped, then the
initial conditions are t = 0, v = 0, y = o. Substituting these values in
(c) and (d), we find
(e) Cl = -40g, C2 = -1600g.

Hence the two required equations are

(f) v = 40g(1 - t
e- / 40 ), y = 40g(t + 40e-t / 40 - 40).

Comment 16.25. We see from (f), that as t --+ 00, the velocity
" --+ 40g. This means that when a resisting force is present, the velocity
does not increase indefinitely with time but approaches a limiting value
beyond which it will not increase. This limiting velocity is called the
terminal velocity of the falling body. In this example, it is 40g ft/sec.
LeUOD 16A VERTICAL MOTION 143

Example 16.26. The problem and initial conditions are the same as in
Example 16.24 excepting that the force of the air resistance is assumed to
be proportional to the second power of the velocity. Find the velocity of
the body as a function of time and also the terminal velocity of the body.
Solution. Here the force of the air resistance is (m/40)v 2 • Hence
(a) of Example 16.24 must be modified to read
dv m 2 dv 40g - v2 dv 1
(a) m dt = my - 40 v , dt = 40 40g - v2 = 40 dt.

Integrating the last equation in (a) and inserting the initial conditions as
limits of integration, we obtain

(b)
v
i-o (2V1Qg)2
dv
- v2
_
-
1.-
40
itt==o dt
.
I ts solution is

(c) 1 log (2V1Qg + v) = 1.- t 2V1Qg

--=---- -
+ v _
e(..JI0g/10)t .
4V1Qg 2v'IOg - v 40 ' 2V1Qg - v
Solving the last equation for v, we obtain

(d) v = 2v'IQg
e(.,JIOg/IO)t - 1) ,
( e(.,J IOg/ 10) t +1
which gives the velocity of the body as a function of t.
As t -+ 00, we see from (d) tha t v -+ 2v'IOY. This is the terminal
velocity of the body.
Example 16.27. A raindrop falls from a motionless cloud. Find its
velocity as a function of the distance it falls. Assume it is subject to a
resisting force which is proportional to the second power of the velocity.
Also find its terminal velocity.
Solu tion. Taking the downward direction as positive, the differential
equation of motion (16.17) must be modified to read

(a) dv 2
mdt=my-kv.

where k > 0 is a proportionality constant. Since we wish to find v as a

function of the distance y, we replace dv/dt by its equal as given in (16.11).
Hence (a) becomes
dv 2 v dv dy
(b) mv-
dy = my - kv , my - kv 2 = m·
If the cloud is taken as the origin, then the initial conditions are y = 0,
144 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

v = O. Integration of (b) and insertion of the initial conditions give

(c)
f. " ,,=0 mg
V
-
dv
kv 2
= 1.. 111
m 11==0
dy
'
whose solution is
(d) - log (my ;, kv 2
) = !,
mg _ kv 2 = mge- 2klll"',

v2 = 7' (1 - e-2klllm).

As t 00,the distance y the raindrop falls approaches infinity, and as

y 00, we see from (d) that v2 mg/k. Hence the terminal velocity is

(e) T. V. = Vmg/k.
Note. Since the body is falling and the downward direction is positive,
the positive square root must be taken for the velocity in the last equation
of (d).
Comment 16.28. The terminal or limiting velocity has no y in it,
and is therefore independent of the height from which the raindrop falls.
It is also independent of the initial velocity. From actual experience we
know that a raindrop reaches its limiting velocity in a finite and not in
an infinite time. This is because other factors also operate to slow the
raindrop's velocity.

Comment 16.29. A body falling in water encounters a resistance

just as does the body falling in air. If the magnitude of the velocity is small,
the resistance of the water is approximately proportional to the first
power of the velocity. The differential equation of motion (16.17) there-
fore becomes, with the downward direction positive,
dv
(a) m dt = mg - kv,

which is similar to (a) of Example 16.24.

Example 16.3. A man with a parachute jumps at a great height from
an airplane moving horizontally. After 10 seconds, he opens his para-
chute. Find his velocity at the end of 15 seconds and his terminal velocity
(i.e., the approximate velocity with which he will float to the ground).
Assume that the combined weight of man and parachute is 160 pounds,
and the force of the air resistance is proportional to the first power of the
velocity, equaling tV when the parachute is closed and 10v when it is
opened.
Leeson 16A VERTICAL MOTION 145

Solution. For the first 10 seconds of fall, the differential equation of

motion (16.17) of the man is, with poBitive direction downward,
dv
(a) m dt = mg - Iv·
Here the downward force mg is equal to 160 pounds and the mass m =
160/32. Hence (a) becomes

160 dv = 160 - Iv dv 1
(b)
32 dt ' dt + 10 v = 32.

Its solution, by the method of Lesson lIB, is

(c) v = 320 + ce- o. u .

If we take the origin at the point of jump, then t = 0, v = O. Hence by
(c), we find c = -320 so that
(d) v = 320(1 - e- o. U ).

When t = 10
(e) v = 320(1 - e- 1) = 320(0.6321) = 202.3 ft/sec.

Starting with the tenth second, the differential equation (16.17) be-
comes (remember the resistance is now 10v)
160 dv dv
(f) 32 dt = 160 - 10v, dt + 2v = 32,

whose solution is
(g) v = 16 + ce- 2t •
Inserting in (g) the initial condition which, by (e), is t = 0, v = 202.3,
we find c = 186.3. Hence (g) becomes

(h) v = 16 + 186.3e-2t •
When t = 5, i.e., 5 seconds after the parachute opens and 15 seconds
after his jump,
(i) v = 16 + 186.3e- 10 = 16 + 186.3(0.000045)
= 16 + 0.008 = 16.008 ft/sec.

The tenninal velocity is [in (h) let t 00] 16 ft/sec. We see from (i),
therefore, that only 5 seconds after the parachute is opened, the man is
already floating to earth with a practically steady velocity of 16 ft/sec.
146 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

Comment 16.31. In deriving formula (16.17) for a vertically falling

body, we ignored the distance y of the object above the earth's surface,
since we assumed it to be relatively small in comparison with the radius
R of the earth. If, however, the dis-
tance of the object is very far above
f
+ the earth's surface, then this distance
cannot be thus ignored. In this case
(16.14) becomes
R Mm
(16.32) F= - G - , r2
r=O where M is the mass of the earth con-
sidered as being 'concentrated at its
Figure 16.33 center and r is the distance of the
body of mass m from this center (Fig.
16.33). Replacing in (16.32) the value of F as given in (16.1), we obtain

dv Mm dv GM
(16.34) m- = -G--, dt=-rT·
dt r2

Since G and M are constants, we can replace GM by a new constant k.

There results
dv k k
- -,
(16.35) - --,
dt r2 r2

where v = dr/dt. From (16.35) we deduce that the acceleration of

a body in the gravitational field of the earth varies inversely as the square
of the distance of the body from the center of the earth.

,Example 16.36. A body is shot straight up from the surface of the

earth with an initial velocity Vo. Assuming no air resistance, find:
1. The velocity v of the body as a function of the distance r from the
center of the earth.
2. Its velocity when it is 4000 miles above the earth's surface.
3. How high the body will rise.
4. The magnitude of the initial velocity Vo in order that the body may
escape the earth, i.e., in order that it may never return to the earth.
5. The time t as a function of the distance r of the body from the earth's
center.
Solution (Fig. 16.361). We take the origin at the center of the earth,
and call R the radius of the earth. Then, by (16.35),

dv k
(a) a -- -dt -- - r2 ·
Le8son 16A VERTICAL MOTION 147

Substituting in (a), the initial conditions r = R, a = -g, we find k = gR2.

Hence (a) becomes
(b)

Since we wish to find v as a function of the distance r, we replace dv / dt by

its equivalent value as given in (16.11). Hence (b) becomes
dv gR 2 gR 2
(c) v-=---,
dr r2 V dv = - T"2 dr.
Integration of (c) and insertion of the initial conditions v = vo, r = R,
.
gIve
(d)
1"
11=110
vdv = -gR
2 dr
2'
r==R r
lr
whose solution is

(e) v2 = vo 2 + - 2gR = vo 2 + 2gR - 1).

Hence the answer to question 1 is

(f) v = ± + 2gR - 1) i
the positive sign is to be used when the body is rising, the negative sign
when it is falling. When the body is 4000 miles above the earth's surface,

t+
Surface of the earth
r=R
r

r=O,
center of earth

Figure 16.361

r == 8000 (R = 4000 miles approximately). Inserting this value in (f), we

obtain
(g) v = ± + 2gR (S!O - 1) = ± VV02 - 4000g,

which is the answer to question 2.

148 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

The body will continue to rise until v = O. Hence by (f)

(h) o= vo 2 + 2gR (-r R

- )
1 , r = 2g R2gR2
- Vo 2 '

which is the distance the body will rise above the center of the earth if
fired with an initial velocity Vo. Subtracting R from this value will give
the distance the body will rise above the earth's 8urface. This is the
answer to question 3.
The body will escape the earth, i.e., it will never return to the earth,
if r increases with time. This means we want r to become infinite as the
velocity v of the body approaches zero. By (e) we see that if vo 2 = 2gR,
then r -+ 00 as v -+ o. Hence the answer to question 4 is
(i) Vo = V2gR = V(2)(32)(4000)(5280)
= 36,765 ft/sec = 7 mi/sec,* approx.,
= 25,100 mi/hr, approx.,
which is the escape velocity of a body if air resistance is ignored.
The answer to question 5 is somewhat more difficult to obtain. In (e)
replace v by dr/dt and let
(j) b = vo 2 - 2gR.
Hence (e) becomes

(k) v = dr
dt
= ±
r
+b= ± +r br = ± .!r var + br2.
When the body is rising, the velocity is positive and we can, therefore,
write (k) as
(1) dt = r dr _ (a 2br) dr _ _a + dr
Var + br2 - 2b Var + br2 2b Var br2 +
If we assume b < 0, i.e., if we assume [see (j)] vo 2 < 2gR, so that the
body cannot escape the earth, then integration of (1) gives

(m) t = c + 1b v'ar + br2 - a

...;=b Arc sin (-2br - a) , b < 0,
2b -b a
where a,b have the values given in (j).
Substituting in (m) the initial conditions t = 0, r = R, we obtain

(n) c = -
1
bvaR 2 a sin (-2bR -
+ b:R + 2b...;=bAre a) , b < O.
-b a
With this value of c, (m) defines t as a function of r for a rising body.
·With g = 32 ft/sec 2, Vo = 6.96 mi/sec instead of 7 mi/sec. However, g is closer to
32.17 ft/sec 2• With this value of g, Vo = 6.98 mi/sec.
Lesson 16A VERTICAL MOTION 149

Remark. When the body is falling, v is negative. Hence for a falling

body, we must, in (k), take
(0) v= -
r
+ b'
in order to arrive at an equation comparable to (1) above. Or, if you wish,
you may use formula (d) of Example 16.38 following. It gives the time of
a falling body as a function of T with initial conditions t = 0, v = 0,
r = roo
If we assume that b = 0, i.e., if we assume v0 2 = 2gR [see (j)] so that
the body will escape the earth, then (e) becomes
dr V2g R rl/2 dr = V2g R dt, 1r 3/2 = V2g Rt + C.
(p) v = dt = r1/2 '

When t = 0, r = R. Hence c = iR3/2. Therefore the last equation in

(p) becomes
(q)

Comment 16.37. 1. Note from (16.35) that, as T --+ 00, the accelera-
tion d 2rIdt 2 due to the gravitational force of the earth approaches zero.
This means that the influence of the earth's gravitational field, although
never zero, becomes insignificant.
2. From (e) we observe that when v0 2 = 2gR, the escape velocity of
the body, the velocity equation reduces to v2 = 2gR 2 1r. Hence as r gets
larger, the velocity of the body will continue to get smaller until such time
as it enters the gravitational field of another heavenly body. And if
v0 2 > 2gR so that v0 2 - 2gR equals a positive constant k 2, then the
2gR 2
velocity equation (e) reduces to v = k 2 2
+ r
.
Hence as r --+ 00,
v --+ k.
3. From equation (q) above, which expresses time as a function of r
with Vo 2 = 2gR, we see that t also approaches infinity as r --+ 00.
Example 16.38. A body falls from interstellar space at a distance
To from the center of the earth. Find:

1. Its velocity v as a function of the distance r, where r is measured

from the center of the earth.
2. Its velocity when it reaches the surface of the earth.
3. The time t as a function of the distance r.
Take the earth's center as the origin and the outward direction as positive.
Solution. The differential equation of motion of the body is the same
as that of (c) in the previous example. Integration of this equation and
150 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

insertion of the initial conditions gives

11 2 jr dr
(a) vdv = -gR 2·
/.11=0 r=ro r
Its solution is
(b) v2 = 2gR 2 - - - ,
r ro
(1 1)
which is the answer to question 1.
When r = R, i.e., when the body is at the surface of the earth, its
velocity is, by (b),
(c) V = 2gR - 2gR2 ·
-
ro
The negative sign is needed because the body is moving toward the earth
and the outward direction is positive. This is the answer to question 2.
From (c) we see that if ro is very large, i.e., if the body is extremely far
away from the earth, then 2gR2/ro is very close to zero, and Ivl is extremely
close to, but less than V2gR. This means that a body falling from outer
space can never exceed a velocity equal to V2gR. As we saw in (i) of
Example 16.36, V2gR = 25,100 miles/hour. Hence, if air re8istance is
ignored, a body falling from interstellar space will have a velocity at the
earth's surface which differs extremely little from 25,100 miles/hour.
We leave it to you as an exercise to show that the answer to question 3 is

(d) t =
RV2g
[vror - r2 + ro2 (r2 _Arc sin 2r ro- ro)]

=
RV2g
( vror - r2 + -Arc
ro
2
cos 2r -
ro
ro) ·

Hint. Start with (b) and follow the method we used in Example 16.36
to find the answer to question 5. You do not need to make the substitu-
tions (j) of Example 16.36.
Comment 16.39. In solving the two previous problems, we assumed
no air resistance. Since there is air resistance, the escape velocity vo
would have to be sufficiently greater than V2gR = 25,100 miles/hour to
overcome this resistance. If, however, the body emerged from the earth's
atmosphere, which is rare 100 miles above its surface, * with a velocity
equal to or perhaps very slightly more than 25,100 miles/hour, it would
escape the earth. Conversely, the formula for the velocity of a body
falling from outer space will give fairly acourate results until the object
reaches the earth's atmosphere or about 100 miles from its surface. Con-
*A calculation made from an analysis of one of our satellite's orbits shows that the
density of atmosphere at 932 miles above the earth is one thousand million millionths
the density of air at sea level.
Lesson 16A-Exerclse 151

sidering the great distances involved, 100 trliles is relatively insignificant,

but its importance is tremendous. It complicates the whole problem of
exit and reentry of satellites.

EXERCISE 16A
I. Verify the accuracy of the answer as given in the text to question 3 of
Example 16.38.
In problems 2-5, assume no air resistance and that the object is
near the earth's surface.
2. A ball is thrown vertically upward from the ground with an initial velocity
. of 80 ft/sec.
(a) Find its velocity and distance equations as functions of time. Take the
origin at the point where the ball is thrown and the upward direction
as positive.
(b) What are its velocity and height at the end of 1 sec?
(c) How long and how high will it rise?
(d) When will it reach the ground and with what velocity?
(e) Would the results differ if the ball were a projectile weighing 5 tons?
3. A man leans over the side of a bridge and drops a stone. His stop watch
shows that the stone touched the water in 2.1 seconds. How high is the
bridge above the water?
4. A ball is given a downward velocity of 8 ' ft/sec from a height of 120 ft above
the ground.
(a) When will it reach the ground and with what velocity will it strike the
ground?
(b) How much lower must one stand in order to drop a ball and have it reach
the ground at the same time as the first ball?
(c) With what velocity will the second ball strike the ground?
(d) What is the significance of the negative sign of -3 seconds obtained
in (a)? Hint. Substitute t = -3 in your velocity equation. Then solve
this problem: If a ball is thrown upward from the ground with a veloc-
ity of 88 ft/sec, how high will it go; at what height will its velocity be
8 ft/sec downward; when will it reach this height and velocity?
S. A person, 81 ft above the ground, drops an object. With what velocity must
a second person 180 ft above the ground throw an object straight down in
order that both objects reach the ground at the same time?
In the problems below, weight in pounds is equal mass in slugs times
the acceleration of gravity in feet per second per second, i.e.,
(16.391) W = mg, m = W/g.
6. A man weighs 160 lb on earth.
(a) What is his mass?
(b) The acceleration of gravity on the surface of the moon is aprroximately
one-sixth that of the earth. What will he weigh on the surface of the
moon?
(c) Find formulas comparable to the velocity and distance equations
(16.19) and (16.2) for the surface of the moon.
152 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

7. A ball thrown vertically upward from the surface of the earth with a velocity
of 64 ft/sec will reach a maximum height of 64 ft in 2 sec (verify it). If
thrown with the same velocity on the surface of the moon, find, by means
of the formulas developed in 6(c), comparable figures for the maximum height
reached and the time needed to attain this height.
8. If a man can high-jump 5 ft on earth, how high will he jump on the moon
and how much longer will he be in the air as compared with the time in the
air on the earth? Assume his center of gravity is 2 ft from the top of the
bar. Hint. See problem 7.
9. A man whose weight is 160 lb is in an elevator which is descending with an
acceleration of 2 ft/sec 2 • What is his weight while riding in the elevator?
Hint. Use (16.391); remember his mass is constant, and when the elevator
accelerates down, he accelerates up.
In problems 10-17 and 20,21, assume that the force R of the air re-
sistance is proportional to the first power of the velocity, i.e.,
R = kv, and that the falling or rising body is near the earth's surface.
10. A body of mass m is dropped from a great height.
(a) Find its velocity and the distance it falls as functions of time. Take
the positive direction as downward and the origin at the point where
the body is dropped. Hint. In (a) of Example 16.24 replace m/40 by k.
Use your results to check the accuracy of the ans\\yers given in (f).
(b) What is its terminal velocity?
II. Solve problem 10 if the body initially is given a downward velocity of
Vo ft/sec. What is its terminal velocity? COlnparc with 10(b) above. Note
that the initial velocity does not affect the terminal velocity.
12. A body weighing 1921b is dropped from a great height. The proportionality
constant k of the air resistance is 12.
(a) Find its velocity and the distance it falls as a function of time. Solve
independently. Use results obtained in 10 only as a check.
(b) What is its terminal velocity?
(c) How far does it fall in 10 sec? What is its velocity at that moment?
13. Solve problem 12 if the body is given an initial downward velocity of
170 ft/sec instead of being dropped. Solve independently. Use the results
obtained in 11 only as a check.
14. When a paratrooper falls freely from a great height before opening his chute,
his terminal velocity is approximately 175 ft/sec. Assume a paratrooper
and his chute together weigh 200 lb.
(a) Find the proportionality factor k of the air resistance. Hint. Use the
formula for T.V. found in 10(b).
(b) Find his velocity and the distance he falls as a function of time.
(c) What is his velocity at the end of 81 sec, 17! sec, 26!sec, 32i sec?
(d) How far has he fallen in 321 sec?
IS. Assume the paratrooper of problem 14 opens his chute when he has reached
his terminal velocity of 175 ft/sec, and that his chute is designed to give
him a safe landing speed of 16 ft/sec.
(a) What is the new value of k? Hint. Use the formula for T.V. found in
problem 11.
(b) Find his velocity and the distance he falls after he opens his chute as
functions of time.
Lesson 16A-Exercise 153

(c) What is his velocity at the end of 1 sec, 2 sec, 3 sec, 4 sec, 5 sec?
_ (d) How far has he fallen in 5 sec?
(e) Do your answers in (c) and (d) suggest a safe height at which he can
open his chute?
(f) If he opens his chute at a height of 1040 ft, in how many seconds does
he reach the ground?
16. A paratrooper jumps from a plane flying horizontally at a great height.
When he feels that he has reached a steady velocity (i.e., when he has
reached his terminal velocity), he opens his chute. Assume this steady veloc-
ity is 180 ft/sec.
(a) Find his velocity and the distance he falls as a function of time before
the chute opens. Hint. Use formula for T.V. found in 10(b) and solve
for mike
(b) Calculate his velocity at the end of II! sec, 221 sec, 331 sec, 45 sec.
(c) How far has he fallen in 45 sec?
17. If the force of the air resistance is 50 lb when a body is falling at a velocity
of 25 ft/sec, what is the value of the proportionality constant k of the air
resistance? For this k, find the terminal velocity of a falling body weighing
100 lb. If the body has an initial velocity of 20 ft/sec, find its velocity and
distance equations as functions of time.
18. A body weighing 96 lb begins to sink as soon as it is placed in water. Two
forces act on it to oppose its motion, an upward force due to the buoyancy
of the object and a force due to the resistance of the water. Assume the
buoyant force is 12 lb and the resistance of the water is 6v. Take the origin
on the surface of the water and downward direction as positive.
(a) Find the velocity and position of the body as functions of time.
(b) Find its terminal velocity.
19. The specific gravity of a body is defined as the ratio of its weight to the
weight of an equal volume of water. Assume a body is released from the surface
of a medium whose specific gravity is one-fourth that of the body and that the
resistance offered by the medium is mv /3.
(a) Find the velocity of the body as a function of time. Hint. The specific
gravity of the medium equal to i that of the body, implies that the medi-
um's upward buoyant force is one-fourth the weight of the body.
(b) Find its terminal velocity.
20. A body of mass m is shot straight up from the ground with an initial velocity
of Vo ft/sec.
(a) Find the velocity and position of the body as functions of time. Take
the positive direction upwards and the origin on the ground.
(b) How high will the body rise and when will it reach this maximum height?
21. An object weighing 64 lb is shot straight up from the ground with an initial
velocity of 96 ft/sec. Assume the force of the air resistance is 4v.
(a) Find the velocity and position of the object as functions of time.
(b) How high will the body rise and when will it reach this maximum
height? Check the accuracy of your answers in (a) and (b) with the
formulas obtained in problem 20.
(c) When and with what velocity will it strike the ground? Note that the
down trip takes longer than the up trip, whereas when there is no air
154 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

resistance both times are the same. Note also that the return velocity
is smaller than the initial velocity.
(d) Do you believe you would get the same answer for the velocity of the
falling body when it strikes the ground and for the time of the down
trip if you used the formulas for a falling body as found in problem 10,
with y having the value determined in (b)? Try it.
In problems 22-30, assume that the force R of air resistance is pro-
portional to the second power of the velocity, i.e., R = kv 2 , and
that the falling or rising body is near the earth's surface.
22. In Example 16.27, we found the velocity v of a falling body as a function of
the distance fallen. Starting with equation (a) of this example, find the
velocity and distance of a falling body as functions of time. Find its terminal
velocity. Compare with (e) of Example 16.27.
23. A body of mass m falls from a great height. Its initial velocity is Vo ft/sec.
Find:
(a) Its velocity as a function of the distance fallen. Take the positive direc-
tion downwards and the origin at the point of fall.
(b) Its velocity as a function of time.
(c) Its terminal velocity. Compare with (e) of Example 16.27 and with
problem 22. Note that the initial velocity does not affect the terminal
velocity.
24. A body of mass m is fired vertically upward from the ground at an initial
velocity of Vo ft/sec.
(a) Find its velocity as a function of its height. Take positive direction
upwards and origin on ground.
(b) Find its velocity as a function of time.
(c) Find its height as a function of time.
(d) When will it reach maximum height?
(e) How high will it rise?
NOTE. These formulas are valid only while the body is risin, When it begins
to fall, formulas developed in Example 16.27 and problem 22 must be used.
Compare with problem 21 where ,ve used one formula to calculate the time
for a round trip.
25. With what velocity will the body of problem 24 return to the earth and how
long will it take for its descent? Hint. Read note in problem 24. To find
the velocity, use (d) of Example 16.27 with y equal to the value of the
maximum height as found in problem 24( e). Note that the returning velocity
is less than the initial velocity Vo. To find the time of descent, use the formula
for y as found in problem 22. Here it is not easy to see that the time of
descent is longer than the time of ascent.
26. The velocity of a parachutist at the moment his chute opens is 160 ft/sec.
The force of the air resistance is mv 2 /8. Find:
(a) His subsequent velocity and distance as functions of time. Take the
origin at the point where the chute opens and the positive direction
do\vnwards.
(b) His velocity 1 sec after his chute opens; 2 sec after.
(c) His terminal velocity.
(d) How far he falls in the first second; in the second second.
(e) Approximately when he reaches the ground if he is 1064 ft above the
earth when his chute opens.
Lesson 16A-Exercise 155

27. A body of mass m is dropped from a plane flying horizontally 1 mile above
the earth. The force of the air resistance is 2mk 2v2. The terminal velocity
of the body is 100 ft/sec. Find:
(a) The value of the constant k. (Hint. T.V. = Ymg/k; replace k by 2mk2.)
(b) The velocity of the body a.s a function of time.
(c) The velocity of the body at the end of 3 sec.
(d) When the body reaches a velocity of 60 ft/sec.
28. A body falls from a great height. Its terminal velocity is 10 ft/sec. (a) Find
its velocity and distance equations as functions of time. Hint. T.V. =
Y mg/k. Solve for kim. (b) Find its velocity equation as a function of dis-
tance.
29. A paratrooper and his chute, which together weigh 192 lb, drop from an
airplane moving horizontally. He opens his chute at the end of 10 sec.
Assuming the proportionality constant of air resistance is 1/120 when the
chute is closed and 4/3 when it. is open, find:
(a) His velocity as a function of time before the chute is opened.
(b) His terminal velocity before the chute is opened.
(c) His velocity at the end of the first 10 sec.
(d) His velocity as a function of time after the chute is opened.
(e) His terminal velocity after the chute is opened.
(f) His velocity at the end of 15 sec, i.e., his velocity 5 sec after the chute is
opened.
Solve independently and then check your results with the formulas found in
problems 22 and 23.
30. A man and his parachute weigh 192 lb. Assume that a safe landing velocity
is 16 ft/sec and that air resistance is proportional to the square of the velocity,
equaling! lb for each square foot of cross-sectional area of the parachute
when it is moving at 20 ft/sec at right angles to the direction of motion.
What must the cross-sectional area of a parachute be in order that the para-
trooper land safely? llint. First find the force of the air resistance. Then
find k such that T.V. = 16. Then find the number of square feet of para-
chute that will make the force of the air resistance equal to kv 2 •
In problems 31-39, assume no air resistance and that the object is
far enough from the earth so that equation (16.35) applies. Use the fol-
lowing data: R = 4000 miles, g = 32 ft/sec 2 , V2iR = 6.96 mi/sec.
31. With what velocity must a rocket be fired in order to reach a height of 400
mi above the earth; 4000 mi above the earth? Solve independently. Check
your results with (h) of Example 16.36.
32. The "air" 200 mi above the earth is so thin that it will hardly slow a space
vehicle. It is called the F -2 region of the atmosphere. What velocity should
a rocket have at 200 miles above the earth, if all its fuel is exhausted at that
point, in order to go another 3800 mi?
33. A body is shot to a height of 400 mi and then starts to fall. What is its
velocity (a) when it has fallen 200 mi and (b) when it is at the surface of
the earth? Hint. Use (a) of Example 16.38 with ro = R +400 = 4400.
Note that the velocity, when it reaches the ground, is the same as the
velocity required to propel it 400 miles upward. See problem 31.
34. Assume a body falls from rest at a distance of 61 R, i.e., at a distance of
244,000 mi from the center of the earth (equivalent to the moon's distance
from the center of the earth). With what velocity and in how many hours
will it reach the earth?
156 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

35. A body is fired straight up with an initial velocity equal to the escape veloc-
ity v2gR. Find:
(a) The velocity of the projectile as a function of the distance r from the
center of the earth.
(b) When it will have reached 244,000 mi, the distance of the moon from
the center of the earth.
Hint. Use (e) and (q) of Example 16.36.
36. A body is fired straight up with an initial velocity Vo whose magnitude is less
than escape velocity. When will it reach its maximum height? Hint. The
maximum height the body will reach is given in (h) of Example 16.36.
Express this value of r in terms of a and b as defined in (j). Then use (m)
and (n). NOTE. This time equation is valid only for a rising particle. If
you wish to compute its return time, you must use the time equation given
in (d) of Example 16.38, with ro equal to the height from which it begins
its fall.
37. Show that if Vo is very much less than the escape velocity v2gR, then the
time for the object to reach its maximum height, as given in problem 36,
is approximately vo/g. Hint. Replace the Arc sin function by its series ex-
pansion, Arc sin x = x + x 3 /6 + · · ..Then eliminate vo 2 and higher
powers of Vo. To see some justification for this elimination, let Vo = io
mi/sec in the time equation of problem 36.
38. A body is shot straight up from the surface of the moon with an initial
velocity Vo. (a) Find the velocity v of the body as a function of its distance
rm from the center of the moon. (b) Find the escape velocity of the body.
The radius of the moon is approximately 1080 mi; the acceleration of gravity
on the surface of the moon is approximately one-sixth that of the earth.
Take the outward direction from the moon as positive.
39. (a) Prove that if a particle were placed approximately nine-tenths of the
distance D from the center of the earth to the center of the moon on a line
connecting moon to earth, Fig. 16.392, the particle would be at rest, Le., the
gravitational pulls of moon and earth on a particle placed nine-tenths of the
distance from the center of earth to the center of the moon are equal. As-
sume the mass of the moon is 1/81 the mass of the earth. Hint. Apply
(16.32) to both earth and moon. Then equate the two forces.

,.......---- r ------il......--D - r

Moon

9D
10

Figure 16.392

(b) Set up the differential equation of motion of a particle shot from the
surface of the earth toward the moon, considered fixed, taking into account
both the earth's and moon's gravitational attractions. Solve the equation
Lesson 16A-Answers 157

with t = 0, v = vo. Take the positive direction as outward from the earth,
and the orgin at the earth's center.
(c) At what velocity must the particle be fired in order to reach the
neutral point? Hint. In the velocity equation found in (b), you want
v = 0, when r = faD. Assume Rm 2 = 6R2/81, D = 61R R/4, +
g", = g/6, where R", is the radius of the moon and g", is the acceleration due to
the force of gravity of the moon. Note the small effect of the moon's gravita-
tional attraction.
(d) At what velocity must a projectile be fired in order to reach the moon?
40. If a body were dropped in a hole bored through the center of the earth, it
would be attracted toward the center with a force directly proportional to
the distance of the body from the center.
(a) With what velocity will it pass the center?
(b) When will it reach the other end of the hole?
NOTE. The motion of the body is known as simple harmonic motion. A
fuller discussion of this motion can be found in Lesson 28.

ANSWERS 16A
2. (a) v = -32t +
80, y = -16t 2 80t. +(b) 48 ft/sec, 64 ft.
(c) 21 sec, 100 ft. (d) 5 sec, -80 ft/sec. (e) No.
3. 70.56 ft.
4. (a) 21 sec, 88 ft/sec. (b) 20 ft. (c) 80 ft/sec.
(d) If 3 sec ago, the ball were thrown upward from the ground with a
velocity of 88 ft/sec, it would reach a height of 121 ft and have a velocity
of 8 ft/sec as it passed the 120 ft point on its way down.
5. 44 ft/sec.
16t
6. (a) 5 slugs. (b) 261 lb. (c) v = - 3 Cl, y = -it +
2
CIt C2. + +
7. 384 ft, 12 sec.
8. 15 ft if one assumes he scales the bar horizontally so that he raises his
center of gravity 2 ft on earth; 6 times as long.
9. 150 lb.
2
10• () = mg (1 _ -k tl m) = mg t+ m g ( -k tl m _ 1)
a v k e ,Y k k2 e .
(b) T.V. = mg/k.
II• V =
mg (1
k -ktlm)
- e + voe,
-ktlm

Y = ":: t- - :0) (1 _ e- kllm

).

T.V. = mg/k, same as in 10(b).

12. (a) v = 16(1 - e- 2t ), y = 16t +
8(e- 2t - 1). (b) 16 ft/sec.
(c) y = 8(19 +
e- ) = 152 ft, v = 16(1 - e- ) = 16 ft/sec.
20 20

13. (a) v = 16(1 - e- 2t ) +

170e- 2t = 16 +
154e- 2t , y = 16t 77(1 - e- 2t ). +
(b) 16 ft/sec. (c) 237 ft, 16 ft/sec.
14. (a) k = -to
2
(b) v = 175 (1 - e-3211 17 5
), Y = 175t + 1;: (e -32 If 1 7 5 - 1).
(c) 139.7 ft/ser, 167.9, 173.6,174.6. (d) 4791 ft.
158 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

15. (a) 12.5. (b) v = 16(1- e-21 ) + 175e-21, y = 16t + 1;9 (I __ e-21).
(c) 37.5 ft/sec, 18.9, 16.4, 16.05, 16.01. (d) 159.5 ft.
(e) Over 160 ft. (f) Approx. 60 sec.

16. (a) v = 180(1 - e -81/45), y = 180t + I::

2
(e- 81145 - 1).
(b) v(II!) = 180(1 - e- 2 ) = 155.6 ft/sec, 176.7, 179.6, 179.9.
(c) Approx. 7100 ft.
17. k = 2; T.V. = 50 ft/sec; v = 50 - 30e -11,/50 = 50 - 30e -0.64,;
y = 50t - 1500 (1 _ e- II '/50) = 50t _ 375 (1 _ e- O.64 ,).
g 8
18. (a) v = 14(1 - e- 2t ), y = 14[t - 1(1 - e- 2 '»).
(b) T.V. = 14 ft/sec.
19. (a) v = 72(1 - e-t/3 ). (b) T.V. = 72 ft/sec.

20• ()
a v
gm ( -ktlm
=k e
1)
-
-ktlm + voe ,

y = (gm2 + mvo)
(1 _ e-k'lm) _ gm t
k2 k k ·
(b) y = -
m ( vo - -gm log kvo + gm) ,t = -m log kvo + gm ·
k k gm k gm
21. (a) v = 16(7e-2t - 1), y = 56(1 - e- 2t ) - 16t.
(b) y = 8(6 - log 7) = 32.4 ft, t = i log 7 = 0.97 sec.
(c) 3.5 sec approx., -16 ft/sec" (d) Same answer.
2.Jklll m' 1
22. v = vmg/ktanhygk/mt = ymg/k e .r:-:- - ,
e2 'V kill m' +1
m m e.J,Iklm' + e-.J,Iklm'
y = k log cosh v' gk/ m t = k log 2 '
T.V. = Vmg/k ft/sec.
For definitions of cosh and tanh, see (18.91) and (18.92).
a v2
23• () = kmg (1 - e-2kJ/1 m) + Vo 2e-2k,,1 m• S·Ince the POSI"t·lve d·Irec t"Ion 18
•

downward, the positive square root must be taken for v when the body
is falling.

(b) v'k II + v'mii = v'k 110 + v'mii i.J gkl mi.

v'k V - v'mii v'k Vo - v'mii
(c) T.V. = Vmg/k ft/sec.
a v2
24• () = kmg (e- 2k1ll m - 1) + Vo 2e-2kJ/lm• S·Ince the POSI"t·IVe d·lrect·Ion 18
·

upward, the positive square root must be taken for v when the body is
rising.
(b) t = y--m-/k-g (Arc tan yk/mg Vo - Arc tan yk/mg v),
or v = Vgm/k tan (c - ykg/m t), where c = Arc tan yk/gm Vo.
The formula for the time t is valid only for a rising body.
Lesson 16A-Answers 159

m I [cos (c - Vkg/rn
(c) y=-og t)J .
k cos c
(d) t = cVm7kY.
m
(e) y = k log sec c.

25. v = Y mg/(kv02 + mg) Vo ft/sec,

t = Y m/gk cosh -1 sec c or cosh V gk/m t = sec c,
where c = Arc tan Vk/gm vo. For definition of cosh, see (18.91).
11 ge- +
26. (a) v = 16 11 _ ge- 4t ,y = 16t
4t
+ 8 log (5.5-- 4 ' ).
4.5e
(b) 16.49 ft/sec, 16.009 ft/sec. (c) 16 ft/sec.
(d) 29.5 ft., 16.1 ft. (e) 66 sec.
27. (a) k = 1/25.
e0.64t - 1
(b) " = 100 eO•641 + 1·
(c) 74.4 ft/sec. (d) 2.2 sec.
e6.4t - 1
28. (a) v = 10 tanh (3.2t) = 10 ,
e6 .
4t
+1
100
y == - log cosh 3.2t = -
100
log
e3 •2 , +2 e-3 .2 '
g g
(b) v = HWI - e- 0 .64y•
e2,..[10/15 - 1
29. (a) v = 151.8..[10 .
e 2 , 10/15 + 1
(b) T.V. == 151.8 ft/sec. (c) 147.4 ft/sec.
16t/3
(d) v = 12 1.18e + 1.
1.18e16t/3 - 1
(e) 12 ft/sec. (f) 12 ft/sec.
30. 600 sq ft.
31. 2.10 mi/sec, 4.92 mi/sec.
32. 4.7 mi/sec.
33. (a) 1.45 mi/sec. (b) 2.10 mi/sec.
34. 6.9 mi/sec, 119 hours.
35. (a) v = Ry2g/r, (b) 50.6 hours.
36. t = - !yaR+ bR2 -
b
a
2bv'=b 2
['I' - Arcsin -2bRa - aJ.

t = gR 2
(2gR - V0 2)3/2
[VOY2gR - v0 + Arc cos (gR - v
gR gR
2
)J
02

2
gR [VOV2 gR - v0 2 . vo]
- (2gR - 1102 )3/2 gR 2 Arc sm V2gR · +
160 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

38. (a) v2 = vo 2 + gRm (Rm -

a rm
1), where Rm is the radius of the moon.

(b) Vo = v'''''-gR-m-/-a = 1.48 mi/sec.

2 2
b) v dv = -gR + gmRm ,
39. ( dr r2 (D - r)2
2 2 2
v = 2gR-
2 +
2gmRm +
Vo 2 - 2g R - 2gmRm
r D-r D-R
(c) vo = y'2gR (0.99) = 99 percent of the escape velocity of the earth
when the gravitational pull of the moon is ignored.
(d) The velocity must have a positive value when it reaches the neutral
point. See (c).
40. (a) 4.9 mi/sec. (b) 42.5 min.

LESSON 16B. Horizontal Motion. If a body moves in a horizontal

direction, as on a table or platform, a frictional force develops, which
operates to stop the progress of the motion. This frictional force is due to
1. The gravitational force of the earth pressing the body to the platform.
2. The smoothness or roughness of the surface of the platform. This
quality of the surface, i.e., its roughness or smoothness, is characterized
by means of a letter J.L, called the coefficient of friction of the surface.

Definition 16.4. The frictional force of a body moving on a hori-

zontal surface is, by definition, equal to the product of the coefficient of
friction J.L of the body and the gravitational force mg, i.e., frictional
force = J.L(mg).
Comment 16.41. From experience we know that it requires a greater
force to begin the movement of an object than it does to keep it moving.
There are thus two coefficients of friction, one called static friction,
which operates at the start of the motion, the other called sliding friction,
which operates after the motion has pegun.

In addition to the frictional force, a body also may be subject to a

resisting force due to the air or other medium in which it moves.

Example 16.411. An object on a sled is pulled by a force of 10 pounds

across a frozen pond. Object and sled weigh 64 pounds. The coefficients
of static and sliding friction are negligible. However, the force of the air
resistance is twice the velocity of the sled. If the sled starts from rest,
find its velocity at the end of 5 seconds and the distance it has traveled in
that time. What is its terminal velocity?

Solution. Here m = 64/32 = 2 and the force of the air resistance is

given as 2v pounds. Hence the differential equation of motion of the sled
Lesson 16B HORIZONTAL MOTION 161

is (remember mass X acceleration of a body = net force acting upon it)

dv dv
(a) 2- = 10 - 2v
elt ' elt +v= 5.
Its solution is
(b)

The initial condition is t = 0, v = O. Hence Cl = -5, and (b) becomes

(c) v = 5(1 - e- t ).
When t = 5,
(d) v = 5(1 - e- 5 ) = 5(1 - 0.0067) = 5(0.9933) = 4.97 ft/sec.
By (c) the terminal velocity is found to be 5 ft/sec.
To find the distance x traveled in time t, we integrate (c) to obtain
(e)

When t = 0 and x = 0 (taking the origin at the starting position), we

find, from (e), C2 = -5. Hence (e) becomes
(f) x = 5 (t + e-t - 1).
And when t = 5,
(g) x = 5(5 + e- 5 - 1) = 5(4 + 0.0067) = 20 ft approx.

Example 16.42. A boy weighing 75 pounds runs for a slide and reaches
it at a velocity of 10 ft/sec. If the coefficient of sliding friction JJ between
his shoes and the ice is 1/25, how far will he slide? Ignore wind resistance.
Solution. By Definition 16.4, the frictional force is 1/25· 75 = 3
pounds. Since this is the only force which is opposing the motion and the
mass of the boy is 75/32, the differential equation of motion becomes
75 dv elv 32
(a) 32 elt = -3, - -.
elt = 25
By (16.11), we can write (a) as
elv 32 32
(b) v- = - - , velv= --elx
elx 25 25 '
where x is the distance measured from the beginning of the slide. Integra-
tion of (b) and insertion of the initial and final conditions results in

(c)
0 32
velv = - -
12: elx.
/ 11== 10 25 %==0
162 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

Its solution is

(d) 50 = Hx, x = 39 ft approx.

Example 16.43. Solve the previous problem if a wind is blowing

against the boy with a force to his velocity.

Solution. The differential equation (a) above must now be modified

to read [with dv/dt replaced by its equal v(dv/dx), - see (16.11)],

75 dv v dv 32
(a) -v-
32 dx
= -3 - v
' v + 3 = - 75 dx,

1 0
v=lO
(
1-
V
3
+3
) 32
dv = - -
75
1%
%=0
dx.

The solution of (a) is

(b) -3 log 3 - 10 + 3 log 13 = -Hx, x = 13.1 ft approx.

Example 16.44. A boat is being towed at the rate of 18 ft sec. At

the instant when the towing line is cast off, a man takes up the oars and
begins to row with a force of 20 pounds in the direction of the moving
boat. If man and boat together weigh 480 pounds, and the resistance is
equal to!v pounds, find the speed of the boat at the end of 30 seconds.

Solution. The net force acting on the boat at the instant t = 0 when
the towing line is cast off is the man's force of 20 pounds, less the resisting
force of !v pounds. The mass of man and boat is 480/32 = 15. Hence
the differential equation of motion is

dv 7 4
(a) -+-v=_·
dt 60 3

Its solution by Lesson lIB is

(b) v = ce- 7t / 60 + ¥.
Substituting in (b) the initial conditions t = 0, v = 18, we find c = 46/7.
Hence (b) becomes
(c) v = ¥e- 7t / 60 + ¥.
When t = 30,
(d) v = ¥e- 7 / 2 + ¥ = 11.6 ft/sec.
Lesson 16B-Exercise 163

EXERCISE 16B
1. A boy pulls a sled, on which a parcel has been placed, with a constant force
of F lb. The sled and parcel together weigh mg lb. The frictional force of
the ice on the runners is negligible. However, the force of the air resistance
is k times the velocity of the sled. If the sled starts from rest, find:
(a) Its velocity as a function of time.
(b) Its distance as.a function of time.
(c) Its distance as a function of velocity.
(d) Its terminal velocity.
2. In problem 1, assume sled and parcel weigh 96 lb, that air resistance is t
times the velocity and that the boy is moving the sled at a constant rate of
4.5 ft/sec) (Le., the terminal velocity of the sled is 4.5 ft/sec).
(a) Find the constant force which the boy is applying to the sled.
(b) Find th'e velocity and distance equations as functions of time.
(c) How far has the sled moved in 5 min?
Solve independently and then check your answers with formulas found in
problem 1.
3. A boy weighing mg lb runs for a slide and reaches it with a velocity of Vo ft/sec.
The coefficient of sliding friction between his shoes and the ice is T. (a) Find
his velocity as a function of distance. Ignore air resistance. (b) How far
will he slide?
4. A boy weighing 80 lb runs for a slide and reaches it with a velocity of 12 ft/sec.
The coefficient of sliding friction between his shoes and the ice is 1/20 and
the wind blows against him with a force equal to twice his velocity. (a) Find
his distance as a function of his velocity. (b) How far will he slide?
5. A 32,000-ton ship starting from rest begins to move because of the actions
of its propellers that exert a forward force of 120,000 lb. The force of the
water resistance is 5000v. Find the velocity of the ship as a function of time
and its terminal velocity.
6. The brakes are applied to a car, traveling on a slippery road, when it has
slowed down to a speed of 6 mi/hr = 8.8 ft/sec. It slides 80 ft before coming
to a stop. Compute the coefficient of sliding friction between tires and street.
Neglect the force of air resistance.
7. A body weighing 50lb rests on a table whose coefficient of sliding friction
is 1/25. The body is attached by a string to a weight of 14 lb that hangs
vertically over the table. At the moment the system is released, the 50-1b
body is 15 ft from the edge of the table. Assume no other forces are operating.
When and with what velocity does the body leave the table? Hint. The total
mass of the system is 64/32.
8. The forward thrust of an airplane due to its propellers is F lb. The air re-
sistance is kv 2 • Find the terminal velocity of the plane.
9. An 8-lb body starting from rest is being pulled along a surface, whose co-
efficient of sliding friction is 1, by a force that is equal to twice the distance
of the body from its starting point x = o. Air resistance is v2 /8. Find the
velocity of the body as a function of its distance. Hint. The resulting differ-
ential equation is linear in v2 •
164 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

10. A man and his boat weigh 320 lb. The man exerts a force of 16 lb on the
oars. The resistance of the water is twice the speed. Find:
(a) The velocity of the boat as a function of time.
(b) Its speed after 5 sec.
(c) Its terminal velocity.
II. A man and his boat weigh 400 lb. At the moment the man picks up his oars
to row, the boat is moving at the rate of 22 ft/sec. If the resistance of the
water is 2v lb and the oars exert a constant force of 15 lb, find the velocity
of the boat as a function of time; also find the terminal velocity of the boat.

ANSWERS 16D
F -k'lm
1. (a) " =- k (1 - e ).
F(
(b) z == k t +
me-k'lm
k -
m)
k .
(c) x (-v _
= m
k
Flog F -
k F
kv) .
(d) T.V. = F/k ft/sec.
2. (a) lIb.
(b) v = \(1 - e-
tI27
), x = i(t +
27e-
t127
- 27).
(c) 1228.5 ft.
3. (a) v2 = v0 2 - 2rgx. (b) x = v0 2/2rg.
2+v
4. (a) v - 2 log 14 = 12 - tx. (b) 10.1.
5. v = 24(1 - e-tI400), T.V. = 24 ft/sec.
6. 0.015. 8. T.V. = v'F7k ft/sec.
7. t = 2.2 sec, v = 13.4 ft/sec.
2t
9. ,,2 = 16(z - 2 +2e-j.
10. 0
(a) v = 8(1 - e- . ). (b) v = 5.1 ft/sec. (c) T.V. = 8 ft/sec.
II. v = 1(15 +
2ge-4tI25); T.V. = 15/2 ft/sec.

LESSON 16C. Inclined Motion. Quantities that have both magni-

tude and direction, such as force, velocity, acceleration, are called vector
quantities. It is the usual custom to
y
represent a vector quantity by a line with
an arrowhead at one end. The magnitude
P of the vector is given by the length of the
/ .....- - - - -.... R
./'
./ line and its direction by the inclination
/'" of the line. In Fig. 16.5, PQ represents a
8
vector quantity. With PQ as hypotenuse,
o x
we construct a right triangle whose sides
Figure 16.5 are parallel to the x and y axes. Then
the vectors PR and RQ are called respec-
tively the x component and the y component of the vector PQ. Their
respective magnitudes are given by
Lesson 16C INCLINED MOTION 165

(16.51) IPRI = IPQ cos 81, IRQI = IPQ sin 81,

and their respective directions by the directions of the arrowheads.
Comment 16.511. A vector formerly was written with an arrow over
it to distinguish it from a line segment. The current practice, and the one
we shall adopt, is to use bold face type.
A vector may also be broken up into two or more components in any
directions. In Fig. 16.52, we have broken up the vector PQ into four
component vectors PR, RS, ST, TQ. Note that each
vector begins where the other leaves off and that the
final vector's arrowhead touches the original vector's T
arrowhead.
If a body moves along a line which is inclined to
the horizontal, the effective force causing it to move P
downhill is that component of the gravitational force
acting in a direction parallel to the motion. The forces R
opposing the motion are the frictional force and the Figure 16.52
wind resistance. Here the frictional force is equal to
the product of the coefficient of friction J.L and the component of the
gravitational force acting in a direction perpendicular to the motion.
Example 16.53. A toboggan with two people on it weighs 520 pounds.
It moves down a slope whose gradient is 5/12. If the coefficient of sliding
friction is 1/50, and the force of the wind resistance is 5 times the velocity,
find the time it will take the toboggan to reach the bottom of a 650-ft
long incline. What would the terminal velocity be if the slope were of
infinite length?
Solution. Let a be the angle of incline of the slide. Since its gra-
dient = 5/12, tan a = 5/12. Hence (Fig. 16.54a) sin a = 5/13 and cos
a = 12/13. The gravitational force, which is the combined weight of sled

(0,0)

12
(a) (b)
Figure 16.54

and two people, equals 520 pounds. Therefore the magnitude of the com-
ponent of the gravitational force in the direction of the incline (Fig.
16.54b) is 520 sin a = 520(5/13) = 200 pounds. The magnitude of the
166 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

component of the gravitational force perpendicular to the incline is

520 cos a = 520(12/13) = 480 pounds. Since the coefficient of sliding
friction is 1/50, the sliding frictional force is 480(1/50) = 9.6 pounds.
The mass of the body is 520/32. Hence the differential equation of motion
is (remember mass X acceleration = net force acting on the body)
520 dv dv 4
(a) 32 dt = 200 - 9.6 - 5v = 190.4 - 5v, dt + 13 v = 11.7.
Its solution by the method of Lesson lIB is
(b)
The initial conditions are t = 0, v = O. Hence Cl = -38, and (b) becomes
(c) v = 38(1 - e-4t/13).

Integration of (c) gives

(d) 8 = 38(t + lj-e- 4t / 13 ) + C2.
If we take the origin at the starting point of the toboggan, then t = 0,
8 = O. Therefore by (d), C2 = -123.5. Hence (d) becomes

(e) 8 = 38(t + 1fe-4t/13) - 123.5.

When 8 = 650, we obtain from (e)
(f)
whose solution is t = 20.4 seconds approximately. This is the time it
will take the toboggan to reach the bottom of the incline. -
From (b), the terminal velocity is 38 ft/sec (let t (0).

EXERCISE 16C

In the problems below, take the origin at the start of the motion and
the positive direction in the initial direction of motion.
1. A toboggan with four boys on it weighs 400 lb. It slides down a slope with
a 300 incline. Find the position and velocity of the toboggan as functions of
time if it starts from rest and the coefficient of sliding friction is -k. If the
slope is 123.2 ft long, what is the velocity of the toboggan when it reaches
the bottom? Neglect air resistance.
2. A body weighing 400 lb is shot up a 300 incline with an initial velocity of
50 ft/sec. The coefficient of sliding friction is -to.
(a) Find the position and
velocity of the body as functions of time. (b) How long and how far will
the body move before coming to rest? Neglect air resistance.
3. Two particles start from rest and from the same point on the circumference
of a circle in a vertical plane. One moves along a vertical diameter; the other
along a chord of the circle (any chord will do). If the only force acting on the
particles is that of gravity, prove that both particles reach the circumference
of the circle at the same time.
Le8son 16C-Exercise 167

4. A body weighing W lb slides down a slope with an a O incline. Its initial

velocity is Vo ft/sec. The coefficient of sliding friction is T. Ignore air re-
sistance.
(a) Express the velocity and the distance of the body as functions of time.
(b) Express its velocity as a function of distance.
(c) Express the time as a funetion of the distance.
Use these formulas to verify the accuracy of your answers to 1.
S. A body weighing W lb is shot up a slope with an a O incline. Its initial velocity
is Vo ft/sec. The coefficient of sliding friction is T. Ignore air resistance.
(a) Find the position and velocity of the body as functions of time.
(b) How long and how far will the body move before coming to rest?
Use these formulas to verify the accuracy of your answers to 2.
6. A body weighing W lb, starting from rest, slides down a slope with an a O
incline. The coefficient of sliding friction is T and the force of the air re-
sistance is ltv lb.
(a) Express the velocity and the distance of the body as functions of time.
(b) Express its distance as a function of the velocity.
(c) What is its terminal velocity?
7. A boy and his sled weigh 96 lb. Starting from rest, he slides down an incline
whose gradient is i. The coefficient of sliding friction is fi, and the force of
the air resistance is twice the velocity.
(a) Find the velocity and the distance of the sled as functions of time.
(b) What is his terminal velocity?
(c) When will he reach the bottom of the slope if it is 272 ft long?
(d) What is his velocity at the bottom?
Solve independently . Use the formulas found in problem 6 to verify the
accuracy of your answers.
8. A body weighing W lb is shot up an a O incline with a velocity of Vo ft/sec.
The coefficient of sliding friction is T and the force of the air resistance is
kv lb.
(a) Express the velocity and the distance of the body as functions of time.
(b) How long and how far will it go?
9. A body weighing 64 lb is shot up an incline, whose gradient is t, with a
velocity of 96 ft/sec. The coefficient of sliding friction is ! and the force of
the air resistance is tov lb.
(a) Find the velocity and the distance of the body as functions of time.
(b) How long and how far will it go?
Solve independently. Use the formulas found in problem 8 to verify the
accuracy of your answers.
10. A toboggan with two people on it weighs 300 lb. It starts from rest down a
slope, i mile long, from a height 200 ft above a horizontal level. The co-
efficient of sliding friction is Tb and the force of the wind resistance is
proportional to the square of the velocity. When the velocity is 30 ft/sec,
this force is 6 lb.
(a) Find the velocity of the toboggan as a function of the distance and of
the time.
(b) With what velocity will the toboggan reach the bottom of the slide?
(c) When will it reach the bottom?
(d) What would its terminal velocity be if the slide were infinite in length?
168 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

ANSWERS 16C
1. v = 15.4t,8 = 7.7t2, V = 61.7 ft/sec.
2. (a) v = 50 - 18.77t, 8 = 50t - 9.39t 2 •
(b) t = 2.7 sec, 8 = 66.6 ft.
4. (a) v = vo + g(sin a - cos a)t; = vot +
T 8 (sin: a - r cos a)t 2 •
(b) v 2 = v0 2 + 2g(sin a - r cos a)8.
vv02 + 2g8(sin a - r cos a) - Vo
( c) t = .
g(sin a - r cos a)
5. (a) v = Vo - g(sin a + r cos a)t, t$ = vot - (sin a + r cos a)t2 •
(b) t = vo/g(sin a + r cos a) sec, 8 = v0 2 /2g(sin a + r cos a).
W (sln
6. (a) v = k . a - r cos a )(1 - e-ktlm) ,

8 = -W (sln
. a - r cos a ) ( t + -me -ktlm - - m) .
k k k

(b) 8 = (-v - : log A lev) , where A = mg(sin a - r cos a).

(c) T.V. = (sin a - T cos a).

7. (a) v = 27.2(1 - e- 2t/ 3 ), 8 = 27.2(t + ie- 2t/3 - i).
(b) T.V. = 27.2 ft/sec. (c) 11.5 sec. (d) 27.2 ft/sec.

8. (a ) v =
W(sin a +k r cos a) ( -ktlm
e -
1) +
voe ,
-ktlm

8= W(sin a t T cos a) (1 _ e- kt1m ) _ t] + ';0 (1 _ e- ktlm ).

(b) t = m log (1
k + W('sln a r cos a ») sec,
m g(sina + rcosa) 10 (1 + .
2
8 = mvo _ kvo ) ft.
k k 2 g W (sln a +
r cos a)
9. In the formulas: W = 64, r = 1, k = to, sin a = I, cos a = t,
Vo = 96, m = 64/32 = 2.
eO.l05t - 1 2 -0.00148
10. (a) v = 74.1 eO.IOSI + 1 'v = 5484(1 - e ).

(b) 68 ft/sec. (c) 30 sec, approx. (d) 74.1 ft/sec.

LESSON 17. Pursuit Curves. Relative Pursuit Curves.

LESSON 17A. The path traced by a body which

Pursuit Curves.
always moves in the direction of a fixed point or of another moving object
is called a pursuit curve.
Example 17.1. A pilot always keeps the nose of his plane pointed
toward a city T due west of his starting point. If his speed is v miles per
Lesson 17A PURSUIT CURVES 169

hour and a wind is blowing from the south at the rate of w miles per hour,
find the equation of the plane's path. Assume that it starts from a flying
field which is at a distance a miles from T.

Solution (Fig. 17.11). Let P(x,y) be the position of the plane at any
time t. The vector representing its speed has magnitude v and is pointed
toward T. Call 8 the angle this vector makes with the horizontal line con-
necting the flying field and T. The wind vector points due north and has a
magnitude w. The diagonal of the parallelogram formed by the vectors v

y
Actual direction
of plane

T(O,O) (a,O) X
t4----- x

Figure 17.11

and w then represents the actual direction and magnitude of the plane's
velocity at time t. Since this direction is changing instantaneously, it must
be tangent to the path of the plane at P. It must therefore be the slope
dy/dx of the desired curve. Our problem then is to find an equation which
expresses dy / dx as a function of x and y.
The respective components of the airplane's velocity in the x and y
directions are
dx dy _ -v sin 8.
(a) dt - -v cos 8, dt

Hence the effective velocity of the plane in the y direction, taking the
wind's velocity into account, is

dy
(b) -v sin 8 + w.
dt

From Fig. 17.11, we see that

x
(c) sin 8 = -;::::::::::::y=== cos 8 = -;::=====
VX2 + y2 VX2' + y2
170 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

Substituting these values in the first equation of (a) and in (b) we obtain

(d)
dx -vx dy vy +
dt = VX2 y2+ dt = - V X2 y2 + w.

Division of the second equation in (d) by the first gives

dy -vy + WV x 2 + y2
y _ W
V
v'X2 + y2
(e) dx = -vx = ----x---

For convenience, we let

w
(f) k =-,
v
so that k represents the ratio of the speeds of wind and plane. Hence (e)
now becomes
(g) x dy = (y - kVx 2 + y2) dx.
This equation is of the homogeneous type discussed in Lesson 7. It can
therefore be solved by the method outlined there. Easier perhaps is to
make use of the integrable combination (10.31). The necessary steps are
outlined below without further comment. (Initial conditions are t = 0,
x = a, y = 0, y' = y/x = 0.)

(h) x dy y dx = _ : VI + (y/x) 2 dx, d(y/x) _ _ k dx

VI + (Y/X)2 x'

(i)
1 "1'"
I//X=O
d(y/x)
vI + (Y/X)2
= _1'" X==(I
k dx
X '

(j) log + VI + (Y/X)2) = -k log:'

Let
y
(k) u =-.
x
Then (j) becomes
(1) log (u + vI + u 2 ) = -k log x ,
a

(m) u + VI + u 2 = (:)-le,
(n) I + u2 =:( )-2k - ( )-k 2 u + u 2,
(0) ( )-k =:( )-2k -
2: u 1,
Lesson 17A PURSUIT CURVES 171

(p)

Replacing u by its value in (k), we obtain

Replacing k by its value in (f), we have finally

(r) y = ; [(:Y-<w/v> _ (:y+<w/vl

as the equation of the path.
Comment 17.111. Equation (r) enables us to obtain interesting con-
clusions in regard to the path of the plane.
Case 1. Wind speed w = plane speed v. In this case (r) becomes

(s) Y = a(
2 1 -
X2) '
a2

which is the equation of a parabola, Fig. 17.12. Note that the plane will
never reach its destination T.
y
w
-v =2

w
- v =1
.,--
___ _ _...a (a,O)
J'
I I

T(O, 0) (a,O) X
Figure 17.12 Figure 17.13

Case 2. Wind speed w > plane speed v. In this case w/v > 1 so
that 1 - (w/v) < O. Hence as x _ 0, l-(w/v> [= <W/V>-l] _ 00.

We see from (r), therefore, that as x 0, y 00. Again the plane will
never reach T. A rough graph of its path is shown in Fig. 17.13 with
w = 2v.
172 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

Case 3. Wind speed W < plane speed v. In this case w/v < 1, so
that 1 - (w/v) > o. We see from (r), therefore that when x = 0,
y = o. Hence the plane will reach the town T. A rough graph of its path
is shown in Fig. 17.14, with v = 2w.
y w 1
-v ="2

T(O, 0) (a,O) X

Figure 17.14

Example 17.2. Solve the problem of Example 17.1 by using polar

coordinates.

Solution (Fig. 17.21). The components of the wind velocity w in the

radial and transverse directions are respectively [for the derivation of
these formulas, see Lesson 34, (34.2)].

dr . 8 d8
(a) dt=wsln, r dt = W cos 8.

y
Actual velocity
of plane

T(O,O) (a,O) X

Figure 17.21

Hence the effective velocity of the plane in the radial direction, taking
into account the wind's speed wand the plane's speed v, is
dr
(b) dt = -v + W sin 8.
Lesson 17A PURSUIT CURVES 173

Dividing (b) by the second equation in (a), we obtain

dr -v + w sin 8 v
(c) r d8 = w cos 8 - - w sec 8 + tan 8.
Let
(d) k= - -.
v
w
Then (c) becomes
dr
(e)
r
(k sec 8 + tan 8) d8.
Its solution, by integration, is

(f) log r + log c = k log (sec 8 + tan 8) - log cos 8,

which can be written as

(g) cr = (sec 8 + tan 8)k sec 8.

At t = 0, r = a, 8 = o. Substituting these values in (g), we have

(h) ca = 1, c = l/a.

The equation of the path is, therefore, after replacing k by its value in (d),

(i) r = a(sec 8 + tan 8) -'V/W sec 8,

or
(j) r cos 8 = a(sec 8 + tan 8)-'V/w.
Example 17.3. Solve the problem of Example 17.1 if the wind is
blowing with a velocity w in a direction which makes an angle a with the
vertical, Fig. 17.31.
y
Direction of
the wind

w cos a

T(O,O) w sin a (a,O) X

Figure 17.31

Solution. We shall give two methods by which this problem may be

solved.
174 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

Method 1. Choose the axes so that the direction of the wind becomes the
yaxis (Fig. 17.32). The initial conditions at t = 0, therefore, become

(a) x = a cos a, y=aSlna.

.
Direction
of wind
y

Position of plane
at t==O
Po(a cos a, a sin a)

a sin (¥

a cos a
T(O,O) x
Figure 17.32

The differential equations of motion, however, are exactly the same as

those in Example 17.1, namely

(b) -dx = -v cos 8 dy

dt = . +
-v sm 8 w.
dt '

Proceeding just as we did in Example 17.1, we obtain,

(c) log (u + vI + u 2 ) = -k log x + logc, k = wlv.

But now at t = 0, x = a cos a, U = y/x = (a sin a)/(a cos a) = tan a.
Substituting these values in (c), and solving for c, we obtain

(d) c = (tan a + sec a) (a cos a)".

Hence (c) becomes, after replacing u by its value y/x,

(e) log + y2) + log x" = log [(tan a + sec a)(a cos a)"J.
Simplification of (e) gives

(f) x"-l(y + VX2 + y2) = (tan a + sec a) (a cos a)", k -= wlv.

Replacing k by its value w/v, we obtain finally as the equation of the
plane's path
(g) x(tIJ/fI)-l(y + v'X2 + y2) = (tan a + sec a) (a cos a)tlJlfI.
Lesson 17A-Exercise 175

We leave it to you as an exercise to draw rough graphs of its path as

we did in Comment 17.11. Remember tan a, sec a, cos a and a are con-
stants.
Direction of the wind
y

x
T(O,O) (a,O) x
Figure 17.33

Method 2 (Fig. 17.33). From Fig. 17.33, we see that the resultant of
the effective velocities of wind and plane in the x and y directions are
respectively
dx
(h) dt = -v cos 8 + w sin a, : = -v sin 8 + W cos a.

Dividing the second equation in (h) by the first we obtain

dy -v sin 8 + w cos a
(i)
dx - -v cos 8 + w sin a
In (i) replace sin 8 by its value y/vx 2 + y2 and cos (J by X/VX2 + y2.
It then simplifies to

(j) -vy
( Vx 2 + y2
+ w cos a) dx = (- vx
Vx + y2
2
+ W sin a) dy,
which is a homogeneous equation. Remember v, w cos a, and w sin a are
constants. Hence it can be solved by the method of Lesson 7. The initial
conditions are x = a, y = o. We leave it to you as an exercise to com-
plete.
EXERCISE 17A
I. A man swims across a river 100 ft wide, always heading for a tree directly
across from his starting point. He can swim at the rate of 3 ft/sec.
(a) Find the equation of his path if the current is carrying him downstream
at the rate of 1 ft/sec; 3 ft/sec; 4 ft/sec.
(b) Draw the graph of each equation.
176 PROBLEMs LEADING TO FIRST ORDER EQUATIONS Chapter 3

Solve independently and check your answers with (r) of Example 17.1.
2. In problem 1 show that the man will reach the tree on the opposite bank if
the current is 1 ft/sec, that he will reach a point on the opposite shore
50 ft from the tree if the current is 3 ft/sec, that he will never reach the op-
posite shore, if the current is 4 ft/sec.
3. Solve problem 1 by using polar coordinates. Solve independently and check
your answers with (j) of Example 17.2.
4. An insect steps on the edge of a turntable of radius tJ that is rotating at a
constant angular velocitya. It moves straight toward the center of the table
at a constant velocity "0. Find the equation of its path in polar coordinates,
relative to axes fixed in space. (Hint. If B is the angle through which the
turntable has rotated in time t and r is the distance of the insect from the
center at that moment, then dB/dt == a, r(d8/dt) == ra.) Draw a graph
of the equation if tJ == 10 It, a == 1r/4 radians/sec, "0 == 1 It/sec. How
many revolutions will the table have made by the time the insect reaches the
center?
5. Assume in problem 4 that the insect always moves in a direction parallel to
the diameter drawn through the point where he steps on the table.
(a) Find the equation of its path relative to axes fixed in space.
(b) What kind of curve is it? Hint. Change the equation to rectangular
coordinates if the polar form is unfamiliar to you.
6. A boy stands at A, Fig. 17.34. By means of a string l ft long, he holds a
boat, which is in the water at B. He starts walking in a direction per-
pendicular to AB, always keeping the string taut. Find the equation of the
boat's path. Ignore the height of the boy above the horizontal and assume
that the string is always tangent to the path. The resulting curve is known
as a tractrix. Hint. The slope of the tractrix is: tan 8 == dy / th.
y B(O,l)

A(O, 0) x
Figure 17.34

7. A pilot always keeps the nose of his plane pointed toward a city which is 400
mi due west of him. A wind is blowing from the south at the rate of 20 mi/hr.
His speed is 300 mi/hr. Find the equation of his path. Solve independently,
by using both rectangular and polar coordinates. Check the accuracy of your
answers with (r) of Example 17.1 and (j) of Example 17.2.
8. A pilot always keeps the nose of his plane pointed toward a city which is tJ
mi due north of him. His speed is " mi/hr and a wind is blowing from the
west at w mi/hr. Find the equation of his path.
9. Solve problem 8 if the city is a mi due south of the pilot.
10. A pilot always keeps the nose of his plane pointed toward a city which is
300 mi due west of him. A 25-mi/hr wind is blowing in a direction whose
Lesson 17B RELATIVE PURSUIT CURVE 177

slope is t. His speed is 200 mi/hr. Find the equation of his path. Solve in-
dependently. Use both methods outlined in this lesson. Check the accuracy
of your differential equations with (b) and (j) of Example 17.3.
11. Solve problem 10 if the wind is blowing in a direction whose slope is -to
Use method 2 of Example 17.3. Solve independently and then check the
accuracy of your differential equation with (j) of this example.
12. Assume in problem 4 that the insect moves straight toward a light which is
fixed in space directly above the end of the diameter drawn through the
point where it steps on the table. Find the differential equation of its path
in polar coordinates.
ANSWERS 17A

1. 11 = 50 - x
2
= -200(y - 50);

11 - 50 [ -113 -
3. r = 100(1 - sin 8)/(1 +
sin 8)2; r = 100/(1 + sin 8);
r cos 8 = 100(sec 8 +
tan 8) -3/4.
4. r = a - Vo 8 (origin at center). Ii revolutions.
a
5. (a) 2vo(r sin 8) = a(a 2 - r2). (b) Circle: center (O,-vo/a),
radius v' a 2 + vo 2 /a 2 •
6. x = llog l + v'12 - y2 _ v'l2 _ y2.
Y

7. 11 = 200 - ; cos 8= 4OO(sec 8+ tan 8)-15,

T

8. x = i - 1+<..1.)] with positive directions to the east and

to the south.
9. Same as 8 with positive directions to the east and to the north.
10. First method: solution is y v'x 2+ +
y2 = 2(240) 118x 7/8.
Second method: Differential equation is
(200y - 20v'x 2 +
y2) dx - (200x - 15v'x 2 y2) dy = + o.
II. Differential equation is
(200y - 20v''--x2-+-y-2) dx - (200x + I5v'x 2 + y2) dy = o.
12. dr = _ vor cos .(8 -1/1) ,where 1/1 is the angle made by the original
d8 ra +
Vo sIn (8 - 1/1)
diameter and a line drawn from its end (i.e., where the light is) to the insect's
position.

LESSON 17B. Relative Pursuit Curve.If the origin of a coordi-

nate system is not fixed on the ground but is attached to a pursued object,
then the path described by a pursuer, moving always in the direction of
the object he is pursuing, is called a relative pursuit curve. It is, in
178 PROBLEMS' LEADING TO FmST ORDER EQUATIONS Chapter 3
I

effect, the path traced by the pursuing body as seen by an observer in

the pursued object.
Example 17.4. A fighter plane whose speed is Vp is chasing a bomber
plane whose speed is VB. The nose of the fighter plane is always pointed
toward the bomber which is flying in a direction making an angle with
the horizontal. Find the path traced by the fighter as plotted by an
observer in the bomber.
Solution (Fig. 17.41). Call (0,0) the origin of a coordinate system
fixed on the ground, and (0,0) the origin of a coordinate system which is
moving with the bomber plane. The position of the fighter plane at any
y P(x, Y) =(x, y) = position of fighter at time t

l'F sin (11" - 8)

( VB cos fJ
Q(O,O)== (xB' YB) = position of
bomber at time t
(0,0) x

Figure 17.41

time t is therefore given by two sets of coordinates, one (x,y) with respect
to the fixed ground origin (0,0) and the other, (%,11) with respect to the
moving origin (0,0).
As measured from an observer on the ground, the effective velocities of
the fighter plane in the x and y directions are (see Fig. 17.41)
dx
(a) dt - V F cos (r - 8) z:::: - V F cos 8,

= -VFsin{r - 8) = -VFsin8.
If two planes are approaching each other along a straight line, then to
an observer in one plane, it will seem as if he were standing still and the
other plane were coming toward him with a speed equal to the sum of
the speeds of the two planes. If the two planes are moving away from
Lesson 17B RELATIVE PURSUIT CURVE 179

each other along a straight line, then it will seem to an observer in one
plane as if he were standing still and the other plane were going away
from him with a speed equal to the sum of the speeds of the two planes.
If, however, the planes are going in the same direction along a straight
line, then to an observer in a pursued plane it will seem, if his speed is
slower than the other, as if he were standing still and the other plane
were coming toward him with a speed equal to the difference of the speeds
of the two planes. If his speed is greater than the other, then it will seem
to him as if he were standing still and the other plane were going away
from him with a speed equal to the difference of the speeds of the two
planes.
Here the horizontal component of fighter and bomber velocities are in
the same direction. Hence (see Fig. 17.41), to an observer in the bomber
(we assume the fighter's component is greater than the bomber's), it is as
if he were standing still, and the fighter plane were coming toward him in
the positive x direction with a velocity equal to the difference of their
two x components of velocity, Le., the rate of change of x to an observer
in the bomber is
d%
(b) dt = -VF cos 8 - VB cos {3.

The vertical components of fighter and bomber velocities are in opposite

directions and pointed toward each other. Hence, the velocity of the
fighter relative to an observer in the bomber, is in the negative y direc-
tion and is the sum of their two vertical velocities, i.e., the rate of change
of 'f} is
(c) = -(VF sin 8 + VB sin (3).

Dividing (c) by (b), we obtain

(d) -d'f} = V F sin 8+ V B sin {3

,
d% VF cos 8 + V B cos {3
which has the same form as (i) of Example 17.3. Remember VF , VB sin {3,
and VB cos {3 are constants. Hence it can be solved by the method sug-
gested there. We leave it to you as an exercise to complete. The initial con-
ditions will depend on the distance and direction of the fighter plane from
the bomber plane at t = 0, i.e., at the moment when the pursuit began.
Example 17.5. Solve the problem of Example 17.4 by using polar
coordinates.
Solution (Fig. 17.51). To an observer in the bomber, when he moves
to the right it appears to him as if he were standing still and the fighter
plane were moving to the left. Hence, to an observer in the bomber,
180 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3
I

when he moves with a velocity VB, represented by the lower vector in

Fig. 17.51, it appears to him as if he were standing still and the fighter is
position of fighter at any time t

Direction of bomber's flight

o
Figure 17.51

moving with a velocity VB represented by the upper vector in this figure.

The radial and transverse components of the upper VB are
dr
(a) dt = -VB cos (8 - (3), r = VB sin (8 - (3).

As seen from an observer in the bomber, therefore, the effective velocity

of the fighter in the radial direction is (remember it is the sum of the two
velocities in the same negative r direction)
dr
(b) dt = -VF - VB cos (8 - (3),

and in the transverse direction is

d8
(c) r- = VB sin (8 - (3).
dt

Dividing (b) by (c), we obtain

dr cos (8 - (3) VF
(d) r d8 = - sin (8 - (3) - VB csc (8 - (3).

I ts solution is

(e) log r + log c

= -log sin (8 - (3) - log [csc (8 - (3) - cot (8 - (3)],

which simplifies to
Lesson 17B RELATIVE PURSUIT CURVE 181

(f) CT = [csc (8 - (j) - cot (8 - {j)]-VFIVB

sin (8 -
1
- sin (8 - (j)
[1 -cos (8 -
sin (8 -
1
[sin (8 - - cos (8 _
NOTE. The initial conditions are t = 0, T = To, 8 = 80 •

Comment 17.52. As viewed by an observer in the bomber, it is as if

he were standing still, and the path of the fighter as given in (f) (which is
the resultant of both planes' motions) were due entirely to the fighter
plane's movements.

Comment 17.521. By means of (f), we can draw a rough graph of

the fighter plane's path, as seen from the observer in the bomber.
Case 1. If VB = VF, i.e., if the ground speed of both planes are the
same, then (f) reduces to
1
(g) CT = ,
1 - cos (8 -
which is the equation of a parabola [see (34.68) and comment following].
A rough graph of the path of the fighter plane as viewed from the
bomber is given for this case in Fig. 17.53. It is evident that the fighter
will never reach the bomber.

Fighter path

Figure 17.53
182 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

CaseS. IfVF > VB,thenl - (VF/VB ) < 0, and (f) can be written as

[sin (8 -
(h) CT =
[1 - cos (8 -

A rough graph of the fighter plane as viewed from the bomber is given for
this case in Fig. 17.54 (with V F = 2VB).

Figure 17.54

EXERCISE 17B
1. An airplane A is flying with a speed of 200 mi/hr in a direction whose slope is
t. A plane B, 50 miles due north of him, starts in pursuit. Plane B, whose
speed is 300 mi/hr, always keeps the nose of his plane pointed toward A. Find
the equation of B's path as observed from A. Use rectangular and polar co-
ordinates. Hint. In polar coordinates at t = 0, 8 = '1'/2, tan = t, r = 50.
2. Solve problem 1, in polar coordinates only, if A is flying in a southeasterly
direction and plane B, at the instant he starts in pursuit, is 50 miles north-
east of A. Hint. At t = 0, 8 = '1'/4, = -'1'/4, r = 50. Make a figure
corresponding to 17.51. Be very careful of signs.
3. Solve problem 1, in polar coordinates only, if A is flying in a northeasterly
direction and plane B, at the instant he starts in pursuit, is 50 miles south-
east of A. Hint. At t = 0, 8 = -'1'/4, = '1'/4, r = 50. Make a figure
corresponding to Fig. 17.51. Be very careful of signs.
Lesson 17M A. FLOW OF WATER THROUGH AN ORIFICE 183

ANSWERS 17B"
1. Differential equation in rectangular coordinates is

dii 300y +12OV'x2 + y2 with t = 0, x = 0, y = 50.

d! = aOO! + 16oV!2 + y2
Solution in polar coordinate is:

5M (4 sin 8 - a cos 8)112

r =
(5 - 4 cos 8 - a sin 8) 3/2 ·

_
2• r -
5M (sin 8 + cos 8)112 • 3 _ 5M (cos 8 - sin 8)112 •
3/2. r - 3/2
(0 - cos 8+ sin 8) (0 - cos 8 - sin 8)

LESSON 17M.
MISCELLANEOUS TYPES OF PROBLEMS LEADING TO
EQUATIONS OF THE FIRST ORDER
A. Flow of Water Through an Orifice. When water flows from a
tank through a small hole in its bottom, it has been proved that the rate
of flow of water is proportional to the area of the hole and the square root
of the height of the water in the tank. Hence
dV
(17.6) -dt = -ka-1i,
v n, dV = -kav'h dt,
where
V is the number of cubic feet of water in the tank at time t sec,
k is a positive proportionality constant,
a is the area of the hole in square feet,
h is the height in feet of the water above the hole at time t.
The minus sign is necessary because the volume of water is decreasing.
If A is the cross-sectional area of the water surface at time t, then
dV = A dh. SUbstituting this value of dV in the second equation of
(17.6) and taking k = 4.8, a figure determined experimentally as valid
under certain conditions, we obtain
(17.61) A dh = -4.8av'h dt,
where
A is the cross-sectional area in square feet of the water surface
at time t sec,
h is the height in feet of the water level above the hole or orifice
at time t,
r is the radius in fE'et. of thp. orifice.
184 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

With the aid of (17.61), solve the following problems.

1. A tank, whose cross section measures 3 ft X 4 ft, is filled with water to a
height of 9 ft. It has a hole at the bottom of radius 1 in. (a) When will the
tank be empty? (b) When will the water be 5 ft high? Hint. At t = 0,
h = 9 and in (17.61) r = -h, A = 12.
2. A water container, whose circular cross section is 6 ft in diameter and
whose height is 8 ft, is filled with water. It has a hole at the bottom of radius
1 in. When will the tank be empty?
3. Solve problem 2 if the tank rests on supports above ground so that its 8 ft
height is now in a horizontal direction, Fig. 17.62, and the hole of radius 1 in.
is in its bottom.

A=16x

8 ,
I

6
h
,,
I

\
\
\
\
\ ,
"
Figure 17.62

4. A water tank, in the shape of a conical funnel, with its apex at the bottom and
vertical axis, is 12 ft across the top and 18 ft high. It has a hole at its apex
of radius 1 in. When will the tank be empty if initially it is filled with water?
5. A water tank, in the shape of a paraboloid of revolution, measures 6 ft in
diameter at the top and is 3 ft deep. A hole at the bottom is 1 in. in diameter.
When will the tank be empty if initially it is filled with water?
6. A cylindrical tank is 12 ft in diameter and 9 ft high. Water flows into the
tank at the rate of 11"/10 ft 3 / sec. It has a hole of radius 1 in. at the bottom.
When will the tank be full if initially it is empty? Hint. Adjust the first
equation in (17.6) to reflect the increase in volume due to the water intake.
Remember dV = A dh. After simplification, you should obtain

r 9

Jh=O
dh
12 - Vh -
- -1-f dt
t

4320 t=O •

If you have trouble integrating, see hint given in answer.

R B. First Order Linear Electric Circuit.

The differential equation which results when
an applied electromotive force E, an inductor
L L, and a resistor R are connected in series is,
:F'ig. 17.621,

Figure 17.621
(17.63) IJ + Ri = E.
Lesson 17M c. STEADY STATE FLOW OF HEAT 185

For an understanding of this equation and for the meaning of these tenns,
read Lesson 30. The units we shall adopt are
ohms for the coefficient of resistance of the resistor,
henries the coefficient of inductance of the inductor,
volts for The electromotive force (also written as emf),
amperes for the current.
7. Solve the differential equation (17.63), if E = Eo and at t = 0, the current
i = io. What is the limiting current as t -+ oo?
8. Solve (17.63) if E = Eo sin wt and at t = 0, i = io.
9. An inductance of 3 henries and a rebistance of 30 ohms are connected in series
with an emf of 150 volts. At t = 0, i = o. Find the current when t = 0.01
sec.
10. An inductance of 2 henries and a resistance of 20 ohms are connected in series
with an emf of 100 sin 150t volts. At t = 0, i = o. Find the current when
t = 0.01 sec.
II. When an emf is disconnected from a circuit in which a current is flowing, Le.,
at t = 0, E = 0, the current is called an induced current. The term
L di/dt is then referred to as an induced electromotive force. At the
moment an emf is disconnected from a circuit in which there is a resistance
of 30 ohms and an inductance of 6 henries, the current i = 15 amp. (a) Find
the current equation as a function of time. (b) When will the current be
7.5 amp?
C. Steady State Flow of Heat. When the inner and outer walls of
a body, as for example the inner and outer walls of a house or of a pipe, are
maintained at different constant temperatures, heat will flow from the
wanner wall to the colder one. When each surface parallel to a wall has
attained a constant temperature, we say that the flow of heat has reached
a steady state. In a steady state flow of heat, therefore, each surface
parallel to a wall, because its temperature is now constant, is called an
isothermal surface. Isothermal surfaces at different distances from an
interior wall will, of course, have different temperatures. In many cases
the temperature of an isothennal surface is a function only of its distance
x from an interior wall, and the rate of flow of heat in a unit time across
such a surface is proportional both to the area of the surface and to dT / dx,
where T is the. temperature of the isothermal surface. Hence
dT
(17.64) Q = -kA-,
dx
where
Q is the rate of flow of calories * of heat in 1 sec across an isothennal
surface,
k, the proportionality constant, is called the thermal conductivjty
of the material that is between the walls,
*A calorie is equal to the amount of heat required to change the temperature of 1
gram of water 1 degree centigrade.
186 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

A is the area in square centimeters of an isothermal surface,

T is the temperature in centigrade degrees of the isothermal surface,
x is the distance in centimeters of the isothermal surface from an
interior wall.

The negative sign is used to indicate that heat flows from the interior
wall of higher temperature to the exterior wall of lower temperature.
With the help of (17.64), solve the following problems.
12. The inner and outer radii of a hollow spherical shell are 4 cm and 9 cm respec-
tively. The thermal conductivity of the material between the walls is
0.75 cal/deg-cm-sec. The inner surface is kept at a constant temperature of
100°C and the outer surface at O°C. Find:
(a) The rate of heat loss per second flowing outward through the exterior
of the shell.
(b) The temperature T of a surface 5 cm from the center.
Hint. In (17.64) A = 411"r 2 ,
where r, which replaces x in the formula, is the
radius of an isothermal surface. Initial conditions are: r = 4, T = 100;
r = 9, T = o.
Note, from answer, that Q has a constant value. In this example, it is
216011" cal/sec. Hence, through each isothermal surface, this much heat
escapes each second. If a surface is nearer the center of the sphere, so that
its area is smaller than the area of a surface farther away, more heat per unit
area will escape each second from the nearer surface than from the farther
one. The total loss of heat through each surface is, however, the same.
13. A steam pipe of negligible thickness has an inside radius of 6 cm. It is in-
sulated with 3 cm coating of magnesia, whose thermal conductivity is
0.000175. 1"he interior of the pipe is maintained at 100°C and the outer
surfa ce at O°C. Find:
(a) The rate of heat loss per second flowing outward from each meter length
of pipe.
(b) The temperature of an isothermal surface whose radius is 8 cm.
II into In (I 7.64), .t1 = 211"r( 100), where r, which replaces x in the formula,
is the radius of an isothermal surface.

D. Pressure-Atmospheric and Oceanic. We consider a column

of air of cross-sectional area A, of height dh and at a distance h units from
the surface of the earth, I"'ig. 17.65. simplicity, we assume the earth
is a plane, the air is at rest, and is unaffected by changes of latitude,
longitude, and temperature. The positive direction is upwards. Let p be
the pressure* on this column of air, due to the weight of all the air above
it. Hence the total force P on a column of air of cross-sectional area A
is P == A p. By differentiation \ve obtain
(a) dP == A dp,

·Pressure is the force acting perpendicularly on a unit area of surface.

Lesson 17M D. ATMOSPHERIC AND OCEANIC PRESSURE 187

which gives the change of the total force P for a change in the pressure p.
Let p be the weight of a unit volume of air. Therefore the weight of a col-
umn of air of height dh and cross-sectional area A is
(b) (A dh)p.
The decrease in total force as one goes from the height h to the height
h + dh must be equal to the weight of
the column of air of thickness dh. Hence
equating the negative of (a) with (b), we A
obtain
dp
t+
(17.66) -A dp = pA dh,
dh
-p,
h
Surface of
which states -that the rate of change of
air pressure with height is equal to the the earth
negative of the weight of a unit volume of
air at that height. The units we shall use
are: pounds per cubic foot for p, pounds Figure 17.65
per square foot for p, and feet for height.
If p is oceanic pressure and h is the depth below sea level, then (17.66)
becomes
dp
(17.67) dh = p,

where p is the weight per cubic foot of ocean water.

With the aid of (17.66) and (17.67), solve the following problems.
14. Assume that the air is an isothermal gas. Therefore by Boyle's law, its
pressure and density are related by the formula
(17.671) p = kp.
(a) Determine the pressure of the air as a function of the height h above
the earth if at sea level the air pressure is 14.7Ib/sq in. Hint. In (17.66),
replace p by kp.
(b) What is the value of the constant k in (17.671) if p = 0.081 lb/cu ft at
sea level?
(c) What is the air pressure at 10,000 ft, at 15,000 ft, at 70,000 ft, at 50 mi?
(d) Show that the pressure is zero only when h is infinite.
(e) Express the density of air as a function of height. Hint. In (17.66) re-
place dp by dp/k as obtained from (17.671).
(f) What is the weight of the air at an altitude of 1000 ft, of 5000 ft, of
10,000 ft, of 50,000 ft?
15. If air expands adiabatically, i.e., without gaining or losing heat, then
p = kp1.4.
(a) Find the pressure of the air as a function of height. A.ssume that at sea
level p = 14.7 lb/sq in. and p = 0.081 lb/cu ft. Hint. First find the
value of k. Then in (17.66) replace p by (p/k)5/7.
188 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

(b) How high is the atmosphere, i.e., at what height is p = o? Hint. In

(17.66) replace dp by 1.4kpo.4 dp. Solve for p. Then find h when p = o.
16. Assume that the weight p of a cubic foot of sea water, under a pressure of
p Ib/ft 2 is given by the formula
(17.672) p = k(1 + 2 · 10-8 p) Ib/ft3 ,
and is 64 Ib/ft 3 at sea level.
(a) Find the value of k. Hint. At sea level p = o.
(b) Find the pressure of sea water as a function of its depth below sea level.
Hint. In (17.67), replace p by its value as given in (17.672).
(c) Find the weight per cubic foot of sea water as a function of depth.
Hint. In (17.67), replace dp by 108 dp/2k as determined from (17.672).
(d) What is the pressure and density of sea water at 20,000 ft below sea level?
17. If the earth is assumed spherical instead of planar, show that (17.66) be-
comes
dp 2p
(17.68) dh = -p - R h +
where R is the radius of the earth. Hint. See Fig. 17.69. The volume of a
thin spherical shell of air at a distance r units from the center of the earth
and thickness dr is 411"r2 dr. Its weight, therefore, is (411"r 2 dr)p. The total

Figure 17.69

force P on this spherical surface due to the weight of the air outside it is
P = 411"r2p. Hence dP = 411"(r2 dp + 2rp dr). Set -dP, which is the de-
crease in the total force P as one goes from the distance r to the distance
r+ dr, equal to the weight of the spherical shell.

E. Rope or Chain Around a Cylinder. When a rope or chain,

assumed uniform, flexible, and inextensible, is wound around a rough
cylindrical post, a small force applied at one end of the rope or chain can
control a much larger force applied at the other end. For example, a man
can hold in check a large weight by winding a rope attached to it, a suffi-
cient number of times about a pole.
For a post, whose axis is horizontal, it has been proved that the differ-
ential equation
dT
(17.7)
d8
arCCOS 8 +
p, sin 8) + p,T
Lesson 17M F. MOTION OF A COMPLEX SYSTEM 189

expresses the tension Tin the rope or chain at a; point P on it, when the rope
or chain is just on the verge of slipping, see Fig. 17.71. In equation (17.7),
T is the pull or tension on the rope at any point P on it, in pounds,
T is the radius of the cylinder, in feet,
J.L is the coefficient of friction between rope and post,
ais the weight of the rope or chain in pounds per foot,
8 is the radial angle of P.
With the help of (17.7), solve the following problems.
18. Show that the solution of (17.7) is

(17.72) T = 1 +ar p.2 [(1 - p. )

2 •
SIn 8 - 2p. cos 8]
+ ce.
pol

19. A chain weighs a lb/ft. It hangs over a circular cylinder with horizontal
axis and radius r ft. One end of the chain is at A, Fig. 17.71. How far must
the other end extend below D, so that the chain is on the verge of slipping.
Hint. The initial conditions are: 8 = 0, T = 0, and 8 = 1f', T = la,
where l is the length of the portion of the chain overhanging at D.
20. A chain weighs a lb/ft. It hangs over a circular cylinder with horizontal
axis and radius r ft. One end of the chain is at B, Fig. 17.71, the other end
just reaches to D. What is the least value
of p. so that the chain will not slip? Hint.
The initial conditions are: 8 = 11"/2, T = 0,
B
--..----
and if the chain is not to slip, then at 8 = 11",
T :::::: o.
21. A chain weighs lib/ft. It hangs over a
circular cylinder with horizontal axis and D
radius ! ft. One end of the chain reaches
three quarters around the top to C, Fig.
17.71; the other end extends below D. The l
coefficient of friction between chain and
cylinder is !. If the chain is on the verge of
slipping, find the length l of the overhang.
22. If the axis of the cylinder is vertical, then Figure 17.71
the rope's or chain's weight which now acts
vertically so that it does not press down upon the cylinder, has little effective
force. Hence, in formula (17.7), a which is the weight per unit length of
rope, may be taken to be zero. The formula then simplifies to
dT
(17.73) d8 = p.T.
With the help of (17.73), solve the following problem. A longshoreman is
holding a ship by means of a hawser wound around a vertical post. The
ship is pulling on the rope with a force of 5 tons. If the coefficient of friction
between rope and post is 1, and the man exerts a force of 50 lb to hold the
ship, approximately how many turns of rope is he using? Hint. Initial
condition is 8 = 0, T = 50. We want 8 when T = 10,000.
F. Motion of a Complex System. Solve problems 23-27 below, by
use of Newton's law of motion F == ma == mv(dv/dy) , where F is the
190 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

alge braic sum of all the forces acting on a body of mass n" and a is the
acceleration of the center of gravity of the body.
23. A 24-ft chain weighing 8 lb/ft hangs over a frictionless support, which is
more than 24 ft above the floor. Initially the chain is held at rest with 10 ft
overhanging on one side of the support and 14 ft on the other side, Fig. 17.74.

y = 12
•

14ft. f+
2{:}Y y = 0, equilibrium
position
} -2
-y

Figure 17.74

How long after its release and with what velocity will the chain leave the
support? Hint. When 12 ft of overhang on each side of the support,
the chain is in equilibrium. Call this position y = o. Then if the distance of
one end of the chain from equilibrium is y, the distance of the other end is -y.
The effective force moving the chain is therefore 2y8. The mass of the chain
is 248/32. Initial conditions are t = 0, v = 0, y = 2.
24. In problem 23, assume that the 14-ft overhang just touches the floor. Com-
pute the velocity with which the other end will leave the support.
25. A chain is 12 ft long. Six feet of the chain are held extended on a frictionless
flat table which is more than 12 ft above the ground, the other 6 ft hangs
over the table. When and with what velocity will the end of the chain leave
the table after its release? Hint. m = 128/32, F = y8, where y is the dis-
tance of the overhanging part of the chain from the edge of the table and 8
is its weight per foot. Initial conditions are t = 0, y = 6, v = o.
26. Assume, in problem 25, that the table is only 4 ft above the ground. With
what velocity will the chain of problem 25 leave the table?
27. It has been proved that when a mass particle slides without friction down a
fixed curved path whose equation is y = f(x), its differential equation of
motion, with upward direction positive, is
(17.75) v dv = -g dy,
where g is the acceleration due to gravity, v is the velocity of the particle along
the curve, and y is the vertical position of the particle at time t. Therefore,
v = ds/dt, where s is the distance the particle has moved along the curved
path. Show that. the solution of (17.75), with initial conditions t = 0,
v = vo, y = Yo, IS

(17.76) v2
= (dS)
dt
2
= VO 2 + 2g(yo - y).
Lesson 17M G. VARIABLE MASS. ROCKET MOTION. 191

By means of the substitution in (17.76) of ds/dt = VI +(dy/dx)2 dx/dt,

the equation of the path y = I(x) and the initial condition Yo = I(xo), show
that (17.76) becomes
2
(17.77)
dX)2 = vo +
2g[l(xo) - I(x)] .
( dt 1 +
[f'(X)]2
The solution of (17.77) will give x as a function of t. With x known, we can
determine y by means of the given equation of the path y = I(x).
Use (17.77) to solve the following problem. A particle moves along a
smooth wire, shaped in the form of the parabola x 2 = -y. Initially it is at
the origin and has a velocity of 4 ft/sec. Find the position of the particle at
the end of 5 sec. Hint. I(x) = -x 2, I(xo) = 0, I' (x) = -2x.

G. Variable Mass. Rocket Motion. In the straight line motion

problems thus far considered, the mass of a particle or of a body remained
constant throughout the motion. If, however, the mass itself is also chang-
ing with time, decreasing or increasing, then Newton's second law of
motion no longer holds and must be modified. It has been proved, in the
case of a body of variable mass moving in a straight line, that the differen-
tial equation governing its motion is given by

(17.78)
where
m is the mass of the body at time t,
v is the velocity of the body at time t,
F is the algebraic sum of all the forces acting on the body at time t,
dm is the mass joining or leaving the body in the time interval dt,
u is the velocity of dm at the moment it joins or leaves the body,
relative to an observer stationed on the body.
Note that (17.78) differs from Newton's second law of motion by the
term u dmldt.
If a mass dm leaves the system in the time interval dt, dmldt will be a
negative quantity; if it joins the system, dml dt will be a positive quantity.
With the aid of (17.78), solve the following problems.
28. A rocket, which weighs 32M lb and contains fuel weighing 32mo lb, is pro-
pelled straight up from the surface of the earth by burning 32k lb of fuel per
second and expelling it backwards at a constant velocity of A ft/sec relative
to an observer on the rocket. Assume that the only force acting on the rocket
is that of gravity. Find the velocity of the rocket and the distance it travels
as functions of time. Take positive direction upward. Hint. In (17.78), the
variable mass m at time t is m = M + mo - kt; therefore dm/dt = -k.
The relative velocity u of dm is -A. The force of gravity at time tis F =
- (it! + mo - kt)g. Answers are

(17.79) v = -gt - A log (1 - M mo t), 0 < t< M o

m ,
192 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

(17.8) y = At - !gt
2
+ (M + mo - kt) log (1 - M mo t) l

O <t
= <
M+mo.
k
NOTE. If the rocket moves in free space so that it is not subject to the
gravitational force of the earth, then F = 0 in (17.78) and g = 0 in (17.79)
and (17.8).
29. (a) Show that, when the rocket's fuel of problem 28 is exhausted, it has
reached a theoretical height of

(17.81) A mo g
Y = -k- - 2
(mo)2
k + TAM log ( M +M mo) .
Hint. The mass mo of the fuel will be exhausted in time t - mo/k. Sub-
stitute this value in (17.8).
(b) Show that its velocity at that moment is

(17.82) v = - g;O - Alog (M mo) .

30. A rocket of mass M, containing fuel of mass mo, falls to the earth from a
great height. It burns an amount k of its mass per second and ejects it down-
ward with a constant velocity relative to an observer on the rocket of Aft/sec.
Find the distance it falls in time t. Take positive downward direction. Hint.
See problem 28.
31. A rocket and its fuel have mass mo. At the moment it starts to burn an
amount k of its mass per second, it is moving with a velocity Vo. The fuel
is ejected backwards with just enough velocity, so that the ejected fuel is
motionless in space. Find the subsequent velocity and distance equations
of the rocket as functions of time. Hint. For the ejected fuel to be motion-
less relative to an observer on the earth, the backward velocity of the fuel
must equal the forward velocity of the rocket; remember, the fuel at the
instant of ejection has the same forward velocity v as the rocket itself.
Relative to an observer on the rocket, however, it will seem to him as if he
were standing still and the ejected fuel moving away from him at the
rate of -v ft/sec, where +v ft/sec is his own velocity in the positive direc-
tion. Therefore u in (17.78) is -v. The variable mass m = mo - kt
and dm/dt = -k.
32. Solve problem 31 if the rocket were moving in free space. Hint. F = 0
in (17.78).
33. A body moves in a straight line in free space with a velocity of vo ft/sec.
Initially its mass is mo
and as it moves it adds to its mass k slugs per second.
Find its velocity and distance equations as functions of time. Hint. In
(17.78), F = 0 and assuming the added mass dm is stationary in space, then
its velocity relative to an observer on the body is -v, where v is the velocity
of the body. See problem 31.
34. A spherical raindrop falls under the influence of gravity. Its mass increases
by the addition of stationary moisture particles at a rate which is propor-
tional to its surface area. Initially its radius r = roo Find its acceleration,
velocity, and distance equations as functions of time. Take positive direction
downward. Show that if initially ro = 0, the acceleration has the constant
Lesson 17M H. ROTATION OF A LIQUID 193

value g/4. For hints, see answer section. First try to solve without making
use of these hints.
35. A chain unwinds from a coil held at rest. It falls straight down under the
influence of gravity, which is the only acting force. Initially l ft of the chain
are unwound. Find its velocity as a function of time. Take positive direction
down ward. For hints see answer section.
36. Solve problem 35, if the chain must first slide along a frictionless plane in-
clined at an angle 8 with the horizontal before dropping straight down.
Assume that the coil is held at rest at a distance l ft from one end of the plane
and that initially one end of the chain is at the end of the plane. For hints,
see answer section.

H. Rotation of the Liquid in a Cylinder.

37. A vessel of water is rotated about a vertical axis with a constant angular
velocity w. Show that when the water is motionless relative to the vessel, the
surface of the water assumes the form of a paraboloid of revolution. Find
the equation of the curve made by a vertical cross section through the axis of
the cylinder. Hint. See Fig. 17.9. Two forces act on a particle of water at P:

t+
Tangent to the surface

X axis

Figure 17.9

one downward due to the weight mg of the particle; the other mw 2 x due to
the centrifugal force* of revolution. Since the particle of water is motionless,
the resultant of these two forces must be perpendicular to the surface of the
water at P; if it were not, then the resultant would itself have a component
of force in a direction tangent to the curve and thus cause the particle of
water at P to move. You can therefore equate tan 8 = dy/dx with mw 2 x/mg.
38. Assume that the rotating vessel of problem 37 is a cylinder containing a liquid
whose weight per unit volume is p. If the pressure on the axis of the vessel is
Po, show that the pressure at the surface of the liquid at a distance r from
the axis is given by the differential equation dp/dr = pw2 r/g. Then show
that its solution is P = Po +(pw 2r2/2g). Hint. As in problem 37, use the
fact that an element of volume is in equilibrium under the pressures on its
surfaces and centrifugal force.

*For a definition of centrifugal force see Lesson 30M, A.

194 PROBLEMS LEADING TO FIRST ORDER EQUATIONS Chapter 3

39. Assume the cylinder of problem 38 contains a gas whose weight per unit
volume is p and which obeys Boyle's law p = kp, where p is pressure and
k a constant. Show that p = poew2r2/2lcfl. Hint. Replace p by p/k in the
formula given in problem 38.

ANSWERS 17M.
I.(a) 36/11" min. (b) 12(3 - v5)/1I" min.
2.18v2 min. 4. 30.5 min.
3.32v6/1I" min. 5. 12v3 min.
6.Make the substitution u = 12 - vh; 65 min.
· ,=
7 · R (1 - e-BtIL) +., o
Eo -BtIL
e; • E 0 /R • ,=
+
sln wt -
8· ,· = w2L2Eo R2 [R' wL cos wt + wL e -Btl L] + ,· o-Btl
eL.

9. 0.476 amp. II. (a) i = 15e- 5t • (b) 0.14 sec.

10. 0.299 amp. 12. (a) Q = 216011" cal/sec. (b) 64°C.
13. (a) 8.611" cal/sec. (b) 29°C.
14. (a) p = 14.7e- lb/sq in. = 2117e-lch lb/sq ft.
lch
(b) 0.000038.
(c) 10.0Ib/sq in., 8.3 lb/sq in., 1.0 lb/sq in., 0.00060 lb/sq in.
(e) p = 0.081e-o.o00038h.
(f) 0.078 lb/ cu ft, 0.067, 0.055, 0.012.
15. (a) p2/7 = po2/7 - i-k- 5/7h, where Po = 14.7 lb/sq in. = 2116.8 lb/sq ft
and k = 2116.8(0.081) -7/5.
(b) h = = = 91,500 ft = 171 mi.
16. (a) k = 64. (b) p = 5·107(e128h1108 - 1). (c) p = 64e128h1108.
2
(d) 1,296,500Ib/ft , 65.7 Ib/ft •3

19. l = 1 (1 +
e""). Note that for a fixed IL, l is a function only of T.
20. 2p = (1 - p,2)e",,/2, p, = 0.7324.
21. l = 0.63 ft.
22. 8 = 15.9 radians = slightly more than 2! turns of rope about the post.
23. v = -tV2IO ft/sec; t = vi log (6 + Va5) sec.
24. 18.1 ft/sec.
25. 17.0 ft/sec; vi log (2 +
v3) sec.
26. 15.3 ft/sec.
27. x = 20 ft, Y = -400 ft, or x = -20 ft, y = -400 ft.

30. y = igt
2
At - : (M +
- kt) log mo - (1 - ! ,M
O=< t
<
M
k
mo. +
g 2kmovo - gmo 2
31. v = 2k (mo - kt) +
2k(mo - kt) ,
g 2 2 2kmovo - gmo 2 ( k)
y = 4k2 (2m okt - k t ) - 2k2 log 1 - mo t ·
_ movo 1 mo
32. v = movo/(mo - kt), y - k og mo _ kt

33. v = movo/(mo + kt), y

movo I
=- - og mo + kt ·
k mo
Lesson 17M-Answers 195

+
34. First show, see Exercise l5D, 12, that T = TO kt/8, where T is the radius of
the raindrop at time t, k is a proportionality constant and 8 is the mass per
cubic foot of water. Hence dT = k dt/8. The velocity u of the particle dm
is -v where v is the velocity of the raindrop at time t, since relative to an
observer moving with the raindrop, it is as if he were stationary and the
particle were moving to him, in the negative upward direction, with a
velocity equal to his own velocity at the moment the moisture particle at-
taches itself to the raindrop. The variable mass m at time tis m = 4...,.38/3;
dm/dt = k41rT2. Answers are

Acceleration a = (1 + ") ,

.
VelocIty v = k8 4
g (
T -
T04)
T3 '

8 (8)2
k ( 2+ 2)
4
. g TO
Dlstance y = T ,2 - 2TO •

Replacing T by TO + kt/8 in the above equations will give acceleration,

velocity, and distance equations as functions of time.
35. The nlass of chain at time tis (l + y)8, where 8 is the mass of chain per unit
+
length and Z y is the distance fallen in time t. Therefore dm/dt =
8 dy/dt = 8v. The velocity of a piece dm of the chain as it leaves the spool,
relative to a person moving with the chain is -v, where v is the velocity of
the chain at time t. Change dv/dt in (17.78) to v dv/dy. Answer is

1)2(1 + y)2 = [(I + y)3 _ Z3].

+
36. See problem 35. Acting force now is [(l sin 8) 8 y8]g, where y is the vertical
distance the chain has fallen in time t. Answer is
1)2(1 + y2) = [3ly sin 8(21 + y) + l(3l + 2y»),
37. y = w2z 2 /2g, if the origin is taken at the lowest point of the parabola.
 Chapter 4

Linear Differential Equations

of Order Greater Than One

Introductory Remarks. In Lesson 11, we introduced the important

linear differential equation of the first order. The linear differential equa-
tion of higher order which we shall discuss in this chapter has even greater
importance. Motions of pendulums, of elastic springs, of falling bodies,
the flow of electric currents, and many more such types of problems are
intimately related to the solution of a linear differential equation of order
greater than one.
Definition 18.1. A linear differential equation of order n is an
equation which can be written in the form

(18.11) In(x)y(n) + In-I (x)y(n-t> + · · · + II (x)y' + !o(x)y = Q(x),

where lo(x), 11 (x), · · · , In(x), and Q(x) are each continuous functions of x
defined on a common interval I and In(x) 1'= 0 in I. *
Note that in a linear differential equation of order n, y and each of its
derivatives have exponent one. It cannot have, for example, terms
such as y2 or (y')1/2 or [y(n)]3.

Definition 18.12. If Q(x) 1'= 0 on I, (18.11) is called a nonhomo-

geneous linear differential equation of order n. If, in (18.11), Q(x) ==
o on I, the resulting equation
(18.13) In(x)y(n) + In-I (x)y(n-t> + · · · + 11 (x)y' + lo(x)y = 0

is called a homogeneous linear differential equation of order n.

*The notation "/n(X) = 0 in I" (or on I or over I) means /n(X) = 0 for every x in 1
(or on I or over I) and is read as "/n(X) is identically zero in 1 (or on I or over I)."
The symbol/n(x) ¢ 0 in I means /n(X) is not equal to zero for every x in 1 (or on 1 or
over 1), although it may equal zero for 80me x in 1. It is read as "/n(X) is not identically
zero in 1 (or on 1 or over I)."
196
Lesson 18A COMPLEX NUMBERS 197

Remark. Do not confuse the term homogeneous as used in Lesson 7

with the above use of the term.
For a clearer understanding of the solutions of linear differential equa-
tions of order n you should be familiar with:
1. Complex numbers and complex functions.
2. The meaning of the linear independence of a set of functions.
We shall, therefore, briefly discuss these topics in this and the next lesson.

LESSON 18. Complex Numbers and Complex Functions.

LESSON 18A. Complex Numhers. We shall not attempt to enter

into a lengthy discussion of the theory of the number system, but will
indicate only briefly the important facts we shall need. The real num-
ber system consists of:
1. The rational numbers; examples are the positive integers, zero, the
negative integers, fractions formed with integers such as f, -t, and f.
2. The irrational numbers; examples are 0, 7r, and e.

Definition 18.2. A pure imaginary number is the product of a

real number and a number i which is defined by the relation i 2 = -1.
Examples of pure imaginary numbers are ai, -5i, 0 i, and ..vr i.
Definition 18.21. A complex number is one which can be written
in the form a + bi, where a and b are real numbers.
It is evident that the complex numbers include the real numbers since
every real number can be written as a + Oi. They also include the pure
imaginary numbers, since every pure imaginary number can be written
as 0 + bi.
Definition 18.22. The a in the complex number z = a + bi is called
the real part of Z; the b the imaginary part of z.
Note that b, the imaginary part of the complex number, is itself real.
Remark. The complex numbers developed historically because of the
necessity of solving equations of the type
x2 +1= 0, x2 + 2x + 2 = o.
If we require x to be real, then these equations have no solutions; if, how-
ever, x may be complex, then the solutions are respectively x = ±i and
x = -1 ± i. In fact, if we admit complex numbers, we can make the
following very important assertion. It is so important that it has been
labeled the "Fundamental Theorem of Algebra."
198 HIGHER ORDER LINEAR DIFFERENTIAL EQUATIONS Chapter 4.

Theorem 18.23. Every equation of the form

(18.24) anx n +
an_lxn- I + ···
+ alx ao + = 0, an 0,
where the a's are complex numbers has at least one root and not more than 11,

distinct roots. Its left Bide can be written as

(18.25) an(x - rl) (x - r2) · · · (x - rn),
where the r's are complex numbers which need not be distinct.
Comment 18.26. Although Theorem 18.23 tells us that (18.24) has 11,
roots, it does not tell us how to find them. If the coefficients are real
numbers, then you may have learned Horner's or Newton's method for
finding approximate real roots. Or you can set y equal to the left side of
(18.24) and draw a careful graph. The approximate values of the coordi-
nates of the points where the curve crosses the x axis will also give the
real roots of (18.24).
To find the imaginary roots of (18.24) is a more difficult matter. How-
ever, there are methods available for approximating such roots. *
Definition 18.3. If z = a + bi, then the conjugate of :6, written
as z, is Z = a - bi.
To form the conjugate of a complex number, change the sign of the
coefficient of i. For example if z = 2 + 3i, Z = 2 - 3i; if z = 2 - 3i,
! = 2 + 3i.
Definition 18.31. If z = a + bi, then the absolute value of :6,
written as Izl, is Izl = v'a 2 + b2 •
For example, if z = 2 + 3i, then
Imaginary axis Izi = v'22 + 3 2 = v'I3; if z = 2 -
3i, then Izl = v'22 + (-3)2 =
e2+Si v'I3. (NOTE. By our definition Izl
is always a non-negative number.)
If, in a rectangular coordinate
system, we label the x axis as the
(0,0) Real axis real axis and the y axis as the im-
aginary axis, we may represent any
complex number z = a + bi graphi-
- 2 - Si e e2-Si cally. We plot a along the real axis
and b along the imaginary axis. In
Fig. 18.32, we have plotted the
Figure 18.32 numbers 2 + 3i, 2 - 3i, -2 + 3i,
-2 - 3i.
By the graphical method of representing complex numbers, it is easy to
establish relationships between a complex number and the polar coordi-
*w. E. Milne, Numerical Calculus, Princeton University Press, 1949.
LeUOD 18A COMPLEX NUMBERS 199

nates of its point. In Fig. 18.33, we have represented graphically the com-
plex number 2 = X + iy and the polar coordinate (r,8) of its point. If
Z=X + iy=(r, 8)
Izi = r==.J x + y2
2

(0,0) x

Figure 18.33

you examine this figure carefully, you will have no trouble establishing
the following relationships.
(18.4) 121 = r = v' x 2 + y2,
(18.41) x = r cos 8,
(18.42) y = r sin 8.
Definition 18.5. If in
(18.51) z = x + yi,
(called the rectangular form of z)) we replace x by (18.41), and y by
(18.42), we obtain

(18.52) 2 = r cos 8 + ir sin 8 = r(cos 8 + i sin 8),

(called the polar form of z).
Definition 18.53. The polar angle 8 in Fig. 18.33 is called the Argu-
ment of z, written as Arg z. In general, Arg z is defined to be the smallest
positive angle satisfying the two equalities

(18.54) cos 8 = 1:1 ' sin 8 = 1;1 ·

Fonnulas (18.4) to (18.42) plus (18.52) and (18.54) enable us to change
a complex number from its rectangular form to its polar form and vice
versa.
Remark. Note that we have written arg 2 with a capital A to desig-
nate the smallest or principal value of 6. If written with a small a, then
arg z = Arg z + 2nr , n = 1,2,3 ...
Esample 18.55. Find Arg 2 and the polar form of 2, if the rectangular
fonn of 2 is
(a) 2 = 1 - i.
200 HIGHER ORDER LINEAR DIFFERENTIAL EQUATIONS Chapter 4.

Solution. Comparing (a) with (18.51), we see that x = 1 and y =

-1. Therefore by (18.4)
(b) /z/ = r = V12 + (-1)2 = V2.
By (18.54)
1 . -1
(c) cos 8 = - , sin 8 = - .
V2 V2
Hence 8 is a fourth quadrant angle. By Definition 18.53, therefore,
7r
(d) Argz = T·
Finally by (18.52), and making use of (b) and (d), we have

(e) 7r +
_ /n2 ( cosT .· 7r)

which is the polar form of z.

Example 18.56. Find the rectangular form of z if its polar fonn is

(a) z = V2 (cos ; + i sin ;) ,

Solution. Comparing (a) with (18.52), we see that
r
(b) r = V2, 8 =-.
3
Hence by (18.41) and (18.42),

(c) x = V2cos 7r = V2 and y = V2sin 7r = V6,

3 2 3 2
By (18.51), the rectangular form of z is therefore

(d) z = !(v'2 + V6 i).

LESSON 18B. Algebra of Complex Numbers. Complex numbers
would not be of much use if rules were not available by which we could
add, subtract, multiply, and divide them. The rules which have been
laid down for these operations follow the ordinary rules of algebra with
the number i 2 replaced by its agreed value -1. If Zl = a + bi and
Z2 = C +
di are two complex numbers, then by definition,

(18.6) Zl + Z2 = (a + bi) + (c + di) = (a + c) + (b + d)i,

(18.61) Zl - Z2 = (a + bi) - (c + di) = (a - c) + (b - d)i,
Lesson 18C COMPLEX FUNCTIONS 201

(18.62) ZlZ2 = (a + bi)(e

+ di) = ae + adi + bei + bdi 2
= + (ad + be)i,
(ae - bd)
Zl a -+- bi e - di (ae -+- bd) + (be - ad)i
(18.63)
Z2 = e + di . e - di = e2 + d 2
ae -+- bd be - ad .
e2 + d .,... O.
2
= -e2-+-d-2 + e2 + d 2
Example 18.64. If Zl = 3 + 5i and Z2 = 1 - 3i, compute Zl + Z2,
Zl - Z2, ZlZ2, Zl/Z2'

Solutions.
Zl + Z2 = (3 + 5i) + (1 - 3i) = 4 + 2i.
Zl - Z2 = (3 + 5i) - (1 - 3i) = 2 + Si.
ZlZ2 = (3 + 5i)(1 - 3i) = (3 + 15) + (-9 + 5)i = 18 - 4i.
Zl = 3 + 5i. 1 + 3i = (3 - 15) + (9 + 5)i = _ 12 + 14 i
Z2 1 - 3i 1 + 3i 1+9 10 10'

LESSON 18C. Exponential, Trigonometric, and Hyperbolic Func-

tions of Complex Numbers. If x represents a real number, i.e., if x
can take on only real values, then the Maclaurin series expansions (which
you studied in the calculus) for eX, sin x and cos x are

X x2 x3
(18.7) eX = 1 + I! + 2! + 3! + ... , -00 < x < 00,
. x3 x5
(18.71) SIn x = x - 3! + 5! - . · ., -00 < x < 00,
x2 X4
(18.72) cos X = 1 - 2! + 4! - .." - 00 < x < 00.
Each of these series is valid for all values of x, i.e., each series converges
for all x. If Z represents a complex number, i.e., if Z can take on complex
values, we define
Z Z2 z3
(18.73) e' = 1+ -I! + -2! + -3! + · .. '
. Z3 z5
(18.74) slnz=z--+--···
3! 5! '
Z2 Z4
(18.75) cos Z = 1- - -2! + 4!
- - .·· .
It has been proved that each of these series also converges for all z.
202 HIGHER ORDER LINEAR DIFFERENTIAL EQUATIONS Chapter 4.

Example 18.76. Find the series expansion for cos i.

Solution. By (18.75) with z = i and i 2 = -1, we have

cos i = 1 + 1:..
2! + 1:.. 6! + · · · ·
+ 1:..
(a)
41

Note that cos i turns out to be a real number.

By means of (18.73) to (18.75), we can prove the following identities.
These are the ones you will meet most frequently, and the ones we shall
need.
(18.8)
(18.81)
(18.82) ei , = cos z + i sin z,
(18.83) e-i , = cos z - i sin z,

(18.84) . 1 (i, -i,)

SIn z = 2i e - e ,
(18.85) cos z = !(ei , + e-i ,),
(18.86) e' 0 for any value of z.

Proofs of (18.8) to (18.86). We give below an outline of the proofs of

(18.8) to (18.86). Rigorous proofs would be beyond the scope of this text.

Proof of (18.8). In (18.73) replace z by zero.

Proof of (18.81). By (18.73)

2 3
e'l = 1 + ZlI! + 2!
+ 3! + ... '
2 3
e'2 = 1 + Z2 + Z2 + Z2 + ...
I! 2! 3!
Since each series also converges absolutely, we may multiply them to
obtain
Lesson 18-Exercise 203

Proof of (18.82). Replace z by iz in (18.73). There results, with

2
i = -1,
• ·2 2 ·3 3 ·4 4
i
e- = 1 + ,z
I!
+ 2!
+ 3!
+ 4!
+ ...
Z2 iz 3 z4
= 1 + iz - 2! - aT + 4! + · · ·

= (1 - + +. · .) + i (2 - +. · .)
= cos z + i sin z [by (18.74) and (18.75)].
Proof of (18.83). Replace z by -iz in (18.73) and proceed as above.
Or replace z by -z in (18.82) and then note by (18.74) and (18.75) that
sin (-z) = -sin z; cos (-z) = cos z.
Proof oj (18.84). Subtract (18.83) from (18.82) and solve for sin z.
Proof of (18.85). Add (18.82) and (18.83) and solve for cos z.
Proof of (18.86). By (18.81) and (18.8)

Since the product e8 e-8 = 1, e8 cannot have the value zero for any z.
The combinations

appear so frequently in problems that, for convenience, symbols have

been introduced to represent them. These are

(18.9)

(18.91)
-.e
8
+ e- 8

coshz= 2 '

sinh z e8 - e- 8
(18.92) tanhz=
cosh z
=
e- + e-- ·
These three functions are called respectively the hyperbolic sine, the
hyperbolic cosine, and the hyperbolic tangent of z.

EXERCISE 18
1. Find the conjugate of each of the following complex numbers. (a) 1 + i.
(b) 3. (c) -3i. (d) 5 - 6i. (e) -2. (f) 2i. (g) -3 - 4i.
2. Find the absolute value of each of the complex numbers in 1 above.
204 HIGHER ORDER LINEAR DIFFERENTIAL EQUATIONS Chapter 4

3. Find Arg Z and the polar form of Z of each of the complex numbers in 1 above.
4. Find the rectangular form of z, if its polar form is:

(a) 2 (cos : + i sin · (e) v'2 (cos £r + i sin £r).

(b) 3 (cos 2; + 2;) · i sin (f) va (cos ir + i sin ir).

(c) 4(cos + i sin '1').
'I'

(d) 5 (cos ; + i sin ;).

5. Given Zl = -2 + i, Z2 = 3 - 2i. Find: (a) Zl + Z2. (b) Zl - Z2.

(c) ZlZ2. (d) Zl/Z2.
6. If Zl and Z2 are two complex numbers, prove:

(a) Zl + Z2 = Zl + Z2. (c) Zlz2 = Zl Z2.

(b) -Zl = -Zl. (d) Zl/Z2 = Zl/Z2, Z2 :F o.
7. With the help of 6, prove that if Zl, Z2, Z3 are complex numbers and
Z = ZlZ2 + ZI(Z2 + Z3) + Zl + Z2 - Z3 + ZI/Z2, Z2 0,
then

NOTE. If you have proved 7, then you have proved that if Z is obtained
by adding, subtracting, multiplying, and dividing complex numbers Zl, Z2,
Z3, .:...: ._'then z can be obtained by performing the same arithmetic operations
on Zl, Z2, Z3, • • • •
8. With the help of 7, prove that if r is a root of
(a)
then if is a root of

(b) anX
n
+ an_IX n
-
1
+ . · . + alx + ao = O.
9. With the help of 8, prove that if the coefficients in 8(a) are real and r is a
root of 8(a), then if is also a root of 8(a). NOTE. If you have succeeded in
proving 9, then you have proved the following very important theorem. If
the coefficients in 8(a) are real, then its imaginary roots must occur in con-
jugate pairs.
10. Prove that the product of two conjugate complex numbers is a positive real
number.

ANSWERS 18
1. (a) 1 - i. (b) 3. (c) 3i. (d) 5 + 6i (e) -2 (f) -2i.
(g) -3 4i. +
2. (a) V2. (b) 3. (c) 3. (d) V61. (e) 2. (f) 2. (g) 5.

3. (a) : ' v'2 (cos + i sin ·

Lesson 19A LINEAR INDEPENDENCE OF FUNCTIONS 205

(b) 0, 3(cos 0° + i sin 0°).

3'1' 3 ( 3'1' + . . 311")
( ) 2'
C cos 2 ." 2 . Sln

(d) 309°48', V6I(cos 309°48' + i sin 309°48').

(e) '1', 2(cos + i sin '1').
'I'

(f) !, 2 (cos 'I' + i sin '1') .

2 2 2
(g) 233°8', 5(cos 233°8' + i sin 233°8').
4. (a) V2 + iV2. (b) -J + itVa. (c) -4. (d) 5i.
(e) -1 - i. (f) -Va i. (g) !Va - if.
5. (a) 1 - i. (b) -5 + 3i. (c) -4 + 7i. (d) -fa - ila .
LESSON 19. Linear Independence of Functions. The Linear
Differential Equation of Order n.

LESSON 19A. Linear Independence of Functions.

Definition 19.1. A set of functions II (x), 12(X), · • • , In(x), each de-
fined on a common interval I, is called linearly dependent on I, if
there exists a set of constants CI, C2, • • • , Cn, not all zero, such that

(19.11)

for every x in I. If no such set of constants el, C2, • • • , Cn exists, then the set
of functions is called linearly independent.
Definition 19.12. The left side of (19.11) is called a linear combina-
tion of the set of functions/l(x), 12(X), • • • ,In(x).
Example 19.13. For each of the following sets of functions, determine
whether it is linearly dependent or independent.

1. x, -2x, -3x, 4x, I: -00 <x < 00 •

2 . x P , x q, P :F q, x > 0, x < 0.
3. ePx , eqX , p :F q, I: -00 <x< 00 .

4. eX, 0, sin x, 1, I: -00 <X< 00 .

Solution. First we observe that all the functions in each set are
defined on their respective intervals. Second we form for each set the
linear combination called for in (19.11).
The linear combination of the set of functions in 1, equated to zero, is

(a)

We must now ask and answer this question. Does a set of c's exist, not
all zero, which will make (a) a true equation for every x in the interval
206 HIGHER ORDER LINEAR DIFFERENTIAL EQUATIONS Chapter'

- 00 < x < oo? Rewriting (a) as

(b)

we see immediately there are an infinite number of sets of c's which will
satisfy (b), and therefore (a), for every x in I. For example,

(c) Ct = 2, C3 = 0,
Ct = 5, C3 = 1,
are two such sets. Hence by Definition 19.1 the set of functions in 1 is
linearly dependent.
The linear combination of the set of functions in 2, equated to zero, is

(d)

If we assume x P and Xf are linearly dependent functions in I, then by

Definition 19.1, there must exist constants Cl and C2 both not zero such
that (d) is an identity in x. Since Ct and C2 are not both zero, we may
choose one of them, say Cl ¢ o. Dividing (d) by CtXf, we obtain
P-f C2
(e) X = - - , p ¢ q.
Ct

The value of the left side of (e) varies with each x in the interval x > 0,
x < o. The right side, however, is a constant for fixed values of Cl and
C2. Hence to assume that x P , Xf are linearly dependent leads to a contra-
diction. They must therefore be linearly independent.
The proof that the functions in 3, eP %, ef:Z, p q are linearly independent
is practically identical with the proof that x P and Xf, p ¢ q are linearly
independent. It has therefore been left to you as an exercise; see Exercise
19,1.
The linear combination of the set of functions in 4, equated to zero, is

(f)

Choose Ct = Ca = C4 = 0 and C2 = any number not zero. Then (f) will

read
(g) 0 · eX + C2 • 0 0 · sin x +
0 · 1 = 0, +
which is a true equation for every x in I. Hence the given set is linearly
dependent.
Comment 19.14. Whenever a set of functions contains zero for one
of its members, the set must be linearly dependent. All you have to do is
to choose for every C the value zero, except the one which is the coefficient
of zero.
Lesson 19B THE LINEAR DIFFERENTIAL EQUATION OF ORDER n 207

If a set of functions were selected at random, one could work a long time
trying to find a set of constants CI, C2, • • • , Cn, not all zero, which would
make (19.11) true. However, a failure to find such a set would not of
itself permit one to assert that the given set of functions was independent,
since the possible values assignable to the set of c's are infinite in number.
It is conceivable that such a set of constants exist, not easily discoverable,
for which (19.11) holds. Fortunately there are tests available to deter-
mine whether a set of functions is linearly dependent or independent. A
discussion of these tests will be found in Lesson 63B and Exercise 63,5.
You may be wondering what linear dependence and linear independence
of a set of functions has to do with solving a linear differential equation of
order n. It turns out that a homogeneous linear differential equation has
as many linearly independent solutions as the order of its equation. (For
its proof, see Theorems 19.3 and 65.4.) If the homogeneous linear differ-
ential equation, therefore, is of order n, we must find not only n solutions
but must also be sure that these n solutions are linearly independent.
And as we shall show, the linear combination of these n solutions will be
its true general solution. For instance, if we solved a fourth order homo-
geneous linear differential equation and thought the four functions x, -2x,
-3x, 4x in Example 19.13-1 were its four distinct and different solutions,
we would be very much mistaken. All of them can be included in the one
solution y = cx. Hence we would have to hunt for three more solutions.
On the other hand, the functions x and x 2 , which we proved were linearly
independent (see Example 19.13-2 with p = 1, q = 2), could be two dis-
tinct solutions of a homogeneous linear differential equation of order two.
In that event, the linear combination CIX + C2X2 would be its general
solution.
Hence in solving an nth order homogeneous linear differential equation,
we must not only find n solutions, but must also show that they are
linearly independent.

LESSON 19B. The Linear Differential Equation of Order n.

We have already defined in 18.1 the linear differential equation of order n.
In connection with this equation, there are two extremely important
theorems. The proof of the first one, Theorem 19.2 below, is too com-
plicated to be given at this stage and would only lead us too far afield.
In order, therefore, not to delay the study of methods of solving nth
order linear differential equations more than necessary, we have post-
poned its proof to Lesson 65. The proof of the second one, however,
Theorem 19.3 below, is given in this lesson.
Theorem 19.2. If fo(x) , fl (x), · · · , fn(x) and Q(x) are each continuous
functions of x on a common interval I, and fn(x) 0 when x is in I, then
208 HIGHER ORDER LINEAR DIFFERENTIAL EQUATIONS Chapter"

the linear differential equation

(19.21) fn(x)y(n) + fn-I (x)y(n-U + · · · + fl (x)y' + fo(x)y = Q(x)

has one and only one solution,

(19.22) Y = y(x),

satisfying the set of initial conditions

(19.23) y(xo) = Yo, Y (n-U(x O

) =nY- I ,

where Xo is in I, and Yo, Yl, · · · , Yn-I are constants.

This theorem is an existence and a uniqueness theorem. It is an existence
theorem because it gives the conditions under which a solution of (19.21)
satisfying (19.23) must exist. It is also a uniqueness theorem because it
gives conditions under which the solution of (19.21) satisfying (19.23) is
unique. As remarked previously, the proof of this very important theorem
has been postponed to Lesson 65; see Theorem 65.2.
Right now we shall content ourselves with the proofs of three important
properties of linear differential equations that we shall need for a clearer
understanding of the remaining lessons of this chapter. These properties
are incorporated in the following theorem.
Theorem 19.3. If fo(x), fl (x), · · · , fn(x) and Q(x) are each continuous
functions of x on a common interval I and fn(x) 0 when x is in I, then:
1. The homogeneous linear differential equation
(19.31) fn(x)y(n) + fn-I (x)y(n-I) + · · · + fl (x)y' + fo(x)y = 0
has n linearly independent solutions YI (x), Y2(X), · · · , Yn(x).
2. The linear combination of these n solutions
(19.32)
where Cl, C2, · · · , Cn is a set of n arbitrary constants, is also a solutioo, oj
(19.31). It is an n-parameter family of solutions of (19.31). (We shall
explain later the significance of the subscript c in 'Ye.)
3. The function
(19.33) y(x) = yc(x) + yp(x),
where yc(x) is defined in (19.32) and yp(x) is a particular solution of the
nonhomogeneous linear differential equation corresponding to (19.31),
namely
(19.34) f n(x )y(n) + fn - I (x )y(n-t> + ··· + f I (x )y' +f 0 (x)y = Q(x),

is an n-parameter family of solutions of (19.34).

Lesson 19B THE LINEAR DIFFERENTIAL EQUATION OF ORDER 11, 209

Proof of 1. We shall, for the present, limit the proof of 1 to the special
case where the coefficients lo{x), • • • , I", (x) are constants. The proof will
consist in actually showing, in lessons 20 to 22 which follow, how to find
these n linearly independent solutions. The proof, for the case where the
coefficients are not constants, has been deferred to Lesson 65, Theorem 65.4.

Proof of 2. By hypothesis each function Yl, Y2, • • • , Yn is a solution

of (19.31). Hence each satisfies (19.31). This means, by Definition 3.4
and the Remark at the bottom of page 22, that

(a) In{X)YI("') + 1",_I{X)YI(n-O + .. · + Il{X)YI' + lo{x)YI = 0,

In(X)Y2("') + 1",-1 (X)Y2("'-0 + · · · + II{X)Y2' + IO(X)Y2 = 0,
...................................
I", (X)Yn (",) + In-I (X)y", + · · · + II (X)y",' + IO{x)Yn = o.
(",-1)

Multiply the first equation of (a) by Cl, the second by C2, • • • , the last by
c"" where Cl, C2, • • • , Cn are n arbitrary constants, and add them. The
result is

(b) In{X)[CIYI (n) + C2Y2 + · · · + CnYn (n)]

(11,)

+ 1",-1 {X)[CIYl (n-O + C2Y2 + · · · + C",Yn (n-I)] + · · ·

(11,-0

+ lo{x)[cIYI + C2Y2 + · · · + e",y",] = o.

This last equation can be written as·
(c) + C2Y2 + · · · + cny](n)
In{X)[CIYI
+ In-I (X)[CIYl + · · · + CnYn](n-1) + · · ·
+ IO{X)[CIYl + · · · + cnYn] = o.
By (19.32) each quantity inside the brackets is Ye. Hence (c) is

(d) In {X)Yc (n) + In-I {X)Yc (n-O + · · · + lo{x)yc = 0,

which says that Ye satisfies (19.31). It is therefore a solution. Since it is

a linear combination of n independent solutions and contains n parameters,
it is an n-parameter family of solutions.
Proof of 3. By hypothesis, YP is a particular solution of (19.34).
Hence, by Definition 3.4,

(e) In {x)yp (n) + In-l {x)yp (n-1) + · · · + 11 (X)Yp' + lo{x)yp = Q(X).

*For example,
ux
"( ) + vx
"() == £l2u(x) + £l2v(x) +
== £l2(u
dx2
v) = (
U
+ v.
)"
210 HIGHER ORDER LINEAR DIFFERENTIAL EQUATIONS Chapter 4

Adding (d) and (e) we obtain

(f) fn(x)(Ye + Yp)(n) + fn-l (x) (Ye + Yp)(n-U
+ . · · + fo(x)(Ye + Yp) = Q(x).
This last equation says that (Ye + Yp) satisfies (19.34) and is therefore a
solution. !i'urther, since Y = Ye + YP contains n parameters, it is an
n-parameter family of solutions of (19.34).
Definition 19.4. The solution Ye(x) in (19.32) of the homogeneous
equation (19.31) is called the complementary function of (19.34).
Hence the use of the subscript c in Ye(x).
Remark. The subscript p in Yp(x) of (19.33) is used to distinguish it
from the Ye part of the n-parameter family of solutions.
Comment 19.41. We shall prove in Lesson 65, see Theorem 65.5,
that the function Ye(x) of (19.32), which is an n-parameter family of solu-
tions of (19.31), is in fact its true general solution in accordance with our
Definition 4.7. Every particular solution of (19.31) can be obtained from
it by properly choosing the n arbitrary constants. We therefore shall refer
to this n-parameter family of solutions as a general solution. We infer fur-
ther from this theorem that if one person has obtained the n independent
solutions Yl, Y2, · · • , Yn (whose linear combination is the general solution)
by using one method, then it is not possible for a second person using an-
other method to find a general solution ,vhich is essentially different
from it. The second person's solution will be obtainable from the first by
a proper choice of the constants Cl, C2, • • • , Cn in (19.32).
A similar statement can be made for the function y(x) in (19.33). The
proof will be found in Theorem 65.6. We shall therefore refer to it as the
general solution of (19.34).
EXERCISE 19
I. Prove that if p :;C q, the functions ePX and eqx are linearly independent.
Hint. Follow the method used in proving the linear independence of the
functions in Example 19.13-2.
2. Prove that the functions eGX and xeGX are linearly independent. Hint. Make
use of Example 19.13-2 with p = 0, q = 1, and the ·fact that e· 0 for
all z.
3. Prove that the functions sin x, 0, cos x are linearly dependent.
4. Prove that the functions 3e 2x and -2e 2x are linearly dependent.
It is extremely important that you prove the statements in Exercises 5 to
7 below. Follow the method used to prove statements 2 and 3 of Theorem
19.3.
5. If YP is a solution of
(19.5) fn(x)y(n) + · · · + fl(X)y' + fo(x)y = Q(x),

then Ayp is a solution of (19.5) with Q(x) replaced by AQ(x).

Lesson 20 LINEAR EQUATION WITH CONSTANT COEFFICIENTS 211

6. Principle of Superposition. (Also see Comment 24.25.) If YPI is a solution

of (19.5) with Q(x) replaced by Ql (x) and YP2 is a solution of (19.5) with
Q(x) replaced by Q2(X), then YP = YPI +
YP2 is a solution of
In(x)y(n) + · · · + 11 (x)y' + lo(x)y = Ql(X) + Q2(X).
7. If yp(x) = u(x) + iv(x) is a solution of
(19.51) In(x)y(n) + · · · + 11 (x)y' + lo(x)y = R(x) + is(x),
where lo(x), • • • , In (x) are real functions of x, then
(a) the real part of YP' i.e., u(x), is a solution of
In(x)y(n) + · · · + II (x)y' + lo(x)y = R(x),
(b) the imaginary part of YP' i.e., vex), is a solution of
In(x)y(n) + · · · + 11 (X)y' + lo(x)y = SeX).

Hint. Two complex numbers are equal if and only if their real parts are
equal and their imaginary parts are equal.
8. Assume that YP = x 2 is a solution of y" + y' - 2y = 2(1 + x - x 2 ).
Use 5 above to find a particular solution of y" + y' - 2y = 6(1 + x - x 2 ).
Answer: y p = 3x 2 • Verify the correctness of this result.
9. Assume that y PI = 1 + x is a solution of
y" - y' +y = x,
and YP2 - e2x is a solution of
y" - y' +y = 3e 2x •
Use 6 above to find a particular solution of y" - y' y = x + +
3e2x •
Answer: y p = 1 + x + e2x • Verify the correctness of this result.
10. Assume that YP = (to cos x - fa sin x) +
i(-to sin x +
fa cos x) is a solu-
tion of y" - 3y' + 2y = eix = cos x + i sin x. Use 7 above to find a
particular solution of (a) y" - 3y' + 2y = cos x, (b) y" - 3y' 2y = +
sin x. Ans. (a) y(x) = -to cos x - fa sin x. (b) y(x) = to sin x+ 1 0 cos x. 3

Verify the correctness of each of these results.

II. Prove that two functions are linearly dependent if one is a constant multiple
of the other.
12. Assume II, 12, la are three linearly independent functions. Show that the
addition to the set of one of these functions, say II, makes the new set linearly
dependent.
13. Prove that a set of functions II, 12, · · · , In, is linearly dependent if two
functions of the set are the same.
14. Prove that a set of functions, 11, 12, · · · ,In, is linearly dependent, if a subset,
i.e., if a part of the set, is linearly dependent.

LESSON 20. Solution of the Homogeneous Linear Differential

Equation of Order n with Constant Coefficients.

LESSON 20A. General Form of Its Solutions. In actual practice,

equations of the type (19.31), where the coefficients are functions of x
with no restrictions placed on their simplicity or complexity, do not
usually have solutions expressible in terms of elementary functions. And
212 HIGHER ORDER LINEAR DIFFERENTIAL EQUATIONS Chapter"

even when they do, it is in general extremely difficult to find them. If,
however, each coefficient in (19.31) is a constant, then solutions in terms
of elementary functions can be readily obtained. For the next few lessons,
therefore, we shall concentrate on solving the differential equation
(20.1)
where ao, ai, · · · , an are constants and an o.
Without going into the question of motivation, let us guess that a
possible solution of (20.1) has the form
(20.11)
We now ask ourselves this question. For what value of m will (20.11) be
a solution of (20.1)" By Definition 3.4, it must be a value for which
d n mx+ dn-l
+ d
(20.12) an dxn e mx+
an-l dxn- 1 e · · · al dx emx+ aoemx = 0·

Since the kth derivative of emx = mkemx , we may rewrite (20.12) as

(20.13) anmnemx + an_lmn-lemX + ··· +almemx +
aoemx = o.
By (18.86), emx 0 for all m and x. We therefore can divide (20.13) by
it to obtain
(20.14)
We at last have the answer to our question. Each value of m for which
(20.14) is true will make y = emz a solution of (20.1).
But (20.14) is an algebraic equation in m of degree n, and therefore, by
the fundamental theorem of algebra (see Theorem 18.23), it has at least
one and not more than n distinct roots. Let us call these n roots mI,
m2, · · · , m n , where the m's need not all be distinct. Then each function

(20.15)
is a solution of (20.1).
Definition 20.16. Equation (20.14) is called the characteristic
equation of (20.1).
NOTE. The characteristic equation (20.14) is easily obtainable from
(20.1). Replace y by m and the order of the derivative by a numerically
equal exponent.
In solving the characteristic equation (20.14), the following three pos-
sibilities may occur.
1. All its roots are distinct and real.
2. All its roots are real but some of its roots repeat.
3. All its roots are imaginary.
Lesson 20B ROOTS REAL AND DISTINCT 213

We shall discuss each of the above three possibilities separately. Other

possibilities may also occur as, for example, when all roots are distinct
but some are real and some are imaginary. It will be made apparent
why such Qther possible combinations do not require special consideration.
Remark. We have already commented, see Comment 18.26, in regard
to the difficulty, in general, of finding the n roots of the characteristic
equation (20.14). If an equation of this type, of degree greater than two,
were written at random, the probability is very high that it would have
irrational or complex roots which would be difficult and laborious to find.
Since this is a differential equations text and not an algebra text, the
examples used for illustration and exercises have been carefully chosen
so that their roots can be readily found. In practical problems, however,
finding the roots of the resulting characteristic equation may not be and
usually will not be an easy task. We cannot emphasize this point too
strongly.

LESSON 20B. Roots of the Characteristic Equation (20.14) Real

and Distinct. If the n roots mt, m2, ••• , m.", of the characteristic
equation (20.14) are distinct, then the n solutions of (20.1), namely,
Y2-- e , yn-- emnx
m2x
(20.2) ... ,

are linearly independent functions. (For proof when n = 2, see Exercise

19,1. For proof when n > 2, see Example 64.2.) Hence by Theorem
19.3 [see in particular (19.32)] and Comment 19.41,
(20.21)
is the general solution of (20.1).
Example 20.22. Find the general solution of
(a) y'" + 2y" - y' - 2y = O.
Solution. By Definition 20.16, the characteristic equation of (a) is
(b) ma + 2m 2 - m - 2 = 0,
whose roots are
(c) ma = -2.
Hence by (20.21) the general solution of (a) is
(d)
Example 20.23. Find the particular solution y(x) of
(a) y" - 3y' + 2y = 0,
for which Yc(O) = 1, yc'(O) = O.
214 HIGHER ORDER LINEAR DIFFERENTIAL EQUATIONS Chapter"

Solution. By Definition 20.16, the characteristic equation of (a) is

(b) m2 - 3m +2= 0,
whose roots are
(c)

Hence by (20.21) the general solution of (a) is

(d)

By differentiation of (d), we obtain

(e)

Substituting the given initial conditions x = 0, Yc = 1, Y/ =

and (e), there results
° in, (d)

(f) 1 + C2,
= Cl

o = Cl + 2C2.
Solving (f) simultaneously for Cl and C2, we find Cl = 2, C2 = -1.
Substituting these values in (d), we obtain the required particular solution
(g)

LESSON 20C. Roots of Characteristic Equation (20.14) Real but

Some Multiple. If two or more roots of the characteristic equation
(20.14) are alike, then the functions (20.15) formed with each of these 11,
roots are not linearly independent. For example, the characteristic equa-
tion of
(a) y" - 4y' + 4y = 0
.
IS

(b) m2 - 4m +4 = 0,

which has the double root m = 2. It is easy to show that the two func-
tions Yl = e 2z and Y2 = e 2z are linearly dependent. Form the linear
combination
(c)

and take Cl = 1, C2 = -1. Hence the general solution of (a) could not
be Y = cle
2z
+ (Remember, a general solution of a second order
C2 e2z •
linear differential equation is a linear combination of two linearly inde-
pendent solutions.) Actually we can write the solution as
(d)
Lesson 20C ROOTS REAL BUT SOME MULTIPLE 215

from which we see that we really have only one solution and not two.
We must therefore search for a second solution, independent of e2z •
To generalize matters for the second order linear differential equation,
we confine our attention to the equation
(20.3) y" - 2ay' + a 2y = 0,
whose characteristic equation
(20.31) m2 - 2am + a2 = 0
has the double root m = a. Let
(20.32)

where u is a function of x. We now ask ourselves the usual question.

What must u look like for (20.32) to be a solution of (20.3)? We know,
by Definition 3.4, that (20.32) will be a solution of (20.3) if
(20.33) (ueClZ)" - 2a(uetlZ )' + a2 ( ueClZ) = o.
Performing the indicated differentiations in (20.33), we obtain
(20.34)

which simplifies to
(20.35)

By (18.86), eClz o. Hence, (20.35) will be true if and only if

(20.36) u" = o.
of (20.36) twice gives
(20.37)

We now have the answer to our question. If, in (20.32), u has the value
(20.37), then
(20.38) Yc = (Cl + c2 x )eClZ
will be a solution (20.3). It will be the general solution of (20.3) provided
the two functions eClZ and xetlZ are linearly independent. The proof that
they are indeed linearly independent was left to you as an exercise; see
Exercise 19,2. Hence by Theorem 19.3 and Comment 19.41, (20.38)
is the general solution of (20.3).
In general it can be shown that if the characteristic equation (20.14)
has a root m = a, which repeats n times, then the general solution of
(20.1) is
(20.4) Yc = (Cl + C2X + cax2 + · · · + cnxn-l)eClz.
216 HIGHER ORDER LINEAR DIFFERENTIAL EQUATIONS Chapter 4.

And if, for example, the characteristic equation (20.14) can be written in
the form
(20.41)

which implies that its roots are

m = 0 twice, m = a three times, m = -b four times, m = -c once,

then the general solution of its related differential equation is
(20.42) Yc = Cl + C2X + (ca + C4X + csx2)eClZ
+ (C6 + C7 X + CSX 2 + C9 X 3) e -bz + CIOe.
-cz

Observe that with each n-fold root p, ePz is mUltiplied by a linear com-
bination of powers of x beginning with XO and ending with x,,-I.
Example 20.43. Find the general solution of

(a) y(4) - 3y" + 2y' = o.

Solution. The characteristic equation of (a) is
(b)
whose roots are
(c) m = 0, m = 1, m=1 , m = -2.

Since the root 1 appears twice, the general solution of (a), by (20.42), is

(d)

Example 20.44. Find the particular solution of

(a) y" - 2y' + y = 0,

for which Yc(O) = 1, y/ (0) = o.

Solution. The characteristic equation of (a) is

(b) m2 - 2m + 1 = 0,

whose roots are m = 1 twice. Hence the general solution of (a), by

(20.38), is
(c) Yc = (Cl + c2x)e z .

Differentiation of (c) gives

(d)
Lesson 20D ROOTS IMAGINARY 217

To find the particular solution for which x = 0, Yc = 1, Yc' = 0, we

substitute these values in (c) and (d). The result is

(e)

The simultaneous solution of (e) for CI and C2 gives CI = 1, C2 = -1.

Substituting these values in (c), we obtain the required particular solution

(f) Y = (1 - x)e z .

LESSON 20D. Some or All Roots of the Characteristic Equation

(20.14) Imaginary. If the constant coefficients in the characteristic
equation (20.14) are real, then (see Exercise 18,9) any imaginary roots it
may have must occur in conjugate pairs. Hence if a +
i{3 is one root,
another root must be a - i{3. Assume now that a +
i{3 and a - i{3 are
two imaginary roots of the characteristic equation of a second order linear
differential equation. Then its general solution by (20.21) is

(20.5) Yc = +
= +
= +
By (18.82) and (18.83)

(20.51) = cos {3x + i sin {3x; = cos {3x - i sin (3x.

Substituting (20.51) in the last equation of (20.5) and simplifying the

result gives

(20.52)

We now replace the constant CI' + C2' by a new constant CI and the con-
stant i(CI' - C2') by a new constant C2. Then (20.52) becomes

(20.53)

which is a second form of the general solution (20.5).

A third form of writing the general solution (20.5), more useful for
practical purposes, is obtained as follows. Write the equality

(20.54) CI cos {3x+ C2 sin {3x

= YCI 2 + C2 2 ( CI cos {3x + C2 sin (3x) •
YC12+C22 V C1 2 +C2 2
218 HIGHER ORDER LINEAR DIFFERENTIAL EQUATIONS Chapter'

Figure 20.55

We see from Fig. 20.55 that

(20.56)

The substitution of these values in the right side of (20.54) gives

(20.57)
Cl cos {3x + C2 sin {3x = V C1 2 + C2 2 (sin 8 cos {3x + cos 8 sin (3x)
= VC12 + C2 2 sin ({3x + 8).
Prove as an exercise that if we had interchanged the positions of Cl
and C2 in Fig. 20.55, equation (20.54) would have become
(20.58)
Cl cos {3x + C2 sin {3x = VC1 2 + C2 2 (cos 8 cos {3x + sin 8 sin (3x)
= VC12 + C2 2 cos ({3x - 8).

Replacing in (20.57) and in (20.58) a new constant C for the constant

v'C1 2 + C22, we may write the general solution (20.53) in either of the
forms

(20.59) Yc = ce
az
sin ({3x + a) or Yc = ceaz cos ({3x - 8).

NOTE. The 8 in the first equation is not the same as the 8 in the second
equation.
Hence in the case of complex roots, the general solution of a linear
differential equation of order two, whose characteristic equation has the
conjugate roots a + {3i and a - {3i, can be written in any of the following
forms:

(20.6) (a) Yc = cle(a+i,,>z + C2e(a-i,,>z,

(b) Yc = eaz(cl cos {3x + C2 sin (3x),
(c) Yc = ceaz sin ({3x + 8),
(d) Yc = ce
az
cos ({3x - 8).

The significance of the two arbitrary constants C and 8 which appear in

(20.6) (c) and (d) will be discussed in Lesson 28.
Lesson 20D ROOTS IMAGINARY 219

Example 20.61. Find the general solution of

(a) y" - 2y' + 2y = O.
Solution. The characteristic equation of (a) is
(b) m2 - 2m +2 = 0,

whose roots are 1 ± i. Hence in (20.6), a = 1, (3 = 1. The general

solution of (a), therefore, can be written in any of the following forms:
(c) Yc = cleU+i>z + C2eU - i>z,
Yc = eZ(cl cos x + C2 sin x),
Yc = ceZ sin (x + 8),
Yc = cez cos (x - 8).

If the linear differential equation

(20.7)
has a conjugate pair of repeated imaginary roots, i.e., if a + i{3 and
a - i{3 each occurs twice as a root, then by (20.38) the general solution
of (20.7) is
(20.71) Yc = (C1 + c2x)e(a+i,,>z + (ca + c4x )e(a-i,,>z.
The equivalent forms are
(20.72) Yc =
ea%[(cl +
C2X) cos {3x + (ea + C4X) sin {3x],
Yc = eaz [c1 sin ({3x 81 )+ + C2X SIn- ({3x + 82)],
Yc = ea%[cl cos ({3x - 81 ) + C2X cos ({3x - 82 )].

Example 20.73. Find the general solution of

(a) y(4) + 2y" + 1 = O.

Solution. The characteristic equation of (a) is

(b) m4 + 2m2 + 1 = 0;
whose roots are m = ±i twice. Hence in (20.71) and (20.72) a = 0,
{3 = 1. The general solution of (a) therefore can be written in any of the
following forms:
(c) Yc = (C1 + c2x )eiz + (ca + c4x )e-iz ,
Yc = (C1 + C2 X) cos x + (ca + C4X) sin x,
Yc = C1 sin (x + 81 ) + C2X sin (x + 82),
Yc = C1 cos (x - 81 ) + C2X cos (x - 82 ).
220 HIGHER ORDER LINEAR DIFFERENTIAL EQUATIONS Chapter 4.

Example 20.74. Find the general solution of

(a) y'" - 12y" + 22y' - 20y = 0
Solution. The characteristic equation of (a) is
(b) mS - 8m 2 + 22m - 20 = 0
whose roots are 2, 3 ± i. The general solution of (a), therefore, can be
written in any of the following forms:
(c) Yc = cle 2z + C2e(S+i)z + cse(S-i)z,
Yc = cle
2z
+ e3z (c2 cos x + Cs sin x),
Yc = cle
2z
+ C2e3z cos (x - 8),
Yc ==- cle 2z + C2e3z sin (x + 8).

EXERCISE 20
Find the general solution of each of the following equations.
I. y" + 2y' = O. S. 6y" - 11y' + 4y = O.
2. y" - 3y'+ 2y = O. 6. y" + 2y' - Y = O.
3. y" - y = O. 7. y'" +y" - lOy' - 6y = O.
4. y'" + y" - 6y' = o. 8. y(4) - y'" - 4y" +
4y' = O.
9. y(4) + 4y'" + y" - 4y' - 2y = o.
10. y(4) - a 2 y = 0, a > O. IS. 3y'" +5y" +
y' - y = O.
II. y" - 2ky' - 2y = o. 16. y'" - 6y" +
12y' - 8y = O.
12. y" + '4ky' - 12k 2 y = O. 17. y" - 2ay' +
a 2 y == O.
13. y(4) = O. 18. y(4) +3y'" = O.
14. y" + +
4y' 4y = O. 19. y(4) - 2y" == O.
20. y(4) + 2y'" - 11y" - 12y' + 36y == O.
21. 36y(4) - 37y" + 4y' + 5y == o. 26. y" - 4y'+ 20y == O.
22. y(4) - 8y" + 16y = O. 27. y(4) + 4y" + 4y == O.
23. y" - 2y' + 5y = O. 28. y'" + 8y == O.
24. y" - y' + y = o. 29. y(4) + 4y" = O.
2S. y(4) + 5y" + 6y = o. 30. yes) + 2y'"+ = o.
y'

For each of the following equations find a particular solution which

satisfies the given initial conditions.
31. y" = 0, y(l) = 2, y'(I) = -1.
32. y" + +
4y' 4y = 0, y(O) = 1, y'(O) = 1.
33. y" - 2y' + 5y = 0, y(O) = 2, y'(O) = 4.
34. y" - 4y' + 20y = 0, y( 1r/2) = 0, y'( 1r/2) = 1.
3S. 3y'" + +
5y" y' - y = 0, y(O) = 0, y'(O) = 1, y"(O) - -1.

ANSWERS 20
1. y = Cl +z c2e-2z2z• 4. y = Cl +z/2 c2e2z +4z/3c3e-3z•
2. y = cle + c2e • S. y = cle + c2e •
3. y = cle z + c2e- z • 6. y = cle(-l+v'2)z + c2e(-1-{2)z.
Lesson 21A METHOD OF UNDETERMINED COEFFICIENTS 221

7. Y = cle3z + + c3e(-2-{2)z.
Cl + c2e + c3e + c4e-2z •
z 2z
B. y =
9. Y = cle z + c2e- z + +
10. Y = cle{;,z + c2e-..fO.z + C3 cos va x + C4 sin va x.
II. Y = cle(Ic+v'k 2 +2)z + c2e(Ic-.Jk2 + 2 )z.

12. Y = cle-
6kz
+ c2e2kz • 16. Y = (Cl + C2X + c3 x2 )e2z •
Cl + C2X + C3x2 + C4X • 17. Y = (Cl + c2 x )eGz •
3
13. Y =
14. Y = (Cl + c2 x )e- 2z • lB. Y = Cl + C2X + C3x2 + c4e- 3z •
15. Y = cle z/3 + (C2 + c3 x )e-z • 19. Y = Cl + C2X+ +
(Cl + c2 x )e + (C3 + c4x)e- 3z •
2z
20. Y =
21. Y = cle- z + c2e z/2 + c3e Sz/6 + c4e- z/3 •
(Cl + c2 x )e + (C3 + c4x)e- •
2z 2z
22. Y =
23. Y = eZ(cl cos 2x + C2 sin 2x).
24. Y = Cle
l+.J3i Z
2 + C2e l-.J3i
2
Z
= e
z/2 (
C3 cos
va
2 x + C4 SIn. va )
2 x •
25. Y = Cl cos va X + C2 sin va x + C3 cos v2 x + C4 sin v2 x.
26. Y = e2z (cl cos 4x + C2 sin 4x).
27. Y = (Cl + C2X) cos v2 x + (C3 + C4X) sin v2 x.
2B. Y = cle- 2z + e [c2 cos va x + C3 sin va xJ.
Z

29. Y = Cl + C2X + C3 cos 2x + C4 sin 2x.

30. Y = Cl + C2 sin x + C3 cos X + C4X sin x + CsX cos x.
31. Y = 3 - x.
32. Y = (1 + 3x)e-2z •
33. Y = e (2 cos 2x + sin 2x).
Z

34. Y = !e2z - r
sin 4x.
9 z/3
35. Y = 16 e
+ (x"4 - 9)
16 e •
-z

LESSON 21. Solution of the Nonhomogeneous Linear

Differential Equation of Order n with
Constant Coefficients.

LESSON 21A. Solution by the Method of Undetermined Coeffi-

cients. By Theorem 19.3 and Comment 19.41, the-general solution of
the differential equation
(21.1)

where an '¢ 0 and Q(x) ¢ 0 in an interval I, is

(21.11) y(x) = Yc(x) + Yp(x),
where yc(x), the complementary function, is the general solution of the
related homogeneous equation of (21.1) and yp(x) is a particular solution
222 HIGHER ORDER LINEAR DIFFERENTIAL EQUATIONS Chapter'

of (21.1). In Lesson 20, we showed how to find Yc. There remains the
problem of finding Yp.
The procedure we are about to describe for finding YP is called the
method of undetermined coefficients. It can be used only if Q(x)
consists of a sum of terms each of which has a finite number of
independent derivatives. This restriction implies that Q(x) can only con-
tain terms such as a, x1r., eo,x, sin ax, cos ax, and combinations ofsuch
terms, where a is a constant and k is a positive integer. See Exercise 21,2.
For example, the successive derivatives of sin 2x are
2 cos 2x, -4 sin 2x, -8 cos 2x, etc.
However, only the set consisting of sin 2x and 2 cos 2x is linearly inde-
pendent. The addition of any succeeding derivative makes the set linearly
dependent. Verify it. The linearly independent derivatives-of x 3 are
3x 2 , 6x, 6.
The addition to this set of the next derivative, which is zero, makes the
set linearly dependent. See Comment 19.14. However the function X-I,
for example, has an infinite number of linearly independent derivatives.
To find YP by the method of undetermined coefficients, it is necessary
to compare the terms of Q(x) in (21.1) with those of the complementary
function Yc. In making this comparison, a number of different possibilities
may occur, each of which we consider separately in the cases below.
Case 1. No term of Q(x) in (21.1) is the same as a term oj Yc. In this
case, a particular solution YP of (21.1) will be a linear combination of the
terms in Q(x) and all its linearly independent derivatives.
Example 21.2. Find the general solution of
(a) y" + 4y' + 4y = 4x 2 + 6ex •
The complementary function of (a) is
(b)
(Verify it.) Since Q(x), which is the right side of (a), has no term in common
with Yc this case applies. A particular solution YP will therefore be a
linear combination of Q(x) and all its linearly independent derivatives.
These are, ignoring constant coefficients, x 2 , x, 1, eX. Hence the trial
solution YP must be a linear combination of these functions, namely
(c) yp = AX2 + Bx + C + Dex ,
where A, B, C, D are to be determined. Successive derivatives of (c) are
(d) YP' = 2Ax + B + De x,
(e) YP" = 2A + De x •
Lesson 21A METHOD OF UNDETERMINED COEFFICIENTS 223

As we have repeatedly remarked, (c) will be a solution of (a) if the sub-

stitution of (c), (d), and (e) in (a) will make it an identity in x. Hence
(c) will be a solution of (a) if

(f) 2A + Dez + 4(2Ax + B + De Z

)

+ 4(Ax 2 + Bx + C + DeZ ) = 4x 2 + 6e Z
•

Simplification of (f) gives

(g) 4Ax2 + (SA + 4B)x + (2A + 4B + 4C) + 9Dez = 4x + 6ez • 2

We now ask ourselves the question: What values shall we assign to A, B,

C, and D to make (g) an identity in x? We proved in Example 19.13,
that x, x 2 are linearly independent functions. The proof can be extended
to show that xo, x, x 2 are also linearly independent functions. Hence the
answer to our question is: values which will make each coefficient of like
powers of x zero. This means that the following equalities must hold.

(h) 4A = 4 ,
SA + 4B = 0,
2A + 4B + 4C = 0,
9D = 6.
Solving (h) simultaneously, there results

(i) A=1 , - -2 ,
B- C = t, D = i.
Substituting these values in (c), we obtain

(j)

which is a particular solution of (a). Hence by (21.11), the general solu-

tion of (a) is (b) +
(j), namely,
(k)
Example 21,.21. Find the general solution of

(a) y" - 3y' + 2y = 2xe 3z + 3 sin x.

Solution. The complementary function of (a) is

(b)

(Verify it.) Since Q(x), which is the right side of (a), has no term in com-
mon with Yc, a particular solution YP will be a linear combination of Q(x)
and all its linearly independent derivatives. These are, ignoring constant
224 HIGHER ORDER LINEAR DIFFERENTIAL EQUATIONS Chapter 4

coefficients, xe 3x , e3x, sin x, cos x. Therefore the trial solution YP must be

of the form
(c) YP = Axe 3x + Be 3x +
C sin x +
D cos x.
Successive derivatives of (c) are

(d) yp' = 3Axe 3x + Ae 3x + 3Be 3x + C cos x - D sin x,

(e) YP" = 9Axe 3x + 6Ae 3x + 9Be 3x - C sin x - D cos x.
The function defined by (c) will be a solution of (a) if the substitution of
(c), (d), and (e) in (a) will result in an identity in x. Making these sub-
stitutions and simplifying the resulting expression, we obtain

(f) 2Axe 3x + (3A + 2B)e 3x + (C + 3D) sin x

+ (D - 3C) cos x = 2xe 3x + 3 sin x.
Equation (f) will be an identity in x if the coefficients of like terms on
each side of the equal sign have the same value. Hence we must have
(g) 2A = 2,
3A + 2B = 0,
C + 3D = 3,
-3C + D = O.
Solving (g) simultaneously, there results

(h) A=1 , B = -I, C = to, D -- yo.

9

Substituting these values in (c), we obtain

(i) + to- sin x + to- cos x.

yp = xe 3x - !e 3x
Hence by (21.11) the general solution of (a) is (b) + (i), namely

Case 2. Q(x) in (21.1) contains a term which, ignoring constant coeffi-

cients, is Xk times a term u(x) of Yc, where k is zero or a positive integer.
In this case a particular solution YP of (21.1) will be a linear combination
of Xk+1u(x) and all its linearly independent derivatives (ignoring constant
coefficients). If in addition Q(x) contains terms which belong to Case 1,
then the proper terms called for by this case must be included in Yp.
Example 21.3. Find the general solution of

(a) y" - 3y' + 2y = 2X2 + 3e 2x.

Lesson 2lA METHOD OF UNDETERMINED COEFFICIENTS 225

Solution. The complementary function of (a) is

(b)

(Verify it.) Comparing Q(x), which is the right side of (a), with (b), we
see that Q(x) contains the term e2z which, ignoring constant coefficients,
is X O times the same term in Ye. Hence for this term, YP must contain a
linear combination of xO+le 2z and all its linearly independent derivatives.
Q(x) also has the term x 2 which belongs to Case 1. For this term, there-
fore, yp must include a linear combination of it and all its linearly inde-
pendent derivatives. In forming the linear combination of these functions
and their linearly independent derivatives, we may omit the function e2z
since it already appears in Ye; see Exercise 21,1. Hence the trial solution
YP must be of the form

(c) YP = AX2 + Bx + C + Dxe 2z .

Successive derivatives of (c) are

(d) YP' = 2Ax + B + 2Dxe 2z + De 2z ,

(e) YP" = 2A + 4Dxe 2z + 4De 2z .

Substituting (c), (d), (e) in (a) and simplifying, we see that (c) will be a
solution of (a) if

(f) 2Ax2 + (2B - 6A)x + (2A - 3B + 2C) + De 2z = 2X2 + 3e 2z .

Equation (f) will be an identity in x if the coefficients of like terms on
each side of the equal sign have the same value. Hence we must have

(g) 2A = 2, 2B - 6A = 0, 2A - 3B + 2C = 0, D = 3.

From (g) we find

(h) A = 1, B = 3, C = t, D = 3.

Substituting these values in (c), there results

(i)

Combining this solution with (b), we obtain for the general solution of (a)

{j)

Example 21.31. Find a general solution of

(a) Y" - 3y' + 2y = xe 2z + sin x.

226 HIGHER ORDER LINEAR DIFFERENTIAL EQUATIONS Chapter 4.

Solution. The complementary function of (a) is

(b) Yc = cle
x + C2 e2x •
Comparing Q(x) which is the right side of (a), with (b), we see that Q(x)
contains a term xe 2x which, ignoring constant coefficients, is x times a
term e2x in Yc. For this term, therefore, yp must be a linear combination
of xl+1e 2x = x 2e2x and all its independent derivatives. In addition we
note Q(x) contains a term sin x which belongs to Case 1. For this term,
therefore, YP must include a linear combination of it and its independent
derivatives. In forming the linear combination of all these functions and
their independent derivatives, we may omit the function e2x since it
already appears in Yc. Hence YP must be of the form
(c)

Successive derivatives of (c) are

(d) yp' = 2Ax 2e2x + 2Axe + 2Bxe + Be + C cos x
2x 2x 2x
- D sin x,
(e) yp" = 4Ax 2e2x + 8Axe 2x + 2Ae 2x + 4Bxe 2x + 4Be 2x
- C sin x - D cos x.
Substituting (c), (d), (e) in (a) and simplifying the result, we see that (c)
will be a solution of (a) if
(f) 2Axe 2x + (2A + B)e 2x + (C + 3D) sin x
+ (D - 3C) cos x = xe 2x
+ sin x.
Equating the coefficients of like terms on each side of the equal sign, we
find that
(g) 2A = 1, 2A +B = 0, C + 3D = 1, D - 3C = o.
From (g), we obtain
(h) A = I, B =-1 , C= to, D = fu-.
Substituting these values in (c), there results
(i)

The general solution of (a) is therefore the sum of (b) and (i).
Example 21.32. Find a general solution of
(a) y" +
y = sin a x.

Solution. The complementary function of (a) is

(b) yc = Cl sin x + C2 cos x.
Lesson 21A METHOD OF UNDETERMINED COEFFICIENTS 227

By (IS.S4),
• 3 _ (e iZ - e- iZ )3 _ e3iz _ e- 3iz 3(e iz _ e- iz )
(c) SIn x - 2I. - -
SI. + SI.

= -! sin 3x + t sin x.
Q(x), which is the right side of (c), with (b), we see that Q(x)
contains a term, which, ignoring constant coefficients, is XO times a term
sin x in Yc. Hence the trial solution yp must be of the form

(d) YP = A sin 3x + B cos 3x + Cx sin x + Dx cos x.

Successive derivatives of (d) are

(e) yp' = 3A cos 3x - 3B sin 3x + Cx cos x + C sin x

- Dx sin x + D cos x.
yp" = -9A sin 3x - 9B cos 3x - Cx sin x + 2C cos x
- Dx cos x - 2D sin x.

Substituting (c), (d), (e) in (a), we obtain

(f) -SA sin 3x - SB cos 3x + 2C cos x - 2D sin x

= -1 sin 3x + t sin x.
Equating coefficients of like terms on each side of the equal sign, there
results
(g) -SA = -1, B = 0, C=o , D= -I.
From (g), we have

(h) A = n, D = -I.
Substituting these values in (d), we obtain
(i) YP = n sin 3x - Ix cos x.

A general solution of (a) is therefore the sum of (b) and (i).

Case 3. This case is applicable only if both of the following condi-
tions are fulfilled.
A. The characteristic equation of the given differential equation (21.1)
has an r multiple root.
B. Q(x) contains a term which, ignoring constant coefficients, is times x"
a term u(x) in Yc, where u(x) was obtained from the r multiple root.
In this case, a particular solution YP will be a linear combination of
xt+ru(x) and all its linearly independent derivatives. If in addition Q(x)
228 HIGHER ORDER LINEAR DIFFERENTIAL EQUATIONS Chapter"

contains terms which belong to Cases 1 and 2, then the proper terms
called for by these cases must also be added to Yp.
Example 21.4. Find the general solution of
(a) y" + 4y' + 4y = 3xe- 2z •
Solution. The complementary function of (a) is
(b)

We observe first that the characteristic equation of (a), namely m 2 +

4m +
4 = 0, has a multiple root, m = -2. Secondly we observe that
Q(x) which is the right side of (a), contains the term xe- 2z which is x
times the term e- 2z in Ye (or alternately xe- 2z of Q(x) is XO times the term
xe- 2z in Ye), and that this term in Ye came from a multiple root. Hence,
by the above remarks under Case 3, T = 2, k = 1, and T +
k = 3, (or
alternately T = 2, k = 0 and T
3 2
+ k = 2). Therefore, YP must be a linear
combination of x e- ,. and all its linearly independent derivatives [or
alternately x 2 (xe- 2z ), which yields the same x 3e- 2z , and its derivatives].
In forming this linear combination, we may omit the functions e- 2z and
xe- 2z since they already appear in Ye. Hence YP must be of the form
(c)

The successive derivatives of (c) are

(d) YP' = -2Ax 3e- 2z + 3Ax 2e- 2z - 2Bx 2e- 2z + 2Bxe- 2z ,
(e) yp" = 4Ax 3e- 2z - 12Ax 2e- 2z + 6Axe- + 4Bx e-
2z 2 2z

- 8Bxe- 2z +
2Be- 2z •

Substituting (c), (d), (e) in (a) and simplifying the result, we see that (c)
will be a solution of (a) if
/
(f)

Equating coefficients of like terms on each side of the equal Signt there
results
(g) A = i, B = o.
Substituting these values in (c), we obtain
(h)

The general solution of (a) is, therefore, the sum of (b) and (h), namely

(i)
Lesson 2lA METHOD OF UNDETERMINED COEFFICIENTS 229

Example 21.41. Find the general solution of

(a) Y" + 4y' + 4y = 3e- 2x •
Solution. The complementary function of (a) is
(b)

The characteristic equation of (a), namely m 2 + 4m + 4 = 0, has a

multiple root m = -2. The function Q(x), which is the right side of
(a), contains the term e- 2x which is xO times a term in Ye (or alternately
2x
X-I times the term xe- in Ye). Since this term in Ye came from a mul-
tiple root, this Case 3 applies. Hence by the remarks under Case 3,
r = 2, k = 0, and r + k = 2 (or alternately r = 2, k = -1 and
r+ k = 1). Therefore YP must be a linear combination of x 2e- 2x and
all its linearly independent derivatives [or alternately x(xe- 2x ) , which
yields the same x 2e- 2x ]. In forming this linear combination, we may
omit the terms e- 2x and xe- 2x since they already appear in Ye. Hence YP
must be of the form
(c) YP = Ax 2e- 2x , YP' = 2Axe- 2x - 2Ax 2e- 2x ,
YP" = 2Ae- 2x - 8Axe- 2x + 4Ax 2e- 2x •
Substituting (c) in (a) and simplifying the result, we see that YP will be a
solution of (a) if
(d)
Substituting this value of A in the first equation of (c), we obtain
(e) - 3 x 2e- 2x .
YP-2-
The general solution of (a) is therefore the sum of (b) and (e).
Comment 21.42. The function which we have labeled YP' since it does
not contain arbitrary constants, has been correctly called, by our Defini-
tion 4.66, a particular solution of (21.1). There are, of course, infinitely
many other particular solutions of the differential equation, one for each
set of values of the arbitrary constants in the Ye part of the general solu-
tion Y = Ye + Yp. These constants are determined in the usual way by
inserting the initial conditions in the general solution y. Do not confuse,
therefore, a particular solution obtained by the method of undetermined
coefficients and the particular solution which will satisfy given initial
conditions. See example below.
Example 21.43. Find the particular solution of
(a) Y" - 3y' + 2y = 6e- x
for which yeO) = 1, y'(O) = 2.
230 HIGHER ORDER LINEAR DIFFERENTIAL EQUATIONS Chapter 4.
.
Solution. By the methods outlined previously, verify that
(b)

However this YP is not the particular solution which satisfies the initial
conditions. To find it, we must first write the general solution, and then
substitute the initial conditions in it and in its derivative. The general
solution of (a) and its derivative are, by (b),
(c)

Substituting in (c) the initial conditions x = 0, y = 1, y' = 2, we obtain

(d) 1 = Cl + + 1,
C2

2 = Cl +2C2 -1,

whose solutions are Cl = -3, C2 = 3. Hence the particular solution of

(a) which satisfies the given initial conditions is, by (c) and these values
of CI, C2,
(e)

LESSON 21B. Solution by the Use of Complex Variables. There

is another way of solving certain types of nonhomogeneous linear equa-
tions with constant coefficients. If in
(21.5)
the a's are real, Q(x) a complex-valued function (i.e., a function which
can take on complex values) and yp(x) is a solution of (21.5), then (see
Exercise 19,7):
1. The real part of YP is a solution of (21.5) with Q(x) replaced by its
real part.
2. The imaginary part of YP is a solution of (21.5) with Q(x) replaced by
its imaginary part.
Remark. Statements 1 and 2 above would still be valid if the coeffi-
cients in (21.5) were real, continuous functions of x instead of constants.
Example 21.51. Find a particular solution of
(a) y" - 3y' + 2y = sin x.
Solution. Instead of solving (a), let us solve the differential equation
(b) y" - 3y' + 2y = eiz .
By (18.82), eiz = cos x + i sin x. Its imaginary part is, therefore, sin x.
Hence by 2 above, the imaginary part of a particular solution YP of (b)
Le8son 21-Exercise 231
will be a solution of (a). A particular solution of (b), using the method of
undetermined coefficients is [take for the trial solution YP = (A + Bi)eiz ],
(c) YP = -frIeiz + foie iz
= + i sin x) + (cos x + i sin x)
rlr(cos x

= to cos x - to- sin x + i(fisin x + to cos x).

The imaginary part of the solution yp is -h sin x + to- cos x. Hence a

particular solution of (a) is
(d) YP = i7I sin x + to- cos x.
Question. What would a particular solution of (a) be, if in it sin x
were replaced by cos x? [Ans. The real part of yp, namely yp = -h cos x
- to sin x.]
EXERCISE 21
I. Prove that any term which is in the complementary function Yc need not
be included in the trial solution yp. (Hint. Show that the coefficients of
this term will always add to zero.)
2. Prove that if F(x) is a function with a finite number of linearly independent
derivatives, i.e., if F(n)(x), F(n-l)(x), · · · , F'(x), F(x) are linearly independ-
ent functions, where n is a finite number, then F(x) consists only of such terms
as a, x lc , eGz , sin ax, cos ax, and combinations of such terms, where a is a
constant and k is a positive integer. Hint. Set the linear combination of these
functions equal to zero, i.e., set
CnF(n)(x) + Cn_1F(n-l)(x) + · · · + ClF'(x) + CoF(x) = 0,
where the C's are not all zero, and then show, by Lesson 20, that the only
functions F(x) that can satisfy this equation are those stated.
Find the general solution of each of the following equations.
3. y" + 3y' + 2y = 4. z 10. y" - 2y' - 8y = 9xez +
10e-z •
4. y" + 3y' + 2y = 12e • 11. y" - 3y' = 2e 2z sin x.
5. y" + 3y' + 2y = e • iz 12. y(4) - 2y" + y = x - sin x.
6. y" + 3y' + 2y = sin x. 13. y" + y' = x +
2 2x.
7. y" + 3y' + 2y = cos x. 14. y" + y' = x+ sin 2x.
8. y" + 3y' + 2y = 8+ 6e + 2 sin x.
z
15. y" + y = 4x sin x.
9. y" + y' + y = x 2 • 16. y" + 4y = x sin 2x.
17. y" + + y = x 2e-
2y' z
•
18. y'"+ + 3y' + y2z=
3y" 2e- z - x 2 e- z •
19. y" + 3y' + 2y = e- + x 2• 21. y" + y' - 6y = x + e2z •
20. y" - 3y' + 2y = xe- Z
• 22. y" +y = sin x+ e- •
Z

23. y'" - 3y" + 3y' - y = e

Z
•

24. y" +
y = sin x. Hint. sin 2 x = I - I cos 2x.
2
25. y'" - y' = e2z sin 2 x.
26. yes) +
2y'" +
y' = 2x sin x + +
cos x. Hint. Solve yes) 2y'" + + y' =
2x +iz
e ; see Example 21.51.
27. y" +
y = sin 2x sin x. Hint. sin 2x sin x = I cos x - I cos ax.
232 HIGHER ORDER LINEAR DIFFERENTIAL EQUATIONS Chapter"

For each of the following equations, find a particular solution whi'Ch

satisfies the given initial conditions.
28. y" - 5y' - 6y = e3z, yeO) = 2, y'(O) = 1.
29. y" - y' - 2y = 5 sin x, yeO) = 1, y' (0) = -l.
30. y'" - 2y" + y' = 2e z + 2x, yeO) = 0, y' (0) = 0, y" (0) = o.
31. y" + 9y = 8 cos x, y(r/2) = -1, y'(r/2) = 1.
32. y" - 5y' + 6y = eZ (2x - 3), yeO) = 1, y'(O) = 3.
33. y" - 3y' + 2y = e- z, yeO) = 1, y'(O) = -1.

ANSWERS 21
3. y = cle- 2z + C2e- z + 2.
4. y = cle- 2z + c2e-zz + 2e iz Z
•

5. y = cle- 2z + c2e- z + to(e - 3ieiz ).

6. y = cle-2z + c2e- + toCsin x - 3 cos x).
2z + c2e-z + to(3 sin x + cos x).
7. y = cle-
8. y = cle- 2z + c2e- z + 4 + e + !(sin x - 3 cos x).
Z

9. y = e-z/2 ( CI cos va
2 X + C2 SIn . va
2 x) + x 2 - 2x.
10. y = cle4z + c2e- 2z - xe Z
2e- z.
2z
-

n. Cl + c2e (3 sin x + cos x).

3z
y = - e
5
12. y = (CI + c2x)e
z+ (C3 + C4X)e-z + x - -4-·
SIn x
3
13. Y = Cl + c2e- + .
z

2
14. y = CI + c2e- + x2 - x - (2 sin 2x + cos 2x).
z

15. y = Cl cos X + C2 sin x - x(x cos x - sin x).

16. y = CI cos 2x + C2 sin 2x - :6 (2x cos 2x - sin 2x).
4 -z
17. y = Cle -z+ C2xe -z+ x e
12·
3 -z
18. y = Cle -z + C2xe -z + C3X 2e-z + X 60
e (20 - 2)
x ·
2
19. y = Cle
-2z + C2e -z + 2x - 23x + "47 - xe -2z•
20. y = cle 2z + c2e z + is(6xe- + 5e- Z Z

2z
).

21. y = Cle
-3z + 2z X
C2e -"6 - 36
1 + xe
-5- .

22. y =
. x
CI cos X + C2 SIn x - 2 cos x
+1-z2e ·
23. Y = (Cl + C2X + C3x2 + x;) e Z

. 1 cos 2x
24. y = CI cos X + C2 SIn x + 2 + 6 .

25• Y -_ CI
+ C2e z+ C3e-z + (..!.
12
+ 9 cos 2x 520
- 7 sin 2X) 2z
e •
Lesson 22B THE METHOD OF VABlATION OF PARAMETERS 233

26. Y = Cl + C2 sin x + Ca cos x + C4X sin x + csX cos x + x2

2
+ X8 (cos x - sin x).
27• y .=Cl cos X + . +
C2 SIn x 'x. ax
4 SIn x cos16·+
28. y = He 6z + He- z - ne3z.
29. y = le 2z + te-Z - i sin x + i cos x.
30. y = x2 + + +
4x 4 (x 2 - 4)ez •
31. y = cos x + +
i cos ax sin ax.
32. y = e + xe Z •
2z

33. 6y = -10e2z + + 15ez e- Z •

LESSON 22. Solution of the Nonhomogeneous Linear

Differential Equation by the Method of
Variation of Parameters.

LESSON 22A. In the previous lessons of

Introductory Remarks.
this chapter, we showed how to solve the linear differential equation
(22.1) any(n) + an_ly(n-O + · · · + alY' + aoy = Q(x), an 0,

where:
1. The coefficients are constants.
2. Q(x) is a function which has a finite number of linearly independent
derivatives.
You may be wondering whether either or both of these restrictions may
be removed.
In regard to the first restriction there are very few types of linear equa-
tions with nonconstant coefficients whose solutions can be expressed in
terms of elementary functions and for which standard methods of obtain-
ing them, if they do exist, are available. In Lesson 23, we 'shall describe
a method by which a general solution of a second order linear differential
equation with nonconstant coefficients can be found provided one solution
is known. Again therefore, the equation must be of a special type so that
the needed one solution can be discovered.
As for the second restriction, it is possible to solve (22.1) even when
Q(x) has an infinite number of linearly independent derivatives. The
method used is known by the name of "variation of parameters" and is
discussed below.

LESSON 22B. For con-

The Method of Variation of Parameters.
venience and clarity, we restrict our attention to the second order linear
equation with constant coefficients,

(22.2)
234 HIGHER ORDER LINEAR DIFFERENTIAL EQUATIONS Chapter 4.

where Q(x) is a continuous function of x on an interval I and is ;EO on I.

If the two linearly independent solutions of the related homogeneous
equation
(22.21)

are known, then it is possible to find a particular solution of (22.2) by a

method called variation of parameters, even when Q(x) contains terms
whose linearly independent derivatives are infinite in number. In de-
scribing this method, we assume therefore, that you would have no trouble
in finding the two linearly independent solutions Yl and Y2 of (22.21).
With them we form the equation
(22.22)

where Ul and U2 are unknown functions of x which are to be determined.

The successive derivatives of (22.22) are

(22.23) YP' = + Ul'Yl + U2'Y2 + U2Y2'

UlYl'
= (UIYl' + U2Y2') + (Ul'Yl + U2'Y2),
(22.24) YP" = (UIYl" + U2Y2") + (Ul'Yl' + U2'Y2') + (Ul'Yl + U2'Y2)'.

Substituting the above values of YP' YP', and YP" in (22.2), we see that Y'P
will be a solution of (22.2) if

(22.25) ( " + U2Y2 ") + a2 (Ut ,Yl , + U2 'Y2 ')

a2 UlYl
+ a2(ul'Yl + U2'Y2)' + al (UlYl' + U2Y2')
+ al(ul'Yl + U2'Y2) + aO(ulYl + U2Y2) = Q(x).

This equation can be written as

(22.26) ul(a2Yl" + alYl' + aOYl) + u2(a2Y2" + atY2' + aOY2)

+ a2(ul'Yl' + U2'Y2') + a2(ul'Yl + U2'Y2)'
+ al (Ul'Yl + U2'Y2) = Q(x).
Since Yl and Y2 are assumed to be solutions of (22.21), the quantities in
the first two parentheses in equal zero. The remaining three terms
will equal Q(x) if we choose Ul and U2 such that
(22.27) Ut'Yt + U2'Y2 = 0,

U 1'Y t , + 'Y 2
U2 Q(x)
, _- - -·
a2
The pair of equations in (22.27) can be solved for Ut' and U2' in terms of
the other functions by the ordinary algebraic methods with which you
Lesson 22B THE METHOD OF VARIATION OF PARAMETERS 235

are familiar. Or, if you are acquainted with determinants (see Lessons 31
and 63), the solutions of (22.27) are
0 Y2 Yl 0
Q(x) Q(x)
Y2' Yl'
a2 a2
(22.28) u'
1 -- u'
2 --
Yl Y2 Yl Y2
Yl' Y2' Yl' Y2'

These equations (22.28) will always give solutions for Ul' and U2' provided
the denominator determinant o. We shall prove in Lesson 64 that, if
Yl and Y2 are linearly independent solutions of (22.21), then this de-
nominator is never zero.
Integration of (22.28) will enable us to determine Ul and U2. The sub-
stitution of these values in (22.22) will give a particular solution yp of
(22.2).
Comment 22.29. Since we seek a particular solution YP' constants of
integration may be omitted when integrating Ul' and U2'.
Comment 22.291. If the nonhomogeneous linear differential equation
is of order n > 2, then it can be shown that

(22.3)

will be a particular solution of the equation, where Yl, Y2, · · · , Yn are the
n independent solutions of its related homogeneous equation, and Ul',
U2', ••• , un' are the functions obtained by solving simultaneously the
following set of equations:

(22.31) Ul'Yl + U2'Y2 + · · . + un'Yn = 0,

Ul'Yl' + U2'Y2' + · · · + Un'Yn' = 0,
..............................
U1'Y 1(n-1) + U2'Y 2 (n-l) + ••• + Un 'Y n (n-1) -_ Q(x)

In (22.31), an is the coefficient of yen) in the given differential equation.

Again we remark that since we seek a particular solution YP, arbitrary
constants may be omitted when integrating Ul', U2', ... , un' to find
Ul, U2, • · · , Un·
Comment 22.32. The method of variation of parameters can also be
used when Q(x) has a finite number of linearly independent derivatives.
In the above description of this method, the only requirement placed on
Q(x) is that it be a continuous function of x. You will find however, that
236 HIGHER ORDER LINEAR DIFFERENTIAL EQUATIONS Chapter"

if Q(x) has a finite number of linearly independent derivatives, the method

of undetermined coefficients explained in Lesson 21 will usually be easier
to use. To show you, however, that the method of variation of parameters
will also work in this case, we have included in the examples below one
which was solved previously by the method of undetermined coefficients.
Comment 22.33. The proof we gave above to arrive at (22.27) would
also have been valid if the constant coefficients in (22.2) and (22.21) were
replaced by continuous functions of x. The method of variation of pa-
rameters can be used, therefore, to find a particular solution of the equation
(22.34)
provided we know two independent solutions Yt and Y2 of the related
homogeneous equation
(22.35) f2(X)Y" + ft(x)y' + fo(x)y = o.
Example 22.4. Find the general solution of
(a) Y" - 3y' + 2y = sin e- X.
(N OTE. This equation cannot be solved by the method of undetermined
coefficients explained in Lesson 21. Here Q(x) = sin e-z , which has an
infinite number of linearly independent derivatives.)
Solution. The roots of the characteristic equation of (a) are m = 1,
m = 2. Hence the complementary function of (a) is
(b)

The two linearly independent solutions of the related homogeneous equa-

tion of (a) are therefore
(c)

Substituting these values and their derivatives in (22.27), we obtain, with

a2 = 1, (remember a2 is the coefficient of Y")

(d) +
Ut'e Z u2'e 2z = 0,
+
ut'e Z u2'(2e 2Z ) = sin e-Z.
Solving (d) for Ut' and U2', there results
(e) Ut' = -e-Z sin e-z ,
Therefore

(f) Ul = f sin e-X(_e-Z) dx, U2 = -

f · e-Z( -e -Z) dx.
e- zSIn
Lesson 22B THE METHOD OF VARIATION OF PARAMETERS 237

Hence (in the integrands, let u = e- x, du = -e- X dx)

(g) Ul = -cos e-x , U2 = -sin e- X + e- X cos e- X.

Substituting (c) and (g) in (22.22), we obtain
(h) YP = -(cos e-X)eX + (e- X cos e- X - sin e-x)e 2x
= _e 2x sin e- x.

Combining (b) and (h) gives

(i)
which is the general solution of (a).
Example 22.41. Find the general 'solution of
(a) y" + 4y' + 4y = 3xe- 2x .

(NOTE. We have already solved this example by the method of undeter-

mined coefficients. See Example 21.4.)
Solution. The complementary function of (a) is
(b) Yc = cle- 2x + C2xe- 2x .
Therefore the two independent solutions of the related homogeneous
equation of (a) are
(c) Y 1 -- e- 2x ,
Substituting these values and their derivatives in (22.27), we obtain
(d) ul'e- 2x +
u2'(xe- 2X ) = 0,
ul'(-2e- 2x ) +
u2'(-2xe- 2x e- 2x ) + = 3xe- 2x .

Solving (d) for Ul' and U2', there results

(e)
Hence,
(f)
Substituting (c) and (f) in (22.22), we have

(g) YP = _x 3e- 2x + {xe- 2X ) = ie- 2Xx 3 •

Combining (b) and (g) we obtain for the general solution of (a)
(h)
which is the same as the one we found previously in Example 21.4.
238 HIGHER ORDER LINEAR DIFFERENTIAL EQUATIONS Chapter'

Remark. The method of variation of parameters has one advantage

over the method of undetermined coefficients. In the variation of param-
eter method, there is no need to concern oneself with the different cases
encountered in the method of undetermined coefficients. In the above
example, the characteristic equation has a repeated root and Q(x) con-
tains a term xe- 2z , which is x times the term e- 2z of Ye.
Example 22.42. Find the general solution of

(a) y" +y = tan x,

'Tr
2
1('
--<x<_·
2
(N OTE. This equation cannot be solved by the method of Lesson 21.
Here Q(x) = tan x which has an infinite number of linearly independent
deriva tives.)
Solution. The complementary function of (a) is

(b) Ye = Ct cos X + C2 sin x.

The two independent solutions of the related homogeneous equation of
(a) are therefore

(c) Yt = cos x,
.
Y2 = sin x.

Substituting these values and their derivatives in (22.27), we obtain

(d) cos x + U2' sin x = 0,

Ut'

Ut'(-sin x) + U2 ' cos X = tan x.

Solving (d) for Ul ' and U2', there results

sin 2 x 1 - cos 2
(e) Ul
I
- -
cos x
-
cos x
X
- -sec x + cos X, U2' = sin x.

Hence

(f) Ut = -log (sec x + tan x) + sin x, U2 = -cos x.

Substituting (c) and (f) in (22.22), we have

(g) yp = -cos x log (sec x + tan x) + sin x cos x - sin X COB x
'Tr 1('
- -cos x log (sec x + tan x), - 2 < x < 2 ·

Combining (b) and (g), we obtain for the general solution of (a),
(h)
'Tr 1('
Y= Cl cos X + C2 sin x - cos x log (sec x + tan x), --<x<_·
2 2
Lesson 22B THE METHOD OF VARIATION OF PARAMETERS 239

In Fig. 22.43, we have plotted a graph, using polar coordinates, with x

as the polar angle and y as the radius vector, of a particular solution of
(a) obtained by setting Cl = C2 = 0 in (h).

Figure 22.43

Example 22.44. If

(a) Yl = x and Y2 = X-I

are two solutions of the differential equation

(b) x 2 y" + xy' - y = 0,

find the general solution of

(c) x 2 y" + xy' - y = x, x o.

Solution. (See Comment 22.33.) Substituting the given two solu-
tions and their derivatives in (22.27), it becomes

(d)

Solutions of (d) for Ul' and U2', by any method you wish to choose, are
, 1 x
(e) Ul =-,
2x
U 2' - - -.
2
Hence
1
(f) Ul = 2 log x, - -.
4
Substituting (a) and (f) in (22.22) gives
x x
(g) yp = 2 log x - 4·
240 HIGHER ORDER LINEAR DIFFERENTIAL EQUATIONS Chapter 4.

Combining (g) with (a), we obtain for the general solution of (c),

(h) y = c} IX + C2X-1 + ; log x - : I

which simplifies to
(i)

EXERCISE 22

Use the method of variation of parameters to find the general solution

of each of the following equations.
1. y" +
y = sec x. 9. y" + 2y' + y = e-Z log x.
2. y" +
y = cot x. 10. y" + y = csc 2x.
3. y" +
y = sec 3 x. II. y" + y = tan x.
4. y" - y = sin 2 x. 12. y" + 2y' + y = e-z/x.
5. y" +
y = sin 2 x. 13. y" + y = sec x csc x.
6. y" + 3y' + 2y = 12e z. 14. y" - 2y' + y = log x.
7. y" + 2y' +
y = x 2e- z. 15. y" - 3y' + 2y == cos
8. y" +
y = 4x sin x.

Use the method of variation of parameters to find the general solution

of each of the following equations. Solutions for the related homogeneous
equation are shown alongside each equation.
16. x 2y" - xy' +
y = x; YI = x, Y2 = x log x.

17. y" - y'z

2 2 2
+
x 2 Y == x log x; Yl = x, Y2 = X •
18• y + x y' - 4y = x;
x 2" 3 YI = X 2 ,Y2 = x12 .
19. x 2y" + xy' - y = x 2e- z ; YI = x, Y2 = x-I.
20. 2x2y" + 3xy' - y = x-I; YI = x1l2, Y2 == x-I.

ANSWERS 22
I. y = CI cos X +
C2 sin x +
x sin x +
cos x log cos x.
2. y = CI cos X + C2 sin x + sin x log (csc x - cot x).
3. y = CI cos X + C2 sin x + 1 tan x sin x.
4. y = +
cle z c2e- z - t sin 2 x-f·
5. y = CI cos X +
C2 sin x +
t cos 2x 1. +
6. y = cle- 2z +
c2e- z +z
2e .
4 -z
7. y = Cle -z+ C2xe -z+xe12 .
8. y = CI cos X + C2 sin x - x 2 cos X + x sin x.
9. y = tx e-2 Z
(210g x - 3) (CI + +
c2x)e-z •
10. y = CI cos X +C2 sin x +
sin x log sin x - x cos x.
11. y = CI cos X +C2 sin x +
sin x log (sec x tan x) - + 2.
Le8son 23 LINEAR EQUATION WITH NONCONSTANT COEFFICIENTS 241

12. Y == + C2X + Xlog x).

13. Y == Cl cos X + C2 sin x + sin x log (esc x - cot x) -
cos x log (sec x + tan x).
14. Y == (Cl + + x - f).
15. Y == + - cos
x
16. Y == CIX + C2X log x + 2 (log x) 2·
17. Y == CIX + C2X2 + lx3 10g x - fx 3 •
3
18. 11 = CIX
2
+ C2X- + X5
2
•

19. Y = + C2X- 1 +
CIX + x-I).
20. y = Clx 1l2 + C2X- 1 - ix- 1 log x.
LESSON 23. Solution of the Linear Differential Equation with
Nonconstant Coefficients. Reduction of
Order Method.

LESSON 23A. Introductory Remarks. We are at last ready to ex-

amine the general linear differential equation
(23.1) In(x)y("') + 1",-1 (x)y(",-u + · · · + 11 (X)y' + lo(x)y = Q(x),

and its related homogeneous equation

(23.11) I",(x)y("') + 1",-1 (x)y(n-U. + · · · + I. (x)y' + lo(x)y = 0,
where lo(x), 11 (x), · · · , I",(x), Q(x) are each continuous functions of x on
a common interval I and I",(x) 0 when x is in I.
We have already remarked that in most cases the solutions of (23.1)
will not be expressible in terms of elementary functions. However, even
when they are, no standard method is known of finding them, as is the
case when the coefficients in (23.1) or (23.11) are constants, unless the
coefficient functions I",(x) are of a very special type; see, for example,
Exercise 23,18. For an unrestricted nth order equation (23.11) that has a
solution expressible in terms of elementary functions, the best you can
hope for by the use of a standard method is to find one independent solu-
tion, if the other n - 1 independent solutions are known. And for (23.1),
the best you can obtain from a standard method is to find one independent
solution of (23.11) and a particular solution of (23.1), again provided the
other n - 1 independent solutions of (23.11) are known.
You can see, therefore, that even in showing you a standard method
for finding a general solution of only the second order equation
(23.12) 12(X)y" + 11 (x)y' + lo(x)y = Q(x),
it is essential that the functions lo(x), 11 (x), 12(X) be of such a character
that the needed first solution of its related homogeneous equation can be
discovered.
242 HIGHER ORDER LINEAR DIFFERENTIAL EQUATIONS Chapter 4.

Comment 23.13. The function Y == 0 always satisfies (23.11). Since

this solution is of no value, it has been appropriately called the trivial
solution.

LESSON 23B. Solution of the Linear Differential Equation with

Nonconstant Coefficients by the Reduction of Order Method. As
remarked in Lesson 23A, we assume that we have been able to find a non-
trivial solution Yt of the homogeneous equation
(23.14) f2(X)Y" + ft(x)y' + fo(x)y = o.
The method by which we shall obtain a second independent solution of
(23.14), as well as a particular solution of the related nonhomogeneous
equation
(23.15) 12(X)Y" + II (x)Y' + Jo(x)y = Q(x),
is called the reduction of order method.
Let Y2(X) be a second solution of (23.14) and assume that it will have
the form
(23.2) Y2(X) = Yl(X) f u(x) dx,

where u(x) is an unknown function of x which is to be determined. The

derivatives of (23.2) are
(23.21) Y2'(X) = YIU + YI' f u(x) dx,

(23.22) Y2"(X) = YIU' + YI'U + Yl'U + YI" f u(x) dx.

Substituting the above values for Y2, Y2', and Y2" in (23.14), we see, by
Definition 3.4, that Y2 will be a solution of (23.14) if

(23.23) 12(X) [YIU' + 2Yl'U + Yl" f u(x) dX]

+ h(x) [YIU + Yl' f u(x) dx] + lo(x) [Yl(X) f u(x) dx] = O.
We can rewrite this expression as

(23.24) (!2(X)Yl" + h(X)Yl' + IO(X)Yl) f u(x) dx

+ 12(X)YIU' + [2/2(X)Yt' + Jl(X)Yl]U = O.
Since we have assumed that YI is a solution of (23.14), the quantity in
the first parenthesis of (23.24) is zero. Hence (23.24) reduces to
(23.25)
Lesson 23B REDUCTION OF ORDER METHOD 243

Multiplying (23.25) by dx/[uf2{X)Yl], we obtain

(23.26) du + 2 dYl = -fl{X) dx

u Yl f2{X) .
Integration of (23.26) gives
(23.27) log u + 2 log Yl = - ! fl{X)
f2{X) dx,

Iog (UYI 2) = - ! ff2{X)

1 (x) d
x,

2
- f/ 1 (:&) d:&

UY1 = e /2(:&) ,

- f/ 1 (:&) d:&

U = e /2(:&) /YI 2 •
Substituting this value of U in (23.2), we obtain for the second solution of
(23.12)
- f/
f
1 (:&) d:&
e /2(:&)
(23.28) Y2 = Yl Y1 2 dx

We shall prove in Lesson 64, see Example 64.22, that this second solution
Y2 is linearly independent of Yl.
Comment 23.29. We do not wish to imply that the integral in (23.28)
will yield an elementary function. In most cases it will not, since in most
cases, we repeat, the given differential equation does not have a solution
which can be expressed in terms of elementary functions. Further the
solution Yl need not itself be an elementary function, although in the
examples below and in the exercises, we have carefully chosen differential
equations which have at least one solution expressible in terms of the
elementary functions.
Example 23.291. Find the general solution of
(a) x 2 y" + xy' - y = 0, x 0,
given that Y = x is a solution of (a).
Solution. Comparing (a) with (23.14), we see that f2{X) = x 2 ,
fl{X) = x. Hence by (23.28) with Yl = x, we obtain

(b) Y2 = X
f
-
f
e x 2:&
-d:&
:&2

dx = x f -log:&

e x2 dx = X f x- 3 dx
-2 -1
-
- xX
--2-
- - --.
2
X
244 HIGHER ORDER LINEAR DIFFERENTIAL EQUATIONS Chapter 4

Hence the general solution of (a) is

(c)

Comment 23.292. Although formulas are useful, it is not necessary

to memorize (23.28). We shall now solve the same Example 23.29 by
using the substitution (23.2). With Yl = x, (23.2) becomes

(d) Y2 = x f u(x) dx.

Differentiating (d) twice, we obtain

(e) Y2' = xu + f u(x) dx, Y2" = xu' + 2u.

Substituting (d) and (e) in (a), we have

(f) X2(XU' + 2u) + x [ xu + f u(x) dxJ - x f u dx = 0,

which simplifies to the separable equation

(g) xu' + 3u = 0, x o.
Its solution is
(h) u = x- 3 •
Substituting (h) in (d), there results

(i) Y2 = X fx -3
dx = X
-2-1
x --2 = -
-2
.
X

Hence the general solution of (a) is, as found previously,

(j)

The same substitution (23.2), namely

(23.3) y(x) = Yl(X) f u(x) dx,

in the nonhomogeneous equation (23.15) will yield not only a second

independent solution of (23.14) but also a particular solution of (23.15). As
before, we differentiate (23.3) twice and substitute the values of y, y', and
y" in (23.15). The left side of the resulting equation will be exactly the
same as (23.25) found previously. Its right side, however, will have Q(x)
in it instead of zero. Hence in place of (23.25), we would obtain
(23.31)
Lesson 23B REDUCTION OF ORDER METHOD 245

an equation which is now linear in u and therefore solvable by the method

of Lesson lIB. The substitution of this value of u in (23.3), will make
y(x) a solution of (23.15).

Comment 23.311. Two integrations will be involved in this pro-

cedure; one when solving for u in (23.31), the other when integrating u in
(23.3). There will, therefore, be two constants of integration. If we in-
clude these two constants, we shall obtain not only a second independent
solution Y2(X) of (23.14) and a particular solution Yp(x) of (23.15), but
also the first solution Yl (x) with which we started. Hence by this method
the substitution (23.3) will give the general solution of (23.15).
Example 23.32. Given that Y = x is a solution of

x 2 y" + xy' - Y = 0, x 0,

find the general solution of

(a) x 2 y" + xy' - y = x, x o.

Solution. By (23.3), with one solution Yl(X) = x, we have

(b) y(x) = x f u(x) dx.

Its derivatives are

(c) y'(x) = xu + udx, f

y"(x) = xu' + u + u.

Substituting (b) and (c) in (a) gives

(d)
2
x (xu' + 2u) + x (xu + f u dx) - x f u dx = x,

which simplifies to
(e)

This equation is linear in u. By Lesson lIB, its integrating factor is

e!3z- dz = e 10g z3 = x 3 • Hence, by (11.19),
1

(f)

U = - X
-1
+ Cl,X- 3 •
2
246 HIGHER ORDER LINEAR DIFFERENTIAL EQUATIONS Chapter 4.

Substituting this value of u in (b), we obtain

(g) y(x) = x f (x 2 I + CI/x-a) dx = x [! log x - CI' X2 2 + C2]

which is equivalent to
(h) y(x) = CIX-
l
+ C2X + ; log x.
Comment 23.33. We could, if we had wished, have obtained (e)
directly from (23.31). Comparing (a) with (23.15), we see that f2(X) = x 2,
fl(X) = x and Q(x) = x. The Yl in this formula is the given solution x.
Substituting these values in (23.31) will yield (e). Verify it.

EXERCISE 23

Use the reduction of order method to find the general solution of each
of the following equations. One solution of the homogeneous equation
is shown alongside each equation.
I. x 2y" - xy'+ y = 0, Yl = X.
2• y" - -2 Y'+ -2
2 Y = 0, Yl = X.
X x
3. (2x 2 + l)y" - 4xy' +
4y = 0, Yl = x.
4. y" + (x 2 - X)y' - (x - l)y = 0, Y 1 = x.

5• y"+ (x"2 - X1),y - y = 0, Yl = X2•

6. 2x2y" + 3xy' - y = 0, Yl = x l/2 •
7. y" + [f(xj - l]y' - f(x)y = 0, Yl = e Z
•

8. y" + xf(x)y' - f(x)y = 0, Yl = x.

9. x 2y" - xy' + y = x, Yl = X.

I O. y" - -2 Y'+ -2
2 Y = Iog x, Yl = X X.
X x
II. x 2y" + xy' - 4y = x 3, Yl = x 2•
12. x 2y" + xy' - y = x 2e- Yl = x. z
,

13• 2X 2Y" + 3xy' - Y = -,

1 Yl = X 112 •
x
14. x 2y" - 2y = 2x, Yl = x 2.
15. y" + (x 2 - l)y' - x 2y = 0, Yl = eZ •
16. x 2y" - 2y = 2x 2, Yl = x 2.
17. x 2y" +
xy' - y = 1, y 1 = x.
18. The differential equation
(23.4) a3(x - xO)3 y'" + a2(x - xO)2y" + al(x - xo)y' + aoy = Q(x), x Xo,
Lesson 23-Exercise 247

where ao, aI, a2, as, Xo are constants, and Q(x) is a continuous function of x,
is known as an Euler equation. Prove that the substitution
x - Xo = eU , U = log (x - xo),
will transform (23.4) into a linear equation with constant coefficients. Hint.

dy = dy dx = (x _ xo) dy ,
du dz du dx
2
tty dy 2 d y
- - - - (x - xo) - 2,
du2 du - dx

tty
du s - 3 du 2 + 2 du = (x - xo) dz3·

Use the substitution given in problem 18 above to solve equations 19-21.

19. (x - 3)2y" + (x - 3)y' +
y = x, x 3.
20. x 2y"+ xy' = 0, x O.
21. 3
x y'"+ 2
2x y" - xy' +
y = x, x o.
22. Prove that the substitutions

will transform the linear equation

g2(X)y" + gl(X)y' + go(x)y = 0

into the Riccati equation (see also Exercise 11,25)

, go(x) gl(X) 2
Yl = - - - - - - Yl - Yl •
g2(X) g2(X)
Hence if Yl is a solution of the Riccati equation, y = ef fil dz is a solution of
the linear equation.
23. Use the substitutions given in 22 to transform the linear equation xy" -
y' - x3 y = 0 into a Riccati equation. Find, by trial, a solution of the
Riccati equation; then find a solution of the linear equation. With one
solution of the linear equation known, find its general solution.
24. Prove that the substitution
1 du
y - f2(X)U dz' f2(:1;) 0,

y' = _ u" + U' (f2 U' + f2'U) ,

f2U f22u2
will transform the Riccati equation,

into the second order linear equation

f2(X)U" - [f2'(X) + fl(X)f2(X)]U' + fo(x)ff2(X)]2u = o.
248 HIGHER ORDER LINEAR DIFFERENTIAL EQUATIONS Chapter 4.

25. A second order linear differential equation

(23.5) 12(X)Y" + 11 (X)y' + lo(x)y = Q(X),

is said to be exact if it can be written as

(23.51) ! [M(X) + N(X)Y] = Q(x),

i.e., if its left side can be written as the derivative of a first order linear
expression. A necessary and sufficient condition that the equation be exact is
that
+ 10 = o.
2
(23.52) d '12 _ dll
dx 2 dx
If the equation is exact, then M(x) and N(x) of (23.51) are given by
(23.53) M(x) = 12(X) and N(x) = Il(X) - 12'(x).
Show that the following equation is exact and solve.
(x 2 - 2x)y" + 4(x - l)y' + 2y = e2z •
2z
2 e
Ans. (x - 2x)y = 4 + C1X + C2.
26. A function k(x) is called an integrating factor of an inexact differential
equation, if after multiplication of the equation by it, the resulting equation
is exact. A necessary and sufficient condition that k(x) be an integrating
factor of the linear differential equation
(23.6) 12(X)Y" + 11 (x) y'''lo (x)y = Q(x)

is that it be a solution of the differential equation

2
d d
(23.61) dx 2 (12k) - dx (Ilk) + 10k = o.
Find an integrating factor of the following equation, then mUltiply the
equation by the integrating factor and use (23.52) to prove that the resulting
equation is exact.
x 3 y" - (2x 3 - 6x 2)y' - 3(x 3 + 2x2 - 2x)y = O.
Ans. k(x) = eZ ; also k(x) = e-3z •

ANSWERS 23
I. y = C1X + C2X log x.
2. y = C1X + C2x2.
3. y = CIX + c2(2x 2 - 1).
z3 z2

4. Y = CIX + C2X f e
--+-
3 2
x2 ch.

I
-z2/4
2 2 e
5. y = C1X + C2X :.;3 dx.
Lesson 23-Answers 249

6. Y = CIXl/ 2 + C2X- 1 •
7• Y = Cle
+ C2e e dx.

8. Y = CIX + C2X e x2 • f
x
CIX + C2X log x + 2 (log x) •
2
9. Y =
10. Y = OlX + C2X 2 + lx3 log x - fx 3 •
3
CIX + C2X- + x5 .
2 2
II. Y =
12. Y = CIX + C2X- 1 + + x-I).
13. Y = CIXl/ + C2X- -
2 1 ix- 1 log x.
14. Y = CIX 2 + C2X- 1 - x.

15• Y = Cle
+ C2e e dx.

CIX + C2X- 1 + ix log x.

2 2
16. Y =
17. Y = CIX - C2X- 1 - 1.
2
19. d
du 2 Y+ Y = e "'+ 3,
Y = Cl cos [log (x - 3)] + C2 sin [log (x - 3)J + i+
2
d y
20. du 2 = 0, Y = Cl log x + C2.
3 2
d y
21. du 3 -
d y
du 2 -
dy
du +y = e,
u
y = CIX
+ C2X log X + C3X -1 + "4X (log x )2 •
23. Riccati equation: Yl
I
= X
2
+ -Ylx - 2
Yl . Yl
'
= x, Y =e

General solution:
 Chapter 5

Operators and Laplace Transforms

In the next and succeeding lessons we shall prove certain theorems by

means of mathematical induction. We digress momentarily, therefore, to
explain the meaning of a "proof by mathematical induction."
Suppose we wish to prove a statement about the positive integers as,
for example, that the sum of the first n odd integers is n 2 , Le., suppose
we wish to prove
(a) 1 + 3 + 5 + 7 + ··· + (2n - 1) = n2•

By direct substitution we can verify that (a) is true when n = 1, n = 2,

n = 3, for then (a) reduces to the respective identities 1 = 12 , 1 + 3 = 2 2 ,
1 + 3 + 5 = 3 2 • Since it would be impossible to continue in this manner
to check the accuracy of (a) for every n (there are just too many of them),
we reason as follows. If by assuming (a) is valid when n = k, where k is
an integer, we can then prove that (a) is valid when n = k + 1, it will
follow that (a) will be true for every n. For then the validity of (a) when
n = 3 will insure its validity when n = 4. The validity of (a) when
n = 4 will insure its validity when n = 5, etc., ad infinitum. And since
we have already shown by direct substitution that (a) is valid when
n = 3, it follows that (a) is valid when n = 4, etc., ad infinitum.
We shall now use this reasoning to prove that (a) is true for every n.
We have already shown that (a) is true when n = 1, n = 2, and n = 3.
We now assume that (a) is true when n = k. Hence we assume that
(b) 1+3 + 5 + 7 + · · · + (2k - 1) = k
2

is a true equality. Let us add to both sides of (b) the next odd number in
the series, namely (2k - 1 + 2) = 2k + 1. Equation (b) then becomes

(c) 1 + 3 + 5 + 7 + · · · + (2k - 1) + (2k + 1)

= k + 2k + 1 = + 1)2.
2
(k

Since we have assumed (b) is true, it follows that (c) is true. But if in
250
Lesson 24A DEFINITION OF AN OPERATOR. LINEAR PROPERTY 251

(a), we let n = k + 1, we obtain

(d) 1 + 3 + 5 + 7 + ··· + (2k + 1) = (k +
which agrees with (c). Therefore, if (b) is true so also is (d). Hence we
have shown that if (a) is true when n = k, it is true when n = k 1. +
But (a) is true when n = 3. It is therefore true when n = 4. Since
(a) is true when n = 4, it is therefore true when n = 5, etc., ad infinitum.
A proof which uses the above method of reasoning is called a proof by
induction.
Comment 24.1. Every proof by induction must consist of these two
parts. First you must show that the assertion is true when n = 1 (or
for whatever other letter appears in the formula). Second you must show
that if the assertion is true when n = k, where k is an integer, it will
then be true when n = k + 1.

LESSON 24. Differential and Polynomial Operators.

LESSON 24A. Definition of an Operator. Linear Property of

Polynomial Operators. An operator is a mathematical device which
converts one function into another. For example, the operation of differ-
entiation is an operator since it converts a differentiable function I(x) into
a new function f'(x). The operation of integrating f:o
f(t) dt is also an
operator. It converts an integrable function f(t) into a new function F(x).
Because the derivative at one time was known as a differential coeffi-
cient, the letter D, which we now introduce to denote the operation of
differentiation, is called a differential operator. Hence, if y is an nth
order differentiable function, then

(24.11) DOy = y, Dy = y',

(Other letters may also be used in place of y.) For example, if y(x) = x 3,
then DOy = x 3, Dy = dy/dx = 3x 2, D2y = d 2y/dx 2 = 6x, D3y =
d 3y/dx 3 = 6, D4y = d 4y/dx4 = 0; if r(8) = sin 8 +
8 2 , then DOr =
sin 8 + 8 2 , Dr = dr/d8 = cos 8 + 28; D2r = d 2r/d8 2 = -sin 8 + 2; if
x(t) = t 2, then DOx = t 2, Dx = dx/dt = 2t, D 2x = d 2x/dt 2 = 2, D3 X .
=0
By forming a linear combination of differential operators of orders 0 to
n, we obtain the expression

(24.12) P(D) = ao + aiD + a2D2 + · · · + anDn, an 0,

where ao, ai, · · · , an are constants.

Because of the resemblance of P(D) to a polynomial, we shall refer to
it as a polynomial operator of order n. Its meaning is given in the
following definition.
252 OPERATORS AND LAPLACE TRANSFORMS Chapter 5

Definition 24.13. Let P(D) be the polynomial operator (24.12) of

order n and let y be an nth order differentiable function. Then we define
P(D)y to mean
(24.14) P(D)y = (anDn '+ · · · + aID + ao)Y
= anDny + · · · + alDy + aoY·
By (24.11), we can also write (24;.14) as
(24.15)

Hence, by (24.15), we can write the linear equation with constant coeffi-
cients,

as
(24.17) P(D)y = Q(x),

where P(D) is the polynomial operator (24.12).

Theorem 24.2. If P(D) is the polynomial operator (24.12) and Yb Y2
are two nth order differentiable functions, then
(24.21)

'where CI and C2 are constants.

Proof.In the left side of (24.21) replace P(D) by its value as given
in (24.12). There results
(a) (anDn + an_1Dn-1 + · · · + alD + aO)(CIYI + C2Y2).
By Definition 24.13, we can write (a) as
(b) anDn(ClYl + C2Y2) + an_1Dn-I(CIYI + C2Y2) + · · ·
+ alD(cIYI + C2Y2) + aO(cIYI + C2Y2).
By (24.11) and the property of derivatives,
(c) Dk(CIYI + C2Y2) = CIy(k) + C2y(k) = clDkYI + C2DkY2,
for each k = 0, 1, 2, · · · ,n. Hence (b) becomes
(d) an(c1DnYl + C2DnY2) + an-l (clDn-I Y1 + C2Dn-1Y2)
+ · · · + al (clDYI + C2 DY2) + aO(cIYI + C2Y2)
= cl(anDn + an_1D n- 1 + · · · + aiD + aO)YI
+ c2(anDn + an_1Dn-l + · · · + aiD + aO)Y2
= clP(D)Yl +
c2 P (D)Y2,
Le••on 24A DEFINITION OF AN OPERATOR. LINEAR PROPERTY 253

which is the same as the right side of (24.21). We have thus shown that
the left side of (24.21) is equal to its right side.
By repeated application of (24.21), it can be proved that

(24.22) P(D)(CIYl + C2Y2 + · · · + CnYn)

= c1P(D)Yl + c2 P (D)Y2 + · · · + cnP(D)y",
where Cl, C2, ... , Cft are constants, and each of Yl, Y2, • · • , y" is an nth
order differentiable function.
Example 24.221. Use Definition 24.13 to evaluate

(a) (D2 - 3D + 5)(2x 3 + e2z + sin x).

Solution. Here P(D) = D2 - 3D + 5 and the function Y of (24.14)
is 2x 3 + e2z + sin x. Therefore, by (24.14) and (24.11),

(b) (D 2 - 3D + 5)(2x 3 + e2z + sin x)

= D2(2x 3 + e2z + sin x) - 3D(2x 3 + e2z + sin x)
+ 5(2x 3 + e2z + sin x)
= 12x + 4e 2z - sin x - 18x 2 - 6e 2z - 3 cos x
+10x 3 +5e 2z + 5 sin x
= 10x 3 - 18x 2 + 12x + 3e 2z + 4 sin x - 3 cos x.

Example 24.222. Use (24.22) to evaluate

(a) (D 2 - 3D + 5)(2x 3 + e2z + sin x).

NOTE. This example is the same as Example 24.221 above.
Solution. By (24.22), with P(D) = D2 - 3D + 5,
(b) (D2 - 3D + 5)(2x 3 + e2z + sin x)
= (D 2 - 3D + 5)(2x 3 ) + (D 2 - 3D + 5)e 2z
+ (D 2 - 3D + 5) sin x
= 2(6x - 9x 2 + 5x 3 ) + (4e 2z - 6e 2z + 5e 2z )
+ (-sin x - 3 cos x + 5 sin x)
= 10x 3 - 18x 2 + 12x + 3e 2x + 4 sin x - 3 cos x,

the same result obtained previously.

Definition 24.23. An operator which has the property (24.21) is
called a linear operator. Hence the polynomial operator (24.12) is
linear.
254 OPERATORS AND LAPLACE TRANSFORMS Chapter 5

Comment 24.24. With the aid of the linear property of the poly-
nomial operator P(D) of (24.12) we can easily prove the following two
assertions, proved previously in Theorem 19.3.
1. If Y 1, Y2, · · · , Yn are n solutions of the homogeneous linear equation
P(D)y = 0, then Yc = CIYI + C2Y2 + · · · + CnY", is also a solution.
2. If Yc is a solution of the homogeneous linear equation P(D)y = 0, and
YP is a particular solution of the nonhomogeneous equation P(D)y =
Q(x), then y = Yc + yp is a solution of P(D)y = Q(x).
Proof of 1. Since Yb Y2, • • • , Yn are each solutions of P(D)y = 0, we
have
(a) P(D)YI = 0, P(D)Y2 = 0, · · · , P(D)y", = 0.
Hence also
(b)

where CI, C2, • • • , Cn are constants. Adding all equations in (b) and making
use of (24.22), we obtain

(c)

which implies that Yc = CIYI + C2Y2 + · · · + c",y", is a solution of

P(D)y = O.
Proof of 2. By hypothesis Yc is a solution of P(D)y = 0, and Yp is a
solution of P(D)y = Q(x). Therefore
(d) P(D)yc = 0 and P(D)yp = Q(x).

Adding the two equations in (d) and making use of (24.21), we obtain

(e) P(D)(yc + yp) = Q(x),

which implies that y = Yc + yp is a solution of P(D)y = Q(x).

Comment 24.25. Principle of Superposition. In place of the
linear differential equation

(a)

where P(D) is a polynomial operator (24.12), let us write the n equations

(b) P(D)y = Q2, · · • , P(D)y = Q",.

Let YIp, Y2p, · • · , Ynp be respective particular solutions of the n equations

of (b). Therefore

(c) P(D)YIp = QI, P(D)Y2p = Q2, • • • , P(D)Ynp = Q",.

Lesson 24B ALGEBRAIC PROPERTIES OF POLYNOMIAL OPERATORS 255

Adding all the equations in (c) and making use of (24.22), there results

(d) P(D)(YIP + Y2p + · · · + Ynp) = QI + Q2 + · · · + Qn,

which implies that

(e) YP = YIp + Y2p + · · · + Ynp

is a solution of (a).
We have thus shown that a particular solution YP of (a) can be obtained
by summing the particular solutions YIp, Y2p, • • • , Ynp of the n equations
of (b). The principle used in this method of obtaining a particular solu-
tion of (a) is known as the principle of superposition.
Example 24.26. Use the principle of superposition to find a particular
solution of the equation

(a)

Solution. By Comment 24.25, a solution YP of (a) is the sum of the

particular solutions of each of the following equations.

(D2 + l)y = 3.

Particular solutions of each of these equations are respectively-write

(D 2 + l)y as y" + Y and use the method of Lesson 21-

(c) YIp = x 2 - 2, Yap = 3.

Hence a particular solution of (a) is

LESSON 24B. Algebraic Properties of Polynomial Operators.

Whenever PleD) and P 2(D) appear in this lesson, we assume that

(24.3) PleD) = anDn + an_IDn-1 + · · · aID + ao, +

P 2 (D) = bmD'" + bm_IDm-1 + · · · + bID + bo,

are two polynomial operators of orders nand m respectively, n m.

Whenever Y appears we assume it is an nth order differentiable function,
defined on an interval I.
Definition 24.31. The sum of two polynomial operators P I (D) and
p 2 (D) is defined by the relation
(24.32)
256 OPERA,TOI\B ANI) LAPLACE TRANSFORMS ChapterS

Theorem 24.33. Let PI(D) and P 2(D) be two polynomial operators of

tMfortrU! (£4.8). Then PI and P 2 can be addedjWlt as if they were ordinary
polynomials, i.e., we can add the coejJicients of like orders of D.

Proof. By (24.32), (24.3) and the rules of differentiation,

(a) (PI + P 2 )y = (a"D" + ... + a2D2 + aiD + ao)Y

+ (b",D'" + ... + b D2 + hD + bo)Y
2

= (anD" +. . + b",D'" + .. + a2D2 + b2D2

+
aiD + biD + ao + bo)Y
= [anD" + ... + b",D'" + ... + (a2 + b 2)D 2

+
(al + bl)D + (ao + bo)]y.
Example 24.331. If

(a) PleD) = 3D s + D2 + 3D - 1, Pz(D) = 5D2 - 7D + 3,

find PI(D) +P 2 (D).

Solution. By Theorem 24.33,

As an exercise, prove that the following identities follow from Definition

24.31.
(24.34) PleD) + P!I(D) = P 2(D) + PI (D)
(commutative law of addition).
PleD) + [P 2 (D) + Pa(D)] = [PI (D) + P 2(D)J + Ps(D)
= PI (D) + P 2 (D) + Pa(D)
(associative law of addition).

Definition 24.35. The product of a function hex) by a polynomial

operator P(D) is defined by the relation

(24.36) [h(x)P(D»)y = h(x)[P(D)y].

E%ample 24.37. Evaluate

(al

Solution. Comparing (a) with (24.36), we see that hex) = 2x 2 ,

P(D) = 3D 2 + 1, y(x) = e ax . Therefore, by (24.36), (24.14) and (24.11),

(b) [2X2(3D2 + 1)leS'" = 2xZ[(3DZ + I)ea,,)

= 2x 2[3(ge S") + ea,,] = 2x 2(2&3") = 56x 2e 3x .
Lesson 24B ALGEBRAIC PROPERTIES OF POLYNOMIAL OPERATORS 257

Example 24.38. Evaluate

(a) (3x 3 + 2)[(D z + 3D + 2) + (2D Z

- 1)]e 2 :e.

Solution. By Theorem 24.33 and (24.36), we can write (a) as

Carrying out the indicated differentiations in (b), we obtain

(c) (3x 3 + 2)(12 + 6 + 1)e 2:e = 1ge z:e(3x 3 + 2).

Comment 24.381. By (24.32), we may also write

(24.39) h(X)[PI(D) +P 2 (D)]y = h(x)[PI(D)y + Pz(D)y]

= h(x)[PI(D)y] + h(x)[Pz(D)y].
Example 24.4. Evaluate (a) of Example 24.38, by use of (24.39).

Solution.
(a) (3x 3 + 2)[(D z + 3D + 2) + (2D z - 1)]e z:e
= (3x + 2)[(D z + 3D + 2)e z:e] + (3x + 2)[(2D
3 3 Z
- l)e 2:e]
= (3x 3 + 2)(4 + 6 + 2)e z:e + (3x 3 + 2)(8 - l)e z:e
= 1ge z:e(3x 3 + 2).
Definition 24.41. The product of two polynomial operators P I (D)
and Pz(D) is! defined by the relation

(24.42)

Example 24.43. Evaluate

(a) [(3D z - 5D + 3)(D z - 2)](x 3 + 2x).

Solution. By (24.42), (24.14) and (24.11),

(b) (3D z - 5D + 3)[(D Z

- 2)(x 3 +
2x)]
z +
= (3D - 5D 3)(6x - 2x 3 - 4x)
= 3(-12x) - 5(6 - 6x z - 4) + 18x - 6x 3 - 12x
= -6x 3 +
30x z - 30x - 10.

Prove as exercises that the following identity follows from Definition

24.41,
(24.44) PI (D)[P z(D)P 3 (D)] = [PI (D)P z(D)]P 3 (D)
= P I (D)P z(D)P 3 (D)
(associative law of multiplication),
258 OPERATORS AND LAPLACE TRANSFORMS Chapter 5

and the following identity from Definitions 24.31 and 24.41,

(24.45) P I (D)[P 2(D) + P 3 (D)] = P I (D)P 2(D)+ P I (D)P3 (D)

(distributive law of multiplication).
Theorem 24.46. If

(24.47) P(D) = anDn + an_IDn- + ... + aiD + ao,

1
a" ¢ 0,

where ao, a I, ••• , an are constants, then

(24.48)

where r}, r2, ... , rn are the real or imaginary roots of the characteristic equa-
tion (20.14) of P(D)y = 0, i.e., a polynomial operator with constant coeffi-
cients can be factored just as if it were an ordinary polynomial.
Proof. We shall prove the theorem only for n = 2, i.e., we shall
prove

In the equalities which follow, we have indicated the reason after each
step. Be sure to refer to these numbers. We start with the right side of
(a) and show that it yields the left side.

By
(b) [(D - rl)(D - r2)]y
= (D - rl)[(D - r2)y] (24.42),
= (D - rl)(Dy - r2Y) (24.14),
= (D - rl)Dy - (D - rl)(rzy) (24.21),
= D2y - rlDy - r2Dy + rlr2Y (24.14),
= D2y - +
(rl r2)Dy +rlr2Y Theorem 24.33,
= [D 2 - h + r2)D +
rlr2]y (24.14).

Corollary 24.481. If P(D) is the polynomial operator (24.47), then

(24.482)

where PleD) and P 2(D) may be composite factors of P(D), i.e., PI and P 2
may be products of factors of (24.48).
The proof follows from Theorem 24.46.
Theorem 24.49. The commutative law of multiplication is valid for
polynomial operators, i.e.,
Lesson 24B ALGEBRAIC PROPERTIES OF POLYNOMIAL OPERATORS 259

Proof. In the proof of Theorem 24.46, interchange the subscripts 1

and 2 of rl and r2' Since the final formula on the right of (b) in the proof
will remain the same, the equality (24.5) follows.

Example 24.51. Evaluate (D2 - 2D -3) (sin x + x 2).

Solution. Method 1. By application of Theorem 24.2.
(a) (D 2 - 2D - 3)(sin x + x 2)
= (D 2 - 2D - 3) sin x +
(D2 - 2D - 3)x 2
= -sin x - 2 cos x - 3 sin x +
2 - 4x - 3x 2
= 2 - 4x - 3x 2 - 4 sin x - 2 cos x.

Method 2. By application of Theorem 24.46 and (24.42).

(b) (D2 - 2D - 3) (sin x + x2)
= [(D + I)(D - 3)](sin x +
x 2)
(D + 1)[(D - 3)(sin x +
x 2 )]
= (D + +
l)(cos x 2x - 3 sin x - 3x 2)
= -sin x + 2 - 3 cos x - 6x + cos x
+ 2x - 3 sin x - 3x 2
= 2 - 4x - 3x 2 - 4 sin x - 2 cos x.

Definition 24.52. If

(24.521)

is a polynomial operator of order n, then

(24.522) P(D + a) = an(D + a)n + ... + al(D + a) + ao,

where a is a constant, i.e., (24.522) is the polynomial operator obtained
by replacing Din (24.521) by D a. +
Summary 24.523. * Polynomial operators can be added, multiplied,
factored, and multiplied by a constant, just as if they were ordinary
polynomials. Furthermore, the

(24.53) Commutative law: P 1 P 2 = P 2Pt,

(24.54) Associative law: P 1 (P2 P a) = (PIP 2 )Pa = P 1P 2P a,
(24.55) Distributive law: P 1 (P2 +
P a) = P 1 P 2 P1Pa +
are all valid.

"Some of the properties here summarized do not apply to polynomial operators with
nonoonstant coefficients. See, for example, Exercise 24, 17 and 18.
260 OPERATORS AND LAPLACE TRANSFORMS Chapter 5

LESSON 24C. Exponential Shift Theorem for Polynomial Opera-

tors. If the function to be operated on has the special form ue''''', where
u is an nth order differentiable function of x defined on an interval I,
then the following theorem, called the exponential shift theoreDl
because the formula we shall obtain shifts the position of the exponential
e ...., will aid materially in evaluating P(D)ue"'z.
Theorem 24.56. (Exponential Shift TheoreDl) If
(24.57)
an ¢ 0, is a polynomial operator with constant coefficients and u(x) is an
on
nth order differentiable function of x defined an interval I, then
(24.58) az
P(D)(ue ) = e"'ZP(D +a)u,
where a is a constant.
Proof. We shall first prove by induction that the theorem is true for
the special polynomial operator P(D) = Dk. By Comment 24.1 we must
show that:
1. (24.58) is true when k = 1, i.e., when P(D) = D.
2. If (24.58) is valid when k = n, then it is true when k = n·+ 1, i.e.,
if (24.58) is valid when P(D) = Dn, then it is true when P(D) = D"H
Proof of 1. When P(D) = D, the left side of (24.58) is D(ue"'Z). By
(24.11) and (24.14),
(a) D(ueaz ) = aueaz +
e"""u' = eaz(u' +
au) = e"'Z(D +
a)u.
By Definition 24.52, if P(D) = D, then P(D +
a) = D + a.Substitut-
ing these values of D and D +
a in the first and last terms of (a), we
obtain (24.58).
Proof of 2. We assume that (24.58) is true when P(D) = D". We
must then prove that the theorem is true when P(D) = Dn+l. By
Definition (24.52), if P(D) = Dn, P(D +a) = (D +
a)n. Substituting
these values of P(D) and P(D +
a) in (24.58), we obtain
(b)
which, by our assumption, is a true equality. Operating on (b) with D,
we obtain
By
(c) D[Dn(ue"'z)) = D[eaz(D +
a)nu) (b),
= eazD(D +
a)nu +ae"'Z(D +
a)nu (24.11),
+ +
= eaz[D(D a)n a(D a)")u + (24.39),
= eaz[(D +
a)n(D +a))u Corollary 24.481,
+
= eaz(D a)nHu. Corollary 24.481.
Lesson 24C EXPONENTIAL SHIFT THEOREM-POLYNOMIAL OPERATORS 261

The left side of (c) is Dn+l(ueaz ). Hence we ha.ve shown by (c) that the
validity of (b) leads to the validity of
(d)
It therefore follows by Comment 24.1, that
(24.59)
for every k.
Let P(D) be the operator (24.57). Then
By
(e) P(D) (ue"Z)
= a"D"(ue az ) + ... + aID(ueaz ) + aoueaz (24.14),
= eaz[a,,(D + + ... + al(D + a) + aolu
a)" (24.59),
= eazP(D +a)u Definition 24.52.

Corollary 24.6.
(24.61)

Proof. Let P(D) = (D - a)". Then by Definition 24.52, P(D + a)

= (D + a - a)" = Dn. Substituting these values of P(D) and P(D + a)
in (24.58), we obtain (24.61).
Corollary 24.7. If c is a constant, and P(D) is the polynomial operator
(24.57), then

(24.71) P(D)(ceaz ) = ceazP(a).

Proof. By (24.58) and Definition 24.52,

(a) P(D)ceaz = eazP(D + a)c

= eaz[an(D + a)" + an_I(D + a)n-l + ... + aolc.
By Theorem 24.46, (24.42) and (24.14),

(b) (D + a)kc = (D + a)k-I(D + a)c = (D + a)k-Iac

= (D + a)k-2(D + a)(ac) = (D+ a)k-2(a 2c)

Hence, by (b), we can write the second equation in (a) as

(c) P(D)ceaz = eaz[a"a" + an_Ia,,-1 + ... + ala + ale.

By Definition 24.52, (c) is equivalent to

P(D)ceaz = eazP(a)c = ceazP(a).

262 OPERATORS AND LAPLACE TRANSFORMS Chapter 5

Example 24.8. Evaluate

(a) (D 2 + 2D + 3) (e 2" sin x).
Solution. Comparing (a) with (24.58) we see that P(D) = D2 +
2D +
3, a = 2, U = sin x. Hence by Definition 24.52, P{D 2) = +
(D +
2)2 2(D + 2) + +
3 = D2 6D + +
11. Therefore by (24.58),

(b) (D 2 + 2D + 3) {e " sin x) =

2 +
e2"(D 2 6D +
11) sin x
= e2"(10 sin x +
6 cos x).
Example 24.81. Evaluate
(a)
Solution. Comparing (a) with (24.58), we see that P(D) = D2 -
D +
3, a = -2, U = x 3 • Hence by Definition 24.52, P{D - 2) =
{D - 2)2 - (D - 2) 3 = D2 - 5D + +
9. Therefore, by (24.58),
(b) (D2 - D + 3){x 3e- 2,,) = e- 2"{D2 - 5D + 9)x 3
= e- 2"(6x - 15x 2 + 9x 3 ).
Example 24.82. Evaluate
(a)
Solution. Comparing (a) with (24.61), we see that a = 2, U = sin:l:,
n = 3. Hence by (24.61),

(b) (D - 2)3(e 2 " sin x) = e 2"D3 sinx = -e 2"cosx.

Example 24.83. Evaluate
(a)

Solution. Comparing (a) with (24.71), we see that P{D) = D3 -

3D 2 +
2, c = 5, a = -4. Therefore P{ -4) = -64 - 48 2 = -110. +
Hence by (24.71)
(b) (D 3 - 3D 2 + 2)(5e-4z) = 5e- 4 ,,( -110) = -550e- b •

LESSON 24D. Solution of a Linear Differential Equation with

Constant Coefficients by Means of Polynomial Operators. In
Lessons 21 and 22 we outlined methods for finding the complementary
function Yc and a particular solution YP of the nonhomogeneous linear
equation,
(24.9) any(n) + an_ly(n-ll + ... + alY' + ao = Q{x), an o.
In this lesson we shall solve (24.9) by means of polynomial operators.
We illustrate the method by means of examples:
Lesson 24D SOLUTION OF LINEAR EQUATION BY POLYNOMIAL OPERATORS 263

Example 24.91. Find the general solution of

(a) y'" + 2y" - y' - 2y = e2z •

Solution. In operator notation, (a) can be written as

(b) (D 3 + 2D2 - D - 2)y = e 2z •
By Theorem 24.46, (b) is equivalent to
(c) (D - I)(D + I)(D + 2)y = e2z •
Let
(d) u = (D + I)(D + 2)y.
Then (c) becomes
(e) (D - l)u = e2z,
which is a first order linear differential equation in u. Its solution, by
Lesson lIB, is
(f)

Substituting this value of u in (d) gives

(g) (D + I)(D + 2)y = e2z + clez •
Let
(h) v = (D + 2)y.
Then (e) can be written as
(i)

an equation linear in v. By Lesson llB, its solution is

Substituting this value of v in (h) gives

(k)

an equation linear in y. Its solution, by Lesson llB, is

(1) y = tre 2Z + 6
eZ + C2e-z + cse-2z ,
which can be written as
(m)
264 OPERATORS AND LAPLACE TRANSFORMS Chapter 5

Comment 24.92. The solution (m) could have been obtained much
more easily if we had found Yc by means of Lesson 20, and used the above
method to find only the particular solution e 2x /12. The solution (f)
would then have read

the solution (j)

and the solution (1),

Yr> = ne 2x
•

The roots of the characteristic equation are, by (c), 1, -1, -2. We

could, therefore, easily have written the complementary function

Example 24.93. Find the general solution of

(a) y" +Y= eX.

Solution. In operator notation, (a) can be written as
(b) (D 2 + l)y = eX.

By Theorem 24.46, (b) is equivalent to

(c) (D + i)(D - i)y = eX.
Let
(d) u = (D - i)Y.
Then (c) becomes
(e) (D + i)u = eX, u' + iu = eX,

an equation linear in u. Its solution, by Lesson lIB, is

(f) U = 1 +1 i e + X
Cl
,-ix
e .
Hence (d) becomes
(g) ,.
Y - = 1 +1 i e + Cl ,-ix
X
e ,
whose solution is
(h)

This last equation can be written with new parameters as

(i)
Lesson 24-Exercise 265

Again we remark that (i) could have been obtained more easily, if we
had found Yc by means of Lesson 20 and used the above method to find
only the particular solution e'" /2.
In general, if the nonhomogeneous linear differential equation (24.9)
of order n is expressed as
(24.94)
where rb r2, ... , rn are the roots of its characteristic equation, then a
general solution (or, if arbitrary constants of integration are ignored, a
particular solution) can be obtained as follows. Let
(24.95)
Then (24.94) can be written as
(24.96) (D - rl)u = Q(x),
an equation linear in u. If its solution u(x) can be found, substituting it
in (24.95) will give
(24.97) (D - r2)(D - ra) ... (D - rn)Y = u(x).
Let
(24.98) v = (D - ra) .•. (D - rn)Y.
Then (24.97) becomes
(24.99)
an equation linear in v. If a solution for vex) can be found, substituting
this value in (24.98) will give
(24.991) (D - ra) ... (D - rn)Y = vex).
The repetition of the above process an additional (n - 2) times will
eventually lead to a solution for y.

EXERCISE 24
1. Prove by induction that
12 22+ + + ... + n2
32 = n(n + 1)(2n + 1) .
6
2. Find DOy, Dy, D2y, Day for each of the following:
(a) Vex) = 3x 2 , (b) vex) = 3 sin 2x, (c) Vex) = vx.
s. Find DOr, Dr, D2r for each of the following:
(a) reO) = cos 0 + tan O. (b) reO) = 82 + sin O.
1
(c) rt6) = 82'
266 OPERATORS AND LAPLACE TRANSFORMS Chapter 5

4. Find DOx, Dx, D 2x for each of the following:

(a) x(t) = t 2 + + 1.
3t (b) x(t) = a cos 8t + b sin Bt.
(c) x(t) = a cos (8t + b).
5. Use Definition 24.13 to evaluate each of the following:
(a) (D2 - 2D - 3) cos 2x. (b) (D2 - 6D + 5)2e3z
(c) (D' - 2D2)4x3 . (d) (D2 - 4D 4)(x 2 + + X + 1).
6. Use (24.22) to evaluate each of the following:
(a) D3(cos ax + sin bx). (b)2 (D2 + D) (3ez + 2x3).
(c) (D2 - 2D + 4)(xe + 5x + 2).
Z

7. Use the principle of superposition to find a particular solution of each of the

following equations.
(a) y" + 3y' +
2y = 8 6ez + +
2 sin x.
(b) y" - 2y' - 8y = 9xeZ +
IOe- Z •
(c) y" + 2y' +
lOy = eX 2. +
8. Evaluate PleD) + P2(D), where
(a) PleD) = D2 + 2D - 1, P2(D) = 3D3 + 4D2 - D + 3.
(b) PleD) = D4 - 2D2, P2(D) = D2 - 6D + 5.
9. Evaluate PleD) + P2(D) + P3(D), where
(a) PleD) = D2 - 2D - 3, P2(D) = D2 - D,
P 3(D) = D' +
2D3 - 2D2 3D 4. + +
(b) PleD) = 3D3 2D2 + +
1, P2(D) = 2D4 D3 + + 2D2 - 3D + 2,
P3(D) = 3Ds - 4D2 7. +
10. Evaluate by means of (24.36)
(a) [x 2(D2 + 1)](2e z ). (b) [(x - 1)(D3 + D2)](e2z + x 2).
(c) [e- z (3D + 4)] (tan x + 1). (d) [(x 2 + 3x + 4)(D3 + 1)](x3 + 2).
II. Evaluate two ways, by Theorem 24.33 and by (24.39),
(a) x 2 [(D2 + +
1) D](2e- 2x ).
(b) (x - 1)[(D3 D2)+ +
(2D2 - 6D 5)](x3 2). + +
(c) sin x[(2D2 - D 1) + +(D2 +
D - 1)] cot x.
12. Evaluate two ways, by (24.42) and by Theorem 24.46,
(a) (D - 1)(D +
1)(3e-·).
(b) (D2 - D +
1)(D 1)(3x3 + 2x2 - + 5x - 6).
(c) (D2 +1)(D - 3)(cos x 2e3z ). +
13. Evaluate, by means of Theorem 24.56,
(a) (D2 - 2D - 3)(e2z cos 2x). (b) (D2 - D +
3)(3x2e- 2z ).
(c) (D - 2)2(e-' tan x). (d) (D2 +
2D)(3e 2z esc x).
(e) (D2 - 2D +
6) (e- 3• log x).
14. Evaluate, by means of Corollary 24.6,
(a) (D - l)e x sin x. (b) (D + l)e- cos Z
x.
(c) (D - 2)2(e 2 'log x). (d) (D + 2)2(e- 3z2• tan x).
(e) (D - 3)2(e3• Arc sin x). (f) (D + 3)2(e- cot x).
Lesson 24-Answers 267

IS. Evaluate, by means of Corollary 24.7,

(a) (D3 + 3D + 1)(5e2,,). (b) (D3 + 3D + 1)(2e- 2,,).
(c) (D - 2)3(3e4,,). (d) (D2 - 3D 7)5e8 ". +
16. Prove Theorem 24.46 for n = 3. Hint. (D - TI)(D - T2)(D - T3)
D3 - (TI + T2 + T3)D2 + (TIT2 + TIT3 + T2T3)D - TIT2T3'
17. Prove Theorem 24.46 is false if the coefficients of P(D) in (24.47) are func-
tions of x. Hint. Show, for example, that (x 2D2 - 1) (xD - l)(xD + I)
by evaluating (x 2D2 - 1)e2" and (xD - 1)[(xD + 1)e2"1.
18. Prove the commutative law (24.5) is false if the coefficients of P(D) in
(24.47) are functions of x. Hint. Show, for example, that (D2+ xD)(3D)y
3D(D2 + xD)y, where y is a second order differentiable function of x.
Find the general solution of each of the following differential equations.
Follow the method of Lesson 24D.
19. y" - y' - 2y = e". 29. y'" - 3y' +2y = e-".
20. y" +
3y' +2y = 12e". 30. y" + 4y = 4x sin 2x.
21. y" +
3y' +2y = e'''. 31. +
y'" - 3y" 3y' - y = e".
22. y" + 3y' + 2y = sin x. 32. y'" - y' = e2 " sin 2 x.
23. y" +
3y' +2y = COB x. 33. 4y" - 5y' = x 2e-"
24. y" + 3y' + 2y = 8+ 6e"+ 2 sin x. 34. +
ylll - y" y' - y = 4e- 3,,+ 2x4.
25. y" - 2y' - 8y = lOe-" +
9xe". 35. y'" - 5y" + 8y' - 4y = 3e 2".
26. y" - 3y' = 2e 2" sin x. 36. Prove the equalities in (24.34).
27. y(4) - 2y" + y = x - sin x. 37. Prove the equality in (24.44).
28. y" +
y' = x 2 +
2x. 38. Prove the equality in (24.45).

ANSWERS 24
2. (a) 3x 2, 6x, 6, O. (b) 3 sin 2x, 6 cos 2x, -12 sin 2x, -24 cos 2x.
(c) '\IX, ix- I12 , -ix-3/2 , fx- 5/2 .
3. (a) cos 6 + tan 6, -sin 6 + sec 2 6, -cos 6 + 2 sec 2 6 tan 6.
(b) 62 + sin 6, 26 + COB 6, 2 - sin 6. (c) 6- 2, -26- 3 , 66-4 •
4. (a) t 2 + 3t + 1, 2t + 3, 2. (b) a cos 6t + b sin 6t, -a6 sin 6t + b6 cos 6t,
-a62 cos 6t - b6 2 sin 6t.
(c) a cos (6t + b), -a6 sin (6t + b), -a6 2 cos (6t + b).
5. (a) 4 sin 2x - 7 cos 2x. (b) -8e 3". (c) -48x. (d) 4x 2 - 4x + 2.
6. fa) a 3 sin ax - b3 cos bx. (b) 6e" + 6x 2 + 12x.
(c) 3xe" + 18 - 20x + 20x2•
7. (a) y" = e" + 4 + i(sin x - 3 cos x).
(b) y" = -xe" - 2e-". (c) y" = x 3 /3.

8.
(c)
(a)
y" = -he" + i·
3D3 + 5D2 + D + 2. (b) D4 - D2 - 6D + 5.
9. (a) D4 + 2D3 + 1. (b) 3D5 + 2D4 + 4D3 - 3D + 10.
10. (a) 4x 2e". (b) 2(x - 1)(6e2" + 1). (c) e-"(3 sec 2 x + 4 tan x + 4).
(d) (x 2 + 3x + 4)(x3 + 8).
II. (a) 6x 2e- 2". (b) (x - 1)(5x3 - 18x2 + +
18x 16).
(c) 6 cot x esc x.
12. (a) O. (b) 3x +3 2x2 - 5x + 12. (c) O.
13. (a) e2"(-4 sin 2x - 7 cos 2x). (b) 3e- 2'(2 - lOx -t- 9x 2).
(c) e-'(2 sec 2 x tan x - 6 sec 2 x + 9 tan x).
(d) 3e 2" esc x(csc 2 x + cot 2 x - 6 cot x + 8).
(e) e- 3 "(21 log x - x- 2 - 8x- I ).
268 OPERATORS AND LAPLACE TRANSFORMS Chapter 5

14. (a) eZ cos x. (b) -e-Z sin x. (c) _e 2z x- 2.

(d) 2e- 2z sec 2 x tan x. (e) xe3z (1 - x 2) -3/2.
(f) 2e-3z csc 2 x cot x.
15. (a) 75e 2z . (b) -26e- 2z . (c) 24e4z . (d) 235eSz.
19. y = cle-z + c2e 2z - le z.
20. y = 2ez + cle- 2z +
c2e-z .
21. y = n(e'" - 3ie"') + qe- 2z +
C2e- z .
22. y = n(sin x - 3 cos x) + qe- 2z +
c2e- z.
23. y = n(3 sin x +cos x) + cle- 2z +
c2e-z .
24. y = 4 + +
eZ i(sin x - 3 cos x) +
cle- 2z +
c2e- z .
25. y = qe4z +
2z
c2e- 2z - xeZ - 2e-Z.

26. y = - ; - (3 sin x + cos x) + CI + c2e3z. .

27. y = (q + c2x)e"'+ (C3 + c4x)e -z+ x - -4-'
smx
3
28. y = q + C2e + 3 .
-z

29. y = (q + c2x)e z + c3e- 2z + ie- z .

30. y = CI cos 2x + C2 sin 2x - (2x cos 2x.- sin 2x).

31. y = (q + C2X + C3x2 + x;) e'.

32. Y = CI+ C2e + cae + 121 + 9 cos 2x 520
z -z (- 7 sin 2X) 2.
e .
-z
33. y = q + C2e
5z14
+ (81x + 234x + 266) e729'
2

34. y = qe + sin x + cos X -

Z
C2 C3 + 2x4 + 8x3 + 48).
2 2z
35. y (q + c2x)e2z + C3e + -3x2e - .
Z

LESSON 25. Inverse Operators.

In Lesson 21, we found, by the method of undetermined coefficients, a

particular solution of the nth order linear differential equation
(25.1) P(D)y = Q(x),
where P(D) is the polynomial operator
(25.11)
and Q(x) is a function which consists only of such terms as b, x", eax,
sin ax, cos ax, and a finite number of combinations of such terms. Here
a and b are constants and k is a positive integer. In this lesson we shall
show how inverse operators may furnish a relatively easy and quick
method for obtaining this same particular solution.
Let
(25.12) Yc = CIYI CnYn + ... +
Lesson 25A MEANING OF AN INVERSE OPERATOR 269

be the complementary function of (25.1), i.e., let Ye be the general solu-

tion of P(D)y = 0, and let Yp' be a particular solution of (25.1). Therefore
the general solution of (25.1), by Theorem 19.3, is
(25.13) Y = Ye + Yp·
By following the method of undetermined coefficients as outlined in
Lesson 21, we obtained a particular solution YP that contained no term
which was a constant multiple of a term in Ye. However, there are in-
finitely many other particular solutions of (25.1). By Definition 4.66,
each solution which satisfies (25.1) and does not contain arbitrary con-
stants is a particular solution of (25.1). For example, the complementary
function of the differential equation
(a)
is
(b)

A particular solution of (a), found by the method of undetermined coeffi-

cients, is
(c) YP = -x 2 - 2.
Therefore the general solution of (a) is
(d)
You can verify that the following solutions, obtained by assigning arbi-
trary values to the constants Cl and C2 of (d) are, by Definition 4.66, also
particular solutions of (a).
(e) YP = _x 2 - 2, YP = _x 2 - 2 - 3ez ,
YP = _x 2 - 2 + 6e z - e-z , YP = _x 2 - 2 + 3e-z , etc.
Note, however, that every two functions differ from each other by terms
which are constant multiples of terms in Ye. In the proof of Theorem
65.6, we show that this observation holds for all particular solutions of
(25.1).
I.,n the remainder of this lesson and the next, whenever we refer to a par-
ticular solution YP of (25.1), we shall mean that particular solution from
which all constant multiples of terms in Ye have been eliminated. For in-
stance, in the above example, our particular solution of (a) is _x 2 - 2.

LESSON 25A. Meaning of an Inverse Operator.

Definition 25.2. Let P(D)y = Q(x), where P(D) is the polynomial
operator (25.11) and Q(x) is the special function consisting only of such
270 OPERATORS AND LAPLACE TRANSFORMS Chapter 5

terms as b, x", ea"" sin ax, cos ax, and a finite number of combinations of
such terms, * where a, b are constants and k is a positive integer. Then the
inverse operator of P(D), written as P-l(D) or I/P(D), is defined as
an operator which, when operating on Q(x), will give the particular solu-
tion YP of (25.1) that contains no constant multiples of a term in the
complementary function Ye, i.e.,
1
(25.21) P-1(D)Q(x) = YP or P(D) Q(x) = YP,

where YP is the particular solution of P(D)y = Q(x) that contains no

constant multiple of a term in Ye.
Comment 25.22. If we are given P(D) and Q(x), we now know how
to find P-1(D)Q(x). By Definition 25.2, it is the particular solution YP
of P(D)y = Q(x) that contains no constant multiples of terms in Ye.
Example 25.23. Evaluate

Solution. By Definition 25.2, (D 2 - 3D + 2)-lX = YP, where YP is

a particular solution of

(b) (D 2 - 3D + 2)y = x, Y" - 3y' + 2y = x,

that contains no constant multiples of terms in Ye.

By the method of Lesson 21 or 24D, we obtain the particular solution
x 3
(c) YP = 2" + 4·
Hence
(d)

Comment 25.24. By Definition 25.2, we conclude that

(25.25) D-nQ(x) == integrating Q(x) n times and ignoring

constants of integration.

Proof. By Definition 25.2, D-Q(x) = YP' where YP is the particular

solution of
(a)

that contains no constant multiples of terms in the complementary func-

"This restriction on Q(x) is a drastic one. Our definition, however, would not be mean-
ingful if Q(z) were not thus restricted, since it might then be difficult to exhibit !lp ex-
plicitly or implicitly in terms of elementary functions.
Lesson 25A MEANING OF AN INVERSE OPERATOR 271

tion Yc of (a). The complementary function of (a) is

(b)

and these terms result from retaining the constants of integration when
integrating (a) n times.

Example 25.251. Evaluate

(a) D- 2 (2x + 3).
Solution. By (25.25),

(b) D- 2 (2x + 3) = D- 1 f (2x + 3) dx = D- 1 (X 2 + 3x) = + .

+ 3x 2 /2 is a particular solution of
You can verify that x 3 /3
(c) y" = 2x + 3,

and that the complementary function of (c) is Yc = Cl + C2X.

Comment 25.26. We draw another important conclusion from Defi-

nition 25.2, namely that if P(D)y = 0, then
(25.27) YP = P-1(D)(0) = 0 or YP = (0) = O.

Proof. By Definition 25.2, P-1(D)(0) = YP' where YP is the particular

solution of
(a) P(D)y = 0

that contains no constant multiple of a term in the complementary func-

tion Yc of (a). This particular solution is YP = O.

Theorem 25.28.

(25.29) P(D)[P-1(D)Q] = Q or P(D) QJ = Q.

Proof. Let YP be a particular solution of
(a) P(D)y = Q.
Therefore
(b) P(D)yp = Q.
By (a) and Definition 25.2,
1
(c) YP = P(D) Q.

In (b) replace YP by its value in (c). The result is (25.29).

272 OPERATORS AND LAPLACE TRANSFORMS Chapter 5

LESSON 25B. Solution of (25.1) by Means of Inverse Operators.

As stipulated at the beginning of Lesson 25, the function Q(x) of (25.1)
may contain only such terms as b, xk, eaz , sin ax, cos ax, or a finite com-
bination of such terms, where a and b are constants and k is a positive
integer. We shall first consider each of these functions individually and
then combinations of them.
1. If Q(x) = bxl< and P(D) = D - ao, then (25.1) becomes

(a) (D - ao)Y = bxl<,

The complementary function of (a) is Yc = ceaoz . Hence a trial solution

YP' by the method of Lesson 21A, is

(b) YP = AIXk + A 2x k - 1 + A aXk - 2 + A 4x k - a + ... + Akx + A k+l .

Differentiation of (b) gives
(c) yp' = A1kxk - 1 + A 2(k - l)x k - 2 + Aa(k - 2)xk - a + ... + A k.
Therefore, by (b) and (c), YP will be a solution of (a) if

(d) yP' - aoyp = -aOAlx k + (A1k - aOA 2)x k - 1

+ [A 2(k - 1) - aoAalxk-2
+ [Aa(k - 2) - aoA41xk-a + ... + (Ak - aOA k+ l )
= bXk.

Equation (d) will be an identity in x, if

(e) -aOAI = b, Al = - -aob.

A _ A1k _ bk
2 - ao - - ao2·
bk(k - 1)
aoa
bk(k - l)(k - 2)
ao 4

Substituting (e) in (b), we have

(f) YP -_ - -b [ x k + -k x k-l + k(k - 2 1) X

k-2 +
· • • +-
k!],
k ao o.
ao ao ao ao
Lenon25B SOLUTION OF (25.1) BY MEANS OF INVERSE OPERATORS 273

We shall now prove that the same particular solution results if we

formally expand 1/(D - ao) in ascending powers of D and then perform
the necessary differentiations. By Definition 25.2, a particular solution
of (a) is
1,. 1 k
(g) YP = D - ao (bx ) = ( D ) (bx )
-ao 1 - -
ao

where the last series was obtained by ordinary division. Note that in
making this division, it is not necessary to go beyond the DkI ao k term since
DHl x" = O. Performing the indicated differentiations, we have

(h) YP = - !. [x" + !. X,.-I + k(k - 1) X"-2 + ... + ao" , ao ¢ 0,

ao ao ao 2
which is the same as (f). In general, it has been proved that if

(i)

then
(25.3) YP = 1 (bx " )
P(D)

= ( al a2 1 2 an
ao l+-D+-D + .. ·+-D
n) (bx")
ao ao ao
= -aob (1 + biD + b2 D 2 + ... + b"D" )x," ao ¢ O.

where (1 biD + b2D2 + + ... +

b"Dk)lao is the series expansion of the
inverse operator 1/P(D) obtained by ordinary division.
If k = 0, then (i) becomes P(D)y = b. Hence, by (25.3),
1 b
(25.31) YP = P(D) b = ao' ao ¢ O.

Example 25.32. Find a particular solution of

(a) y" - 2y' - 3y = 5, (D 2 - 2D - 3)y = 5.

Solution. Comparing (a) with (i) above we see that ao = - 3, b = 5,

k = O. Therefore, by (25.31),
(b)
-5
yp=a'
274 OPERATORS AND LAPLACE TRANSFORMS Chapter 5

Example 25.33. Find a particular solution of

(a) 4y" - 3y' + 9y = 5x 2,

Solution. Here P(D) = 4D2 - 3D + 9, ao = 9, b = 5. Hence, by

(25.3),
1 2
(b) YP = ( ) (5x )
9 1 _ D
3
+ !D2
9

= \)5 ( 1 + 3D - 3D2) x 2 = l(x

2
+ 1x - 1)·
Note that we did not need to go beyond the D2 term since D3(X 2) = O.

2. If Q(x) = bx" and P(D) = anD" + ... +

aID, so that 0, ao =
then D is a factor of P(D). Therefore, by Theorem 24.46, we can write
P(D) = D(anDn-1 + ... +
a2D +
al), where al O. If both ao = 0
and al = 0, then D2 is a factor of P(D), so that we can write P(D) =
D 2(anD n- 2 + ... +a3D +a2). In general, let Dr be a factor of P(D).
Then P(D)y = bxk can be written as

(a) P(D)y = DT(anDn-T + ... + aT+ID + aT)y = bx\ ar O.

Therefore, by Definition 25.2,

(b) YP = D T(D
an n-T + . . 1. + ar+l D + aT) (bx k) , aT O.

We shall now stipulate that the inverse operator in (b) means

(25.34) YP = 1 [ D
-Dr an n T
+ . . . 1+ aT+l D + aT (bx), k ]
aT O.

Comment 25.35. Since polynomial operators commute, we could also

have written, in place of (b) above,

(c)
YP = (anDn-T + ... +1 aT+ID + aT)DT (bXk)
1
=
n-T + ...
an D I+ aT+l + aT [D T (bi')]'
D ar O.

In effect we would now be integrating first, see (25.25), and then differ-
entiating. Although no harm results, following this order may introduce
terms in the solution YP that are constant multiples of terms in Ye. In
that event, we merely eliminate such terms. As exercises, follow the order
of procedure given in (c) to find a particular solution of each of the two
Lesson 25B SOLUTION OF (25.1) BY MEANS OF INVERSE OPERATORS 275

examples below. In both cases you will obtain a term in YP that is a

constant multiple of a term in Ye'
Example 25.36. Find a particular solution of
(a) Y" - 2y' = 5, (D2 - 2D)y = 5.

Solution. Here P(D) = D2 - 2D = D(D - 2). Therefore by Defi-

nition 25.2 and (25.34)
(b)

By (25.31), with b = 5, ao = -2,

1
(c) --(5) = -
D - 2
i·
Substituting (c) in (b), and then applying (25.25) to the result, we obtain

(d) YP = i (- = -!x.
Example 25.37. Find a particular solution of
(a)

Solution. Here P(D) = D 5 - D3 = D 3(D 2 - 1). Therefore, by

Definition 25.2 and (25.34),

(b) yp = 1[1 2 ]
D3 D2 _ 1 (2x) .

By (25.3), with ao = -1, b = 2,

(c) D2 1 (2X2) = -(1 D2) (2X2) = -2(1 + D2)X 2

_2(X2 + 2).
Substituting (c) in (b), and then applying (25.25) to the result, we obtain

(d) YP 1 [-2(x 2 + 2)1 =

= -D3 (X5 + -X3) .
-2 -
60 3
3. If Q(x) = bea", then (25.1) becomes P(D)y = beGX • We shall now
prove that a particular solution of this equation is
1 GX beGX
(25.4) YP = P(D) be = P(a) , P(a) O.

Note that the a in Pea) is the same as the exponent a in eGX •

276 OPERATORS AND LAPLACE TRANSFORMS Chapter 5

Proof. By (25.1) and (25.11), P(D)y = be"'" is equivalent to

(a) (anDn + an_IDn-1 + ... + aID + ao)Y == be""'.

Since P(a) 0, (D - a) cannot be a factor of P(D). This means that a

cannot be a root of the characteristic equation of (a). This in turn implies
that the complementary function of (a) cannot have a term e"'" in it.
Hence the trial solution YP of (a), by Case 1 of Lesson 21A, is

(b)

Differentiating (b) n times, we obtain

(c)

The substitution of (b) and (c) in (a) gives

(d) ananAe"'" +
an_Ian-IAe"'" + ... +
alaAe"'" +
aoAe"'" = be""',
Ae""'(anan + n
an_Ia - 1 + ... +
ala ao) = be""'. +
By Definition 24.52, the quantity in parenthesis is P(a). The last equa-
tion in (d) therefore simplifies to
b
(e) AP(a) ,= b, A = P{a) .

Substituting this value of A in (b) gives the expression on the extreme

right of (25.4). Hence be"'" /P(a) is a particular solution of P(D)y = be""'.
Example 25.41. Find a particular solution of

(a) ylll - y" + y' + y = 3e- 2""

Solution. Here P(D) = D3 - D2 + D + 1. Therefore, by (25.4),

with b = 3, a = -2,
1 -2'" 3e- 2'" 3e- 2'"
(b) YP = P(D) 3e = P(-2) = (-2)3 - (_2)2 - 2 +1
= _i.\e-2"'.
4. If Q(x) = b sin ax or b cos ax, no special difficulty arises. For, by
(18.84) and (18.85), we can change these functions to their exponential
equivalents and use (25.4). Easier, perhaps, is to apply the method of
Lesson 21B. We shall use this latter method to solve the following problem.
Example 25.42. Find a particular solution of

(a) y" - 3y' + 2y = 3 sin 2x, (D 2 - 3D + 2)y = 3 sin 2x.

Le8son 25B SOLUTION OF (25.1) BY MEANS OF INVERSE OPERATORS 277

Solution. Since e 2' z = cos 2x + i sin 2x, the imaginary part of a

particular solution of
(b) (D 2 - 3D + 2)y = 3e 2' z

will be a solution of (a). Here P(D) = D2 - 3D + 2. Therefore, by

(25.4), with b = 3, a = 2i,

3e 2iz 3e 2iz 3e 2iz (1 - 3i)

(c) yp = P(2i) = (2i)2 - 6i + 2 = -2(1 + 3i) . (1 - 3i)

= 3(1_-;03i) e 2iz = -h(1 - 3i)(cos 2x + i sin 2x)

= -h[(cos 2x + 3 sin 2x) + i(sin 2x - 3 cos 2x)].

The imaginary part of (c) is

(d) YP = *(3 cos 2x - sin 2x),

which is a particular solution of (a).

5. Exponential Shift Theorem for Inverse Operators.

Theorem 25.5. If P(D)y = ue az , where P(D) is a polynomial opera-

tor of order n and u is a polynomial in x, then

1 az az 1
(25.51) YP = P(D) ue = e P(D + a) u.
Proof.
(a)

By applying (25.29) to each side of (a), we can write (a) as

az
(b) P(D) (ue )] = eazP(D + a) [P(D 1+ a) u] .

By (24.58), e"ZP(D + a)u = P(D)(eazu). Applying this equality to the

right side of (b), with P(D 1+ a) u playing the role of u, we obtain, by (b),

(c) P(D)
az
(ue )] = P(D) [e
az
P(D 1+ a) u J.
By (24.21), we can write (c) as

(d)
278 OPERATORS AND LAPLACE TRANSFORMS Chapter 5

Let y represent the quantity in brackets. Then, by (d) and Comment 25.26,

(e) YP = (ue .... ) - etJZ P(D 1+ a) u 1= (0) = 0,

from which we deduce,
1 ( ....) .... 1
(f) YP = P(D) ue = e P(D + a) u.
Example 25.52. Find a particular solution of
(a) y" - 2y' - 3y = x 2e2", (D 2 - 2D - 3)y = x 2e 2".
Solution. Here P(D) = D2 - 2D - 3. Therefore
By
(b) 1 ( 2 2.. ) 1 ( 2 2.. ) Def.25.2,
YP = P(D) x e = D2 _ 2D _ 3 x e

2.. 1 2
=e P(D + 2) x (25.51),

2.. 1 2
Def.24.52,
=e (D + 2)2 _ 2(D + 2) _ 3x

2.. 1 2
=e D2 + 2D _ 3x Theorems 24.46
and 24.33,

(25.3),

Def. 24.13
and (24.11)
6. FOl'Dlula (25.4) can be used only if P(a) ,.r, O. What if P(a) = 01
If P(a) = 0, then (D - a) is a factor of P(D). Assume (D - a)F is a
factor of P(D). Therefore we can write
P(D) = (D - a)FF(D), F(a),.r, o.
We shall now prove that if P(D) = (D - a)FF(D), where F(a) ,.r, 0, and
P(D)y = be...., then
1 tJZ 1 .... bzFe....
(25.6) YP = P(D) (be ) == (D _ a)FF(D) (be ) = rlF(a) , F(a),.r, o.

Proof. We stipulate that the inverse operator in (25.6) shall mean

tJZ
(a) (D a)F (be )] .
Leuon25B SOLUTION OF (25.1) BY MEANS OF INVERSE OPERATORS 279

(Remark. As commented in 25.35, no hann results, other than more labor,

if the order of performing the inverse operations is interchanged. As re-
marked there, using this inverted order may introduce terms in Yt> that
are constant multiples of terms in Ye. In that event we merely eliminate
such terms.) By (25.4),
1 az beGS
(b) F(D) be = F(a)' F(a) o.
SUbstituting (b) in (a), we obtain

(c) 1
(D - a)r F(a) e
[b GS] .
By Definition 24.52, if P(D) = (D - aY, then P(D + a) = (D + a - aY
= Dr. In (25.51), let P(D) = (D - aY and u = b/F(a). Then (c) is
equivalent to
(d) as 1 [ b ]
e Dr F(a) .

By (25.25), D-r[b/F(a)] means integrate [b/F(a)] r times, ignoring con-

stants of integration. The first integration gives bx/F(a), the second
bx2/2!F(a), the third bx 3/3!F(a), and finally the rth integration gives
bxr /r!F(a). Hence (d) becomes

(e)

which is the same as the last expression in (25.6).

Example 25.61. Find a particular solution of

(a) y'" - 5ylI + 8y' - 4y = 3e 2s ,

Solution. Here P(D) = D3 - 5D2 + 8D - 4. Therefore, by Defi-

nition 25.2,
(b) 2S
YP = D3 _ 5D21+ 8D _ 4 (3e ).

If we now attempted to apply (25.4) to the right side of (b), we would

find that P(a), which here equals P(2), is zero. Hence we must resort to
(25.6). Since (D 3 - 5D2 + 8D - 4) = (D - 2)2(D - 1), we have by
(b),
(c) 1 3 2s
YP = (D - 2)2(D - 1) e .

Comparing (c) with (25.6), we see that b = 3, a = 2, r = 2, F(D) =

280 OPERATORS AND LAPLACE TRANSFORMS Chapter 5

D - 1 so that F(a) = F(2) = 2 - 1 = 1. Hence by (25.6), (c) becomes

(d)

Comment 25.62. The above solution (d) could also have been found
by the method of Lesson 21, Case 3. A good way to discover how much
easier the above method is, is to try to obtain (d) by means of that lesson.
Remark. As an exercise, show that if in place of (c), we had inverted
the order of the inverse operators and written

(e) YP = D 1[(D 2)2 (3e2:l:)],

and solved it by first applying (25.51) and then (25.4), the solution YP
would have contained additional terms that were constant multiples of
terms in Ye.
Example 25.63. Find a particular solution of
(a) y" + 4y' + 4y = 5e- 2",.
Solution. Here P(D) = D2 + 4D + 4 = (D + 2)2. Therefore, by
Definition 25.2,
(b) YP = (D +12)2(1) (5e-2",) .
If we now attempted to apply (25.4) to the right side of (b), we would
find P(a) = P( -2) = O. Hence we must resort to (25.6). Comparing
(b) with (25.6), we see that a = -2, r = 2, F(D) = 1, b = 5. Hence,
by (25.6) and (b),
(c)

7. Let Q(x) = Ql(X) +

Q2(X) + ... +
Qn(x). Therefore P(D)y =
Q(x) = Ql(X) +Q2(X) + ... +
Qn(X). We showed in comment 24.25,
that a solution YP of P(D)y = Q(x) is the sum of the respective particular
solutions of P(D)y = QiJ P(D)y = Q2, .• ', P(D)y = Qn. It follows,
therefore, that if P(D)y = Q, then

(25.7)

Because of (25.7) and the rules developed in this lesson, we are now in
a position to solve a linear differential equation

(25.71) any(n) + ... + alY' + aoy = Q(x),

Lesson 25B SOLUTION OF (25.1) BY MEANS OF INVERSE OPERATORS 281

where Q(x) may contain such terms as b, x\ eG >:, sin ax, cos ax, and com-
binations of such terms. Here a and b are constants and k is a positive
integer.
Example 25.72. Find the general solution of

(a) y" - y = x3 + 3x - 4, (D 2 - l)y = x 3 + 3x - 4.

Solution. Here P(D) = D2 - l. Therefore by Definition 25.2 and
(25.3)
(b)

By (25.7), we may apply 1/(D 2 - 1) = -(1 + D2) to each of the

terms in Q(x), namely to x 3, 3x, -5. Hence, by (b) and (25.7),

(c) yp = _(x 3 + 3x - 4 + 6x) = _(x 3 + 9x - 4).

The complementary function of (a) is

(d)

The general solution of (a) is therefore the sum of (c) and (d).
Example 25.73. Find the general solution of

(a) +
ylll - y" y' - y = 4e- 3 >: + 2x\
+
(D - D2 D - l)y = 4e- 3>:
3 + 2X4.
Solution. Here P(D) = D3 - D2 +D - l. Therefore by (25.7)

3 4
(b) YP = D3 _ D2\ D _ 1 (4e- >:) + D3 _ D2\ D _ 1 (2x ).

By (25.4)
1 4 -3>:
(c) 4e -3>: = .,--""":--__-"=e..-;;--;---;--;::-c--_=_
D3 - D2 +D - 1 (-3)3 - (-3)2 + (-3) - 1

By (25.3)

(d) -(1 _ D D2 _ D3) (2X4) = -(1 + D + D4)(2x4)

= -(2x 4 + 8x3 + 48).
Therefore, by (b), (c), and (d),

(e) YP = - (
-3>:
e 10 + 2X4 + 8x 3 + 48) .
282 OPERATORS AND LAPLACE TRANSFORMS ChapterS

The roots of the characteristic equation of (a) are I, i, -i. Hence

(f) Ye = cle'" + C2 sin x + Ca cos x.
The general solution of (a), is therefore the sum of (e) and (f).

EXERCISE 25
1. Evaluate (see Comment
(a) (D - 3)-I(X3 +
3x - 5). (b) (D - 1)-lx2 •
(c) (D2 - 3D +
2)-1 sin 2x. (d) (D2 - 1)-1(2x).
(e) (4D2 - 5D)-I(x 2e- z ).

2. Evaluate (see Comment 25.24).

(a) D-I(2x +
3). (b) D-3x . (d) D-2(2 sin 22:).
Find a particular solution of each of the following differential equations
by means of the inverse operator; use the appropriate method outlined in
numbers 1 to 7 of Lesson 25B.
3. y"+3y'+2y = 4. 14. y" - y = sin x. Hint. See 15.
4. y" + +
y' y = x 2. 15; y" - y = cos x.
5. y" - y = 2x. Hint. For 14 and 15, solve
6. y" - 3y' + 2y = x. y"-y=e....
7. y' - y = 3x 2 • +
16. y" - 3y' 2y = 3 sin x.
8. y" + y' = x 2 + 2x. +
17. y" a2 y = sin ax. Hint. See 18.
9. y'" - y" = 2x 3 • +
18. y" a2 y = cos ax.
10. +
y(4) - y'" y" = 6. Hint. For 17 and 18, solve y" +
U. y" + + 3y' 2y = 12ez . a2 y = e""", a F O.
12. y" + + 3y' 2y = e.... 19. 4y" - 5y' = x 2e- z .
13. y" + y = 3e- 2z . + +
20. y" y' y = 3x 2 ez•
21. y" - 2y' - 8y = 9xe • Z

22. y'"+ + +3y" 3y' y = e- z (2 - x 2).

23. y" + 4y = 4x sin 2x. Hint. See 24.
24. y" + 4y = 4x cos 2x.
Hint. For 23 and 24, solve y" + 4y = 4xe 2iz .
25. y" - y = 2ez •
26. y" + y' - 2y = 3e- 2z .
27. y'" - 5y" + 8y' - 4y = eZ •
28. y(4) - 3y'" - 6y" + 28y' - 24y = e2z .
29. y'" - lly" + 39y' - 45y = e3z .
30. y" - 2y' + y = 7e z .
31. y'" +y' = sin x. Hint. See 32.
32. y'" +y' = cos x.
Hint. For 31 and 32, solve y'" + y' = e....
33. +
y'" - 3y" 3y' - y = 2ez .
34. y" + + 3y' 2y = 8 + +
6e z 2 sin x.
35. y" + + 3y' 2y = 2(e-2z + x 2).
36. + +
y(5) 2y'" y' = 2x + + sin x cos x.
37. y' - 3y = x 3 + 3x - 5.
38. Solve Examples 25.36 and 25.37 by following the order suggested in Com-
ment 25.35. Show that the results differ from the text answers, if they do
differ, by a term which is a constant multiple of a term in Ye.
Lesson 26A PARTIAL FRACTION EXPANSION THEOREM 283

39. Solve Example 25.61, by interchanging the order of the operators in the
denominator of (c). Show that your answer differs from the text answer by
a term which is a constant mUltiple of a term in Ye'

ANSWERS 25
I. (a) -*(9x 3 + 9x + 33x -
2
34). (b) -(x 2 + 2x + 2).
(c) fo- cos 2x - * sin 2x. (d) -2x.
e-"
(e) 729 (81x
2
+ 234x + 266).
2. (a) x 2 + 3x. (b) x 4 /24. (c) e3 " 13. (d) -! sin 2x.
3. yp = 2. 21. YP = -xe".
3 -"
2 X e 2
4. x - 2x. 22. yp = 60 (20 - x ).
5. -2x. 23. yp = (sin 2x - 2x cos 2x).
x 3
6. YP = 2+4' 24. yp = (2x sin 2x + cos 2x).
7. YP = -3(x 2 + 2x + 2). 25. yp = xe".
B. yp = x 3 /3. 26. yp =

9. yp = -2(;: + + x 3 + 3X2).
10. YP = 3x 2 • 2B. yp = x 3 e2 "/30.
n. yp = 2e". 29. yp = -x 2e3"/4.
12. yp =
1 - 3, '"
--w-
e . 30. yp = 7x e"/2.
2

13. yp = ie-2 ". 31. yp = -!x sin x.

14. yp = -! sin x. 32. yp = -!x cos x.
IS. yp = -! cos x. 33. yp = x 3 e"/3.
16. yp = toCsin x + 3 cos x). 34. YP = 4+ e"+!(sin x - 3 cos x).
x
17. yp = - 2a cos ax. 35. YP = x
2
- 3x +t - 2xe-
2
".
2
lB. yp = :a sin ax. 36. yp = x2 + (cos x - sin x).

19. See lee). 37. See 1(a).

20. yp = te"(3x 2 - 6x + 4).
LESSON 26. Solution of a Linear Differential Equation by Means
of the Partial Fraction Expansion
of Inverse Operators.

LESSON 26A. Partial Fraction Expansion Theorem. In algebra,

an expression of the type
(a)
284 OPERATORS AND LAPLACE TRANSFORMS Chapter 5

can be simplified by the usual method of finding the least common de-
nominator. There results
5x - 1
(b) x 2 - 1·

Conversely, if we start with (b) we can expand it into the partial frac-
tions (a) by means of the partial fraction expansion theorelD of
algebra. We do this by factoring the denominator of (b), and then deter-
mining A and B so that
5x-l A B
(c)
x2 - 1 = x-I + xn·
Putting the right side of (c) under one common denominator, the equation
becomes
(d)
5x - 1 Ax + +
A Bx - B
x2 - 1 = x2 - 1

Since the denominators on both sides of the equal sign are alike, the A's
and B's must be chosen so that their respective numerators are also alike,
i.e., so that
(e) x(A + B) + (A - B) "" 5x - 1.

Equation (e) will be an identity in x if the coefficients of like powers of x

on both sides of the identity sign are equal. Hence we must have

(f) A +B = 5,
A - B = -1.

Solving (f) simultaneously, we find A = 2 and B = 3. Substituting these

values in (c) gives us the partial fraction expansion of (b), namely

(g) 5x - 1 = _2_+_3_.
x2 - 1 x-I x 1 +
We thus are able to go from (a) to (b) or from (b) to (a).
Comment 26.1. The general rule for determining the form of each
numerator in a partial fraction expansion of a quotient P(x)/Q(x), where
P(x) is of degree less than Q(x), will be evident from the following example.
Let
(h)

and let P(x) be a polynomial of degree less than Q(x). Then the partial
Lesson 26A PARTIAL FRACTION EXPANSION THEOREM 285

fraction expansion of P(x)/Q(x) will have the form

2
(i) P(x) = Bx + ex +
D Ex + +
F + + Gx2 + H
Q(x) x a + x3 b x2+ C + (x + C)2
I J K
+ x + d + (x + d)2 + (x + d)3
Note that:
1. Each numerator is a polynomial of degree one less than the degree of
the term inside the parenthesis of its denominator.
2. A term such as (x 2 +
C)2 has the exponent two outside the paren-
thesis. Hence (x 2 +
c) appears twice in the denominator, once as
(x 2 +
c), the second time as (x 2 C)2. +
3. A term such as (x +
d)3 has exponent three outside the parenthesis.
Hence (x +
d) appears three times in the denominator, once as (x d), +
the second time as (x +
d)2, and the third time as (x d)3. +
Comment 26.11. If the numerator of a fraction is a constant and
the denominator has distinct zeros, then there is a neat method which
will quickly give the numerators of a partial fraction expansion. Let

(a) f(x) = (x - T1)(X - T2) ... (x - TI;) ... (x - Tn),

where the zeros T1, T2, ... , Tn are distinct. By (a)

By comment 26.1, the partial fraction expansion of l/f(x) will have the
form
1 A1 A2 Ak An
(c) -=-
f(x)
-+-
x - T1
- + ... +--+
X - T2 X - Tk
... +-_.
X - Tn
Multiply (c) by (x - Tk). Since by (b), f(Tk) = 0, there results

(d) x - Tk = A 1(x - Tk) + ... + Ak + ... + An(x - Tk)

f(x) - f(Tk) x - T1 X- Tn
Let x --> Tk. The left side of (d) will approach 1/!'(Tk) and its right side
will approach AI;. Hence,
1
(e) Ak = f'(Tk)' k = 1, 2, ... , n.

Substituting (e) in (c), we obtain

1 1 1 1
(26.12) f(x) f'(T1)(X - T1) + !'(T2)(X - T2) + ... + f'(Tn)(X - Tn)
286 OPERATORS AND LAPLACE TRANSFORMS Chapter 5

Example 26.13. Find the partial fraction expansion of

3
(a) x2 - 1·

Solution. Comparing (a) with (26.12), we see that I(x) = x 2 - 1 =

(x - l)(x + 1). Therefore, rt = 1, r2 = -1, I'(x) = 2x, I'(rl) =
1'(1) = 2,I'(r2) = 1'( -1) = -2. Hence, by (26.12),

(b) _3_ _ 3 [ 1 _ 1 ].
x2 - 1 - 2(x - 1) 2(x + 1)
Example 26.131. Find the partial fraction expansion of
3
(a)
(x - 2)(X2 + X + 1)
Solution. We can if we wish factor

and then use (26.12). Or we can use the rules of partial fraction expansion
given in Comment 26.1. Using this latter method, we have

3 Bx+C
(b) (x _ 2)(X2 + X + 1) x - 2 + + x2 X 1

AX2 + Ax + A + Bx 2 + Cx - 2Bx - 2C
(x - 2)(x 2 + X + 1)
For (b) to be an identity in x, the numerator in the last fraction must
equal 3. Hence we must choose A, B, C, so that

(c) (A + B)x + (A 2
- 2B + C)x + (A - 2C) == 3.

Equating coefficients of like powers of x on both sides of the identity sign,

we obtain

(d) A +B = 0, A - 2B+C = 0, A - 2C = 3.

Solving (d) simultaneously for A, B, C gives

(e) A = t, B = -t, C= -l.

Substituting these values in (b), we have

(f)
3 3 3(x + 3)
(x - 2)(x 2 + X + 1) 7(x - 2) 7(x 2 + + 1)
X
Lesson 26A PARTIAL FRACTION EXPANSION THEOREM 287

Example 26.132. Find the partial fraction expansion of

1
(a)
(x + l)(x - 1)2
Solution. Here F(x) = (x + l)(x -
1)2 has a repeated root. Hence
we cannot use (26.12), but must fall back on the rules of partial fraction
expansion given in Comment 26.1. Therefore,
1 A B C
(b)
(x + l)(x - 1)2 = X+ 1 + x - I + (x - 1)2
A(x - 1)2 + B(x 2
- 1) + C(x + 1)
"'" (x - 1)2(x + 1) .

For (b) to be an identity in x, the numerator in the last fraction must

equal one. Hence we must choose A, B, and C so that
(c) A(x - 1)2 + B(x 2 - 1) + C(x + 1) "'" 1.
Instead of equating coefficients of like powers of x and then solving for
A, B, C as we did in the previous example, an alternate simpler method in
this case, is to let x = 1 in (c). There results
(d) 2C = 1,
If we let x = -1 in (c), we obtain

(e) 4A = 1, A=t.
If we let x = 0 in (c), we obtain

(f) A-B+C=1.

With A = :1, C = !, we find B = -1. Substituting these values in

(b), we obtain
1 1 1 1
(g)
(x + l)(x - 1)2 - 4(x + 1) 4(x - 1) + 2(x - 1)2

Comment 26.14. Analogously it can be shown that:

1. If YP = Q(x), then the inverse operator can also be expanded

into partial fractions just as if it were an ordinary polynomial.
2. Applying each member of a. partial fraction expansion of IIP(D) to
Q(x) and adding the results will give the same answer as will applying
IIP(D) to Q(x).
3. If P(D) has distinct factors, then it is permissible to take advantage
of (26.12) to find its partial fraction expansion.
288 OPERATORS AND LAPLACE TRANSFORMS Chapter 5

LESSON 26B. First Method of Solving a Linear Equation by

Means of the Partial Fraction Expansion of Inverse Operators.
We illustrate the method by means of examples.
Example 26.15. Find the general solution of

(a) y'" - 5y" + 8y' - 4y = 2e 4 %,

Solution. Here

(b) P(D) = D3 - 5D2 + 8D - 4 = (D - 2)2(D - 1).

Therefore, by Definition 25.2,

Following the partial fraction expansion method outlined above, we find

1 1 1 1
(d) (D - 2)2(D - 1) = D - 1 - D - 2 + (D - 2)2

Hence by (d) and Comment 26.14, (c) can be written as

(e) YP = D
4
1 (2e %) - D
4
2 (2e %) + (D 4
2)2 (2e %).

Applying (25.4) to each term on the right of (e), we obtain, with b = 2,

a = 4, and P(D) equal to the respective denominators,

(f)

which is a particular solution of (a). The roots of the characteristic equa-

tion of (a) are 2, 2, l. Hence

(g)

The general solution of (a) is therefore the sum of (f) and (g).
Example 26.16. Find a particular solution of

(a)

Solution. Here P(D) = D2 - l. Therefore by Definition 25.2,

3
(b) YP = D2 1 (2e %) = (D _ + 1) (2e %).
3

By Comment 26.14 and (26.12), with f(x) == P(D) = D2 - 1, f'(x) ==

Lesson 26B SOLUTION BY MEANS OF A PARTIAL FRACTION EXPANSION 289

P'(D) = 2D, Tl = 1, T2 = -1, P'(Tl) = 2, P'(T2) = -2, we can write

(b) as
1 (2 3",) 1
(c) YP = 2(D - 1) e - 2(D + 1) (2e3"').
Applying (25.4) to each term in the right of (c), we obtain

(d)

Comment 26.2. If an inverse operator is of the second order and has

distinct factors, we can generalize its partial fraction expansion. By (26.12),
if f(x) = x 2 - (Tl T2)X + +
TIT2 = (x - Tl)(X - T2), then
1 1 1 1
(a)
f(x) = - Tl)(X - T2) = f'(Tl)(X - + !'(T2)(X - T2)'
Tl)

where rl and T2 are distinct. Here f'(x) = 2x - (Tl + r2). Hence

f'(Tl) = 2Tl - (rl T2) = Tl - + r2 and f'(r2) = 2r2 - (Tl + T2) =
-rl + T2. Therefore (a) becomes
1
(b)

Analogously,
(26.21)

which may be taken as a formula for the partial fraction expansion of a

second order inverse operator whose zeros Tl and T2 are distinct.
Example 26.22. Find a particular solution of

(a) Y" + 2y' + 2y = 3xe"', (D 2 + 2D + 2)y = 3xe"'.

Solution. Here

(b) P(D) = (D 2 + 2D + 2).

Therefore by Definition 25.2

(c) + 2 (3xe"').
yp = D2 +
The roots of D2 + 2D + 2 are rl = -1 + i, T2 = -1 - i. Hence,

(d) Tl - T2 = -1 + i + 1 + i = 2i.
290 OPERATORS AND LAPLACE TRANSFORMS Chapter 5

Therefore, by (26.21) and Comment 26.14, we can write (c) as

(e) YP = [D + ; _ i (3xe Z
) - D + 11 + i (3Xe Z
)] •

Applying (25.51) to each term on the right of (e), we obtain, with u = 3x,
a = 1,

(f) YP = iJ Z
e D + ; _ i (3x) -
Z
e D + + i (3X)].
By (25.3), the series expansions of the inverse operators in (f) are [write
1/(D + 2 - i) = 1/[(2 - i)(1 +
D/(2 - i»] and then use ordinary
division]

2 i (1 - 2 i + .. -) and 2 i (1 - 2 i + .. -) .
Therefore, by (25.3), (f) is equal to
Z

(g) YP = --;
e -- 3 .( 1
2l 2 - l
[ - -D
-.
2 - l
+ . .. ) X

__3 (1 _ 2+i
2+i
...
z
3e [ x 1 x 1]
= 2i 2 - i - (2 - i) 2 - 2 + i + (2 + i) 2
z
3e (2i 8i) 3 Z 12 Z
= 2i "5 x - 25 = 5 xe - 25 e ,

which is a particular solution of (a).

LESSON 26C. A Second Method of Solving a Linear Equation by

Means of the Partial Fraction Expansion of Inverse Operators.

Example 26.3. Find a particular solution of

(a) Y" - 3y' + 2y = sin x, (D 2 - 3D + 2)y = sin x.

Solution. Here
(b) P(D) = D2 - 3D + 2.
Therefore, by Definition 25.2,
1 .
(c) yp = D2 _ 3D + 2 sm x,
where YP is a particular solution of (a). The zeros of D2 - 3D + 2 are
Lesson 26-Exerciae 291

rl = 2, r2 = 1. Hence,

By (26.21) and Comment 26.14, (c) becomes

1. 1.
(d) YP = D _ 2 sm x - D _ 1 sm x.
Let
1 . -1 .
(e) Ylp = D _ 2 smx, Y2p = D _ 1 sm x.
Therefore
(f) YP = Y1p + Y2p·

By Definition 25.2, Y1p and Y2p are particular solutions respectively of

(g) (D - 2)y = sin x, (D - l)y = -sin x.

A particular solution of each equation in (g) is respectively

(h) Yip = -
2sinx + cos x
5 Y2p =
sin x +2 cos x
Substituting these values in (f), we obtain
(i) yp = -1 sin x - 1- cos x + ,sin x + 'cos x
= ta sin x + fa cos x,
which is a particular solution of (a).
Comment 26.31. If we had solved (a) by the method of polynomial
operators, as outlined in Lesson 24C, we would have written (a) as

(D - 2)(D - l)y = Q(x),

and then, in effect, solved two linear equations in succession. By the above
method, we solve two linear equations independently.

EXERCISE 26
1. Find the partial fraction expansion of each of the following.

(a) - - .
2
(b) (x _ 2)(x2 -
3
1)
(c) 2x + 1.
x2 - 1 x2 - 1
2
(d) 2:/:
x 2 -1
+ 1. (e) (x _ 1)2
x
(f) (x2 +
x+1
1)(x - 1)
292 OPERATORS AND LAPLACE TRANSFORMS ChapterS

Find a particular solution of each of the following equations. Use

methods of Lessons 26B or C.
2. y" +y = Z2 +
e2",. 9. y'" - 3y"+ 3y' - y = eS •
3. y" - 2y' - 3y = 3e"". 10. y'" - y" + y' - y = 4e-3s 2Z4. +
4. y" - 2y' - 3y = 3 sin z. II. y'" - y" = 2z 3.
S. y" - 2y' - 3y = 3 cos z. 12. y(4) - y'"+ y" = 6.
6. y" + 4y = 2e-"'. 13. y(4) - 3y'" - 6y" +
28y' - 24y == els•
7. y" - 3y' +
2y = ze-S • 14. y'" - lly"+ 39y' - 45y = e3s•
8. y" - y' - 6y = z e2 ",. + IS. y(S) + 2y'" + y' = 2z sin z +cos z. +
ANSWERS 26
1 1 1 3 1
I. (8) Z _ 1 - z + 1• (b) z - 2 - 2(z - 1) + 2(z + 1)
3
(c) 2(z _ 1) + 2(z + 1)
1
(d) 2 +"2 3(1
z - 1 - .z
1)
+1 .
1 1 -z 1
+ (z - 1)2
(e) z _ 1 (0 z2 +1+z - 1.
2. yp = 2 + !e
Z2 - 2",.

3. -I\(2 cos z + sin z) - iI\(2 sin z - cosz).

4.,5. See 3.

II. yp = -2 (;: + z3 + 3Z2) •

12. yp = 3z2 •
13. yp = z3e 2"'/30.
14. yp = -z 2 e3 "'/4.
2
15. yp = Z2 + z8 (cos z - sin z).

LESSON 27. The Laplace Transform. Gamma Function.

LESSON 27A. Improper Integral. Definition of a Laplace Trans-

form. For a clearer understanding of the material of this lesson, a
knowledge of the meaning of the improper integral lex) dx is essential. 1:
We shall therefore briefly review this subject for you.
Let I(x) be a continuous function on the interval I: 0 x h,
h > O. If, as h -+ 00, the definite integral fo" I(x) dx approaches a finite
limit K, we say the improper integral 1:
I(x) dx exists and converges
to this value K. In that event we write

(27.1)
10
r I(x) dx = "_co
lim rIo I(x) dx = K.
10
If the limit on the right does not exist, we say the improper integral on
the left diverges and does not exist.
Lesson 27A IMPROPER INTEGRAL. LAPLACE TRANSFORM 293

Letf(x) be a continuous function on the interval I: 0 < x h, h > o.

If as h -+ 00 and E -+ 0, the definite integral feh f(x) dx approaches a
finite limit L, we say the improper integral f(x) dx exists and con- 1:
verges to this value L. In that event we write

(27.11)
Jr
0
f(x) dx = lim
A-+CIO
0-->0
lh e
f(x) dx = L.

If the limit on the right side does not exist, we say the improper integral
on the left diverges and does not exist.
Example 27.111. Determine whether the following integral exists.

(a) 10'" x ! 1 dx.

Solution.

(b) 10 hX ! 1 dx = log (h + 1).

As h -+ 00, log (h + 1) -+ 00. Hence the improper integral (a) diverges
and does not exist.
Example 27.112. Evaluate

(a) 10'" x 2 1 dx.

Solution.

(b) lim
h-->",
1
h

0 X
2 +1 1 dx = lim (Arc tan h -
h ___
Arc tan 0) = i·
Hence, by (27.1), the integral (a) exists and converges to 7r/2. We can
therefore write

o x2
1 '1"
7r
1 dx = "2. +
The following additional information will also be needed.

(27.113) (a) lim e-'''' = 0, if 8 > o.

Z-+QO 8
-.'"
(b) lim
Z-+CIO 8
= 0, if 8 > o.
(c) lim xe-'''' = 0, if 8 > o.
"'___ 8
(d) lim x"e-'''' = 0, if 8 > 0, n real.
"'---
294 OPERATORS AND LAPLACE TRANSFORMS Chapter 5

Finally we shall need the following theorem, which we state without

proof.
Theorem 27.12. If the improper integral

(27.121) i" e-,xf(x) dx, 0 x < 00,

converges for a value of s = so, then it converges for every s > so.
The integral (27.121), if it exists, is a function of s. It is called the
Laplace transform of f(x) and is written as L[f(x)]. Hence the following
definition:
Definition 27.13. Let f(x) be defined on the interval I: 0 x < 00.
Then the Laplace transforlD off(x) is defined by

(27.14) L[f(x)] = F(s) = i"e-,xf(X) dx,

where it is assumed that f(x) is a function for which the integral on the
right exists for some value of s.
Example 27.15. Find the Laplace transform of the functionf(x) = 1,
x o.
Solution. By (27.14), with f(x) = 1,

(a) L[I] = F(s) = 10" e-' x

dx

r e-'
h
= lim x
dx
h-.oo } 0

.
= h........
11m ( - -1e
s
_ax)lh
1)
0
-.h
= lim
h........
(
s S

= !,
s
if s > 0 [by (27.113)(a)].

Example 27.151. Find the Laplace transform of the function f(x) = £,

x 0, where x is the function defined by y = x, see (6.15).
Solution. By (27.14), with f(x) = X,

(a) L[x] = F(s) = i"xe-ex dx

= lim
h-+oo
f 0
h

xe-
ax
dx
Leuon 27B PROPERTIES OF THE LAPLACE TRANSFORM 295

= 1· [-.
1Dle . (
h .....'"
- -x- 2 "
8 8
l)r 0

= lim
(
he-·h
__ -.h
_ _ _e_+_
2
1) 2
h.....'" 8 8 8

= 1
82
, if 8 > 0 [by (27.113)(c) and (b)l.

LESSON 27B. Properties of the Laplace Transform.

Theorem 27.16. If the Laplace tran8form of It (x) converges for 8 > 8J

and the Laplace transform of h(x) converges for s > S2, then for 8 greater
'than the larger of 81 and S2,

(27.17) L[Ct!1 + c2h1 = cI L [lt1 + c2 L [h1,

where Cl and C2 are constants, i.e., the Laplace transformation is a linear
operator.
Proof. By (27.14) and the hypothesis'of the theorem,

(a) cI L [lt1 = Cl 10'"e-U:1t (x) dx, s> SI,

c2L (f21 = c21o"'e-""fz(x) dx, 8> S2·

Hence by Theorem 27.12, both of the above integrals exist for all 8 > 81
and S2' Therefore fbr s > SI and S2,

(b) cIL(fl1 + c2L[h1 = 10'"e-u'ct!l(X) dx + lo"'e-U:cz!2(X) dx

= Io"'e-""[Cdl + c2h1 dx = L[Ct!1 + c2h1.
By repeated application of (27.17), it can be proved that

(27.171) L[cdl + c2h + ... + c"f"l

= cIL[lt1 + c2L[h1 + ... + c"L(f,,].
We state the following theorem without proof.

Theorem 27.172. If It (x) and h(x) are each continuous functions of x

and/t = 12, then L[lt1 = L[h]· Conversely if L[fd = L[f21, and/t'!2 are
each continuou8 functions of x, then It = f 2.
296 OPERATORS AND LAPLACE TRANSFORMS Chapter 5

Definition 27.18. If F is the Laplace transform of a continuous func-

tion f, i.e., if
L[f) = F,
then the inverse Laplace transform of F, written as L -l[F), is f, i.e.,
L-1[F) = f·
The inverse Laplace transform, in other words, recovers the continuous
function f when F is given.
In Example 27.151 we showed that L[x) = l/s2. Hence the continuous
function which is the inverse transform of l/s2 is X, i.e., L-1 [1/8 2) = X.
Theorem 27.19. The inverse Laplace transformation is a linear opera-
tor, i.e.,
(27.191)
Proof. Let
(a)

where It and /z are continuous functions. Therefore by Definition 27.18,

(b)
By (27.17)
(c)

which, by (a), can be written as

(d)

By Definition 27.18, we obtain from (d),

(e) L-1[c1F 1 + C2F2) = cdl + C2/z'
Hence by (b), (e) becomes
(f) L-1[clF l + C2F2) = c1L-1[Fd + C2L-1[F2).
LESSON 27C. Solution of a Linear Equation with Constant Co-
efficients by Means of a Laplace Transform. The method we are
about to describe for solving the linear equation
(27.2) any(nl(x) + an_ly(n-ll(x) + ... + alY'(x) + aoY = f(x),
where ao, at. ... , an are constants and an 0, is known by the name of
the Laplace transform method. As the name suggests, it attains its
objective by transforming one function into another. However, unlike the
differential operator, the Laplace transform accomplishes the transforma-
Lesson 27C SOLUTION BY MEANS OF A LAPLACE TRANSFORM 297

tion by means of the integral in (27.14). We remark that the Laplace

method has one advantage over the other methods thus far studied for
solving (27.2): it will immediately give a particular solution of (27.2)
sati8fying given initial conditions.
By multiplying (27.2) by e-U and integrating the result from zero to
infinity, we obtain

(27.21) l"'e-""[a".y cnl + a,,_lyCn-ll + ... + alY' + aoY] dx

o ('"
= )0 e-u'f(x) dx, 8 > so,
which is equivalent to

(27.22) a". 10'"e-U;ycn l dx + an-l 10'"e-S:"yCn-1l dx + ...

+ al 10'"e-""y' dx + ao10'"e-Uy dx = 10'"e-""f(x) dx, s > 80·

By (27.14), we note that each integral in (27.22) is a Laplace transform.

Hence the equation can be written as
(27.23) anL[yCnl] + a"_lL[yCn-ll] + ... + a1L[y']
+ aoL[y] = L[f(x)], 8 > so.

Comment 27.24. Equation (27.23) is easily obtained from (27.2).

Insert an L after each constant coefficient in (27.2) and place brackets
around y(x), y'(x), ... , yCn\x) and f(x).
Our next task is to evaluate L[yCnl). By (27.14),

(27.241) L[ycnl] = i"'e-UyCnl dx.

If n = 0, (27.241) becomes (remember, n is an order, not an exponent)

(27.25) L[y) = i"'e-Uy dx, s > so.

If n = 1, (27.241) becomes

(27.26) L[y') = 10'"e-"'y' dx, s > so·

Integrating (27.26) by parts, we obtain with. u = e-''', dv = y'dx,

(27.27) L[y'] = lim + 8 10('"e-'''y dx

= lim [e-'''y(h) - y(O)] + 81'"0 e-Uy dx, s > 80.
298 OPERATORS AND LAPLACE TRANSFORMS Chapter 5

We now make an additional assumption that y(x), which is the solution

we seek, is a function such that *
(27.28) lim e-SXy(/')(x) = 0, k = 0, 1, 2, ... , n - 1, 8 > so.

Then (27.27) becomes, with the help of (27.25),

(27.29) L[y'] = -y(O) + sL[y].
If n = 2, (27.241) becomes
(27.3) L[y"] = 10'"e-It/:y" dx, s > 80·

The integral in (27.3) can be evaluated by two successive integrations by

parts. However an easier method is to make use of (27.29). If in it we
replace y by y', we obtain
(27.301) L[y"] = -y'(O) + sL[y'].
By (27.29), we can write (27.301) as
(27.31) L[y"] = -y'(O) + s(-y(O) + sL[yJ)
= s2L[y] - [y'(O) + sy(O)].
Similarly, by (27.31),
(27.311) L[y"'] = s2L[y'] - [y"(O) + sy'(O)].
By (27.29), (27.311) becomes
(27.32) L[y"'] = S2( -y(O) + sL[yJ) - [y"(O) +
sy'(O)]
= s3L[y] - [y"(O) + sy'(O) 2+
8 y(0)].

And in general it can be shown that

(27.33) L[y(nl] = snL[y] - [y(n-ll(o) + sy(n-2l(0) + ...
+ sn-2 y'(0) + sn-Iy(O)].
Hence, by (27.33), we can now write (27.23) as
(27.4) ansnL[y] - an[y(n-Il(o) + sy(n-2l(0) + ...
+ sn-2 y '(0) + sn-Iy(O)]
+ an_ISn-IL[y] - a n_l[y(n-2l(o) + sy(n-3l(o) + ...
+ sn-3 y '(O) + sn-2y(0)]

+a2s2L[y] - a2[y'(O) + sy(O)]

+aIsL[y] - aIY(O)
+aoL[y] = L[f(x)]

·See D. V. Widder, Advanced Calculus, Prentice-Hall (1961), for proof that the solu-
tion y(x), obtained by the Laplace transform method, is a function which satisfies
(27.28).
Lesson 27C SOLUTION BY MEANS OF A LAPLACE TlIANSFOllM 299

Collecting coefficients of like terms, (27.40) becomes

(27.41) [ans n + an_ISn- + ... + a2s2 + als + ao]L[y]

1

- [ans n- l + a n_IS n- 2 + ... + a2s + ally (0)

- [a nsn- + an_ISn- 3 + ... + a3s + a2]y'(0)
2

- [ans + a n_l]y(n-2)(0)
- any(n-l>(o) = L[f(x)].

Examine (27.41) carefully. L[y] is the Laplace transform of the solution

y(x) we seek, but which we do not know as yet; y(O), y'(O), ..• , y(n-l)(o)
are constants given by initial conditions; L[f(x)] is the Laplace transform
of the function f(x) which appears in the given linear equation (27.2).
Just as there are integral tables, there are tables of Laplace transforms
which will give L(f(x)]. By Definition 27.13, L[f(x)] is a function of s.
Now look again at (27.41). By solving it for L[y], the right side of the
resulting equation will be wholly a function of s. Let us call this function
G(s). The problem of finding the particular solution y(x) is thus reduced
to one of hunting in the tables for that continuous function y whose
Laplace transform is G(s), Le., L[y] = G(s); y = L -I[G(S)]. In short,
the Laplace transform method has changed the original differential equa-
tion involving derivatives, to an algebraic equation involving a function
of s.
Note. If the initial conditions give the values of y, y', y", ... , y(n-l)
at x = Xo ¢ 0, it is always possible to translate the axes by letting
x = +
xo, so that = 0 when x = Xo. The given differential equation
can then be solved in terms of and replaced afterwards by x - Xo.
In solving linear equations by the Laplace transform method, we may
use either the two equations (27.23) and (27.33), or the equivalent equa-
tion (27.41).
Comment 27.42. If f(x) = 0, then by (27.14)

(27.43) L[O] = i"'e-'%O dx = o.

Example 27.431. Use the method of Laplace transforms to solve
(a) y' + 2y = 0,
for which y(O) = 2.
Solution. Method 1. By use of (27.41). Comparing (a) with (27.2),
we see that n = 1, al = 1, ao = 2. By (27.43), L[O] = O. Hence (a)
becomes, with the aid of the given initial condition y(O) = 2 and (27.41),

(b) (s + 2)L[y] - (1)(2) = o.

300 OPERATORS AND LAPLACE TRANSFORMS Chapter 5

Therefore
(c) L[y] = - 2- .
s+2
Referring to a table of Laplace transforms (there is a short one at the end
of Lesson 27D) we find, see (27.82),

(d) therefore L[2e-

2
"'] = s ! 2.

Hence by (c), (d) and Theorem 27.172,

(e)

which is the required solution.

Method 2. By use of (27.23) and (27.33). As noted in comment 27.24, we
can write (a) as
(f) L[y'] + 2L(y) = L[O].

By (27.33), or directly from (27.29), (f) becomes

(g) sL[y] - yeO) + 2L[y] = L[O].

Solving (g) for L[y] and noting from (27.43) that L[O] = 0, and from the
initial conditions that yeO) = 2, we obtain
2
(h) L[y] = s + 2'
which is the same· as (c) above.

Example 27.44. Use the method of Laplace transforms to solve

(a) y" + 2y' + y = 1,

for which yeO) = 2, y'(O) = -2.

Solution. Comparing (a) with (27.2), we see that n = 2, a2 = 1,
at = 2, ao = 1. Hence by (27.41), (a) can be written as

(b) (S2 + 2s + I)L[y] - (s + 2)y(0) - y'(O) = L[I].

In Example 27.15, we found that L[I] = l/s. Substituting this value and
the initial conditions in (b), and then solving for L[y], we obtain

(c) L[ ] _ 2S2 + +
2s 1 _! _1_ _ 1
Y - s(s +
1)2 - s +s+1 (s + 1)2
Lesson 27C SOLUTION BY MEANS OF A LAPLACE TRANSFORM 301

Referring to a table of Laplace transforms we find, see (27.8), (27.82) and

(27.83),
Z
(d) L[I] = L[e-",] = s 1' L[xe- ] = (s : 1)2

Hence by (27.171),
(e ) L[1 + e-z - xe
-Z] 1
= "8
+ s +1 1 - (s +1 1)2
Therefore by (c), (e) and Theorem 27.172,
(f)
Alternate Method of Solution. As noted in Comment 27.24, we can
write (a) as
(g) L[y"] + 2L[y'] + L[y] = L[I].
By (27.33), or directly from (27.31) and (27.29), (g) becomes
(h) s2L[y] - y'(O) - sy(O) - 2y(O) + 2sL[y] + L[y] = L[I],
which simplifies to (b) above.
27.45. Solve
(a) y" + 3y' + 2y = 12e 2z ,
for which yeO) = 1; y'(O) = -1.
Solution. By (27.41), (a) can be written as
(b) (S2 + 3s + 2)L[y] - (s + 3)y(O) - y'(O) = L[12e 2Z ] = 12L[e 2Z ].
From a table of Laplace transforms, we find [see (27.82)], L[e 2z ] =
l/(s - 2). Substituting this value and the initial conditions in (b), we
obtain
(c)
8
2
8 + 3 3 1
L[y] = (s + 2)(s + l)(s - 2) = s + 2 - s + 1 + s - 2·

Referring to a table of Laplace transforms we find, see (27.82),

2Z
(d) L[e- ] = s 2'
Hence by (27.171),
(e) L[3e- 2Z _ 3e-z + e2z] = _3_ _ _
s+2
3_
8+1
+ _1_.
8-2
Therefore by (c) and (e)
(f)
302 OPERATORS AND LAPLACE TRANSFORMS Chapter 5

Alternate Method of Solution. As noted in Comment 27.24, we can

write (a) as
(g) L[y") + 3L[y') + 2L[y] = 12L[e 2 ,.].
By (27.31) and (27.29), (g) becomes
(h) 82 L[y) - y'(O) - sy(O) - 3y(O) + 38L[y] + 2L[y] = 12L[e 2 ,.],
which reduces to (b) above.

LESSON 27D. Construction of a Table of Laplace Transforms.

In this lesson, we shall find the Laplace transforms of a few simple functions.

(27.5) L[k] = re-'''kdx = k lim r"e-'''dx

10 "...... 10
= k lim
" ......
[-e-''']'' =
80
k lim
" ......
[-e-'" +.!]
8 8

= 8
if 8 > 0 [by (27.113)(a)].

(27.51) L[x"] = r"e-""x" dx = lim r"e-'''x'' dx

10 "...... 10
" ,,-I ( 1) ,,-2
= lim e-'''' ( -x _
2
_ n n -
3
X
"...... 8 8 8

_ .•• _ nIx _
8" 8,,+1
I"0
nl
= 8,,+1' 8 > 0, n = 1, 2,

[If 8 > 0, then by (27.113), e-'''h'' -+ 0 as h -+ 00, and when x = 0,

each term excepting the last equals zero.]

(27.52) L[e""') =
10r" e-''''e''''' dx = 10r" e<"-·>,,, dx = ,.....
lim [6<"->"'1"
.. a-8Jo
= _1_ lim (e<"-'>" - 1) = _1_, if 8 > a.
a-8"...... 8-a
By (18.84) and (27.52),
(27.53) L[sin ax) = L [e'"'' -;/_.""']

= ;iC ia - 8 iii)
1(2ai)
= 2i 8 2 a2 + = 82 +a a 2 •
Lesson 27D CONSTRUCTION OF A TABLE OF LAPLACE TRANSFORMS 303

Many theorems exist which will aid in the computations of the Laplace
transforms of more complicated functions. Unfortunately we cannot enter
into a detailed study of them. However, to show you the power of these
theorems, we shall state below a relatively simple one, and use it to find
the Laplace transforms of functions which otherwise would be more diffi-
cult to obtain.
Theorem 27.6. If
(27.61) F(8) = L[f(x)] = fo"'e-U:f(x) dx, 8 > 80,

then
(27.62) F'(8) = -L[xf(x)] = - fo"'e-""xf(X) dx, 8 > 80,

F"(8) = L[x 2f(x)] = f'" e-U:x2f(x) dx, 8 > 80,

. °
;(n l (8) = (_l) nL[xnf(x)] = (-I)nfo"'e-·"'xnf(X) dx, 8> 8o,

Note. Each Laplace transform in (27.62) can be obtained by differ-

entiating the previous function of 8 with respect to 8. The theorem in
effect states that if F(8) = L[f(x)], then one can find the Laplace trans-
form of xf(x) by differentiating -F(8); of x 2f(x) by differentiating F(8)
twice, etc.
E%ample 27.63. Compute

(a) 1. L[x sin ax], 2. L[x 2 sin ax].

Solution. By (27.53)
(b) L[sin ax] =
8
2 +a a 2 = F(8).

Hence by (b) and Theorem 27.6,

(c) L[x sin ax] = -F'(8) = (8 2 '

and
2 2
F"( ) 2a(38
(d) L[x 2sm
'
ax
]
= 8 = (8 2 +-a 2 )3a ) •
Another theorem, called the Faltung theorem., which is helpful in
evaluating integrals and one which we shall prove, is the following.
Theorem 27.7. If

(27.71) F(8) = L[f(x)] and G(8) = L[g(x)],

304 OPERATORS AND LAPLACE TRANSFORMS Chapter 5

then
(27.72) L Lf"f(X - t)g(t) dt] = L [.l"'f(t)g(x - t) dt]

= L[f(x)] . L[g(x)] = F(s) • G(s).

Proof. Let u = x - t. Therefore with x constant, du = -dt,·
u = 0 when t = x, and u = x when t = o. Making these substitutions
in the first integral in (27.72), it becomes

(27.73) L [1 0
- f(u)g(x - u) dU] = L Lf f(u)g(x - u) dU].

In the second integral of (27.73), replace the dummy variable of integra.-

tion u by t. The result is the second integral in (27.72). Hence we have
proved the first equality in (27.72).
By (27.14),

(27.74) L f(x - t)g(t) dt] = e-a", f(x - t)g(t) dt] dx.

= r [['"
}"'-O },_O
e-a"'f(x - t)g(t) dt] dx.

In Fig. 27.741, the shaded part indicates the region in the (x,t) plane
over which the integrations on the
right of (27.74) take place. The first
(-.-)
integration from t = 0 to t = x will
give the area of the vertical rectangle
of width dx.The second integration
from x = 0 to x = 00 will give the
area of the entire shaded region.
Let us now invert the order of in-
tegration, i.e., let us integrate first
over dx and then over dt. A first in-
t= tegration, see Fig. 27.741, from x = t
1IWJu..wu..w..lJ...Ju..wu..w........:u..uJ.LL._- to x = 00, will give the area of the
x
horizontal rectangle of width dt. A
Figure 27.741 second integration from t = 0 to
t = 00 will give the area of the entire
shaded region. Hence we can write (27.74) as

(27.75) L [t"
},_O f(x - t)g(t) dt] r [r - t)g(t) dx] dt
= },_O }"'_' e-a"'f(x

= r g(t) [r e-a"'f(x - t) dx] dt.

it_a }"'_'
Leuon27D CONSTRUCTION OF A TABLE OF LAPLACE TRANSFORMS 305

In the last integral of (27.75), let w = x - t. Then dw = dx (remember

t is a constant in this integration) f(x - t) = few) and e-''" = e-·(tD+ll.
Also w = 0 when x = t, and w -+ 00 when x -+ 00. Making all these
substitutions in this integral and then changing w, the dummy variable of
integration, back to x, we obtain

(27.76) L [I'"
1-0
f(x - t)g(t) de] = I'"'
1_0 1.._
get) [('"' e-·'"e-·tf(x)
0
dx] dt
= 10'"' e-"g(t) dtfo'"' e-u'f(x) dx

= 10'"' e-'''g(x) dx10'"' e-""f(x) dx

= L[g(x)] . L[J(x)].
By (27.71), the last expression on the right of (27.76) is G(s)F(s).
Example 27.77. Use Theorem 27.7 to show that

(a) 1 ('" ( - t)at b dt = albl a+b+l

'1 0 x (a+b+l)lx,
a> -1, b > -1, x > o.
(1 ab alb!
2. 10 (1 - e) t dt = (a b + +1)1'

Solution. Let
(b) f(x) = x a,
Then
(c) f(x - t) = (x - t)a,
By (27.72)

(d) L [10'" f(x - t)g(t) de] = L[f(x)] . L[g(x)].

Substituting (b) and (c) in (d), we obtain

a
(e) L [10'" (x - tte b dt] = L(x ) . L(x b).

From a table of Laplace transforms, we find [see (27.81)]

a b a! b! alb!
(f) L[x 1. L[x 1 = saH' Sb+l = sa+b+2

= L[ alb! a+b+l]
(a +b+ I)! x .
306 OPERATORS AND LAPLACE TRANSFORMS Chapter 5

Replacing the right side of (e) by its value in (f), we have by Theorem
27.172,
(g) {'" ( _ t)"t b dt = alb! ,,+b+1
10 x (a+b+l)!x.

To prove (a) 2, set x = 1 in (g).

Remark. The functions in (a) are known as beta functions.

Short Table of Laplace Transforrns

If f(x) = Then L[J(x)] = F(8)

k
(27.8) k -,
8
8> 0

(27.81) x
. n!
8,,+1' 8 > 0, n = 1, 2, ••.
1
(27.82) - - , 8>
8-a
a
.. az n!
(27.83) X e (8 _ a),,+1' 8 > a, n = 1, 2, ...
a
sin ax
(27.84)
82 +a 2
8
(27.85) cos ax
82 + a 2

2as
(27.86) x sin ax
(82 + a2 )2
2
8 - a2
(27.87) x cos ax
(82 + a2 )2
b
eO'" sin bx
(27.88)
(8 - a)2 + b2
eO., cos bx
s - a
a)2 + b2
(27.89)
(s -
(27.9) i'" f(x - t)g(t) dt L[J(x)] . L[g(x)] = F(8)G(s)

LESSON 27E. The GarnIna Function. By means of a function

called the gamma function, it is possible to give meaning to the factorial
function n!-ordinarily defined only for positive integers-when n is any
number except a negative integer. Formulas such as (27.81) and (27.83)
will then have meaning when n = 0 or ! or -t or 2:1, etc. We digress
momentarily, therefore, to give you those essentials of the gamma func-
tion which we shall need for our present and future purposes.
Lesson 27E THE GAMMA FUNCTION 307

Definition 27.91. The gaDlDla function of k, written as r(k), is

defined by the improper integral

(27.92) r(k) = 10'' 'X"'-l e -Z dx, k > O.

If k = 1, (27.92) becomes

(27.93) rei) = re-z

J0 dx = lim
A-+oo
= 1.

Integration of (27.92) by parts gives, with u = e-z , dv = Io

X -
1
dx,
lo
(27.94) r(k) = [e-;x ]: + i fo""x"'e- z dx, k > o.
By (27.113), the first term in the right of (27.94) approaches zero as
h - 00, and is zero when x = O. The second term by (27.92) equals
r(k + I)/k. Substituting these values in (27.94), we obtain

(27.95) r(k) = ir(k + 1), k> O.

Hence by (27.95)
(27.96) r(k + 1) = kr(k), k > O.

By (27.93), rei) = 1. Therefore by (27.96), when

(27.961) k = 1: r(2) = Ir(I) = 1 = I!
k = 2: r(3) = 2r(2) = 2 . 1 = 2!
k = 3: r(4) = 3r(3) = 3 . 2 . 1 = 3!
k = 4: r(5) = 4r(4) = 4·3·2·1 = 4!
And in general, when k = n, where n is a positive integer,
(27.97) r(n + 1) = n!
By (27.95)
(27.971) r(k) = r(k -: 1), k F- O.

In (27.971) replace k by k + 1. There results

2
(27.972) r(k + 1) = . : 1 ), k F- -1.

Now substitute in (27.971) the value of r(k +

1) as given (27.972). We
thus obtain
(27.973)
r(k
r(k) = k(k
2) +
+
1)' k F- 0, -1.
308 OPERATORS AND LAPLACE TRANSFORMS Chapter 5

If in (27.972) we replace k by k + 1, there results

3
(27.974) r(k + 2) = 2 ), k -2.

Now substitute (27.974) in (27.973). There results

(27.975) r(k) = k(k

r(k+ 3)+ 2)'
+ l)(k k 0, -1, -2.

In general it can be shown that

r(k + n)
(27.98) r(k) = k(k + 1)(k + 2) ... (k + n - 1)'
k 0, -1, -2, ... , -(n - 1).

By (27.971) and (27.98) we can extend the definition of r(k), which,

by (27.92), was defined only for k > 0, to include negative values of k,
provided k 0, -1, -2,···. For example, if k = -t, then by (27.971),
we can define
(27.981) r( -t) = -2rw·
If, therefore, we know the value of r(t), we then also know the value of
r( -t). And if we know the value of r( -t), we then also know the value
of r( -I). For by (27.971), we can define

(27.982) r( -I) = -!r( -t), etc.

Tables of values of the gamma function exist just as they do for sin x,
log x, or e"'. From such tables we find, for example, r(t) = y-:;r. Hence
by (27.981) and (27.982),
(27.983) r( -t) = -2y-:;r.
r( -I) = -!( -2v,r) = ty'1r, etc.
By means of the gamma function and its extended definition, we are
thus able to give meaning to nl when n is any number excepting a nega-
tive integer. For, by (27.97), tables of values of the gamma function, and
(27.96),
(27.984) 01 = r(l) = 1.
(-t)! = rw = v,r.
(-I)! = r(-t) = -2v,r.
WI = r(!) = trw = t0r.
Lesson 27E THE GAMMA FUNCTION 309

In Fig. 27.985, we have drawn a graph of the gamma function.

u 3

-1 2 3 4 k

-1

-2

-3

(1-4
Figure 27.985

Example 27.986. Compute

(a) L[X- 1/ 2 ].
Solution. By (27.14)
(b) L[X- 1 / 2 ] = 10'"e-"·x- 1 2
/ dx.

You will find in a table of integrals, that the value of the improper in-
tegral on the right of (b) is r(-!)/vs. By (27.984), r(!) = (-!)' = ..;;i.
Hence (b) becomes
(c)

Example 27.987. Compute

(8) L[Xn - 1 2
/ ], n = 1, 2, 3, ....
Solution. By (c) of Example 27.986
(b) L[X- 1 / 2 ] = ..;;i S-1/2 = F(s).
310 OPERATORS AND LAPLACE TRANSFORMS Chapter 5

Therefore by (b) and (27.62) of Theorem 27.6,

(c) L[Xl/2] = L[X(X- 1/ 2)] = -F'(s) = "7! S-3/2,

L[X3/2] = L[X2(X-l/2)] = F"(S) = S-5/2,

L[X5/2] = L[X3(X-l/2)] = -F"'(s) = 3· S-7/2.

And in general
(d) L[Xn - I
/
2] = L[xn (X- 1 / 2)] = (-I)"r")(s)

_ 1· 3· 5· .. (2n - 1)V"i -(n+l/2)

- 2" s , n = 1, 2, 3, ....

Example 27.988. Compute

(a) L[x-1/ 2ea",].
Solution. By (27.14)
(b) L[x- 1/ 2ea",] = i GO

e-(8-<J)Zx- I / 2 dx.

The integral on the right of (b) has the value rw/vs=a. By (27.984)
rw = (-i)! = v'1i-, Hence (b) becomes

(c) L[X -1/2 ea",] = V"i , s > a.

Vs=a
Additional Table of Laplace Transforms

If J(x) Then L[J(x)] = F(s) =

(27.99) -112 ..;;

x
Vs
,,-112 1·3·5· .. (2n - 1)V;: -(,,+112)
(27.991) x 2" 8, n = 1,2··'
n!
(27.992) x" s"+1 ' n > -1

(27.993) X
-112 4%
e
v;:
v'8="Ci
n-1I2 4% 1· 3·5· .. (2n - 1)V;: -(,,+112)
(27.994) X e 2" (s - a) , n - 1, 2, ...

n!
(27.995) X" e4% (s _ a),,+1 ' n > -1
Le880D 27-Exercise 311

Following exactly the method outlined in Example 27.987, you will find

(27.989) L[X,,-1/2e""'j = 1 . 3 . 5· .. J:n - nV1r (8 _ a)-(,,+1/2l,

8 > a, n = 1,2,3"" .
Remark. We see, therefore, that with the gamma function, the Laplace
transforms of x" and x"el n as given in (27.81) and (27.83) now also have
meaning when n > -1.

EXERCISE 27
I. Evaluate those of the following improper integrals which converge.

(a) 1'" x dx
(b) 1'"
0 (1
2x dx
+ x2)2
(c) ('" dx .
10 x2

1'"
o "';x2+ 1
dx
(d) 0 (x + 1)3
Find the Laplace transform of each of the functions 2-5.
2. sinh ax, x 51: O. 3. cosh ax, x 51: O. 4. ax + b, x 51: o.
5. I(x) = k, 0;:;;; x < 1.
= 0, x 51: 1.
With the aid of Theorem 27.6, find the Laplace transform of each of the
following functions.
6. x sinh ax. 7. x cosh ax. 8. x cos ax.
9. xe"Z. 10. x 2e"z. ll. x 3 eaz •

Use the method of Laplace transforms to find a solution of each of the

following differential equations satisfying the given initial conditions.
12. y' - Y = 0, y(O) = 1.
13. y' - Y = eZ , y(O) = 1.
14. y'+ y = e-z , y(O) = 1.
IS. y"+ + 4y' 4y = 0, y(O) = 1, y'(O) = 1.
16. y" - + 2y' 5y = 0, y(O) = 2, y' (0) = 4.
17. 3y'"+ + 5y" y' - y = 0, y(O) = 0, y'(O) = 1, y"(O) = -1.
18. y" - 5y' - 6y = e3z , y(O) = 2, y'(O) = 1.
19. y" - y' - 2y = 5 sin x, y(O) = 1, y'(O) =-1.
20. y'" - 2y" + y' = 2ez +
2x, y(O) = 0, y'(O) = 0, y"(O) = O.
21. y" + +y' y = x 2 , y(O) = 1, y'(O) = 1.
22. Evaluate each of the following gamma functions.
(a) r(6). (b) r(7). (c) r(-5/2). (d) r(5/2). (e) r(7/2).
23. Evaluate each of the following factorial functions.
(a) (-i)! (b) (-i)! (c) (t)! (d) (i) I
24. Verify the correctness of (27.989). Hint. See the solution of Example 27.987.
312 OPERATORS AND LAPLACE TRANSFORMS Chapter 5

25. In Example 27.986, we used the fact that

10'" x- 1I2
e-U: dx = rw/v'B, 8 > O.

Prove it. Hint. Make the substitution u = 8X. Then use (27.92).
26. Prove, in general, that

roo ,,-,zdx _ r(n+ 1)

Jo x e - 8,,+1 ' n> -1,8> O.

See hint in 25.

ANSWERS 27
I. (a) Diverges. (b) 1. (c) Diverges. (d) i.

2. _a2
a
> a.
8
> a. 4. -8-
a+
b8
82 ,8 3. 82 _ a2 ' 8 2 ' 8> O.

5.
k(1 - e-')
8'
6
• (82 _
2as
a2 )2
7 8
• (82 -
2
+aa22)2
2 2
8 - a 9 1 10 2 31
B. (82 + a2)2 . • (8 - a)2 • (8 - a)3 II. (8 _ a)4

12. y = eZ. 13. y = (x l)e • + z

14. y = (x l)e- z •+ 15. y = (1 + 3x)e-2z •

16. y = (2 cos 2x sin 2x)ez •+
I7. Y = 16 e
9 Z/3+(X"4 - 9) 16 e
-z'
.
lB. y = H-e 6z + He-z - -he3z•
19. y = !e2z + 1e-z - i sin x+ i cos x.
20. y = x 2 + 4x + 4 + (x 2 - 4)ez •
21. y = Va x + -3-
e-z/2 ( cos 2 7Va sm
. Va2 x) + x2 - x2.
22. (a) 5! (b) 6! (c) -h....,r.i (d) i0f (e) 1.j-0i.
23. (a) t0f (b) -h....,r.i (c) hli (d) ¥-Y1r.
 Chapter 6

Problems Leading
to Linear Differential Equations
of Order Two

In this chapter we shall consider the motion of a particle whose equa-

tion of motion satisfies a differential equation of the form
2
d x
(a) dt 2 + 2r dx 2
dt + 000 x = f(t),

where f(t) is a continuous function of t defined on an interval I, and r

and 000 are positive constants. If f(t) == 0, then (a) simplifies to
d 2x
(b) dt 2 + 2r dx 2
dt + 000 x = O.

If r = 0, then (a) becomes

d2 x
(c) dt 2 + 000 2x = f(t).
Finally if both r = 0 and f(t) == 0, then (a) becomes
2
d x
(d) dt 2 + 000 2x = O.
In the lessons which follow we shall name and discuss each of these four
important equations (a), (b), (c), and (d).

LESSON 28. Undamped Motion.

LESSON 28A. Free Undamped Motion. (Simple Harmonic Mo-

tion.) Many objects have a natural vibratory motion, oscillating back
313
314 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

and forth about a fixed point of equilibrium. A particle oscillating in this

manner in a medium in which the resistance or damping factor is neg-
ligible is said to execute free undamped motion, more commonly called
simple harmonic motion. Two examples are a displaced helical spring and
a pendulum. There are various ways of defining this motion. Ours will
be the following.
Definition 28.1. A particle will be said to execute simple harmonic
motion if its equation of motion satisfies a differential equation of the
form
(28.11)

where Wo is a positive constant, and x gives the position of the particle as

a function of the time t.
By the method of Lesson 20D, you can verify that the solution of
(28.11) is

(28.12) x = CI cos wot + C2 sin wot.

By (20.57) we can also write the solution (28.12) in the form

(28.13)
or by (20.58) with +0 replacing - 0, in the form

(28.14) x = VC12 + C2 2 cos (wot + 0) = C cos (wot + 0).

Hence an equivalent definition of simple harmonic motion is the following.
Definition 28.141. Simple harmonic motion is the motion of a
particle whose position x as a function of the time t is given by any of the
equations (28.12), (28.13), (28.14).
Example 28.15. A particle moving on a straight line is attracted to
the origin by a force F. If the force of attraction is proportional to the
distance x of the particle from the origin, show that the particle will
execute simple harmonic motion. Describe the motion.
Solution. By hypothesis

(a) F = -kx,

where k > 0 is a proportionality constant. The negative sign is necessary

because when the particle is at Pi, see Fig. 28.16, x is positive and F
acts in a negative direction; when it is at P2, x is negative and F acts in
a positive direction. F and x, therefore, always have opposite signs.
Lesson 28A FREE UNDAMPED MOTION. (SIMPLE HARMONIC MOTION) 315

Hence by (16.1) with 8 replaced by x, (a) becomes

d 2x d 2x k
(b) F= m dt 2 = -kx, dt 2 = --x.
m

±.
I I I I
-c p. 0 PI c

Figure 28.16

Since k and m are positive constants, we may in (b) replace kim by a

new constant wo 2 • The differential equation of motion (b) is then
d2 x
(c) dt 2 + Wo
2
x = 0,

which is the same as (28.11). By Definition 28.1, therefore, the particle

executes simple harmonic motion.
Description of the Motion. The solution of (c), by (28.13), is

(d) x = c sin (wot + 0).

Differentiation of (d) gives
dx
(e) dt = v = cwo cos (wot + 0),
where v is the velocity of the particle. Since the value of the sine of an
angle lies between -1 and 1, we see from (d) that Ixl Icl. Hence the
particle can never go beyond the points c and -c, Fig. 28.16. These
points are thus the maximum displacements of the particle from the
origin O. When Ixl = Icl, we have, by (d), Isin (wot 0)1 = 1, which +
implies that cos (wot +/I) = o. Therefore, when Ixl = Icl the velocity v,
by (e), is zero. We have thus shown that the velocity of the particle at
the end points ± c, is zero.
When x = 0, i.e., when the particle is at the origin, then, by (d),
sin (wot+ 0) = 0, from which it follows that Icos (wot 6)1 = 1. Since +
the value of the cosine of an angle lies between -1 and 1, we see, by (e),
that the particle reaches its maximum speed Ivl = Icwol, when it is at
the origin. For values of x between 0 and Icl, you can verify, by means
of (d) and (e), that the speed of the particle will be between 0 and its
maximum value Icwol; its speed increasing as the particle goes from c,
where its speed is zero, to the origin, where its speed is maximum. Mter
crossing the origin, its speed decreases until its velocity is again zero at
-c. The particle will now move in the other direction-remember the
force which is directed toward the origin never ceases to act on the par-
316 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

ticle-its speed increasing until it reaches its maximum speed at 0, and

then decreasing until it is zero again at c. The particle thus oscillates back
and forth, moving in an endless cycle from c to -c to c.
Example 28.17. A particle P moves on the circumference of a circle
of radius c with angular velocity Wo radians per second. Call Q the point
of projection of P on a diameter of the circle. Show that the point Q
executes simple harmonic motion.
Solution. See Fig. 28.18. Let
Po(xo,Yo) be the position of the particle at time t = 0,
P(x,y) be the position of the particle at any later time t,
& be the central angle formed by a diameter, taken to be the x axis,
and the radius OP o.

P(x,y)

x axis

Figure 28.18

In time t the central angle through which the particle has rotated is wot.
(For example if Wo = 2 radians/sec, then in i second, P has swept out a
central angle equal to 1 radian; at the end of 2 seconds the central angle
swept out is 4 radians; at the end of t seconds, the central angle swept

particle at t = °
out is 2t radians.) The projections on the diameter of the positions of the
and t = t, are respectively, Qo(xo,O) and Q(x,O). It is
evident from Fig. 28.18 that
(28.2) x = C cos (wot + &),
an equation which expresses the position of the particle's projection Q on
a diameter, as a function of the time t. A comparison of (28.2) with (28.14)
shows that they are alike. Hence the point Q executes simple harmonic
motion.
The alternate form (28.13) may be obtained by measuring the initial
angle 0 from the y axis instead of from the x axis, Fig. 28.22. From the
figure we see that
(28.21) x = c cos & + wot)) = -c sin (wot + &)
== C sin (wot + &).
Lesson 28B DEFINITIONS: SIMPLE HARMONIC MOTION 317

yaxis

1r
2

:taxis

Figure 28.22

Comment 28.23. Description of the Motion. As P moves once

around the circle, its projection Q (Fig. 28.18) moves to one extremity of
the diameter, changes direction and goes to the other extremity, changes
direction again and returns to its original position headed in the same
starting direction. Note the similarity of Q's motion to the motion of the
particle in Example 28.15.

LESSON 28B. Definitions in Connection with Simple Harmonic

Motion. For convenience, we recopy the three solutions of (28.11).
Each, by Definition 28.141, is the equation of motion of a particle exe-
cuting simple harmonic motion.

(28.24) x = Cl + C2 sin wot.

cos wot
(28.25) x = c sin (wot + 8).
(28.26) x = c cos (wot + 8).

If you will refer to Figs. 28.16 and 28.18, and reread the "description of
the motion" paragraphs of Example 28.15 and Comment 28.23, the
definitions which follow will be more meaningful to you.
Definition 28.3. The center 0 of the line segment on which the par-
ticle moves back and forth is called the equilibrium position of the
particle's motion.
Definition 28.31. The absolute value of the constant c in (28.25) and
in (28.26) is called the amplitude of the motion. It is the farthest dis-
placement of the particle from its equilibrium position.
Comment 28.32. By (28.13), Icl = VC12 +
C22. If therefore the solu-
tion of (28.11) is written in the form (28.24), then the amplitude of the
motion is VC12 + C2 2 •
318 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

Definition 28.33. The constant a in (28.25) and in (28.26) is called

the phase or the phase angle of x. [Note that when t = °
in (28.26),
x = c cos a, and from Fig. 28.18, we observe that c cos a is the initial
position (xo,O) of the particle's projection.]
If, in (28.26), t = 0, 27r/wo, 47r/wo, ... , 2n7r/wo, n an integer, then for
each t, x = c cos a. This means that in time 27r/wo, the particle has
made one complete revolution around the circle and is back at its starting
position, headed in the same starting direction. Or equivalently, the par-
ticle has made one complete oscillation along a diameter or a line segment
and is back at its starting position headed in the same starting direction.
Hence the following definition.

Definition 28.34. The constant

27r
(28.35) T=-,
Wo

where Wo is the constant in (28.24), (28.25), and (28.26), is called the

period of the motion. It is the time it takes the particle to make one
complete oscillation about its equilibrium position. If the constant Wo is
the angular velocity of a particle moving on the circumference of a circle,
then the period is the time required for the particle to make one com-
plete revolution or equivalently the time it takes the particle's projection
on the diameter to make one complete oscillation about the center of the
circle.
For example, if a particle makes two complete revolutions, or equiva-
lently two complete oscillations, in one second, i.e., if Wo = 47r rad/sec,
then its period by (28.35) is ! second; if it makes t of a revolution or of a
complete oscillation in one second, i.e., if Wo = 7r/2 rad/sec, then its period
is 4 seconds. Note that the reciprocal of T gives the number of complete
revolutions or oscillations made by the particle in one second. For example
when the period T, which is the time required to make one complete
revolution or oscillation, is !, the particle makes two complete oscilla-
tions in one second; when the period T = 4, the particle makes t of a.
complete oscillation in one second. Hence the following definition.

Definition 28.36. The constant

1 Wo
(28.37) p=-=-
T 27r

is called the natural (undamped) frequency of the motion. It gives

the number of complete revolutions or cycles made by the particle in a
unit of time, or equivalently it gives the number of complete oscillations
made by the particle on a line segment in a unit of time.
Les80n 28B DEFINITIONS: SIMPLE HARMONIC MOTION 319

An alternative definition of natural undamped frequency that is often

used is the following.
Definition 28.38. The constant Wo in (28.24) to (28.26), (which can
be thought of as the angular velocity of a particle moving on the circum-
ference of a circle) is also called the natural (undamped) frequency
of the motion.
Comment 28.39. We shall use II when we wish to express the fre-
quency of the motion in cycles per unit of time and use Wo when we wish
to express it in radians per unit of time. In the example after Definition
28.34 where Wo = 411" rad/sec, the frequency II, by (28.37), is thus 2
cycles/sec; the frequency Wo is 411" radians/sec.
Comment 28.4. A Summary. The differential equation of motion of
a particle executing simple harmonic motion has the form
d 2x
dt 2 + Wo 2x = O.

Its solutions may be written in any of the following forms.

x = Cl cos wot + C2 sin wot,

x = C cos (wot + 6),
x = C sin (wot + 6).

They are the equations of motion of a particle executing simple harmonic

motion. The motion can be looked at as the motion of the projection on a
diameter, of a particle moving with angular velocity Wo on the circum-
ference of a circle of radius c. Or it may be looked at as the motion of a
particle attracted to an origin by a force which is proportional to the dis-
tance of the particle from the origin. The particle moves back and forth
forever across its equilibrium position. The farthest position reached by
the particle from its equilibrium position is given by c. Its speed at c
and -c is zero; at the origin its speed is greatest. The time required to
make a complete oscillation is T = 211"/wo; the number of complete oscilla-
tions in a unit of time is l/T.
Example 28.5. A particle executes simple harmonic motion. The
natural (undamped) frequency of the motion is 4 rad/sec. If the object
starts from the equilibrium position with a velocity of 4 ft/sec, find:
1. The equation of motion of the object.
2. The amplitude of the motion.
3. The phase angle.
4. The period of the motion.
5. The frequency of the motion in cycles per second.
320 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

Solution. Since the particle executes simple hannonic motion, its

equation of motion, by (28.26), is
(a) x = e cos (wot + a), v = : = -woe sin (wot + a).
The frequency of 4 rad/sec implies, by Definition 28.38, Wo = 4. Hence
(a) becomes
(b) x = e cos (4t + a), v = -4e sin (4t + a).
The initial conditions are t = 0, x = 0, v = 4. Substituting these values
in the two equations in (b), we obtain

(c) 0= e cos a, 4 = -4e sin a.

Since the amplitude e 0, we find from the first equation in (c), a =
±7r/2. With this value of a, the second equation in (c) gives e = =Fl.
Hence the equation of motion (b) becomes

(d) x = cos (4t - i) or x = _·cos (4t + i) ,

which is the answer to 1. The answers to the remaining questions are:
2. Amplitude of the motion, by Definition 28.31, = 1 foot.
3. The phase angle, by Definition 28.33, = -7r/2 radians.
4. The period of the motion, by Definition 28.34, = 7r/2 seconds.
5. The frequency of the motion in cycles per second, by Definition
28.36, equals 2/7r cps.
Example 28.51. A particle executes simple hannonic motion. The
amplitude and period of the motion are 6 feet and 7r/4 seconds respec-
tively. Find the velocity of the particle as it crosses the point x = -3 feet.
Solution. Since the particle executes simple hannonic motion, its
equation of motion, by (28.26), is

(a) x = e cos (wot + a).

The amplitude of 6 feet implies, by Definition 28.31, that e = 6 (or -6).
The period of 7r/4 seconds implies, by Definition 28.34, that
27r 7r
(b) - = -4' Wo = 8.
Wo

Hence the equation of motion (a) becomes, using e = 6,

(c) x = 6 cos (8t + a), = -48 sin (8t + a).

Les8on8 28A and B-Exercise 321

When x = -3, we find from the first equation in (c),

(d) cos (8t + a) = -I, 8t + a= 120° or 240°,
and from the second equation in (c)

(e) v = = -48 sin (120° or 240°)

= -48(±hl3) = =F24V3 ft/see.
The plus sign is to be used if the body is moving in the positive direction,
the minus sign if the body is moving in the negative direction.

EXERCISE 28A AND B

1. Verify the accuracy of each of the solutions of (28.11) as given in (28.12),
(28.13), and (28.14).
2. If a ball were dropped in a hole bored through the center of the earth, it
would be attracted toward the center with a force directly proportional to
the distance of the body from the center. Find: (a) the equation of motion,
(b) the amplitude of the motion, (c) the period, and (d) the frequency of the
motion. NOTE. This problem was originally included in Exercise 16A, 40.
3. A particle executes simple harmonic motion. Its period is 211" sec. If the
particle starts from the position x = -4 ft with a velocity of 4 ft/sec, find:
(a) the equation of motion, (b) the amplitude of the motion, (c) the frequency
of the motion, (d) the phase angle, and (e) the time when the particle first
crosses the equilibrium position.
4. A particle executes simple harmonic motion. At t = 0, its velocity is zero
and it is 5 ft from the equilibrium position. At t = t, its velocity is again 0
and its position is again 5 ft.
(a) Find its position and velocity as functions of time.
(b) Find its frequency and amplitude.
(c) When and with what velocity does it first cross the equilibrium position?
5. A particle executes simple harmonic motion. At the end of each ! sec, it
passes through the equilibrium position with a velocity of ±8 ft/sec.
(a) Find its equation of motion.
(b) Find the period, frequency and amplitude of the motion.
6. A particle executes simple harmonic motion. Its amplitude is 10 ft and its
frequency 2 cps.
(a) Find its equation of motion.
(b) With what velocity does it pass the 5-ft mark?
7. A particle executes simple harmonic motion. Its period is a7l" sec and its
velocity at t = 0, when it crosses the point x = Xl, is ±VI. Find its equa-
tion of motion.
8. A particle executes simple harmonic motion. Its frequency is 3 cps. At
t = 1 sec, it is 3 ft from the equilibrium position and moving with a velocity
of 6 ft/sec. Finds its equation of motion.
9. A particle executes simple harmonic motion. When it is 2 ft from its equilib-
rium position, its velocity is 6 ft/sec; when it is 3 ft, its velocity is 4 ft/sec.
Find the period of its motion, also its frequency.
322 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

10. A particle moves on the circumference of a circle of radius 6 ft with angular

velocity ."./3 rad/sec. At t = 0, it makes a central angle of 150· with a fixed
diameter, taken as the x axis.
(a) Find the equation of motion of its projection on the diameter.
(b) With what velocity does its projection cross the center of the circle?
(c) What are its period, amplitude, frequency, phase angle?
(d) Where on the diameter is the particle's projection at t = o?
11. A particle moves on the circumference of a circle of radius 4 ft. Its projection
on a diameter, taken as the x axis, has a period of 2 sec. At x = 4, its velocity
is zero. Find the equation of motion of the projection.
12. A particle weighing 8 Ib moves on a straight line. It is attracted to the origin
by a force F that is proportional to the particle's distance from the origin. If
this force is 6 Ib at a distance of -2 ft, find the natural frequency of the
system.
13. A particle of mass m moving in a straight line is repelled from the origin 0
by a force F. If the force is proportional to the distance of the particle from
0, find the position of the particle as a function of time. NOTE. The particle
no longer executes simple harmonic motion.
14. If at t = 0 the velocity of the particle of problem 13 is zero and it is 10 ft
from the origin, find the position and velocity of the particle as functions of
time.
15. At t = 0 the velocity of the particle of problem 13 is -aVk ft/sec and it
is a feet from the origin. If the repelling force has the value kx, find the
position of the particle as a function of time. Show that if m < 1, the
particle will never reach the origin; that if m = 1, the particle will approach
the origin but never reach it.
16. A particle executes simple harmonic motion. At t = 0, it is -5 ft from its
equilibrium position, its velocity is 6 ft/sec and its acceleration is 10 ft/sec 2 •
Find its equation of motion and its amplitude.
17. Let lei be the amplitude of a particle executing simple harmonic motion. We
proved in the text, see description of the motion, Example 28.15, that the
particle has zero speed at the points ±e and maximum speed at the equilib-
rium position x = o.
(a) Prove that the particle has maximum acceleration at x = ±e and has
zero acceleration at x = o. Hint. Take the second derivative of the
equation of motion (28.25).
(b) What is the magnitude of the maximum velocity and of the maximum
acceleration?
(c) If v'" is the maximum velocity and a", is the maximum acceleration of a
particle, show that its period is T = 27rv",/a", and its amplitude is
A = v",2/a"..
18. Two points A and B on the surface of the earth are connected by a friction-
less, straight tube. A particle placed at A is attracted toward the center of
the earth with a force directly proportional to the distance of the particle
from the center.
(a) Show that the particle executes simple harmonic motion inside the tube.
Hint. Take the origin at the center of the tube and let x be the distance
of the particle from the center at any time t. Call b the distance of the
center of the tube from the center of the earth and R the radius of the
Lesson 28C EXAMPLES: SIMPLE HARMONIC MOTION 323

earth. Take a diameter of the earth parallel to the tube as a polar axis
and let (r,B) be the polar coordinates of the particle at any time t. Then
show that the component of force moving the particle is F cos Band
that cos B = x/r. Initial conditions are t = 0, x = 'l/R2 - 62,
dx/dt = O.
(b) Show that the equation of motion is
x = ...; R2 - b2 cos"';k/m t,
where k is a proportionality constant and m is the mass of the particle.
(c) Show that for a given mass, the time to get from A to B is the same for
any two points A and B on the surface of the earth. Hint. Show that
the period of the motion is a constant.

ANSWERS 28A AND B

2. (a) x = (4000)(5280) cos (t/812) in ft. (b) 4000 mi. (c) 85 min.
(d) 0.7 oscillation/hr.

3. (a) x = 4...;2 sin (t - . (b) 4'1/2" ft. (c) 1 rad/sec or cps.

(d) -1f/4. (e) 1f/4 sec.
4. (a) x = 5 cos (81ft); v = -401f sin (81ft). (b) 8 ... rad/sec, 5 ft.
(c) is sec, -401f ft/sec.
5. (a) x = ; sin (4; t) . (b) i sec, 41f/3 rad/sec, or! cps, 6/1f ft.
6. (a) x = 10 sin (41ft + 8). (b) ::I::201f'1/3 ft/sec.
7. x = "';X1 2+ a2v12/4 sin (2t/a + 8), where sin 8 = Xt!"';X1 2 + a2v12/4.
8. x = "';9 + 1/1f 2 sin (61ft + 8), where 8 = Arc sin (31f/"';91f 2 + 1).
9. 1f sec, 2 rad/sec or 1/1f cps.
10. (a) x = 6 cos (1ft/3 + 51f/6). (b) ::I:: 21f ft/ sec.
(c) 6 sec, 6 ft, i cps, 51f/6 rad. (d) -3'1/3 ft from center.
11. x = 4 cos (1ft).
12. '1/12 rad/sec or 'l/3h.. cps.
13. x = cle.jklm' + c2e-..;v.;;,), where k is a proportionality constant.
14. x = 5(e.jklm' + e-.jklm,), v = 5'1/k/m (e.jklm' - e-.jklm,).
15. x = [(1 - Vm)e-w;;., + (1 + Vm)e--w;;.,).
16. x = 3...;2 sin ...;2 t - 5 cos ...;2 t; ...;43 ft.
17. (b) v = cwo, a = cwo2, where c is the amplitude and wo is the frequency of
the motion in radians per unit of time.

LESSON 28C. Examples of Particles Executing Simple Harmonic

Motion. Harmonic Oscillators. A dynamical system which vibrates
with simple harmonic motion is called a harmonic oscillator. Below we
give two examples of harmonic oscillators.
324 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

Exalllple A. The Motion of a Particle Attached to an Elastic

Helical Spring. Hooke's Law. The unstretched natural length of an
elastic helical spring is lo feet, Fig. 28.6(a). A weight w pounds is attached
to it and brought to rest, Fig. 28(b). Because of the stretch due to the

]I
Equilibrium
position of the
spring

(a) (b)
Figure 28.6

attached weight, a force or tension is created in the spring which tries to

restore the spring to its original unstretched, natural length. By Hooke's
law, this upward force of the spring is proportional to the distance l,
by which the spring has been stretched. Hence,

(28.61) The upward force of the spring = ki,

where k > 0 is a proportionality constant, called the spring constant

or the stiffness coefficient of the spring. The downward force acting on
the spring is the weight w pounds that is attached to it. If the spring
is on the surface of the earth, then w = mg, where m is the mass in slugs
of the attached weight and g is the acceleration due to the gravitational force
of the earth in feet per second per second. Since the spring is in equilib-
rium, the upward force must equal the downward force. Hence by (28.61)

(28.62) ki = mg.

Let y = 0 be the equilibrium position of the spring with the weight w

pounds attached to it. If the spring with this weight attached is now
stretched an additional distance y, Fig. 28.6(c), then the following forces
will be acting on the spring.
Lesson 28C EXAMPLES: SIMPLE fuRMONIC MOTION 325

1. An upward force due to the tension of the spring which, by Hooke's

law, is now k(l y). +
2. A downward force due to the weight w pounds attached to the spring
which is equal to mg.
By Newton's second law of motion, the net force acting on a systetI1 is
equal to the mass of the system times its acceleration. Hence, with tha
positive direction taken as downward,

d2 y
(28.621) m dt 2 = mg - k(l + y) = mg - kl - kyo

By (28.62), (28.621) simplifies to

(28.63)

which is the differential equation of motion of an elastic helical spring.

Since it has the same form as (28.11), we now know that a displaced
helical spring with weight attached and with no resistance will execute
simple harmonic motion about the equilibrium position y = o. The
solution of (28.63) is

(28.64) y = c cos (v'k/m t + 8) or y = c sin (v'k/m t + 8).

Example 28.65. A 5-pound body attached to an elastic helical spring
stretches it 4 inches. Mter it comes to rest, it is stretched an additional
6 inches and released. Find its equation of motion, period, frequency, and
amplitude. Take the second for a unit of time.
Solution. Since 5 pounds stretches the spring 4 inches = i foot, we
have by (28.62), with mg = 5, l = i,
k
(a) -=5
3 '
k = 15,

which gives the stiffness coefficient of the spring. With g = 32, the mass
m of the attached body is n\. Therefore the differential equation of
motion, by (28.63), is
5 d2y
(b) 32 dt2 = -15y,

Its solution, by (28.64), or by solving it independently, is

(c) y = c cos (v'96 t + 8), = -v'96 c sin (v'96 t + 8).

The initial conditions are t = 0, y = i, v = dy/dt = o. Inserting these
326 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

values in both equations of (c), we obtain

(d) i = ccos &,
o= -v'96 c sin

Since c, the amplitude, cannot equal 0, the second equation in (d) implies
& = O. With this value of &, the first equation in (d) gives c = i. Sub-
stituting these values in the first equation of (c), we find for the equation
of motion
(e) y = i cos v'96 t.
The period of the motion, by Definition 28.34, is therefore 211"/v'96
seconds; the frequency, by Definition 28.36, is V96/211"" cps, or by Defini-
tion 28.38, V96 rad/sec; the amplitude, by Definition 28.31, is i foot.
Example 28.66. A body attached to an elastic helical spring executes
simple harmonic motion. The natural (undamped) frequency of the mo-
tion is 2 cps and its amplitude is 1 foot. Find the velocity of the body as
it passes the point y = i foot.
Solution. Since the body attached to the helical spring is executing
simple harmonic motion, its equation of motion by (28.64) is

(a) y = ccos (vk/mt + &).

The amplitude of 1 foot implies, by Definition 28.31, that c = 1. The
frequency of 2 cps implies, by Definition 28.36, that 2 = vk/m/211"", or
vk/m = 411"". Substituting these values in (a) it becomes

(b) y = cos (411""t + &), v = = -411"" sin (4m + &).

When y = i, we obtain from the first equation in (b)
(c) i = cos (411""t + &),
Hence, by (c), when y = i,
(d) sin (411""t + &) = ± 1·
Substituting (d) in the second equation of (b), we obtain

(e) v = -dy = ±2 V3
311""
dt '
which is the velocity of the body as it passes the point y = i. The plus
sign is to be used if the body is moving downward; the minus sign if it is
moving upward.
Lesson 28C EXAMPLES: SIMPLE HARMONIC MOTION 327

EX8IDpie B. The Motion of 8 SilDple PendululD. A displaced

simple pendulum of length l, with weight w = mg attached (Fig. 28.7)
will execute an oscillatory motion. If the angle of swing is very small,
then as we shall show, the motion will closely approximate a simple har-
monic motion.
The effective force F which moves the pendulum is the component of the
weight acting in a direction tangent to the arc of swing. From Fig. 28.7,
(Jis positive if the weight w is
to the right of 0; negative
if to the left of O.
w - d6 / dt is positive if the weight
w moves counterclockwise;
negative if it moves clockwise.
F is positive if it acts in a direction
to move the pendulum
counterclockwise; negative
if the direction is clockwise.
Note that when 6 is positive,
F is negative; when (J is
negative, F is positive.

Figure 28.7

we see that the value of this component is mg sin 6, where 8 is the angle
of swing and m and g have their usual meanings. (The tension in the
string and the component of the weight in the direction of the string do
not affect the motion.) Therefore by Newton's second law
dv . dv .
(28.71) F = m dt = -mg sm 8, dt = -gsm 8.

The distance s over which the weight moves along the arc is
ds d8
(28.72) s = [8, therefore dt = l dt .

Since v = ds/dt, we obtain from the second equation in (28.72),

d8 dv d2 8
(28.73) v = l dt .. dt = l dt 2 •

Substituting in (28.71) the value of dv/dt as given in (28.73), we have

d 28
(28.74) l dt 2 + g sm. 8 = o.
Equation (28.74) does not as yet have the form (28.11) of simple harmonic
motion. However, we know from the calculus that sin 8 = 8 - 83 /3! +
328 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

- .. '. We may therefore write (28.74) as

d 20 g03 gOS
(28.75) l-+gO
dt 2
- -+-
3! 5! - ... = 0
.
Hence, for a pendulum swinging through a small angle, it is not unreason-
able to assume that the error made in linearizing the equation by dropping
terms in 0 3 and higher powers of 0 will also be small. Hence (28.75)
becomes
,
(28.76)

But this last equation is now the differential equation of a particle exe-
cuting simple harmonic motion. The solution of (28.76) is

(28.77) o = c cos (...;U7i t + 8),

where c is the amplitude of 0 in radians.
E%ample 28.771. A simple pendulum whose length is 5 ft swings with
an amplitude of n radian. Find the period of the pendulum and its
velocity as it crosses the equilibrium position, i.e., the position when
0=0.
Solution. The given amplitude of n
radian implies that c = n in
(28.77). Therefore, with 1 = 5 and g = 32, (28.77) becomes

(a) + 8),
o = n cos (V32/5 t
= -nv32/5 sin (V32/5 t + 8).

The period of the pendulum, by Definition 28.34, is

(b) T = = 'iVlOsec.
When 0 = 0, we have by (a),
(c) 0= cos (V32/5 t + 8),
which implies that
(d) sin (V32/5 t + 8) = ± l.
Substituting (d) in the second equation of (a), there results

(e)

By the first equation in (28.73) and (e), we obtain with 1 = 5,

(f) v = 5(±IsVlO) = ±iVlO ft/sec,

Lell80n 28C-Exercise 329

which is the velocity of the pendulum as it crosses the central position.

The plus sign is to be used if the weight is moving counterclockwise; the
minus sign if it is moving clockwise.

EXERCISE 28C
1. Verify the accuracy of the solution of (28.63) as given in (28.64).
2. Verify the accuracy of the solution of (28.76) as given in (28.77).
3. Solve (28.63) for y as a function of t if at t = 0, y = Yo, v - Vo.
4. Show that the sum y2 + , where y is the solution of (28.63) as
given in (28.64), is a constant. NOTE. The first term gives the potential
energy of a mass attached to a helical spring displaced a distance y from
its equilibrium position. The second term gives its kinetic energy. Since
the sum of the two energies is a constant, a mass oscillating on a helical
spring with simple harmonic motion satisfies the law of the conservation
of energy.
5. Prove the law of the conservation of energy for the undamped helical spring
(see problem 4) by multiplying (28.63) by dy/dt and then integrating the
resulting equation with respect to t. For hint, see answer.
6. A 12-lb body attached to a helical spring stretches it 6 in. After it
to rest, it is stretched an additional 4 in. and released. Find the equation df
motion of the body; also the period, frequency, and amplitude of the motiQ'n.
With what velocity will it cross the equilibrium position?
7. A lO-lb body stretches a spring 3 in. After it comes to rest, it is stretched an
additional 6 in. and released.
(a) Find the position and velocity of the body at the end of 1 sec.
(b) When and with what velocity will the object first pass the equilibrium
position?
(c) What is the velocity of the body when it is -3 in. from equilibrium?
(d) When will its velocity be 2 ft/sec and where will it be at that instant?
(e) What are the period, frequency, and amplitude of the motion?
8. A spring is stretched 8 in. by a 16-lb weight. The weight is then removed
and a 24-lb weight attached. After the system is brought to rest, the spring
is stretched an additional 10 in. and released with a downward velocity of
6 ft/sec.
(a) Find the equation of motion, period, frequency, amplitude and phase
angle.
(b) Answer the same questions if the weight were given an upward velocity
of 6 ft/sec instead of a downward one.
9. A heavy rubber band of natural length lo is stretched 2 ft when a weight W
is attached to it. Find the equation of motion and the maximum stretch of
the band, if at the point lo the weight is given a downward velocity of 6 ft/sec.
Assume the rubber band obeys Hooke's law.
10. The spring constant of a helical spring is 27. When a weight of 96 lb is at-
tached, it vibrates with an amplitude of 3/2 ft.
(a) What are its period and frequency?
(b) With what velocity does it pass the point y -1 ft?
330 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

11. A piece of steel wire is stretched 8 in. when a 128-lb weight is attached to it.
After it is brought to rest, it is displaced from its equilibrium position. Find
the frequency of its motion. Assume the wire obeys Hooke's law.
12. A helical spring is stretched 2 in. when a 5-lb weight is attached to it. If the
5-lb weight is removed and a weight of W Ib attached, the spring oscillates
with a frequency of 6 cps. Find the weight W.
13. When two springs, suspended parallel to each other, are connected by a bar
at their bottom, they act as if they were one spring. The spring constant of
the system is then equal to the sum of the spring constant of each spring.
Assume a weight of 8 Ib will stretch one spring 2 in. and the other spring 3 in.
(a) What is the spring constant of each spring?
(b) If the two springs are suspended parallel to each other and connected
by a bar at their bottom, what is the spring constant of the system?
(c) If a weight of 8 Ib is attached to the system, brought to rest, and then
released after being displaced 6 in., find the amplitude, period, and
frequency of the motion.
14. When two springs are connected in series, i.e., when one spring is attached
to the end of the other, the spring constant k of the system is given by the
formula
1
-
k
= -+-,
1
kl
1
k2
where kl and k2 are the respective spring constants of each spring. Assume
the springs of problem 13 are attached in series.
(a) What is the spring constant of the system?
(b) If a weight of 8 Ib is attached to the bottom spring, brought to rest,
displaced 6 in, and then released, find the amplitude, period, and fre-
quency of the motion.
The solutions to problems 15-17 will be facilitated if reference is made
to the solution given in problem 3.
15. A spring is stretched l ft by a weight Wand brought to rest. It is then given
a downward velocity of vo ft/sec.
(a) Find the lowest point reached by the weight and the elapsed time.
(b) Find the amplitude, frequency, and period of the motion.
16. A spring is stretched l ft by a weight Wand brought to rest. It is then
stretched an additional a ft and given a downward velocity Vo.
(a) Find the amplitude, frequency, and period of the motion.
(b) Answer the same questions if the initial velocity is vo upward instead of
downward.
17. A helical spring has a period of 4 sec when a 16-lb weight is attached. If
the 16-lb weight is removed and a weight W attached, the spring oscillates
with a period of 3 sec. Find the weight W.
In problems 18-28, it is assumed that 8, the angle a pendulum makes
with the vertical, i.e., the angle the pendulum makes with its equilibrium
position, is sufficiently small so that equations (28.76) and (28.77) are
applicable. Remember that in these problems angular velocity w = d8jdt,
Lesson 28C-Exercise 331

and linear velocity v = l d8/dt = lw, where l is the length of the pendulum
from point of support to the center of mass of the bob. For positive and
negative directions, see Fig. 28.7.
18. A simple pendulum of length I ft is given an angular velocity of wo rad/sec
from the position 8 = 80.
(a) Find the position of the bob of the pendulum as a function of time.
(b) Find amplitude, period, frequency, and phase angle.
(c) With what velocity, angular and linear, does the pendulum cross the
equilibrium position?
19. A simple pendulum of length I ft is released from the position 8 = 80.
(a) Find the position of the pendulum as a function of time.
(b) Find amplitude, period, and frequency. Compare results with problem 18.
(c) With what velocity, angular and linear, does the pendulum cross the
equilibrium position?
20. A simple pendulum of length 2 ft is released from the position 8 = n rad
at time t = o.
(a) Find the value of 8 when t = 11'/16.
(b) When and with what velocity, angular and linear, does the pendulum
cross the equilibrium position?
21. A simple pendulum of length 6 in. is given an angular velocity of magnitude
! rad/sec toward the vertical from the position 8 = trJ roo. Find the equa-
tion of motion, amplitude, period, frequency, and phase angle. Hint. At
t = 0, w = -! rad/sec.
22. A simple pendulum of length 2 ft is given an angular velocity of ! rad/sec
toward the vertical from the position 8 = -trJ rad. Find the equation of
motion and the amplitude.
23. Solve problem 22, if the pendulum is given an angular velocity of ! rad/sec
away from the vertical from the position 8 = trJ rad.
24. A simple pendulum of length I ft is started by giving it an angular velocity
of wo rad/sec from its equilibrium position.
(a) Find the equation of motion.
(b) Find the value of 8 and the time when the pendulum reaches its maximum
displacement.
25. A clock pendulum is regulated so that 1 sec elapses each time it passes its
equilibrium position. (Its period is therefore 2 sec.) What is the length of
the pendulum?
26. It is desired that a clock pendulum cross its equilibrium position each
second and have an amplitude of n rad. What should its angular velocity
be as it crosses the equilibrium position? Hint. The length of the pendulum
has already been found in 25, the equation of motion in problem 24.
27. The amplitude of a pendulum is 0.05 rad. What fraction of its period has
elapsed when 8 = 0.025 rad? For hint, see answer.
28. A clock pendulum has a length of 6 in. Each time it crosses the equilibrium
position, it makes a tick. How many ticks will it make in 1 hr?
29. In the pendulum example in the text and in the problems above, the weight
of the wire supporting the mass was considered negligible and therefore ig-
nored. If, however, this weight is not negligible, then, instead of (28.71),
332 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

the differential equation of motion of the pendulum and bob is

(28.772) I = -(mg sin 8)l,

where I is the moment of inertia* of the pendulum and bob about an axis
which passes through the point of suspension and is perpendicular to the
plane of rotation of the pendulum; m is the mass of pendulum and bob;
l is the length of the pendulum from the point of support to the center of
gravity of pendulum and bob. [For the case of a pendulum of negligible
mass, the moment of inertia I of a particle of mass m attached at a distance
l from the axis of rotation is ml 2 • Substitution of this value of I in (28.772)
will yield (28.74).)
With the aid of (28.772), find the equation of motion and the period of a
pendulum consisting of a uniform rod of length 1 and swinging through a
small angle. Hint. The moment of inertia I of a uniform rod of length l
and mass m about an axis through one end is given by
1

I = lim
ft--+oo
t
i-I
fJxl dx; = 10 fJx2 dx = mt with 8 = mil.

Since the rod is uniform, its mass may be considered as concentrated at its
center, i.e., its center of gravity is at its geometric center.

In problems 30-34, we no longer assume that the angle of swing is

small. Hence we cannot, in (28.74), replace sin 8 by 8.
30. Solve (28.74) for d8/dt with the initial conditions t = 0, 8 = 80, w = o.
Hint. Multiply the equation by 2d8/dt and then make use of the identity

Answer is
(28.78) w = d8/dt = ±V2i/ivcos 8 - cos 80.
31. By integrating (28.78), remember the initial condition is t 0, 8 = 80,
show that

(28.79) t(8) = ± Vl/2g (' du .

1'0 vcos u - cos 80
Since at t 0, 8 = 80, w = 0, one-half of a period elapses when 8 = -80.
Use this fact, (28.79), and the fact that cos u is an even function, i.e.,
cos (-u) cos u, to show that the period T of the pendulum is given by

(28.791)
T
2" =
(f1'0
'\/29
du
-Iovcosu -.cos80 '

T = 2v'2 vT79 r'o du

10 vcosu - cos 80

·For definition of moment of inertia, see Lesson 30M-B.

Lesson 28C-Exercise 333

In (28.791), replace cos u by its equal 1 - 2 sin 2 i, cos 80 by its equal

1 - 2 sin 2 Show that the resulting equation is

(28.7911) T = 2..Ji7Y to V sin

J0 2 (80/2) -
du
sin2 (u/2)
Finally in (28.7911), make the substitutions

(28.7912) sin = sin sin </>, ! cos du = sin cos </> d</>.

Show that when u = 0, </> = 0; when u = 80, </> = .../2 and that (28.7911)
therefore becomes
. ,2

(28.7913) T . 'l/g
= 4v "
1 VI -
o k 2 sin 2 </>
, k = sm 2"'
. 80

The integral in (28.7913) is known as the complete elliptic integral of the

first kind. It cannot be expressed in terms of elementary functions, but
it can be evaluated* by writing the integrand as a power series

(28.7914) (1 - k 2sm",
. 2 ",)-112 = 1 + l.k 2
2 • 2",
5m ", + N1 . 3 k4 . 4",
SIn", + ••••

Substituting the above series in (28.7913) and carrying out the integration,
we obtain, since, when n is an even positive integer,

r '2 . n _ 1· 3· 5· .. (n - 1) "II"
Jo sm </> d</> - 2.4.6 ... n 2"'
(28.7915) T = 2nIf7Y [1 + k
2

22
+ (12.4
.3)2 k4 + ...J'
which gives the actual period of a pendulum.
By (28.77), the period of a pendulum, approximated by replacing sin 8
by 8, is
(28.7916)

Assume the initial displacement of a pendulum at t = 0, is 80 = 4°. Then

by (28.7913), k = sin 2° = 0.0349 and k 2 = 0.00122. Hence by (28.7915),
the more exact period of the pendulum is

(28.7917) T (exact) = 2nIf7Y (1 + + + .. .)

= 2nIf7Y (1.000305) sec,
and by (28.7916) the approximate period is
(28.7918) T (approx.) = 2"11"vl/g sec.

*A table of values of this integral can be found, just as one can find a table of values
of sin X; see, for example, Pierce's Tables.
334 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

By (28.7917) and (28.7918), we see, therefore. that

(28.7919) T (exact) = 1.000305T (approx.).
Suppose now, we design a clock with pendulum of sufficient length l so
that its period, using the approximation formula (28.7918) is 2 sec, and its
amplitude is 4,0. Therefore, by (28.7919), its more exact period is 2.00061 sec.
This means that in each 2 sec the clock is incorrect by 0.00061 sec, its second
hand has moved 2 sec whereas it should have moved 2.00061 sec; the clock
is thus too slow by this much. Calculate the error in the clock at the end
of each 24 hr. Ana. 26 sec slow.
32. Solve (28.74) for d8/dt with the initial conditions t == 0, 8 = 0, w = woo
See hint in 30. Answer is

(28.792) d8)
( -dt
2
= wo
2
- -2g (1 -
l
COB 8) = wo 2[ 1 - - 2g2 (1 -
lw0
cos 8) ] •

33. In (28.792), make the substitution

(28.793) k2 = and sin2 f!.. = t(1 - cos 8).

lw0 2 2
Show that at time t, remember at t 0,8 = 0,

(28.7931) t(8) = ::I:

l ' du
::I:.!.
wo
['/2
Jo v'1 -
dtP
k 2 sin2 cP
o waYl - k 2 sin 2 (u/2)
The integral in (28.7931) is called an incomplete elliptic intep'al of the
first kind.
34. In the text, we derived the differential equation for a pendulum of negligible
weight swinging along the arc of a circle. The resulting equation (28.74) was
not simple harmonic. Assume now that the end of the pendulum moves
along a curve which is given by the parametric equations
(a) x = a(28) + a sin 28,
y = a - a cos 28,
where a is a positive constant and 8 is the inclination of the tangent to the
curve at a point P, Fig. 28.794. Hence, the differential equation of motion
of the pendulum is

m = -mg sin 8.

Replacing" by its equal dB/dt, we obtain

2
(b) d 8 • 8
dt 2 = -g sm ,

where 8 is the distance along the curve measured from the lowest position O.
It can be shown that the distance B along an arc of the curve (a) from its
lowest point is B = 4a sin 8. Solve for sin 8 and substitute this value in (b).
Show that the resulting equation is now simple harmonic. Solve for 8 as a
function of t, find the period of the motion and establish the fact that the
period is a constant.
Lesson 28C-Exercise 335

The parametric equations (a) above are the equations of an inverted

cycloid traced by a point on a circle of radius a. (When a circle rolls along
a straight line, the curve traced by a point P on its circumference, is called
a cycloid.) In (a) above, 28 is the central angle made by lines drawn from
the center of the circle, one vertical, the other to P. It can then be proved
that the inclination of the tangent to the cycloid at P is 8. If, therefore,
you can devise a pendulum whose bob will move along the path of an in-
verted cycloid, the pendulum will swing in simple harmonic motion regard-
less of the angle of swing, and its period will be independent of its amplitude.

X=a(28)+a sin 20
{ y=a-acos28

Figure 28.794

Archimedes' principle states that a body partially or totally sub-

merged in a liquid is buoyed up by a force equal to the weight of the
liquid displaced. For example, if a 5-pound body floats on water, then
the weight of the water displaced is 5 pounds, i.e.,

V (in ft 3 of water displaced) X 62.5 (weight of a ft 3 of water) = 5.

The fact that the body is resting or floating on water implies that the
downward force due to the weight W = mg of the body must be the
same as the upward buoyant force of the water, namely the weight of
water displaced.
Call y = 0 the equilibrium position of the lower end of a body when it
is floating in a liquid. If the body is now depressed a distance y from its
equilibrium position, an additional upward force will act on the body due
to the additional liquid displaced. If the body's cross-sectional area is A,
then the volume of additional liquid displaced is Ay and the weight of
this additional liquid displaced is (Ay)p, where p is the weight per unit
volume of the liquid. Since this additional upward force is the net force
acting on the depressed body, we have, by Newton's second law of mo-
tion, with positive direction downward,

(28.795)
336 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

A floating body, therefore, when depressed from its equilibrium position

executes simple harmonic motion.
With the aid of (28.795), solve the following problems:
35. A body of mass m has cross-sectional area A sq ft. It is depressed Yo ft from
its equilibrium position in a liquid whose weight per ft 3 is p and released.
(a) Show that its equation of motion is
(28.796) y = Yo co8'\/ pAlm t,
where y is the distance of the body from equilibrium. Hint. At t = 0,
Y = Yo," = O. (b) If the body is a right circular cylinder with vertical axis
and radius T, show that its period is

(28.797)
T
rm:K.
'\jp-
36. A cylindrical buoy of radius 2 ft and weighing 64 lb floats in water with its
axis vertical. It is depressed t ft and released.
(a) Find its equation of motion, amplitude, and period.
(b) How far below the water line is the equilibrium position.
37. A cubic block of wood weighing 96 lb has cross-sectional area 4 ft. It is de-
pressed slightly in a liquid which weighs 50 lb/cu ft and released. What is
its period of oscillation?
38. A cylindrical buoy of radius 1 ft floats in water with its axis vertical. Its
period, when depressed and released, is 4 sec. What is its weight?
39. A cubical block of wood is 2 ft on a side. When depressed and released in
water, it oscillates with a period of 1 sec; What is the specific gravity of the
wood? Hint. Use (28.796) to find the mass m of the cubical block. Then
find its weight per cubic foot. Then compare this figure with the weight of a
cubic foot of water.
40. When a cylinder with vertical axis is depressed and then released in water,
it vibrates with a period To sec. When water is replaced by another liquid,
it vibrates with a period of 2To sec. Find the weight of the liquid per cubic
foot. Hint. Use (28.797) to find p.
41. A body whose cross-sectional area is A, displaces, when in equilibrium, l ft
of a liquid. The body is then depressed and released. Show that its differen-
tial equation of motion is

(28.798)

where y is the distance of the body from equilibrium at time t. Hint. In

equilibrium, the weight of the body equals the weight of the water displaced.
The weight of displaced water is (Al)p. Therefore Alp = mg. Solve for p
and substitute in (28.795).
42. Solve (28.798). What is the period of vibration?
43. A cylinder with vertical axes vibrates in water with a period of 2 sec. How
far from the surface is its equilibrium position? Hint. Make use of the
solution of (28.798) as found in problem 42.
44. A spherical body of radius T is in equilibrium when half of it is submerged in
a liquid. It is given a displacement and released. Find its differential equa-
tion of motion and, if the displacement is small in comparison with the
radius T. also find its approximate period. For hints see answer.
Lesson 28C-Answe1'll 337

ANSWERS 28C

3. Y = vov:m7k sin t) + Yo cos (vk/m t)

"'" VY02 + (m/k)v02 sin (vk/m t + 8),

where 8 = Arc sin Yo

VY02 + (m/k)v0 2
5. To integrate the first term make the substitution u = dy/dt, du =
(d 2 y / dt 2 ) dt.
6. y = ! cos 8t, 1(/4 sec, 8 rad/sec, or 4/1( cps,! ft, ± ! ft/sec.
7. (a) 0.16 ft, -5.37 ft/sec. (b) V21(/32 sec, -5.66 ft/see,
(c) ±2v6 ft/see, (d) 0.31 sec, -0.47 ft,
(e) V21(/8 sec, 8v2rad/see, 1 ft.
30. 5./n VIal . . /n
8. (a) y = -4- sm (40 t) +"6 cos (4v 2 t) "'" 60 sm (4v 2 t + 0.67),

-4- sec, 40 rad/see, VIal/M ft, 0.67 rad.

(b) y = - sin (40 t) + cos (40 t); remaining answers are the
same as in (a).
9. y = ! sin 4t - 2 cos 4t. Amplitude is 5/2 ft. Maximum stretch is 9/2 ft.
10. (a) 211"/3 sec, 3 rad/see or 3/21( cps. (b) ±!v5 ft/see.
ll. 4v3 rad/sec.
20
12. 31(2 lb.
13. (a) 48 and 32. (b) 80. (c) 1 ft, vS 1(/20 sec, 4vS/1r cps.
14. (a) 19.2. (b) 1 ft, 21(/V76.8 sec, V76.8/21( cps.
15. (a) y = vov m/k = voVTlU ft, t = iV m/k = i VTlU sec.

(b) voVTlU ft,....,!kJm = v'u/i rad/see, 21r\1'T7i sec.

16. (a) A = va2 + lv0 2 /g, period and frequency are the same as in 15.
(b) Same answers as in (a).
17. 9 lb.

18. (a) 8 = 80 eosv'u/it+woVTlUsinv'u/it "'" V802 + (l/g)w0 2 sin (v'u/it+ 8),

where 8 = Arcsin (80/V80 2 + (l/g)w02).
(b) V80 2 + (l/g)w0 2 ft, 2rVTlU sec, (1/21()v'u/i cps.
8 is given in (a).
(c) w = rad/see, v = ± Vlg802 + l2w02 ft/see.
19. (a) 8 = 80 cos v'u/i t.
(b) A = 80 rad, T = 21r\1'T7i sec, II = v'u/i cps.
21(
(c) W = ±80v'u/i rad/see, v = ±80vUl ft/sec.
338 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

20. (a) 8 = 0/24 rad. (b) t = 1(2n + 1)11", n = 0, 1, .. " ::I::! rad/sec,
::1::1 ft/sec.
21. 8 = trJ cos 8t - -h sin 8t ... v89/80 sin (8t - 4.15), v89/80 ft, 11"/4 sec,
4/11" cps, 4.15 rad.
22. 8 = -trJ cos 4t + 1 sin 4t, A = V 41/40 ft.
23.8 = trJcos4t+ 1 sin 4t, A = v41/40ft.
24. (a) 8 = wov'Vu sin (V(jji t). (b) 8 = wov'Vu rad, t = (1I"/2)v'Vu sec.
25. 32/11"2 ft. 26. 0.21 rad/sec.
27. T /6. Hint. In the solution ,iven in problem 18, 80 = 0.05 and take wo = O.
Find t when 8 = 0.025.
28. 9167.
29. 8 = C1 cos V3g/2l t + C2 sin V3g/2l t, T = 2rv'2l73U sec.
34. 8 = C1 sin (t/2)v'"i/O. + C2 cos (t/2)v'"i/O., T = 4'lI"v'aiU.
36. (a) y = ! cos A = ! ft, T = fv;:;5 sec. (b) 0.08 ft.
37. 0.77 sec. 38. 2546 lb. 39. 0.405. 40. One-fourth the weight of
water.
42. y = Ccos Vi7i t, T = sec.
43. g/.".2 ft = 3t ft approx.
44. Let y be the displaced distance of the sphere from equilibrium. The buoyant
force of the liquid, therefore, is equal to the weight of liquid displaced, i.e.,
it is the weight of a volume of liquid equal to one-half the volume of the
sphere minus the volume of a segment of a sphere of height r - y. The
volume of a segment of a sphere is (1I"h2/3)(3r - h), where r is the radius
of the sphere and h is the height of the segment. And since the body is in
equilibrium when only one-half of it is submerged, the weight of the liquid
per unit volume must be 2p, where p is the weight per unit volume of the
sphere. The differential equation motion is
2
d y = _
dt 2
f!.
2 r
[3 l!: _(l!:)3].
r
If the displacement 'iI is small in comparison with r, then it is reasonable to
assume that the error made in linearizing the equation ,py dropping (y/r)3
will also be small. The approximate period is therefore T = 211"V2r/3g.

LESSON 28D. Forced Undamped Motion. The motion of a particle

of mass m that satisfies a differential equation of the form
2
d y d 2y 1
(28.8) m dt?: + mwo 2y = f(t), dt 2 + Wo 2 y = m f(t)
where f(t) is a forcing function attached to the system and Wo is defined
in 28.38, is called forced undamped motion, in contrast to the free un-
damped motion (i.e., simple harmonic motion) when f(t) ... O. Let us
assume that the forcing function f(t) = mF sin (wt +
(J) where F is a con-
stant. Then (28.8) becomes
(28.81) + wo 2 y = F sin (wt + (J).
Lesson 28D FORCED UNDAMPED MOTION 339

If we set the left side of (28.81) equal to zero, and solve the resulting
homogenous equation, we obtain the complementary function

(28.82) Yo = e sin (wot + &).

A particular solution yp of (28.81) will then depend on the relative
values of the natural (undamped) frequency Wo of the system and the
iUlpressed frequency w of the forcing function mF sin (wt (3). We +
shall treat each of the two possibilities in the two cases below.

Case 1. Col ColO. If w Wo, then a particular solution of (28.81) is

(28.83) YP = Wo
2 F
- w
2 sin (wt + (3).
Hence, by (28.82) and (28.83), the general solution of (28.81) is

(28.84) y = e sin (wot + &) + Wo 2 F- w

2 sin (wt + (3).
The motion of the system is now the sum of two separate and distinct
motions, each of which is simple harmonic. The displacement or depar-
ture of the particle from its equilibrium position is therefore the sum of
two separate (harmonic) displacements with respective amplitudes of e
and F/(wo 2 - w2 ). The maximum value of the displacement or depar-
ture, however, cannot exceed lei +
1F/(wo 2 - w2 )1. Since all these letters
denote constants, the displacement or departure has finite magnitude. A
motion in which the displacement or departure of a particle from its
equilibrium position remains finite with time is called a stable Ulotion.
If, however, Wo and ware nearly alike, then (wo 2 - w2 ) will be small,
and since this term appears in the denominator of (28.84), the departure
or displacement of the particle from equilibrium will be large, i.e., the
vibrations of the system will be big, and if sufficiently large, a breakdown
of the system may result.
The behavior of the particle's motion is again influenced by two mo-
tions with different frequencies, the natural (undamped) frequency ColO
and the forcing frequency w. If wo/w is a rational number, say 3, then
the motion due to Yo will make three revolutions while the motion due to
YP is making only one. Therefore, during the time interval 27r/w (the
period of one revolution due to the YP motion), the motion of the system
will be erratic. However, at the end of this time interval, the position of
the particle due to the Yc and YP motions will again be its starting one,
and the system will again repeat its erratic behavior. The motion of the
system will thus have an appearance somewhat like that shown in Fig.
340 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

28.85. If, however, wo/w is an irrational number, then the motion will
not have a repetitive pattern.

t== 47r
co co

Figure 28.85

Case 2. (.01 = (.010' If w = wo, the differential equation of motion

(28.81) becomes
(28.9) + wo 2 y = F sin (wot + fJ).
Now, however, a term in the Yo part of the solution as given in (28.82),
agrees, except for phase and constant coefficient, with the function on the
right of (28.9). Hence the trial function YP' by Lesson 21A, Case 2, must
be of the form

(28.91) YP = At sin (wot + fJ) + Bt cos (wot + fJ).

Following the method described in Lesson 21A, Case 2, we find
F
(28.92) YP = - -2 t cos (wot
Wo
+ fJ).
Hence the general solution of (28.9), by (28.82) and (28.92), is

(28.93) y = c sin (wot + 8) - 2F t cos (wot

Wo
+ fJ).
The maximum departure or displacement of the motion is iel + 12wo t I·
The presence of the variable t in the second term implies that the depar-
ture or displacement due to this part of the motion increases with time,
see Fig. 28.94. A motion in which the departure or displacement increases
beyond all bounds as time passes is called an unstable motion. In such
cases, a mechanical breakdown of the system is bound to occur. This
condition, where w, the frequency of the forcing function, equals wo, the
Lesson 28D FORCED UNDAMPED MOTION 341

natural (undamped) frequency of the system, is known as undamped

resonance, and Wo is called the undamped resonant frequency.
y

Figure 28.94

Comment 28.941. In engineering circles, the functionf(t) of (28.8) is

referred to as the input of the system, the solution y(t) of (28.8) as the
output of the system.
Example 28.95. The differential equation of motion of a system is

(a) y" + 4y = cos 2t.

Find its equation of motion. Is the motion stable or unstable? What is

the undamped resonant frequency?
Solution. Setting the left side of (a) equal to zero, and solving, we
obtain the complementary function

(b) yc = c cos (2t + a).

Since the right side of (a) agrees, except for phase, with the complementary
function (b), the trial function YP' by Lesson 21A, Case 2, must be of the
form
(c) YP = At sin 2t + Bt cos 2t.
Following the method outlined in this lesson, we find

(d) YP = it sin 2t.

Hence the general solution of (a) is

(e) y = c cos (2t + &) + it sin 2t.
The presence of the factor tj4 in the amplitude of the second term implies
that the displacement increases with time. The motion is therefore un-
342 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

stable. A comparison of (a) with (b) also shows that the condition of
undamped resonance is present since the frequency of the forcing function
and the natural (undamped) frequency of the system are both 2 radians
per unit of time. Hence the undamped resonant frequency is 2 rad/unit
of time.
Example 28.951. The differential equation of motion of a system is

(a) y" + 4y = 6 sin t.

Find its equation of motion. Is the motion stable or unstable?

Solution. The general solution of (a), by any method you wish to
use, is
(b) y = e cos (2t + &) + 2 sin t.
The maximum displacement is lei + 2, a finite quantity. Hence the motion
is stable.
Example 28.96. A 16-lb weight stretches a spring 6 in. A forcing
function J(t) = 10 sin 2t is attached to the system. Find the equation
of motion. What is the maximum displacement of the weight? Is the
motion stable or unstable? What frequency of the forcing function would
produce resonance?
Solution Because of the presence of the forcing function, (28.63)
must be modified to read
d 2y
(a) m dt 2 + ky = 10 sin 2t.

Here mg = 16, m = H = i. Since 16 pounds stretches the spring 6

inches, we have, by (28.62),
(b) ik = 16, k = 32.

Hence (a) becomes

1 d2y d 2y .
(c) 2 dt2 + 32y = 10 sin 2t, dt 2 + 64y = 20 sm 2t.

The general solution of (c) is

(d) y = e cos (8t + &) + 1 sin 2t.
Its maximum displacement is lei + 1, a finite quantity.
Therefore the
motion is stable. To produce resonance, the frequency of the forcing
function would have to be 8 rad/unit of time.
Lesson 28D-Exercise

EXERCISE 28D
1. Verify the accuracy of the solution of (28.81) as given in (28.84).
2. Verify the accuracy of the particular solution of (28.9) as given in (28.92).
3. (a) Solve (28.81) with w wo and initial conditions t = 0, Y = Yo,
v = Vo. Use for the Yc solution the form Yc = Cl sin wot +C2 cos wot.
(b) Write the solution if fJ = O.
4. In (28.81), replace sin (wt + fJ) by cos wt. Solve the equation, with w wo,
and initial conditions t = 0, y = Yo, v = Vo.
5. Solve (28.9) with fJ = 0 and initial conditions t = 0, y = Yo, v = vo.
When a forcing function f(t) is attached to a helical spring, the differ-
ential equation of motion, as given in (28.63), must be modified to read
2
d y
(28.961) m dt 2 + ky = f(t).

Use this equation to solve the helical spring problems below. Take positive
direction downward.
6. A 4-lb body stretches a helical spring 1 in. A forcing function f(t) =
t sin 8V6 t is attached to the system. Find the equation of motion. Is the
motion stable or unstable? What is the undamped resonant frequency?
7. A 16-lb body stretches a helical spring 4 in. A forcing function f(t) ==
sin 4y'6 t is attached to the system. After it is brought to rest, it is displaced
6 in. and given a downward velocity of 4 ft/sec. Find the equation of motion
of the body. Is the motion stable or unstable? What is the undamped
resonant frequency? Hint. At t = 0, y = I, II = 4.
8. An 8-lb body stretches a helical spring 2 ft. After it is brought to rest, a
forcing function f(t) = sin 6t is attached to the system causing it to vibrate.
Find:
(a) Distance and velocity as functions of time (Hint. At t = 0, Y = 0,
II = 0).
(b) The natural frequency of the system.
(c) The forcing frequency.
(d) The maximum possible displacement or departure of the body from its
equilibrium position.
(e) Whether the motion is stable or unstable.
9. Answer the same questions as in problem 8, if at t = 0, the body is held at
rest 4 ft below the equilibrium position and then given a velocity of -3 ft/sec.
In addition, write thc complementary function in the form Yc = c cos (kt+ 8).
10. Solve (28.8) if f(t) = Ft where F is a constant. Is the motion stable or
unstable?
ll. A particle weighing 16lb and moving on a horizontal line, is attracted to an
origin 0 by a force which is proportional to its distance from O. When the
particle is at x = -2, this force is 91b. In addition, a forcing functionf(t) =
sin 3t is impressed on the system. If at t = 0, x = 2, II = 0, find (a) the
equation of motion of the particle and (b) the resonant frequency of the
system.
12. In problem 11, change sin 3t to cos 2t. Find: (a) the equation of motion,
(b) the natural frequency of the system, (c) the forcing frequency, and
(d) the maximum possible displacement of the particle from O.
344 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

13. A body attached to a helical spring oscillates with a period of 7r18 sec. A
forcing function attached to the system produces resonance. What is the
frequency of the forcing function?
14. A mass m is attached to a helical spring whose spring constant is k. At
t = 0, it is brought to rest and a constant forcing function J(t) = lib Ib
is impressed on the system. After b sec, the force is removed.
(a) Find the position y of the mass as a function of time. Hint. First solve

°
with J(t) = lib. Find y(b) and y'(b). Now solve the equation with
J(t) = and initial conditions t = b, y = y(b), dYldt = y'(b).
Remark. After the input or forcing function lib is removed, note that it
still is possible to have an output yet). Note, too, that if b is small, say 1150,
thenJ(t) = 501b, but that it acts for only 1150 sec. It is as if the mass were
given a sudden blow by a force that was immediately removed. Finally note
that the output or response function yet) is continuous for t 6; 0, even though
the input or forcing function J(t) is discontinuous. The latter can be written
as

J(t) = f ° t b,
10, t > b.

If b = 0, J(t) does not exist. In engineering circles, however, the fictitious

forcing function J(t) which results when b = 0, is called a unit iRlpulse;
in physics it is called a Dirac 6-function.
(b) Show that as b -> 0, the solution y(t)-Iet us call it yo(t)-approaches
yo(t) = (l/kblklm sin vklm t.
Hint. Use the fact that lim (sin 818) = 1. Now prove that yo(t) satis-
'--.0
fies the equation m(d2Yldt 2 ) ky = + °
with initial conditions t = 0,
Y = 0, dYldt = 11m. The function yo(t) is called the bnpulsive re-
sponse or the response of the systeRl to a unit iRlpulse.
15. A 16-lb weight stretches a spring 8 ft. At t = 0, it is brought to rest and a
forcing function J(t), defined by

J(t) = {el, ° t 1,
0, t > 1
is impressed on the system. Find the equation of motion. (See hint in 14.)
A forcing function which has a different formula for a different time
interval is called an interRlittent force.
16. (a) Show that the solution of
2
d y 2
dt 2 + woY F sin (wt), w wo,

with initial conditions t = 0, Y = 0, dYldt = 0, is

(28.97) y = F
w0 2 - w2
(sm
. wt - -w.)
sm wot .
wo
Lesson 28D-Exercise 345

(b) Show that if w = wo +

E, where E > 0 is assumed to be small, then
the solution (28.97) becomes

(28.971) y (2wo : ) [sin wot - sin (wo

EE + E)tl + (2Wo : )EWO sin wot.
.
(c) Usmg th·d . sm
e l entity . A - . B =
sm A + B Sin
2 cos --2- . --2-
A - B , sh ow

that if we ignore the last term on the right of (28.971)-we shall refer
to it again later-(28.971) can be written as

(28.972) Y = -
2 wO
( F E)
+2 E
E). d]
[ 2 cos ( wo + 2 t sm"2 .

(d) As we pointed out in this lesson of the text, the phenomenon of un-
damped resonance occurs when w = woo The amplitude of the motion
then increases with time so that an unstable motion results. In this
problem, we have taken w = wo E, E +
0, so that w woo However,
as E --+ 0, w approaches the resonant frequency WOo We now make the
assumption that E, although not zero, is very small in comparison with
W00 Hence we commit a relatively small error if in (28.972) we replace
Wo + E/2 by W00 Show that then (28.972) becomes

(28.973) Y = (-
F. d)
EWO sm"2 cos wot.

(e) We can get an idea of the appearance of the graph of the motion given
by (28.973), if we look at the function defined by it, as a harmonic
motion cos wot with a time varying amplitude
F d·
(28.974) A = --sin-·
EWO 2
Equation (28.974) itself defines a simple harmonic motion whose period
is 411"/ E. Since E is assumed small, the period of A is large. This means
that the amplitude of (28.973) is varying slowly. It is, therefore, called
appropriately a slowly varying amplitude, and the function cos wot is
said to be amplitude modulated. The graph of A is given by the
broken lines in Fig. 28.975.
y

1.:.11 ,. "'

2wo" . . . . 2wo 2wo

- 1':01 'V,
--
Figure 28.975
346 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

By (28.973), show that for each value of t such that cos wot = =FI,
yet) = ± (F / EWo) sin (Et/2). Hence show that the graph of the solution
yet) will touch the upper part of the dotted curve in Fig. 28.975, for each
value of t such that cos wot = - I and touch the lower part of the
dotted curve for each value of t such that cos wot = 1.
(f) Show that yet) = 0 when

t = ... , (2n + I)'/r

2wo 2wo 2wo 2wo
and that the period of cos wot is therefore 2r/wo. Since wo is much
greater than E, this period 2r/wo is much smaller than the period 4r/E
of the slowly varying amplitude A of (28.974), whose graph is represented
by the dotted lines in Fig. 28.975. Show, therefore, that the graph of
yet) will thus resemble the solid lines shown in Fig. 28.975.
The variations in the amplitude of yet) are known as beats. When
the amplitude is largest, the sound is loudest. This phenomenon of beats
can be heard when two tuning forks with almost but not identical fre-
quencies are set into vibration simultaneously. Note that the last term
in (28.971), which was omitted in arriving at (28.973), represents a simple
harmonic motion. It does not affect the phenomenon of beats.
Remark 1. Each musical note in an instrument has a definite frequency
associated with it. When a standard note and its corresponding musical
note are sounded at the same time, beats will result if their frequencies differ
slightly, i.e., if they are not in tune. When the musical note of the instrument
is adjusted so that beats disappear, the musical note is then in tune with the
standard note. Can you see how this result can be used to tune an instru-
ment?
Remark 2. We assumed in this problem w = wo E. Therefore, as +
E -> 0, w -> wo, the natural frequency of the system. Hence the undamped
resonant case discussed in the text is the limit of the undamped modulated
vibrations discussed in this problem.

ANSWERS 28D

3. (a) y J...
wo
(v o - cos 13
wo - w 2) sin wot + (YO - sin 13
wo - w
2) cos wot
+ wo2 F- w
2 sin (wt + 13).
(b) y = J...
wo
(vo - 2
Fw
wo-w
2) sin wot + Yo cos wot + wo-w
2F 2sin wt.
4. y = Vo sin wot
wo
+ (YO - 2 F 2) cos wot
wo - w
+wo 2 -F w2 cos wt.

5. y
F +2 2wovo sm. wot + yo cos wot -
2
Ft cos wot
2 .
wo wo
6. y c sin (8v6 t + 8) - t cos (8v6 t). Unstable. Wo = 8v6 rad/sec.
Lesson 29A FREE DAMPED MOTION. (DAMPED HARMONIC MOTION) 347

7. y =
1 +9616V6.sm 4v.;;;6 t + ! cos 4v. ;;;6 t - V6 . ;;;
24 t cos 4v 6 t.
Unstable. wo = 4V6 rad/see.
8. (a) y = n(3 sin 4t - 2 sin 6t), v = !(eos 4t - cos 6t).
(b) Wo = 4 rad/see. (c) w = 6 rad/see. (d)! ft. (e) Stable.
9. (a) y = -1rJ sin 4t +
4 cos 4t - ! sin 6t
== 4.03 sin (4t + - ! sin 6t, where = 11" - Arc sin 80/80.5.
v = -16.1 cos (4t + - ! cos 6t.
(b), (c), (e) same as in 8. (d) 4.03 ! = 4.23 ft.+
10. y = Cl cos wot +
C2 sin wot+ Ft/w 2 • Unstable.
II. (a) x = i sin 3t +
2 cos 3t - !t cos 3t. (b) wo = 3 rad/see.
12. (a) x =
(d) i+
I
cos 2t
= 2.
+
! cos 3t. (b) 3 rad/see. (c) 2 rad/see.
13. 16 rad/see.

14. (1 - eosVk/m t), 0 t b,

y(t) [cosvk/m(t - b) - cosVk/mtJ

= sin [vk/m (t ]sinvk/m t > b.

15. _ !sin2t _ icos2t, 0 t 1,

!(2e sin 2 + e cos 2 - 1) sin 2t

y(t)

+ i(2e cos 2 - e sin 2 - 2) cos 2t, t > 1.

LESSON 29. Damped Motion.

In the previous lesson, we ignored the important factor of resistance or
damping. In this lesson we shall discuss the more realistic motion of a
particle that is subject to a resistance or damping force. We shall assume,
for illustrative purposes, that the resisting force is proportional to the
first power of the velocity. Frequently it will not be. In such cases more
complicated methods, beyond the scope of this text, will be needed to
solve the resulting differential equation.

LESSON 29A. Free Damped Motion. (Damped Harmonic Motion),

Definition 29.1. A particle will be said to execute free damped mo-
tion, more commonly called damped harmonic motion, if its equation
of motion satisfies a differential equation of the form
d 2y d 2y
(29.11) m dt 2 + 2mr dy
dt + mwo
2
y = 0, dt 2 + 2r dy
dt + Wo
2
y = 0,
 348 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

where the coefficient 2mr > 0 is called the coefficient of resistance of

the system. As before Wo is the natural (undamped) frequency of the
system and m is the mass of the particle.
The characteristic equation of (29.11) is m 2 + 2rm + wo 2 = 0, whose
roots are
(29.12)
The solution of (29.11) will thus depend on the character of the roots of
(29.12), i.e., whether they are real, imaginary, or multiple. We shall con-
sider each case separately.
Case 1. r2 > lIJo2 • If r2 > wo 2 , the roots in (29.12) are real and
unequal. Hence the solution of (29.11) is

(29.13)
Since both exponents in (29.13) are negative quantities (verify it) we can
write (29.13) as
(29.14)
If Cl ¢ 0, C2 ¢ 0, and ClI C2
have the same sign, then because e" > 0
for all z, there is no value of t for which y = O. Hence, in this case, the
graph of (29.14) cannot cross the t axis. If, however, Cl ¢ 0, C2 ¢ 0, and
ClI C2 have opposite signs, then setting y = 0 in (29.14) and solving it for
t will determine the t intercepts of its graph. Therefore setting y = 0 in
(29.14), we obtain
(29.15) e(A-B)t = C2, _
Cl

(A - B)t = log ( -C2) ,

t = __I-log (-C2) .
A - B Cl

From (29.15), we deduce that there can be only one value of t for which
y = O. We have thus shown that the curve representing the motion
given by (29.14) can cross the t axis once at most. Further, by (27.113),
y --+ 0 as t --+ 00. [Remember A and B in (29.14) are negative.]
Differentiation of (29.14) gives

(29.16) dy
dt = Cl
A eAt + C2 B eBt, A < 0, B < O.

Since this equation has the same form as (29.14), it, too, can have, at
most, only one value of t for which dy/dt = O. Hence the curve deter-
mined by (29.14) can have at most only one maximum or minimum point.
Leeson 29A FREE DAMPED MOTION. (DAMPED ILuwONIC MOTION) 349

The motion is therefore MnOscillatory and dies out with time. In Fig.
29.17 we have drawn graphs of a few possible motions.
y y

Figure 29.17

Comment 29.18. In this case, where r2 > "'0 2 , the resisting or damp-
ing force represented by r overpowers the restoring force represented by
"'0 and hence prevents oscillations. The system is called overdamped.
Example 29.19. A helical spring is stretched 32 inches by an object
weighing 2 pounds, and brought to rest. It is then given an additional
pull of 1 ft and released. If the spring is immersed in a medium whose
coefficient of resistance is find the equation of motion of the object.
Assume the resisting force is proportional to the first power of the ve-
locity. Also draw a rough graph of the motion.
Solution. Because of the presence of a resisting factor, whose coeffi-
cient of resistance is t, the differential equation of motion (28.63) for the
helical spring must be changed to read
d 2y 1 dy
(a) m dt 2 = -ky - 2 dt .

In this example, since 2 pounds stretches the spring 32 inches =t feet,

we have, by (28.62),
(b) tk = 2, k = i.
The mass m of the object is '* = n. Hence (a) becomes
1 d 2y 1 dy 3 d 2y dy
(c) 16 dt2 + +
2 dt 4 Y = 0, dt 2 + 8 dt + 12y = 0,
whose general solution is
(d)
The initial conditions are t = 0, y = 1, dy/dt = O. Substituting these
values in (c), we obtain
(e) 1 = Cl + C2,
o= -2Cl - 6C2.
350 PRoBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

The solution of (e) is Cl = t, /;2 = -,. Hence (d) becomes

(f)

Setting y = 0 in (f), we find t = -1 log 3 = -0.27. Setting y' = 0,

we find t = 0 and by the first equation in (f), y = 1 when t = O. Hence
the curve has a maximum at t = 0, y = 1. Setting y" = 0, we find
the curve has an inflection point at t = 1 log 3 = 0.27, Y = 0.77. A
rough graph of the motion is given in Fig. 29.191. The motion is non-
oscillatory. The maximum displacement occurs at t = 0, i.e., at the
beginning of its motion; the displacement then gradually dies out.
y

,,,'
,,"
-0.27 0.27

Figure 29.191

Case 2. r2 = flJrt If r2 = w02, the roots of (29.12) are - r twice.

Hence the solution of (29.11), by Lesson 20C, is
(29.2) y = cle-r! + C2te-r!,
y' = -rcle-rl + C2e-r! - rC2te-r!.

Since r > 0, by (27.113), both e-r ! and te-rl -+ 0 as t -+ 00. And as in the
previous case, there is only one value of t at most for which y and y' = O.
Therefore as in the previous case, the motion is nonoscillatory, and dies
out with time. The graphs of some of its possible motions are similar to
those shown in Fig. 29.17.
Comment 29.21. In this case, where r = Wo, the resisting or damp-
ing force represented by r is just as strong as the restoring force repre-
sented by Wo and hence prevents oscillations. For this reason the system
is said to be critically damped.
Case 3. r2 < Wo2. If r2 < w0 2, the roots in (29.12) are imaginary and
can be written as
(29.3)

The solution of (29.11), by Lesson 20D, is therefore

(29.31) y = ce-r ! sin (v'w02 - r2 t + 8).
Lesson 29A FREE DAMPED MOTION. (DAMPED HARMONIC MOTION) 351

Because of the sine tenn in the solution, the motion is oscillatory.

The damped ampHtude of the motion is ce- r ! and since r > 0, this
factor decreases as t increases and approaches zero as t approaches 00.
Hence with time, the particle vibrates with smaller and smaller oscilla-
tions about its equilibrium position.
Each function defined in (29.31) is not periodic since its values do not
repeat. However, because the motion is oscillatory, we say the function
is damped periodic and define its damped period to be the time it
takes the particle, starting at the equilibrium position, to make one com-
plete oscillation. Hence its damped period is said to be

(29.32)

The damped frequency of the motion is v' 000 2 - r2 radians per unit
of time, or v' 000 2 - r2/27r cycles per unit of time.
The exponential tenn e-r ! is called appropriately the damping factor.
Since this factor decreases with time, the motion eventually dies down.
When t = 1/r, the damping factor is lie. The time it takes the damping
factor to reach this value lie is called the time constant. Hence the
time constant T = 1/r.
A graph of the function defined by (29.31) is given in Fig. 29.33. It is
an oscillatory motion whose amplitude decreases with time.
y

Figure 29.33

Comment 29.34. In this case, where r2 < 000 2 , the damping force
represented by r is weaker than the restoring force represented by 000
and thus cannot prevent oscillations. For this reason the system is called
underdamped.
352 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

Example 29.35. If the coefficient of resistance in example 29.19 is

i instead of !, find:
1. The equation of motion of the system.
2. The damping factor.
3. The damped amplitude of the motion.
4. The damped perioo of the motion.
5. The damped frequency of the motion.
6. The time constant.
Solution. The differential equation of motion (c) in example 29.19
now becomes
2
(a) d y
dt 2
+ 6 dy
dt
+ 12 = 0
.
Its solution is
(b)
Differentiation of (b) gives
(c) y' = -3ce- 3t sin (0 t + 6) + c0 e- 3t cos (Va t + 6).
The initial conditions are t = 0, y = 1, y' = o. Substituting these values
in (b) and (c), we obtain
(d) 1 = c sin 6,
o = -3c sin 6 + v'3 c cos 6.
Substituting in the second equation of (d), the value of c as given in the
first equation, we obtain
(e) o= -3 + v'3 cot 6,
1r 71r
6="6 or 6'
Hence, by the first equation in (d), choosing sin 6 = !, we have,
(f) c = 2.
The equation of motion (b) therefore becomes

(g) y = 2e- 3t sin ( v'3 t + i) ,

which is the answer to 1. The answers to the remaining questions follow.
2. The damping factor is e- 3t .
3. The damped amplitude of the motion is 2e- 3t feet.
Lesson 29A-Exercise 353

4. The damped period of the motion is 211"/0 seconds.

5. The damped frequency of the motion is 0 rad/sec == 0 cps.
211"
6. The time constant T = -1 sec.

EXERCISE 29A
1. Verify the accuracy of the solution of (29.11) with r2 > wo 2 , as given in
(29.13).
2. Verify that each of the exponents in (29.13) is a negative quantity.
3. Verify the accuracy of the solution (d) of Example 29.19.
4. Verify the accuracy of the solution of (29.11) with r2 = wo 2 , as given in
(29.2)
5. Verify the accuracy of the solution of (29.11) with r2 < wo 2 , as given in
(29.31).
6. Verify the accuracy of the solution (b) of Example 29.35.
7. Show that the system whose differential equation is
d2 y
dt 2 + 2a dydt + b2 y = 0, a > 0,

is: (a) overdamped and the motion not oscillatory if a 2 > b2 , (b) critically
damped and the motion not oscillatory if a 2 = b2 , (c) underdamped and
the motion oscillatory if a 2 < b2 •
8. (a) Solve the differential equation
2
d y dy
(29.36) dt 2 + 2a dt + by = 0.
Note that here b is not squared as in 7.
(b) Show that the motion of the system is stable only if a > and b > 0.
(For definitions of stable and unstable, see Lesson 28D). Hint. Show
°
that if a > 0, b > 0, each independent solution of (29.36) approaches
zero as t -> 00. Hence the distance y from equilibrium approaches zero.
Consider each other possibility a > 0, b < 0; a < 0, b > 0; a < 0,
b < 0, and show that in each case y -> 00 as t -> 00.
°
(c) Show that if a < 0, b > and a2 < b, the motion, although unstable,
is oscillatory; if a > 0, b < 0, or if a < 0, b < 0, or if a < 0, b > 0,
and in each case a2 > b, the motion, although unstable, is not oscillatory.
9. With the help of the answers to problems 7 and 8, determine, without solving,
whether the motion of the system, whose differential equation is:
2 2
d y dy d y
(a) dt 2 - dt - 2y = 0. (e) dt 2 + 4 dydt - 4y = 0.
2 2
d y dy d y dy
(b) dt2 - 3 dt + 2y = 0. (f) dt 2 - 2 dt + 5y = 0.

h +2
(c) dt 2 dt + 5y = 0.
h + 6 dt + 6y
(g) dt 2 = 0.
2
(d) !!...!+
dt 2
4 dy + 4y
dt
= 0.
354 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

is stable or unstable; oscillatory or not oscillatory. Also determine whether

the system is underdamped, critically damped, or overdamped. Check your
answer by solving each equation. Draw a rough graph of each motion.
10. A particle moves on a straight line according to the law
2
d x dx
dt 2 + 2r dt + x = 0,

where r is a constant and x is the displacement of the particle from its equilib-
rium position.
(a) For what values of r will the motion be stable; unstable; oscillatory; not
oscillatory. For what values of r will the system be underdamped; criti-
cally damped; overdamped.
(b) Cheek your answers by solving the equation with r = !, r = 1, r = 2,
r = -t, r = -l.
(c) For what value of r will the motion be oscillatory and have a damped
period equal to 311"?
(d) Is there a value of r that will make the damped period less than 211'?
n. A particle moves on a straight line in accordance with the law
2
d x
dt 2 + 4 dx
dt + 13x = 0.

At t = 0, x = 0, v = 12 ft/sec.
(a) Solve thc equation for x as a function of t.
(b) What is the damping factor, the damped amplitude, the damped period,
the damped frequency, the time constant?
(c) Find the time required for the damped amplitude-and hence also for
thc damping factor-to decrease by 50 percent. Hint. The damped
amplitude is 4e- 2t • When t = 0, 4e- 21 = 4. You want t so that
4e- 2t = 2.
(d) What percentage of its original value has the damping factor, and there-
fore the damped amplitude, after one half period has elapsed? Hint.
The damped period is 211'/3. Evaluate e-2t when t = 11'/3. What is
the damped amplitude at that instant?
(e) Where is the particle and with what velocity is it moving when t
11'/6 sec?
(f) Draw a rough graph of the curve.
12. A particle moves in a straight line in accordance with the law
2
d x
dt2
+ 10 dx
dt
+ 16 X=.
0

At t = 0, x = 1 ft, v = 4 ft/sec.
(a) Find the equation of motion.
(b) Is the motion oscillatory'?
(c) What is the maximum value of x? When does x attain this maximum
value?
(d) Draw a rough graph of the curve. Does the curve cross the t axis for
t> o?
Lesson 29A-Exercise 355

13. A particle is executing damped harmonic motion. In 10 sec, the damping

factor has decreased by 80 percent. Its damped period is 2 sec. Find the
differential equation of motion. ,
14. A particle of mass m moves in a straight line. It is attracted toward the
origin by a force equal to k times its distance from the origin. The resistance
is 2R times the velocity. Find the maximum value of m so that the motion
will not be oscillatory.
15. A particle moves in a straight line in accordance with the law
d2x
dt 2 + 6 dx
dt - 16x = o.
At t = 0, the particle is at x = 2 ft and moving to the left with a velocity
of 10 ft/sec.
(a) When will the particle change direction and go to the right?
(b) Will it ever change direction again?
When the damping or resisting factor of a system is not negligible, the
differential equation (28.63) for the helical spring must be modified to read,
with downward direction positive,
2
d y
(29.37) m dt 2 + r dydt + ky = 0,

where we have assumed that the force of resistance is proportional to the

first power of the velocity and r > 0 is the coefficient of resistance of the
system. Note that here r replaces 2mr of (29.11).
Use (29.37) to solve the following problems, 16-25.
16. A weight of 16lb stretches a helical spring If ft. The coefficient of resistance
of the spring is 2. After it is brought to rest, it is given a velocity of 12 ft/sec.
(a) Find the equation of motion. Draw a rough graph of the motion.
(b) Find damping factor, damped amplitude, damped period, damped
frequency, time cunstant.
(c) When will the weight stop for the first time and change direction? How
far from equilibrium will it then be?
(d) When will it stop for the second time? How far from equilibrium will
it be?
(e) Write a formula which will give the times when the weight crosses the
equilibrium position and for the times of its successive stops.
17. A 16-lb weight stretches a spring 6 in. -The coefficient of resistance is 8. After
the spring is brought to rest, it is stretched an additional 3 in. and released.
Find the equation of motion. Draw a rough graph of the motion.
18. In problem 17, change the coefficient of resistance to 10. Find the equation
of motion. Draw a rough graph of the motion.
19. (a) Solve (29.37) if r2 < 4km and the initial conditions are t = 0, y = Yo,
" = O.
(b) What is the damped period of the motion?
(c) When will the damping factor, and therefore the damped amplitude, be
p percent of its initial value? Hint. The damping factor is e-rI/2m.
At t = 0, the damping factor is e-r l/ 2m = 1 = 100 percent. Therefore
want t such that e-rl/2m = p/lOO. Hence -rt/2m = log (p/l00), t =
-(2m/r) log (p/l00).
356 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

(d) Call the time obtained in (c) to sec. Therefore the damping factor at the
end of to sec is e- r ' oI2m and this damping factor, and therefore also the
damped amplitude, is p percent of its value at t = O. Show that at the
end of every period of to sec, the new damping factor is p percent of its
value at the beginning of the period. Hint. Show that e-r('o+'o)/2m =
p2/104, i.e., show that it is p2/104 of the original damped period and
hence is p percent of the damped period at the end of to sec. Or you can
let e-r' oI2m = 100 percent. Then want tl such that e-r .,/2m = p/100.
Find h = -(2m/r) log p/lOO as in (c).
(e) When t = T, the damped period of the motion, the damping factor is
e-rTI2m. It has a definite value, say q percent of the value of the damping
factor at t = O. Show that at the end of each period of T sec, the damp-
ing factor, and therefore the damped amplitude, is q percent of the damp-
ing factor at the beginning of the period. Hint. See (d) above. This
constant percentage, therefore, gives the percentage decrease in the
displacement of a particle from equilibrium at the end of a period as
compared with its displacement at the beginning of a period. Hence, the
damped amplitude at the end of a period of T sec = q percent of the
damped amplitude at the beginning of that period. Therefore,
I (Damped amplitude at the beginning of a period of T sec)
(29.38)
og Damped amplitude at the end of that period
= log (100/q) = a constant D.

'fhe constant D is called the logarithmic decrement. It is, as equa-

tion (29.38) shows, the constant positive difference between the
logarithm of the damped amplitude at the beginning of a period of T
sec and the logarithm of the damped amplitude at the end of that period.
(f) Find the logarithmic decrement of this problem. Hint. In (29.38) sub-
stitute the damped amplitude when t = 0 and when t = T as found
in (b).
(g) When will the body first reach the equilibrium position?
20. A 20-lb weight stretches a spring 3 in. After it comes to rest, it is given an
additional stretch of 2 in. and released. The internal resistance of the spring
is negligible but the resistance due to the air is 1/50 of its velocity.
(a) Find the equation of motion and draw a rough graph of its motion.
(b) Find the damped amplitude, damping factor, damped period, damped
frequency, time constant.
(c) When will the damping factor have decreased by 50 percent?
(d) Over what time intervals will the damping factor at the end of anJnter-
val be 50 percent of its value at the beginning of the interval?
(e) What percentage of its original value does the damping factor have, and
therefore also the amplitude, at the end of a period? Note by (e) of
problem 19 that the damping factor at the end of any period is this same
percentage of the damping factor at the beginning of that period.
(f) Find the logarithmic decrement.
(g) When does the particle first cross the equilibrium position?
21. The oscillatory motion of a spring is given by
2
d y
dt2 + 2a dy
dt + by = 0, a
2
< b.
Lesson 29A-Exercise 357

It is observed that the damping factor has decreased by 80 percent in 10 sec

and that its damped period is 2 sec. Find the values of a and b.
22. The natural frequency of a spring is 1 cps. After the spring is immersed in a
resisting medium, its frequency is reduced to i cps.
(a) What is the damping factor?
(b) What is the differential equation of motion?
23. In problem 21, find a and b if the period of the motion is 2 sec and the loga-
rithmic decrement is !.
24. The differential equation of motion of a body attached to a helical spring is
given by (29.37).
(a) Solve the equation if its mass m = r2I 4k and at t = 0, Y = YO, v = vo.
(b) Is the motion oscillatory or not oscillatory?
(c) When will it reach its maximum displacement from equilibrium?
(d) Show that from its maximum displacement it will move toward
equilibrium position but never reach it. Hint. Show that Y -+
t -+ 00 [see (27.113»).
° as
the

(e) Show that if rvo = -2kyo, the body will never change its direction but
will move continually toward the equilibrium position.
25. The differential equation of motion of a body attached to a helical spring is
given by (29.37).
(a) Solve the equation if r2 > 4km and at t = 0, y = 0, v = vo.
(b) Is the motion oscillatory or not oscillatory?
(c) Show that the solution can also be written in the form

2mvo -r./2m.
e
h vr2 2- 4km
t.
y=
vr2 - 4km
sm
m
Hint. See (18.9).
(d) When will the body reach its maximum displacement from equilibrium?
(e) Show that from its maximum displacement, it will move toward the
equilibrium position but never reach it.
We have included below only a few pendulum problems because of the similar-
ity in form of the pendulum equation and the helical spring equation-compare
(29.37) with (29.381) below. The same questions asked for the spring could be
asked for the pendulum. All one need do to obtain a solution for the pendulum
is to replace k in the previous answers by mgll and y by 8. Remember, linear
velocity v = 1 d81dt = lw, where w is angular velocity.
26. A simple pendulum of length l, with weight mg attached, swings in a medium
which offers a resisting force proportional to the first power of the linear
velocity. Show that the differential equation of motion is
2
(29.381)
dt 2
+ ..!:..mdtd8 + 1 8 = 0,

where r is the coefficient of resistance of the system. Hint. Adjust (28.71)

to take into account the resisting force, and remember linear velocity
d8
v = 1 dt .
358 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

27. (a) Show that the pendulum in problem 26 is overdamped and the motion
not oscillatory if r2/4m 2 > gil; critically damped and the motion not
oscillatory if r2/4m 2 = gil; underdamped and the motion oscillatory
if r2/4m 2 < gil.
(b) Solve (29.381) if at I = 0, (J = 0, w = wo and r2/4m 2 < gil.
28. A weight of 2 lb is attached to a pendulum 16 ft long. Find the smallest
positive value of the coefficient of resistance r for which the pendulum will not
oscillate.
29. A weight of 4 lb is attached to a pendulum swinging in a medium which
offers a resistance of one-eighth of the linear velocity. It is desired that the
period of the pendulum be 211". How long must the pendulum be?
30. (a) Solve (29.381) if the pendulum is released from the position (J = (Jo.
Assume oscillatory motion.
(b) When will the pendulum first reach the equilibrium position?

ANSWERS 29A
7. y = cle<-·+v'·2_b 2). +
c2e<-·-v'·2_b2)., b2 < a 2,
y = (Cl +
c21)e-·', b2 = a 2,
y = e-·'(cl cos Vb 2 - a 2 1+ C2 sin v'b 2 - a2 I), a 2 < b2.
8. y = cle<-,,+v'·2_b). + b < a2,
y = (Cl +
c21)e-·', b = a 2,
y = e-·'(ci cos 1+ C2 sin I), a 2 < b.
9. (a) Unstable, not oscillatory, overdamped, y = cle 2• +
c2e-'.
(b) Unstable, not oscillatory, overdamped, y = cle 2• +
c2e'.
(c) Stable, oscillatory, underdamped, y = ce-' sin (21 +
05).
(d) Stable, not oscillatory, critically damped, y = cle- 2• +
c2te- 2••
(e) Unstable, not oscillatory, overdamped, y = cle<-2+2{2lt +
c2e<-2-{2) •.
(f) Unstable, oscillatory, underdamped, y = ce' cos (21 +
05).
(g) Stable, not oscillatory, overdamped, y = ce<-3+-y'3)t +
c2e<-3-.,f3) ••

10. (a) Stable only if r 0; underdamped and oscillatory if 0 < r < 1;

critically damped and not oscillatory if r = 1; overdamped and not
oscillatory if r > 1; unstable and oscillatory if -1 < r < 0; unstable
and not oscillatory if r -1.
(b) y = ce- tl2 sin (Va 1/2 6),+
y = (Cl + c21)e-',
y = cle<-2+v'3lt + C2 e <-2-v'3).,
y = qe· /2 sin (Va 1/2 + 6), --------
y = (C! + c21)e'.
(c) r'= V5/3. (d) No.
11. (a) x = 4e- 2• sin 31. (b) e- 2., 4e- 2• ft, 211"/3 sec, 3 rad/sec or 3/2'11" cps,
! sec. (c) I = ! log 2 sec = 0.35 sec. (d) 12.3 percent, 0.49 ft.
(e) 1.4 ft, -2.8 ft/sec.
12. (a) x = 2e- 2 • - e- 8 '. (b) No. (c) x 1.19 ft, I = 0.12 sec.
(d) No.
2
d y
13. dl 2 + 0.322 dydl + 9.896y = O.
2
14. R /k.
15. (a) 0.223 sec. (b) No.
Lesson 29B FORCED MOTION WITH DAMPING 359

16. (a) y = 3e- 2t sin 4t. (b) e- 2t , 3e- 2t ft, 7r/2 sec, 4 rad/sec, ! sec.
(c) 0.277 sec, 1.54 ft. (d) 0.277 +
'IT/4 sec, -0.32 ft.
(e) n'IT/4 sec; 0.277 +
n'IT/4 sec, n = 0, 1,2, ....
8t
17. y = e- (t + 2t). 18. Y = te- - he- 16t •
4t
19• ()a y -- Yo '\JI4km4km_ r2 e-,tl2m cos (v4km2m- r2 t + 8) '
where 8 = Arc tan (-r /V 4km - r2).
(b) T = 4'ITm/V 4km - r2 sec.
(f) log decrement = 2u/V 4km - r2.
(g) t = m( 'IT - 28) /V 4km - r2 sec.
20. (a) y = ie-o.oI6t cos (Sy2 t +
8), approximately, where tan 8 = -0.0014,
8 = -0.0014 radian,
(b) ie-o.oI6t ft, e-O.OI6t, 'lTy2/S sec, Sy2 rad/sec, or 4y2/'IT cps, 62! sec.
(c) and (d) 43.3 sec. (e) 99 percent. (f) 0.009. (g) 0.14 sec.
21. a = 0.161, b = 9.S96.
22. (a) e-4·68t. (b) y" +
9.37y' +
39.5y = O.
23. a = 0.1, b = 9.SS.

24. (a) y = [YO + (V + YO

O 2: ) t] e-
2ktl
,.
(b) Not oscillatory. (c) t = vor2/2k(rvo + 2kyo) sec.
mvo -,tI2m( d,2-4kmI2m -t,J,2-4kml2m)
25. (a) y = e e -e
v'r2 - 4km
(b) Not oscillatory
vr2 - 4km v'r2 - 4km
(d) tanh 2 t = ,
m r
t = 2m tanh -I yr;.2=--_----,47'km-.
v'r2 - 4km r
I l -,tI2m. /g r2
27. (b) e= 2wom '\J 4m2g _ lr 2 e sm 1- 4m2 t.
1
28. r = . rn lb-sec/ft. 29. l = 25.6 ft.
4v2
30. In answers to 19(a) and (g), replace k by mg/l and y bye.

LESSON 29B. Forced Motion with Dam.ping. The motion of a

particle that satisfies the differential equation
d2y
(29.4) m dt 2 + 2mr dt + mwo 2y = f(t),
2
d y 1
dt2 + 2r dy
dt + Wo
2
y = mf(t),

where, as before, 2mr is the coefficient of resistance of the system, Wo is

the natural (undamped) frequency of the system, m is the mass of the
860 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

particle and f(t) is a forcing function attached to the system, is called

forced daDlped Dlotion in contrast to the free damped motion (i.e.,
damped harmonic motion) when f(t) == O. In engineering circles, f(t) is
called the input of the system and the solution y(t) of (29.4) the output
of the system. Let us assume the forcing function f(t) = mF sin (wt (3) +
where F is a constant. Then (29.4) becomes
2
d y
(29.41) dt 2 + 2r dy
dt + Wo y =
2 •
F sm wt
(
+ (3).
The different possible complementary functions y. obtained by setting
the left side of (29.41) equal to zero and solving it will be the same as.
those given in the three cases of Lesson 29A. The trial solution YP for all
such solutions y. is
(29.42) YP = A sin (wt + (3) + B cos (wt + IJ).
Following the method outlined in Lesson 21A, we find that
F(w0 2 _ w 2)
(29.43) A= [(w0 2 - w2)2 + (2rw)2] ,
B _ -F(2rw)
- [(w0 2 - w 2)2 +
(2rw)2]
Let (see Fig. 29.44),

Figure 29.44

(29.45)

Substituting these values in (29.43) and the resulting expressions for A

and B in (29.42), we obtain
F
(29.46) yp = V(w0 2 - w 2)2 + (2rw)2
X [cos a sin (wt + (3). - sin a cos (wt + (3)].
Lellson 29B FORCED MOTION WITH DAMPING 361

Hence the general solution of (29.41) is

(29.47) y = y. + F sin (wt + (3 - a),

v'(w0 2 - 00 2)2 + (2rw)2
where y. is anyone of the functions given in Lesson 29A. As we saw there,
the motion due to the y. part of the solution (29.47), in all cases, whether
oscillatory or nonoscillatory, dies out with time. For this reason this part
of the motion has been called appropriately the transient lIlotion. The
equation of motion (29.47) is thus a complicated one only for the time in
which the transient motion is effective. Thereafter the motion will be
due entirely to the YP part of the solution as given by the second term on
the right of (29.47). This part of the motion has therefore been appro-
priately named the steady state lIlotion.
Comment 29.48. In many physical problems, the transient motion is
the least important part of the motion. However, there are cases where it
is of major importance.
By (29.41) and (29.47), we see that the steady state motion has the
same frequency as the forcing function I(t) , namely 00 rad/sec, but is out
of phase with it and that the amplitude of the steady state motion is

(29.5)

If '" = "'0 [the condition for (undamped) resonance], the amplitude re-
duces to the interesting form
F
(29.51) A=-·
2rwo
If '" ;t6 "'0, then by differentiating (29.5) with respect to 00 and setting
the resulting expression for dA/dw equal to zero, we obtain
(29.52)
from which we find
(29.53) 00
2 = 000
2 - 2r2,
Hence if a resisting force is present, and if 00, the frequency of the forcing
function, is not equal to 000, the natural (undamped) frequency of a sys-
tem, then, for fixed F, the amplitude A of the steady state motion will
be a maximum if 00 has the value given in (29.53). A forcing function I(t) ,
having this frequency 00, is then said to be in resonance with the system.
Substituting this value of 00 in (29.5), we find that the maximum ampli-
tude is
(29.531)
362 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

Assume now that 2r, the coefficient of resistance of a system per unit
mass, is small. Hence we commit a small error if we omit the r2 term
in (29.531). We thus obtain
F
:0:;--,
(29.532)
2rwo

the same amplitude obtained in (29.51) when w = woo Further, we

showed in Lesson 29A, Case 3, that the natural (damped) frequency of a
system is v' wo 2 - r2, which, for small r, is close to the resonant fre-
quency v' wo 2 - 2r2, i.e., it is close to the frequency which will produce
the maximum amplitude.
We infer from all the above remarks that if a resisting force is present
and w, the fr.equency of the forcing function f(t), equals wo, the natural
(undamped) frequency of a system, or is close to v'wo2 - r2, the natural
(damped) frequency of the system, then the amplitude of the system is inversely
proportional to the damping or resisting factor 2r. Hence if 2r is small,
A will be large, and tremendous vibrations may be produced. That is
why soldiers crossing a bridge may be ordered to break step (although
the chances are that this precaution is unnecessary), for it is feared that
if the frequency which they create with their footbeat is the same as the
natural (undamped) frequency of the bridge, or near its damped frequency,
and if in addition the internal resistance of the bridge is small, the vibra-
tions may become so large as to cause a breakage. The walls of Jericho,
so some assert, came tumbling down because the sound the trumpeteers
made with their trumpets caused a wave motion whose frequency equaled
the natural (undamped) frequency of the walls. Students at Cornell
University used to find it amusing either to create a wave motion in the
old suspension bridge over the gorge or to get it to swing violently from
side to side. They would march across it in a straight line with a rhythmic
beat or walk with a sailor's gait, first emphasizing one side, then the other.
To timid souls, however, it was never very amusing-terrifying would be
'l. more descriptive word. On such occasions, it was impossible to walk
across the bridge with an even step or in a straightline, depending on
whether the bridge was waving or swinging.
We cite two more examples of this phenomenon and ones which you
can easily experience or may have already experienced.
1. A swing, with a child seated on it, when displaced from its equi-
librium position, will move back and forth across the equilibrium position
with a natural (damped) frequency. If you now apply a force to the
swing with a frequency close to this natural (damped) frequency, then for
a fixed F and small r, the maximum amplitude will equal, approximately,
F /2rwo. Hence, if r is small, the amplitude of swing will be large. If you
want a still larger amplitude, you must increase F.
Lesson 29B FORCED MOTION WITH DAMPING 363

2. When you jump off a diving board, the end of the board will vibrate
about its equilibrium position with a natural (damped) frequency. If
instead of jumping off, you now jump up and down above the end of the
board with a frequency near this natural (damped) frequency, you will
be able to make the magnitude of the oscillation large. If r is small, the
maximum amplitude, for a given F, will equal, approximately, F /2rwo.
The ratio
(29.54) M = Amplitude of Y1!. ,
F/wo2
where F and wo 2 are given in (29.41), is called the nlagnification ratio
of the systenl or the anlplification ratio of the systenl. By (29.54)
and (29.47), this magnification ratio is

(29.55)

Since Wo is fixed, the amplification ratio of a system depends on the fre-

quency w of the forcing function J(t) and the coefficient of resistance per
unit mass 2r. In practical applications where w is also fixed, the resistance
2r is made large if one wishes the magnifying response to be small as, for
example, in vibrations of machinery and in shock absorbers; the resistance
2r is made small, if one wishes the response to be large, as, for example,
in a radio receiver.
If in (29.55), we let
r
(29.56) JJ. = and 11=-,
Wo Wo
the equation becomes
1
(29.561)
M =

The quantity JJ., by (29.56), is thus the ratio of the impressed or input
frequency w to the natural (undamped) frequency woo The quantity II
may be looked at as measuring the amount of damping present for a
fixed woo For each fixed value of II, M is a function of JJ.. Hence it is pos-
sible to draw a graph of the magnification M for each such fixed value
of II. For example, if II is i, then, by (29.561),

(29.562)
364 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

If II = 0, which implies by (29.56) that r = 0, then, by (29.561),

1
(29.563) M(p,) = 1 _ p,2'

By (29.563), we see that as p, -+ 1, which implies by (29.56) that

M -+ 00.
'" -+ "'0,

Example 29.564. A forcing function f(t) = i cos 2t is applied to the

motion given in Example 29.19. Find the steady state motion and the
amplification ratio of the system. Is resonance possible?
Solution. With f(t) = i cos 2t, the differential equation of motion
(c) in Example 29.19 becomes
1 d 2y 1 dy 3 5
(a) 16 dt2 + 2 dt + 4Y = 2 cos 2t,
d 2y
(b) dt 2 + 8 dy
dt + 12y = 40 cos 2t.

A particular solution of (b) is

(c) YP = 2 sin 2t + cos 2t,

which is the steady state motion. By Comment 28.32, the amplitude of
the motion defined by (c) is V22 +
12 = 0. Comparing (b) with
(29.41), we see that ",2 = 12, 2r = 8, F = 40. Therefore by (29.54),
the magnification ratio of the system is

(d)
V5 '3V5
M = 40/12 = 10'
And since "'0
2
= 12 < 2r2 = 32, resonance is not possible. See (29.53).

Comment 29.6. For easy reference, we have listed in the table on

page 365 the different differential equations discussed thus far in this
chapter, and the pertinent information related to each. -------

EXERCISE 29B
1. Verify the values of A and B as given in (29.43).
2. Verify the solution (29.46).
3. Verify (29.53).
4. Verify the accuracy of the solution (c) of Example 29.564.
5. For what value of CAl will the magnification ratio as given in (29.55) be a
maximum? Find this maximum value.
6. A particle moves according to the law
d2 y
dt 2 + 2 dydt + 9y .
= 5 sm 2t.
 Differential Equations Discussed in Lessons 28 and 29

.
E quatlOn Sit'
0 u IOn ofName
Motion AmpI'Itud e of Type
Motion Frequency

Y.=QCoswot+c2sinwot Simple. y'C12+ C2 2 wo rad or

dt 2 harmonIC c motIOn. wo
y. = c sin (wot + 0) or free Stable. 2'11' cycles per
undamped C • •
y. = c cos (wot+ 0) motion. UnIt tIme.

2
d y
dt2
+ wo2y Y = y. as given above + Forced
c
+ w02 F_ w2
Oscillatory.
Stable.
F . motIon.
= F sin (wt+ (3) w02 _ w2 sm (wt+ (3), wo w

F F (Resonance)
y = y. - 2- t cos (wot+ (3), c - -2 t Unstable
wo w o '
wo = W

2
d y dy = c e<-'+""-..02)1 Damped None Non-
dt2 +2r dt y. I (_'_""-"0')1 oscillatory.
2_ 0 + C2e ,r > Wo· motIOn Stable.
+
y. = cle-"+C2te-",r = wo damt,Ped
mo IOn.

y.=ce-"sin(Vw02-r2t+o) ce-" Oscillatory. y'w02-r2rad

, Stable. or
r < wo r--o----=-
y'w02 - r2
2'11' cycles
per unit of time.
d2Y + dy + 2
-dt 2 2r -dt wo y
+ Fsin (wt+{3 -a)
y -- y • • I( 2 _ 2)2 + (2)2
Forced
d d
+ F Oscillatory. Of steady state
• I(e2 _
S
2)2 + (2)2 ta ble. mo t'lOn, w rad or
. v wo w rw motIOn. v wo w rw w/2'11' cycles per
= F sm (wt + (3) unit time.
366 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

(a) Find the steady state motion; also the amplitude, period, and frequency
of the steady state motion.
(b) What is the magnification ratio of the system?
(c) What frequency of the forcing function will produce resonance?
7. A particle moves according to the law
2
d y
m dt 2 + 2mr dydt + mwo2 y = f(t).

(a) Find the equation of motion if f(t) = mF cos wt. Assume r2 < w0 2.
(b) What is the transient motion; the steady state motion?
(c) What is the amplification ratio of the system?
8. A particle moves in accordance with the law
d 2y
dt 2 + 4 dydt + 16y = f(t).

(a) What frequency of the function f(t) will make the period of the steady
state motion T/3?
(b) What frequency of the function f(t) will produce resonance?
9. A particle moves according to the law
2

dt 2
+ 5 dydt + 6Y = e-'sin 2t.

(a) Solve for y as a function of t.

(b) What is the input; the output?
(c) Describe the motion.
10. In Exercise 29A, 8, we asked you to show that the motion of a particle whose
differential equation is + 2a + by = 0, is stable only if a > 0,
b > O. Since the addition to the equation of a function f(t) does not affect
the complementary function y., it follows that a > 0, b > 0 is also a
necessary condition for the stability of the motion of a particle whose
differential equation is
d2y
dt 2 + 2a dydt + by = f(t).
Prove that it is not a sufficient condition by solving the equation
2

+ 5 + 6y = 12e',

and then showing that the solution y(t) --+ 00 as t --+ 00.

When the damping or resisting factor of a system is not negligible and

a forcing function f(t) is attached to it, the differential equation (28.63)
for the helical spring must be modified to read [See also (29.37).]
2
d y
(29.7) m dt 2 + r dy
dt + ky = J(t),
Lesson 29B-Exercise 367

where r is the coefficient of resistance of the system.- Use (29.7) to solve

the next two problems.
11. A 16-lb weight stretches a spring 1 ft. The spring is immersed in a medium
whose coefficient of resistance is 4. After the spring is brought to rest, a
forcing function 10 sin 2t is applied to the system.
(a) Find the equation of motion.
(b) What is the transient motion; the steady state motion?
(c) Find the amplitude, period, and frequency of the steady state motion.
(d) What is the magnification ratio of the system?
12. A 16-lb weight stretches a spring 6 in. Its coefficient of resistance is 2. The
16-lb weight is removed, replaced by a 64-lb weight and brought to rest.
At t = 0, a forcing function 8 cos 4t is applied to the system. Find the
steady state motion and the amplification ratio of the system.
13. In (29.4), let

f(t) = m(AI sin WIt + A2 sin w2t + ... + A .. sin w..t),

so that n different oscillations are impressed on the system.
(a) Find the steady state motion. Hint. Use the superposition principle,
see Comment 24.25; also Exercise 19,6.
(b) What is the magnification ratio due to the input mAl sin WIt, to
mA2 sin w2t, ... , to mA .. sin w..tt A glance at the denominator of each
magnification ratio term will show that those terms with frequencies
close to wo will be magnified to a much larger extent than those with
frequencies farther away. A system of this kind thus acts as a filter.
It responds to those vibrations with frequencies near wo and ignores
those vibrations with frequencies not near woo
14. In Exercise 28D, 14, we introduced the discontinuous unit impulse function

f(t) f o t b,
to,
=
t > b.
Solve the equation
2
d y
dt 2 + 2 dydt + 2y = f(t),

for y as a function of t, where f(t) is the above function and initial conditions
are t = 0, Y = 0, y' = O. Hint. First solve withf(t) = lib. Find y(b) and
y'(b). Then solve the equation with f(t) = 0 and initial conditions t = b,
y = y(b), dyldt = y'(b).
15. Solve problem 14 if
f(t) = {1l_1,
0, t > 1.
Hint. See suggestions given in 14.
368 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

ANSWERS 29B
S. W = V w0 2 - 2r2, the same value of w that makes the amplitude a maxi-
mum, see (29.53); M(w) = 1/2rv'w02 - r2.

6. (a) yp = _5_ sin (2t - a), where a = Arc tan t,

V41
5V41, 'If, 2 rad/sec. (b) 1V41. (c) ....n.
7. (a) y = ce- rt cos (v'w02 _ r2 t + 6) + F cos (wt - a) , with
v' (w0 2 - ( 2)2 (2rw)2 +
r2 < w0 2 and a given by (29.45). (b) First term on right of (a); second
term on right of (a). (c) Same as (29.55).
8. (a) w = 6. (b) w = vS.
-t
9. ( a ) y = cte -2t + C2e -3t - e20 (.
sm 2t + 3 cos 2t).
(b) e-' sin 2t; solution yet) as given in (a).
(c) Each term in yet) approaches zero as t --> co. The complementary
function is not oscillatory; the particular solution, however, is damped
oscillatory since y --> 0 as t --> co.
10. Y = cle- 2 ' + c2e- + e' --> co as t --> co.
3'
-4t
II. (a) y = e26 (sin 4t + 8 cos 4t) + h(7 sin 2t - 4 cos 2t).
(b) First term in (a); second term in (a). (c) V65/13, 'If sec, 1/'If cps.
(d) 8V65/65.
12. yp = sin 4t; 4.
13. (a) yp = Al sin (Wit - al) + ... + A" sin (w"t - a,,) ,
V(w0 2 - (12)2 + (2rwI)2 V(w0 2 - w,,2)2 + (2rw,,)2
where ai, i = 1, .•• , n, is defined as in (29.45).
2 2
(b) "'0 "'0,
V(w0 2 - (12)2 + (2rwI)2'···' V(w02 - w,,2/,2 + (2rwn)2
14.
;b [1 - e-'(sin t + cos t)], 0 t b.
y = -t
e
2b
[{eb(sin b + cos b) - I} sin t

+ {eb(cos b - sin b) - I} cos tl, t > b.

t
IS. e- (1 - cos t), 0 t 1.
y
{
e-t[sin 1 sin t + (cos 1 - 1) cos tl
= e-t[cos (t - 1) - cos tl, t > 1.
Lesson 30A SIMPLE ELECTRIC CIRCUIT 369

LESSON 30. Electric Circuits. Analog Computation.

By Newton's laws of motion we were able to set up a relationship among

active forces in a mechanical system. Analogous laws, known as Kirch-
hoff's (1824-1887) laws, make it likewise possible for us to set up a rela-
tionship among those forces which supply and use energy in an electrical
system. In Lesson 30A below, we state one of these laws and apply it to
a simple electric circuit.

LESSON 30A. Simple Electric Circuit. In the simple electric circuit

which we have diagrammed in Fig. 30.1, the source of energy in the cir-
cuit is marked E. It may be a cell, battery, or generator. It supplies the
energy in the form of an electrical flow of charged particles. The velocity
of the particles is called a current. However, the energy source will
produce this flow only when the key at A is moved to B. The circuit is
then said to be closed. The electromotive
force of the battery or other source of
energy, usually written as emf, is defined
as numerically equal to the energy sup-
L
plied by the battery or source when one
unit charge is carried around the complete
circuit. For example, if three units of
energy are supplied by a source when one
unit charge is carried around the complete Figure 30.1
circuit, then its emf is three units. There
are, for the electrical system, as in the mechanical one, different systems of
units in use. In the one we shall adopt, the unit of emf is called a volt.
The other three elements in the circuit labeled R, L, and C are users of
energy. In nontechnical terms, this means that a certain amount of energy
is needed to move the electrical flow of charged particles across these
barriers. We express the energy each uses by giving the voltage drop
across it. *
From the physicist, we learn that:

(30.11) the voltage drop across a resistor (R in figure) = Ri,

the voltage drop across an inductor (L in figure) = L ,

the voltage drop across a capacitor (C in figure) = b q,

·The voltage drop across each element is easily measured by means of an instrument
called a voltmeter. All one need do is to connect one wire of the voltmeter to one side
of the element, another wire to the other side, and then read how far a pointer moves.
370 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

provided:
the resistance R of the resistor is measured in ohms,
the coefficient of inductance L of the inductor is measured in
henrys,
the capacitance C of the capacitor is measured in farads,
the charge q in the circuit is measured in coulombs,
the current i in the circuit, which is defined to be the rate of change
of the charge q, or the velocity of q, i.e.,
. dq
(30.12) = dt'
is measured in amperes.
The resistor, as the name implies, resists the How of the charged par-
ticles, and thus energy is needed to move the particles across it. The
inductor's job is to keep the rate of How of the charj!:ed particles as near
constant as possible. It thus opposes an increa!"'J or a decrease in the
current. The capacitor stores charged particles and thus interrupts the
electrical How. When the accumulated charges become too numerous for
its capacity, the charged particles leap across the gap (that is when the
spark occurs) and the particles then continue their course in the circuit.
Kirchhoft"s second law states that the sum of the voltage drops in a
closed circuit is equal to the electromotive force of the source of energy
E(t). Hence, by (30.11),

(30.13) Ri +L + bq = E(t).

By (30.12), we can write (30.13) as

d2q
(30.14) L dt 2 + R dq 1
dt + C q = F(t),

which is the differential equation of motion J

the charge q in the circuit
as a function of the time t.
To find the current i in the circuit as a function of the time t, we can
either solve (30.14) for q and take its derivative, or we can differentiate
(30.13) to obtain, with the help of (30.12), the differential equation
d 2i
(30.15) L dt 2 + R didt + C1. = d
dt E(t),
and then solve (30.15) for i.
Assume
(30.16) E(t) = F sin (wt + (J); therefore i E(t) = Fw cos (wt + (J).
Lesson 30A SIMPLE ELECTRIC CIRCUIT 371

Then (30.15) becomes

d 2i di 1.
(30.17) L dt 2 + R dt + at = Fwcos (wt + (J),
d 2i R di 1. Fw
dt2 + L dt + CL t = L cos (wt + (J).
Its solution, by any method you wish to use, is, assuming the roots of the
characteristic equation are imaginary,

(30.18) i = Ae-(R I2L)t sin R2C2 t + 6)

2
+ F C [RWC sin (wt + (J) + (1 - CL( ) cos (wt + (J)] •
w (RWC)2 + (1 - CL(2)2

Let (Fig. 30.211)

1 - CLw 2
(30.19) sin a = 2;=+=:: :;(;: :1=-==;C;=:L=W=::;2)=;::2
RwC
cos a = V(RwC)2 + (1 - CL( 2)2
Then the ip part of (30.18) can be written as
. FwC
(30.2)
tp = 2;=+===(;':=1=-==;C===L=w===2);=;::2
X [sin (wt + (J) cos a + cos (wt + (J) sin a]

FwC [sin (wt + (J + a)].

V(RwC)2 + (1 - CL( 2)2
Hence the solution (30.18) becomes

(30.21) i = Ae-(RI2L)t sin R2C2 t + 6)

+ FwC sin (wt + (J + a)

V(RwC)2 + (1 - CLw2)2

The current in the circuit, therefore consists of two parts, a damped

harmonic motion due to the ic part of the solution and a simple harmonic
372 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

motion due to the i p part. As in the mechanical case, the presence of the
damping factor e-<R/2Llt causes the current due to the ic part of the solu-
tion to die out in time. [If we had assumed real roots of the characteristic
equation of (30.17), instead of im-
aginary ones, the current due to the
ic part of the solution would still die
2
out in time. See the solutions for
1 - CLw each of the different cases in the cor-
responding mechanical case, Lesson
RwC 29A.J The ic part of the solution is
therefore called appropriately the
Figure 30.211 transient current. The current
equation (30.21) will thus be a com-
plicated one only for the time in which the transient Current is effective.
Thereafter the current will be determined entirely by the i p part of the
solution. The i p current has therefore been named, also appropriately,
the steady state current.
Comment 30.212. As in the mechanical case, the function E(t) of
(30.14) or E(t) of (30.15) is called the input of the system; the solution

of each equation the output of the respective system.

By (30.21) and (30.16), we see that the steady state current has the
same frequency as that of the energy source E(t), namely w rad/sec, but
is out of phase with it.
The amplitude of the steady state current is, by (30.21),
FwC F
(30.22) A
V(RWC)2 + (1 - CLW 2)2

The denominator of the last expression in (30.22) is called the iin-

pedance Z of the circuit. When its value is a minimum, the amplitude
A is a maximum. To find the value of w that will make Z a minimum, for
fixed R, C, and L, we differentiate the impedance equation

(30.23) Z = R2 + - Lw) 2
with respect to wand set dZ/dw equal to zero. The result is [square Z in
(30.23) and then differentiate with respect to wJ

(30.24) o= 2 (J....
we - LW) ( - Cw
_1 - 2 L)
'
Lesson 30A SIMPLE ELECTRIC CIRCUIT 373

from which we obtain

(30.25) 2 1
w = CL' W = v 1/CL.

For this value of w, the impedance Z of the current will be a minimum,

the amplitude A will be a maximum, and as in the mechanical system, we
say the electromotive force is in resonance with the circuit.
Substituting in the first equation of (30.22), this resonant value of
w 2 = 1/CL as given in (30.25), we obtain for the maximum value of the
amplitude,

(30.26) A = !:..
R
From (30.26) we observe that when resonance occurs, the maximum value
of the amplitude A is inversely proportional to the resistance R. Hence
when R is small, the maximum value of A is large, and when R is large,
the maximum amplitude is small. The condition of resonance Lherefore is
always dangerous unless the resistance R is sufficiently large to prevent a
breakdown of the circuit. And if R = 0, a breakdown is bound to occur.
If we fix F, R, L, and w, then by (30.22), the amplitude A of the steady
state current is a function of the capacitance C. If C = 0, A = 0, and
if C is adjusted so that is equal to the frequency w of E(t), then
A will be largest. Hence by adjusting C, we can make the amplitude of
the steady state current small or large. In a public address system when
we want a large amplification ratio [this means, by (29.54), that we want
the amplitude A of Yv to be relatively large], or in a home radio set when
we want a lower amplification ratio, we adjust the capacitance C accord-
ingly by turning a dial.

Example 30.27. A capacitor whose capacitance is 2/1010 farad, an

inductor whose coefficient of inductance is ..Jo henry, and a resistor whose
resistance is 1 ohm are connected in series. If at t = 0, i = 0 and the
charge on the capacitor is 1 coulomb, find the charge and the current in
the circuit due to the discharge of the capacitor when t = 0.01 second.
Solution. Here E(t) = 0, C = TIro, L = ..Jo, and R = 1. Hence
(30.14) becomes

1 d 2q dq
(a) 20 dt2 + 1 dt + 505q = 0,

Its solution is
(b) q = e-10t(cl sin lOOt + C2 cos lOOt).
374 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

Therefore
(c) i = = -lOe-10t(cl sin lOOt + C2 cos lOOt)
+ e-10t(100Cl cos lOOt - 100C2 sin lOOt).
The initial conditions are t = 0, q = I, i = o. Substituting these values
in (b) and (c), we obtain
(d) I = C2,

0= -lOc2 + 100cl, Cl = 0.1.

In (b) and (c) replace Cl and C2 by these values. Then when t = 0.01
there results
(e) q(O.OI) = e-o.1(0.1 sin I +
cos 1) = 0.57 coulomb,
i(O.OI) = -lOe-O.1(0.1 sin I cos I) + +
e-o.1(10 cos I - 100 sin I)
= -76.9 amperes.
The negative current indicates that the condenser is discharging, i.e.,
the charged particles are moving in a direction opposite to the one in
which they moved when the capacitor was being charged.
Example 30.3. To the circuit of the previous problem is added a
source of energy whose electromotive force E = 50 sin 120t. Change the
capacitance of the capacitor to 2 X 10- 3 farad. At t = 0 seconds, the
switch is closed. If at that instant there is no charge on the capacitor and
no current in the circuit, find:
1. The equation of motion of the steady state current after the switch
is closed.
2. The amplitude of the steady state current.
3. The frequency of the steady state current.
4. The value of the capacitance which will make the amplitude of the
steady state current a maximum.

Solution. Here E(t) = (50 sin 120t) = 6000 cos 120t. Using the
figures for L. and R as given in Example 30.27 and of C as given above,
the differential equation of motion (30.15) becomes
1 d 2i 3
(a) 20 dt2 + didt + 210 •
= 6000 cos 120t,
2
d 2i
dt
+ 20 didt + 104i = 120,000 cos 120t.

Its steady state solution is

(b) i p = 11.5 sin 120t - 21 cos 120t.
Leason 30B ANALOG COMPUTATION 375

By Comment 28.32, the amplitude of the steady state current is

(c) A = Vll.5 2 + 212 = 23.9.
The frequency w of the steady state current is 120 rad/sec, the same as
the frequency of the source of energy E(t).
By (30.25), the amplitude of the steady state current will be a maximum
if C has a value such that Vl/CL = w, i.e., when
1 20 1
(d) C = Lw2 = 120 2 = 720 farad.
LESSON 30B. Analog Com.putation. We recopy below the differen-
tial equation of motion (28.63) of a mechanical system with the coefficient
of resistance and forcing function tenns added, and the differential equa-
tions (30.14) and (30.15) of an electrical system.
2
(30.4) m ddt 2y + r dy
dt + k y = ]i'.
sm wt.
2
d q dq 1
(30.41) L dt 2 + R dt + C q = E(t).

d 2i
(30.42) L dt 2 + R didt + C1. = d
de [E(t)].
When placed underneath each other in this manner, the similarity in
fonn of the two systems is striking. It should be evident to you that if
in an electric circuit, Fig. 30.43 (a) , we insert a resistor R = r, an inductor
L = m, a capacitor C = l/k, and a source of energy E = F sin wt (or

!
= spring constant

L=m I Forcing function

E
, =F sin (wt)
m
Dashpot whose

L-_---l.
I of
resIstance IS r
(a) (b)
Figure 30.43

-Fw cos wt), the solution q of (30.41) [or i of (30.42)] will be the same
as the solution y of (30.4). By solving the electrical system, it is then
possible to determine the motion of a corresponding mechanical system,
such as the one pictured in Fig. 30.43(b). Since it is usually less expensive
376 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

and easier to set up a simple electric circuit than it is to construct a me-

chanical system, the importance of this fortunate coincidence should be
evident to you. This method, which is now well developed, of computing
the motion of a mechanical system from a simple electric circuit is known
as analog computation.
However, because of the current accessibility to high-speed digital
computers, the most accurate and least expensive method at present of
computing the motion of a mechanical system is to use such a computer.

EXERCISE 30
1. Verify the accuracy of the solution of (30.17) as given in (30.18).
2. Verify the accuracy of the solution (b) of Example 30.27.
3. Verify the accuracy of the solution (b) of Example 30.3.
In the problems below, it is assumed, when not explicitly stated, that
the coefficient of inductance L of the inductor is measured in henrys, the
resistance R of the resistor is measured in ohms, the capacitance C of the
capacitor is measured in farads, the charge q is in coulombs, the current i
is in amperes and the emf of the source of energy is in volts.
4. If the emf i.e., if the source of energy, is missing from the circuit, then the
differential equations (30.14) and (30.15) become respectively
d 2q dq 1
(30.5) Ldt2+Rdt+Cq = 0,
2
(30.51) L di
dt2
+ R
dt C
. = o.
J

(a) What is the natural (undamped) frequency of vibrations of current and

charge? Hint. Set R = o.
(b) For what values or R will the charge and current subside to zero without
oscillating; for what values of R will they oscillate before subsiding to
zero?
(c) Find the general solutions for q and i as functions of time if R2 = 4L/C.
To what mechanical case is this situation comparable?
(d) Find q and i as functions of time if at t = 0, q = qo and i = o. Assume
R2 < 4L/C.
5. For a certain LRO electric circuit, L = !, C = m.
(a) For what values of R will the current subside to zero without oscillating
after the emf is removed from the circuit; for what value of R will it
subside to zero with oscillations?
(b) What is the natural (undamped) frequency of the system?
6. A capacitor whose capacitance is 10- 5 farad, an inductor whose coefficient
of inductance is 10 henrys, and a resistor whose resistance is 3 ohms are
connected in series. At t = 0, i = 0 and the charge on the capacitor is
0.5 coulomb. Find the charge and current in the circuit as functions of time
due to the discharge of the capacitor.
Lesson 3O-Exercise 377

7. If a resistance is missing from the circuit, then by (30.14)

2
d q
(30.52) L dt 2 + Cq = E(t).

Equation (30.52) is the differential equation of the harmonic oscillator for

the electric current and corresponds to the forced undamped motion of the
mechanical system, see Lesson 28D.
(a) Solve for q and i as functions of time if E(t) = 0 and t = 0, q = qO,
i = O. What is the natural (undamped) frequency of the system?
(b) Solve for q and i as functions of time if E(t) = a constant emf E and
t = 0, q = 0, i = O.
(c) Solve for q and i as functions of time if E(t) = E sin wt and t = 0,
q = 0, i = 0 (two cases). What value of w will produce (undamped)
resonance?
8. (a) Find q and i as functions of time if in (30.52) C = 10-4, L = 1,
E(t) = 100, and at t = 0, q = 0, i = O.
(b) What is the natural (undamped) frequency of the system?
(c) What is the value of the current when t = 0.02 sec?
9. (a) Find q and i as functions of time if in (30.52) C = 10-4 , L 1,
E(t) = 100 sin 50t and at t = 0, q = 0, i = O.
(b) What is the value of the current when t = 0.02 sec?
(c) What is the maximum value of the current?
(d) What is the natural (undamped) frequency of the system?
10. If the capacitance is missing from the circuit, then by (30.13),

(30.53) L + Ri = E(t).

(a) Find i as a function of t if E(t) is a constant emf E and at t = 0, i = O.

What is the transient current, the steady state current?
(b) Find i as a function of t if E(t) = E sin wt and at t = 0, i = O. What is
the transient current, the steady state current?
ll. Find i as a function of t if in (30.53) R = 20, L = 0.1, and
(a) E(t) = 10, (b) E(t) = 100 sin 50t.
12. An inductor of L henries, a resistor of R ohms, and a capacitor of C farads
are connected in series to a battery whose emf is E volts.
(a) Find q and i as functions of time. Assume R2 < 4L/C.
(b) What is the frequency of the transient charge and current?
(c) Is there a steady state charge, a steady state current?

Hint. By (30.14), the differential equation is L :t; +

2
R + q = E.
13. (a) Find q and i as functions of time if in (30.14) and (30.15), L = 1, R 5,
C= 10-4, E(t) = 50, and at t = 0, when the switch is closed, q = 0,
i = O.
(b) What is the frequency of the transient charge and current?
(c) What is the steady state charge?
14. (a) Find the steady state current if, in (30.15), L = io, R = 5, C =
4 X 10-4 , dE/dt = 200 cos lOOt, and if at t = 0, when the switch is
closed, q = 0, i = O.
378 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

(b) What is the amplitude and frequency of the steady state current?
(c) For what value of the capacitance will the amplitude be a maximum?
(d) What should the frequency of the input E(t) be in order that it be in
resonance with the system?
(e) What is the maximum value of the amplitude for this resonant fre-
quency?
(f) What is the impedance of the system?
15. Find the steady state charge and the steady state current if, in (30.14)
and (30.15), L = to, R = 20, C = lO-4, E = 100 cos 200t. In regard
to the steady state current, answer all questions (b) to (f) of 14.
16. In (30.15), let
E(t) = EI sin wIt + E2 sin w2t + ... + E" sin w"t,
so that n different frequencies are impressed on an electric system.
(a) Show that the steady state current is

(30.54) i.

where ai, i = 1, ... , n, is defined as in (30.19). Hint. Use the super-

position principle, see Comment 24.25; also Exercise 19,6.
(b) Show that the amplitude of the steady state current due to the input
Ek sin wkt is
Ek
Ak = ---;:========:::; 2
R2 + - LWk)

We proved in the text that Ak will be largest when vl/CL is equal to

the frequency Wk. Hence by adjusting C until vl/CL = Wk, we can make
the amplitude of the response or output due to the input Ek sin wkt
larger than the amplitudes due to the other inputs. The electrical system
will thus act as a filter, responding to those inputs whose frequencies
are near vl/CL and ignoring those inputs whose frequencies are farther
away. If the inputs, for example, are coming from different radio statiops
which are broadcasting at different frequencies, you tune your radio to
one of them by turning a dial and adjusting the capacitance until the
amplitude of the output is greatest for that station's input. The ampli-
tude Ak also has Ek in the numerator. Hence for good reception from
station k, you would want its Ek to be larger, i.e., more powerful, than
the E of other stations and the frequencies of the other stations to be
not too close to Wk. Compare this problem with Exercise 2913, 13.

ANSWERS 30
4. (a) .jl/CL rad/sec. (b) No oscillations if R2 !1: 4L/C, oscillations if
R2 < 4£/C. (c) y = e-RtI2L(CI +
C21), critically damped case. NOTE.
Here y = q or i.
Lesson SO-Exercise 379

/ L
(d) q = 2qo '\J4L _ CR2 e
-RtI2L.
sm '\JCL -
(/1 R2
4L2 t + 8) ,

8 = - l;i = dq/dt.

5. (a) R 40, no oscillations; R < 40 oscillations. (b) 40 rad/sec.

6. q = !e-3 !120 sin (lOOt +
8) approximately, where 8 = Arc tan (2000/3) ap-
proximately; i = -/oe- a'/20 sin (lOOt Ii) + +
50e- 3 !120 cos (lOOt Ii). +
7. (a) q = qo cos V1/CL t; i = - sin V1/CL t, V1/CL rad/sec.
v'CL
(b) q = CE(l - cos V1/CL t), i = dq/dt.
(c) q = 1 (sin wt - w..j(fL sin V1/CL t), w ;;o!! 1/..j(fL;

i dq/dt,
=
EC.
q = - 8m
1
-- t - -E V. r;:;-;T 1
C/ L t cos - - t, w = l/VCL;
2 v'CL 2 v'CL
i = dq/dt; w = 1/..j(fL rad/sec.
8. (a) q = Iio(l - cos lOOt), i = sin lOOt. (b) 100 rad/sec.
(c) 0.909 amp.
9. (a) q = -h(sin SOt - ! sin lOOt); i = !(cos SOt - cos lOOt).
(b) i(0.02) = 0.638 amp. (c) max Iii = amp. t
(d) 100 rad/sec.
. E (1 -R'IL). E -RtIL. E
10• ()
a t = R - e ; 1t = - Ii e ; t. = Ii'

(b) .
t = R2+E L 2w2 (R smwt
. - L wcoswt + L we-RtIL) ,
ELw -R'IL
it = R2 + L2w2 e

t. = R2:L 2w2 (R sin wt - Lw cos wt).

II. (a) i = !<1 - e- 200 '). (b) i = H(4 sin SOt - cos SOt + e-200').
12. (a) q = qc +
EC, where qc is the same as ic in (30.18),
. dq
1 = dt'

(b) V4CL - R2C2/4TCL cps.

(c) q. = EC. There is no steady state current. In taking the derivative of q,
the constant EC vanishes.
13. (a) q = -1O-3 e-2.5'(0.125 sin 99.97t +
5 cos 99.97t) 0.005, +
i = dq/dt = 0.500e- 2.5 ' sin 99.97t.
(b) 99.97/21r = 15.9 cps.
(c) 0.005. For a short time the charge on the capacitor will oscillate about
this figure and approach this figure as t -> 00.
14. (a) i. = -s\(sin lOOt +
4 cos lOOt). (b) -s\VI7, 100/211' cps.
(c) C = 2 X 10-3 farad. (d) 1000 rad/sec. (e) 2/5 amp.
(f) 5VI7 ohms.
380 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

15. (a) q. = 5 X 1O-3 (sin 200t +

2 cos 200t), i, = cos 200t - 2 sin 200t.
(b) v'5, 200/211" cps. (c) 5 X 10-4 • (d) 200v'5. (e) 5 amp.
(f) 20v'5.

LESSON 30M.
MISCELLANEOUS TYPES OF PROBLEMS LEADING TO
LINEAR EQUATIONS OF THE SECOND ORDER
A. ProbleDls Involving a Centrifugal Force. When a body is
whirled in a circle at the end of string, a force, directed toward the center
of the circle, must be exerted to prevent the body from flying off; the
faster the rotation, the more powerful the force. Since this force is di-
rected toward the center of the path, it has been called the centripetal
force or central force. And since the body remains in its path, there
must be an outward force in the opposite direction equal to the central
force. This force is called the centrifugal force. It has been proved
that the centripetal force required to hold a mass m in a circular path of
radius r, moving with a linear velocity v is
2
(30.6)
r
Hence this formula must also give the centrifugal force of the mass m.
The linear velocity v of the particle is v = r dO/dt, where 0 is the central
angle measured in radians through which the particle is rotated. Substi-
tuting this value of v in (30.6), we obtain

(30.61) C.F. = rm ( r dO)2

dt

where w is the angular velocity of the particle.

With the help of (30.61), solve the following problems.
1. A smooth straight tube rotates in a vertical plane about its mid-point with
constant angular velocity w. A particle of mass m inside the tube is free to
slide without friction.
(a) Find the differential equation of motion of the particle. Hint. There
are two forces acting on the particle at time t, see Fig. 30.62.
(b) Solve the equation with t = 0, r = ro, dr/dt = vo.
(c) From the introductory remarks, it is clear that if the particle is too far
from 0 or if dr/dt is too great, the particle will fly off from an end of the
tube; for certain values of rand dr/dt, it will not. Find values of the
initial conditions ro and vo = dr/dt so that the particle will execute
simple harmonic motion. Write the resulting equation of motion for
these values. Can you identify it? Draw the figure.
2. Solve problem 1, if the tube rotates in a horizontal plane about a vertical
axis. Assume at t = 0, r = 0 and dr/dt = vo.
3. Solve problem 1, if at t = 0, r = 0, dr/dt = O.
Lesson 30M B. ROLLING BODIES 381

Figure 30.62

B. Rolling Bodies. Newton's first law of motion states that a body

at rest or moving with uniform velocity will remain in the respective
state of rest or motion unless a force acts on it. We say the body has
inertia, i.e., it resists having its status changed. Similarly, a body at
rest or rotating about an axis with a constant angular velocity will re-
main in the respective state of rest or rotation, unless a torque or a
moment of force acts on it; for dcfinition of torque, see (30.63) below.
In this case we say the body has rotational inertia, also called moment
of inertia. By definition, the moment of inertia I of a particle is

(30.621)

where m is the mass of the particle located x units from the axis of rotation.
In the calculus, you were taught how to calculate the moment of inertia of
different bodies. For example, for a solid cylinder of radius r and mass m
rotating about an axis coinciding with the axis of the cylinder, I = mr 2 /2.
It is as if the entire mass of the cylinder were concentrated at a distance
r 2 /2 units from the axis.
We define the torque or moment of force L as follows, see Fig. 30.631.

(30.63) L = x times the component of the force F acting at

right angles to the line joining the axis of rota-
tion and the point P where F is being applied;
x is the distance between the axis and P.

o x p

/ IFI
2j--compo_, of F.,
rIght angles to OP
Axis of rotation
1 to plane of paper

Figure 30.631
382 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

Finally, it has been proved that corresponding to the law F = mass X

acceleration governing the linear motion oi a body, the law governing the
rotational motion of a body is given by
d 28 dw
(30.64) L = 101 == I dt 2 == I dt '
where a is the angular acceleration of the body, w is its angular velocity,
and 8 is the central angle through which the body has rotated from 8 = O.
With the help of equations (30.621) to (30.64), solve the following
problems.
4. A cord is wound a few turns around a solid cylindrical spool of mass m and
radius r. One end of the cord is attached to the ceiling. See Fig. 30.65. At
t = 0, the spool, which is being held against the ceiling with axis horizontal,
is released.
Cord
attached
here
F P F
t = 0, y = 0, 0 = °
y = rO

y axis
mg

Figure 30.65
(a) If gravity is the only acting force, find the differential equation of motion
of the spool. Hint. By Newton's law, mass X acceleration of body must
equal the net force acting on the ·body. These forces are F and mg as
shown in Fig. 30.65. By (30.63), (30.64) and the fact that 1 = mr2/2
for a solid cylinder rotating about its axis, we have
mr2 d28
Fr = 101 ="2" dt 2 '

When the cylinder has rolled through a central angle 8 so that the point
of the spool initially at A is now at B, the distance y from the ceiling is r8.
Therefore,

Hence F = (m/2)(d 2y/dt2).

(b) Solve the differential equation for y as a function of t. Remember at
t = 0, Y = Q, dy/dt = O.
5. Answer questions (a) and (b) of problem 4 if there is a resisting force due to
friction and air of (m/BO)(dy/dt). (c) What is the limiting velocity?
Leason 30M D. BENDING OF BEAMS 383

6. At t = 0, a solid cylinder of radius r and mass m is placed at the top of an

incline and released, Fig. 30.66. Assume it rolls without slipping and that a
frictional force F acts to oppose the motion.
B

Figure 30.66

(a) Find the differential equation of motion. Hint. The only difference
between this problem and 4 is that mg is replaced by mg sin cx.
(b) Solve the differential equation. Remember at t = 0, 0 = 0, ds/dt = o.

c. Twisting Bodies. When a spring is stretched, a force results,

proportional to the amount of stretch, that tries to restore the spring to
its original natural length. Similarly, when a hanging wire is twisted by
rotating a bob about it as an axis, where the bob is rigidly attached to it
at one end, a torque or moment of force results that tries to restore the
wire to its original position. This torque L is, in many cases, proportional
to the angle (J through which the bob is turned. By (30.64), therefore,
2
d (J
(30.67) I dt 2 = -k(J,

where k is called the torsional stiffness constant. The negative sign

is necessary, because when (J is turning clockwise, the torque acts counter-
clockwise; hence torque and (J have opposite signs.
7. (a) Solve (30.67) for () as a function of time if the torque is equal in mag-
nitude to the angle (J, i.e., k = 1.
(b) If the bob returns to its equilibrium position at the end of each 1 second,
find the moment of inertia of the bob with respect to the wire as an axis.
Assume the mass of the wire is negligible.

D. Bending of Beams. We consider a beam with the following

properties.
1. It is relatively long in comparison with its width and thickness.
2. Every cross section is uniform.
3. The center of gravity of each cross section lies on a straight line, called
the axis of the beam. It is the line joining (0,0) to (L,O) in Fig. 30.7.
384 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

If a beam merely rests on supports at its two ends, it is called a simple

beam and it is said to be simply supported at the ends. If a beam is
supported only at one end, as for example, when it is embedded in masonry
at one end and hangs freely at the other end, it is called a cantilever beam.
In Fig. 30.7, we have drawn a simple beam of length L ft with rectangu-
lar cross sections whose centers of gravity lie in the geometrical center of

Figure 30.7

the rectangle. (However, beams may have other shaped cross sections, as
long as all cross sections are uniform and the center of gravity of each lies
on a straight line.) We may look on such a beam as composed of fibers
parallel to the axis of the beam, each of whose length is L ft. When a
load is distributed along a simple beam, a sag develops so that the fibers
on one side of the beam are compressed and fibers on the other side are

Figure 30.71

stretched, Fig. 30.71. It follows, therefore, that somewhere between the

two sides, a neutral surface exists that is neither stretched nor com-
pressed (shaded area in Fig. 30.71), i.e., it retains its original length L.
The intersection of this neutral surface with a vertical plane through the
of the beam is called the elastic cqrve of the beam. It is the curve
jpining (0,0) to (L,O) in Fig. 30.71.
I It has been proved in mechanics that

EI
(30.72) M(x) = If '
where:
M(x) is the bending moment at any cross section A, x units
from one end of the beam. The bending moment at A is
defined as the algebraic sum of all the moments of force
Lesson 30M D. BENDING OF BEAMS 385

acting on only one side of A about an axis through the

center of the cross section A, marked CD in the figure.
(For definition of a moment of force, see (30.63) above.)
I is the moment of inertia of the cross section A about its
center axis CD (for definition of moment of inertia, see
(30.621) above).
R is the radius of curvature of the elastic curve of the beam.
E is a proportionality constant, called Young's modulus or
modulus of elasticity. It is dependent only on the material
of which the beam is made.
The radius of curvature is given by the formula
R = [1 + (y')2J3/2jy".
Its substitution in (30.72) gives
(30.73) M(x) = Ely"[l +
(y,)2J-3/ 2
= Ely"[1 - i(y')2 + ¥(y')4 - ... J.
Since the bending is usually slight, y' is very small. Hence it is not unrea-
sonable to assume that we commit a small error if in (30.73) we neglect
(y,)2 and higher powers of y'. Equation (30.73) thus simplifies to
(30.74) M(x) = Ely",
which is the differential equation of the elastic curve of the beam.
We shall arbitrarily assume that an upward force gives a positive moment
and that a downward force gives a negative moment.
With the help of (30.74), solve the following problems.
8. A horizontal beam of length 2L ft is simply supported at its ends. The
weight of the beam is evenly distributed and equals w lb/ft.
(a) Find the equation of the elastic curve. Hint. See Fig. 30.75. The total
weight of the beam is 2Lw lb. Therefore the upward force at each end

Lw

P(x,O) (L,O)
},.:..y---------:::;..
...

f - - - - - - - = - - - 2L - x

Figure 30.75

is Lw lb. Since the beam is uniform, we can consider the weight of the
beam from (0,0) to P(x,O) as concentrated at its mid-point (x/2, 0).
Hence the downward force at this mid-point is wx lb. The bending
386 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

moment M(x) at P is therefore-remember the bending moment is the

algebraic sum of all moments of force acting on one side of the cross
section A whose axis goes through P-
X) wx2
M(x) = (Lw)x - (wx) ( 2 = Lwx - 2·

Substitute this value of M(x) in (30.74) and solve. The initial conditions
are X = 0, y = 0; x = L, y' = O. (Note there is also a third initial
condition x = 2L, Y = o. However, the three are not mutually inde-
pendent. Use of any two of the three will result in a solution which
satisfies the third condition. Verify this statement.)
(b) What is the maximum sag? Hint. The maximum sag occurs when
x = L.
(c) Show that the same bending moment at P results, if the forces to the
right of P were used. Hint. The bending moment at P due to the forces
on the right is
M(x) = Lw(2L - x) - [(2L _ x)wl (2L;- x) .
Simplify the right side.
9. A horizontal beam of length 2L ft is simply supported at its ends and carries
a weight W Ib at its center. If the weight of the beam is negligible compared
to W, find the equation of the elastic curve and the sag at the center. See
Fig. 30.76. Two cases must be considered.

Figure 30.76

Case 1. If P is to the left of the mid-point, the only active force to the left
of P is W /2. Hence the bending moment at P is

(a) M(x) = x which can be written as iWL - iW(L - x), O;:i! x < L.
Case 2. If P is to the right of the mid-point, then the active forces to the left
of Pare W /2 upward and W downward. Hence the bending moment
at P is

(b) M(x) = l; x - W(x - L) which can be written as iWL + iW(L - x),

L < x;:i! 2L.
Cases 1 and 2 can therefore be treated as one if we write
(c) M(x) = iWL =F iW(L - x),
where it is understood that the minus sign is to be used when 0 ;:i! x < L
and the plus sign when L < x ;:i! 2L. When x = L, (a), (b), and (c) are
Leuon30M D. BENDING OF BEAMS 387

the same. Hence the solution obtained by using (c) is also valid when x = L.
Initial conditions are x = 0, y = 0; x = 2L, Y = O. (A third initial
condition x = L, y' = 0 is not independent of the other two. Verify that
it satisfies the derivative of the solution.)
10. Solve problem 9, if the weight of the beam is not negligible and is w Ib/ft.
Hint. Follow all the instructions given in 8 and 9. As in 9 there will be two
cases, one when P is to left of center, the other when P is to right of center.
The bending moment at P due to the forces to the left of Pare

!.l(x) = (WL + l:) x _ w;2

=
2
wLx - !wx - !W(L - x) + !WL, 0 x < L,
if P is to left of the center, and

M(x) (WL + x - - W(x - L)

wLx - !wx
2
+ !W(L - x) + !WL, L < x 2L,
if P is to right of the center.
When x = L, both bending moments are the same. Thus both cases can
be combined if you take
M(x) = wLx - !wx 2 =F !W(L - x) !WL, +
where it is understood that the minus sign is to be used when 0 x < L
and the plus sign when L < x 2L.
II. A horizontal beam of length 30 ft is simply supported at its ends and carries
a weight of 360 lb at its center. If the weight of the beam is negligible, find
the equation of the elastic curve for each half beam. What is its sag? Solve
independently. Check your results with solutions given in problem 9.
12. A simply supported horizontal beam of length 2L carries a weight W Ib
attached to it at a distance 2L/3 from one end. Assume the weight of the
beam is negligible. Find the equation of the elastic curve. Hint. The end
of the beam closer to the weight now supports 2W/3 lb; the other end sup-
ports only W /3 lb. Two cases will be needed as in 9 and 10, one if P is to
the left of W, the other if P is to the right of W. The bending moment at P
using forces to the left of Pare
2W 2L
M(x) = 3 x, 0 x < '"3'
if P is to the left of W, and

M(x) = 2: x _ W (x _ 2:) , 2L
'"3 < x 2L,
if P is to the right of W.
Initial conditions are x = 0, y = 0; x = 2L, y = O. Note also, since the
elastic curve is continuous at x = 2L/3 and has a tangent there, that when
x = 2L/3, the value of y and the value of the derivative y' for each of the
two curves must be the same. These conditions are known respectively as
the condition of continuity of the curve and the condition of continuity
388 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

of the slope. You will need to use these facts in order to evaluate some of
the constants of integration.
13. A horizontal beam of length 2L ft and of uniform weight w Ib/ft is embedded
in concrete at both ends. Find the equation of the elastic curve and the
maximum sag. Take the origin at one end of the beam. Here in addition to
the usual moments found in problem 8, there is an additional moment of
force at each end acting to keep the beam horizontal, i.e., the masonry at
each end prevents the beam in its immediate neighborhood from sagging.
Call this unknown moment of force M. The initial conditions are x = 0,
y = O· x = 0 y' = O· x = L y' = o· x = 2L y = o· x = 2L y' = 0
There five 'sets of codditions. Use of say the first three, wili
enable you to evaluate M and the constants of integration. Verify that the
resulting equation satisfies the other two initial conditions.
14. Solve problem 13, if the beam also supports a weight W at its center. Hint.
Here, in addition to the usual moments found in problem 10, there is a mo-
ment M at each end. As in problem 10, there are two cases to be considered,
one when P is to the left of center, the other when P is to the right of center.
It will be easier to treat each case separately instead of combining them as
we did in 9 and 10. Initial conditions are x = 0, y = 0; x = 0, y' = 0;
x = L, y' = 0; x = 2L, y = 0; x = 2L, y' = O. Note that the condition
x = L, y' = 0, applies to each case, since the' curve is continuous at x = L.
There is one more condition than you need. However, it is not independent
of the others. Verify that the one you omit satisfies the solution.

(0,0) 2L-x (2L,0)

Q
(2L-x)w

Figure 30.77

15. A cantilever beam of length 2L and of uniform weight w Ib/ft is embedded in

concrete at one end. Find the equation of its elastic curve and the maximum
deflection. See Fig. 30.77. In this case, it will be easier to consider moments
of force to the right of P. Since the beam is uniform, there is a downward
force at the center of PQ. And since this is the only acting force to the
right of P, the bending moment at P is

M(x) = -w(2L -
2L - x)
x) ( - - 2 - = -
w
2' (2L - x) .
2

Substitute this value in (30.74). Initial conditions are x = 0, y = 0; x = 0,

y' = O.
16. A cantilever beam of length 2L and of negligible weight supports a load of
W Ib at its center.
(a) Find the equation of its elastic curve, the deflection at its center, and its
maximum deflection. Hint. See Fig. 30.78. There are two cases to be
Lesson 30M D. BENDING OF BEAMS 389

considered. Take moments to right of P. Initial conditions are x = 0,

y = 0; x = 0, y' = 0. When P is to the right of center, there are no
forces to the right of P. Hence the bending moment 1If(x) = °
at P.

the slope y' for the case °

You will also need to use the fact that when x = L, the solution y and
x < L must agree respectively with the
solution y and the slope y' for the case L < x 2L.

(0,0)
•

Figure 30.78

(b) Find the maximum deflection and the equation of the elastic curve if the
weight W were placed at the end of the beam. Hint. Here there is only
one case to consider, namely, p·to the left of W. The only force to the
right of P contributing to the bending moment 1If(x) at P is the weight W.
17. Solve problem 16(a) if the weight of the beam is not negligible and is w lb/ft.
Hint. There will be two cases as in 16. The bending moment at P will be
the sum of the bending moments given in 15 and 16. And remember when
x = L, the solution y and the slope y' must agree for the two cases.
18. A horizontal beam of length 2L is embedded in concrete at one end and is
simply supported at the other end with both ends at the same level. A weight
W is suspended at its mid-point and the beam itself weighs w lb/ft. Find the
equation of the elastic curve. Take the origin at the embedded end. Hint.
We need two cases, Case 1 when P is to the left of the mid-point; Case 2
when P is to the right of the mid-point. See Fig. 30.79. Take moments to
the right of P. Call F the unknown upward force at (2L,0). Initial condi-
tions are x = 0, y = 0; x = 0, y' = 0; x = 2L, Y = 0. And remember,
when x = L, the solution y and the slope y' must agree for both cases.
M=F(2L-x) M=F(2L-x)
2L-x
-2-
(x,O)

Figure 30.79

19. A spring board, fixed at one end only, may be considered as a cantilever beam.
It is desired that its maximum deflection be 1 ft when a 240-lb man steps on
the end. If the board is 20 ft long and weighs 5 lb/ft, find the value of the
constant EI. Hint. The formula for maximum deflection is the sum of the
maximum deflections given in 15 and 16(b). And remember in these formulas
L is one-half the length of the board.
390 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

ANSWERS 30M.
d2T 2
1. (a) m dt 2 = mrw - mg sin wt.

(b) T = ro
e ..1 + e-wl
2 +
( 2WV o _
2w 2
g) e..1 - 2 e-wl + 2wg.sm wt
2

= TO cosh wt + 2WV;w;- g sinh wt + sin wt.

(c) TO = 0, vo =
Resulting equation is li:J.
g. 8g . t
r = 2w2 sm w = 2w 2 sm .

In polar coordinates it is the equation of a circle with center at (g/4w2, 7r/2)

and radius g/4w 2 •
vo ( .. I
2.r=2we - .. I)
-e .

3• r = 4wg 2 (2 sm
. wt - e..I + e-WI) .
2 2
d y d y 2
4. (a) m dt 2 = mg - F, dt 2 = ig. (b) y = gt /3.

3 d2y 1 dy
5. (a) "2 dt2 + 80 dt = g.
(b) y = 9600g(e-1/120 - 1) + 80gt, v = _80ge-1/120 + 80g. (c) 80g.
2 2
d s. m d s 3 d 2s .
6. (a) m dt 2 = mgsma - F, F = "2 dt 2 ' "2 dt2 = gsma.
2
(b) s = (g sin a}t /3.
7. (a) 8 cr cos (Vlli t
= o. (b) I = 1/(647r ). + 2

wx 2 3 3
8. (a) Ely = 24 (4Lx - x - 8L ), 0 x 2L.
4
(b) sag = 5wL /24E I ft.
lV 2 3 2
9. Ely = 12 [3Lx =F (L - x) - 6L x + L],
3
0 x 2L;
Wx 2 2
Ely = 12 (x - 3L), 0 x L;
lVx W
Ely = 12 2
(x - 3L )
2
+ '6 (L - x),
3
L x 2L;
3
sag = lVL /6EI ft.
10. Ely = sum of the results obtained in 8 and 9; sag = sum of the results
obtained in 8 and 9.
U. Ely = 30x(x 2 - 675), 0 x 15,
= 30x(x 2 - 675} + 60(15 - x)3, 15 x 30;
sag = 202,500/ EI ft.
Lenon 30M-An8wers 391

Wx 2 2
12. Ely = 81 (9x - 20L), 0 x 2L/3;

Ely = (9x 2 _ 20L 2) _ (x _ 2t x 2L.

2
13. Ely" = M + Lwx - 2' M = -L 2 w/3;
2
Ely = (4Lx - x 2 - 4L2) = _ x 2(2L _ X)2;

sag = wL4/24EI ft.

14. Moment M = -iwL -

2
4WL ;

Ely" = (WL + x - w;2 + M, 0 x Li

W
4L x ) + 24 (2x -
W 3 4. 22 3 2
Ely = 24 (4Lx - x - 3Lx), 0 x Li

Ely" = (WL + x - w;2 - W(x - L) + M, L x 2L,

Ely = -
W 3
(4Lx - x - 4L x )
4 22
+ -W [2x 3 - 4(x -
3
L) - 3Lx ],
2

24 24 L x 2L,
sag = (wL4. +
WL 3)/24EI.
w 4. 3 4.
15. Ely = 24 [16L - 32L x - (2L - x) ]
W 3 2 2 4
= 24 (8Lx - 24L x - x ).
Maximum deflection = 2wL4./EI.

16. (a) Ely =

W [L 3
6" 2
- 3L x - (L - x) ]
3

W 3 2
= 6" (x - 3Lx), 0 x Li

W 3 2
Ely = 6" (L - 3L x), L x 2L.
Deflection at mid-point WL3/3 Eli maximum deflection 5WL3/6EI.
3 2 3
(b) Ely = (x - 6Lx ), maximum deflection 8WL /3EI.
W 3
W 3
17. Ely = 24 (8Lx - 24L x - x )
22 4.
+ 6" (x - 3Lx),
2
0 x Li

W 3
Ely = 24 (8Lx - 24L x - x ) + 6"
W 3 22
2
(L - 3L x), L x
4.
2L.
Deflection at midpoint: (17w£4 + 8WL3)/24EI,' deflection at end point:
(12wL4 + 5WL3)/6EI.
392 PROBLEMS LEADING TO LINEAR EQUATIONS OF ORDER Two Chapter 6

5W
18. F = £wL+ 16·
Ely" = -W(L - x) -
w
"2 (2L - x) + F(2L -
2
X)i
w W
2x ) + 96 (llx
3 22 4 3 2
Ely = 48 (10Lx - 12L x - - 18Lx ),
o x L;
w
Ely" = -"2 (2L - x)
2
+ F(2L - x),

Ely = .:!!!.. (lOLx 3 - 12L 2 x 2 - 2X4)

48
W
+ 3
96 (16L - 48L x
2
+ 30Lx 2
-
3
5x), L x 2L.
19. 740,000.
 Chapter 7

Systems of Differential Equations.

Linearization of First Order Systems

LESSON 31. Solution of a System of Differential Equations.

LESSON 31A. Meaning of a Solution of a System of Differential

Equations. In algebra it is frequently necessary to solve a system of
simultaneous equations of the type

(a) 2x + 3y = 5, (b) x 2 - 3y = 2, (c) x + 3y - 2z = 15,

x - y = 15; x + y2 = 5; x - y +z = 7,
3x + 2y - Z = 12.
Similarly, it is frequently necessary to solve a system of differential equa-
tions of the type
dx
(d) x dt + y dy
2
dt + 2x - 3y = 3t,
2
d x dy dy t
dt 2 + 2y dt + 3 dt - 3x = e,

where x and yare dependent variables and t is an independent variable.

In Lessons 33 and 34, we discuss and solve numerous physical problems
which give rise to such systems.
A solution of an algebraic system of two equations is a pair of values of x
and y such that this pair satisfies both equations. For example, x = 10,
y = -5 is a solution of (a), since this pair of values satisfies both equa-
tions. Analogously we say that the pair of functions x(t), yet), each de-
fined on a common interval I, is a solution of the system (d), if this pair
satisfies both equations identically on I, i.e., if in each equation of (d) an
identity results when x is replaced by x(t), y by yet), and their respective
derivatives by x'(t), y'(t), etc.
The extension of the meaning of a solution of a system of three or more
equations should be apparent.
393
394 SYSTEMS. LINEARIZATION OF FIRST ORDER SYSTEMS Chapter 7

LESSON 3IB. Definition and Solution of a SystelD of First Order

Equations.
Definition 31.1. The pair of equations
dx dy
(31.11) dt = ft(x,y,t), dt = !2(x,y,t),

where ft and!2 are functions of x, y, t, defined on a common set S, is called

a systelD of two first order equations. A solution of (31.11) will
then be a pair of functions x(t), y(t), each defined on a common interval I
contained in S, satisfying both equations of (31.11) identically.
A generalization of this type of system is given in the following definition.
Definition 31.12. The system of n equations

dYl
(31.13) de = ft(YbY2, ... , y,.,t),

dY2
de = !2(Yl,Y2, ... , y .. ,t),

where ft, ... , In are each functions of Yb Y2, ... , y .. , t, defined on a com-
mon set S, is called a systelD of n first order equations.
Definition 31.14. A solution of the system (31.13) is a set of func-
tions Yl(t), Y2(t), ... , Yn(t), each defined on a common interval I con-
tained in S, satisfying all equations of (31.13) identically.
Comment 31.141. In Lesson 62, you will find a criterion, Theorem
62.12, which gives a sufficient condition for the existence and uniquene88
of a solution of the system (31.13) satisfying the n initial conditions,
(31.15)

Comment 31.16. In-Lesson 62B, we show how a nonlinear differen-

tial equation of order greater than one can be reduced to a system of first
order equations. If, therefore, we can devise methods for finding solutions
of the system (31.13), then theoretically any nonlinear differential equation
can be solved. In Lesson 34, in fact, we solve certain special types of second
order nonlinear differential equations by reducing them to a system of
two first order equations.
Comment 31.17. Solutions of systems of first order equations will
not in general be expressible explicitly or implicitly in terms of elementary
functions. Only a few very special first order systems will have such solu-
Lesson3lB SOLUTION OF A FIRST ORDER SYSTEM 395

tions. Even this simple looking pair of first order equations,

dy
dt = x

cannot be solved in terms of elementary functions. In later lessons, we

shall show you how a pair of first order equations may sometimes be
solved by means of series methods, numerical methods, and by a method
known as Picard's methorl of successive approximations. We wish to im-
press on you that the exarr.ples of first order systems which we have solved
below are of a very special kind, artificially designed to enable us to ob-
tain solutions in terms of elementary functions.
Example 31.18. Solve the first order system
dx t dy y
(a)
dt = x2 '
dt = t2' x ¢ 0, t ¢ o.
(Note that each equation is of the separable type.)
Solution. From the first equation, we obtain

(b)

and from the second, provided y ¢ 0,

(c) log y = - f+ C2',

The pair of functions defined by (b) and (c) is a solution of the system (a).
Example 31.19. Solve the first order system

dx 2t dy x2 - y
(a) dt=2e, dt = --t-' t ¢ O.

(Note that the first equation has no x or y in it.)

Solution. Solving the first equation in (a), we obtain

(b)

Substituting (b) in the second equation of (a), there results

(c)
dy y e4t + 2c}e 21 + C}2
dt +7 = t

which is a first order linear equation in y. Its solution, by any method

396 SYSTEMS. LINEARIZATION OF FIRST ORDER SYSTEMS Chapter 7

you wish to choose, is

(d) yt = f + 2cle2t + C12) dt = le 4t + Cle2t + C1 2t + C2,

(e'lt

y = (le 4t + Cle2t + C12t + C2)t-I, t ¢ o.

Your can verify that the pair of functions defined by (b) and (d) satisfies
both equations of (a) and is, therefore, a solution of the system.

LESSON 3Ie. Definition and Solution of a System of Linear

First Order Equations. A special type of first order system is one in
which the functionsiI(x,y,t) andh(x,y,t) of (31.11) are linear in x and y.
This means that each equation of the system has the form
dx
(31.2) dt = 11 (t)x + gl (t)y + hI (t),
dy
dt = h(t)x + g2(t)y + h 2(t).
A pair of equations of this type is called a system of two linear first
order equations.. Note that x and y both have the exponent one, but
that no such restriction is placed on the independent variable t. A general-
ization of this type of system is given in the following definition.
Definition 31.21. The system of n equations

(31.22) dIeI = 111 (t)Yl + iI2(t)Y2 + ... + iIn(t)Yn + Ql (t),

dIe2 = hl(t)Yl + Idt)Y2 + ... + hn(t)Yn + Q2(t),

is called a system of n linear first order equations.

Comment 31.23. In Lesson 62C you will find a criterion, Theorem
62.3, which gives a sufficient condition for the existence and uniqueness of a
solution of the system (31.22) satisfying the initial conditions,
(31.24)
Comment 31.24. No standard method is known of finding a solution
in terms of elementary functions, if one exists, of a general linear first order
system (31.22). If, however, all coefficients /ij(t) , i = 1, ... , n; j =
1, ... , n, are constants, then standard methods of solution are available.
These methods are discussed in Lesson 31D, where the system (31.22)
with constant coefficients, is included in the larger class of systems of
LeSSOD 3Ie SOLUTION OF A LINEAR FIRST ORDER SYSTEM 397

linear equations with constant coefficients of order greater than or equal

to one. In the examples solved below, where the coefficients are not con-
stants, we again impress upon you the fact that they have been artificially
selected to yield elementary functions for solutions.
Example 31.25. Solve the linear first order system
dx dy
(a) -=2xt-x
dt ' dt = 2yt + x.
(Note that the first equation has no y in it.)
Solution. The first equation in (a) is the separable type discussed in
Lesson 6C. By the method outlined there, we obtain the general solution
(b)
Substituting (b) in the second equation of (a), there results
dy
(c) dt = 2ty + Cle 12_1
,

which is a first order equation, linear in y. Its solution, by the method of

Lesson lIB, is
(d) ye _12 = Cl f e -Idt = -cle- I + C2,
l2 I
y = e (C2 - Cle- ).

You can verify that the pair of functions defined by (b) and (d) satisfies
both equations of (a) and is therefore a solution of the system (a).
Example 31.26. Solve the linear first order system

dx _ 221 dy x - y
(a) dt - e , dt = -t-'

Solution. The solution of the first equation is

(b) x = e
21
+ Cl'
Substituting (b) in the second equation of (a), there results
dy e
21
- y + CI dy y e 21 + CI
(c) dt = t dt +T= t .

The solution of (c), by the method of Lesson lIB, is

(d) yt = f (e
21
+ Cl) dt, y = (!e 21 + CIt + C2)t- l
•

The pair of functions defined by (b) and (d) is a solution of thesystem(a).

398 SYSTEMS. LINEARIZATION OF FIRST ORDER SYSTEMS Chapter 7

LESSON 3ID. Solution of a SysteDl of Linear Equations with

Constant Coefficients by the Use of Operators. Nondegenerate
Case. The pair of equations

(31.3) fl(D)x + gl(D)y = h1(t),

f2(D)x+ g2(D)y = h2(t),

where D is the operator d/dt and the coefficients of x and yare polynomial
operators as defined in Lesson 24A, is called a systeDl of two linear
differential equations. A generalization of this type of system is given
in the following definition.
Definition 31.31. The system of equations

(31.32) Pll(D)Yl + PdD)Y2 + ... + P1n(D)Yn = ht(t),

P 21 (D)Yt + P 22 (D)Y2 + ... + P 2n (D)Yn = h2(t),

Pnt(D)Yt + P n2 (D)Y2 + ... + Pnn(D)Yn = hn(t),

where D is the operator d/dt and the coefficients of y., Y2, ... , Yn are poly-
nomial operators, is called a systelll of n linear differential equations.
Definition 31.321. A solution of the linear system (31.32) is a set of
functions Yt (t), Y2(t), ... , Yn(t), each defined on a common interval I,
satisfying all equations of the system (31.32) identically; the solution is a
general addition, the set of functions Yt (t), .•• , Yn(t) contains the
correct number of arbitrary constants; see Theorem 31.33 below.
Examples of systems of linear equations are

(a) (2D + 3)x + (5D - l)y = e',

(D - l)x + (3D + l)y = sin t;
(b) (D + 3D - l)x + Y = 6 + t 2,
2
(D + 2)x - (D 2 + D)y = t;
(c) (3D2 + l)x + D3y - (D + l)z = t 2 + 2,
(D - l)x + (D2 + l)y + (D 2 - 2)z = e' ,
Dx - Dy - (D3 + 5)z = 2t + 5.

Systems of linear equations with constant coefficients lend themselves

readily to solutions by means of operators and, if t 0, by Laplace trans-
forms. Although we shall confine our attention primarily to a system of
two linear equations and touch briefly on a system of three linear equations,
the extension of the method of solution to a larger system of linear differen-
Lesson 3ID SOLUTION OF A LINEAR SYSTEM BY OPERATORS 399

tial equations with constant coefficients will be made apparent. In this

lesson we shall solve a linear system by the use of operators; in Lesson 3lH
by means of the Laplace transform.
Since polynomial operators with constant coefficients obey all the rules
of algebra summarized in 24.523, the method we shall be able to use for
solving a system (31.32) will be similar to that used in solving an algebraic
system of simultaneous equations. There are, however, two important
differences between the two systems.
1. The operator symbol D does not represent a numerical quantity such
as 2, 0, etc. It represents a differential operator, operating on a
function. Hence the order in which operators are written is important.
2. Solutions of algebraic systems do not usually have arbitrary constants;
general solutions of systems of linear differential equations usually do.
By Definition 31.321, a general solution, in addition to satisfying the
system, must contain the correct number of such constants. The
relevant theorem needed in this connection for the two equation
system (31.3) is the following.
Theorem 31.33. The number of arbitrary constants in the general solu-
tion x(t), yet) of the linear system (31.3) is equal to the order of
(31.34)
provided It (D)(l2(D) - (II (D)h(D) O.
The proof of the theorem has been deferred to Lesson 31E.
Comment 31.341. If the difference in (31.34) is zero, the system is
called degenerate. We shall discuss the degenerate case in Lesson 31F.
Comment 31.35. The quantity alb2 - a2bl, where aI, a2, bl , b2
are constants, appears so frequently in mathematical literature that it
has been given a special name. It is called a determinant. * A de-
terminant is also usually written as

hence, Ibal
l
b
b2
2
1 == a l b2 - a2bl.

Since (31.34) has the same form as a determinant, we shall also refer to it
as a determinant, and write
(31.36)

*The solution of the pair of equations a,x +

b,y = c" a.x +
b.y _ c. is x =
(c,b. - c.bt>/(a,b. - a.bt>, y = (a,c. - a2c,)/(a,b. - aob,). Note that the
quantity (a,b o - aob,) appears in the denominator of both equations and thus de-
termines whether a solution exists; see Lesson 63A. Hence the name determinant.
400 SYSTEMS. LINEARIZATION OF FIRST ORDER SYSTEMS Chapter 7

Observe that the members of the left side of (31.36) are the coefficients
of (31.3) written in the same relative position in which they appear there
and that each term on the right side is a cross product of these coefficients
in a definite order with a minus sign between them. We shall therefore
refer to (31.36) as the deterlninant of the systeln (31.3). Using this
terminology, we can say, by Comment 31.341, that the system (31.3) is
degenerate if its determinant (31.36) is zero.
For the remainder of this Lesson 31D, we consider only nondegenerate
systems.
The usual procedure followed to solve the algebraic system
(a) 2x +
3y = 7,
3x - 2y = 4,
is to multiply the first by 3, the second by -2, and then add the two. In
this way we eliminate x, thus obtaining
13y = 13, y = 1.
To find x, we can again start with (a) and eliminate y, or we can substitute
y = 1 in either of the two equations. By either method, we find x = 2.
This pair of values x = 2, Y = 1 satisfies both equations and is therefore
a solution of (a).
We follow an identical procedure in solving the system (31.3). Multiply-
ing the first by h(D), the second by -!teD) and adding the two resulting
equations will eliminate x and yield a linear differential equation in y
which can be solved by previous methods. Substituting this value of y
in either of the two given equations will enable us to find x(t). [Or we can
eliminate y in the given equations and solve for x(t).J The
feature of this standard method is that in multiplying each equation by a
polynomial operator, we usually raise the order of the given equations and
thus introduce superfluous constants in the pair of functions x(t), yet).
It then becomes necessary, if the pair is to be a general solution of (31.3),
to determine how these constants are related: usually a tedious task. We
shall show by examples how to eliminate such superfluous constants and
thus obtain the general solution of (31.3). Later we shall describe a method
which will immediately give the correct number of constants in the pair
of functions x(t), yet), and hence will immediately yield a general solution
of (31.3)-see Lesson 3lE.
Example 31.361. Solve the system
ax
(a) 2 dt - x + dy
dt + 4y = 1,
ax dy
dt-dt=t-1.
Lesson 31D SOLUTION OF A LINEAR SYSTEM BY OPERATORS 401

Solution. In operator notation, with D = d/dt, we can write (a) as

(b) (2D - l)x (D + + 4)y = 1,
Dx- Dy=t-1.
Following the procedure outlined above, we multiply the first equation
by D, the second by (D +
4) and add the two. There results
(c) (3D2 + 3D)x = D(I) + (D + 4)(t - 1) = 4t - 3.
The general solution of (c), by any of the previous methods discussed, is
(d)
Substituting (d) in the second equation of (b), we obtain
(e)
which simplifies to
(f) Dy = -C2e- I + -t - -4.
3 3
Integration of (f) gives its general solution,
2
(g) yet) = C2e-' + - it + Ca.
By Comment 31.35, the determinant of (b) is (2D - 1)(-D) -
D(D + 4) = -3D2 - 3D which is of order two. Hence, by Theorem
31.33, the general solution of (b) must contain only two arbitrary con-
stants. But the pair of functions x(t), yet) as given in (d) and (g) respec-
tively has three. To find a relationship among the three constants, we use
the fact, see Definition 31.321, that the solution of a system of equations
is a set of functions which satisfies each equation of the system identically.
Since yet) was obtained by substituting x(t) in the second equation of (b),
we know that the pair of functions x(t), yet) will satisfy this equation
identically. Hence we substitute (d) and (g) in the first equation. Making
these substitutions, there results
(h) (2D - 1)(c} +c2e-l+it2 -it) + (D+4) (C2e-'+i - it+ca) = 1.

Performing the indicated operations in (h), we obtain

(i) -6 - c} + 4ca = 1, 4ca = c} + 7, Ca = c} t 7 .
Hence if Ca = (c} + 7)/4, (h) will be an identity in t. Substituting this
value in (g), we obtain the corrected function

+ -+
2
(j) y ( t) = C2e
_I
+ -t6 - 4
-t
3
- 7.
Cl
4
402 SySTEMS. LINEAlUZATION OF FmsT OJmER SYSTEMS Chapter '1

The pair of functions x(t), yet) defined by (d) and (j) now contains the
correct number of two arbitrary constants. You can verify that this pair
of functions satisfies both equations of (a) and is, therefore, by Definition
31.321, its general solution.
Esample 31.37. Solve the system

(a)
dx
dt - x + dy
dt + y = 0,
dx dy
2 +2x+2 -2y=t.
dt dt
Solution. In operator notation, we can write (a) as
(b) (D - I)x + (D + I)y = 0,
(2D 2)x + + (2D - 2)y = t.
Multiplying the first equation in (b) by (2D + 2), the second by
-(D - 1) and adding the two, there results
(c) [(2D + 2)(D + 1) - (D - I)(2D - 2)]y = -(D -I)t,
SDy = t- 1,
t 1 t2 t
Dy=S-S' yet) = 16 - S + Cl·
Substituting this value of yet) in the first equation of (b), we obtain
t2 t ]
(d) (D - I)x + (D + 1) [ 16 - S + Cl = 0,

which simplifies to the first order linear equation

t2
(e) (D - I)x = - 16 + S1 - Cl·

The general solution of (e) is

t2
(f) x(t) = 16 + St + Cl + C2e'.
The determinant of (b), by Comment 31.35, is (D - I)(2D - 2) -
(D +I)(2D +
2) = -SD which is of order one. Hence by Theorem
31.33, the pair of functions x(t), yet) as given in (f) and (c) respectively
should have only one arbitrary constant. But the pair has two. To find
the relationship between the constants we proceed as we did in the pre-
vious example. Substituting (f) and (c) in the second equation of (b) (we
have already used the first), there results

(g) 2(D + 1) + + Cl + C2e'] + 2(D - 1) - + Cl] = t,

Le8SOD 3ID SOLUTION OF A LINEAR SYSTEM BY OPERATORS 403

which simplifies to
(h)

Equation (h) will be an identity in t only if C2 = o. Substituting this

value in (f) we obtain the corrected function
t2 t
(i) x(t) = 16 + 8 + Ct·
The pair of functions x(t), y(t) defined in (i) and (c) respectively is, by
Definition 31.321, the general solution of (a).
Example 31.38. Solve the system
(a) (D + 3)x + (D + l)y = el ,
(D + l)x + (D - l)y = t.
Solution. Multiplying the first by -(D - 1), the second by (D + 1)
and adding the two, we obtain
(b) [_(D2 + 2D - 3)+ (D2 + 2D + 1)]x = 1 + t,
4x = 1 + t, x(t) = !<t + 1).
Substituting this value of x(t) in the first equation of \a), there results

(D + l)y = e' - 1 - tt.

The general solution of (c) by any method you wish to choose is

,
(d) y(t) = cle- ' + - ! - tt.

The determinant of (a) is (D +

3)(D - 1) - (D 1)2 = -4 which+
is of order zero. Hence, by Theorem 31.33, the pair of functions x(t), y(t)
should have no arbitrary constant. But the pair as given in (b) and (d)
has one. We know the pair satisfies the first equation of (a) identically,
since we used it to find y(t). We therefore substitute x(t) and y(t) in the
second equation of (a). There results

(e) (D + 1) + + (D - 1) (c1e- 1 + - - tt) = t.

!4+4!.. +4! - cle-' + 2-4 - Cte-t - 2+4! + tt = t'
which simplifies to
(f)
404 SYSTEMS. LINEARIZATION OF FIRST ORDER SYSTEMS Chapter 7

Hence (f) will be an identity in t, only if Cl = o. Substituting Cl = 0

in (d) gives the corrected function
et 1 3
(g) y(t) = "2 - "4 - "4 t.
By Definition 31.321, the pair of functions x(t), y(t) defined by (b) and (g)
is the general solution of (a).
Example 31.39. Solve the system
d 2x dy
(a) dt 2 - 4x + dt = 0,
2
dx d y
-4 dt + dt 2 + 2y = O.

Solution. In operator notation we can write (a) as

(b) (D 2 - 4)x + Dy = 0,
(-4D)x + (D2 + 2)y = O.

Multiply the first equation by (4D), the second by (D 2 - 4) and add the
two. There results

(c) [4D 2 + (D 2 + 2)(D2 - 4»)y = 0, (Dol. + 2D2 - 8)y = O.

The general solution of (c) is

(d) y(t) = cle{2t + cze-{2t + Ca cos 2t + Col. sin 2t.

Substituting (d) in the second equation of (b) gives

(e) -4Dx + + - 4ca cos 2t - 4C4 sin 2t

+ + 2c2e--v'2t + 2ca cos 2t + 2C4 sin 2t = 0,

Dx = cle-v'2 t + C2e--v'2t - tca cos 2t - tC4 sin 2t,

X (t) = 2V2 Cle -v'2t -

V2 cze
2 - .l.'
sm 2t + .l.
cos + C5'
2t

The determinant of (b), by Comment 31.35, is (D 2 - - 4)(D 2 + 2)

D( -4D) which is of order four. Hence by Theorem 31.33, the pair of
functions x(t), y(t) should have four arbitrary constants. But the pair 8'S
given in (d) and (e) has five. You can verify that the substitution of x(t),
y(t) in the first equation of (b)-we have already used the second-will
be an identity if C5 = O. Hence this pair of functions, with C5 = 0, is the
general solution of (a).
Lesson 3lE AN EQUIVALENT TRIANGULAR SYSTEM 405

LESSON 3IE. An Equivalent Triangular System. The standard

method outlined above for solving the system (31.3) can be tedious and time
consuming, if by its use the pair of functions thus obtained has more than
the required number of constants. A method therefore which would im-
mediately give the correct number of constants in the pair of functions
should indeed be welcomed. We shall now develop such a method.
For convenience we recopy the system (31.3),
(31.4) fl(D)x + gl(D)y = hl(t),
I2(D)x + g2(D)y = h2(t).

From it, we obtain a new system in the following manner. We retain either
one of the equations in (31.4). Let us say we retain the first. The second is
then changed as follows. Multiply the one retained, here the first, by any
arbitrary operator k(D) and add it to the second. The new system thus
becomes
(31.41) Jx(D)x+ gl(D)y = hl(t),
[Jx(D)k(D) + I2(D)]x + [gl(D)k(D) + g2(D)]y = k(D)hl(t) + h 2(t).
The new system (31.41) is equivalent to the first system (31.4) in the sense
that a pair of functions x(t), yet) which satisfies the first system will also
satisfy the second system, and conversely, a solution of the system (31.41)
will satisfy the system (31.4).
By Comment 31.35, the determinant of (31.4) is

(31.42)

and the determinant of (31.41) is

(31.43)

Hence we see that the determinants of both systems are the same. There-
fore the order of the determinant of our original system (31.4) must be
the same as the order of the determinant of the equivalent system (31.41).
Let us assume momentarily-we shall show you later how to do it-
that by always retaining one equation in a 8ystem and changing the other
in the manner described above, we can obtain a new system, equivalent to
(31.4), in the form
(31.44) F1(D)x = H1(t),
F 2(D)x + G2(D)y = H 2(t).

Hence the determinant of the system (31.44) is the same as the deter-
minant of the system (31.4) and the orders of their determinants are also
the same.
406 SYSTEMS. LINEARIZATION OF FIRST ORDER SYSTEMS Chapter 7

We call such an equivalent system, i.e., one in which a coefficient of x or

of y is zero, an equivalent triangular system. But now note. If we
solve the system (31.44) for x, using the first equation in (31.44), its gen-
eral solution x(t) will have as many arbitrary constants as the order of
Fl(D). Substituting this solution in the second equation of (31.44) will
result in a linear equation in y whose order is equal to that of G2 (D).
Hence, in solving this equation for y, w.e shall introduce additional con-
stants equal to the order of G2 (D). Therefore the total number of arbi-
trary constants in the pair of functions x(t), yet) of (31.44), if solved first
for x and then for y, will equal the sum of the orders of Fl(D) and G2 (D).
But, by Comment 31.35, the determinant of (31.44) is F l (D)G 2 (D) whose
order is exactly the sum of the orders of Fl and G2 • (Remember in multi-
plying differential operators, the orders are added just as if they were
exponents, for example D2D3 = D 5 .) We have thus shown that for the
system (31.44) the pair of functions x(t), yet), if the system is solved first
for x(t) and then for yet), will contain the correct number of constants
(since the pair satisfies each equation identically), and that this number
equals the order of the determinant of (31.44). But the order of the deter-
minant of (31.44) is the same as the order of the determinant of the original
system (31.4) from which it came. We have therefore not only proved
that a solution of (31.44) will also be a solution of (31.4) and contain the
correct number of arbitrary constants, but have at the same time proved
Theorem 31.33.
If the original system is already in the special form
(31.45) (a) fl(D)x = hl(t),
g2(D)y = h 2(t),
or in the triangular form
(31.45) (b) ft(D)x = hl(t),
f2(D)x + g2(D)y = h 2(t),

then the pair of functions x(t), yet) obtained by solving the system will
contain the correct number of constants. In solving (b), it is essential to
obtain x(t) first and then yet). If you solve first for y by eliminating x,
superfluous constants may be introduced.
Example 31.46. Solve the system
(a) (D + l)x = t,
D2y = e2t •
Solution. The general solution of the first equation in (a), by the
method of Lesson HB, is
(b) x(t) = t - 1 + cle-t .
Le880D 3lE AN EQUIVALENT TRIANGULAR SYSTEM 407

The general solution of the second equation in (a) is

(c) 2t
y(t) = ie + C2t + C3'
You can verify, by Theorem 31.33, that the pair of functions in (b) and
(c) has the correct number of three constants and is therefore the general
solution of (a).
Example 31.47. Solve the system
(a) +
(3D 2 3D)x = 4t - 3,
(D - l)x - D2y = t 2.

Solution. The general solution of the first equation in (a), by any

of the previous methods discussed, is
(b)
Hence
(c) (D - l)x = -C2e-t + it - i - Cl - C2e-t - it 2 + it
= -i - Cl - 2c2e-t + ¥t - it2.
Substituting (c) in the second equation of (a), there results
(d) D2y = -i - Cl - 2c2e-t ¥t - jt 2. +
Integrating (d) twice, we obtain the general solution

(e) y (t) = C4 + cat - 7

(6 + 2' t + at3
Cl) 2 .l..J..
- -.L 4
nt - 2C2e -I.

The pair of functions defined in (b) and (e) is the general solution of (a).
You can verify, by Theorem 31.33, that this pair of functions contains the
correct number of four arbitrary constants.
Before proceeding to the general system, we consider one more special
type, namely one in which a coefficient of x or y is a constant. Hence the
system is of the form
(31.48) ft(D)x + ky = h1(t),
h(D)x + g2(D)y = h2(t).
In this case it will be possible to obtain an equivalent triangular system in
one step by the following procedure. Retain the equation which contains the
constant coefficient-in (31.48) it is the first-and change the second by
multiplying the first by -g2(D)/k and adding it to the second.
Example 31.49. Solve the system
(a) (3D - l)x +
4y = t,
Dx-Dy=t-1.
408 SYSTEMS. LINEARIZATION OF FIRST ORDER SYSTEMS Chapter 7

Solution. Following the procedure outlined above we copy the first

equation, and obtain a second equation by multiplying the first by Dj4
and adding it to the second. We thus obtain the equivalent triangular
system
(b) (3D - l)x + 4y = t,
2 D
i(3D + 3D)x = "4 (t) + t - 1 = t - t.
The general solution of the second equation in (b), by any of the methods
previously discussed, is
(c)
Substituting this value of x in the first equation in (b), we obtain
(d)
which simplifies to
(e) y(t) = Cl t 7 + C2e-' + - ;t.

The determinant of (a), by (31.36), is

(3D - 1)(-D) - 4D = -3D2 - 3D,
whose order is two. Note that the pair of functions x(t), y(t) contains the
correct number of two constants and is therefore the general solution of (a).
31.5. Solve the system
(a) (D + 4)x + Dy = 1,
(D - 2)x +
y = t2 •
Solution. We copy the second equation and obtain a new first equation
by multiplying the second by -D and adding it to the first. We thus
obtain the equivalent triangular system
(b) (- D2 + 3D + 4)x = -2t + 1,
(D - 2)x +y = t2 •
The general solution of the first equation in (b) by any of the methods
previously discussed is
t 5
t = Cle + C2 e - 2 + 8·
41 -I
(c) x ()

Substituting (c) in the second equation of (b) gives

(d) (D - 2) ( cle + C2e- I - 2t + 85) + y = t 2 ,

41

y(t) = -2cle'" + 3c2e-I + t 2 - t + i.

LeSSOD 3IE AN EQUIVALENT TRIANGULAR SYSTEM 409

The detenninant of (a) is (D +

4) - D(D - 2) = - D2 3D 4, + +
which is of order two. Note that the pair of functions x(t), y(t) contains
the correct number of two constants, and is therefore the general solution
of (a).
Example 31.51. Solve the system

(a) (D + l)x + y = e' ,

(D 2 + l)x + (D - l)y = t.
Solution. We copy the first equation, and change the second by multi-
plying the first by -(D - 1) and adding it to the second equation. We
thus obtain the equivalent system
(b) (D + l)x + y = el ,

2x = - (D - l)e' +t= t, x(t) =

Substituting in the first equation of (b), the value of x as found in the

second equation, there results
,It
(c) y(t) = e - - --.
2 2
The determinant of (a), by (31.36), is (D l)(D - 1) - (D 2 + 1) = +
D2 - 1 - D2 - 1 = -2, which is of order zero. Note that the pair
of functions x(t), y(t) contains this correct number of zero constants, and
is therefore the general solution of (a).

We shall now show you a method by which you can reduce a general
system to an equivalent triangular one. We demonstrate the method by
an example. Consider the system

(31.52) (D 3 - D2 + 2D + l)x + (D 4 + 3D2 + l)y = 0,

(D - 2)x + (D3 - 3)y = O.

Of the four polynomial operators in (31.52), concentrate on the one of

lowest order. In the above example it is D - 2. Retain the equation in
which it appears, multiply the equation by - D2 and add it to the first.
The resulting equivalent system is

(D2 + 2D + 1)x + (-D 5 + D4 + 6D 2 + l)y = 0,

(D - 2)x + (D3 - 3)y = O.

Note that by this process, we were able to reduce the order of the co-
efficient of x in the first equation from three to two. The polynomial
operator of lowest order in this new system is again (D - 2). We there-
410 SYSTEMS. LINEARIZATION OF FIRST ORDER SYSTEMS Chapter 7

fore again retain the equation in which it appears, multiply it by - D

and add it to the first. We thus obtain the equivalent system

(4D +
l)x + (-D5 + 6D2+ 3D + l)y = 0,
(D - 2)x + (D3 - 3)y = O.

Note that the order of the coefficient of x in the first equation has been
reduced from two to one. There are now two polynomial operators with
the same lowest order. We can retain either one. Because the coefficient
of D in (D - 2) is one, it will be found easier to retain this equation.
Multiply it by -4 and add it to the first equation. We thus obtain the
equivalent system

9x + (- D5 - 4D3 + 6D 2 + 3D + 13)y = 0,
(D - 2)x + (D 3 - 3)y = O.

The system is now of the form (31.48). Therefore, retaining the first equa-
tion, multiplying it by -(D - 2)/9 and adding it to the second will
finally give the equivalent triangular system

9x + (- D5 - 4D 3+ 6D 2 + 3D + 13)y = 0,
[ (D 5 + 4D 3 - 6D 2 - 3D - 13) - 9 - + D 3 - 3] y =
D-2 O.

In the manner described above, it is always possible to reduce the order

of the coefficient of one polynomial operator in the system to zero. When
that point has been reached, the system will be in the form (31.48). One
more step will then give the required equivalent triangular system.
Comment 31.53. Use of the above method will always enable you to
obtain an equivalent triangular system. Use of a little ingenuity may at
times enable you to obtain it sooner. For example, if the given system is,
or if in the course of your work, it becomes,

D3 X + (D 2 - 2)y = e l
,

-2D 3x + (D + 3)y = 4t + 2,
then retaining the first, multiplying it by two and adding it to the second
will give you an equivalent triangular system immediately, even though
D3 is not the polynomial operator of lowest order.
Example 31.531. Solve the system

(a) (D + l)x + (D + l)y =1,

Dy=t-1.
Lesson 3lE AN EQUIVALEN1 TRIANGULAR SYSTEM 411

Solution. Here there are three polynomial operators of the same

lowest order. It will be found easiest to retain the 8econd and change the
first by adding the two equations. There results the equivalent system

(b) (D2 + D + I)x + y = t,

D 2x - Dy = t - 1.

The system is now of the form (31.48). We therefore copy the fir8t equa-
tion and change the second by multiplying the first by D and adding it to
the second. We thus obtain the equivalent triangular system

(c) (D 2 + + I)x + y
D = t,
(D 3+ + D)x
2D2 = t.

A general solution of the second equation is

(d) x(t) = Cl + C2e-t + C3te-t + - 2t.

Substituting this value of x(t) in the first equation of (c), we obtain

(e) yet) = t - (D 2 + D + I) (Cl + C2e-t + C3te-t + - 2t)

= t - (C2e-t - 2C3e-t+ C3te-t + 1 - C2e-t + C3e-t

- C3te-t + t - 2 + Cl + C2e-t + C3te-t + - 2t) ,
yet) = - [Cl + C2e-' + C3( _e- t + te- t) + - 2t - 1].
The determinant of (a), by (31.36), is (D +
1)(-D) (D +
1)(D2) +
which is of order three. Note that our pair of functions x(t), yet) contains
the correct number of three constants, and is therefore the general solution
of (a).
Example 31.54. Solve the system
(a) (2D - I)x +(D + 4)y =1,
Dx - Dy = t -,).

Solution. Here all four polynomial operators in (a) are of the same
order one. We can therefore retain either equati9n of (a). It will, however,
be found easier to retain the 8econd and change first by adding the
second to it. We thus obtain the equivalent system

(b) (3D - l)x +

4y = t,
Dx-Dy=t-1.
412 SYSTEMS. LINEARIZATION OF FIRST ORDER SYSTEMS Chapter 7

The system is now of the form (31.48). We therefore retain the first equation
and change the second by multiplying the first by D/4 and adding it to
the second. There results the equivalent triangular system
(c) (3D - l)x 4y= t, +
(3D 2 +
3D)x = 4t - 3.
The general solution of the second equation in (c) is
(d) x(t) = Cl C2e-t +
Jt 2 - it. +
Substituting this value of x in the first equation of (c), we obtain
(e) (3D - 1)(cl C2e-t+ +
Jt 2 - it) - t = -4y,
which simplifies to

(f) y ( t) = C2e
-I
2
4
+ 6"t - gt +-7 '
+ -Cl 4
The pair of functions (d) and (f) are the same as those obtained in Ex-
ample 31.361 by using the standard method of elimination. Note how much
easier the above method is. Note too that we obtained the correct number
of two constants immediately.
Example 31.55. Solve the system
(a) (D - + (D + l)y =
l)x 0,
t
(D + l)x + (D - l)y = 2'
Solution. We retain the first equation and change the second by multi-
plying the first by -1 and adding it to the second. We thus obtain the
equivalent system
(b) (D - 1)x + (D + l)y = 0,
t
2x - 2y =-.
2
Retain the second equation and change the first by multiplying the second
by (D + 1)/2 and adding to the first. There results the equivalent
triangular system
(c) 2Dx = (D t 1) t = t(1 + t),
t
2x-2y=-'
2
Integrating the first equation in (c), we obtain
t2 t t2
(d) 2x = "8 + 4" + 2c, x(t) = 16 + 8"t + c.
Lesson 31F DEGENERATE CASE 413

Substituting (d) in the second equation of (c), we obtain

t2 t t t2 t
(e) 2y = 8" + 4 - 2 + 2c, y(t) = 16 - 8" + c.
These are the same functions obtained in Example 31.37. Note how much
easier the above method is over the previous one. Note, too, that we ob-
tained the correct number of one constant immediately.
Example 31.56. Solve the system
(a) (D + 3)x + (D + l)y = el ,
(D + l)x + (D - l)y = t.

Solution. We retain the second equation and change the first by multi-
plying the second by -1 and adding it to the first. We thus obtain the
equivalent system
(b) 2x + 2y = e' - t,
(D + l)x + (D - l)y = t.

Retain the first equation and change the second by multiplying the first by
-(D - 1)/2 and adding it to the second. There results the equivalent
triangular system
(c) 2$ + 2y = e' - t,
2x = -!(D - l)(e ' - t) +t = !(t + 1).
From the second equation of (c), we obtain

(d) x(t) = :Ht + 1).

Subtracting the second equation in (c) from the first, we obtain

(e) 2y = I
e - - --,
2 2
3t 1
y(t) = -e'2 - -3t
4
- -41 .
These are the same functions x(t), y(t) obtained in Example 31.38. Note
that by this method we obtained the correct number of zero constants
immediately.

LESSON 3IF. Degenerate Case. ft(D)g2(D) - gl(D)12(D) = O. An

algebraic system of equations may be degenerate. In such cases the system
will have no solutions or infinitely many solutions. For example, the
systems
(a) 2x + 3y = 5, (b) 2x + 3y = 5,
2x + 3y = 7; 4x + 6y = 7,
414 SYSTEMS. LINEARIZATION OF FIRST ORDER SYSTEMS Chapter 7

have no solutions. On the other hand, each of the systems

(c) 2x + 3y = 5, (d) 2x + 3y = 0,
4x + 6y = 10; 4x + 6y = 0,
has infinitely many solutions. Note that in all four examples, the deter-
minant formed by the coefficients of x and y is zero. Note also that there
are no solutions, when in trying to eliminate x or y, the right side does not
also reduce to zero. There are infinitely many solutions when the right
side reduces to zero.
Similarly, we call the system of linear differential equations (31.4) de-
generate whenever its determinant

!I(D) gl(D) I = ! (D) gz (D) - (D)! (D)

I!z(D) gz(D) 1 gl 2

is zero. As in the algebraic system, there will be no solutions if, in trying

to eliminate x or y, the right side of the system is not zerc.; there will be
infinitely many if the right side is zero.
Example 31.6. Show that the system
(a) Dx-Dy=t,
Dx - Dy = t 2,
is degenerate. Find the number of solutions it has.
Solution. By (31.36), the determinant of (a) is _D2 - (_Dz) = o.
Hence the system is degenerate. Since the right side does not reduce to
zero when we eliminate x or y, it has no solutions. This example corresponds
to (a) at the bottom of page 413.
Example 31.61. Show that the system
(a) Dx - Dy = t,
4Dx - 4Dy = 4t,
is degenerate. Find the number of solutions it has.
Solution. The determinant of (a) is -4D 2 - (_4D2) = O. Hence
the system is degenerate. Its right side, however, reduces to zero when
we eliminate x or y. In this case there are infinitely many solutions of the
system. For example, the pair, x = + e, y = 5cis a solution; the pair

x = + ell Y = - + e2 is a solution. [In either equation, define

x(t) arbitrarily, and solve for yet). This pair of functions will also satisfy
the other equation.] This example corresponds to (c) at top of page.
Lesson 31G SYSTEMS OF THREE LINEAR EQUATIONS 415

Example 31.7. Show that the system

(a) (D + l)x + (D + l)y = 0,
(D - l)x + (D - l)y = 0,
is degenerate. Find its solutions.
Solution. The determinant of (a) is
(b)
D +1 D +
11 = (D + I)(D - 1) - (D + I)(D - 1) = O.
ID - l D-l
Hence the system (a) is degenerate. By (24.21), we can write (a) as
(c) (D + 1)(x + y) = 0,
(D - l)(x + y) = O.
+ +
Let u = x y. Therefore (c) becomes (D l)u = 0, and (D - l)u = O.
The solution of the first equation is u = cle- I; the solution of the second
equation is u = c2e l. Hence the solution of the system (c) is
(d)

However, if CI ¢ 0 and C2 ¢ 0, then the pair of solutions of (d) is in-

consistent, i.e., there are no functions x(t) and yet) that will satisfy them
simultaneously. If however, CI = C2 = 0, then the infinitely many func-
tions, such that
(e) x +y = 0 or x(t) = -yet)

will satisfy the system (c).

LESSON 31G. Systellls of Three Linear Equations. We discuss

briefly the system of three linear equations:
(31.71) h(D)x + (/1(D)y + hl(D)z = kl(t),
h(D)x + (/2(D)y + h 2 (D)z = k 2(t),
Ja(D)x + (/a(D)y + ha(D)z = ka(t).

Its determinant is defined as

(31.72) h(D) (/I(D) hI(D)
h(D) (/2(D) h 2(D) == h(/2h a + i2(/3h l + fa(/Ih 2 - h(/3h2
Ja(D) (/3(D) h 3(D) - i2(/Ih a - Ja(/2 h l.

The method of finding a general solution of the system (31.71) follows

the same rules outlined previously in solving a two-equation system. The
number of constants in the general solution of (31.71), i.e., in the set of
416 SYSTEMS. LINEARIZATION OF FIRST OBDER SYSTEMS Chapter 7

functions x{t), y{t), z{t) satisfying (31.71), as in the case of a two-equation

system, must equal the order of the determinant (31.72), provided this
determinant is not zero.
The usual standard method for solving the system (31.71) is to eliminate
one of the variables, say z, in the same manner we eliminate this variable
in an algebraic system of three equations. The system (31.71) can thus
be reduced to the system
(31.73) F 1 {D)x + G1{D)y = ll{t),
F 2 {D)x + G2 {D)y =
Finally by eliminating, in the usual manner, one variable from the system
(31.73), say y, we obtain the single linear equation
(31.74) F{D)x = k{t),
which we can solve for x. Substituting this value of x in either equation of
(31. 73) will enable us to solve for y. Substituting both values x and y
in anyone of the equations in (31.71) will enable us to solve for z. Since
in this process, we usually must multiply by an operator in order to effect
the desired eliminations, the number of constants in the set of functions
x{t), y{t), z{t) will generally be more than required. We then have to go
through the tedious process of finding what relationship exists among the
constants by substituting the three functions x{t), y{t), z{t) in either of the
two equations of (31.71) which were not used to find z{t). Since the three
functions must satisfy each equation in (31.71) identically, we can thus
determine from the equation a relationship among the constants, just as
we did in the two-equation system.
Fortunately it is always possible, in exactly the same manner described
for the two-equation system, to reduce the system (31.71) to an equivalent
triangular system of the form
(31.75) F 1 {D)x = k 1 {t),
F 2 {D)x + G2 {D)y = k 2 {t),
F 3 {D)x + Gs{D)y + Hs{D){z) = ks{t)·
If solutions are then obtained in the order x{t), y{t), z{t) starting with the
first equation in (31. 75), this set of functions will contain the c;lorrect
number of constants required in the general solution of the system (31.71).
Example 31.76. Reduce the following system to an equivalent tri-
angular one and determine the number of arbitrary constants needed in the
general solution.
(a) {D - 2)x + y - z= t,
-x + {2D + l)y + 2z = I,
2x + 6y + Dz = O.
Lesson 3IG SYSTEMS OF THREE LINEAR EQUATIONS 417

Solution. Here we can retain anyone of the three equations. We

decide to retain the first equation and change the second equation by multi-
plying the first one by 2 and adding it to the second. We change the third
equation by multiplying the first by D and adding it to the third. There
results the equivalent system
(b) (D - 2)x + y - z = t,
(2D - 5)x + (2D + 3)y = 2t + 1,
(D 2 - 2D + 2)x + (D + 6)y = 1.
We retain the first and third equations and change the second by multiplying
the third by -2 and adding it to the second. We thus obtain the equiva-
lent system
(c) (D - 2)x + y - z = t,
(-2D 2 +
6D - 9)x - 9y = 2t - 1,
(D2 - 2D + 2)x + (D + 6)y = 1.

We retain the first and second equations and change the third by multiplying
the second by (D + 6)/9 and adding it to the third. We thus finally obtain
the equivalent triangular system
(d) (D - 2)x+ y - z = t,
(-2D2 + 6D - 9)x - 9y = 2t - 1,
(-2D 3 + 3D 2 + 9D - 36)x = 12t + 5.
Solving the third equation for x will give three constants. Substituting this
value of x in the second equation will give an equation in y of order zero.
Hence the solution yet) will contain no new constants. Substituting the
solutions x(t), yet) in the first equation will give an equation in z that is
also of order zero and hence the solution z(t) will not contain any new
constants. We will thus have three arbitrary constants in the set of func-
tions x(t), yet), z(t). The determinant of (a), by (31.72), is also of order
three, and hence only three constants are needed in the general solution
of (a). We see, therefore, that by using this method the set of functions
obtained will contain the correct number of constants.
The extension of the standard method of solution to systems of linear
equations of order higher than three should now be apparent to you. To
determine the number of arbitrary constants in the general solution, you
will need to know the order of the determinant of the system. For this
information we refer you to any book on determinants and matrices. If,
however, you solve the system by first reducing it to an equivalent tri-
angular one, your solution will always contain the correct number of
constants.
418 SYSTEMS. LINEARIZATION OF FIRST ORDER SYSTEMS Chapter 7

LESSON 31H. Solution of a System of Linear Differential Equa-

tions with Constant Coefficients by Means of Laplace Transforms.
A system of linear differential equations with constant coefficients can
also be solved by means of Laplace transfonns. Unlike the operational
method, however, the Laplace transfonn method can be used only if initial
conditions are given and the interval over which the solutions are valid is
o t < 00. On the other hand, the Laplace method has an advantage
over the preceding operational one in that it will immediately yield a
particular solution satisfying given initial conditions without the necessity
of evaluating arbitrary constants.
Although we have confined our attention to a system of two differential
equations with two dependent variables, the extension of the method to
a larger system will be apparent.
Example 31.8. Solve the system
d 2x dy
(a) dt 2 - 4x + dt = 0,
2
dx
-4 dt + ddt y2 + 2y = 0,

for which x(O) = 0, x'(O) = 1, yeO) = -1, y'(O) = 2.

Solution. In Laplace notation, we can write (a) as-see Comment
27.24-
(b) L[x"(t)] - + L[y'(t)] = 0,
4L[x(t)]
-4L[x'(t)] + L[y"(t)] + 2L[y(t)] = O.

By (27.33), or directly from (27.29) and (27.31), (b) becomes

(c) 82L[x(t)] - x'(O) - 8X(0) - 4L[x(t)] + 8L[y(t)] - yeO) = 0,
-48L[x(t)] + 4x(0) + 82L[y(t)] - y'(O) - 8Y(0) + 2L[y(t)] = O.
Inserting the initial conditions in (c) and simplifying the resulting eqqa-
tions, we obtain
(d) (8 2 - 4)L[x] + 8L[y] = 1 - 1 = 0,
-48L[x] + (8 2 + 2)L[y] = 2 - 8.

Just as with operators, we now solve (d) as if they were ordinary algebraic
expressions in L[x] and L[y]. Multiply the first equation in' (d) by 48,
the second by 8 2 - 4, and add the two. The result is
(e) (8 4 + 28 2
- 8)L[y] = _8 3 + 28 + 48 -
2
8.
Hence
(f) L[] = _8
3
+ 28 2 + 48 - 8= ! [1 + V2 + 1 - V2 _ 8(8 - 2)] .
y (8 2 + 4)(8 2 - 2) 6 8 + V2 8 - V2 82 + 4
Leeson 31H SOLUTION OF A LINEAR SYSTEM BY LAPLACE TRANSFORMS 419

Referring to a table of Laplace transforms, we find that

(g) L[(1 + = 1 + 0,
8+ 0
L[(1 - = 1- 0,
8 - 0
L[-8 cos 2t] = 82
-88
+22 ' L[8 sin 2t] = 8 82 ! 22
Therefore, by (f) and (g),
(h) yet) = i[(1 + 0)e-v'2 t + (1 - 0)e-{2t - 8 cos 2t + 8 sin 2t].
Again as with operators, two methods are available for finding x(t).
We can start with (d) again and eliminate yet) or we can substitute (f)
in either equation of (d) and solve for L[x(t)]. Using this latter method,
we obtain by (f) and the first equation in (d)

(i) ( 2 ]
8 - 4)L[x + 8
(_8 + 28 + 48 -
3
(8 2 + 4)(8 2 _ 2)
2
8) = o.
G) L[x] _ 8(8 - 2)2(8 + 2) _ 8(8 - 2)
- (8 - 2)(8 + 2)(8 2 + 4)(8 2 - 2) - (8 2 + 4)(8 2 - 2)

= - 1 [ 2 - 0 + 2 + 0 _ 4 (8+2)].
12 8 - 0 8+ 0 82 + 4
Referring to a table of Laplace transforms, we find that

(k) - 0)ev'2 t + (2 + - 4 cos 2t - 4 sin 2t]

equals the expression on the extreme right of (j). Therefore, by (j) and (k),

(1) :t(t) = --;; [(2 - 0)e-{2t + (2 + 0)e--{2t - 4 cos 2t - 4 sin 2t].

Example 31.81. Solve the system (see Example 31.5)

dx dy
(a) dt + 4x + de = 1,

dx _ 2x
de
+y = t2
'
for which x(O) = 2 and yeO) = -1.
Solution. In Laplace notation (a) can be written as

(b) L[x'] + 4L[x] + L[y'] = L[I],

L[x'] - 2L[x] + L[y] = L[t 2 ].
420 SYSTEMS. LINEARIZATION OF FIRST ORDER SYSTEMS Chapter 7

By (27.33) or directly from (27.29), (b) becomes

(c) sL[x(t)] - xeD) + 4L[x(t)] + sL[y(t)] - y(D) = L(l),
sL[x(t)] - xeD) - 2L[x(t)] + L[y(t)] = L(t 2 ).

Referring to a table of Laplace transforms, we find that

1
(d) L(l) = -,
s
Substituting these values and the initial conditions in (c), and simplifying
the resulting equations, we obtain
(e) (s + 4)L[x(t)] + sL[y(t)] = + 1,
2
(8 - 2)L[x(t)] + L[y(t)] = S3 + 2.
Multiplying the second equation in (e) by -s and adding it to the first,
there results
212
(f) (8 +4)L(x) (-s +2s)L(x) = "8 +1 - 82 - 28. +
Solving (f) for L[x(t)] gives
2s 3 - S2 - S 2 + (s + 1)(2s2 - 3s + 2)
(g) L[x(t)] = 82(S2 - 38 - 4) = 82(S - 4)(s + 1)
28 2 - 38 +2 5 1 11
S2(S - 4) = 88 - 28 2 + 8(s - 4)
Referring to a table of Laplace transforms, we find that

(h)
8 88
L[.!!e
8
4t
] = 11
8(s - 4)
.

Hence,

(i) L [£ - + 1; e = 4t
] ;8 - 2!2 + 8(s 4) .

Therefore by (g) and (i)

(j) x(t) = £- + 181 e 4t

•

To find yet), we substitute (g) in either equation of (e). You can verify that

(k)
NOTE. These are the same results you would have obtained if you had
inserted the initial conditions in (c) and (d) of Example 31.5 where we
solved this same problem by means of operators. Verify it.
Lesson 31-Exercise 421

EXERCISE 31

Solve each of the following systems of equations.

dx 2 dy
1. dt = -x, dt = -yo

2• dx -I
dt = 3e ,
dy
Cit = x
+
y.
dx 2 dy 2
3· dt =xt, dt=yt.
dx dy dz
4. dt = 2t, dt = 3x 2t, dt = x + + 4y + t.
dx I dy x - Y
5. dt = e, dt = - t - ·

6• dx
dt = x
+.smt, dy
dt = t - y.
dx __ 3
7• dt x
+ 2e 31, dx
+ dydt -
dt
.
3y = sm 2t.
dx dy
8. dt = y, dt = -x + 2y.
dx dy
9. 3 dt + 3x + I
2y = e, 4x - 3 dt 3y = 3t. +
2
d x . dx 2
d y
10. dt 2 + 4x = 3 sm t, dt - dt 2 +
y = 2 cos t.
2
d x dy dx
11. dt 2 - 4x - 2 dt +
y = t, 2 dt x dt 2 =+ + o.
dxdy I dx dy
12.2dt+dt-x=e,3dt+2dt+y=t.
d 2x
2
d y dx dy
13. dt 2 +x - dt 2 - y = -cos 2t, 2 dt - dt - y = o.
2
d x dy dx dy
14. dt2 - dt = 1 - t, dt 2 dt = 4e
I
+ x. +
2
d x 2
d y dx dy
15. dt2 - x +
dt 2 +
y = 0, dt 2x dt + +
2y = O. +
2 2
dx dy d x d y dy 2
16. dt +
dt +
y = t, dt2 dt 2 dt +x + + +
y = t .

Verify that each of the following systems 17-20 is degenerate

(D = djdt). Find solutions if they exist.
17. Dx +2Dy = e l ,
Dx+ 2Dy = t.
18. Dx - Dy = el ,
3Dx - 3Dy = 3e' .
19. (D2 - 1)x +
(D2 - lJy = o.
(D2 + 4)x + (D2 + 4)y =- O.
422 SYSTEMS. LINEARIZATION OF FIRST ORDER SYSTE!.IS Chapter 7

20. (D - 2)x + (D - 2)y = t,

(D 3)x+ + (D + 3)y = t.
21. Reduce the following system to an equivalent triangular one. How many
arbitrary constants should the general solution contain?
(D - 2)x + 3y - z = t,
-x + (D + 3)y + 2z = 1,
2x + 39y + (D - 4)z = o.
22. Solve each of the following systems of equations.
(a) (D - l)x = 0,
-x +
(D - 3)y = 0,
-x + y +
(D - 2)z = O.
(b) (D - l)x = 0,
3x 2(D + l)y + = 0,
2y +
(2D - 3)z = O.
(c) (D - 2)x = 0,
-3x (D + 2)y + = 0,
-2y +
(D - 3)z = o.
(d) Dx - y +
z = 0,
-x +
(D - l)y = 0,
-x + (D - l)z = o.

Solve each of the following systems of equations by the method of the

Laplace transform.
dx
23. dt - y = t,
dy
x - dt = I, x(O) = 2, y(O) 1.
dx I
24.3 +3x+2y=e,
dt
dy
4x - 3 dt 3y = 3t, + x(O) 1, y(O) -1.
dx
25. dt + dydt - 4y = I,
x+ - 3y = t2, x(O) = 2, y(O) -2.
d2x dy
26'dt2-dt=l-t,

+2 = 4e' + x, x(O) = 0, y(O) = 0, x'(O) 1.

ANSWERS 31

1. X-I = t CI,+ y = c2e- l •

2. x = -3e- 1 + CI, y = ie-I - CI + cze l
•

3. X-I = -!(t 2 + CI), y = cze 13/3 •

Lesson 3I-Answers 423

4. X = t2 +
C}, y = t 3 t2 3clt C2, + + +
t4.+ it3
III = (6C} i)t 2 + (CI + "+ +
4C2)t + C3.
5. X = e' CI, +
y = t-Ie' CI C2t- l . + +
6. X = cle' - i(sin t cos t), y = t - 1 + + c2e-'.
7. X = (2t + cI)e 3"
Y = - 3.sin 2t is
2 cos 2t - i'(3t 2 + 3C}t + 2t + C2).
8. X = Cite' + c2e', Y = cite' + (C} + c2)e'.
9. X = clel/ 3 + c2e-'/3 - 6t, Y = -2clel/3 - c2e-'/3 + ie' + 9t + 9.
10. X = sin t + C} sin 2t + C2 cos 2t,
Y = i cos t - + iC2 sin 2t + c3e' + o4e-'.
iC} cos 2t
11. x = CI sin t +
C2 cos t + c3e' + c4e-',
y = (CI - 2C2) sin t + (2cI + C2) cos t - 3c3e' + c4e-' + t + 2.
12. x = cle' + c2e-' + ite' + 1, y = t - 2 - <t + cI)e' - 3c2e-' - ite'.
13. x = cle' + C2 sin t + C3 cos t - n(3 cos 2t + 4 sin 2t),
y = C}e' + (C2 - C3) sin t + (C2 + C3) cos t - -h(2 cos 2t + sin 2t).
14. x = Cje-' + c2e'/2 + 2e' + 2t,
_, C2 ,/2 , t2
Y = -cle + '2 e + 2e +"2 - t + C3·
15. x - 5ce- 2', y = -3ce- 2 '.
16. x = t 2 +
t - 1, Y = -to
17. No solutions.
18. Infinitely many. Define x(t) arbitrarily and solve for y(t).
19. Infinitely many; x(t) = -y(t).
20. No solutions.
21. One such equivalent triangular system is
(D - 2)x + 3y - III = t,
(2D - 5)x + (.6 + 9)y = 2t 1, +
(D2 - 12D 25)x+ = -lOt - 2.
General solution should contain three arbitrary constants.
3. C}. 2' 3CI.
22. (a) x = cle', Y = C2e - 2" e, III = C3e - '"2 e - C2e3' .
-3CI • + e _. 3c1 • + 2C2 -. + 3./2
(b) x = C}e', Y = -4- e C2 , III = - '"2 e "'5 e C3e.
(c) x = 4C}e 2 ', y = 3C}e 2 , + 5c2e-2., III = -6C}e 2 ' - 2c2e- 2' + c3e3'.
(d) x = CI +C2e', y = -C} + (C2t + c3)e', III +
= -C} (C2t - C2 + c3)e'.
23. x = ie' +
ie-', y = ie' - ie-' - t.
24. x = lje· /3 - lje-1/3 - 6t, Y = -1ge'/3 + lje-1/3 + 9t + 9 + ie'.

25. x = i 2
e ' - 4te
2
' +t + +
2
y = - e ' - 4te • + +
2 2

26. x = tt-' - lfe'/2 + 2e' + 2t,

4 _, 5 ./2 t2
y=-a e -a e +2e+"2- t +1.
,
424 SYSTEMS. LINEARIZATION OF FiRST ORDER SYSTEMS Chapter 7

LESSON 32. Linearization of First Order Systems.

Let the x and y components of the velocity of a particle be given by
dx dy
(32.1) dt = f(x,y), dt = g(x,y),

where the functions f(x,y) and g(x,y) are assumed to have continuous
partial derivatives for all (x,y). This system of two first order equations
may, in many problems, be difficult to solve. What we do in such cases
is to find the first three terms of the Taylor series expansion of f(x,y) and
g(x,y) , and then eliminate the parameter t between them. The Taylor
series expansions of these functions about the point (0,0) are, see Lesson 38,

(32.11) dt = f(x,y) = ao + alx + a2Y + a3x 2 + a"xy + a5Y 2 +"',

dy
dt = (
g x,y) = bo + blx + b2y + baX2 + b"xy + b6 y 2 + ....
Using only the first three terms in each expansion, and dividing the second
by the first, we obtain

(32.12)
dy bo + blx + b2y
dx = ao + atx + a2Y .
The solution of this equation will give what is called a fir8t approximation
to the path of the particle. '" The equation (32.12) is now of the type with
linear coefficients discussed in Lesson 8. As commented in 8.26, the result-
ing implicit solution of (32.12) will frequently be so complicated as to be
of little practical use in determining the path of the particle. We shall
now show how important information as to the character EJf the particle's
motion can often be obtained more easily from the differential equation
(32.12) itself than from its complicated implicit solution.
In Lesson 8B, we showed how by a translation of axes, it is always possi-
ble to transform a differential equation of the form (32.12) with linear
coefficients (assuming these represent nonparallel lines) to one of the form
(all! + blU) dx +(a2l! + °
b2U) dy = with homogeneous coefficients. All
we need do is to translate the origin to the point of intersection of the two
nonparallel lines represented by these linear coefficients. Hence we shall
assume in what follows that this shift of origin has been made. We there-
fore need consider only equations of the form
dy alx + bly l + b2 y
(32.13)
= a2x + b2y' at
a2 bb2 and a2 x ,- 0.

·It is po88ible for the first approximation to be very wide of the actual path of the
particle becaUBe the omitted terms are relevant and important.
Lesson 32-Case 2 LINEARIZATION OF FIRST ORDER SYSTEMS 425

To save writing, we define

(32.14)

Note that D is the determinant

at bll
Ia2
b2
whose elements are the coefficients of x and yin (32.13).
We consider separately each of the cases resulting from different values
of at, a2, b l , b2 • In all cases, dyjdx is meaningless at the origin (0,0).
Hence, in all cases, no solution will lie on this point. However, every other
point in the plane is an ordinary point, including points in a neighborhood
of the origin, no matter how small. Hence, by Definition 5.41, the origin is,
in all cases, a singular point. For convenience, however, we shall say
"solutions through the origin." This expression is to be interpreted to mean
solutions through points in a neighborhood of the origin, but not through the
origin itself.
Case 1. al = b2 = °
and b t = a2. In this case (32.13) simplifies to

(32.15) dy
dx
= J!..,
x
x ¢ °'
whose solutions are

(32.16) y = ex, x ¢ 0.

It is a family of straight lines through the origin (0,0). Thus for this
special case, the family of integral curves of (32.13) is easily drawn.
Case 2. bl = -a2' In this case, (32.13) becomes
dy
(32.17)
dx
Its solution by the method of Lesson 7B, is

(32.18)

Since b l = -a2, we obtain by (32.14),

(32.181)

which is also the discriminant of the quadratic equation (32.18).

From analytic geometry we know the following.
1. If D °
< and e ¢ 0, (32.18) is a family of ellipses with center at the
origin; if e = 0, only the point (0,0) satisfies the equation. But the
426 SYSTEMS. LINEARIZATION OF FIRST ORDER SYSTEMS Chapter 7

point (0,0) is excluded since, by (32.17), dy/dx is meaningless there.

Hence the family of integral curves of (32.13) will resemble those
shown in Fig. 32.19(a).
y

(a)

Figure 32.19

2. If D > 0 and c r6 0, (32.18) is a family of hyperbolas with center at

the origin; if c = 0, (32.18) degenerates into two straight lines through
the origin which are the asymptotes of the family of hyperbolas. Hence
the-family of integral curves of (32.13) will resemble those shown in
Fig. 32.19(b).
For this special Case 2, therefore, the family of integral curves of
(32.13) can also be easily drawn.
Case 3. General case, a l b2 r6 a2bl, with no other restrictions placed on
aI, a2, b}, b2. Let P(x,y) be a point on an integral curve of (32.13),
y

(a) (b)
Figure 32.2
Lesson 32-Case 3 LINEARIZATION OF FIRST ORDER SYSTEMS 427

Fig. 32.2(a). The slope of the tangent line CP to the curve is therefore y'.
Assume this line does not go through the origin O. Draw OB parallel to
CPo Its slope is also y'. From Fig. 32.2(a), we see that (tan <t:AOB =
y' = AB/x; therefore AB = xy')

(32.21) OC = AP - AB = y - xy'.

If therefore OC lies above the x axis, then y - xy' > o. Further we know
from the calculus that if y" > 0 at a point P of a curve, then the curve
in a neighborhood of P is concave upward and a tangent at P lies below
the curve. We have thus proved that if at a point P(x,y) on an integral
curve, y" > 0 and y - xy' > 0, then the tangent line at P separates
curve and origin in a neighborhood of P.
A similar conclusion can be proved if y" < 0 and y - xy' < 0, see
Fig. 32.2(b).
If, however, y" and y - xy' have different signs at P(x,y), then origin
and an integral curve near P will lie on the same side of the tangent line
at P. For example, in Fig. 32.2(a), draw at P an integral curve which is
concave downward so that y" < o.
By differentiation of our original differential equation (32.13), we obtain
after simplification,
(32.22)

By (32.13), the denominator of the fraction in (32.22) is not equal to zero.

Hence it is always positive since it is squared. Its numerator by (32.14)
is D. If therefore
(32.221) D > 0, y" and (y - xy') have the same sign,
D < 0, y" and (y - xy') have opposite signs.
Because of (32.221) and the remarks after (32.21), we can now assert
the following.
Comment 32.23. If y - xy' '¢ 0, i.e., if the tangent line at a point
P of an integral curve of (32.13) does not go through the origin, and if
D of (32.14) > 0, then curve and origin, in a neighborhood of P, are
separated by the tangent at P; if D < 0, then curve and origin, in a
neighborhood of P, lie on the same side of the tangent at P.
If the tangent at a point P of an integral curve of (32.13) goes through
the origin, then it is evident from Fig. 32.2(a) or (b), that OC = 0 and
therefore by (32.21), that y - xy' = O. This means that if (32.13) has
rectilinear solutions (i.e., straight line solutions) through the origin, these
solutions must satisfy the equation y - xy' = O. For example, the straight
line solutions through the origin which we found in Case 1, namely y = ex,
satisfy this requirement. For then y' = c and y - xy' = ex - xc = o.
428 SYSTEMS. LINEARIZATION OF FIRST ORDER SYSTEMS Chapter 7

The question we now ask is the following. Are there, for the general
ca8e, rectilinear solutions of (32.13) which go through the origin? The
answer is: all those solutions which satisfy the equation y - XV' = 0.
From (32.13) we find, after simplification,
, b2y2 + (a2 - b1)xy - alx 2
(32.231) y - xy = a2 x + b2y •
Hence the locus of those points (x,y) which will make the numerator of
(32.231) zero will make y - XV' = 0. Such loci will then be rectilinear
solutions of (32.13). To find them, we set
(32.24)
Let us call the discriminant of (32.24). Then
(32.25) = (a2 - b 1 )2 +
4a 1b2 = a2 2 - 2a2bl + b 2 + 4a b2
1 1

+ +
= a2 2 2a2bl b1 2 - 4a2bl 4a 1b2 +
+ +
= (a2 b 1)2 4(a 1b2 - a2bl)'
By (32.14), we can write (32.25) as
(32.26) = (a2 + b )2 + 4D.
1

From algebra, we know that if

(32.27) (a) > 0, then (32.24) will have two real, distinct factors, say
(ax by)(cx+ dy) = 0,+
(b) = 0, then (32.24) will have one real repeated factor,
(c) < 0, then (32.24) has only imaginary factors.
In case (a), there will thus be two rectilinear integral curves of (32.13)
through the origin. These two lines will separate the plane into four regions
in which all other integral curves of (32.13), for which y - XV' ¢ 0, will
lie.
In case (b), there will be only one rectilinear integral curve through the
origin. This line will divide the plane into two regions in which all other
integral curves of (32.13), for which y - xy' ¢ 0, will lie.
In case (c), there will be no real rectilinear solution of (32.13).
By (32.26), we note further that when D > 0, is also> 0. But when
D < 0, may take on any of the three values in (32.27). We have then
four possibilities to consider in the general Case 3: one when D > OJ
three when D < 0.
Case 3-1. D of (32.14) > 0, [and therefore of (32.26) > 0]. [NOTE.
If D > °
and also b 1 = -a2, then the special Case 2 applies; see 2 after
(32.181) of that case.] For this Case 3-1 we know from the remarks above,
that (32.13) has two rectilinear solutions, each of which is a factor of
Lesson 32-Case 3-1 LINEARIZATION OF FIRST ORDER SYSTEMS 429

(32.24). All points on these two lines will make y - xy' = o. All other
integral curves for which y - xy' 0, will lie in the four regions made by
the two rectilinear solutions. And since D > 0, we know by Comment
32.23, that a tangent line drawn at any point of an integral curve will sepa-
rate curve and origin. The integral curves will thus have the general
appearance of those shown in Fig. 32.281. The rectilinear solutions are
asymptotes of the integral curves. The origin itself, however, is a singular
point. The curves, while not hyperbolas, will have thcir general appear-
ance.
Example 32.28. Discuss, without solving the equation, the character
of the solutions of
, 4x - y
(a) y =2x+y·

Solution. Comparing (a) with (32.13), we see that al = 4, b 1 = -1,

a2 = 2, bz = 1. By (32.14), D = 4 + 2 = 6 > O. Hence this Case 3-1
applies. There should therefore be, and as we shall now show, there are
two factors of (32.24) each of which is a rectilinear solution of (a). Sub-
stituting in (32.24) the above values of all a2, b 1 , b 2, we obtain

(b) y2 + 3xy - 4x 2 = 0, (y + 4x)(y - x) = O.

Hence the two rectilinear solutions of (a) are

(c) y + 4x = 0, y - x = O.

Figure 32.281
430 SYSTEMS. LINEARIZATION OF FIRST ORDER SYSTEMS Chapter 7

[Verify that the functions defined in (c) are solutions of (a).] They are
shown in heavy lines in Fig. 32.281. Note that a tangent drawn at a point
of a nonrectilinear integral curve separates curve and origin.
Case 3-2. D of (32.14) < O. [NOTE. If D < 0 and also b1 = -a2,
then the special Case 2 applies; see 1 after (32.181) of that case.] For this
Case 3-2, as remarked earlier, of (32.26) may take on any of the three
values in (32.27). We shall therefore need to consider each of these three
possibilities separately. Before doing so, however, it will be necessary for
us to have additional information. We therefore digress momentarily in
order to obtain this information.
The family of solutions of (32.13) has slope y'. Let Yl' be the slope of
an isogonal trajectory family which cuts this given family in a positive <ta,
measured counter clockwise from Y' to Yl', Fig. 32.3.

y y' z slope of given family

yj z slope of isogonal
trajectory family

o x
Figure 32.3

By a formula in analytic geometry, therefore,

Yt' - Y'
(32.31) tan a = 1 + Yt'y'
Solving (32.31) for y', we obtain
, Yl' - tan a
(32.32)
Y = 1 + (tan a)Yt'
Replacing y' in (32.32) by its value as given in (32.13), there results

Yl' - tan a alx + b1y

(32.33)
1 + (tan a)Yl' = a2X + b2y·
The solution of (32.33) for Yl' is
, (a2 tan a al)x + + (b 2 tan a + b1)y
(32.34) Yl = (a2 - al tan a)x + (b 2 - b1 tan a)y ,
which has the same form as (32.13). We have thus proved that the differ-
ential equation of a family of isogonal trajectories, making an <ta with
the given family of solutions of (32.13), has the same form as (32.13).
Lesson 32-Case 3-2(a) LINEARIZATION OF FIRST ORDER SYSTEMS 431

The determinant Dr, whose elements are the coefficients of x and y in

(32.34), is
(32.35)
Dr = (a2 tan a +al)(b 2 - b1 tan a) - (b 2 tan a + b )(a2 -
1 al tan a)
+
= (a 1b2 - a2bl)(tan2 a 1) = D(tan 2 a 1) + = D sec 2 a,

where D is given by (32.14). Since sec 2 a > 0, we see from (32.35) that
Dr and D have the same sign.
By Case 2, the family of solutions of (32.34), which remember is an
isogonal family of trajectories of the family of solutions of (32.13), will be
ellipses or hyperbolas with center at the origin if

(32.36)

i.e., if
(32.37) tan a = a2 +bl .
al - b2
They are, by Case 2, ellipses if Dr < 0 or equivalently, as noted above,
if D < o.
Comment 32.371. We have thus proved that if D < 0 a.nd a is an
angle which satisfies (32.37), then the family of isogonal trajectories which
cuts the family of integral curves of (32.13) in this angle a, measured from
the integral family counterclockwise to the isogonal trajectory family, is
a family of ellipses with center at the origin. If a ;<! 0, each integral curve
will therefore approach the origin in one direction and recede infinitely in
the other direction. The origin itself is a singular point. If a = 0, then the
isogonal trajectory family coincides with the integral family and the
integral family is thus also a family of ellipses as in Case 2. [Note that if
a = 0, then tan a = 0 and by (32.37), a2 = -b l as in Case 2.]

We are now ready to examine each of the three possibilities under

Case 3-2.
Case 3-2(a). D of (32.14) < 0 and Ll of (32.26) > O. In this case,
there will be, by (32.27)(a), two rectilinear solutions of (32.13) through the
origin. These lines will separate the plane into four regions. For a. non-
rectilinear integral curve of (32.13), we now know the following facts.
1. It cannot cross either of the rectilinear solutions.
2. By Comment 32.23, origin and integral curve, in a neighborhood of a
point P on it, lie on the same side of the tangent at P.
3. If a is an angle which satisfies (32.37), then the family of isogonal tra-
jectories cutting the integral family of (32.13) in this angle a, measured
432 SYSTEMS. LINEARIZATION OF FIRST ORDER SYSTEMS Chapter 7

from the integral family counterclockwise to the isogonal trajectory

family, is a family of ellipses, with center at the origin, satisfying (32.34).
In addition, it has been proved
that
4. An integral curve will approach
the origin in a direction tangent
to one of the rectilinear solu-
tions; see Exercise 32,12.
5. A radius vector to a point P
moving along an integral curve
away from the origin will ap-
proach coincidence with the sec-
Figure 32.38
ond rectilinear solution.
The graphs of the integral curves will therefore resemble those shown in
Fig. 32.38. The two heavy lines are the rectilinear solutions.
Example 32.39. Discuss, without solving, the character of the solu-
tions of
(a) y
,
= 2x+4y
.
x-y

Solution. Comparing (a) with (32.13), we see that al = 2, b1 = 4,

a2 = 1, b2 = -1. By (32.14), D = -2 - 4 = -6 < oand by (32.26),
t1 = (1 +
4)2 +
4D = 25 - 24 = 1 > O. Hence this Case 3-2 (a)
applies. Since t1 > 0, there should be, and as we shall now show, there
are two rectilinear solutions of (a), each of which is a factor of (32.24).
Substituting in (32.24) the values of aI, bl> a2, b2 , we obtain
(b) + 3xy + 2X2 = 0, (y + 2x)(y + x) = O.
y2

Each of the lines y + 2x = 0 and y + x = 0 are solutions of (a). (Verify

it.) They are shown in heavy lines in Fig. 32.391. All other integral
curves must lie inside the four regions formed by these two lines.
From (32.37), we find tan a = 5/3, so that a is approximately 59°.
With this value of tan a, (32.34) becomes
,_ (I + 2)x + (-I + 4)y
Yl -
_
-
¥X + ill -_ llx +
7y
,
(1 - lj)x + (-1 - -Ix - ¥y -7x - 23y

which has, as it should, the same form as (32.17). Hence by (32.18), its
solution is
(c) llx 2 + 14xy + 23y2 = c.
LeSSOD 32-Case 3-2(b) LINEARIZATION OF FIRST ORDER SYSTEMS 433

A rotation of the axes through an angle of approximately 65°18' will

eliminate the xy term and (c) will become, approximately,

(d)

which represents, as it also should, a family of ellipses. Some of these

ellipses are shown in Fig. 32.391. Each nonrectilinear integral curve of

Figure 32.391

(a) will cut this family at the constant angle a = 59°, measured counter-
clockwise from the integral curve to the ellipse. We have shown a part of
one such integral curve in the figure.
Case 3-2(b). D of (32.14) < 0 and d of (32.26) = o. In this case,
there will be, by 32.27(b), only one rectilinear integral curve through the
origin. This line will separate the plane into two regions. The same com-
ments we made in the previous lesson in regard to a nonrectilinear integral
curve will apply here. These are:
1. It cannot cross the rectilinear solution.
2. By Comment 32.23, the origin and each integral curve in a neighbor-
hood of a point P on it, lie on the same side of the tangent at P.
3. If a is an angle which satisfies (32.37), then the family of isogonal
trajectories cutting the given family of solutions of (32.13) in this angle
a, measured from the integral family counterclockwise to the trajectory
family, is a family of ellipses, with center at the origin, satisfying
(32.34).
4. An integral curve will approach the origin in a direction tangent to the
rectilinear solution; see Exercise 32,13.
5. A radius vector drawn to a point P moving along the integral curve
away from the origin will approach coincidence with the rectilinear
solution.
434 SYSTEMS. LINEARIZATION OF FIRST ORDER SYSTEMS Chapter 7

The graphs of the integral curves will thus resemble those shown in
Fig. 32.4.

Figure 32.4

Example 32.41. Discuss, without solving, the character of the solu-

tions of
+3y
y , =x - -.
(a)
x-y
Solution. Comparing (a) with (32.13), we see that al = 1, bl = 3,
a2 = 1, b 2 = -1. By (32.14), D = -1 - 3 < 0. By (32.26) A =
(4)2 +4( -4) = 0. Hence this Case 3-2(b) applies. There should be,
and as we shall now show, there is, only one rectilinear solution of (a).
Substituting in (32.24) the values of aI, a2, b l , b 2 give above, we obtain
(b) y2 + 2xy + x 2 = 0, (y + X)2 = 0.
Hence the line y = -x is the only rectilinear solution of (a). We have
left it to you as an exercise, to find the trajectory family of ellipses as we
did in Example 32.39, and the angle a of (32.37) which each integral curve
makes with these ellipses. Draw some of these ellipses and an integral
curve. Its graph should resemble a curve of Fig. 32.4.
Case 3-2(c). D of (32.14-) < 0, A of (32.26) < O. In this case by
(32.27) (c), there are no rectilinear solutions of (32.13). And if a is an angle
satisfying (32.37), then by Comment 32.371, the family of isogonal trajec-
tories which cuts the given family of solutions of (32.13) in this angle a
measured from the integral family counterclockwise to the trajectory family,
is a family of ellipses with center at the origin. If therefore a = 0, the
family of solutions of the original equation (32.13) coincides with the tra-
jectory family and thus are also ellipses with eenter at the origin. Hence
if a = 0, every integral curve encircles the origin.
If, however, a ¢ 0, then each integral curve of (32.13) will cut the isog-
onal trajectory family of ellipses, which are integral curves of (32.34), in
Lesson 32-Case 3-2(c) LINEARIZATION OF FIRST ORDER 435

this constant angle a. We shall now prove for this Case 3-2(c), that when
a ¢ 0, an integral curve will also encircle the origin.
The equation of the family of straight lines through the origin is
(a) Y = mx; y' = m.

Replacing m in the second equation of (a) by its value y/x obtained

from the first equation, we have
(b) y' = m =!!...
x
Let 8 be the angle of intersection of a line of (a) and an integral curve of
(32.13), measured from the integral curve to the line, Fig. 32.42. Then,
by a formula in analytic geometry,
!!.. _ alX + b1y
(c) tan 8 = x a2X + b2 y
1 + '!!. alX + b1y
x a2X +b y 2

By (b) we can write (c) as

(d)

o
Figure 32.42

An integral curve of (32.13), therefore, cuts each line of (a) at the angle 8
given by (d). [For each line of (a), there is a distinct m and therefore by
(d), a definite angle 8.] The numerator of (d) is
(e)
Its discriminant is
(f)

which is exactly the same as the discriminant d of (32.24) [see (32.25)].

Since by our assumption for this Case 3-2(c), d < 0, it follows that the
436 SYSTEMS. LINEARIZATION OF FIRST ORDER SYSTEMS Chapter 7

expression in (f) is less than zero. Hence, as noted in (32.27)(c), there are
no real values of m which will make (e) zero. Since (e) is the numerator
of (d), this means that the angle 8 in (d) is never equal to zero. An integral
curve of (32.13), since it cuts every
straight line through the origin at
an angle 8 0, must therefore en-
circle the origin.
Hence in both cases, a = 0 and
a 0, an integral curve of (32.13)
will go completely around the origin.
In the first case, it will enclose the
origin in the form of an ellipse as in
Fig. 32.19(a); in the second case in
the form of spirals as shown in Fig.
32.43(a) and (b). In both cases the
Figure 32.43 origin is a singular point.

Example 32.44. Discuss, without solving, the character of the solu-

tions of
dy
(a)
dx

(NOTE. This equation was solved in Example 8.25. Its implicit solution
can be found there.)
Solution. A translation of the axes so that the origin is at (-1, 1),
see Example 8.25, transforms (a) to the form

dy -2x+y
(b)
dx x+y

Comparing (b) with (32.13), we see that al = -2, b l = 1, a2 = 1,

b2 = 1. You can verify that D of (32.14) < 0 and d of (32.26) < O.
Hence this Case 3-2(c) applies. By (32.37)

(c) tan a = -i, a = -33°41'.

With this value of tan a, (32.34) becomes

(d) Yl' = (-i - 2)x + (-I + l)y = -8x+ y.

(1 - t)x + (1 + i)y -x+ 5y
Its solution by (32.18) is

(e) 8x 2 - 2xy + 5y2 = c.

Lesson 32-Case 3-2(c} LINEARIZATION OF FIRST ORDER SYSTEMS 437

You can verify that (e) has, as it should, only imaginary factors. A rota-
tion of the axes through an angle
of approximately 73°10' will trans-
form (e) to
y
(f)

which represents a family of ellipses. x

A few of these ellipses are shown in
Fig. 32.45. Each integral curve must
cut each ellipse at the constant angle
-33°41', measured from integral
curve to ellipse. We have also shown
in this figure part of one such inte-
gral curve. Figure 32.45

Comment 32.46. Position of the Particle as a Function of Time.

In the above discussions of the path of a particle, we eliminated the im-
portant element of time. Hence, although we have shown the general
appearance of the path, in the absence of an equation connecting the posi-
tion of the particle with time, we cannot know where the particle is at any
instant. To obtain this information, we must change equation (32.13)
back to the system from which it came, namely

(32.47)

and then solve it by the methods of Lesson 31E. In operator notation, we

can write (32.47) as
(32.48) (D - az)x - bzy = 0,
-alx + (D - b1y) = O.
Following the method outlined in Lesson 31E, we retain the fir8t equation
and change the second by mUltiplying the first by (D - b1)/b z and adding
it to the second. There results the equivalent triangular system
(32.49) (D - az)x - bzy = 0,
(D - b1)(D - az)
[ b
z
- al x
]
= 0.
The second equation in (32.49) reduces to
(32.5)
The characteriRtic equation of (32.5) is
(32.51)
438 SYSTEMS. LINEARIZATION OF FIRST ORDER SYSTEMS Chapter 7

m2 =
a2 + bl - v(a2 -
2
b l )2 + 4a l b2
Hence the solution of (32.5) is
(32.53)
where ml and m2 have the values given in (32.52). Note that the char-
acter of the roots ml and m2, and therefore the solution x, depends on
whether the discriminant (a2 - bl )2 +
4a l b 2 is positive, zero, or negative.
Substituting this value of x in the first equation in (32.48), we obtain
(32.54) b 2y = (D - a2)(clem1t C2em2t) +
= mlcle m1t
+
m2c2em2t - a2clemlt - a2c2em2 '
+
= cI(ml - a2)em1t c2(m2 - a2)em2t.
Hence
(32.55) y = ;2 [cl(ml - a2)e
m1t
+ C2(m2 - a2)e
m2t
].

By Theorem 31.33, the pair of functions x(t), yet) as given in (32.53) and
(32.55), have the correct number of two arbitrary constants. The value of
the constants CI and C2 will depend on the position of the particle at t = O.
The solutions x(t), yet) of (32.53) and (32.55) will then give the position of
the particle at later times t, see, for example, Exercise 32,11.

EXERCISE 32

Discuss without solving the character of the solution of each of the fol-
lowing differential equations. (First determine to which of the different
cases each belongs.) Draw rectilinear solutions where these exist. Draw
a rough graph of an integral curve.
1. dy = y +
1. 6. dy = 3x y. +
dx x-I dx 2x+ y
2. dy = x - 3y. 7. dy = -2x y. +
dx 3x+y dx x+y
3. dy = x - 3y 8. dy = x y . +
dx 3x - lOy dx 3x-y
4. dy = 4x y .+ 9. dy = -x - y.
dx -2x+ y dx x+2y
5. dy = 2x +
y. 10. dy = x - y.
dx 3x+ y dx x +y
Lellson 32-Exercise 439

ll. Assume the original system from which the equation of problem 4 above
came is
dx dy
dt = -2x y,+ dx = y,4x+
and that the initial conditions are x(O} = 1, yeO} = 2. Find the parametric
equations of the path and the values of x and y when t = 1.
12. Prove statement 4 of Case 3-2(a}. Hint. Follow the proof given in
Case 3-2(c}.
13. Prove statement 4 of Case 3-2(b}. Hint. Follow the proof given in
Case 3-2(c).

ANSWERS 32
1. Case 1. 2. Case 2; family of hyperbolas.
3. Case 2; family of ellipses. 4. Case 3-1; y 4x, y -x.
5. Case 3-2(a); y = 0.73x, y = -2.73x.
6. Case 3-1; y = 1.3x, y = -2.3x.
7. Case 3-2(c). 8. Case 3-2(b}; y = x. 9. Case 2; family of ellipses.
10. Case 2; family of hyperbolas.
ll. x(t} = ie- 3t + fe 2t , yet) =-ie- +3t Ve 2 t; x = 4.45, y = 17.71.
 Chapter 8

Problems Giving Rise to Systems

of Equations. Special Types of Second
Order Linear and Nonlinear Equations
Solvable by Reduction to Systems

LESSON 33. Mechanica1. Biologica1. Electrica1 Problems

Giving Rise to Systems of Equations.

LESSON 33A. A Mechanica1 Problem-Coupled Springs. Two

masses ml and m2 rest on a frictionless, horizontal table. They are con-
nected to each other and to two fixed supports by three unstretched
springs 8}, 8 2 , 8 3 with respective spring constants k}, k2' k3' Fig. 33.1(a).

(a)

(b)

Figure 33.1

[For the meaning of a spring constant, see (28.61).] The masses can be
displaced from their equilibrium position by holding, for example, ml>
moving m2 to the right and then releasing both. Figures 33.1(a) and (b)
show the position of the two masses before and after a displacement. We
wish to find the equation of motion of each mass.
440
 Lesson 33A A MECHANICAL PROBLEM-COUPLED SPRINGS 441

Denote by 0 1 the position of ml and by O2 the position of m2, when

the system is at rest. Let Xl be the displacement of ml from its equi-
librium or rest position 0 1 , and X2 the displacement of m2 from its equi-
librium position O2, at time t. Hence when ml is at Xl and m2 is at X2,
Fig. 33.1(b), spring 8 1 has been elongated a distance Xl from its equi-
librium position and spring 8 2 has been elongated a distance X2 - Xl
from its equilibrium position. For example, if Xl = 1 foot and X2 = ! feet,
then spring 8 2 has been elongated only i foot, i.e., a distance X2 - Xl.
Therefore the restoring forces acting on ml, with the positive direction to
the right, are two in number.
1. A spring force due to 8 1 acting to the left. By Hooke's law, which
you will find in Example A of Lesson 28C, this force is -klXl.
2. A spring force due to 8 2 acting to the right. By Hooke's law, this
force is k 2(X2 - Xl). (Remember the spring 8 2 wants to restore itself
to its original size and will therefore pull on ml to the right and on
m2 to the left.)

Hence the differential equation of motion of ml is

d2XI
(33.11) ml dt 2 = -kIXI + k 2(X2 - Xl).

A similar analysis shows that the restoring forces acting on m2 are

two in number.
1. A spring force due to 8 2 acting to the left. By Hooke's law, this force
is -k 2(X2 - Xl).
2. A spring force due to 8 3 acting to the left. Remember 8 3 has been
squeezed a distance X2 and it wants to restore itself to its original size.
Therefore, by Hooke's law, this force is -k3X2.
Hence the differential equation of motion of m2 is
d 2 x2
(33.111) m2 dt 2 = -k 2(X2 - xd - k 3X2.

The solution of the system of linear differential equations (33.11) and

(33.111) will give the equations of motion XI(t) and X2(t) of the coupled
springs diagrammed in Fig. 33.1.
If, in addition to the spring forces, a forcing function FI(t) is attached
to ml and a forcing function F 2(t) is attached to m2 as shown in Fig.
33.121, then the pair of equations (33.11) and (33.111) becomes
d2 x
(33.12) ml (It-f + klXl + k 2(XI - X2) = FI(t),
442 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

+
-+

Figure 33.121

And if we assume that because of friction or because the system is

.
Immerse d m
. a me d'mm, t h ere IS
. a dampmg
. f orce Tl dXl ac t'mg on ml
dt an d
2
a damping force T2 d:e acting on m2 then the system (33.12) becomes

d2Xl dXI ) (
(33.122) ml dt 2 + TI lit + klXI + k 2(Xl - Xz = FI t),

dZX2 dxz ()
mz dt Z + Tzlit + k3XZ + kz(xz - Xl) = Fz t.

If all spring constants have the same value k, then (33.122) simplifies
to the linear system

dZXl dXl k F (
(33.13) ml dt 2 + Tl lit + 2kxI - Xz = 1 t),

In operator notation, we can write (33.13) as

(33.14) + TID + 2k)Xl - kxz =
(m l D 2 Fl(t),
-kXl + (mzDZ + TzD + 2k)xz = Fz(t).

We can eliminate X2 by multiplying the first equation of (33.14) by

(mzDZ +
TzD +
2k), the second by k and adding the two. We thus
obtain
(33.15) {mlm z D4 +
(mlTz mZTl)D3 +
[TlT2 +
2k(ml mz)]D Z + +
+
[2k(Tl + +
Tz)]D 3kZ}Xl = (mzD z +
TzD 2k)Fl(t) kFz(t), + +
an equation which is now linear in Xl' Its general solution can be ob-
tained by the usual methods outlined in previous lessons. When Xl has
been determined, the general solution for Xz can be obtained from the
first equation in (33.14). Note that the general solution of (33.15) for
Xl (t) will contain four constants. Substituting this value of Xl (t) in the
first equation of (33.14) and solving it for X2(t) will add no new constants.
By Theorem 31.33, the correct number of constants required in the gen-
LeMOn 33A-Exercise

eral solution of the system (33.15) is four. Hence the pair of functions
XI(t), x2(l) will contain the correct number of constants.
Comment 33.16. The characteristic equation of the left side of
(33.15) set equal to zero is of the fourth degree. Hence, as we remarked
in Comment 18.26, these roots may not be easy to find, especially if they
are imaginary. Remember these roots are needed in order to write the
complementary function y. of the general solution of (33.15). If however
ml = m2, i.e., if both masses are the same, and rl = r2, then the elimina-
tion of X2 in (33.14) will result in

(33.17) [(m I D2 + rID + 2k)2 - k2]XI

= (m 1 D 2 + rID + 2k)FI(t) + kF2(l).
But now if we set the left side equal to zero, its characteristic equation is
(mlm 2 +
rIm +
2k)2_k2 = 0 which factors into

(33.18)
Its roots can thus be readily obtained by setting each factor equal to zero.
Comment 33.19. If the forcing functions FI(t) and F 2 (l) are iden-
tically zero and if rl !1; 0, r2 !1; 0, then as we showed in Lessons 28A
and 29A, the motion of the system is stable. If the forcing functions are
not identically zero, then the motion is much more complicated and a
breakdown of the system may occur if the phenomenon of resonance is
present, see Exercise 33A,I(b). (If you have forgotten the meaning of
resonance, reread Lessons 28D and 29B.)

EXERCISE 33A
I. (a) Solve the system (33.13) if ml = m2 = mj rl = r2 = 0; FI(t) =
F2(t) = o.
(b) The system in (a) has two frequencies. These are called normal
frequencies. (In simple harmonic motion, a system had only one
natural frequency.) Find the normal frequencies of (a). Resonance
occurs if a forcing function, impressed on the system, has either of
these normal frequencies.
(c) Find the solution if at t = 0, m2 is held fixed and ml is moved to the
right A units and released. Hint. Initial conditions are t = 0, Xl = A,
X2 = 0, dxl/dt = 0, dX2/dt = o.
2. (a) Solve the system given by equations (33.11) and (33.111) if ml = 1,
m2 = 2, kl = 1, k2 = 2, k3 = 3.
(b) What are its normal frequencies?
3. If you write the system of problem l(a) in operator notation, you will obtain
(a) (mD2 + 2k)XI - kX2 = 0, -kXI + (mD2 + 2k)X2 = O.
Since each differential equation of the system is linear, it is reasonable to
444 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

assume that its solutions will have the form

(b)
Show that the substitution of (b) in (a) yields the system
(c) (mw 2 + 2k)Cl - kC2 = 0, -kcr + (mw 2 + 2k)C2 = o.
Note that (c) is obtainable from (a) by replacing D by w, Xl by Cl, X2 by C2.
The system (c) will have nontrivial solutions for Cl and C2 (i.e., solutions
other than Cl = 0, C2 = 0) only if the determinant

(d)
-k I= 0
mw
2
+ 2k '

see Theorem 63.42. Expand the determinant and solve the resulting equa-
tion, called the characteristic equation, for w. Note that the imaginary
part of the solution for w is the same as the normal frequencies found in
l(b). You will have thus proved that the imaginary part of the solution
of w, where w is obtained by the process outlined above, will give the normal
frequencies of the system.
4. Use the method outlined in problem 3 to find the normal frequencies of the
system of problem 2. What is the determinant? Compare with answer to
2(b).
5. (a) Show that when Xl = X2, the system of coupled springs solved in
problem l(a) has only the one normal frequency Vk/m rad/unit time;
that when Xl = -X2, it has only the one normal frequency V3k/m
rad/unit time.
(b) Show that if the initial conditions are so chosen that C3 = C4 = 0 in
the answer to problem l(a), then Xl = X2 and the system will vibrate
with only the one normal frequency Vk/m; if chosen so that Cl =
C2 = 0, then Xl = -X2 and the system will vibrate with only the one
normal frequency V3k/m.
(c) Let Yl = Xl +
X2 and Y2 = Xl - X2. Show that the solution of l(a)
becomes respectively

Yl + 02 cos v'k/m t,
0 1 sin v'k/m t
Y2 03 sin v'3k/m t + 04 cos v'3k/m t.
Note that each equation has only one normal frequency. The first
equation (where Yl = Xl + X2) corresponds to a motion where the two
masses move in coordination to the left and to the right with the same
amplitude; the second equation (where Yl = Xl - X2) corresponds to a
motion where the two masses move in opposite directions with the same
amplitude. These new coordinates Yl and Y2 which replace Xl and X2
are called normal coordinates. They are useful in simplifying certain
problems in the theory of vibrations.
6. (a) S<:llve the system (33.13) if ml = m2 = 2, k = 32, rl = r2 = 0 and
/<\(t) = F2(t) = 4 cos t.
(b) In (a), replace Fl(t) and F2(t) by 4 cos wt. What values of w will produce
resonance?
7. Two masses ml and m2 are connected by two springs Sl and S2 with respec-
tive spring constants kl and k2 as shown in Fig. 33.191. After the system is
brought to rest, the masses are displaced from their equilibrium position.
Lesson 33A-Exercise 445

Assume no damping forces and no forcing functions.

(a) Find the differential equations of motion of the system.
(b) Find the equations of motion of the system if kl = k2 = k and ml
m2 = m.
(c) What are the normal frequencies of the
system?
8. Find the normal frequencies of the system
of problem 7 by means of the method out- - - x, =0, equilibrium
lined in problem 3. What is the determi- m, position of m,
nant? Compare with answer to 7(c).
9. (a) Use the method outlined in problem 3
to find the normal frequencies of the - - X2=0, equilibrium
system shown in Fig. 33.191, i.e., with m2 position of m2
ml m2 and kl k2. See problem 7.
(b) Write the normal frequencies if ml = Figure 33.191
m2 = m .
.10. Solve the system of Fig. 33.191 if kl = 1, k2 = 2, ml = 1, m2 = 2, and if,
'" after the system is'brought to rest, the mass m2 is held fixed, the mass ml
is displaced A units downward and both masses released. Hint. At t = 0,
Xl = A, X2 = 0, dxt/dt = 0, dX2/dt = 0.
11. In the system illustrated in Fig. 33.191, assume that the mass ml is damped
by a shock absorber whose coefficient of resistance is r and that a forcing
function F = Fo sin wt is attached to ml. (a) Show that the differential
equations of motion are
(mlD2 + rD + kl + k2)Xl - k2X2 = Fo sin wt,
-k2Xl + (m2D2 + k2)X2 = 0.
Hint. See answer to problem 7(a). Modify it to include the damping factor
and the forcing function. (b) By eliminating X2, show that
[(mlD2 + rD + kl + k2)(m2D2 + k2) - k22]Xl = Fo(k2 - m2w2) sin wt.
Remark. The general solution of this last equation for Xl will, as usual, be
the sum of a complementary function Xl c and a particular solution Xl". It
can be shown that the motion due to the Xl c part of the solution is a transient
one and subsides with time. If, therefore, k2 and m2 are so chosen that
k2 = m2w 2, then the right side of the equation is zero. Hence Xl" = 0, and the
motion due to the Xl" part of the solution is also zero. This means that if
k2 = m2w 2, the mass ml will in time approach a position Xl and remain
there. Hence, the mass ml will not vibrate. This result has practical use
when it is necessary that the vibrations of an instrument be negligible.

ANSWERS 33A

1. (a) Xl Cl sin vk/m t +

C2 cos Vk/m t + C3 sin V3k/m t
+ C4 cosV3k/m t,

X2 = Cl sin vk/m t +C2 cos vk/m t - C3 sin V3k/m t

- C4 cosV3k/m t.

(b) Vk/m rad/unit time; V3k/m rad/unit time.

446 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter

A
(c) Xl ="2 (cosVk/mt+ cosV3k/mt),
A
X2 = "2 (cosvk/mt - cosV3k/mt).

2. (a) Xl = CI cos (v'l.3t t + §l) + C2 cos (V4.19 t + 62),

X2 = ![1.69C} cos (v'l.3t t + Ih) - 1.l9c2 COS (v'4.i9 t + 62)).
(b) v'l.3t rad/unit time, V4.19 rad/unit time.
4./ w + 3 :-2+ /.
2

-2 2w 5
6. (a) Xl = C} sin 4t +
C2 cos 4t +
Ca sin 4vli t +
C4 cos 4Vli t + h cos t,
X2 = C} sin 4t +
C2 cos 4t - ca sin 4-\/li t - C4 cos 4vli t + h cos t.
(b) w = 4 rad/unit time and w = 4Vli rad/unit time.
2
d xI
7. (a) ml dt 2 = -kIXI + k2(X2 - xi),
2
d x2
m2 dt 2 = -k2(X2 - Xl).

(b) Xl = CI sin VO.38k/m t + C2 cosVO.38k/m t + Ca sin V2.62k/m t

+ C4 cos V2.62k/m t.
X2 = 1.62C} sin VO.38k/m t + 1.62c2 cos VO.38k/m t
- 0.62ca sin V2.62k/m t - 0.62c4 sin V2.62k/m t.
(c) w = VO.38k/m and w = V2.62k/m.
2
8./ mw + 2k I.
-k +k mw
2
9. (a) / mlw + kl + k2 /.
-k2 m2W + k2
w = _! (kl + k2 :l::! '(k1 + k2 + k2)2 _ 4klk2 .
2 ml m2 2'\J ml m2 mlm2

(b) w = ,_ kl - 2k2 :l:: Vkl2 + 4k22 .

'\J 2m

10. Xl = [(3 - va) cos (Va -; 0) t+ (3 + va) cos (Va; 0) t],

X2 = -;0)t - (v'6; 0) tJ.
COs or

Xl = A(0.211 cos 0.518t + 0.789 cos 1.932t),

X2 = A (0.289 cos 0.518t - 0.289 cos 1.932t).
Lesson 33B A BIOLOGICAL PROBLEM 447

LESSON 33B. A Biological Problelll. There is a constant struggle

for survival among different species. One species survives by eating
members of another species; a second by preventing itself from being
eaten. Certain birds live on fish. When, for example, there is an abundant
supply of fish, the population of this bird species flourishes and grows.
When these birds become too numerous and consume too many fish, thus
reducing the fish population, then their own bird population begins to
diminish. As the number of birds decreases, the fish population increases.
And as the fish population increases, the bird population starts increasing,
etc., in an endless cycle of periodic increases and decreases in the respective
populations of the two species.
Problems concerning the rise and fall of the populations of interacting
species have been studied extensively by biologists and mathematicians.
The theoretical mathematical results which we shall develop in this lesson
agree fairly accurately with actual population trends of certain interacting
species. *
We shall consider the problem of determining the populations of two
interacting species: a parasitic species P which hatches its eggs in a host
species H. Unfortunately for the H species, the deposit of an egg in a
member of H causes this member's death. For each 100 of population:
Let Hb denote the number of births per year of the H species,
Hd denote the number of deaths per year of the H species, if no P
species were present, i.e., Hd denotes H's natural death rate,
P d denote the number of natural deaths per year of the P species.
We now make the following assumptions:
1. That Hb, Hd, and Pd are constants. Hence if at time t, x is the popu-
lation of the H species, y the population of the P species, then in time At
(a) Hb tJ.t is the approximate number of births of the host H,

(b) -Hd tJ.t is the approximate number of natural deaths of the

host H,
(c) -Pd At is the approximate number of natur!J,1 deaths of the
parasite P.
2. That the number of eggs per year deposited by the P species result-
ing in the death of the H species is proportional to the probability that
the members of the two species meet. Since this probability depends on
the product xy of the two populations, we assume that in time tJ.t,
(d) -kxytJ.t

·V. A. Kostitzin, Mathematical Biology, London, G. G. Harrap &: Co. Ltd. (1939);
Vito Volterra, Le8 A88ociatiom Biologiquea au Point de Vue Mathemo.tique, Paris,
Hennann &: Co. (1935).
448 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

is the approximate number of deaths of the host H due to the presence of

the parasite P, where k is a proportionality constant. Hence, also, since
each such death of H implies the laying of an egg by P,

(e) kxytJ.t

is the approximate number of births of P.

Therefore in time tJ.t, the approximate changes in the Hand P species
are respectively
x x
(33.192) tJ.x = Hb 100 tJ.t - Hd 100 tJ.t - kxytJ.t,

tJ.y -- k xytJ.t - y tJ.t.

Pd 100

In the first equation of (33.192), let h = (Hb - Hd)/lOO, assumed >0,

i.e., h is the net natural annual percentage increase of the population of H.
In the second equation of (33.192), let p = P d/1OO, i.e., p is the natural
annual percentage death rate of P. Then (33.192) simplifies to

(33.2) tJ.x = hxtJ.t - kxytJ.t,

tJ.y = kxytJ.t - pytJ.t.

Note that we have used the words "approximate changes" in connec-

tion with the system (33.2). It is conceivable that in some time intervals
s:I tJ.t, no births or deaths take place; in
.S! other time intervals, one or more births
I or deaths occur. A graph of the popula-
... .r tion of a species as a function of time is
therefore not a continuous curve. It will
show sudden jumps or drops as in Fig.
33.21. if the population is
Time large, it is convenient to assume that
there is a continuous curve which will ap-
Figure 33.21 proximate the population trend. Further,
one need not stretch one's credulity too
far to believe that the system of equations
dx
(33.22) dt = hx - kxy,
dy
dt = kxy - py,

which result, when in (33.2), we divide each equation by tJ.t and let tJ.t -+ 0,
will give good approximations for large populations and long periods of
time.
Lesson 338 A BIOLOGICAL PROBLEM 449

The equations in (33.22) form a system of two nonlinear first-order

equations which cannot be solved by the methods thus far discussed. In
later chapters we shall describe other methods by which an approximate
solution of the system (33.22) may be obtained, if the numerical values
of h, k, and p are known, see Exercise 39,5.
It is possible, however, to solve the system (33.22) as an implicit func-
tion of x, if we divide the second equation by the first. Making this
division, we obtain
dy y(kx -- p)
(33.23) dx = x(h -- ky) ,

Integration of the second equation in (33.23) gives

(33.24) log y" -- ky = kx -- log x P -- log c,

from which we obtain

(33.25) log cy"x P = ky + kx,

If at t = 0, the population of the host species is Xo and the population
of the parasitic species is Yo, then the last equation in (33.25) becomes

(33.26)

which is the value of the arbitrary constant in the solution.

As we have frequently mentioned, implicit solutions are usually of little
value. Here it would be difficult indeed to plot a graph of (33.25) showing
the population y as a function of x.

Although we cannot at this stage solve the system (33.22) for x and y
as functions of time, nevertheless certain important useful information
can be obtained from it. Suppose, for example, that at a certain moment
of time, the populations of our two interacting species are

h
(33.27) y ="k.

Then by (33.22) dy/dt = 0 and dx/dt = O. Since the rate of change of

the two populations at that moment is zero, there are no increases or
decreases. Let us call these populations the equilibrium populations of x
and y respectively. Let
(33.28) x = + X,
where X and Y represent the respective deviations of the equilibrium
450 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

populations of x and y. Substituting (33.28) in (33.22), we obtain

(33.29) ft +X) = h +X) - k +X) Y) ,

ft Y) = k +X ) Y) - p Y) .

These equations simplify to

dX dY
(33.291) -
dt
= -pY - kXY
'
de = hX + kXY.

If we assume that the deviations X and Yare small, we may, without

serious error, discard the XY terms in (33.291). These equations then
become
dX dY
(33.292)
de
-pY, de = hX.

But now we recognize that the system of equations is exactly the same
as those discussed in Lesson 32 under the heading "Linearization of First
Order Systems." Dividing the second equation in (33.292) by the first,
we have
dY hX
(33.293) - pY·
dX

Its solution, obtained by solving it directly, or by recognizing that this

differential equation comes under Case 2 of Lesson 32 where its solution
is given in (32.18), is
(33.294)

Both p and h are positive quantities. (Remember h is the net natural

percentage increase of the host popu-
lation, assumed >0, and p is the natu-
ral percentage death rate of the para-
site population.) If therefore h = p,
(33.294) is the equation of a family
of circles; if h ¢ p, it is a family of
ellipses. We have thus shown that if
the populations of two interacting
species remain stationary over long
Figure 33.295 periods of time, as they do in many
cases, then the graph of their popu-
lation deviations from equilibrium is a circle or an ellipse, Fig. 33.295.
Note how the mathematical equation (33.294) supports our opening
Lell80n SSC COMPLEX ELECTRICAL CIRCUITS 451

remarks. When the deviation X of the host species equals VCl1i and is
thus largest, the deviation Y of the parasite species is zero, i,e., the para-
site population is normal. But now there are more hosts available in
which the parasites can lay their eggs. The population of the parasite
species thus begins to increase, while that of the host species decreases.
But as the parasites get too numerous, they kill more hosts, the
parasite deviation is maximum and the host deviation is zero. When this
point is reached, there are relatively so many parasites around, that they
begin killing off some of the host's normal population. As the host's
population decreases below normal, the parasitic population, because there
are now insufficient hosts, must also decrease, until finally a point is
reached when its deviation is zero and the host's deviation has its largest
negative value -Vc7h, The host's population is now too low to support
a normal parasitic population. The parasitic deviation therefore begins
to decrease below zero, thus allowing the host's deviation to increase, etc.,
in an endless cycle of periodic increases and decreases.

LESSON 33C. An Electrical Problelll. More COlllplex Circuits.

In Lesson 30, we discussed a simple electric circuit in which the charged
particles moved in one path. A circuit in which the charged particles can
move in different paths will, as we shall show below, give rise to a system
of differential equations. To enable us to write differential equations with
correct signs, we shall adopt the following convention. If a branch of the
circuit contains a source of energy E, we shall indicate by an arrowhead
that the direction of the current is from the side which we label +, to the
side which we label -. In every other branch of the circuit, we put in
arrowheads arbitrarily. It is only essential that the arrowheads remain
fixed in a problem.
Example 33.3. Set up a
system of differential equa-
tions for the current in the
electric circuit shown in Fig.
33.31.
Solution. By Kirch-
hoff's first law, the alge-
braic sum of the currents at
any junction point is zero. Figure 33.31
If we call the current whose
arrowhead is directed toward a junction point positive and the current
whose arrowhead is directed away from a junction point negative, then at
the junction point marked 1 in Fig. 33.31,
(a)
452 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

And at junction point 2,

(b)
Applying Kirchhoff's second law to the left circuit, we obtain
. di2
(c) Rt + L2dt + C1 q = E(t).

If no source of energy is present in a closed circuit, then his second law

states that the algebraic sum of the voltage drops is zero. Applying this
law to the right circuit, we obtain
R . di2
Itl + CI 1
ql - L2 dt = 0.

Differentiating (c) and (d) and replacing i( = dq/dt), wherever it appears

by its value in (a), we have
2
d i2 (di 1 di2) I. . dE
(e) L2 dt2 + R dt +
dt +
C (tl t2) = dt'
)
T
d 2i2 dil I.
L2 de - Rl dt - C tl = 0.
1

The linear system (e) can be solved for il and i2 by the method of Lesson
3ID or E. The current i will then be the sum of these two currents.
Comment 33.311. We obtained two equations in the above problem
by using the left and right circuits shown in Fig. 33.31. It is possible to
obtain additional equations by using other circuits. Each of these, how-
ever, will not be independent of the two obtained in (c) and (d). For
example, we could, if we had wished, taken the circuit going completely
around the border and have thus obtained

Ri + Rlil + dl ql + bq = E.

This equation, however, is the sum of (c) and (d).

Example 33.32. Set up a system of differential equations for the cur-
rents in the electric circuit shown in Fig, 33.33.
Solution. By Kirchhoff's first law, the algebraic sum of the currents
at any junction point is zero. Hence at the junction points correspondingly
marked in Fig. 33.33, we have
(a) CD: il - i2 - i4 = 0, i4 = il - i 2 ;
0: i2 - is - ia = 0, i5 = i2 - ia;
0: ia + i6 - il = 0, i6 = il - i a·
Lesson 33C COMPLEX ELECTRICAL CIRCUITS 453

At junction point 0, i4 +
is - i6 === O. And we see from the sums of CD,
(y, and @, that this condition is satisfied.

C1 CD i2 L2

i.
Hz
q. C.
C ie R6 i6
® qe @
®
is
is
Rs Ls

Figure 33.33

Applying Kirchhoff's second law to the upper left, the upper right and
the lower circuit respectively, we obtain the linear system

(b) Rlil + dl ql + d4 q4 + q6 = E,

di2 + R . + R . -
L 2 de C14 q4 = 0,

- d6 q6 - R6i 6 + La d;: + Rais = O.

Differentiating (b) and replacing d/dt by D and dqn/dt by in, there results

(c) R 1D · + C1 . + C1 . + C1.t6 = dE
dt.'
I 4 6

L2D2i2 + R2Di2 + RS Di6 - d4 i4 = 0,

La D2ia - d i6 - R6Di6 + RaDia = O.

6

Substituting in (c) the values of i 4, i 6 , i6 as given in (a), we obtain

(d) ( RID + Cl
..!... + C4
..!... + C6
..!...) il - ..!... i2
C4
- ..!... ia =
Co
dE,
dt

- d4 i l + (L2D2 + R2D + R6 D + d) i2 - R6 Dia = 0,

- il - RsDi2 + (La D2 + R6 D + Ra D + d6) ia = O.

The system (d) can be solved for ii, i 2 , and ia by the method of Lesson
310.
454 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

Comment 33.34. As remarked in Comment 33.311, other equations

in addition to those found in (b) can be obtained by considering different
circuits. These, however, will not be independent of those in (b). For
example, we could have taken the circuit going completely around the
border. This equation, however, can be obtained from (b). Verify it.
Can you find other possible circuits?

EXERCISE 33C
1. A transformer consists of a primary coil and a secondary coil. The primary
coil has an emf of E(t) volts, a resistance of Rl ohms, and an inductance of
LI henrys. The secondary coil has a resistance of R2 ohms and an inductance
of L2 henrys. Let il and i2 be the respective currents in each coil. It has
been shown that the differential equations for the currents in the coils are
dil
(33.4) Ll at + Rltl. + M Cit
di2
= E(t),

di2
L2 at + R2t2. + M at
dil
= 0,

(a) Solve the system (33.4) if E(t) = °

where M is a constant called the Rlutual inductance.
and M2 < LlL2.
(b) Show that the exponents a and b, as given in the answer section, are
negative quantities and therefore that both il and i2 approach zero as
t-+oo.
(c) Solve the system (33.4) if E(t) is a constant Eo and M2 < LlL2. Show
that as t -+ 0, the current il of the primary coil approaches the steady
state current Eo/RI and the current i2 of the secondary coil approaches
zero as t -+ 00.
2. (a) Set up a system of differential equations for the current in the electric
circuit shown in Fig. 33.41.
..i 'lJlnr'
L i2

E·f • ..
i2
fR'
Figure 33.41

(b) Solve the system with initial conditions t = 0, i = °

when the switch
is closed, and show that the circuit is equivalent to one in which the
resistors Rl, R2 are replaced by one resistor R, whose coefficient of re-
sistance is given by
1 1 1
-=-+-,
R RI R2
Remark. If more resistors R3, ... ,. Rn were inserted in the circuit of
Fig. 33.41, parallel to Rl and R2, it follows from (b) that this circuit is
equivalent to one in which the resistors Rl, R2, ... , Rn are replaced by
Lesson 33C-Exercise 455

one resistor R whose coefficient of resistance is given by

1 1 1 1
R = R1 + R2 + ... + R,:
3. (a) Set up a system of differential equations for the charges in the electric
circuit shown in Fig. 33.42.
(b) Solve the system for q and i as functions of time with initial conditions
t = 0, q1 = q2 = 0 when the switch is closed, and show that the circuit
is equivalent to one in which the capacitors C1, C2 are replaced by one
capacitor whose capacitance is given by C = C1 C2. +
..i ..i2
Eo
1T
"N'v
R
£1
q1T C1 i2 1
Q2T
C
2
i
• •
Figure 33.42

Remark. If more capacitors C3,' ... , C" were inserted in the circuit of
Fig. 33.42, parallel to Cl and C2, it follows from (b) that this circuit is
equivalent to one in which the capacitors C1, C2, ... , C n are replaced by
one capacitor whose capacitance is given by C = C1 C2 C".+ + ... +
In the problems below, set up a system of differential equations for the
current in each of the circuits diagrammed. Solve as many as you wish.
4. See Fig. 33.43. At t = 0, i = 0, q = q1 q2 O.

Eo=50
P.
T Q

C_2·1O- 4

Figure 33.43

5. In problem 4, replace Eo = 50 by E(t) = 100 sin 60t.

6. See Fig. 33.44. At t = 0, i = 0, q = q2 = O.

R=2

Figure 33.44
456 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS ChapterS

7. See Fig. 33.45. At t = 0, i = 0, q2 = 0.

Figure 33.45

S. See Fig. 33.46. Initial conditions are t = 0, i = 0, ql = q2 0.

Figure 33.46-

9. See Fig. 33.47.

Ll i2

i4
q. C4 L2
ie Re R6 i6

i3
is q3 La Ra

Ca

Figure 33.47

10. See Fig. 33.48.

Figure 83.48
Lesson 33C-Answers 457

n. See Fig. 33.49.

I II
E(t) "V

®
Figure 33.49

ANSWERS 33C
1. (a) il = cle
al
+ bl
c2e , where

b _ -(L2Rl + LlR2) ::I:: V(L2Rl + LlR2)2 - 4RlR2(LlL2 - M2)

a, - 2(LlL2 _ M2)
-(L2Rl+LlR2) ::I::V(L2Rl-LlR2)2+4M2RlR2
2(LlL2 - M2)

i2 = [(LlL2 -
2
M )(ac le
al
+ bC2ebl) + L2Rl(cleal + C2 ebl )].
(c) il = cle + c2e bl + (Eo/ Rl); i2 is the same as in l(a).
al

2. (a) (LD + Rl)il + LDi2 = Eo,

(b) i = il + i2 = Eo(Rl + R2) [1 _ e-RIR2t1L(Rl+R2>]
RlR2
== (1 _ e- R • ,L ).

3. (a) (RD + ql + RDq2 = E, ql - 112 == o.

(b) ql = ClEo(1 - e-II R(CI +C2>, q2 = C2 E o(1 _ e-II R(C, +C.> ,

.= . + .
l II l2 = Cit
dql + Cit
dq2 Eo -., R(C, +C2> _ Eo
= If e = R e
-II RC
.

•
't!.
D't + Cl
ql + C
q = E
0, C2 l l
q2 - Cl ql = 0"
,l = II
+ l2..
D il
2
+ D i2 +
2
il + (il + i2) = 0,
{ -1
l2
.
-
1.
-ll = O.
C2 Cl
458 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

5. In 4 replace Eo by 100 sin 60t. Therefore E(t) = 6000 cos 60t.

6. (RD + LD2 + il + ( RD + i2 = -5000 sin lOOt,

Ll D2i l - ( R2D + i2 = O.

7. Ri+ L Ll d;: = E(t),

Ri +L + q2 = E(t) ,
dil I
'-Ll at+ C2 q2 = O.

The three equations are not independent. The third can be obtained by
sUbtracting the second from the first. To solve the system, it is easier to
use the first and third equations.

8. Ll at + ClI ql = Eo, R2i2 + C2I q2 = Eo.

iql dq2
Ll -
dt 2
+ -CI l
ql = Eo, R2 -dt -I q2
+ C2 = Eo.

ql = ClEo ( I - cos _ rr;-T t

I) R
t
sm _ rr;-T ,
V ClLl 2 V ClLl

ClEo. I Eo t
il = - - - sm - - - t - - cos - - - .
VClLl VClLl R2 VClLI

q2 = C2Eo(1 - e-·IR2C2), i2 = !: e-·

IR C
• ••

dill.
9. Ll at C4 q4 +
R616 = E(t), +
.+1
R212 C2 q2 +Ldi2+R'
2 at I q4 = 0,
515 - C4
di3 I
L 3 dt +R·+
313 C3 q3 - R'
6'6 - R'
5115 = 0.

. te'
SubStltu 1,.
dqn and'14 = 1\
= de . ..
- 12, 15 = 12 -
. .. . .
'3, '6 = 11 - 13·
2 2
dil
10. Rl at + L dt2
d il - d i2
L dt2 + C1 (.11 d E()
. ) = iii
- 12 t, R'
IH + R'212 = 0.

II. Junction point @: il - i5 - i2 = 0, is = il - i2.

Junction point @: i2 - i6 - i3 = 0, i6 = i2 - i3.
Junction point @: i3+ i7 - i4 = 0, i7 = i4 - i3.
Junction point 0: i4+ is - il = 0, is = il - i4.
At junction point ®: is +
i6 - i7 - is = 0, which satisfies the four
equations above. Set up, in the usual manner, the differential equation for
Lesson 34A VELOCITY AND ACCELERATION FORMULAS 459

each circuit marked I, II, III, and IV in diagram. For I, it is

Rlil + qs + Ls d;: = E(t).

Although there are other choices of circuits, their equations will not be in-
dependent of the chosen four.

LESSON M. Plane Motions Giving Rise to Systems of Equations.

In Lesson 16, we discussed the motion of a particle constrained to move
along a straight line. In this lesson, we consider the motion of a particle
free to move in a plane.

LESSON MA. Derivation of Velocity and Acceleration Formulas.

If a particle is free to move in a plane, then a change in the direction of
its velocity will be equally as important as a change in the magnitude of
its velocity. Ai3 mentioned in Lesson 16C, quantities in which both mag-
nitude and direction pmy.a role are called vectors.
Since Newton's second law of motion is also applicable to particles
which move in a plane, we have, by (16.1),
dv.
(34.1) F = ma = m dt ... mv.

Remark. A dot over a variable means its derivative with respect to

time; two dots over the variable means its second derivative with respect
to time. Hence:i; ... dx/dt, x ... d 2 x/dt 2 , iJ ... dv/dt, etc.
The mass m is not a vector quantity. We therefore see, by (34.1), that
the acceleration of a particle acted on by a force, not only has magnitude
F /m, but also has the 8ame direction
asF.
In Fig. 34.11, the vector F repre-
sents the magnitude and direction of
a force F. It is convenient to break Fy=Fsin 8
up this vector force into two compo-
nents, one, F"" to represent that part
of the force which accelerates the 0 F,,=Fcos8
..
particle in the x direction, the other,
F II, to represent that part of the force Figure 34.11
which accelerates the particle in the
y direction. If the inclination of the force F is 8, we see from Fig. 34.11,
that '
(34.12) Fx = Fcos 8, FII = Fsin 8.
Since F x = mass times ax, the acceleration of a particle in the x directIon
460 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

and F 1/ = mass times al/' the acceleration of a particle in the y direction,

we obtain from (34.12), the system of equations
d 2x ..
(34.13) F cos 8 = F:z: = ma:z: = m dt 2 "" mx,

In a similar manner we can break up any vector quantity into its x and y
components. In Fig. (34.14) we have shown, for example, the x and y
components of a velocity vector v.

. dy.
' -",-"0' .-.-,

o d:z: •
V:z:=VC08 8- dt =%

Figure 34.14

For certain problems, it is often convenient to use polar coordinates

instead of rectangular coordinates. The vector quantity is then broken up
into two components: one along the radial r direction, the other in a
direction perpendicular to it. In Fig. 34.15, we have broken up the vector
v into these two components Vr and v,.

y=r sin ()

x=rcos ()

Figure 34.15

Let (x,y) be the coordinates of a point P in a rectangular system and

(r,8) its coordinates in a polar system, Fig. 34.15. Then we see from
the figure, that
(34.16) x = r cos 8, y = r sin 8.
Lesson MA-Exercise 461

Differentiation of (34.16) gives

dx dr . d8
(34.17) dt = cos8 -rsm8 ,
dt dt
dy . dr d8
dt=sm8dt+rcos8 dt ·

A second differentiation gives

2 2 2 2
d x d r . dr d8 (d8) • d 8
(34.18) dt 2 = cos 8 dt 2 - 2 sm 8 dt dt - r cos 8 dt - r sm 8 dt 2 '
2 2 2 2
d y . d r
dt 2 = sm 8 dt 2 + 2 cos 8 dr d8
dt dt -
. (d8)
r sm 8 dt + r cos 8 ddt 8 2 •

These formulas, (34.17) and (34.18), are valid for every value of 8. Hence
they must hold in particular when 8 = O. But when 8 = 0, x = rand
the direction perpendicular to r is the y direction. Hence the components
of velocity and acceleration in a radial direction and in a direction per-
pendicular to it are
dx dy
(34.19) Vr = dt' VB = dt'

Substituting the first two equations of (34.19) in (34.17), the second two
in (34.18), we obtain with 8 = 0,
dr. d8.6
(34.2) Vr = dt == r, VB = r dt == rv.
2 2
ddt r r (d8) r•• - r 8 ,
2
(34.21) ar = 2 - dt ==
2
dr d8
aB = 2 dt dt + r ddt 8 == 2r8. + rO.
2

Formulas (34.2) and (34.21) give respectively the components of the

velocity and acceleration vectors of a particle along the radial axis and
in a direction perpendicular to it, at the point P where the curve crosses
the x axis. Since the x axis can be chosen in any direction, these equations
are valid for every point P of the particle's path.

EXERCISE 34A
1. In Fig. 34.22, we have shown the x and y components of a velocity vector v
as well as its components in a radial direction and in a direction perpendicular
to it. With the aid of this diagram, prove (34.2). Hint. First show that
dx dy .
(a) liz = dt = II cos a, lIu = de = II sm a.
462 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

Then show that

(b) vr = V cos (a - 8) = v cos a cos 8 + v sin a sin 8,
dx
= dt cos 8 + dydt sm. 8.
In (b), replace dx/dt and dy/dt by their values as given in (34.17).
y

P(r,8),
........

y=rsin8

x
Figure 34.22

Similarly, show that

(c) v, = v sin (a - 8) = v sin a cos 8 - v cos a sin 8
dy 8 dx. 8
=dt cos -dtsm.
In (c), replace dy/dt and dx/dt by their values as given in (34.17).

P(r,8),

y=r sin 8

(0,0) x
Figure 34.23

2. In Fig. 34.23, we have shown the x and y components of an acceleration vec-

tor a as well as its components in a radial direction and in a direction per-
Lesson 34B THE PLANE MOTION OF A PROJl':CTILE 463

pendicular to it. With the aid of this diagram, prove (34.21). Hint. First
show that
2
d x
(a) a z = dt 2 a cos a, a sin a.
Then show that
(b) a. = a cos (a - 9) = a cos a cos 9 + a sin a sin 9
2 2
d x
= dt 2 cos 9 + ddt2y sm• 9.
In (b), replace d 2 x/dt2 and d 2 y/dt 2 by their values as given in (34.18).
Similarly, show that
(c) a, = a sin (a - 9) = a sin a cos 9 - a cos a sin 9
2 2
d y d x .
= dt 2 cos 9 - dt 2 sm 9.
In (c) replace d 2 y/dt 2 and d 2 x/dt 2 by their values as given in (34.18).
3. A particle of mass m is attracted to a fixed point 0 by a force F. The particle
moves in a plane with a constant speed but not in a straight line. Show that
the particle moves in a circle with center at O. Hint. The component at of
the acceleration vector a in a direction tangent to the path of the particle
measures the change in the speed of the particle. Since this speed is con-
stant, at = o. Hence the acceleration acts only in a direction perpendicular
to the path of the particle. And since F and a are in the same direction, F
and the constant speed Vo are perpendicular to each other. Then show
dy/dx = - x/yo

LESSON 34B. The Plane Motion of a Projectile.

Example 34.3. A particle of mass m is projected from the earth with
a velocity Vo at an angle a with the horizontal. The only force acting on
it is that of gravity. Assuming a level terrain, find:
1. The equation of the particle's path.
2. Its horizontal range.
3. The maximum height it will reach.
4. The value of a for which the range will be a maximum.
5. When the particle will reach the ground.
Solution. Refer to Fig. 34.31. We take the x and y axes in a plane
which is perpendicular to the ground, and which contains the given
velocity vector Vo. The z axis is, as usual, perpendicular to both x and y
axes. Since by assumption the only active force F is that of gravity, the
components of F in the x, y, z directions are respectively F:x; = 0, FII =
-mg, F. = O. Hence
(a) mX = 0, my = -mg, mz = O.
464 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

Integration of (a) gives

(b) V. = Z = Ca.
Since the velocity vector Vo lies in the xy plane, it follows that at t = 0,
i.e., at the moment of projection, see Fig. 34.31,
(c) v., = Vo cos a, VII = Vo sin a, V. = 0,
where a is the angle which Vo makes with the horizontal x axis. Sub-
stituting these values in (b), we find
(d) Cl = Vo cos a, Cz = Vo sin a, Cs = 0.
Hence (b) becomes
(e) V", = :i: = Vo cos a, VII = if = -gt + Vo sin a, V. = z= 0.
Integration of (e) gives
2
t
(f) x = (vo cos a)t + C4, Y = -g"2 + (vo sin a)t + CIl, Z = Ce.

If we now choose our origin at the point where the particle is projected,
then at t = 0, x = 0, y = 0, Z = 0. Substituting these values in (f), we

y 1+
.±.

"
z
Figure 34.31

find C4 = 0, Cs = 0, Ce = 0. The parametric equations of the path of the

particle are therefore
gt Z
(g) x = (vo cos a)t, y = (vo sin a)t - 2"' Z = 0.

Since Z = 0, it follows that a projectile 8ubject only to a gravitational force

move8 in a pla'ne containing the vector Vo.
To find the equation of the path in rectangular coordinates, we eliminate
t by solving the first equation in (g) for t and substituting this value in
the second equation. There results
(h) y = (tan a)x - (
g sec2
x,
2 a)
Lesson 34B-Exercise 465

which is the equation of a parabola through the origin. Since the coeffi-
cient of x 2 is negative, the curve is concave downward.
The horizontal range of the projectile, i.e., the distance from the origin
to the point where the particle strikes the ground, is obtained by setting
Y = 0 in (h), for when y = 0 the particle is at ground level. Equation
(h) then becomes, if a '¢ 7r/2,
2
g x2
(i) o= (sin a)x - -2 2 ' X = g
(2 sin a cos a),
Vo COS a
2
v0 •
X = -sm2a,
g

which gives the horizontal range of the particle.

The maximum height is reached when y' = O. Therefore, differentiat-
ing (h) with respect to x and setting y' = 0, we obtain

g sec 2 a V 2
0= tan a - --2-X,
Vo
X = .JL. sin a
g
COS a.

When x has the value in (j), y by (h) has the value

2 2 2
Vo • 2 Vo • 2 Vo • 2
(k) y=gsm a - 2gsm a= 2gsm a,

which gives the maximum height reached by the particle.

The range will be a maximum when, in (i), a is so chosen that x is a
maximum. Hence differentiating the last equation in (i) with respect to
a, and setting dx/da = 0, we have
2v 2 7r
(I) o= _0_
g
cos 2a, a= -.
4

The range therefore will be a maximum if the projectile is fired at an angle

of 45°. By (i), this maximum range is v02 /g.
The particle will reach the ground when y = o. Setting y = 0 in (g),
we obtain
t2 2vo .
(m) (vo sin a)t - g2 = 0, t = -sIn a,
g
which is the time it takes the particle to reach the ground.

EXERCISE 34B
In problems 1-14, assume the air resistance is negligible.
1. A projectile is fired from the earth with a velocity of 1600 ft/sec at an angle
of 45°. Find the equation of motion, the maximum height reached and the
range of the projectile.
466 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

2. Start with the equation of motion (h) of Example 34.3.

(a) Show that the coordinates of the vertex of the parabola are
2 •
vo smgacosa, Vo 2 sm• 2 a) .
( 2g
(b) Show that the distance of the vertex to the focus is vo 2 cos 2 a/2g, and
therefore that the equation of the directrix is y = vo 2 /2g. Note that the
equation has no a in it. Hence the parabolic orbits of all projectiles
fired with a given velocity have the same directrix regardless of the
angle at which they are fired.
(c) Finally show that this constant height of the directrix above the hori-
zontal is the distance a projectile reaches when fired straight up with an
initial velocity of Vo ft/sec.
If you have succeeded in answering the above questions, you have proved
that every parabolic orbit of a projectile fired with the same velocity Vo
has the same constant directrix whose height is the distance the projectile
would reach if fired vertically.
3. We showed in Example 34.3 that if a projectile is fired at an angle of 45°,
its range will be a maximum and will equal vo 2/g. An artillery piece, whose
muzzle velocity is Vo ft/sec, is located at a distance D < v02/g from an ob-
ject at the same level as itself. Show that there are two angles at which the
artillery piece can be fired and hit the object-one as much greater than
45° as the other is less. Find these angles. Hint. In (h) of Example 34.3,
you want a such that when y = 0, x = D. Make use of the identity
2sinacosa = sin2a = cos(2a -
4. The muzzle velocity of an artillery piece is 800 ft/sec. Assuming a level
terrain, answer the following questions.
(a) An object is 3.8 mi away. Can it be hit?
(b) An object is 15,000 ft away. At what angles must the artillery piece be
fired in order to hit the object?
(c) What is the maximum height reached by the shell of (b)?
(d) When did the shell reach the object?
(e) If a mountain of height 6000 ft is 4000 ft from the artillery piece, is it
still possible to hit the object?
5. A projectile, fired with a velocity of 96 ft/sec, reaches its maximum height
in 2 sec. Assume a level terrain.
(a) Find the angle of projection of the particle. Hint. : = There-
fo're dy/dx = 0 if dy/dt = O.
(b) Find the maximum height reached by the particle.
(c) What is the range of the projectile from the point fired?
6. A projectile is fired from a height of Yo ft above a level terrain, with a velocity
of vo ft/sec and at an angle a with the horizontal. Find:
(a) The equation of the particle's path.
(b) Its horizontal range-take the x axis on ground level.
(c) Its maximum height.
(d) When it will reach the ground.
(e) At what angle and with what velocity it will strike the ground. (Hint.
+
If 8 is the angle, tan 8 = (dy/dt)/(dx/dt) and Ivl = v' (dx/dt) 2 (dy/dt)2).
(f) The value of a that will make the range a maximum.
Lesson 34B-Exercise 467

7. A projectile is fired from a height of 50 ft above a level terrain with a velocity

of 64 ft/sec at an angle of 45°. Answer the questions asked for in 6, with
the exception of (f). In regard to (f), find the angle of projection that will
make the horizontal range a maximum and the value of this maximum range.
8. The maximum range of a projectile when fired on a level terrain is 1000 ft.
(a) What is its muzzle velocity?
(b) What is the maximum horizontal distance it can travel if the projectile
is fired from a 40-ft-high platform? Hint. First find the firing angle for
maximum horizontal range, see problem 6.
9. The maximum distance a boy can throw a ball on level ground is 100 ft.
Neglecting the height of the boy, find
(a) The velocity with which the ball leaves his hand,
(b) The maximum horizontal distance he can throw the ball if he stands on
a roof which is 40 ft above the ground. Hint. First find the throwing
angle for maximum horizontal range, see problem 6.
10. Two athletes, one 6! ft tall, the other 5! ft tall, can each put a shot with the
same velocity of 36 ft/sec. At what Ilngle should the shot leave each athlete's
hand in order to get the maximum horjzontal range? Assume the shot
leaves from heights of 6 ft and 5 ft respectively. How much farther will the
taller athlete's throw go?
n. Answer the questions in problem 6, excepting (f), if a = 0, i.e., if the pro-
jectile is fired horizontally from a distance Yo ft above the horizontal.
12. A projectile is fired with a velocity 110 at an angle a with the horizontal. The
terrain makes an angle fJ with the horizontal.
(a) Find the range of the projectile. Hint. Call R the range of the projectile.
Then the projectile will hit the terrain when x = R cos fJ a'1d y =
R sin fJ. Substitute in (h) of Example 34.3.
(b) Find the value of a which will make the range a maximum. Hint. Make
use of double angle formulas.
(c) What is the maximum range?
13. In problem 3, we gave two angles at which a projectile could be fired in
order to hit an object located within range and on the same level as the firing
weapon.
(a) Solve this same problem if the object to be hit is on the top of a hill
whose angle of inclination is fJ and whose distance D from the firing point
is less than or equal to the range 1I02/g(1 + sinfJ), as given in (c) of
problem 12. Hint. Call (X, Y) the coordinates of the object. Then
D = VX2 + Y2, sinfJ = Y/D, cosfJ = X/D. Replace R by D in
answer to 12(a). Solve for a. Use the fact that
cos A sin B = ![sin (A + B) - sin (A - B)]
and
sin (2a - fJ) = cos (2a - fJ - i) .
(b) Show, by means of the solution found in (a), that it is possible to hit an
object only if its X and Y coordinates satisfy the inequality

,Hint. Use the fact-see answer to (a)-that

468 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

must be 1. Note that if Y = 0, so that object is on a level terrain,

X vo 2 /g as we saw previously.
14. A man is hunting with a gun whose muzzle velocity is 224 ft/sec. He aims
for a bird on the top of a tree 150 ft high and 1500 ft away. Is the bird in
danger of being hit?

In problems 1-14, we ignored air resistance. In the problems below, we

shall assume the projectile is fired from the earth and is subject not only
to a gravitational force but also to an air resistance which is proportional
to the first power of the velocity. We shall also assume that the force of
the air resistance acts in a direction opposite to that of the velocity, i.e.,
that it acts along a tangent to the projectile's path and in a direction to
oppose the motion. Call R the proportionality factor of the air resistance.
15. A projectile is fired on a level terrain at an angle a with the horizontal and
with a velocity of Vo ft/sec.
(a) Find the parametric equations of the particle's path. Hint. Modify
equation (a) of Example 34.3 to take into account the components of
the force of the air resistance in the x and y directions.
(b) What is the maximum height reached by the particle? Hint. Set
dy dy/dt
dx = dx/dt equal to zero.
16. A projectile is fired in a horizontal direction with a velocity of Vo ft/sec
from a height of Yo ft. Find the parametric equations of its path.
17. An anti-aircraft gun fires a shell almost vertically with an initial velocity of
Vo ft/sec. The horizontal component of the air resistance is therefore negligi-
ble. Assume the gun makes an angle a with the horizontal, and the vertical
component of resistance is R dy/dt.
(a) Find the parametric equations of the path of the shell.
(b) Assume the shell weighs 601b, the muzzle velocity is 2000 ft/sec, the
angle of elevation is 80°, and the vertical component of the air resistance
is 1/20 dy/dt. Find the parametric equations of the shell's path, the maxi-
mum height attained by it, and the time required to reach this maximum
height.

ANSWERS 34B
1. y = x - x 2 /80,000,
20,000 ft, 80,000 ft.
'//" \ 1 Dg
3. a = - :%: - Arc cos - 2 .
4 2 v0
4. (a) No. The maximum range is 3.7879 mi. (b) (i::l:: 0.36) rad.
(c) 1693 ft or 8307 ft, approx. (d) 20.6 sec or 45.6 sec, approx.
(e) Yes. When x = 4000 ft, the height y of the projectile is approximately
6498 ft, if the larger angle of elevation is used.
5. (a) a = Arc sin (i). (b) y = 64 ft. (c) x = 143 ft.
2
6. (a) x = (vo cos a)t, y = Yo + (vo sin a)t - i
2
g sec a 2
y Yo+ (tana)x - - 2 2 x.
Vo
Lesson MB-Answers

(b) Range is given by the positive value of x for which

2
g sec a x 2 _ (tan a)x - Yo = O.
2v02
2 • 2
( c) Y = Yo + vo sm a
2g·
(d) t = x/(vo cos a), where x is given by (b),
(e) tan 8 = vosina - gt,
Vo cos a
Ivl = Vvo 2 - 2vo(sin a)gt + (gt)2, where t is given by (d).
(f) sin a = vo/V2(v02 gyo). +
2
7. (a) y = 50 +x
128
. - (b) x = 166.4 ft. (c) 82 ft.
(d) t = 3.7 sec. (e) 122°00', 85.6 ft/sec.
(f) sin a = 0.6, a = 37°, approx., 171 ft, approx.
8. (a) v = 80 Y5ft/sec. (b) 1039 ft, approx. (a = 44°, approx.).
9. (a) 40y2 ft/sec. (b) 134 ft (a = 37°, approx.).
10. Use the answer given in 6(f) to find the angle a for each athlete. Then use
the range equation as given in 6(b).
2
gt g 2
ll. (a) x = vot , Y = Yo - - 2 , y = YO - -2v02· x
(b) x = voV2Yo/g ft.
(c) y = Yo ft.
(d) t = x/vo, where x is given by (b).
(e) tan 8 = -gt/vo, Ivl = Vvo 2 (gt)2. +
2
R R _ 2vo cos a sin (a - fJ)
12. (a ) ange - g cos2 fJ
2
fJ.
(b) a = "4 + 2· (c) Maxlmum range R = g(1 +vosin fJ)
13. (a) a = - fJ ± +- t Arc cos ( gX
2
+Yvo ) .
2

4 2 vo 2VX2 y2+
14. No. See 13(b).
m
IS. (a ) x = R vo cos a ( 1 - e-RI'", ),
m
Y = R (mg
Ii + vosma
. ) (1 - e-RI'",) - mg t,
Ii
where R is the proportionality factor of the air resistance.

(b) t = i log (1 + :: sin a) ,

2
Y = -mvo sm
. a - -m g log ( ]
R R2
+ -RVo.)
mg
sm a .

Yo + R2
2
mvo (1
16. X -- R - e -RI'",) ,Y = m g (1 - e-RI'",) - mg t .
Ii
17. (a) x = (vo cos a)t, y = as in 15(a).
(b) x = 347.3t, Y = 118,861(1 - e-O.02671) - 1200t, 30,153 ft, 36.4 sec.
470 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

LESSON 34C. Definition of a Central Force. Properties of the

Motion of a Particle Subject to a Central Force. Assume a particle
in motion is attracted to a fixed point 0, by a force F. In such cases we
say the particle moves subject to a central force, and call the fixed
point 0 to which the particle is attracted, the center of attraction.
In Example 28.15, we discussed a special central force problem where
the particle moved on a line. In the remainder of this lesson we consider
the motion of a particle subject to a central force where the particle is
free to move in space. We prove below certain properties which are com-
mon to the motions of all particles subject to a central force.
Property A. A Particle in Motion Subject to a Central Force
Moves in a Plane Which Contains the Fixed Point O. We assume a
particle, moving in space, is subject to a central force F. Let the fixed
point 0, toward which the force is directed, be the origin of a coordinate
system. Let P(x,y,z) be the rectangular coordinates of the particle, let r
be the distance of the particle from 0 at time t and let a, fJ, 'Y be the direc-
tion angles of the force F, Fig. 34.4.

Fx=
FC08 IX

x
(a) (b)

Figure 34.4

The components of F in the x,y,z directions are respectively, see Figs.

34.4(a) and (b),
x y
(34.41) F:r; = Fcosa= F-, F" = Fcos fJ = F-,
r r r
Since F", = mi, F" = my, F. = Fi, we by (34.41), the system

(34.42) mx = F::", my = F'It, mZ = F .

r r r
Les80nMC PARTICLE SUBJECT TO A CENTRAL FORCE 471

Multiply the first equation in (34.42) by -y, the second by x and add the
two. There results
(34.43) mxf} - myx = 0, xii - yx = o.
In an analogous manner, we can obtain, respectively, from the second
and third equations in (34.42), and from its first and third equations,

(34.44) yi - zfi = 0, zx - xi = o.
Integrating each of the equations in (34.43) and (34.44) with respect to
time, we obtain

(34.45) xii - yi; = Cl, zi; - XZ = Ca.

[Verify that the derivative of each equation in (34.45) gives the respective
equation in (34.43) or (34.44).] We now multiply the first equation in
(34.45) by z, the second by x, the third by y, and add all three. There
results

(34.46)

which is the equation of a plane through the origin, i.e., through the fixed
point O.
Property B. A Particle in Motion Subject to a Central Force
Satisfies the Law of the Conservation of Angular MOIDentulD. We
assume that a particle in motion is subject to a central force F. By prop-
erty A, the particle moves in a plane. By the definition of a central force,
F is always directed toward a fixed point 0, which we take as the origin
of a polar coordinate system. Let P(r,8) be the coordinates of the par-
ticle's position at time t. Call Fr the component of F acting along the
radial axis r, and F, the component of F acting in a direction perpendicular
to r. Since F always acts toward 0 along a radius vector, its component
F, is zero. Therefore, by (34.21),

(a) F, = ma, = m(2f8 + rll) = 0,

2f8 + rll = o.
If we multiply the last equation in (a) by r, it becomes 2ri8 + r211 = 0,
which is equivalent to
(b)

Integration of (b) gives

(34.47)
472 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

where h is a constant. By definition, tbe angular InOInentUIn of a

particle of mass m rotating about an axis perpendicular to the plane of
its motion is mr2f), where r is the distance of the particle from the axis
of rotation and is its angular velocity about this axis. We see, therefore,
by (34.47), that h is the angular momentum of a particle per unit mass.
And since h is a constant, the equation tells us that the angular momentum
of the particle is conserved. We have thus proved that a particle in motion
subject to a central force satisfies the law of the conservation of angular
InOInentum.
Property C. A Particle in Motion Subject to a Central Force
Sweeps out Equal Areas in Equal Intervals of Time. (NOTE. This
property is essentially a restatement of property B.) The area of a circular

Figure 34.48

sector of radius r and central angle 8 is r2 8/2, Fig. 34.48. Hence when a
radius vector r turns through an infinitesimal angle d8, it sweeps out an
area equal to
(34.481)

Substituting (34.47) in the second equation of (34.481), we obtain, with

initial conditions A = 0, t = 0,

(34.49)
l
A-O
A
dA = it
t_O 2
h
-dt, A
ht
=-.
2
In words the first equation in (34.49) says that the rate of change of the
area A is a constant. (Remember h is a constant.) The last equation
says that equal areas are swept out in equal times. We have thus proved
that a particle in motion subject to a central force sweeps out equal areas
in equal time intervals.

EXERCISE MC
1. (a) By (34.47), = h, where (r,8) are the polar coordinates of a particle
moving subject to a central force. Show that in rectangular coordinates
h = xi; - yi. Hint. tan 8 = y/x. Differentiate with respect to time
and solve for
Lesson 34D FORCE FIELD. POTENTIAL. CONSERVATIVE FIELD 473

(b) Hence show that the areal velocity dA/dt, in rectangular coordinates, is
given by dA/dt = !(xy - yx). Hint. See (34.49).

LESSON 34D. Definitions of Force Field, Potential, Conservative

Field. Conservation of Energy in a Conservative Field. Assume a
force F acts on a unit mass placed at each point (x,y,z) of a region of
space. Hence at each point of the region, we can represent the magnitude
and direction of F by drawing a vector F there. We call F a vector point-
function throughout this region, since it is a vector whose components
along the X, Y, Z axes are functions of the space coordinates X,y,z of
the unit mass. A region of this type is an example of a force field. Its
formal definition follows.

Definition 34.5. A region in space, having the property that at every

one of its points a vector point-function F exists that gives the magnitude
and direction of the force acting on a particle of unit mass placed there,
is called a field of force or a force field.

A region in the neighborhood of the solar system is a field of force. At

each point of the region, a vector point-function exists due to the mem-
bers of the solar system. The region in the neighborhood of a current
bearing wire is a force field. It is called an electromagnetic field.

By Definition 9.23, F", dx + F II dy+ F z dz is called an exact differential

if there exists a function U(x,y,z) such that

(34.51) dU = F",dx + Flldy + Fz dz,

or equivalently, such that

(34.52)
au
-=F""
au
-= F z.
ax az
Definition 34.53. A force field or a field of force is called conserva-
tive if there exists a function U(x,y,z) such that its differential dU satis-
fies (34.51), or equivalently if its partial derivatives with respect to x,y,z
respectively satisfy (34.52), where F"" F y , and Fz are the x,y,z com-
ponents of a force F acting in the field. The function U(x,y,z) itself is
called a force function and the negative of U(x,y,z) is called the potential
or the potential energy of the force field.

Comment 34.54. Not every force field has a potential -U(x,y,z).

Comment 34.6. The potential (- U) may be looked upon as a func-

tion whose partial derivatives with respect to x,y,z give respectively the
components of a force F in the negative x, y, and z directions.
474 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

Example 34.61. A particle moving in a force field is subject to a cen-

tral force F whose magnitude is proportional to its distance r from a fixed
point O. Show that the force field is conservative.
Solution. By hypothesis

where k > 0 is a proportionality constant. The magnitude of the com-

ponents of F in the x,y,z directions are given in (34.41). Hence substitut-
ing (a) in (34.41), we obtain

(b) Fit; = ; (-kr) = -kx, FII = (-kr) -ky,

F. = r (-kr) = -kz.

Let us take for the force function U of Definition 34.53,

2
kr k
(c) U(x,y,z) - 2"" + C == - 2 (x
2
+ y 2 + z2) + C.
Therefore

(d)
au k
- 2 (2x) -kx,
au -ky, au _ -kz.
ax =
ay Tz-
A comparison of (d) with (b) shows that the values on the right of the
equations in (d) are respectively Fit;, F II , F •. Hence by Definition 34.53,
the force field is conservative.
Example 34.62. A particle moving in a force field is subject to a cen-
tral force F whose magnitude is inversely proportional to the square of its
distance r from a fixed point O. Show that the force field is conservative.
Solution. By hypothesis
k
(a) F= --,
r2

where k > 0 is a proportionality constant. Substituting this value of F

in (34.41), we obtain

= r (_ = _ ky kz
(b) F
It; r2
kx ,
r3 F II = --,
r3 F. = - r3 •
Let us take for the force function U of Definition 34.53,
k
(c) U(x,y,z) = -
r
+ C == vx2 + ky2 + Z2 + C.
Lesson MD-Exercise 475

Therefore
(d) au = _ kx = _ kx, au = _ ky, au kz
ax V (X2 + y2 + Z2)3 r 3 ay r 3 iii =- r3 .

A comparison of (d) with (b) shows that the values on the right of the
equations in (d) are respectively Fz , F", F•. Hence by Definition 34.53,
the force field is conservative.
Property D. Conservation of Energy in a Conservative Field.
Let F be a force acting on a particle moving in a conservative field. There-
fore by Definition 34.53, there exists a function U(x,y,z) such that
au au
(34.63) dU = -dx+-dy
ax ay
+ au
-dz
az'
where
au au
ax = Fz , iii = F •.
Since F z = mX, F" = my, F. = mz, we have
au .. au au
(34.64) -=mx ay = my, az
mz.
ax '
Multiplying the first equation in (34.64) by dx, the second by dy, the
third by dz, and adding all three, we obtain, with the help of (34.63),

(34.65) mx dx + my dy + mz dz = dx + dy + dz = dU.

Integration of (34.65) with respect to time gives

+ m% + m;
.2 .2 .2
(34.66) m; = u + c.
The left side of (34.66) is defined as the kinetic energy of a particle.
By Definition 34.53, -U is its potential energy. Hence (34.66) tells us
that the sum of the kinetic and potential energies of a particle in a con-
servative field is a constant. This fact, namely that the sum of kinetic
and potential energies of a particle is a constant, is known as the law of
the conservation of energy. We have thus proved the law of the
conservation of energy for a particle moving in a conservative field.

EXERCISE 34D
1. A particle moving in a force field is subject to a central force F whose magni-
tude is proportional to its distance r2 from a fixed point O. Show that the
force field is conservative.
476 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

2. A particle moving in a force field is subject to a central force F whose mag-

nitude is proportional to its distance r3 from a fixed point O. Show that the
force field is conservative.
3. A particle moving in a force field is subject to a central force F whose mag-
nitude is proportional to its distance rn from a fixed point 0, where n is a
positive number. Show that the force field is conservative.
4. A particle moving in a force field is subject to a central force F whose mag-
nitude is inversely proportional to its distance r from a fixed point O. Show
that the force field is conservative.
5. A particle moving in a force field is subject to a central force F whose mag-
nitude is inversely proportional to its distance r" from a fixed point 0, where
n is a positive number greater than 1. Show that the force field is conservative.
6. Can you think of a force field which is not conservative?

ANSWERS 34D

1. Force function U = -"3k r 3 + C.

k 4
2. Force function U = - 4r + C.
_ _k_ n+1+ C
3. Force function U n+
1r .
k 2
4. Force function U = - 2 logr .
k
5. Force function U =
(n - l)r" 1
6. A field in which energy is being dissipated as the particle moves. For example,
a field in which a resisting force, proportional to velocity, is present cannot
be a conservative field.

LESSON 34E. Path of a Particle in Motion Subject to a Central

Force Whose Magnitude Is Proportional to Its Distance from a
Fixed Point O. We assume that a particle in motion of mass m is sub-
ject to a central force F whose magnitude is proportional to its distance r
from a fixed point O. We already know many facts about the particle.
By properties A, B, C of Lesson 34C, we know that it moves in a plane,
that it satisfies the law of the conservation of angular momentum and
that it sweeps out equal areas in equal times. By Definition 34.5, the
region in which the particle moves is a force field. By Example 34.61, this
field is conservative. Hence by property D following Example 34.62, we
also know that the particle satisfies the law of the conservation of energy.
To find the equation of its path, we take the x,y axes in the plane of
the particle's motion with the origin at the fixed point 0 toward which
the force acts. Let P(x,y) be the coordinates of the position of the particle
at time t in a rectangular system and (r,e) its coordinates in a polar sys-
tem. The components of F in the x and y directions are, by (34.42) (remem-
Lesson ME CENTRAL FORCE PROPORTIONAL TO DISTANCE 477

ber the force is central so that Fz = 0),

Fx =
(34.7) F" = mx=-, F II = my Fy.
r r
By hypothesis
(34.71) F= -k2mr,

where for convenience we have used k 2m for the proportionality constant.

The minus sign is necessary because the force is acting toward 0 and the
positive direction is outward from O. Substituting (34.71) in (34.7), we
obtain the linear system of equations
(34.72)

Their respective solutions, obtained by any of the methods previously dis-

cussed, are
(34.73) x = Cl cos kt + C2 sin kt, y = C3 cos kt + C4 sin kt.
These are the parametric equations of the path. You can verify by refer-
ring to Theorem 31.33 that the pair of functions in (34.73) contains the
correct number of four arbitrary constants. Hence in a specific problem
four initial conditions will be needed, x(O), x'(O), y(O), y'(O). The period
of the motion of the particle, by Definition 28.34, is 27r/k. It is the time
it takes the particle to return to its initial starting position, headed in the
same starting direction.
To find the equation of the path in rectangular coordinates, we must
eliminate the parameter t between the two equations in (34.73). The
easiest way to do this is to first solve them simultaneously for sin kt and
cos kt in terms of x and y. The result is
(34.74) sin kt = C3 X - ClY, cos kt = C4 X - C2Y, ClC4 - C2C3 r6 O.
C2C3 - ClC4 ClC4 - C2C3
Squaring both equations in (34.74) and then adding them, we obtain for
the path of the particle in rectangular coordinates,

(34.75)
1 = (C3X - Cly)2 + (C4X - C2y)2 ,
(ClC4 - C2C3)2
which can be written as
(34.76) (C3 2 + C4 2)X 2 - 2(ClC3 + C2C4)XY
+ (Cl 2 + C2 2)y2 - (ClC4 - C2C3)2 = 0, ClC4 - C2C3 r6 O.
From analytic geometry we know that if the constants in
(34.77) AX2 + 2Bxy + Cy2 + D = 0
478 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

are such that B2 - AC < 0, and

Case 1. D '¢ 0, AD < 0, then the equation represents an ellipse with
center at the origin.
Case 2. D '¢ 0, AD > 0, the equation has no locus.
Case 3. D = 0, the equation represents a single point.
A comparison of (34.76) with (34.77) shows that
(a) B2 - AC = (CIC3 + C2C4)2 - (C3
2
+ C42)(C1 2 + C2 2 )
= -(CIC4 - C2C3)2,
and
(b)

Both of the above expressions are less than zero if CIC4 - C2C3 '¢ 0.
Hence if, in (34.76), CIC" - C2C3, which corresponds to D of (34.77), is
not zero, Case 1 above applies and (34.76) is the equation of an ellipse
with center at the origin.
We have thus proved that the orbit of a particle, attracted to an origin
o by a central force F whose magnitude is proportional to its distance
from 0, is an ellipse with center at the fixed point O. Hence we have also
proved for this case that the force is directed toward the center of
the ellipse.
Comment 34.78. If the particle is constrained to move toward the
fixed point 0 along a radius vector so that 8 is constant, then d8/dt = 0,
F = F" and by (34.21) and (34.71),

This last equation, as we showed in Example 28.15, is, as it should be,

the differential equation of motion of a particle executing simple harmonic
motion.
Example 34.79. A particle weighing 16 pounds is 10 feet from a fixed
point 0 and is given an initial velocity of 15 ft/sec in a direction perpen-
dicular to the x axis. If a central force F acts on the particle with a mag-
nitude which is one-eighth of the distance of the particle from the fixed
point 0, find the equation of its path and the period of the motion.
Solution. We take the origin at the fixed point 0, and the x,y axes
in the plane of the particle's motion. By hypothesis
(a) F = -ir,
and by (34.42),
F", = mX = Fx, Fy
(b)
r
FII = my =-,
r
Lesson 34E CENTRAL FORCE PROPORTIONAL TO DISTANCE 479

where r is the distance of the particle from O. Substituting (a) in (b), we

obtain
X Y
(c) mx
h

= - S' my =
h

- S·
By hypothesis m = -H = ,. Hence (c) becomes

(d) x.. + 4--°' . + y.4-- °,

Y
whose solutions, by any method you wish to choose, are

(e ) x = Cl cos 2
t. t
+ C2 sm 2 ' Y -- C3 cos -2t + Col
.
sm -2t .
Therefore
dx Cl • t t
(f)
dt
- -sm-
2 2
+ -cos-,
C2
2 2
The initial conditions are t = 0, x = 10, Y = 0, dx/dt = 0, dy/dt = 15.
Substituting these values in (e) and (f), we obtain
(g) Cl = 10, C3 = 0,
Hence (e) becomes
t
(h) x = lOcos 2 ' y = 30sin i .
2
By eliminating the parameter t, we obtain
x2 y2
(i) 102 + 30 2 = 1,
which is the equation of an ellipse with center at the origin or fixed point O.
Its graph is shown in Fig. 34.791. The period of the motion, obtained
(0,30)

Vo = 15' !sec

Figure 34.791

from (h), is 411" seconds. It is the time it will take the particle to make a
complete circuit of the ellipse.
480 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

EXERCISE 34E
1. Verify the accuracy of the solution of (34.72) as given in (34.73).
2. Verify the accuracy of (34.74), (34.75), and (34.76).
3. A particle in motion of mass m is subject to a central force F whose magnitude
is proportional to its distance from a fixed point O. Initially it is Xo ft from
the origin and is given a velocity of vo ft/sec in a direction perpendicular to
the x axis.
(a) Find the parametric equations of its path; also the equation of its path
in rectangular coordinates. Take k for the proportionality constant in-
stead of Pm as we did in the text.
(b) For what relative values of Xo, vo will the path be a circle?
4. A body weighing 16 lb is attracted to a fixed point 0 by a force whose magni-
tude is one-eighth the distance of the particle from O. Initially it is 12 ft from
o and given a velocity of vo ft/sec in a direction perpendicular to the x axis.
(a) Find the parametric equations of motion; also the equation of motion in
rectangular coordinates.
(b) What initial velocity vo will make the eccentricity of the orbit t?
S. Solve problem 3(a), if initially the particle is Xo ft from the origin and is given
a velocity of vo ft/sec in a direction making an angle 8 with the x axis.
6. A body weighing 16 Ib is 12 ft from a fixed point O. It is given an initial
velocity of 20 ft/sec in a direction making an angle of 45 0 with the x axis. A
central force acts on the particle with a magnitude equal to one-eighteenth
of the distance of the body from O. Find the equation of its path and the
period of its motion. Verify that the equation satisfies Case 1 after (34.77)
and is therefore the equation of an ellipse.
7. A particle in motion of mass m is repelled from a fixed point 0 with a force
proportional to its distance frum O. Initially it is Xo ft from the origin and is
given a velocity of vo ft/sec in a direction perpendicular to the x axis. (a) Find
the equation of motion. (b) What type conic is it? (c) Show that properties
A, B, C, D of Lessons 34C and D are also valid when the central force is
repelling instead of attracting. '
8. Set up the system of differential equations of motion for the particle of prob-
lem 3 if in addition there is a force of resistance proportional to the velocity.

ANSWERS ME
2 k 2
3. (a) x = xocosVk/mt,y = 1.
Xo mvo
2 2
4. (a) x = 12 cos tt, y 2vo sin it, t44 + 1.
(b) Vo = 3y'3 ft/sec, or 4y'3 ft/sec.
S. x = xocoBVk/mt+voVm/kcos8sinVk/mt,
y = vovm/ksin8sinVk/mt;
(sin 8)x2 -
2
(2 sin 8 cos lJ)xy + (kx
mvo
o: + cos 2 8) l - x0 2 sin 8
2
O.
+
6. x = 12 cos it 30y'2 sin it, y = 30y'2 sin it; 61r sec;
x 2 - 2xy +
1.08y 2 - 144 = o.
Lesson 34F FORCE INVERSELY PROPORTIONAL TO SQUARE OF DISTANCE 481

7. (a) x = xo cosh vk/m t, y = IIOV m/k sinh vk/m t,

x
2
kl
x02 - mll02 = 1.
(b) Hyperbola.
8. mx = -kx - r:i:, my = -ky - ril, where k and r are the proportionality
constants respectively for the force and the resistance.

LESSON 34F. Path of a Particle in Motion Subject to a Central

Force Whose Magnitude Is Inversely Proportional to the Square
of Its Distance from a Fixed Point O. We assume that a particle in
motion, of mass m, is subject to a central force F whose magnitude is
inversely proportional to the square of its distance r from a fixed point O.
By properties A, B, C of Lesson 34C, we know that:
1. The particle moves in a plane.
2. It satisfies the law of the conservation of angular momentum.
3. It sweeps out equal areas in equal times.
By Definition 34.5, the region in which the particle moves is a force field.
Hence, by Example 34.62 and property D following it, we also know that:
4. This field is conservative and the particle therefore satisfies the law of
the conservation of energy.
To find the equation of the path of the particle, we take the x,y axes in
the plane of the particle's motion and the origin at the fixed point 0
toward which the force F is directed. Let P(x,y) be the coordinates of the
position of the particle at time t in a rectangular system and (r,8) its
coordinates in a polar system. By (34.42),

= mx = Fx, .. Fy
(34.8) F",
r
F1/= my =-.
r
By hypothesis
Km
(34.81) F= - --;:-2 ,
where for convenience we have taken Km, K > 0, for the proportionality
constant. The minus sign is necessary because the force acts toward 0
and the positive direction is outward from O. Substituting (34.81) in
(34.8), we obtain the system of equations
.. K =
(34.82) X= --x
3 Y.. -
Ky
r3 .
r '

If in (34.82), we substitute for r its equal vxl! +

y2, the resulting
equations form a nonlinear system which is difficult to solve. It turns
out that the path of the particle can be found more easily by using polar
coordinates. Call Fr the component of F in the radial r direction and F,
482 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

its component in a direction perpendicular to Fr. Then by (34.21),

(34.83) Fr = mar = m(i' - rB2), F, = ma, = m(2i'B + riJ).
Since F is a central force, F, = O. Setting the second equation in (34.83)
equal to zero, and then multiplying it by rim, we find 2ri'8 r 2iJ = 0, +
which is equivalent to

(34.84)

where h is the same constant we introduced in (34.47), i.e., h is the angu-

lar momentum of the particle per unit mass. In the first equation of
(34.83), substitute for Fr its value as given in (34.81) (remember here
Fr == F since the force acts only along r) and for B its value hlr2 as given
in (34.84). We thus obtain
2
Km (.. h ) .. h2 K
(34.85) - 'T2 = m r - r r4 '. r - r3 =-;:2'

Although methods of solving the nonlinear equation (34.85) are given in

both Lessons 35A and 35C which follow-see also Exercise 35,1l-use of
either of these methods will give a solution of t as a function of r. It turns
out to be easier to analyze the path of the particle if we solve the second
equation in (34.85) for r as a function of 8. To accomplish this end, we
use the substitution
1 1
(34.86) U =-, r =-.
r U

[Note that u as defined in (34.86) is the force function U of Example

34.62.] Substituting (34.86) in the l.ast equation of (34.84), we obtain
(34.87) B = hu 2 •
Two differentiations of the second equation in (34.86) give, with the help
of (34.87),
(34.88) i' = _ 2 du = _ 2 du d8 = _ 2 du hu 2 = -h du ,
u dt u d8 dt u d8 d8
2 2 2
i' = -h d u B = d u 2 _h2u2 d u .
d82 -h d82 hu = d82
Substituting the last value of i' of (34.88) and the value of r of (34.86) in
the second equation of (34.85), we obtain
2
(34.89) _h2u2 d u _ h 2u 3 = -Ku 2,
d8 2
which simplifies to the linear equation
d 2u K
(34.891) d8 2 +u = h2 ' h"¢ O.
Lesson 34F FORCE INVERSELY PROPORTIONAL TO SQUARE OF DISTANCE 483

Its solution, by any method you wish to choose, is

(34.892) U = hK2 + C cos (8 - 80 ), h ¢ 0,

where c and 80 are arbitrary constants. Since K, h, and c are constants,

we can write (34.892) in a more useful form by replacing c by a new con-
stant Kelh 2 • We thus obtain
K
(34.893) U = h 2 [1 + e cos (8 - 80 )], h ¢ O.

By (34.86), U = 1/r. If we now make this substitution in (34.893) and

choose our axes so that 80 = 0, the equation simplifies to
h2
(34.894) r = K(l + e cos 8)' h ¢ 0,
which is the equation, in polar coordinates, of the path of a particle
moving subject to a central force whose magnitude varies inversely as the
square of its distance from a fixed point O.

Semi·focal
h2
width=lE
d

Polar axis V(perigee)

·a
Conic is
section

Figure 34.895

We digress momentarily to review for you the proof that (34.894) is

the equation of a conic section whose eccentricity is e, whose semifocal
width is h 2 1K and which has one focus at the origin. In Fig. 34.895, we
have illustrated such a conic. The point on a particle's path that is nearest
the point 0 to which the particle is attracted is called the perigee of the
path. By definition, the ratio rid for every point P on a conic is equal to
its eccentrioity e. Hence, for the two points P and PIon the conic we
have respectively,
r r h2
(a) e = d and d = e
and d l = eK' h ¢ O.
484 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

From the figure, we see that

(b) d 1 = d + r cos 8.

In (b), replace d and d 1 by their values as given in the last two equations
of (a). There results
h2 r h2
(c) -=-+rcos8 r = K(1 + e cos 8)' h 0,
eK e '

which is the same as (34.894). Additionally, we know from analytic

geometry that if
(d) e <
1, the conic is an ellipse,
e = 1, the conic is a parabola,
e > 1, the conic is an hyperbola.

If e = 0, (34.894) becomes r = h 2 1K, which, in polar coordinates, is the

equation of a circle whose radius is h 21K.
Comment 34.896. We have thus proved that, if h 0, the orbit of
a particle, attracted to a fixed point 0 by a central force which satisfies
the inverse square law (34.81), is a conic section with one focus at the
fixed point O. If e < 1, the conic section is an ellipse. Since one focus is
at the fixed point 0, which we also took to be the origin of our coordinate
system, 0 cannot be the center of the ellipse. If therefore a particle, sub-
ject to a force which satisfies the inverse square law (34.81), moves in an
elliptical orbit, the central force is directed toward a focus of the ellipse
and not toward its center. Contrast this result with that obtained in
Lesson 34E. We found there that if F varies directly as the distance r
from 0, then the force is directed toward the center of the ellipse.
Comment 34.8961. If h = 0, then by (34.84), 9 = O. Therefore
8 is a constant. Hence the particle must move on a line. By (34.85),
with h = 0, the differential equation of motion simplifies to
_ K
r= - -
r2
.

For possible methods of solving it, see Lesson 35, also Exercise 35,7
and 22.
Comment 34.897. Determining the Constants of Integration h,
e, 90 of (34.893). The path of the particle, by (34.893), with u replaced
by its equal l/r, is
K
(a) r1 = h2 [1 + e cos (8 - 80 )],
Lesson 34F FORCE INVERSELY PROPORTIONAL TO SQUARE OF DISTANCE 485

where h, e, and 80 are constants which were introduced by integrations.

Assume that when the particle is at the point Po of its path, its distance
from 0 is ro and that it is moving with a velocity Vo in a direction making

Figure 34.898

an angle A with the line joining 0 to Po, Fig. 34.898. For convenience,
we measure subsequent values of the angle 8 from this line OP o. The
initial conditions are, therefore, [for vr , v, values see Fig. 34.898 and (34.2)1
(b) r = ro, 8 = 0, v = Vo, A = angle shown in Fig. 34.898,
Vr = Vo cos A = To, v, = Vo sin A =
The substitution of (b) in (a) gives

(c)
1
-
ro
= hK2 [1 + e cos 801, e cos 80 =
h2
roK - 1.

Let a be the angle measured from the radius vector to a tangent to a

curve at P(r,8), Fig. 34.899. Let v be the velocity of the particle at P.

o
Figure 34.899

Therefore the components of v in the radial direction and in a direction

perpendicular to it are respectively
(d) Vr = V cos a, v, = v sin a.
486 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

By (34.2), v, = l' and VB = r9. Hence (d) becomes

(e) r = V cos a, r9 = v sin a.
By (34.84), r9 = hlr. Therefore the second equation in (e) can be written as
(f) h = TV sin a.
Hence when r = ro, v = Vo and a = A, we obtain, by (f) and (b),
(g)

Differentiation of (a) with respect to 8 gives

1 dr Ke .
(h) - r2d8 = - h2"sm(8 - 80),
Also
Idr Idrdt
- r2d8 = - r2 dt d8
1 dr r2
- r2 dt h [by (34.84)]
1 dr
- Ii dt
1
- Ii, v cos a [by (e) above].

Hence, by equating this last expression with the right side of (h), we
obtain
(i) sin (8 - 80 ) = v cos a.

Inserting in (i) the initial conditions 8 = 0, v = Vo, a = A, Vo cos A =

TO, there results

(j) sin (-8 0 ) = Vo cos A, -esm

. 8
0 =
hvo
K cos
A = h •
KTO'

Squaring the second equations in (c) and (j) and adding them, we have

(k)
2
e = - r+
1 1'0r = - ;:; +1+ 2
cos A.

We can simplify (k) somewhat, by noting from (g) that

(1)

cos 2 A = 1 - r0 2v02'
Lesson 34F FOHCE INVERSELY PROPORTIONAL TO SQUARE OF DISTANCE 487

Substituting the last equation of (I) in (k), we find

2 h4 2h2 h2V02 h4
(m) e = r02K2 - roK 1 + + K2 - r02K2
2h2 h2V02
1 - roK + K2 .

By (g), we can determine h. By (k) or (m), we can then determine e;

remember K is a proportionality constant and not a constant of integra-
tion. With e and h known, we can determine 00 by the second equations
in (c) and (j). After 00 is known, we can then choose our axes to make
00 = 0, and thus obtain (34.894).
Comment 34.9. Energy Considerations Related to the Inverse
Square Law. In property D following Example 34.62, we showed that
the sum of the kinetic and potential energies of a particle moving in a
conservative field is a constant, see (34.66). Therefore, by (34.66),
(a) tmv 2 - U = E,
where we have replaced the constant c by the energy constant E, and the
velocity components, i,y,z, by v. In Example 34.62, we showed that if
F = -k/r2, then the force field is conservative and the potential energy
function - U = -k/r. In the example of this Lesson 34F, F = -Km/r2,
see (34.81). Hence the potential energy function - U = -Km/r. Sub-
stituting this value of U in (a), it becomes

(b) !mv 2 _ Km = E.
r

Inserting in (b), the initial values v = vo and r = ro, we obtain-

remember E is constant for all v and r-

(c) ! mv 0 2 - Km = E.
ro
By (m) of Comment 34.897,

(d)

Multiply (c) by -2h 2 /K 2m. There results·

(e) _ 2h2 E = _ 2h2 (! mv02 _ Km) = _ h2V02 + 2h2 .

K 2m K 2m ro K2 roK
Since the right sides of (d) and (e) are the same, we can equate their left
sides. We have thus shown that

(f) 1 -
2
e = - E = - ! ay E.
488 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

The mass m is a positive quantity; so also is (h/K)2. Hence we conclude

from (f), that if E < 0,
(g) 1 - e2 > 0, e < 1,

and the particle, therefore, must move in an elliptic orbit. If E = 0,

then 1 - e2 = 0, e = 1, and the particle must move in a parabolic
orbit; if E > 0, then e > 1 and the particle must move in a hyperbolic
orbit. We infer from all this that a particle in motion in a conservative
field, whose orbit is elliptic, must initially have had negative energy.
Conversely, if a particle with negative energy is projected into a conserva-
tive field, it will move in an elliptic orbit. Analogous remarks can be
made for the other two types of orbits.
Remark. A particle will have negative energy initially, if its kinetic
energy, which is the first term of (a), is less than the force function U =
Km/r.
Comment 34.91. Equation (34.894) gives the position r of a particle
as a function of 8, h 0. It would be desirable to express r as a function
of t so that we can know where the particle is at any moment. By (34.85)
f' = h 2/r3 - K/r2. As mentioned previously, two methods of solving
this equation will be given in Lesson 35. (Also see Exercise 35,11.) Un-
fortunately, use of either method gives t as a function of r. The problem
of solving the resulting equation for r as a function of t turns out to be
exceedingly difficult.

EXERCISE MF
1. Verify the accuracy of the solution' of (34.891) as given in (34.892).
2. A body weighing 16 lb is 12 ft from a fixed point O. It is given an initial
velocity of 6 ft/sec in a direction perpendicular to the x axis. Find its equation
of motion if it is subject to a central force whose magnitude is equal to 120/r 2 ,
where r is the distance of the particle from O. Hint. Follow the method of the
text. Initial conditions in rectangular coordinates are t = 0, x = xo = 12,
Y = yo = 0, to = 0, flo = Vo = 6. In polar coordinates initial conditions
are t = 0, r = ro = 12, 8 = 80 = 0, fo = 0, = = vo/ro = -h = t.
What is the eccentricity of the path?
3. A body weighing 16 lb is 10 ft from a fixed point O. It is subject to a central
force F whose magnitude is 100/r 2 , where r is the distance of the particle
from O. What initial velocity should be given the particle, in a direction
perpendicular to the x axis, in order that the particle may (a) move in an
elliptic orbit of eccentricity t, (b) move in a circular orbit (hint, orbit is
circular if e = 0), (c) move in a parabolic orbit, (d) move in a hyperbolic
orbit of eccentricity 2?
In the problems below, we shall consider the motion of a satellite of
the earth, where the satellite has been set in motion by being ejected from
a rocket. These problems are central force problems, obeying the inverse
Leason MF-Exercise

square law (34.81). The satellite will be attracted toward the center of
the earth with a force inversely proportional to the square of the distance
r of the satellite from this center. Call R the radius of the earth.
4. In (34.81), replace F by ma, where a is the acceleration of the particle.
(a) Show that
(M.911) K '"' gR2.
Hint. When r '"' R, a -g. =
(b) Show that the equation of motion (34.892) becomes
R2
(34.912) u = gh2 + C cos (8 - 80),

and that (34.894) becomes

2
h
(34.913) r == -g"::R""'2["'-I""",+"--e(';-cos--:8:---8:-07";)) ,
where
(34.914)
5. Assume in (34.913) that at t = 0, 8 = 0, that the last rocket is fired at a
distance r = ro from the center of the earth and that it has ejected the satel-
lite with a velocity Vo in a direction making an angle A with the radius vector
joining the rocket to the center of the earth. See Fig. 34.898. Show that the
constants h, e, and 80 in (34.913) are given respectively by
(34.915) h = r0280,

(34.916) 2 (r03802 _ 1)2 + (r0280 .. )2

e = gR2 gR2 TO ,

(34.917) cos 80 = 3802 -

!e ( r0gR2 1) , .
sm80 1 (r028
= -- ---
0"'0) ,
e gR2
where "'0 and r080 are the initial velocity components of Vo in the radial direc-
tion and in a direction perpendicular to the radial axis. Hint. With K = gR2
as given in (34.911), equation (34.913) is the same as (a) of Comment 34.897.
Make use of (g), (k), (c), and (j) of this comment.
6. Show that the orbit of a satellite of the earth will be a circle if
(34.918)

Hint. By (34.913), the orbit is circular if e = 0; by (34.916) e = 0 if (34.918)

holds.
7. A satellite is ejected by a rocket into a circular orbit 300 mi above the earth's
surface. Find its period of rotation. Hint. The period of the satellite, by
Definition 28.34, is T = 2r/w, where w = 8 is its angular velocity. There-
fore by (34.918), T = (2no/R)vTOTu. Here R = 4000 mi, ro = 4300 mi.
8. If the satellite is very close to the surface of the earth so that ro is very close
to R, then (34.918) can be written as
(34.919)
490 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

Show that the period in this case, for a circular orbit, is approximately 85 min.
See hint in problem 7 for period formula. [Compare with answer to Exercise
28A and B, 2(c).J
9. At a distance TO from the center of the earth, a rocket propels a satellite in 8.
direction perpendicular to the radius vector joining the rocket to the center.
The velocity of the satellite is vo. Hence to = 0, 80 = vo/ro and the angle A
in Fig. 34.898 is 90°.
(a) Show that if 80 2 < gR2/r03, then 80 = '/1'. Hint. In (34.916), to = O.
Solve for e and, since the eccentricity is always positive, choose the proper
sign to make e > 0. Substitute this value of e in (34.917).
(b) Show that the orbit of the satellite is then

Hint. Use (34.913), (34.915), and (34.916).

(c) Show that the satellite is farthest from the center of the earth, called
the apogee of the orbit, when r = ro, 8 = 0, i.e., the apogee is at the
point where the satellite is released. Hint. The distance r is largest when
the denominator in equation (b) above is smallest. The denominator is
smallest when the negative term in it is largest. This negative term is
largest when 8 = 0, cos 8 = 1. Solve for r with cos 8 = 1.
(d) Show that the satellite's perigee, i.e., the point of the satellite's orbit
nearest the center of the earth, occurs when 8 = '/1', and that its distance
from the center of the earth is then r04802/(2gR2 - r03802). See hint
in (c) above.
(e) Show that the satellite will make a complete orbit without hitting the
earth if 80 2 > 2gR3/[r0 3(ro + R)]. Hint. The perigee of the orbit, as
given in (d) above, must be greater than the radius R of the earth.
2
(f) Show that if 80 > gR2/r03, then 80 = 0. See hint in (a) of this problem.
Show that the orbit of the satellite is then

See hint in (b) above. Show that the perigee of the orbit occurs when
r = ro, 8 = 0, i.e., at the point where the satellite is released; that the
apogee of the orbit occurs when 8 = '/I' and that its distance from the
center of the earth is then r0480 2/(2gR2 - r03802), provided 2gR2 >
r03802. See hints in (c) and (d) above. Finally show that if 2gR2
r03802, the orbit will not have an apogee. Hint. The denominator of the
apogee's distance formula above will then be negative or zero.
If 2gR2 = r03802, then r0280 2 = 2gR2/ro. If the satellite is ejected
at or near the surface of the earth so that ro = R, then the last equation
becomes r080 = v'2iiR. But Vo = ro80 so that Vo = v'2iiR which is
the escape velocity of a body fired from the surface of the earth, see (i)
of Example (16.36).
Lesson MG PLANETARY MOTION 491

ANSWERS MF
2. h = rovo = 72, m = t, mk = 120, k = 240,
e = 0.8, r = 21.6/(1 +0.8 cos 8).
3. (a) 5.48 ft/sec. (b) 4.47 ft/sec. (c) 6.32 ft/sec. (d) 7.75 ft/sec.
7. 5700 sec, approximately, or 95 min. The first satellite put into orbit in 1957
by the U.S.S.R., known as the Sputnik, had a nearly circular orbit of 300 miles
above the earth's surface and a period of 96 min.

LESSON MG. Planetary Motion. Newton's law of universal

gravitation states that every two bodies in the universe attract each
other with a force proportional to the product of their masses and in-
versely proportional to the square of the distance separating them. Let
M be the mass of the sun and m the mass of a planet. It can be proved
that we do not commit a serious error if we consider the sun as fixed, its
mass M as concentrated at its center, the planet as a particle, and sun
and planet as isolated bodies. Then by Newton's law of universal gravita-
tion,
GMm
(34.92) F= - -r2 -,

where r is the distance of a planet from the sun's center, and G is a pro-
portionality constant called the gravitational constant. Replacing the
constant GM in (34.92) by a new constant K, it becomes F = -Km/r 2 ,
which is the same equation as (34.81) of Lesson 34F. Since this force F is
directed' toward a fixed point 0, namely the sun's center, it is a central
force. Hence planetary motion is exactly the same as the motion of the
particle discussed in Lesson 34F. We can therefore assert that:
1. A planet moves in a plane.
2. The orbit of a planet is a conic section whose equation, by (34.894), is

h2
r = K(l + e cos 6)' h"¢ 0,

where h is the angular m()mentum of the planet per unit mass, e is the
eccentricity of its orbit and K is the product of the gravitational con-
stant G and the mass M of the sun.
3. The planets satisfy the law of the conservation of angular momentum.
4. The planets sweep out equal areas in equal intervals of time.
5. The force field in which the planets move is conservative; hence the
planets satisfy the law of the conservation of energy.
6. The sun is s.t, one focus of the planet's orbit. [See Comment 34.896.)
Comment 34.93. The Orbits of the Earth and the Other Planets of Our
Solar System Are Ellipses with the Sun at One Focus. Hence for the planets
of our solar system e < 1. This means, as we showed in Comment 34.9,
that each planet, at the beginning of its existence, had negative energy.
492 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

The orbits of comets· which appear after long intervals of time are ex-
tremely elongated ellipses whose eccentricity is near 1, almost close to
parabolas. Those bodies for which e 1 have parabolic or hyperbolic
orbits. They leave the solar system and never return.

EXERCISE 34G
1. Find the equation of motion of a planet of mass m if its distance at perigee,
i.e., its distance nearest the sun, is ro and its velocity there is Vo. Hint. Take
the axis of the ellipse through the perigee. Then at t = 0, r = ro, 8 = 80 = 0,
r = 0, 8 = volro. See Fig. 34.898.
2. Find the approximate equation of Halley's comet. Hint. See footnote at the
bottom of page: e = 0.967, a - c = 0.587, where a is the semimajor axis
and c is the distance of the focus from the center of the elliptic orbit.
3. A comet at rest at an infinite distance away from the sun is attracted toward
the sun in accordance with the inverse square law. If its distance at perigee

Take axis so that 80

8 = 0, U = 1/ro.
°
is ro, find the equation of its path and show that its orbit is parabolic. Hint.
in (34.892). At t = 0,8 = 1r, U = 1/r = 0. When

ANSWERS 34G
2 2
1. r = ro Vo
K + (rov02 - K) cos 8
2 2
2. 31:.41 + 2%.54 = 1. Figures in astronomical units.

3. r = 1 :::s 8' Orbit is parabolic since e, the coefficient of cos 8, is one.

LESSON 34H. Kepler's (1571-1630) Laws of Planetary Motion.

Proof of Newton's Inverse Square Law. Kepler's three laws of plane-
tary motion are:
1. Each planet moves in an elliptical orbit with the sun at one focus.
2. The radius vector connecting sun and planet sweeps out equal areas in
equal times.
3. The square of the period of a planet is proportional to the cube of the
semimajor axis of its orbit.
We have already proved 1 and 2: see numbers 6 and 4 of Lesson 34G.
We shall now prove 3.
Proof of 3. By (34.894), the orbit of a planet is given by
h2
(a) r = K(1 + e cos 8) ,
*The famous Halley's comet has an elliptical orbit whose eccentricity is 0.967. Its
period is 76 years. Its perigee is 0.587 astronomical units (an astronomical unit is the
distance of the earth to the sun, approximately 92,900,000 mi). Since it last visited us
in 1910, it will be again visible in 1986.
 KEPLER'S LAWS. PROOF OF INVERSE SQUARE LAW 493

where one focus is at the origin of a coordinate system, called (0,0) in

Fig. 34.94, and the semifocal width is h2/K, called L in the figure. Let
(0,0) be the center of the ellipse. With respect to the center of the ellipse,
(0, b) (e,L)

1--------+----*':..=:..:!...---t(a,O)
(0,0)

Figure 34.94

let (c,O) be the coordinates of the focus, (a,O), (O,b), be the coordinates of
the ends of the semimajor and semiminor axes respectively. The equation
of the ellipse with respect to its center as an origin is, therefore,

(b)

where c2 = a 2 - b 2 • If x 2 = c2 = a2 - b 2 , then y2 = L2, and there-

fore, by (b),

(c)

The semifocal width L also equals h 2 /K. Substituting this value of L in

the last equation of (c), we obtain

(d)

The first equation in (34.49) holds for every particle subject to a cen-
tral force. It therefore holds for the planets. The last equation in (34.49)
resulted when we took for our initial conditions A = 0, t = 0. Hence if
A = 0, t = 0,
(e) A = ,ht,
gives the area A swept out by a planet in time t. Call T the period of a
planet's orbit, i.e., the time it takes a planet to make a complete circuit
of its orbit. When the planet has made a complete circuit, it has swept
out the area of the ellipse, namely 1rab. Hence when A = 7rab and t = T,
494 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

we obtain by (e)
T _ 21rab,
(f) - h

In (f), replace b2 by its value as given in (d). There results

2 2 2 3
2 41r a 2 ah 41r a
(34.95) T = Ji2'j(=Jr'
Since K(= GM) is a constant, and a is the semimajor axis of the ellipse,
(34.95) says that the square of the period of a planet is proportional to the
cube of the semimajor axis of its orbit.
Proof of Newton's Inverse Square Lawfrom Kepler's Laws. We
have proved Kepler's three laws from Newton's universal law of gravita-
tion. Historically, however, Kepler preceded Newton and hence the
former knew nothing of the law of gravitation. It is indeed remarkable
that Kepler was able to deduce his three laws from an intensive study of
the recordings of the positions of the planets made by direct observations.
It was Newton who used Kepler's laws as a hypothesis to develop his own
universal law of gravitation. Part of his problem was thus the inverse of
the one we solved. He assumed that a planet moves in an elliptical orbit
with the sun at one focus, that it sweeps out equal areas in equal times,
and then set out to prove that the planet must therefore be subject to a
central force directed toward a focus, whose magnitude varies inversely
as the square of the distance of the planet from the sun. The proof follows.
Proof. By Kepler's second law, dAjdt is a constant. Therefore by
the second equation in (34.481),
(a)

where c is a constant. Differentiation of the second equation in (a) and

multiplying the result by a constant mass m, gives
(b) m(2rd + Til) = o.
The second equation of (b), by (34.21), is ma, = O. But ma, is the com-
ponent of force acting on a particle in a direction perpendicular to the
radius vector. Since this component is zero, the force acting on a planet
must be a central one; i.e., the force always acts along a radius vector
toward or away from the sun.
By Kepler's first law, a planet moves in an elliptical orbit with the sun
at one focus. We showed in Lesson 34F, that the equation of an ellipse
in polar coordinates with one focus at the origin is
A
(c) T= ,
1+ecos8
Lesson 34H KEPLER'S LAWS. PROOF OF INVERSE SQUARE LAW 495

where e < 1 is its eccentricity and A is its semifocal width. Differentia-

tion of (c) gives
r = Ae sin 8 (J _ A2 e sin 8 (J
(d)
(1 +e cos 8)2 - (1 +
e cos 8)2 A .
Making use of (c) and (a), we can write (d) as
. _ 2esin8.£._ . 8
(e) r - r A r2 - A sm .

Differentiating (e) and then using (a), we obtain

2
(f) i' = (e cos 8)(J = ;r 2 (e cos 8).

Solving (c) for e cos 8, there results

A - r
(g) ecos 8 = ---.
r
Substituting this value in (f), we obtain
2 2 2
(h) c (A - r) c c
i' = Ar2 --r- = r3 - Ar2'

The component of a force in a radial direction is, by the first equation

in (34.21),
(i) F, = maT = m(i' - r(J2).
In (i), replace i' by its value in (h) and (J by its value in (a). There results

(j) F =
,
m[cr3
2
_
Ar2
_
r3
= _ mc
Ar2
2
•

We !lhowed above that the force acting on a planet is toward or away

from the sun. Since m, c 2 , and A are positive constants, (j) tells us that
the force acting on a planet is directed toward the sun, and its magnitude
is inversely proportional to the square of its distance from the sun.
Comment 34.951. The inverse square law just proved is only part of
the universal law of gravitation. Newton's studies of the gravitational
force of the earth plus his observations of the moon's orbit about the
earth, plus his own genius, enabled him to formulate his famous law of
universal gravitation as stated at the beginning of Lesson 34G.
Comment 34.96. In Lesson 34C, we proved that every particle sub-
ject to a central force obeys Kepler's second law, i.e., it sweeps out equal
areas in equal times. Hence Newton's inverse square law is only a suffi-
cient condition for Kepler's equal area law, not a necessary one. However,
the inverse square law is a necessary and sufficient condition for Kepler's
0&96 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

first law: the orbit of a particle subject to a central force is an ellipse

with the force directed toward a focus. Note that if F is proportional to
r as in Lesson 34E, the orbit is also elliptical, but the force is directed
toward the center of the ellipse.

EXERCISE MH
1. (a) Take 240,000 miles as the semimajor axis of the moon's orbit and its
period as 27.3 days. Use Kepler's thirq law to find the value of the
proportionality constant k for the earth, where k replaces 411"2/ K in
(34.95). Use for units 1000 mi and hour.
(b) In Exercise 34F,7, we found T = 95 min for the period of the circular
orbit of a satellite of the earth 300 mi above its surface. Take 4300 mi
for the semimajor axis of the satellite's orbit and calculate k. Use for
units 1000 mi and hour.
(c) In Exercise 34F,8, we found T = 85 min for the period of the orbit of a
satellite close to the earth's surface. Take 4000 mi for the semimajor axis
of the orbit and calculate k. Use for units 1000 mi and hour.
Compare results in (a), (b), and (c). Am. k = 0.031.
Remark. Next time you read of an earth satellite which has been success-
fully orbited, and its perigee and apogee are given, use this value of k and
Kepler's third law to calculate the period of the orbit and see if it agrees with
the observed period.
2. (a) Use Kepler's third law to calculate the value of the proportionality con-
stant k for the sun. Use for units 1,000,000 mi and day. Take the period
of the earth as 365 days and the semimajor axis of its orbit as 93,000,000
mi. An8. k = 0.165.
(b) Use this value of k to calculate the period of one of the other planets from
its known distance from the sun, or calculate its mean distance from the
sun from its known period.
By (34.891),
(a)

By (34.81) and (34.86),

Km F
(b) F= ---,
r2 K= m - mu 2 •

Hence (a) becomes

d 2 u, Fr2 F
(34.97) dB2 +u = - mh2 = - mh2u,2'

which is the differential equation, in polar coordinates, of the orbit of a par-

ticle subject to a central force. For different values of the force F, there
will be different orbits. Conversely, if the equation r = r(B) in polar
coordinates of the orbit of a particle is known, (34.97) will give the cen-
tral force F which causes the particle to move in this orbit. All one need
Exercise 34M MISCELLANEOUS PROBLEMS LEADING TO SYSTEMS 497

do to find F is to substitute in (34.97) the values of u and d 2uldB2 and

solve for F. (Remember u = l/r.)
Use the above facts and (34.97) to solve the following problems. Assume
in all cases that F is a central force and that the fixed point 0 toward
which the force acts is at the origin.
3. The orbit of a particle is an ellipse with one focus at the origin. Show that the
force F obeys the inverse square law. (NOTE. This assertion has already been
proved in this lesson; see proof of Newton's inverse square law from Kepler's
laws.) Hint. The equation of the orbit in polar coordinates is
A

Therefore

U =
1 + ecos8
A '
Substitute the last two values in (34.97). Solve for F. Remember that m, h,
and A are constants. Ana. F = -mh 2/ Ar2.
4. The orbit of a particle is a circle, with the origin a point of the circumference.
Show that the force F is inversely proportional to the fifth power of the
distance of the particle from the origin. Hint. The equation of the orbit in
polar coordinates is r = 2a cos 8, where a is the radius of the circle. There-
fore sec B = 2a/r = 2au. Also make use of the fact that sec 2 B = 1 + tan 2 B.
An8. F = -8a 2h 2m/r5.
5. The orbit of a particle is an ellipse with the origin at the center of the ellipse.
Show that the force F is proportional to the distance of the particle from the
center, see Lesson 34E. Hint. The equation of the orbit in polar coordinates is
2
r2 = 1 2 8' Follow suggestions in problem 3.
- e cos
Ana. F = -mh 2(1 - e2)r/ A 4.
6. The orbit of a particle is the spiral r = e'. Find the force F.
Ana. F = -2mh 2/r 3 •
7. The orbit of a particle is the lemniscate r2 = a2 cos 2B. Find the force F.
An8. F = -3mh 2a4/r7.
8. The orbit of a particle is the cardioid r = a(1 + cos B). Find the force F.
Ana. F = -3mh 2a/r4.
9. The orbit of a particle is a circle with center at the origin. Find the force F.
Hint. The equation of the orbit is r = a. Ana. F = -mh 2/a 3 •

EXERCISE 34M
MISCELLANEOUS TYPES OF I>ROBLEMS LEADING
TO SYSTEMS OF EQUATIONS
1. A particle moves in a plane. If the x and y components of its velocity are
equal respectively to the y and x coordinates of its position, find the equation
of its path.
2. (a) Solve problem 1, if the word velocity is changed to acceleration. (b) Find
the equation of its path if initially the particle is at the origin and has a
velocity of 15 ft/sec in a direction whose slope is f.
498 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

3. Solve the following system of differential equations. They are used in certain
problems of electron motion.

eE,

where m = the mass of the electron,

e = the charge of the electron,
H = the intensity of the magnetic field,
E = the intensity of the electric field.
Assume that initially the electron is at the origin and its velocity is zero.

In Lesson 30M-C, we discussed the problem of a wire twisted by ro-

tating a bob at one end, where the resulting torque or moment of force
was proportional to the angle of twist, see (30.63), (30.64), and (30.67).
For convenience we recopy (30.67).

d29
(34.971) I dt 2 = -k9,

where k is a proportionality constant, called the torsional stiffness

constant, and 9 is the angle through which the wire has twisted from
an equilibrium position. Make use of (34.971) to solve problems 4-6.
4. Three disks are connected by shafts. The moment of inertia of the two end
disks is I, of the middle disk 21, Fig. 34.98. The torsional stiffness constant
of each of the two shafts connecting the three disks is k. If a torque 2To sin wt

,"",8 2 81

Figure 34.98

is applied to the center disk, find the angular motion of the disks. Assume no
resistance and that initially the disks are at rest and the shafts are in their
untwisted equilibrium position. Hint. Call 91 the angular displacement from
equilibrium at time t of an end disk and 92 the angular displacement from
equilibrium of the middle disk. If the end disks are considered fixed at that
instant, then at time t, the shafts connecting them to the middle disk have
twisted through an angle, 92 - 91. Hence the restoring torque acting on the
middle disk is 2k(92 - 91). If the middle disk is considered as fixed, then the
shaft connecting it to an end disk has twisted through an angle, -(92 - 91).
Hence the restoring torque acting on an end disk is k(91 - 92). Now make
use of (34.971) taking into account the applied torque acting on the center
disk. The differential equations can be found in the answer section.
5. Three disks and a driving wheel are connected by shafts, Fig. 34.99. The right
end disk is free. Each disk has the same moment of inertia I, and the three
shafts have the same torsional stiffness constant, k. Set up the system of
Exercise 34M MISCELLANEOUS PROBLEMS LEADING TO SYSTEMS 499

differential equations for the angular motion 81 (t), 82(t) , 8a(t) of the three
disks from their equilibrium positions due to an angular motion 8 of the driv-
ing wheel. Assume that initially the disks are at rest and the shafts are in
their untwisted equilibrium position. (See hint given in problem 4.)

Figure 34.99

6. Solve problem 5 if there is also a resisting torque operating on the disks

proportional to the first power of their angular velocities. This proportionality
constant is called the torsional resistance constant. Assume the torsional
resistance constant for each disk is R.

ANSWERS 34M
1. x = cle' +
c2e-', Y = cle' - C2e-'j x 2 - y2 c.
2. (a) x = cle' c2e-'+ Ca cos t + C4 sin t, +
y = CIe' +
c2e-' - Ca cos t - C4 sin t.
(b) x = ¥(e' - e-') +! sin t, y = ¥(e' - e-') - i sin t.

3. x
Em (
= eH2 1 - cos m
He )
t , Y
E
= Ii t - eH2 sm
Em. eH
m t.
They are the parametric equations of a cycloid. For the definition of a cycloid,
see Exercise 28C,34.
2
4. 1 = -k(81 - 82), for each end disk,
2
21 = -2k(82 - 81) + 2To sin wt, for the middle disk.
Solutions are,
kTo sin wt
81 = CI + c2t + Ca
• . /iU7Y
smv 2k/1 t + C4
. /iU7Y
cos v 2k/1 t + Iw2(lw2 _ 2k)'
2
- w I) sin wt
82 = CI + c2t - Ca
• _ /iU7Y
sm v 2k/ It - C4
. /iU7Y
cos v 2k/ It + To(k
Iw2(lw2 _ 2k) .

Initial conditions are t = 0, 81 = 0, 82 = 0, 81 = 0, 82 = 0.

2
5. 1 = -k(81 - 8) - k(81 - 82),

2
1 d 82
dt 2
SOO PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

or
(I D2 + 2k)81 - k82 = k8(t) ,
-k81 + (l D2 + 2k)82 - k8a = 0,
-k82 + (I D2 + k)8a = O.
Initial conditions at t = 0 are 81 = 82 = 8a = 81 = 82 = 8a = O.
6. Add the term -R d8I/dt to the right side of the first equation of 5; -R d82/dt
to the right side of the second equation; -R d8a/dt to the right side of the
third equation.

LESSON 35. Special Types of Second Order Linear and Nonlinear

Differential Equations Solvable by Reduction to a
System of Two First Order Equations.

In previous lessons, we outlined standard methods by which solutions

in terms of elementary functions could be obtained for certain types of first
order differential equations and for linear differential equations with con-
stant coefficients of order n > 1. For the nonlinear differential equation
of order n > 1 and for linear equations un'th nonconstant coefficients of
order n > I, such standard methods are available only if the equation
belongs to one of several special kinds. In Lesson 23 and Exercise 23, 18
and 22, we discussed such special kinds of linear equations with noncon-
stant coefficients. In this lesson we discuss three special types of nonlinear
equations of order two for which a standard method of solution is avail-
able. (See also Exercise 35, 23 for an additional type.) These same
methods can, of course, also be used if the equation is linear.

LESSON 35A. Solution of a Second Order Nonlinear Differential

Equation in Which y' and the Independent Variable x Are Absent.
Equations of this type that we shall consider will be those which can be
written in the form
(35.1) y" = f(y),

where f(y) is defined on an interval I: a ;:;;; y ;:;;; b. Note that y' and x
are missing. The substitution u = y', u' = y" will change (35.1) into the
equivalent first order system
dy du
(35.11)
dx
u, dx = f(y)·

The second equation in (35.11) can be solved for u as follows. Multiply

it by 2u to obtain 2uu' = 2uf(y). Replace 2uu' by its equal (d/dx)(u 2 ) ,
and u by dy/dx. Hence
(35.12)
Lesson 35A y' AND INDEPENDENT VARIABLE X ABSENT 501

Therefore
(35.13) d(u 2) = 2f(y) dy.
Integration of (35.13) gives

(35.14) u 2 = 2 ff(y) dy = F(y) + Cl'

Substituting (35.14) in the first equation of (35.11), we have

(35.15) = ± v'F(y) + Cl'

If v'F(y) + Cl 0, we obtain from (35.15)

(35.16) ± f dy
v'F(y) + Cl
= fdX + C2·
Remark. We do not wish to imply that (35.16) will always be in-
tegrable in terms of elementary functions. In the example below, fey) has
been chosen carefully so that it will be. In general, it will not be.
Example 35.17. Solve the nonlinear equation
(a) y" = 4y-3, Y O.
Solution. Following the method outlined above, we substitute u = v',
u' = y" in (a) to obtain the equivalent first order system.

(b) = u, = 4y-3.

Multiplying the second equation in (b) by 2u, there results

(c) 2u du -
dx -
8
y
-3
u,
i..
dx u
( 2) _
-
8
Y
-3 dy
dx"

d(u 2) = 8y-3 dy, u2 = f 8y-3 dy = _4y-2 + CIt Cl > 0,

4
u = ±v'Cl - 4y-2 = ± v'Cly2 -
, Y > -2- or y < --2
-.
Y vc;- vc;-
Substituting this last value of u in the first equation of (b), we have

(d) _--;::::y::::d=.=y=== = dx.

±v'Cly2 - 4
Integration of (d) now gives
1
(e) ± Cl v' Cly2 - 4 = x + C2,
2 -2
Cly2 = (ClX + ClC2)2 + 4, y > . r:-' y < . r:-' Cl > 0.
VCI V CI
502 PRoBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

LESSON 35D. Solution of a Second Order Nonlinear Differential

Equation in Which the Dependent Variable y Is Absent. Equa-
tions of this type that we shall consider will be those which can be written
in the form
(35.2) y" = f(x,y').

Note that y is missing. The substitution u = y', u' = y" will change
(35.2) into the equivalent first order system
dy du
(35.21) dx = u, dx = f(x,u).

The second equation is now a first order equation in u and hence may be
solvable by the methods of Chapter 2. If it is and its solution is u =
u(x) + CI, then by the first equation in (35.21),

(35.22) y = f [u(x) + cil dx + C2·

Example 35.23. Solve the nonlinear equation
(a) y" = X(y,)2.

Solution. Following the method outlined above, we substitute u = y',

u' = y" in (a) to obtain the equivalent first order system
dy du 2
(b) -= u dx=XU.
dx '
If u 'F 0, we can write the second equation in (b) as
du
(c) 2
u
= xdx.

Its solution is u = -2/(x 2 + c), which we write as

(d)

with the understanding that the plus sign is to be used if C > OJ the
minus sign if C < O. Substituting (d) in the first equation of (b), we have
dy 2 -2dx
(e) -= -
dx x2 ± Cl 2
, dy = -X"""2-±-C-I""'2

By integrating (e), we obtain

y = -2 UI Arc tan + C2] and y = - log =;: + C2.

Lesson 35C INDEPENDENT VARIABLE X ABSENT 503

If Cl = 0, then by (e),
2 2
(f) dy = - 2dx,
x
y = -+C,
x
x O.

LESSON 35C. Solution of a Second Order Nonlinear Equation in

Which the Independent Variable x Is Absent. Equations of this
type that we shall consider will be those which can be written in the form

(35.3) y" = !(y,y').

Note that x is missing. If we let u = y', then

/I du du dy du
(35.31) y = dx = dy dx = dy u.

Substituting these values in (35.3) will change it into the equivalent first
order system
(35.32) dy = u du
u dy = !(y,u).
dx '
The second equation in (35.32) is now a first order equation in u and
hence may be solvable by the methods of Chapter 2. If it is and its solu-
tion is u = u(y) + CJ, then by the first equation in (35.32),

(35.33) f Cl = f dx = x + C2'
Example 35.34. Solve the nonlinear equation
(a) yy" = y3 + (y,)2.
Solution. Following the method outlined above, we substitute u = y',
y" = u :: in (a) to obtain the first order system

dy du
(b) dx = u, uy dy = Y
3
+ u 2.
The second equation in (b) can be written as

(c) du _ !y u = y 2u- 1, U 0, y 0,

which is a Bernoulli equation. Hence, following the method outlined in

Lesson llD, we multiply (c) by 2u to obtain
du 2 2
(d) 2u-
dy - -u
y = 2y 2 ,
504 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

which can be written as

(e)

an equation linear in u 2 • Its solution by the method of Lesson liB is

(the integrating factor is e!- 2dlllll = e- 210"" = 1/y2)

(f)
2
uy2 = f 2dy, u = ± YV2y + CI, U ¢ 0, y ¢ 0, 2y + CI > 0.

Substituting these values of u in the first equation of (b), we obtain

y
(g) = ± YV2y + CI ; _-;:::=.d:::== = ± dx, y ¢ 0, 2y + CI > 0.
YV2y + CI

The solution of the second equation in (g), if y ¢ 0, 2y + CI > 0, is

(h) _2_1
. r:- og
V2y +.CIr -
-_ ± x + C2, CI > 0,
VCl vy

.
2
Arc tan
V2y
.
+ CI = ±x + C2, CI < 0.
V -CI V -CI

If Cl = 0, then (f) becomes u = ±v'2 y3/2. Therefore by (b)

(i) :; = ±V2 y3l2, ± y-3/2 dy = V2 dx, T'2y-1/2 = V2 X + C2

(V2x - C2)2y = 4, y ¢ 0.
NOTE. The function y = °
which we had to discard in order to obtain
(h) and (i) also satisfies (a). It is a particular solution of (a) not obtain-
able from the families (h) or (i).

EXERCISE 35

Solve each of the following differential equations.

I. y" = 2yy'. 6. (y+ l)y" = 3(y')2.
2. y3 y" = k. 7. r = -k/r2.
3. yy" = (y')2 - 1. (See Comment 34.8961.)
4. x 2 y" +
XV' = 1. 8. g" = ik
y 2.
(Note that the equation is linear.) 9. y" = 2ky 3.
5. xy" - y' = x 2. 10. yy" + (y')2 - y' = O.
(Note that the equation is linear.)
2
h k
n. r = r3 - T2' (See Comment 34.91.)
12. yy" +(y')3 _ (y')2 = o.
13. yy" - 3(y'? = O.
14. (1 +
x 2)y" (y')2 +
1 = O. +
IS. (1 +
x 2 )y" 2x(y' + +
1) = O. (Note that the equation is linear.)
Lesson 35-Exercise 505

Find a particular solution of each of the following differential equations

satisfying the given initial conditions.
16. (y + 1)y" = 3(y')2, y(1) = 0, y'(1) = -!.
17. y" = y'ell , y(3) = 0, y'(3) = 1.
18. y" = 2yy', yeO) = 1, y'(O) = 2.
19. 2y" = ell, yeO) = 0, y'(O) = 1.
20. x 2y" +
xy' = 1, y(1) = 1, y'(1) = 2. (Note that the equation is linear.)
21. xy" - y' = x 2, y(1) = 0, y'(1) = -1. (Note that the equation is linear.)
22. r = - ,r(O) = 1, teO) = 0, r(O) = -!. See problem 7 above. Hint.
After finding t as a function of r in integral form, substitute either r = cos 2 u
or r = u 2 •
23. A differential equation is said to be homogeneous in x, if by arbitrarily
assigning degree n to xn and degree -n to y(n), each nonzero term of the
equation is of the same degree. In computing the degree of a term, the de-
gree of each of its members is added. For example, x 2 is of degree 2; y is of
degree zero; y" is of degree -2; x 2yy" is of degree zero. Equations homoge-
neous in x may be solvable by the substitutions, see also Exercise 23,18,
x = e'U, u = logx,
2 2y
d y=1-
- dY ),
(d- - -
dx 2 x 2 du 2 du
etc., and then making use of one of the methods of Lesson 35. Verify that
each of the following nonlinear differential equations is homogeneous in x
and solve.
(a) xyy" - 2x(y')2 yy' = O.+
(b) xyy" +
x(y')2 - yy' = O.
(c) xyy" - 2x(y')2 + (y + 1)y' o.

ANSWERS 35
1. y = CI tan [CI(X + C2»).
2. Cly2 = (CIX + +
C}C2)2 k.
3. ClY = sinh (CIX + C2).
(log X)2
4. y = - - 2 - + log x + C2.
CI

3
"3 + c}x + C2.
X 2
5. y =
6. (y + 1)-2 = CIX + C2.
7. t = f 2k CIT dr + C2. To evaluate the integral, let r u 2 • See 22.

8. x = ± f v'k dy
y3 + CI
+ C2.
9. x = ± f v'k dy
y4 + CI
+ C2.
 506 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

10. x = y + crlog (y - cr) + C2; y = c.

± -1
n. t = C2 C! V clr2 + 2kr - h2 ± ---a-'2 k log [ r + -k + + 2kr - h2 ] .
cr C! Cl
12. y = x + Cl log (C2Y).
13. y-2 = CIX C2. +
14. C1 2 y = (cr 2 + 1) log (CIX + 1) - crx + C2.
15. y = Cl Arc tan x - x + C2.
16. (y + 1)-2 = x.
17. y = -log (4 - x).
18. 11 = tan (x + .
19. ey/ 2 = _2_.
2-x
(log x)2
20. y = - - 2 - + 2 log x + 1.
3
x
21. y = 3' - x
2
+ 32 .
22. k = i, cr -1, t = 1 r

± r dr = =F (Arc cosVr+ v'r(l - r»).

NOTE. Since the limits of integration are from 1 to rand r < 1, use of the
plus sign in the integrand will give a negative time; use of the minus sign
will give a positive time.
23. (a) y-l = Cl log x + C2. (b) y2 = CIX 2 + C2.
(c) 2 Arc tan (cry) = crlog x + C2.

LESSON 36. Problems Giving Rise to Special Types of

Second Order Nonlinear Equations.

LESSON 36A. The Suspension Cable. A cable, chain, string, or

similar object supported at two ends is called a suspension cable. It
may support a load attached to it as in the case of a bridge, or it may
hang under its own weight. We consider the latter possibility first. We
shall determine for it the shape of the curve that the cable assumes.
In Fig. 36.1, we have drawn a cable, assumed to be perfectly flexible
and inextensible, supported at two ends, A, B, and hanging under its
own weight. Whether the cable will have the appearance shown in Fig.
36.1(a) or in Fig. 36.1(b) will depend on the length 8 of the cable relative
to the length AB between the points of support. We consider the more
general case shown in Fig. 36.1(a), where the cable at its lowest point
does not necessarily have a horizontal tangent as it does in Fig. 36.1(b).
Let P(x,y) and P1(x y + +
be two neighboring points on the
curve AB, and call the length of the arc between them. The forces
acting on this piece of cable, considered as isolated from the rest of the
system are [refer to Fig. 36.1(a)]:
Lesson 36A THE SUSPENSION CABLE 507

y y
t+
.±.

(0,0) (xo,O) x (0,0) x

(a) (b)
Figure 36.1

1. A tension Tat P directed along the tangent to the curve.

2. A tension T1 at P 1 also directed along the tangent to the curve but in
a direction opposite to that of T.
3. A force due to the weight of the cable of length d8, directed downward.
If we assume a homogeneous cable whose weight is uniformly dis-
tributed and is w pounds per foot, then the weight of the cable of length
d8 is w d8.
By hypothesis, the cable is perfectly flexible and inextensible. And
since it is in equilibrium, the portion PP 1 of the cable will retain its
shape under the action of the three forces 1, 2, 3 above, even if it were
cut at P and Ph just as if it were a rigid body. But the condition for
static equilibrium of the piece of cable PP 1 is that the algebraic sum of
the components of the system of forces acting along an arbitrary direction
be zero. Hence equating to zero, the algebraic sum of the horizontal com-
ponents of the three forces 1, 2, 3 above, we obtain [keep referring to Fig.
36.1(a»).
(36.11) T1 cos 81 - T cos 8 = 0,

where 8 and 8 1 are the angles shown in Fig. 36.1(a).

Equating to zero the vertical components of these forces, we obtain

(36.12) T 1 sin 81 - T sin 8 - wd8 = O.

The change in the horizontal tension (T cos 8), is

(36.13) d(T cos 8) = T1 cos 81 - T cos 8;

508 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

the change in the vertical tension T sin 8, is

(36.14) fl(T sin 8) = Tl sin 81 - T sin 8.

Substituting (36.13) in (36.11) and (36.14) in (36.12), we obtain respec-

tively
(36.15) fl(T cos 8) = 0, fl(T sin 8) = wfls.

Dividing each of the equations in (36.15) by flx and letting flx -+ 0,

there results the system of equations
d
(36.16) dx (T cos 8) = 0, -d ( T ' ) = wds
sln8 -'
dx dx

From the first equation in (36.16), we find that

(36.17) Tcos 8 = H,

where H is a constant. Since T cos 8 is a constant and equal to H, we

may write the second equation in (36.16) as

(36.18) d
dx
(TTcos8
sin 8)
= Ii
w ds
dx'
which simplifies to
d ds
(36.19) dx (tan 8) = k dx '
where
(36.2) k = w.
H

But tan 8 is the slope y' of the curve of the cable at P. Hence

(36.21)

Substituting (36.21) in (36.19) and recalling that the differential ds of the

arc PP I of the cable is given by the formula ds = v'dx 2 dy2 so that +
ds
dx = vI + (dyjdx)2,
we obtain
(36.22)

This is a nonlinear second order equation of the type discussed in Lesson

35B with y missing. Following the method outlined there, we let u = y',
Lesson 36A THE SUSPENSION CABLE 509

u' = y" and thus obtain from (36.22), the system

dy
(36.23) dx = u, -du
dx
= kvl + u 2,
By the method of Lesson 6e, and with the constant of integration taken
to be kxo, the solution of the second equation in (36.23) is
(36.24) log (u + VI + u 2 ) = kx - kxo,
u + VI + u 2 = e"(X-x.).

The second equation in (36.24) can be solved for u by the usual algebraic
means. A simpler method, however, is the following. Take the reciprocal
of each side of the second equation in (36.24) and rationalize the de-
nominator of the resulting left side. There results

(36.25)
Adding (36.25) and the second equation in (36.24), we obtain

(36.26)
By (18.9), we can write (36.26) as

(36.261) u = sinh [k(x - xo)].

Replacing u in (36.261) by its value as given in (36.23) and k by its value

as given in (36.2), we obtain
(36.27) y' = sinh [;; (x - xo) J.
Integration of (36.27) gives

(36.28) Y = (x - x o)] + c,
which is the equation of the curve of a hanging cable supported at two
points P and PI. Since the same fonn of equation (36.28) results if the
points P and PI are taken anywhere on AB, (36.28) is the equation of a
hanging cable supported at two ends A,B. Keep in mind that as P and
PI shift, the values of H and x in (36.28) change.
There are three constants in our solution (36.28), H, xo, and c. From
(36.27), we see that y' = 0 when x = xo, (remember sinh 0 = 0). This
means that Xo is the x coordinate of that point on the curve (36.28) where
its slope is zero. This point is, therefore, a minimum point of the curve.
If when x = xo, we now choose our x axis so that y has the value H/w,
then by (36.28), c = 0, (remember cosh 0 = 1). In Fig. 36.I(a), we
have shown the point (xo, H/w). If we now choose our y axis so that
510 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

y y

,
' . ....... A

H=T t++
---+

(0,0) X (0,0) X
(a) (b)
Figure 36.29

it goes through the point (xo, H/w), then Xo = O. With reference to

these new axes, i.e., where the y axis goes through the lowest point of
the curve (36.28), and the x axis is H/w units below this lowest point,
see Fig. 36.29(a), (36.28) simplifies to

(36.3) y = ! cosh (;; x) ,

and (36.27) simplifies to
(36.31) y' = sinh (;;x).
A curve whose equation has the form (36.3) is called a catenary.
Finally to find H we make use of the fact that the length AB of the
cable is 8. Let a be the x coordinate of the point Ai b the x coordinate of
the point B. Then from the calculus, we know

(36.32) 8 = .f VI + (y')2 dx.

By (36.31), we can write (36.32) as

(36.321) 8 = lab + sinh 2

(;;x) dx

Jar cosh wx dx
b
=
H

= HwSinhllx
W Iba'
Lesson 36A THE SUSPENSION CABLE 511

Therefore
(36.33)

In (36.33), s, w, a, b are given constants. This equation therefore deter-

mines H.
Comment 36.34. Let us assume that s is sufficiently long so that the
cable hangs as in Fig. 36.1(b). Hence the point (xo,H/w) at which the
curve has a horizontal tangent is a point of the cable. The tension T of
the cable at that point is therefore also horizontal. But when the tension
T is horizontal, the angle 8 in Fig. 36.1(a) is zero. And when 8 = 0, we
see from (36.17) that T = H. Hence for this special case, H is the tension
of the cable at its lowest point, see Fig. 36.29(b). And if Xo = 0, the length
8 of a catenary from its lowe8t point (O,H/w) to any point P(x,y) on it, for
this special case, is therefore [in (36.33) take a = 0, b = xl,
H . hW
(36.35) 8 = W sm H X'

Example 36.36. A cable 50 feet long, weighing 5 lb/ft, hangs under

its own weight between two supports 30 feet apart. Find:
1. The equation of the curve.
2. The tension at its lowest point.
3. The sag.

(0,0)

Figure 36.37

Solution (Fig. 36.37). If we choose the origin so that the y axis goes
through the lowest point (O,H /w) of the cable, then by (36.3) the equation
of the curve formed by the cable is

(a)
H w
y = w cosh H Xj

and by (36.35), the length of the cable from x = °

to x = x is
H . hW
(b) s = w sm H X'
512 PROBLEMS: S:rSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

Inserting in (b) the given conditions, w = 5, x = 15, 8 = 25, we find

(c) 25 = sinh 15, 125 = sinh 75 .

H H
Using a table of values of sinh x, the approximate value of H which will
satisfy (c) is
(d) H = 40.8 lb.
This is the tension of the cable at its lowest point. When H = 40.8 and
w = 5, we obtain from (a)
x
(e) y = 8.16 cosh 8.16 '

which is the equation of the curve of the cable. From it we find that
when x = 15,
15
(f) y = 8.16 cosh 8.16 = 8.16(3.22) = 26.28 ft,

and that when x = 0, y = 8.16. Hence the sag of the cable is '26.28 -
8.16 = 18.12 feet.
Comment 36.38. If the specific weight of a body is not uniform but is
a function of x, then its total weight W from x = 0 to x = x is given by

(36.4) W = 10'" f(x) dx,

where f(x) is the specific weight of the body, i.e., its weight per unit
length. For example if f(x) = 50 +
x pounds, then when x = 10 feet,
the specific weight of the body at that point is 60 lb/ftj when x = 10.1
feet the specific weight at that point is 60.1 lb/ft. By (36.4), its total
weight W from x = 0 to say x = 20 feet is
20
W = 10 (50 + x) dx = [50X + = 1200 lb.

Assume that a perfectly flexible cable of negligible weight supports, by

means of vertical rods, a horizontal load such as a bridge, whose specific
weight is f(x), The rods are equally spaced and close enough to form a
continuous system. The weights of the rods are also negligible. We wish
to find the equation of the curve formed by the cable.
Following the steps used to arrive at the equation of the hanging cable
with no load attached, we find that no changes occur until we reach
(36.12). In this equation, Wd8 must be replaced, see Fig. 36.41, by f(X)dX,
where f(3:.) is the specific weight of the horizontal load measured from the
origin, which is taken to be the lowest point of the curve. Continuing
from there on, and replacing Wd8 by f(X)dx, we obtain eventually, in
Lesson 36A THE SUSPENSION CABLE 513

1++
-
Figure 36.41

place of (36.18),
(36.42)
dx
(T sin 8) =
Tcos 8
f(x) ,
H

d(tan 8) = dx,

d = dx.

Integrating the last equation and making use of the fact that when x = 0,
dy/dx = 0, we have
(36.43) = fox f(x) dx.
A second initial condition is x = 0, y = O. By Comment 36.34, the H
in this equation is the tension at the lowest point of the cable.
Example 36.44. A perfectly fleXible suspension cable, attached to two
towers at the same level, 100 feet apart, supports a bridge (assumed rigid)
by means of vertical rods connecting cable and bridge. The rods are
equally spaced and close enough to form an approximately continuous
system. The weights of cable and rods are negligible in comparison with
the weight of the bridge whose specific weight is given by f(x) = 10 +
x 2 /50 lb/ft. If the length of the cable is such that its sag is 25 feet when
the bridge is horizontal, find the equation of the curve in which the cable
hangs and the tension at its lowest point. Assume that the origin is taken
at the lowest point of the cable.
Solution. See Fig. 36.441. By (36.43) and the given specific weight
f(x) = 10 + x 2 /50, we obtain

dy (" (X2) x
3
(a) H dx = )0 10 + 50 dx = lOx + 150 .

Integration of (a) gives

2 X4
(b) Hy = 5x + 600 + C1·
514 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

The substitution in (b) of the initial condition x = 0, y = 0 gives Cl = 0;

the substitution of the initial condition x = 50, Y = 25, gives
4
(c) 25H = 5(50)2 + 600
50 , H = 917,

(0,0) (50,0)

Figure 36.441

which is the tension of the cable at its lowest point. Hence (b) becomes

(d) Y =
1
917 5x
(2 + 600
x4) '
which is the equation of the curve of the cable.

EXERCISE 36A
1. A cable of length 28 and uniform weight w lb/ft hangs from two supports
on the same level, Fig. 36.45. The supports are 2L ft apart with L < 8.
The tension at the lowest point of the cable is H.
y L

d(
t+.±.

(0,0) x
Figure 36.45

(a) Show that the sag d is given by

(36.46) H(
d=;; cosh'H-l' wL )
Hint. Use (36.3). The sag is the difference in values when x = Land
x = o.
Lesson 36A-Exercise 515

(b) Show that the length a of the cable from x = 0 to x = Lis

(36.47) a = !!..
w
sinh wL.
H
Hint. Use (36.35) with x = L.
(c) Show that the tension T at any point of the cable is given by
w
(36.48) T = H cosh Ii x = wy.

Hint. By (36.17) (see also Fig. 36.45) T = H sec 8 = H

vdx dx+ dy2 =
2

HVI +
(dy/dx)2. Replace y' by its value as given in (36.31). Remark.
Since cosh x is an increasing function of x, the tension is a maximum
when x is largest, i.e., at a point of support.
(d) Show that the total weight of the cable is given by
(36.49) W = 2H sinh (wL/H).
Hint. Use (36.47) which gives the length of cable from x = 0, to x = L.
(e) Show that the horizontal tension H is given by
(36.491) e",LlH = (8 + d)/(a - d),

wL lb
H
log [(a + d)/(a - d)] •
Hint. In (36.46) and (36.47), change the hyperbolic functions to their
exponential forms, see (18.9) and (18.91), and then show that the right side
of the first equation in (36.491) simplifies to e",LIH.
(f) Show that the horizontal tension H is also given by
(36.492) H = W(8 2 - d 2 )/2d.
Hint. Use (36.46), (36.47), and the fact that cosh 2 x - sinh 2 x = 1.
2. A telephone wire, weighing 0.04 Ib/ft, is attached to poles at intervals of
300 ft. The sag at the center is 7.5 ft.
(a) Find the tension at the lowest point of the curve. Hint. Use (36.46).
Remember in this formula, the distance between supports is 2L.
(b) What is the length of the wire between poles? Hint. Use (36.47). Re-
member in this formula the length of wire is 2a. Use (36.491) or (36.492)
to check your result.
(c) What is the tension at the supports? Hint. Use (36.48) with x = 150.
(d) Find the equation of the curve. Hint. Use (36.3).
3. A cable 100 ft long, weighing 41b/ft, hangs under its own weight between
two supports on the same level and 80 ft apart. Find:
(a) The equation of the curve.
(b) The tension of the cable at its lowest point; at a point of the cable mid-
way between the lowest point and a point of support; at a point midway
horizontally between the lowest point and a point of support; at a point
of support.
(c) The sag.
(d) The slope of the curve at a point of support.
For hints, see answer section.
516 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

4. A cable hangs under its own weight between two supports on the same level.
The slope of the curve formed by the cable at a point of support is 0.2013.
Its sag at the center of the cable is 12 ft. Find:
(a) The distance between supports.
(b) The length of the cable.
(c) The equation of the curve.
Hint. Use (36.31) to find wL/H and solve for H/w in terms of L. Then use
(36.46), (36.47), (36.3) in that order.
5. A chain 117.5 ft long hangs under its own weight between two supports on
the same level 100 ft apart.
(a) Find the sag of the chain. Hint. Use (36.47) to find H/w. Then use
(36.46).
(b) Find the equation of the curve.
(c) Find the maximum tension if the chain weighs 2Ib/ft. Hint. See re-
mark after (36.48). Use (36.48).
6. A chain hangs under its own weight between two supports on the same level
100 ft apart. Its sag is 7.55 ft.
(a) Find the length of the chain. Hint. Solve (36.46) for H/w. Then use
(36.47).
(b) Find the equation of the curve.
7. A cable hangs under its own weight between two supports on the same level
50 ft apart. The slope of the curve formed by the cable at a point of support
is 0.5211. Find:
(a) The length of the cable.
(b) The sag.
(c) The equation of the curve.
(d) The tension of the cable at its lowest point if it weighs 0.5 lb/ft, i.e.,
find H.
Hint. Use (36.31) to find w/ H. Then use (36.47), (36.46), (36.3) in that
order.
8. A cable, 100 ft long, hangs under its own weight between two supports on
the same level. Its sag at the center is 10 ft.
(a) Find the distance between the supports.
(b) Find the equation of the curve.
Hint. Use (36.491) to find wL/H. Then use (36.46) or (36.47).
9. A cable 200 ft long hangs under its own weight between two supports on the
same level. Its sag at the center is 20 ft. Its maximum tension, see remark
after (36.48), is 60 lb. Find its weight w per foot. For hints, see answer
section.
10. Start with the general equation of the hanging cable as given in (36.28).
It contains three constants, Xo, c, H. By choosing our origin in a special
way, we were able to make Xo = 0, c = O. H was then determined by using
the given length 8 of the cable. This time let us take the origin at the lower
point A. of the hanging cable and call (Xl,Yl) the coordinates of the higher
point B, see Fig. 36.493. Show that the values of the three constants can be
Lesson 36A-ExercilMj 517

Figure 36.493

obtained from the equations

H
(A) 0 = w cosh H xo + c,
W

(36.494) Yl = cosh [N (Xl - XO)] + c,

(B) 8 = .f' VI + {y')2 dx = [Sinh N (Xl - xo) + sinh NXo]
2H [ • h WXl h W{XI - 2XO)]
= til sm 2H cos 2H .
By sUbtracting (A) from (36.494), show that

(36.495) Yl = !!.. [COSh W{XI - xo) _ cosh WX o]

W H H
• h W{XI - 2Xo)]
2H [ • h WXl sm
= til sm 2H 2H .

Finally show by (B) and {36.495)-square both equations and then sub-
tract the second from the first-
2- 2- 4H2. 2(WXl)
8 Yl
w2 smh -2H ,
--

(36.496) sinh (WXl)

2H
= ..!!!...
2H
V,2 - Y12 •
Following the hint given after (36.48), show that the tension at any point
of the cable is given by
W
(36.497) T = H cosh H (x - xo) = w(y - c),

where c is a constant of integration whose value is given by (A) and (36.494).

II. Using the equations found in 10, solve the following problem. See Fig.
36.493. A cable of length 100 ft and uniform weight of 2 lb/ft hangs from
two supports A and B, whose horizontal distance is 50 ft apart. Support
B is 20 ft higher than A. Find;
(a) The equation of the hanging cable. [Hint. Use (36.496), (36.495), and
(36.494) in that order.]
(b) The coordinates of its lowest point, i.e., the point where y' = o.
(c) The tension at both supports and at its lowest point.
518 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

In problems 12-16, assume the weights of the cable and rods are neg-
ligible in comparison with the weight of a horizontal supported load, that
the cable is perfectly flexible, that the rods are equally spaced and close
enough to form a continuous system, and that the origin is at the lowest
point of the cable.
12. A suspension bridge is 2x ft long and is supported by vertical rods. The
weight of the bridge is uniformly distributed and is W Ib per unit horizontal
distance. Find the equation of the curve formed by the cable. Hint. In
(36.43)./(x) = w.
13. Show that the tension T at any point of the cable of problem 12 is given by

(36.498)
See (36.48) and hint following; here y' is obtained from problem 12.
14. A suspension bridge is 200 ft long. The supporting ends of the cable are 50 ft
above the bridge and the center of the bridge is 10 ft below the center of the
cable. The weight of the bridge is uniformly distributed and is w Ib per unit
horizontal distance.
(a) Find the tension at the lowest point of the cable, i.e., find H. Hint.
In the solution to 12, use the fact that x = 100, y = 40.
(b) Find the equation of the curve formed by the cable.
(c) What is the tension at the supports? Hint. See (36.498).
(d) What is the slope of the cable at the supports?
15. A suspension bridge is 2L ft long. The weight of the bridge is uniformly
distributed and is w Ib/ft. The supports of the cable holding the bridge are
a ft above the origin.
(a) Find the tension at its lowest point, i.e., find H. Hint. In the solution
to problem 12, use the fact that when x = L, y = a.
(b) Find the equation of the curve.
(c) What is the tension at the supports? Hint. See (36.498).
16. A suspension bridge is 200 ft long. The sag at its center is 50 ft. The specific
weight of the bridge is given by I(x) = 100 +
x 2 • Find:
(a) The tension at the lowest point, i.e., find H.
(b) The equation of the curve formed by the cable.
17. A uniform, flexible cable weighing Wl Ib/ft supports a horizontal bridge.
The weight of the bridge is uniformly distributed and is W2 Ib/ft. The weight
of the supporting rods is negligible. Show that the differential equation of
the curve of the cable is given by

d2 y
(36.499) H dx 2 = wl'V1 + (dy/dx)2 + W2.
Hint. Combine equations (36.22) and (36.42) with k = Wl/ H and/(x) = W2.
18. If the rods in problem 17 are not of negligible weight and weigh W3 Ib/ft,
show that the differential equation of the curve of the cable is given by
2
(36.499i) H t!J!.
dx 2
= wlV1 + (dy/dx)2 + W2 + W3Y,
Lesson 36A-Exercille 519

where y is the length of a rod. Assume that the rods are so clOSE! together as
to form an approximate continuous system. Hint. The weight due to a rod
of width Ax is way Ax.
19. Holes are bored in the ends of slender uniform rods of varying lengths and
the rods are then strung on a cord of negligible weight. Assume that the rods
just touch each other, that each weighs w lb/sq ft and that their other ends

t+
(O,a) .±.

(0,0) x

Figure 36.4992

lie on a horizontal line, Fig. 36.4992. Find the equation of the curve formed
by the cord. Take the origin at a distance a feet below the lowest point of
the curve. Hint. In (36.15), replace w As by wy Ax. Initial conditions are
II: = 0, Y = a, dy/dz = O.
20. For a nonhomogeneous hanging cable whose weight is a function of its
distance from its lowest point, show that the differential equation of the
curve formed by this hanging cable is
2
d
(36.4993) H = p(s)Vl + (dy/dx)2,
where pes) is the specific weight of the cable, i.e., its weight per unit length
at a distance s units from the lowest point of the cable. Hint. In (36.15),
replace w by pes).
21. If in (36.4993), pes) = aH /(a 2 - x 2 ), and the origin is taken at the lowest
point of the cable so that II: = 0, Y = 0, dy/dx = 0, show that the curve
formed by the hanging cable will be an arc of the circle.
x2 + (y - a)2 = a 2 •

22. An arched bridge is to be built of stone of uniform density. The weight of

the stone is w lb for each square foot of facing. The bridge is so constructed
that at each point of the arch the resultant tension due to the weight of the
stones above it acts in a direction tangent to the arch, see Fig. 36.4994. Find
the equation of the arch. Take the x axis at the top of the bridge, the y axis
through the highest point of the arch, the origin as indicated in the figure
and the positive direction downward. Hint. Since the tension at each point
acts in a direction tangent to the curve, equations (36.15) apply with w As
replaced by wy Ax. Initial conditions are x = 0, y = h, dy/dx = O.
520 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

(0,0) A
E' "
" axis

l+
(O,h)
- .......
;;;;1--'
(",y)
........ ...
yaxis wylb

Figore 36.4994

23. Assume a road is built on top of the masonry of problem 22. The weight
of the road is distributed uniformly and weighs kw lb per linear foot. Find
the equation of the curve. Hint. In (36.15) replace w As by wyAx kw Ax. +
ANSWERS 36A
2. (a) 60 lb. (b) 300.5 ft. (c) 60.3 lb. (d) y = 1500 cosh (x/1500).
3. (a) Use (36.47) to find H. Then use (36.3). (b) Use (36.35) to find x
when 8 = 25. Then use (36.48) with x equal successively to 0; its value
when 8 = 25; 20; 40. (c) Use (36.46). (d) Use (36.31) with x = 40.
4. (a) 239.3 ft. (b) 240.9 ft. (c) y = 598 cosh (x/598).
5. (a) 27.2 ft. (b) y = 50 cosh (x/50). (c) IM.3Ib.
6. (a) 101.5 ft. (b) y = (500/3) cosh (3x/500).
7. (a) 52.11 ft. (b) 6.38 ft. (c) y = 50 cosh (x/50). (d) 25 lb.
8. (a) 97.3. (b) y =- 120 cosh 0.0083x.
9. Use (36.492) to find H/w. Use (36.47) to find L. Use (36.48) with x = L
(write H = w(H/w) and solve for w}; w = 3/13Ib/ft.
11. (a) H = 23.35 approx., Xo = 22.6 approx., C = -41.3 approx.,

y = 11.7 cosh (x - 41.3.

(b) (22.4, -29.7). (c) 82.7 Ib at A, 122.7 lb at B, 23.35 lb at its lowest

point.
2
w x
12. Y = H 2' ' a parabola.
14. (a) H = 125w. (b) y = x 2 /250. (c) T = 25wv'41.
(d) Slope = ::1::0.8.
IS. (a) H = wL2/2a. 2
(b) y = az /L2. (c) (wL/2a}v'L2 + 4a2.
16. (a) H = 530,000/3. (b) y = (3/530,000) (50X2 + .
19. Y =; =
acoshv'w/Hx.
22. Same as in 19 with a replaced by h. Since the tension at each point of the
bridge is tangent to the curve, it follows that no mortar will be needed if
the bridge has the shape of a catenary.
23. y = (h + k) cosh v'w/H x - k.
Lesson 36B A SPECIAL CENTRAL FORCE PROBLEM 521

LESSON 36B. A Special Central Force Problem.

Example 36.5. A particle weighing 16 pounds and 10 feet from a

fixed point is given a velocity of 15 ft/sec in a direction perpendicular to
the x axis. The particle is attracted to the fixed point by a force F whose
magnitude is inversely proportional to the cube of its distance from the
point. If the proportionality constant is 8, find the distance of the particle
from the fixed point as a function of the time.
Solution. Let us take the origin of a coordinate system at the fixed
point. We have already proved in Lesson 340 that a particle subject to a
central force moves in a plane. Since the force F acts only in a direction
along the radius vector, the components of F are, with the proportionality
constant equal to 8,
8
(a) Fr = - -r 3 = mar, F, = 0 = ma,.

Here the mass m = it = !. Hence by (34.21), (a) becomes

(b) (2f8 + riJ) = O.

Mter multiplication by 2r, the second equation in (b) is equivalent to

(c)
from which we obtain
(d)

The initial conditions are t = 0, r = 10, v, = 15. Therefore, by the sec-

ond equation of (34.2), r d8/dt = 15. Substituting these values in (d), we
find Cl = 150. Hence (d) becomes
(e) 8 = 150.
r2

In the first equation of (b) replace 8 by its value as given in (e). There
results

(f) !(I'
2
_150r 3
2
) =
r3
..
r=
1502 -
r3
16 22,484
= - -3 - ,
r
an equation which is of the type discussed in Lesson 35A. As suggested
there, we let u = dr/dt and obtain the system
. 22,484
(g) f = u, u= - -3- .
r
Following the procedure outlined in Lesson 35A, we multiply the second
522 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

equation of (g) by 2u. The remaining steps are given without further
comment.
(h)

At t = 0, r = 10, u = l' = O. Substituting these values in the last

equation of (h), we find C2 = 22,484/100. Hence by (h) and (g),

(i) u=
2
1'2 = -22,484 U2 - = 22,484 (r2 ,
2
dt) l00r2
( dr = 22,484(r 2 - 100) ,
lOr
dt = dr.
149.9v'r 2 - 100
Integration of (i) gives

0) t = v'r2 - 100 + Ca·

At t = 0, r = 10. Therefore Ca = O. Hence (j) becomes
(k) (14.99t)2 = r2 - 100
r2 = 100 + 224.84t2
r = v'100 + 224.84t 2 •
Comment 36.51. Note that for this special central force problem, we
were able to find r as a function of t relatively easily. In Comment 34.91,
we remarked on the difficulty in finding r as a function of t in the case of
a particle moving subject to the inverse square law.

EXERCISE 36B
1. A particle of mass m is attracted to a fixed point 0 by a force F that varies
inversely as the cube of the distance r of the particle from O. Initially the
particle is a units from 0 and is given a velocity Vo in a direction perpendicular
to the x axis. Find:
(a) t as a function of r.
(b) r as a function of t.
(c) r as a function of 8, where 8 is the polar angle.
Take mk for the proportionality constant. Hint. In (c), after you obtain
an equation for;.2 divide the equation by tJ2 = a2vo2/r4 and note that ;./tJ =
dr/d8. Initial conditions are t = 0, r = a, 8 = 0, ;. = 0, vo = atJo.
2. Solve problem 1 if the particle is repelled from instead of being attracted
to O.
Lesson 36C A PURSUIT PROBLEM 523

3. A particle of mass m is attracted to a fixed point 0 by a force F that varies

inversely as the fifth power of the distance r of the particle from O. Take
k 2 m for the proportionality constant. Initially the particle is a units from 0
and is given a velocity vo = k/(a 2VZ) in a direction perpendicular to the
z axis. Find r as a function B. See also Exercise 34H,4. Hint. You should
obtain t 2 = -k2/(2a 2r2) +
k2/(2r4) and 92 = k 2/(2a 2r4). Divide t 2 by 92
andnotethatt/9 = dr/dB. Initialconditionsaret = O,r = a,B = 0, t = 0,
vo = abo = k/(a 2v'2).

ANSWERS 36B
2
2 a 2 2
1. (a) t = 2 2 k (r - a ),
avo -
2 2
(b) r2 = a2 + a vo - k t2 ,
a2

(c) r = asec
Va 2vo2 - k
B.
avo
Note that there are three different possible orbits depending on whether k
is less than, equal to or greater than a2vo 2 •

2. t
2
= a 2t'o2
a
2
+ k (r2 - a 2 ), r2 __ a2 + a2voa22 + k t2,
r = asec
va2vo2 + k B.
avo
3. r = a cos B, which is the equation of a circle through the origin.

LESSON 36C. A Pursuit Problem Leading to a Second Order

Nonlinear Differential Equation. In Lesson 17, we solved pursuit
problems leading to a first order differential equation. We shall now solve
a pursuit problem which leads to a second order nonlinear differential
equation.
Example 36.6. A fighter pilot sights an enemy plane at a distance
and starts in pursuit, always keeping the nose of his pJane in the direction
of the enemy plane. Assume the enemy plane is flying in a straight line
at V E milesjhour, and the fighter plane is flying at V F miles/hour. Find
the equation of the path of the fighter plane as seen by an observer on the
ground. NOTE. This is the same type of problem as the relative pursuit
problem of Lesson 17B. Here, however, the origin is fixed in space and
does not move with the enemy plane.
Solution (Fig. 36.61). Let the origin be the position of the enemy
plane, and Po(xo,Yo) be the position of the fighter plane, at time t = O.
Let the x axis represent the line of flight of the enemy plane. Then at the
end of t seconds, the enemy plane has gone a distance equal to VEt miles
along the x axis, and the fighter plane, which is now at P(x,y), a distance
524 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

V Ft miles along the arc of his path. Hence the fighter plane's distance in
time t is given by

(a) VFt = - [" vI + (dx/dy) 2 dy, t= - -V1F 111 vI + (dx/dy) 2 dy.

i llo 110
(The minus sign is necessary since Yo > y.) At P the fighter's direction
is toward the bomber's position (VEt,O) and is tangent to the curve of pur-

(0,0) (VEt,O) x
Figure 36.61

suit there. Hence the slope of the curve at Pis y' = tan 8 = -y/(VEt - x).
Therefore at P

(b) dy _ y . t=_1 (x_ ydX ).

dx - x - VEt' VE dy

Equating the values of t in (a) and (b), we obtain

(c) _1_
V
(x -
E
y dX)
dy
= - 111 vI
V F 110
+ (dx/dy) 2 dy.

Differentiation of (c) with respect to y gives, by (9.15),

2
(d) V
1 (dX
dy -
E
d x
Y dy2 -
dX)
dy = - V
1
F
vI + (dx/dy)2,
which simplifies to, with V E/VF replaced by a new constant k,
d 2x V
(e) Y dy2 = k-\!I + (dx/dy)'l, k = V;'

This equation is of the type discussed in Lesson 35B, with x and y inter-
changed. Following the procedure outlined there, we let
dx
(f) u.= dy'
Leason 36C A PURSUIT PROBLEM 525

Substituting these values in (e), we obtain the system of first order equa-
tions
(g)
dx
dy = U,

The solution of the second equation in (g) is

(h) log (u = klogy + logcl,

which we can write as
(i)

Squaring (i) and simplifying the result, we obtain

(j)

Replacing U by its value in (f), we have

dx = Cl
2y2k - 1 = ![c k _ y-"].
(k)
dy 2ClY" 2 IY CI

If k = V IdV F, is not equal to one, Le., if the velocities of fighter and

bomber are not the same, then integration of (k) gives

(1)

At t = 0, x = Xo, Y = Yo, dy/dx = Yo/xo, which implies, by (f), that

u = dx/dy = xo/Yo. Substituting these values in (1) and (k), we obtain
respectively
(m)

Xo
(n)
Yo
With the aid of equations (m) and (n), we can obtain, in a particular
problem, the values of CI and C2. The curve of pursuit is given by (1).
If k = 1, then by (k)

(0) dx
dy
= !2 (ClY _
ClY
,

x = -21 (y2
CI -
2
-
1
-log
CI
y + C2) •
526 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

EXERCISE 36C
1. A dog sees a cat and starts in pursuit always running in the direction of the
cat. The dog's speed is v" ft/sec, the cat's v. ft/sec. Assume that initially
the dog is at (O,a), the cat at (0,0) and that the cat runs straight for a tree
located on the x axis.
(a) Let k = v./v". Determine the equation of the dog's path if k 1 and
if k = 1. Hint. Follow the method of Example 36.6. Initial conditions
are t = 0, x = 0, y = a, U = dx/dy = 0.
(b) Show that if k < 1, i.e., if the cat's speed is less than the dog's speed,
the cat will reach the tree in safety pro.vided the distance of the tree from
the origin is <ak/(1 - k 2 ). Hint. Show that when y = 0,
x = ak/(1 - k 2 ) and therefore that dog and cat would meet at this
distance from the origin.
(c) When will the dog reach the cat, assuming the tree's distance is
>ak/(1 - k 2 )?
°
(d) Show that if k > 1, x -+ 00 as y -+ and if k = 1, x -+ 00 as y -+ 0.
In both cases, it is evident that the dog cannot reach the cat.
2. A fighter plane, located at (0,0), sights a bomber at (a,O) and starts in pursuit
always keeping the nose of his plane in the direction of the bomber. Assume
the bomber is flying parallel to the y axis at VB mi/hr, and the fighter plane's
speed is at Vp mi/hr.
(a) Let k = VB/VP. Find the equation of the fighter's path if k 1, and if
k = 1. Hint. Follow method of Example 36.6. Initial conditions are
t = 0, x = 0, y = 0, dy/dx = 0.
(b) Show that if k = VB/VP < 1, i.e., if the bomber's speed is less than the
fighter's speed, the fighter will reach the bomber at x = a, y =
ak/(1 - k 2 ). Hence show that the fighter will reach the bomber in time
t = ak
vB(1 - k 2)
(c) Assume a = 2 mi, Vp = 300 mi/hr, VB = 200 mi/hr. When will the
fighter reach the bomber?

(0,0) x
Figure 36.7

3. Assume that the pursued object does not move along one of the axes, but
along a given curve F(X,Y) = 0, and that at time t, its position is P(X,Y),
Fig. 36.7. The pursuer, as usual, always moves toward the position P of the
pursued object. If the pursuer's position at time t is Q(x,y) and the equation
Lesson 36C-Exercise 527

of his path is y = f(x) , we have, for the equation of the line PQ,
(36.71) (X - x)y' = Y - y.
With Q(x,y) considered as fixed, the differential of (36.71) is
(36.72) (X - x) dy' + y'dX = dY.
The differential of
(36.73) F(X,Y) = 0
is
aF aF
(36.74) ax dX + ay dY = O.
If the speed of Q is k times the speed of P, we have [speed = Ivl Ids/dtl
v(dx/dt)2 +
(dy/dt)2, Idsl = vdx 2 dy2] +
(36.75) Vdx 2 + dy2 = kVdX2 + dY2.
With the help of equations (36.71) to (36.75), solve the following problem.
A pursued plane P flies in a straight line making an angle of 45° with the x
axis. Find the equation of the path of the pursuing plane Q if its speed is
twice that of the pursued plane P. Choose the origin on P's path. Hint. Here
(36.73) is X - Y = 0, (36.74) is dX - dY = 0, (36.75) is

(36.76) VI + (y'P dx = WI + (dY/dX)2 dX = 2v'iTI dX = 2v'2dX,

(36.71) is (X - x)y' = X - y, (36.72) is (X - x) dy' = (1 - y') dX.
(x - X) dy' X - y y- x
Now show that dX = , and y' - 1 = - - - 1 = - - - ,
y-l X-x x-X
y-x (y - x) dy'
x - X = - ' - - I · Hence dX = (y' - 1)2 . Substitute this value of dX
y -
in (36.76) to obtain
2
d
(36.77) (y' - 1)2Vl + (y')2 = 2v'2 (y - x) -.J!..
dx 2
To solve (36.77), make the substitution

(36.78) U = y -x, du = dy _ 1 dy = 1 + du,

dx dx ' dx dx
There should result

(36.79) ( dU)
dx
2
'I + (1 +
'\J du)2 = 2v'2u a2u.
dx dx 2
Since (36.79) does not contain the independent variable x, the method of
Lesson 35C is applicable.

ANSWERS 36C

I. (a) x =
a[
2 k +1 1 (y)k+1
a + k _1 1
(y)l-k]
a - k2 ak_ l' k 1.
528 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

[l-
x=-1 - - - -
2 2a
2
a a l o g - , k=1.
a
y]

1 [(a - X)I+'" a"'(a - X)I-"'] ak

2. (a) y = 2 a"'(l +
k) - 1 _ k +1_ k2 ' k 1.

Y=2
1[1 a
aog a_x + x(x 2
- 2a)]
a'

(c) t = 43.2 sec.

( )3/2
3. 2V2x = 2cl'vi=-; - V2 (y - x) - y + C2.
LESSON 36D. Geometric Problems. In Lesson 13, we solved geo-
metric problems which gave rise to a first order differential equation. In
this lesson, we shall solve a geometric problem giving rise to one of the
specific types of second order equations discussed in Lesson 35.
In rectangular coordinates the radius of curvature R of a curve
y = f(x) at a point P(x,y) on it is given by the formula

(36.8) R=
[1 + (y')2]3/2 .
y"
y y
y>O; also
y">O since
the curve is
concave up
01------0
x
at P(%,y)
y<O; also
y"<O since
the curve is
concave down
o at P(%,y)

(a) (b)
Figure 36.81

H a normal to the curve is drawn from the point P to the x axis, then it
and the radius of curvature have:
1. Opposite directions when y and y" have the same signs, see Fig. 36.81(a)
and (b).
2. The same directions when y and y" have opposite signs, see Fig. 36.82(a)
and (b).
Leuon36D GEOMETRIC PROBLEMS 529

o x

y>O; also y<O; also

y" <0 since
y">O since
the curve is the curve is
concave down concave up
at P(x,y) at P(x,y)

x y

(a) (b)

Figure 36.82

Example 36.83. Find the family of curves whose radius of curvature

is twice the length of the normal segment from a point on the curve to
the x axis, (a) if normal and radius of curvature have opposite directions,
(b) if the two have the same directions.
Solution. The length of the normal segment from a point on the
curve to the x axis is [see (i) of Exercise 13,1] lyy'l (y')21. Hence by +
(36.8) and the hypothesis of the problem,

(a) [1 + = ± 2yy'1 + (y')2,

where the plus sign is to be used when normal and radius of curvature
have opposite directions, the minus sign when they have the same direc-
tion. Division of (a) by [1 +
(y')2] 1/2 gives

(b) 1 + (y')2 = ± 2yy",

an equation which is of the type described in Lesson 35C. In (b), we there-
fore make the substitutions
du
(c) y' = 1.1., y" = 1.1. dy'
thus obtaining
du 2udu ± dy.
(d) 1 + 1.1. 2 = ± 2yu dy , 1 +
1.1. 2 = Y
The solution of (d) is
(e) log (1 + 1.1. 2 ) = ±log y + log C1.
From (e) and (c), we obtain the equation
dy dy
(f) 1 + 1.1. 2 = ClY,
dx
± dx,
y'C1Y - 1
530 PROBLEMS: SYSTEMS. SPECIAL 2ND ORDER EQUATIONS Chapter 8

when normal and radius of curvature have opposite directions, and the
equation
(g) dy = ± - y, v1idy = ± dx
dx y ,
when they have the same directions. The solution of the last equation
in (f) is
2
(h) -YClY -
Cl
1 = ± (x + C2),
which is a family of parabolas with vertices at (-c2,ljCl) and axes parallel
to the yaxis. The solution of the last equation in (g) is (substitute u = v1i)

(i)
which is a family of cycloids. (For definition of a cycloid, see Exercise
28c, 34.)

EXERCISE 36D
1. Find the family of curves whose radius of curvature is equal to the length of
the normal segment from a point on the curve to the x axis and has the same
direction as the normal. Identify the family.
2. Solve problem 1, if the radius of curvature and normal have opposite direc-
tions. Identify the family.
3. Find the family of curves whose radius of curvature has a constant value k.

ANSWERS 36D
1. (x - Cl)2 + y2 = C2. A family of circles with centers at (ct,O) and radius
02·
2. cty = cosh (± CIX +
C2). A family of catenaries.
3. (x - ct)2 +
(y - C2)2 = k 2•
 Chapter 9

Series Methods

Introductory Remarks. As we have repeatedly emphasized, few

differential equations have solutions which can be expressed explicitly or
implicitly in terms of elementary functions. When a solution cannot be
expressed in this way, the problem of finding a solution of a differential
equation is not entirely hopeless. There are available graphical methods,
some of which were described in earlier lessons, numerical methods which
will be discussed in the next chapter and series methods which we shall
consider in this chapter. Furthermore, it is frequently true that an im-
plicit solution in terms of elementary functions is less useful than a series
solution or a numerical one. Implicit solutions are usually such compli-
cated expressions that it is extremely difficult to find values of the de-
pendent variable for given values of the independent variable.
For a clearer understanding of the subject matter of this chapter, it
will be necessary to have a knowledge of Taylor series and to know certain
definitions and theorems from analysis. Hence, before beginning a discus-
sion of series methods for solving differential equations, we shall first re-
view this needed material for you.

LESSON 37. Power Series Solutions of Linear

Differential Equations.

LESSON 37A. Review of Taylor Series and Related Matters. A

series of the form
(37.1) ao + a1(x - xo) + a2(x - XO)2 + a3(x - xO)3 + ... ,
where ao, at. a2, .•• , Xo are constants and x is a variable is called a power
series. A power series may:
1. Converge only for the single value x = Xo.
2. Converge absolutely for values of x in a neighborhood of Xo, i.e., con-
verge for Ix - xol < hi diverge for Ix - xol > h. At the end points
Xo ± h, it may either converge or diverge.
3. Converge absolutely for all values of x, Le., for - 00 < x < 00.
531
532 SERIES METHODS Chapter 9

In cases 2 and 3, the set of values of x for which the power series con-
verges is called the interval of convergence -of the series. In case 2, for
example, if a series also converges for x = Xo ± h, then its interval of con-
vergence is Xo - h x +
Xo h; if it converges only for x = Xo h, but +
not for x = Xo - h, then the interval of convergence is Xo - h < x
Xo +
h; etc. In case 3, the interval of convergence is the entire real axis.
Comment 37.11. Interval of Convergence. In the calculus you were
taught certain tests by which you could determine an interval of con-
vergence of a power series. A simple one and one which is frequently used
is known as the "ratio test." It states that the series

Ul + U2 + Us + . .. + u" + ...
converges absolutely if
(37.12) lim IUn+ll = k < 1.
'Un

We give below examples of each of the three types of series. For con-
venience we have taken Xo = O.
Example 37.13. Determine the interval of convergence of the power
series
(a) 1+ lIx + 2!x 2 + 3!x 3 + ... + n!x" + ....
Solution. Here Un = n!x", Un+l = (n + 1)!x,,+l. Therefore

(b) Un+l/ = / (n + I)! x"+l/ = I(n + l)xl.

/ u" n! x"
For each x F- 0, I(n +
l)xl -+ 00 as n -+ 00. Since this limit F- k < 1,
the series (a), by Comment 37.11, converges only for x = O.
Example 37.14. Determine the interval of convergence of the power
series
(a)

Solution. Here Un = x", U,,+l = X"+l. Therefore

(b)
ft-t>OO
I
lim Un+l / = lim / (n
'Un. n---+oo
+
n
1) X":l/ = Ix!-
x
Hence, by Comment 37.11, the series (a) converges absolutely for each x
whose absolute value is less than one. The interval of convergence, how-
ever, is -1 x < 1, since the series converges for x = -1, but diverges
for x = 1.
 Lesson 37A REVIEW OF TAYLOR SERIES 533

Example 37.15. Determine the interval of convergence of the power

series
(a)

X 2 (,,-I) x 2"
Solution. Here lu,,1 = (2n _ 2)1' 1u,,+11 = (2n)l· Therefore,
2 2
(b) lim / u,,+I/ = lim x " (2n - 2)1 = lim x = 0
"_00 u" ,,_00 (2n)! X 2 ,,-2 ,,_00 2n(2n - 1) ,
for each x. Hence, by Comment 37.11, the series (a) converges absolutely
for all x. Its interval of convergence is therefore the entire real axis.
We now state a number of relevant theorems in connection with power
series.
Theorem 37.16. If a power series (37.1) converges on an interval I:
Ix - xol < R, where R is a positive constant, then the power series defines
a function f(x) which is continuous for each x in I.
Comment 37.17. If one writes a power series, say
(a) 1 + x + x 2 + + ... ,
which is convergent on I: -1 < x < 1, then by Theorem 37.16, the
series (a) defines a function which is continuous on this interval. The
question naturally arises: which function? This question is, in general,
not easy to answer although it is for (a), because the series is a geometric
one. This series, for each x for which Ixl < 1, converges to 1/(1 - x).
Hence,
1
(b) f (x) = T=x = 1 x x
2
+ + + + ... ,
x
3
Ixl < 1.
For example, when x = t, then f(t) = = 2 and the geometric
1 - 2
series on the right converges to 2. But if x = 2, thenf(x) = 1/(1 - 2) =
-1, and the series on the right of (b) certainly does not converge to -1.
Now consider the power series
x3 x5 x7 X 2.. - 1
(c) X - 3" + 5" - 7" + ... + (_1)"-1 2n - 1 + ... ,
whose interval of convergence is also -1 < x < 1. Hence by Theorem
37.16, it defines a function f(x) which is continuous on this interval. In
this case only one familiar with series might recognize that the series (c)
defines the function Arc tan x. It is a fact, however, that many conver-
gent power series cannot be linked to elementary functions for the very
good reason that many convergent power series do not define elementary
functions.
534 SERIES METHODS Chapter 9

The converse of the question raised above is somewhat easier to answer,

i.e., given a continuous function on an interval I, is there a power series
which defines it? We shall state certain theorems below which not only
will give us the answer to this question, but also will show us at the same
time how to find the defining power series, if the function has one.
Theorem 37.2. If f(x) is defined by a power series, i.e., if
(37.21) f(x) = ao + aleX - xo) + a2(x - XO)2
+ a3(x - xO)3 + ... , I: Ix - xol < R,
then
(37.22) f'(X) = al + 2a2(X - Xo) + 3a3(X - XO)2 + ... ,
I: Ix - xol < R,
i.e., the power series obtained by differentiating each term of (37.21) defines
(or converges to) the derivative of f(x) on the same interval I.
Theorem 37.23. If f(x) and g(x) are defined by power series, i.e., if

(37.231) f(x) = ao + aleX - xo) + a2(x - xo)2 + ... , Ix - xol < R

and
(37.232) g(x) = bo + bl(x - xo) + b2(x - XO)2 + ... , Ix - xol < R,
then f(x) = g(x) if and only if
(37.233) ao = bo, al = bl , a2 = b2, ....
Theorem 37.24. If f(x) is defined by a power series, i.e., if

(37.25) f(x) = ao + aleX - xo) + a2(x - XO)2 + .. .

+ an(x - xo)n + ... , Ix - xol < R,
then
(37.26) al = !'(xo),
f"(xo) an = ....
a2 =2"!'···' n.
Proof. By Theorem 37.2, the successive derivatives of (37.25) are
(a) f'(x) al + 2a2(X - xo) + 3a3(X - XO)2 + ...
=
+ nan(x - xo)n-l + ... ,
f"(x) = 2a2 + 3!a3(x - xo) + ... + n(n - 1)an(x - xo)n-2 + ... ,
f"'(x) = 3!a3 + ... + n(n - 1)(n - 2)an(x - xo)n-3 + ... ,
LessoD 37A REVIEW OF TAYLOR SERIES 535

In (37.25) and (a), set x = Xo in each equation. There results

(b)
f"'(xo) = 3!a3, •.. ,
Hence by (b),
(c) ao = f(xo), at = f'(xo),
f"'(xo) tn)(xo)
a3 = -3-1- , ... , an = -n-I-'

If f(x) is defined by a power series, then by (37.25) and (37.26),

(37.27) f(x) = f(xo) + f'(xo)(x - xo) + (x - XO)2

+ flll(X ) (x - XO)3 + ...

+ t")(x )
(x - xo)" + .. " Ix - xol < R.
If Xo = 0, then (37.27) becomes

(37.28) f(x) = f(O) + f'(O)x + x2 + x3 + ...

f(")(O)
+ -,-
n. x" + .. " Ixl < R.
Definition 37.3. The series on the right of (37.27) is called the Taylor
series expansion of f(x) in powers of (x - xo), or in a neighborhood of Xo.
Definition 37.31. The series on the right of (37.28) is called the
Maclaurin series expansion of f(x) in powers of x, or in a neighborhood
of zero.
Example 37.32. Assume the function f(x) = sin x is defined by a
Maclaurin series (37.28) on some interval. Find the series and its interval
of convergence.
Solution. Taking successive derivatives of f(x) = sin x and evaluat-
ing them at x = 0, we obtain
(a) f(x) = sin x, f(O) = 0,
I'(x) =
cos x, 1'(0) = 1,
I"(x) = -sin x, 1"(0) = 0,
flll(X) = -cos x, 1'''(0) = -1,
f(4)(X) = sin x, fW(O) = 0,
t 5
\x) = COB x, f(5)(0) = 1.
536 SERIES METHODS Chapter 9

Substituting the right-hand values of (a) in (37.28), there results

x3 x5 2n 1
(b) sin x = x - 3T + 5! - " • + (_l)n-l (2nX _ - 1)1 + . " .
Here lunl = x 2n - 1 /(2n - I)!; Therefore

(c)
. IUn+l I . I(2n+1)1
x
2n
(2n - I)! I
X2n-1 = . I(2n)(2nx + 1) I= 0,
+1
2

for each x. Hence by Comment 37.11, the series (b) converges for all x.
Comment 37.33. Theorem 37.24 says that if f(x) is defined by a
power series, then the coefficients in the series are given by (37.26). But
it does not tell us whether f(x) can be defined by a power series. Perhaps
it cannot. Here is an example of a function which cannot be thus defined.
Let
(a) f(x) = e- 1/ z2 , x F- 0
= 0, x = O.
The function f(x) is continuous at x = O. Its derivatives at x = 0 are
[use the definition of the derivative 1'(0) = f(h) -;; f(O)]

(b) f'(O) = 0,1"(0) = 0,1"'(0) = 0, ....

Substituting these values in (37.28), we obtain

(c) f(x) = 0 + Ox + ;1 x 2 + ... = 0 + 0 + 0 + ... ,

-00 <x < 00.

For an x F- 0, the right side of (c) certainly does not converge to the
functionf(x) defined by (a). For example, if x = 0.1, then by (a).!(O.l) =
e- 1/ O. 01 • The series on the right side of (c), however, converges to zero
for all x.
If therefore we start with a continuous function f(x) and obtain a
power series whose coefficients are given by (37.26), we still need a theorem
which will tell us whether the power series thus obtained actually defines
or converges to the given function f(x). For this purpose we introduce the
following theorem.
Theorem 37.34. Taylor'8 Theorem. If a function f(x) has derivative8
of all order8 on an interval I: Ix - xol < h, then

(37.35) f(x) = f(xo) + I'(xo)(x - xo) + f"(x ) (x - XO)2 + '"

+ (x - xo)n + Rn(x),
 Leeson 37B SERIES SOLUTION OF A LINEAR EQUATION 537

where R .. (x) , the remainder term or the sum of all terms after the (n + 1)
term, is given by
(37.36)
and X is between Xo and x.

If R .. (x) - 0 as n - 00, then and only then is

(37.37) f(x) = f(xo) 2fL) (x

+ f'(xo)(x - xo) + f"(x - XO)2 + ... ,
Ix - xol < h,
i.e., the infinite series on the right actually converges to and defines the func-
tionf(x) on I: Ix - xol < h.
Remark. The series on the right of (37.35) is called a Taylor series
with remainder.
Definition 37.4. A function f(x) is said to be analytic at a point
x = Xo if it has a Taylor series expansion in powers of (x - xo) valid
for each x in a neighborhood of Xo.
Definition 37.41. A function f(x) is said to be analytic on an in-
terval if it is analytic at each point of the interval.
Examples of Analytic Functions. It has been proved that each of the
functions listed below is analytic on the interval indicated. Its Taylor
series expansion is given by the series on the right.
2 3
x x
(37.42) e
Z
= l+x+21+3f+"', -00 < x < 00;

x2 x. x6
cos x = 1 - 21 + 4T - 6! + ... , -00 < x < 00;

3 5 7
. x x x
smx = x - 3! + 5! - 7! +"', -00 < x < 00;

x 3 x7
"'3 + "5 - '7 + . . .
XII
Arc tan x = x - -1;a; x ;a; 1.

Comment 37.43. Because of (37.26), an analytic function has one

and only one Taylor series expansion. This fact implies that the same
series results no matter what method is used to obtain it. Verify, for
example, that the series 1/(1 - x) = 1 x + x2 +
Ixl < 1, can + .. "
be obtained by Theorem 37.24 as well as by ordinary division.

LESSON 37B. Solution of Linear Differential Equations by Series

Methods. The series methods we shall describe in this lesson are espe-
cially well suited for finding a solution of the linear differential equation
538 SERIES METHODS Chapter 9

with nonconstant coefficients,

(37.5) yen) + fn_l(x)y(n-l) + ... + II (x)y' + fo(x)y = Q(x).

We shall therefore consider this class of equations first, and then proceed
to a discussion of other types of differential equ!J.tions and to systems of
differential equations.
In Theorem 65.2, we state and prove a sufficient condition for the
existence and uniqueness of a solution of (37.5) satisfying n initial condi-
tions. Here we merely state, without proof, a sufficient condition for the
existence of a power series solution of (37.5).
Theorem 37.51. If each function fo(x), fl(x), ... , f .. -I(x), Q(x) in
(37.5) is analytic at x = xo, i.e., if each function has a Taylor series expan-
sion in powers of (x - xo) valid* for Ix - xol < r, then there is a unique
solution y(x) of (37.5) which is also analytic at x = Xo, satisfying the n
initial conditions

(37.52) y(xo) = ao,

i.e., the solution has a Taylor series expansion in powers of (x - xo) also
valid for Ix - xol < T.
Comment 37.53. A polynomial is a finite series. Hence the series is
valid for all x. If therefore the functionsfo(x), II (x), ... , fn-I (x), Q(x) of
(37.5) are each polynomials, then, by Theorem 37.51, every solution of
(37.5) has a Taylor series expansion valid for all x.
Comment 37.54. An existence theorem tells you only whether a solu-
tion of a differential equation exists. It does not tell you how to find the
solution.
First Series Method. By Successive Differentiations. We shall
illustrate by examples the first method of finding a power series solution
of a linear differential equation.
E%ample 37.541. Find by series methods, a particular solution of the
linear equation
(a) y" - (x + l)y' + x 2
y = x
for which y(O) = 1, y'(O) = 1.
Solution. Comparing (a) with (37.5), we see that fo(x) = x 2 , II (x) =
-x - 1, Q(x) = x. Since all these functions are polynomials, the series
solution we shall obtain, by Comment 37.53, is valid for all x. Because

·We shall use the word "valid" to mean that for each x in a neighborhood of XU the
series expansion of the function converges to the value of the function.
Lesson 37B SERIES SOLUTION OF A LINEAR EQUATION 539

we have been given values of the solution and its derivative when x = 0,
we seek a solution in the fonn of the Maclaurin series (37.28). With I(x)
replaced by y(x) it becomes
"(0) 2 y'''(0) 3
(b) y(x) = yeO) +
y'(O)x y 2! x + x + ----ar-
y(4)(0) 4
+-4-'-x + ....
By the initial conditions, x = 0, y = 1, y' = 1. Substituting these values
in (a) gives,
(c) y"(O) - 1 = 0, y"(O) = 1.

In (b), we now know, by the initial conditions and (c), the values of
yeO), y'(O), y"(O). To find the values of succeeding coefficients, we take
successive derivatives of (a) and evaluate them at (0,1). The next two
derivatives are
(d) y"' - (x + l)y" - + x 2y' + 2xy = 1,
y'
y(4) _ (x + l)y'" - 2y" + x 2y" + 4xy' + 2y = O.

Hence when x = 0, y = 1, y' = 1, y" = 1, we obtain from (d)

(e) y1l1(0) = 3,

Substituting in (b), the initial conditions, (c) and (e), we have

(f)

which gives the first five tenns of a series solution of (a) satisfying the
given initial conditions.
Example 37.55. Find by power series methods, a particular solution
of the linear equation
(a) y" +1 x2 y' - 1 x2 Y = 0, Ixl 1,

for which yeO) = 1, y'(O) = 1.

Solution. Since the initial conditions have been given in tenns of
x = 0, we seek a series solution in powers of x. Comparing (a) with
(37.5), we see that/o(x) = -1/(1 - X2),fl(X) = x/(1 - x 2), Q(x) = O.
The Maclaurin series expansion of -1/(1 - x 2 ) is

(b) - _1_2_ -(1

1 - x -
+ x 2 + X4 + x + ...), 6
Ixl < 1.
540 SERIES METHODS Chapter 9

The Maclaurin series expansion of x/(1 - x 2 ) is also valid for Ixl < 1.
Hence by Theorem 37.51, each solution of (a) has a power series expansion
which is valid for Ixl < 1. By the initial conditions, x = 0, y = 1,
y' = 1. Substituting these values in (a) gives

(c) y"(O) = 1.

In the Maclaurin series (37.28), namely

(d) y(x) = y(O) + y'(O)x + + + y<:[O) X4 + ... ,

we now know, by (c) and the initial conditions, the values of y(O), y'(O),
y"(O). To find the values of succeeding coefficients, we multiply (a) by
(1 - x 2 ), then take its successive derivatives and evaluate them at
(0,1). The next two derivatives are

(e) (1 - X2 )ylll - xy" = 0, (1 - X2 )y<4) - 3 xy lll - y" = O.

Hence when x = 0, y = 1, y' = 1, y" = 1, we find from (e)

(f) ylll(O) = 0,

Substituting in (d) the initial conditions, (c) and (f), we obtain

x X42
(g) y(x) = 1 + x + 2" + 24 + ... , -1 < x < 1,

which gives the first five terms of a series solution of (a) satisfying the
given initial conditions.
Example 37.56. Find by power series methods a particular solution
of the linear equation
(a) ylll + X1 y' -
1
x 2 Y = 0, x;t6 0,

for which y(l) = 1, y'(I) = 0, y"(I) = 1.

Solution. Since the initial conditions have been given in terms of
x = 1, we seek a series solution in powers of x-I. Comparing (a) with
(37.5), we see that fo(x) = -1/x 2 , ft(x) = l/x, Q(x) = O. The series
representation of l/x 2 in powers of x - I is 1 - 2(x - 1) + 3(x - 1)2 -
4(x - 1) 3 + ... ,which is valid for 0 < x < 2. The series representation
of l/x in powers of (x - 1) is 1 - (x - 1) +
(x - 1)2 - (x - 1)3 + ... ,
which is also valid for 0 < x < 2. Hence, by Theorem 37.51, each solu-
tion of (a) has a Taylor series expansion in powers of x - I , valid for
o < x < 2. By the initial conditions, x = 1, y = 1, y' = 0, y" = 1.
Lesson 37B SERIES SOLUTIOlf OF A LINEAR EQUATION 541

Substituting these values in (a) gives

(b) y'"(I) +0 - 1 = 0, y"'(I) = 1.

In the Taylor series (37.27), namely

(c) y(x) = y(l) + y'(I)(x - 1) + (x _ 1)2 + (x - 1)3

+ y(4)(I) (x _ 1)4 + y(O)(l) (x _ 1)5 + ...

4! 5! '
we now know the values of y(I), y'(I), y"(I), y"'(I). To find the values
of succeeding coefficients, we multiply (a) by x 2 , then take its successive
derivatives, and evaluate them at (1,1). The next two derivatives are

(d) + 2xy'" + xy" + y' - y' = 0,

x 2y w
xyW + 2y'" + y" = 0, xy(o) + 3y(4) + y'" = o.
When x = 1, y = 1, y' = 0, y" = 1, y'" = 1, we find from (d)

(e)

Substituting in (c), the initial conditions, (b) and (e), we obtain

(f) y(x) = 1 + (x -; 1)2 + (x 1)3 (x - 1)4 + (x - 1)5 _ ...

8 15 '
o· < x < 2,
which gives the first six terms of a series solution of (a) satisfying the
given initial conditions. -
Second Series Method. Undetermined Coefficients. We outline
a second method of obtaining a series solution, one which does not depend
on taking derivatives. This method will therefore be more useful than the
preceding one whenever it becomes too difficult to obtain successive
derivatives.
Example 37.6. Find by power series methods a particular solution of
the linear equation
(a) y" - (x + l)y' + x 2y = X

for which yeO) = 1, y'(O) = 1.

Solution. This example is the same as 37.541. Hence we know that
(a) has a series solution in powers of x valid for all x. A series solution in
powers of x has the form

(b) y(x) = ao + alX + a2x2 + a3x3 + a4x4 + ....

542 SERIES METHODS Chapter 9

By Theorem 37.2, its two successive derivatives, also valid for all x, are
(c) y'(x) = at + 2a2X + 3a3X2 + 4a4X3 + ... ,
y"(X) = 2a2 + 6a3x + 12a4x2 + ....
Substituting (b) and (c) in (a), we see that y(x) will be a solution of (a) if
(d) 2a2 + 6a3x + 12a4x2 + ...
- (x + 1)(at + 2a2X + 3a3x2 + '4a4x3 + ...)
+
x 2(ao + atX + a2x2 + a3x3 + a4x4 + ... ) = x.
Carrying out the indicated operations in (d) and simplifying the result, we
obtain
(e) (2a2 - at) + (6a3 - 2a2 - at - I)x
+ (12a4 - 3a3 - 2a2 + ao)x 2 + ... = O.
Because of Theorem 37.23 [take g(x) = 0 + Ox + Ox 2 + ... ], equation
(e) will be an identity in x if and only if each coefficient is zero. Hence we
must have
at
(f) a2 = 2'
a3 = 2a2 +6at + 1 .
3a3 + 2a2 - ao
12a4 - 3a3 - 2a2 + ao = 0, a4 = 12 .
By (37.26) and the initial conditions yeO) = I, y'(O) = 1, we know, with
Xo= 0, that
(g) ao = yeO) = I, at = y'(O) = 1.
Hence by (f),
!+1-1
(h) a3 = -1, a4 = 12 = s·1
Substituting (g) and (h) in (b), we obtain
x2 x3 x4
(i) y(x) = 1 + x + 2 + 2 + 8" + ' ' ' , -00 < x < 00,

which agrees with (f) of Example 37.541.

Example 37.61. Find by power series methods a particular solution
of the linear equation
()
a y
1/+ 1 x ,
_ x2 Y - 1 -
1
x2 Y = 0,
for which yeO) = 1, y'(O) = 1.
Lesson 37B SERIES SOLUTION OF A LINEAR EQUATION 543

Solution. This example is the same as 37.55. Hence we know that

(a) has a series solution in powers of x valid for JxJ < 1. A series solution
in powers of x has the form
(b)

By Theorem 37.2, its two successive derivatives, also valid for JxJ < 1,
are
(c) y'(x) = al + 2a2x + 3a3x2 + 4a4x3 + 5aSX4 + ... ,
y"(x) = 2a2 + 6a3x + 12a4x2 + 20aijX3 + ....
The function y(x) of (b) will be a solution of (a) if the substitution of (b)
and (c) in (a) yields an identity in x. Mult.iplying(a) by (1 - x 2 ), then
making these substitutions and simplifying the result, we obtain
(d) (2a2 - ao) + 6a3x + (12a4 - a2)x 2 + (20as - 4a3)X3 + ... = O.

Because of Theorem 37.23, equation (d) will be an identity in x, if and

only if each of its coefficients is zero. Hence we must have
a3
(e) a3 = 0, as = 5"' ....
By (37.26) and the initial conditions y(O) = 1, y'(O) = 1, we know, with
Xo= 0, that
(f) ao = y(O) = 1, al = y'(O) = 1.
Hence by (e) and (f)
(g) a3 = 0, as = 0, ....
Substituting (f) and (g) in (b), we have
2
x
(h) y = 1 + x + 2" + 241 X4 + .. " JxJ < 1,
which agrees with (g) of Example 37.55.
37.62. Find by power series methods a particular solution of
1 1
(a) y"'+-y'--y=O
x x2 ,

for which y(l) = 1, y'(I) = 0, y"(I) = 1.

Solution. This example is the same as 37.56. Hence we know that
(a) has a series solution in powers of (x - 1) valid for 0 < x < 2. A
series solution in powers of (x - 1) has the form
S44 SERIES METHODS Chapter 9

By Theorem 37.2, its three successive derivatives, also valid for 0 < x < 2,
are
(c) y' = al + 2a2(X - 1) + 3as(x - 1)2
+ 4a4(X - 1)s + 5a o(x - 1)4 + ... ,
y" = 2a2 + 3!as(x - 1) + 3 . 4a4(X - 1)2
+ 4 ·5a5(X - 1)s + ... ,
y"l = 3!as + 4!a4(x - 1) + 60a o(x - 1)2 + ....
Substituting (b) and (c) in (a), we obtain, after multiplication by x 2 ,
(d) x2[3!aa + 4!a4(x - 1) + 60a o(x - 1)2 + ... J
+ x[al + 2a2(X - 1) + 3as(x - 1)2
+ 4a4(X - l)s + 5a o(x - 1)4 + ... J
-lao + al(x - 1) + a2(x - 1)2 + ...J = 0:

It will be easier to equate coefficients of like powers of x to zero, if we

express x 2 and x in powers of (x - 1). Their respective series are, see
(38.25),
(e) x2 = 1 + 2(x - 1) + (x - 1)2,
X = 1 + (x - 1).

Substituting (e) in (d), and simplifying the resulting expression, we have

(f) (6as + al - ao) + (24a4 + 12as + 2a2)(X - 1)

+ (60a o + 48a4 + gas + a2)(x - 1)2 + ... = O.
Equating each coefficient in (f) to zero, we obtain

(g) aa = ao ai, a4 = (-12as - 2a2),

at; = t6(-48a4 - gas - a2)'

By (37.26) and the initial conditions, we know, with Xo = 1, that

(h) ao = y(l) = 1, al = y'(I) = 0, a2 = b"(I) = 1.

Hence by (g) and (h)

(i) as = l, a4 = -h(-2 - 1) = -1,

Substituting (h) and (i) in (b), we have

G) Y = 1 + (x - 1)2 + (x - 1)3 _ (x - 1)4 + (x - 1)6 + ...

2 6 8 15 '
o< x < 2,
which agrees with (f) of Example 37.56.
Lesson 37B SERIES SOLUTION OF A LINEAR EQUATION 545

Example 37.621. Find by power series methods, a general solution of

the linear equation
(a) y" + (sin x)y' + e"'y = O.

Solution. All coefficients in (a) are analytic at x = O. We shall

therefore seek a series solution of the form

Since, by (37.42),
S
. x3
(c) sm x = x - 3! + X5! - ... , - 00 < x < 00,

2 3
e'" = 1 + x + x2! + x3! + ... , - 00 < x < 00,

each series solution of (a), by Theorem 37.51, is valid for all x. By Theo-
rem 37.2, the two successive derivatives of (b) are

(d) y' = al + 2a2x + 3a3x2 + 4a4x3 + 5asX' + ... ,

y" = 2a2 + 6a3x + 12a4x2 + 20asx 3 + ....

Making the substitutions (b), (c), and (d) in (a), we obtain

(e) (2a2 + 6a3x + 12a4x2 + 20asx 3 + ...)

S
X - 3! + 5! _ ... (al + 2a2x + 3a 3x + ... )
X3 X ) 2 3 4
+ ( + 4a4x +5asx

+ (1 + x + + + .. .)
X (ao + alx + a2x2 + a3x3 + a4x4 + asxs + ... ) = O.

Performing the indicated operations in (e) and simplifying the result, we

have
(f) (2a2 + ao) + (6a3 + 2al + ao)x + (12a 4 + 3a2 + al + x
2

+ ( 20as + 4a3 + a2 + "3 + 6" x + ... = O.

al ao) 3

Equating each coefficient in (f) to zero, we obtain

(g) a3 = -
2al + ao =
6
al
- "3 - 6" '
ao
546 SERIES METHODS Chapter 9

a3 a2 a1 ao
as = - "5 - 20 - 60 - 120

- ( - _ - (- - -
= ao + a1.
20 20
Substituting these values of the a's in (b), there results

(h) y = ao(1 - ;x 2 - ix 3 + -hx' + hx s + ...)

+ a1(x - !X3 - -hx' tox 5 + + ...), -00 < x < 00,

which gives terms to order five of a general solution of (a). Here ao and
a1 are arbitrary constants.
Comment 37.63. In finding a series solution of a differential equa-
tion, it will usually not be easy to write the general term of the series.
In fact, it will, in most cases, be very difficult if not impossible. However,
in each of the above examples where we stopped with a finite number of
terms, it should be evident to you that for each x in the interval of con-
vergence of the series, y(x) can be computed to a desired degree of accuracy
by using sufficient terms, just as e'" or sin x can be computed to a desired
degree of accuracy from their respective series.

EXERCISE 37
1. Find the interval of convergence of each of the following series.
4 6 2,.
(a) x + =-2! + =-3! + ... + n! + ... .
2

(b) 1 + Sx + S2X2 + ... + S,,-1 X.. -1 + ....

(c) 1 + x + 3 + (x + 3)2 + ... + (x + 3),,-1 + ...

22 32 n 2 •

2. Obtain the first three nonzero terms of the Maclaurin series for each of the
following functions. (a) cos x. (b) e. (c) tan x.
Obtain terms to order k of the particular solution of each of the follow-
ing linear equations, where k is the number shown alongside each equation.
Use both methods of this lesson. Also find an interval of convergence of
each series solution.
3. y' - xy+ x 2 = 0, y(o) = 2, k = 5.
4. x 2 y" = + 1, y(l) = 1, y'(I) = 0, k = 4.
X
5. xy" + x y' - 2y = 0, y(l) = 0, y'(I) = 1, k = 4.
2

6. y" + 3xy' + eZy = 2x, y(o) = 1, y'(O) = -1, k = 4.

7. x 2 y" - 2xy' + (log x)y = 0, y(l) = 0, y'(I) = i, k = 5.
8. (1 - x)y'" - 2xy' + 3y = 0, y(O) = 1, y'(O) = -1, y"(O) = 2, k =6.
Lesson 37-Exercise 547

Obtain tenus to order k, in powers of (x - xo) of the general solution

of each of the following equations, where k and Xo are shown .alongside
each equation. Also find an interval of convergence of each series solution.
9. y" - xy' +
2y = 0, k = 7, Xo = O.
10. 2(x2 +8)y" 2xy' +
(x 2)y = 0, k + + = 4, Xo = O.
ll. xy" +
x 2y' - 2y = 0, k = 4, Xo = 1.
12. y" - xy' - y = sin x, k = 5, xo = O.

ANSWERS 37
1. (a) -co <x < co. (b) Ixl < 1. (c) -4 < x < -2.
2 4 2
2. (a) 1 - 2T + 41
X x
- ..•• (b) 1 + x + x2! + ....
x
3
2x
o
(c) x + "3 + 15 + ....
3 4 x°
3. y = 2 + x 2 -"3
X X
+ "4 - 15 + ... , -co < X < co.

2 (x - 1)3 1)4
4. y = 1 + (x - 1) - 2 + (x -
3 - ... , 0 < x < 2.

5. y = x - I - l(x - 1)2 + !(x - 1)3 - !(x - 1)4 + ... , 0 < x < 2.

x2 5x3 x4
6. y = 1 - x - "2 + 6 + "3 - .. " -co < x < co.

7. y = l(x-l)+l(x-l)2+t<x-l)3_ n(x-l)4+ n(x-l)0_ ..• ,

o < x < 2.
2 x3 x4 XO x6
8. y = 1 - x +x -"2 - 12 - 60 - 48 + ... , -1 < x < 1.
3 7
x x
9. y = ao(1 -
2
x ) + al (
x - "6 - XO
120 - 1680 + . .. , )
-co <X < co.

10. y = ao (1 - -:: + + .. .) + al (x - ;: -
-2V2 <
+ .. .),
x < 20.
2 2(X - 1)3 (x - 1)4 )
ll. y = ao ( 1 + (x - 1) - 3 + 3 + ...
+ a1(x _ 1) _ (x -; 1)2 + (x -; 1)3 _ (x 1)4 + .. .). 0 < x < 2.

12. y = ao (1 + 2 + 8 + ...) +
x
2
x4 al (x + x3 + 15
3
x + ...)
5

3 5
x x
+"6 + 40 + ... , -co < x < co.
548 SERIES METHODS Chapter 9

LESSON 38. Series Solution of y' = f(x,y).

The two methods outlined in the previous lesson for finding a series
solution of a linear equation carry over without change to the finding of
a series solution of a general first order equation
(38.1) y' = !(x,y).

However, the determination of the interval of convergence for which the

series solution is valid is a more difficult task. We shall first give a defini-
tion and then state without proof the relevant theorem we shall need in
t'his connection.
Definition 38.11. A function !(x,y) is analytic at a point (xo,Yo) if it
has a Taylor series expansion in powers of (x - xo) and (y - Yo), valid
in some rectangle Ix - xol < b, Iy - yol < c, i.e., !(x,y) is analytic at
(xo,Yo) if

(38.12) !(x,y) = aoo + [aIO(x - xo) +

aOI(y - Yo»)
+ [a20(x - XO)2 +all(x - xo)(y - Yo) +
a02(y - YO)2) + .. "
is valid for each (x,y) in the rectangle Ix - xol < b, Iy - Yol < c which
has (xo,Yo) at its center.
Comment 38.121. The method for finding the coefficients aij in
(38.12) is essentially the same as that used in Theorem 37.24 to find the
coefficients ai in (37.25). These values of aij are
(38.13) aoo = !(xo,Yo);
a!(xo,Yo) a!(xo,Yo)
alO = ax ' aOI = ay ;
2 2
1 a !(xo,Yo) 2 a !(xo,Yo)
a20 = 21 ax2 ' all = 21 axay ,
2
1 a !(xo,Yo)
a02 = 21 ay2
3
3 a !(xo,Yo)
a21 = 3! ay ,
3
3 a 3!(xo,Yo) _ 1 a !(xo,Yo).
al2 = 3! axa y 2 ' a03 - 31 ay3 '

The symbol a"!(xo,Yo)/ax n means evaluate an!(x,y)/ax" when x = Xo,

y = Yo. Similarly, the symbol an!(xo,Yo)/ayn means evaluate a"!(x,y)/ay"
when x = Xo, Y = yo.
In Theorem 58.5, we state and prove a sufficient condition for the exist-
ence and uniqueness of a particular solution of (38.1) satisfying an initial
Lesson 38 SERIES SOLUTION OF y' - J(x,y) 549

condition. Here we merely state without proof a sufficient condition for

the existence of a power series solution of (38.1).
Theorem 38.14. If the function f(x,y) of (38.1) is analytic at (xo,Yo),
i.e., if f(x,y) has a Taylor series expansion in powers of (x - xo) and
(y - Yo), valid for Ix - xol < r, Iy - Yol < r, and if for every (x,y) in
this 2r X 2r rectangle which has (xo,Yo) at its center,

(38.15) If(x,y) I :1! M,

where M is a positive number, then there is a unique particular solution

y(x) of (38.1), analytic at x = xo, satisfying the initial condition y(xo) = Yo,
i.e., the solution has a Taylor series expansion

(38.16) y(x) = y(xo) + y'(xo)(x - xo) + (x - XO)2

+ af-
y"'(x )
(x - xo) 3 + ... ,
valid in an interval about Xo. This interval is at least equal to

(38.17) I: Ix - xol <

where r is given above and M is given in (38.15).
Remark. This theorem is a special case of the more general theorem
on systems stated in Lesson 39. See Theorem 39.12.
Comment 38.18. Formula (38.17) gives a minimum interval on which
the series in (38.16) converges to the solution y(x) satisfying y(xo) = Yo.
The actual interval of convergence may be larger.
Example 38.2. Find by series methods a particular solution of the
nonlinear first order equation
(a)
for which
(b) yeO) = 1.

Solution. Here f(x,y) of Theorem 38.14 is x 2 y2 and Xo = 0, +

Yo = 1. In Comment 38.21 below, we show how to obtain the Taylor
+
series expansion of x 2 y2 in powers of x and (y - 1). This series is
(c) f(x,y) = x 2 + y2 = 1 + 2(y - 1) + x2 + (y - 1)2.

Since the series is finite, it is valid for all x and y. We may therefore choose
for r of Theorem 38.14 any value we please. For an arbitrary r and with
550 SERIES METHODS Chapter 9

Ixl < r, Iy - 11 < r, (c) becomes

(d) Ix 2 + y21
+ 2r + r2 + r2 = 2r2 + 2r + 1,
< 1

which is the maximum value of x 2 + y2 for all (x,y) in a 2r X 2r rectangle

which has (0,1) at its center. It is therefore the M of (38.15). Hence, by
the theorem there is a unique particular solution y(x) of (a), analytic at
x = xo = 0, satisfying (b), i.e., the solution has a series representation
of the form (38.16), namely

(e) y(x) = y(O) + y'(O)x + x2 + x3 + ... ,

valid in an interval about x = o. By (38.17), this interval is at least
equal to [here r is arbitrary, M is given by (d))

(f) I: Ixl < mint, 3(2r2 ; 2r + 1))·

To maximize I, we let u = r/[3(2r 2 2r + +
1)). Taking its derivative
with respect to r, and setting the result equal to zero, we obtain

(g) 0 = 2r2 + 2r + 1 - r(4r + 2), -2r2 +1= 0, r = -1 .

V2
For this value of r, the interval (f) becomes

1 1
(h) I: Ixl < ----,=----=--=-- ------:::_ < 0.069.
3V2(1 + V2 + 1) 6(1 + V2)

By Theorem 38.14, we now know that the series solution we shall obtain
is valid for at least Ixl < 0.069. We shall find this series solution by the
two methods of Lesson 37B.
First Method. By Successive Differentiations. By (b), the initial condi-
tions are x = 0, y = 1. Equation (a) and its next three successive de-
rivatives, evaluated at (0,1), are

(i) y' = x2 + y2, y'(O) = 0 +1= 1;

y" = 2x + 2yy', y"(O) = 0 +2·1·1 = 2;
y'" = 2 + 2yy" + 2(y,)2, y"'(O) = +
2 2 . 1 . 2 2·1 = 8;+
y(4) = 2yy'" + 6y'y", y4(0) = 2 . 1 . 8 +
6 . 1 . 2 = 28.

Substituting (i) in (e), we obtain

(j) y(x) = 1 + x + x 2 + !X3 + ix 4 + ... ,

Leason 38 SERIES SOLUTION OF y' = f(x,y) 551

which gives the first five terms of the series solution of (a) satisfying (b),
valid at least in the interval (h).
Second Method. Undetermined Coefficients. Since Xo = 0, we seek a
power series solution of the form

(k)

Its derivative, by Theorem 37.2, is

(1) y' = al + 2a2x + 3a3x2 + 4a4x3 + 5aSX4 + ....

Substituting (k) and (1) in (a), it becomes

(m) al + 2a2x + 3a3x2 + 4a4x3 + 5aSX4 + ...

= x 2 + (ao + alx + a2x2 + aax3 + ... )2
= x 2 + ao 2 + 2aOalx + (2a Oa2 + al 2)x 2
+ (2aOa3 + 2ala2)x3 + ....
By Theorem 37.23, (m) will be an identity in x, if the coefficient of each
like power of x is zero. Hence we must have

(n) al - ao 2 = 0,
2a2 - 2aOal = 0,
3a3 - 1 - 2aOa2 - al 2 = 0,
4a4 - 2aOa3 - 2ala2 = o.
The initial condition is y(O) = 1. Therefore by (37.26), with Xo = 0,

(0) ao = y(O) = 1.
By (n) and (0)
(p) al = 1, a2 = 1,

a3 =
1+2+1
3
4
= 3"' a4 = i + 2) = i·
Substituting (0) and (p) in (k), there results

(q) y = 1 + x + x2 + tz;3 + ix'" + ... ,

which is the same as (j) above.

Comment 38.21. We shall obtain the series representation off(x,y) =
x2 +
y2, as given in (c) above, in two ways. First we shall obtain it by
use of (38.13). This is a standard method. Starting withf(x,y) = x 2 y2, +
552 SERIES METHODS Chapter 9

we take the successive partial derivatives called for in (38.13). These are
(38.22) f(x,y) = x2 + y2
ilf = 2x ilf = 2y
ilx ' ily ,

iJx2 = 2, iJxiJy = 0, ily2 = 2.

All additional derivatives are zero. Hence by (38.13) and (38.22), with
xo = 0, Yo = 1, we find
(38.23) aOO = 0 + 1 = 1,
alO = 0, aOl = 2,
1 1
a20 = 2! (2) = 1, all = 0, a02 = 2! (2) = 1.

Substituting these values in (38.12), we obtain with xo = 0, Yo = 1,

(38.24) f(x,y) = x2 + y2 = 1 + 2(y - 1) + x 2 + (y - 1)2,

which is the same as (c) above.

A second method of obtaining (38.24) for this particular function
x2 +
y2, one which is much quicker and simpler, is to observe that x 2 is
already in powers of x and that
(38.25) y2 = (y - 1)2 + 2y - 1 = (y - 1)2 + 2(y - 1) + 1.
Comment 38.26. An easy way to find M in the above example,
without the need of (c), is to observe that

(38.27)

If therefore Ixl < rand Iy - 11 < r so that Iyl < r + 1, then by (38.27),
(38.28)

which is the same as (d) above.

38.29. Find by power series methods a general solution, in
powers of x, of the nonlinear equation
(a)

Find an interval of convergence of the series.

Solution. We shall seek a series solution of the form

(b)
LealOn 38 SERIES SOLUTION OF y' = I(x,y) 553

The derivative of (b), in its interval of convergence is, by Theorem 37.2,

(c)

Substituting (b) and (c) in (a), we obtain

(d) al + 2a2x + 3a3x2 + 4a4x3 + '" = x 2 - ao 2

- 2aOalx - (2aOa2 + al 2)x 2 - (2aOa3 + 2ala2)x3 - ....

Equating coefficients of like powers of x, we have

2a2 = -2aOal, a2 = -aOal = ao 3;

3a3 = 1 - 2aOa2 - al 2, a3 = !(1 - 2ao4 - ao 4) = ! - ao 4;
4a4 = -2aOa3 - 2ala2, a4 = ![-2ao(! - ao 4) - 2(-ao 2)(ao 3)]
+
= -lao ao 5 •
Substituting (e) in (b), there results

(f) y' = ao - a0 2x + a0 3x 2 + (! - a0 4)x 3 + (-fao + a05 )x4 + "',

which gives the first five terms of a series solution of (a). Here ao is the
arbitrary constant. By (37.26), it is, for a particular solution, yeO). To
find an interval of convergence of the series solution, we proceed just as
we did above in Example 38.2. Heref(x,y) = x 2 - y2, Xo = 0, Yo 0 is
arbitrary. The Taylor series expansion of x 2 - y2 in powers of x and
(y - Yo) is, see (38.25),

(g) !(x,y) = x2 - y2 = x 2 - [(y - YO)2 + 2yo(y - Yo) + Y02].

Since the series is finite it is valid for all x and y. We may therefore choose
for the r of Theorem 38.14, any value we please. For an arbitrary rand
with Ixl < r, Iy - Yol < r, (g) becomes

(h) I!(x,y) I = Ix 2 - y21 + I(y - Yo)21

Ix 21
+ 2YolY - Yol + Y02 < r2 + r2 + 2Yor + Y02
= 2(r2 + yor) + Y02,
which is an upper bound of x 2 - y2 for all (x,y) in a 2r X 2r rectangle
which has (O,yo) at its center. It is therefore the M of Theorem 38.14.
Hence, by (38.17), there is an interval I, at least equal to

(i) Ixl < min t, 6(r2 + y:r) + 3Y0 2).

on which the series (f) converges.
554 SERIES METHODS Chapter 9

EXERCISE 38
I. Find a series representation of J(x,y) in powers of x and y if (a) J(x,y)
x sin y, (b) J(x,y) = ye z •
2. Find a series representation of J(x,y) in powers of (x - 1), (y - 1) if
(a) J(x,y) = xy, (b) J(x,y) = y log x.
Obtain terms to order k of the particular solution of each of the fol-
lowing differential equations, where k is the number shown alongside each
equation. Use two methods. Also find an interval of convergence of each
series solution.
3. y' x 2 - y2, y(O) = 1, k = 4.
=
4. y' x 2 - y2, y(l) = 1, k = 4.
=
5. y' =
y2 - xy, y(O) = 2, k = 4.
6. y' x2 = +
sin y, y(O) = 7r/2, k = 3. Hint. Replace sin y by its series
expansion in powers of y (for undetermined coefficient method).
7. y' = x +
e , y(O) = 0, k = 4. Hint. Replace e" by its series expansion
V

in powers of y (for undetermined coefficient method).

8. (1 - x - y)y' = 1, y(l) = -2, k = 4. Hint. Write (1 - x - y)y' as
[-(x - 1) - yly' (for undetermined coefficient method).
9. y' = sin (xy) +x 2, y(O) = 3, k = 4.
10. y' = log (xy), y(l) = 1, k = 5.
II. y' = cos (x +y), y(O) = r/2, k = 3.
12. y' = y2 +
ez - 1, y(l) = 1, k = 3.
13. y' = VI +
xy, y(O) = 1, k = 4.
14. y' = 1 +
xl, y(O) = 1, k = 4.
15. y' = cos x +
sin y, y(O) = 0, k = 5. Hint. Replace cos x and sin y by
their series expansions.
Obtain terms to order k, in powers of x, of the general solution of each
of the following differential equations, where k is shown alongside each
equation.
16. y' = x + y2, k = 3.
17. y' = x + .!., k = 4. Hint. Write the equation as yy' = xy + 1.
y

ANSWERS 38

(b) y (1 + x + x + x + ...) .
2 3
I. (a) x (y _ y3
3!
+ y5
5!
_ ...) .
21 31
2. (a) (x - l)(y - 1) + (x - 1) + (y - 1) + 1.
(x - 1)2 _ 1)3 ]
(b) [1 + (y - 1)] [ (x - 1) - 2 + (x 3 - . .. .
3. y = 1 - x + x 2 - ix3 + fx4 - •• " /x/ < 0.069.
(x - 1)3
4. y = 1 + (x - 1)
2
- 3 + (x - 6 1)4 - .. " Ix - 11 < 0.04.
4
2 40 3 307x
5. y = 2+4x+7x +"'3x Ixl < 0.029.
Lesson 39A SERIES SOLUTION OF A FIRST ORDER SYSTEM 555

'If 1 3 r . 1
6. y = "2 + x +"6x + ... , Ixl < 3(r2 + 1) , r arbItrary, Ixl < 6'
3 4
x 5x
7. y =x+ x
2
+ 2" + 12 + ... , Ixl = 3(r +r er ) , r arbitrary.
x-I 3 2 7 3 79 4
8. y = -2+-2-+ 16(x -1) + 64(x -1) + 1024(x -1) + ....
9. y = 3 + ix 2 + lx 3 - Ix4 ± ....
10. y = 1 + l(x - 1)2 + b(x - 1)4 - rlo(x - 1)5 + ....
2 3
'If X x
1I.y="2-2"+"6+ ....
12. y = 1 + 2(x - 1) + i(x - 1)2+ ¥(x - 1)3 + ....
2 3 4
X X x
13. y = l+x+4"+S- 64+ .. ··
x2 2x 3 x4
14. y = l+x+2"+a+2"+···.
245
IS y = x
. + 2 - 24 - 20 + . .. ' Ixl <!:6' r arbitrary .
16. y = ao + ao 2 x + <t + 3
ao )x 2
+ (lao + ao 4o )x3 + ... ,
Ixl < 3[r2 + (1 +r2yo)r + Y02) , r arbitrary.
3 3 3
17• Y = ao
+ 1
ao x
+ ao - 1
2ao 8 x
2 + 3 6ao
- ao x
3 + tao - 15 4 +
24ao 7 x ....
5

LESSON 39. Series Solution of a Nonlinear Differential

Equation of Order Greater Than One and of a
System of First Order Differential Equations.

LESSON 39A. :;eries Solution of a System of Fit'st Order Differ-

ential Equations. In Theorem 62.12, we state a sufficient condition for
the existence and uniqueness of a solution of a system of first order equa-
tions
(39.1) dYI = II (t,YI,Y2,
de ... , Yn ) ,

dY2
de = !2(t,YI,Y2, ••. , Yn),

satisfying the initial conditions

(39.11)
556 SERIES METHODS Chapter 9

We now state without proof a sufficient condition for the existence of a

power series solution of (39.1) satisfying (39.11).
Theorem 39.12. If each function !I, jz, ... , fn of the system (39.1) is
analytic at a point (to, ai, a2, ... , an), i.e., if each function has a Taylor
series expansion in powers of (t - to), (Yl - al), ... , (Yn - an), valid
for It - tol < r, IYl - all < r, ' .. , IYn - ani < r, and if for every
point (t, Yll Y2, ... , Yn) in this (2r)n+l-dimensional rectangle which has
the point (to, all a2, ... , an) at its center,

(39.13) Ih(t, Yll Y2, ... ,Yn)1 < M, i = 1,2,···, n,

where M is a positive constant, then an interval I exists on which there is one
and only one set of functions, Yl (x), Y2(X), ... , Yn(X), each analytic at
t = to, satisfying the given system (39.1) and the initial conditions (39.11),
i.e., each function has a Taylor series expansion in powers of (t - to), namely

(39.14) Yl(t) = al + all(t - to) + al2(t - to)2 + a13(t - to)3 + ... ,

Y2(t) = a2 + a2l (t - to) + a22(t - t o)2 + a23(t - t o)3 + ... ,

valid in an interval about to. This interval is at least equal to

(39.15) I: It - tol < min f' +'

(n 2)M) ,

where r is given above, n is the number of equations in the system (39.1) and
M is given in (39.13).

The coefficients in (39.14) are given by

y,<i>(t )
o
(39.16) aij = -;-.,_.
For example, the coefficient al2 is, by (39.16), Yl"(t o)/2!. Note that it
agrees with the coefficient of (x - xo)2 in Taylor series (37.27). The
coefficient of a23 is, by (39.16), Y2'II(t O)/31 which agrees with the coeffi-
cient of (x - XO)3 in (37.27).
Remork. If we look at the first order equation dy/dx = f(x,y) as if
it were a system of one equation dyi/dt = !I(t,Yl), we see that Theorem
38.14 is only a special case of Theorem 39.12.
Example 39.17. Find by power series methods a particular solution
of the first order system
dx dy
(a) dt = yt, dt = xy,
Lesson 39A SERIES SOLUTION OF A FIRST ORDER SYSTEM 557

for which
(b) x(O) = 1, yeO) = 1.
Solution. Comparing (a) with (39.1) and (b) with (39.11), we see,
with z and y taking the places of Yl and Y2, that
!I = yt, 12 = xy, to = 0, al = 1, a2 = 1.
By Theorem 39.12, we must find series expansions of II and h in powers
of t, (x - 1), (y - 1). These are respectively
(c) !I = yt = t(y - 1) + t,
h = xy = (x - 1)(y - 1) +x +y - 1
= (x - l)(y - 1) + (x - 1) + (y - 1) + 1.
Since these series are finite, they are valid for all t, x, and y. We may
therefore choose the r of Theorem 39.12 arbitrarily. Hence when It I < r,
Ix - 11 < r, Iy - 11 < r, we have by (c),
(d) I!II < r2 + r,
+ r + r + 1 = r2 + 2r + 1.
1121 < r2
For r > 0, (r2 + 2r + 1) > r2 + r. Therefore r2 + 2r + 1 is the M of
(39.13). Hence, by Theorem 39.12, there is a unique pair of functions
z(t), yet), each analytic at t = 0, satisfying the system (a) and the ini-
tial conditions (b), i.e., each function has a series expansion of the form
(39.14), valid in an interval about t = O. By (39.15), this interval is at
least equal to [here n = 2, M is given by the second equation in (d), and
r is arbitrary)
(e)

We find, in the usual manner, that the maximum value of I occurs when
r = 1. Hence I will be a maximum if
(f) I: It I < it; = 0.0625.
We shall find a series solution of (a) by the usual two methods of Lesson
37B.
First Method. By Successive Differentiations. Here the independent
variable is t, the dependent variables z and y. Since the initial conditions
are given at t = 0, we seek series expansions of z(t) and yet) of the form
(37.28), namely
(g) z(t) = z(O) + z'(O)t + z!'(O) t 2 + x"'(O) t3 + Z(4)(0) t" + ...
2 31 41 '
yet) = yeO) + y'(O)t + y"(O) t 2
2
+ y"'(O)
31
t3 + y(4)(0) t 4
41
+ ... .
558 SERIES METHODS Chapter 9

Starting with each equation in (a) and taking its successive derivatives
with respect to t, we obtain
(h) x' = yt, y' = xy;
x" = y + ty', y" = xy' + yx';
x'" = 2y' + ty", y'" = xy" + 2y'x' + yx";
x(4) = 3y" + ty''', y(4) = xy'" + 3y"x' + 3y'x" + yx"';
......... .. ,
We know from the initial conditions that
(i) x(O) = 1, y(O) = 1.

To find the values of succeeding coefficients in (g), we evaluate the de-

rivatives in (h) in the order x', y', x", y", etc. These are with t = 0,
x = 1, y = 1,
(j) x' (0) = 1.0 = 0, y'(O) = 1·1 = 1;
x"(O) = 1+0 = 1, y"(O) = 1·1 + 1·0 = 1;
x"'(O) = 2+0 = 2, y"'(O) = 1 + 0 + 1 = 2;
X(4)(0) = 3+0 = 3, y(")(O) = 2+0+3+2 = 7;

Substituting (i) and (j) in (g), there results

2 3 4
(k) x(t) = 1 + 2"t + 3"t + 8"t + ... ,
2 3
y(t) = 1 + t + 2"t + "3t + 7t"
24 + ....

which give terms to order four of a series solution of (a) satisfying (b),
valid at least in the interval I of (f).
Second Method. By Undetermined Coefficients. Since the initial condi-
tions are given at t = 0, we seek series expansions of the form
(1) x(t) = + alt + a 2t 2 + a3t3 + a"t 4 + ... ,
ao
y(t) = bo + bIt + b2t2 + b3t3 + b"t" + ... .
Differentiating (1) with respect to t and setting each resulting equation
equal to the respective right side of (a), we obtain
(m) al + 2a2t + 3a3t2 + 4a4t3 + ... = yt,
bl + 2b 2t + 3b 3t2 + 4b 4t 3 + ... = xy.

In (m), replace x and y by their values in (1). Hence (m) becomes

Lesson 39B SERlES SOLUTION OF A LINEAR FIRST ORDER SYSTEM 559

(n) al + 2a zt + 3a3tZ + 4a4P + ... = t(bo + bit + b2t2 + b3t3 + ...),

bl + 2b zt + 3b 3tZ + 4b 4t3 +
= (ao + alt + azt2 + a3t3 + .. ·)(b o + bit + bztZ + ... ).

The pnir of functions in (1) will be a solution of (a) if we choose the a's
and b's so that each equation in (n) is an identity in t. Performing the
indicated expansions in (n) and equating for each equation separately,
coefficients of like powers of t, we obtain
(0) a. = 0,
2az = b o, 2b z= aob l+ albo;
3a3 = bit 3b a= aob z+ albl + azbo;
4a4 = b z, 4b4 = a b3 + a1b z + a2bl + a3b
O O'

By (b) and Theorem 37.24,

(p) ao[= z(O)] = 1, bo [= yeO)] = 1.

Hence from (0) and (p), we obtain, calculating the coefficients in the order
alt bit az, b z, etc.

(q) al = 0, b1 = 1; az = lb o= 1, bz = 1(1 + 0) = !;
a, ,b, " b, M. + 0 + .) ,;
a, ib, t, b. ir; + 0 + • + ,) h·
Substituting (p) and (q) in (I), we find

(e)

which is the same as the solution (k) obtained previously.

LESSON 39B. Series Solution of a System. of Lineal" Fint Order

Equations. In Theorem 62.3 we state and prove a sufficient condition
for the existence and uniqueness of a solution of a system of linear first
order equations

(39.2) = fll(t)Yl + f12(t)yZ + ... + fl"(l)y,, + Ql(t),

= hl(t)Yl + fzz(OYz + ... + iz,,(t)y.. + Qz(t),

" f.,(I)y, + f.,(t)y, + ... + f .. (t)y. + Q.(t),

560 SERIES METHODS Chapter 9

satisfying the initial conditions

(39.21)

We now state without proof a sufficient condition for the existence of a

power series solution of (39.2) satisfying (39.21).
Theorem 39.22. If, in (89.1), each fU'T/Ction hi> i = 1, ... , n, j = 1,
... , n and each fU'TICtion Qi, i = 1, " . , n, is analytic at t = to, i.e., if
eachfu'TICtion has a Taylor series expansion in powers of (t - to), validfor
It - tal < r, then there is a unique set of ju'TICtions, YI(t), Y2(t), ..• , y..(t),
each analytic at t = to, satisfying the given system (89.1) and the initial
conditions '(89.11), i.e., each fU'TICtion has a Taylor series expansion in
powers of (t - to), namely
(39.23) YI(t) = at + a11(t - to) + a12(t - to)2 + a13(t - to)3 + ... ,
Y2(t) = a2 + a2l(t - to) + a22(t - t o)2 + a2S(t - to)3 + ... ,

Yn(t) = a.. + a..l(t - to) + a..2(t - to)2 + a..3(t - toy3 + ... ,

valid for It - tol < r. The coejficient8 in (89.18) are given by
y/,>(t o)
(39.24) aii = -j-r -.
YI"'(tO) Y2'(tO)]
[For example, al3 = -3-'-' a2l = -1-'-' etc.

Example 39.25. Find, by power series methods, a partiCUlar solution

of the first order linear system
ch dy 2
(a) dt = X C08 t, dt = t y,
for which
(b) x(O) = 1, y(O) = -1.

Solution. Comparing (a) and (b) with (39.2) and (39.21), we see,
with x and y taking the place of Yt and Y2, that

(c) f11 = cos t, to = 0,

Since f11 and f22 have Maclaurin series expansions, valid for all t, it
follows by Theorem 39.22, that there is a unique pair of functions x(t>,.
y(t), each analytic at t = 0, satisfying the system (a) and the initial con-
ditions (b), i.e., x(t) and y(t) each has a series expansion in powers of t
valid for all t.
Lel80n 39B SERIES SOLUTION OF A LINEAR FIRST ORDER SYSTEM 561

There are the usual two methods available for finding a series solution
of (a) satisfying (b). These have been described in Example 39.17. We
shall use the more difficult one, in this case, of undetermined coefficients.
By Theorem 39.22, the pair of functions x(t), y(t) have series expansions of
the form (39.23). With to = 0, these become

(d) x(t) = ao + alt + a2t2 + aata + ... ,

y(t) = bo + bIt + b2t2 + bat a + ... .
Differentiating (d) with respect to t, and setting each resulting right side
equal to the respective right side of (a), we obtain

(e) al. + 2a2t + 3a at2 + 4a4ta + ... = x cos t,

bl + 2b 2t + 3b at 2 + 4b 4t a + ... = t 2y.

In (e) replace x and y by their values in (d), and replace cos t, see
(37.42), by its Maclaurin series expansion. There results

(f) + 2a2t + 3a3t2 + 4a4t3 + ...

al
= (ao + alt + a2t2 + ...) (1 - + - .. -),
bl + 2b 2t + 3b 3t2 + 4b 4ta + ... = t 2(b o + bIt + b2t2 + bat 3 + ...).
The pair of functions defined in (d) will be a solution of (a) if we choose
the a's and b's in (f) so that each equation is an identity in t. Performing
the indicated operations in (f) and equating for each equation separately,
coefficients of like powers of t, we obtain
(g) al = ao, bl = 0,
2a2 = aI, 2b 2 = 0,
ao
3aa = a2 -"2' 3b a = bo,
al
4a4 = aa - "2' 4b4 = bl .
By (b) and Theorem 37.24,

(h) ao = x(o) = 1, bo = y(o) = -1.

Hence we obtain from (h) and (g), calculating the coefficients in the order
aI, bI, a2, b2, etc.,

(i) al = 1, bl = 0, a3 = .!(! - !) = 0,
b3 = ·!(-l) = -i, b4 = t(O) = 0.
562 SERIES METHODS Chapter 9

Substituting (h) and (i) in (d), we find

2
(j) x(t) = 1 + t + "2t - 8"t' + ... ,
3
t
yet) = -1 - "3 - ...
which give terms to order four of a series solution of (a) satisfying (b),
valid for all t.

LESSON 39C. Series Solution of a Nonlinear Differential Equa-

tion of Order Greater Than One. In Theorem 62.22, we state and
prove a sufficient condition for the existence and uniqueness of a particular
solution of an nth order differential equation

(39.3) d"y = f(x y y' y" ... yCn-lJ)

dxn ' , '" ,
satisfying a set of initial conditions
(39.31) y(xo) = Yo,
y"(xo) = Y2, ... , yCn-ll(xo) = Yn-l·
The proof consists in showing how every nth order differential equation
with given initial conditions can be changed to an equivalent first order
system with equivalent initial conditions. Hence Theorem 62.22 is a by-
product of Theorem 62.12 relating to first order systems. Analogously,
the sufficient condition we now state for the existence of a power series
solution of (39.3) satisfying (39.31) stems from Theorem 39.12 for a first
order system.
Theorem 39.32. If the function f of (39.3) is analytic at a point
(xO,YO,YI, ... , Yn-l), i.e., if f has a Taylor series expansion in powers of
(x - xo), (y - Yo), (y' - YI), ... , (y(n-I) - Yn-l), valid for Ix - xol < r,
Iy - Yol < r, Iy' - yd < r, ... , Iy(n-I) - Yn-ll < r, and if for every
point (x,y,y', ... , yCn-I)) in this (2r)n+l-dimensional rectangle which has
this point (xO,YO,YI, ... , Yn-l) at its center,
(39.33) If(x,y,y', ... ,y(n-I))I < M,

where M is a positive constant, then there is a unique particular solution of

(39.3) analytic at x = Xo satisfying (39.31), i.e., the solution has a Taylor
series expansion in powers of (x - xo), namely

(39.34) y(x) = y(xo) + y'(xo)(x - xo) + (x - XO)2

+Tylll(X )
(x - XO)3 + ... ,
Lesson 39C SERIES SOLUTION OF A NONLINEAR EQUATION, ORDER n 563

valid in an interval about x = xo. Th?:.· interval is at least equal to

(39.35) I: Ix - xol < min (r, (n; 2)M)'
where r is given above, n is the order of the equation (39.3) and M is given
in (39.33).
Example 39.36. Find by series methods a particular solution of the
nonlinear equation
(a)

for which y(I) = 1, y'(I) = 0, y"(I) = 2.

Solution. Comparing (a) with (39.3) we see that n = 3 and f = x 2
+ y2. A comparison of the initial conditions with (39.31) shows that
Xo = 1, Yo = 1, Yl = 0, and Y2 = 2. The series expansion of f in powers
of (x - 1), (y - 1), y', (y" - 2), by (38.25), is
(b) x 2 + y2 = 2 + 2(x - 1) + 2(y - 1) + (x - 1)2 + (y - 1)2.

Since the series is finite, it is valid for all x and y. We may therefore
choose any value we please for r of Theorem 39.32. For an arbitrary r
and with Ix - 11 < r, Iy - 11 < r, (b) becomes
(c) Ix 2 + y21 < 2 + 2r + 2r + r2 + r2 = 2(r2 + 2r + 1),
which is the M of (39.33).
Hence by Theorem 39.32, there is a unique particular solution y(x) of
(a), analytic at x = 1, satisfying the given initial conditions, i.e., the
solution has a Taylor series expansion of the form (39.34), namely
(d) y(x) = y(I) + y'(I)(x - 1)
+ y"(I) (x _ 1)2 + y"'(I) (x _ 1)3 + ...
2! 31 '
valid in an interval about x = 1. By (39.35), this interval is at least
equal to [here n = 3, M is given by (c) and r is arbitrary]
r
(e) I: Ix - 11 < 5[2(r2 + 2r + 1)] .
To maximize I, we let
r
(f)
u = 1O(r2 + 2r + 1) .
Taking its derivative with respect to r and setting the result equal to
zero, we obtain
(g) 0 = (r2 + 2r + 1) - r(2r + 2) = -r2 + 1, r = 1.
564 SERIES METHODS Chapter 9

For this value of r, the interval Iof (e) becomes

(h) Ix - 11 < :k = 0.025.
By Theorem 39.32, we now know that the series solution we shall obtain
is valid for at least Ix - 11 < 0.025. We shall use the usual two methods
to find a series solution.
First Method. By Successive Differentiations. In (d), we know by the
initial conditions that
(i) y(I) = 1, y'(I) = 0, y"(I) = 2.

To find the values of succeeding coefficients in (d), we evaluate (a) and

its successive derivatives at (1,1). These are, with x = 1, y = 1, y' = 0,
y" = 2, /

(j) y'" = x2 + y2, y"'(I) = 1 + 1 = 2;

y(4) = 2x + 2yy', y(4)(I) = 2 + 0 = 2;
y(5) = 2 + 2yy" + 2(y,)2, y(S)(I) = 2 + 2·2 + 0 = 6;
y(6) = 2yy'" + 6y'y", y(6)(1) = 2·2 + 6·0·2 = 4;

Substituting (i) and (j) in (d), we obtain

(k) y = 1 + (x - 1)2 + -!(x - I)a + n(x - 1)4
+ k(x - 1)11 + Th(x - 1)6 + ... ,
which gives the first seven terms of a series solution of (a) satisfying the
initial conditions, valid at least in the interval (h).
Second Method. By Undetermined Coefficients. Since the initial condi-
tions have been given in terms of x = 1, we seek a series solution of the
form

Its next three successive derivatives are

(m) y' = al + 2a2(X - 1) + 3a a(x - 1)2 + 4a4(X - 1)a
+ 5as(x - 1)4 + ... ,
y" = 2a2 + 6aa(x - 1) + 12a4(x - 1)2 + 20all(x - I)a + ... ,
y'" = 6a a + 24a4(X - 1)+ 60a 5(x - 1)2 + ....
Substituting (1) and the last equation of (m) in (a), we obtain
(n) 6a a + 24a4(X - 1) + 60a5(x - 1)2 + ...
= x 2 + lao + aleX - 1) + a2(x - 1)2 + ...)2.
Leason 39C SERIES SOLUTION OF A NONLINEAR EQUATION, ORDER n 565

It will be easier to equate coefficients of like powers of x, if we find the

series expansion of x 2 in powers of (x - 1). By (38.25),

(0) x2 = 1 + 2(x - 1) + (x - 1)2.

In (n) replace x 2 by its value in (0)., perform the indicated multiplication

on the right, and then equate the coefficients of like powers of (x - 1).
There results
(p) 6a a = ao 2 + 1,
24a4 = 2 + 2aOalJ
60a5 = 1 + 2aOa2 + a1 2.

By Theorem 37.24 and the initial conditions,

(q) ao = 1,

Therefore by (p) and (q)

(r) aa = i = l, as = io-(1 + 2) = to.

Substituting (q) and (r) in (l), we obtain

(s) y = 1 + (x - 1)2 + l(x - 1)3 + hex - 1)4 + to(x - 1)5 + ... ,

which agrees with (k) to terms of order five.

Comment 39.37. As explained in Comment 38.26, we can find M
without the need of (b), by observing that Ix 2 y21 Ix 2 1 ly21. + +
Hence if Ix - 11 < r, so that Ixl < r +
1 and if Iy - 11 < r, so that
Iyl < r + 1, then Ix 2 y21 < (r +
1)2 (r +
1)2 = 2(r2 2r+ +1) + +
as in (c) above.

Example 39.38. Find by series methods, a particular solution of the

nonlinear equation

(a) y'" = (x - 1)2 + y2 + y' - 2

for which

(b) y(l) = 1, y'(I) = 0, y"(I) = 2.

Solution. Comparing (a) with (39.3) we see that n = 3 and f =
(x - 1)2 + +
y2 y' - 2. Comparing (b) with (39.31), we see that
Xo = 1, Yo = 1, Yl = 0, Y2 = 2. The series representation of f in powers
of (x - 1), (y - 1), (y' - 0), (y" - 2), by (38.25), is

(c) f(x,y,y') = (x - 1)2 + 1 + 2(y - 1) + (y - 1)2 + y' - 2,

566 SERIES METHODS Chapter 9

which is a finite series, and therefore valid for all x, y, y'. We may there-
fore choose r of Theorem 39.32 arbitrarily. Hence when Ix - 11 < r,
Iy - 11 < r, Iy'l < r, we have by (c)
(d) If(x,y,y') I < r2 + 1-11 + 2r + r2 + r = 2r2 + 3r + 1,
which is the M of (39.33).
Hence, by Theorem 39.32, there is a unique particular solution y(x) of
(a) analytic at x = 1 satisfying (b), i.e., the solution has a Taylor series
expansion of the form (39.34), namely
(e) y(x) = y(l) + y'(l)(x - 1)
+ y"(l) (x _ 1)2 + y"'(1) (x _ 1)3 + ...
2! 3! '
valid in an interval about x = 1. By (39.35), this interval is at least
equal to [here n = 3, M is given by (d) and r is arbitrary]
r
(f) I: Ix - 11 < + 3r + 1) •
5(2r2

To maximize I, we let u = r/5(2r 2 + 3r + 1), take its derivative with

respect to r and set the result equal to zero. We thus obtain
(g) 0 = (2r2 + 3r + 1) - r(4r + 3) = -2r2 + 1, r = 1/0.
For this value of r, the interval I of (f) becomes
(h) I: Ix - 11 < 0.034.
We shall use the usual two methods of finding a series solution of (a)
satisfying (b).
First Method. By Successive Differentiations. In (e), we know by (b)

(i) y(l) = 1, y'(I) = 0, y"(I) = 2.

To find the values of succeeding coefficients in (e), we evaluate (a) and its
successive derivatives at (1,1). These are, with x = 1, y = 1, y' = 0,
y" = 2,

(j) y'" = (x - 1)2 + y2 + y' - 2, y"'(I) = 1- 2 = -1.

y(4) = 2{x - 1) + 2yy' + y", y(4)(1) = 2.
yes) = 2 + 2yy" + 2(y,)2 + y"', y(S\l) = 2 +4 - 1 = 5.

Substituting (i) and (j) in (e), we obtain

(k) y = 1 + (x - 1)2 - l(x - 1)3 + -h(x - 1)4 + -h(x - 1)5 + ... ,

LeHon39C SERIES SOLVTION OF A NONLINEAR EQUATION, ORDER n 567

which are the first six terms of a series solution of (a) satisfying (b), valid
at least in the interval Iof (h).
Second Method. By Undetermined Coefficients. Since the initial condi-
tions have been given in terms of x = I, we seek a series solution of the
form

Its next three successive derivatives are

(m) y'(x) = at + 2a2(X - + 3a3(X - 1)2 + 4a,(x - 1)3

1)
+5all(x -1)'+···,
y"(x) = 2a2 + 6a 3(x - 1) + 12a4(x - 1)2 + 20all(x - 1)3 + ... ,
y"'(x) = 6a a + 24a,(x - 1) + 6Oa (x -
ll 1)2 + ....
Substituting (1) and the first and last equations of (m) in (a), we obtain

(n) 6aa + 24a,(x - 1) + 6Oall(x - 1)2 + .. ,

= (x - 1)2 + lao + aleX - 1) + a2(x - 1)2 + ...]2
+ al + 2a2(X - 1) + 3a a(x - 1)2 + . " - 2.

Equating coefficients of like powers of (x - 1) in (n), we have

(0) 6a a = ao 2 + al - 2,
24a, = 2aOal + 2a2,
60all = 1 + 2aOa2 + al 2+ 3aa.
By (b) and Theorem 37.24, we know that

(p) ao = I,

Therefore by (0) and (p),

(q) aa = t(1 - 2) = -t,

all = iIr{l + 2 - !) = -k.
Substituting (p) and (q) in (1), we obtain

(r) y(x) = 1 + (x -1)2 - t(x -1)3 +J\(x -1)'+-k(x _1)11+ ... ,

which is the same as that obtained previously in (k).
568 SERIES METHODS Chapter 9

EXERCISE 39
Obtain terms to order k of the particular solution of each of the follow-
ing systems of differential equations, where k is the number shown along-
side each equation. Use two methods. Also find an interval of convergence
of each series solution.
dx
1. dt = e
1
+ y, = e-I + x, x(O) = 0, y(O) = 0, k = 4.

2 dx .
• dt = ysmt, =
2
t +x, x (i) = 2,y(i) = l,k = 4.

dx
3·
dt
2
=t+x, dy
dt
= t -l, x(O) = 0, y(O) = 1, k = 4.

dx
4. dt = y
+.smt, dy
dt = x + cos t, x(O) = 0, y(O)
11'
= 2"' k = 4.

5. See (33.22), population problem.

dx dy
dt = hx - kxy, dt = kxy - py, where h, k, and p are constants,
and x(O) = 1, y(O) = 1, k = 3.
dx dy
6. dt = t
2
+x ,
2
dt = yt, x(O) = -1, y(O) = 1, k = 4.
dx dy
7. dt = xyt, dt = x + t, x(O) = 1, y(O) = -1, k = 4.

Obtain terms to order k of the particular solution of each of the follow-

ing differential equations, where k is shown alongside each equation. Use
two methods. Find an interval of convergence.
8. y" = x 2 - y2, y(O) = 1, y'(O) = 0, k = 6.
9. y" = xy - (y')2, y(O) = 2, y'(O) = 1, k = 4.
10. y" = x 2 + sin y, y(O) = y'(O) = 1, k =. 5. i,
Il. y" = cos x + sin y, y(O) = 0, y'(O) = 1, k = 4.
12. y'" = yy' + xy, y(O) = 0, y'(O) = 1, y"(O) = 2, k = 5.
13. y" = y' log y + x, y(O) = 1, y'(O) = 3, k = 4.
14. y'" = y 2 log x + y', y(l) = 1, y'(l) = 0, y"(l) = 1, k = 5.

ANSWERS 39
Lesson 39-Exercise 569

T
It I < 4(r2 + 3r + 1) .
4 x
.
= -11"2 t + t2 + -12
'11"3
t +...
'
y = -11"2 + t + -'11"213
4 6
'11"4
t + -t + - t + ...
48 '
2
It I < 00.
5. x = 1 + (h _ k)t + h - 2hk + kp t2
2
3
h -3hk 2+ 3hkp-hk2 -kp 2 -kp
2+ 2k3 t3 + ...
+ 3! '
2
Y = 1 + (k _ p)t + P - 2kp + hk t2
2
2 2 3 2 2 3
+hk +3kp -2k t3+ ... .

6. x = -1 + 2 3 4
t - t + !t - it + ... , y = 1+ 1t2 + tt4 + ... .

t2
3 4
3 4 t t
7. x = 1- 1t2 + it + tt + .. " y = -1 + t + "2 - "6 + 12+ ....

8. Y = 1 - 1X2 + tx4 - + .. " Ixl < 4(2r2 +T2T + 1) ,Tarbitrary.

3 4
9. y = 2 + x - 1x2 + ix - ix + .. " Ixl < 4(2r 2 ; 4r+ 1) ,T arbitrary.
2 4 5
11" X X X r .
10. y = "2 + x +"2 + 24 - 40 + .. " Ixl < 4(r 2 + 1) ,r arbitrary.
2 3
n. y = x+ x +
2
+ . ", Ixl < i' r arbitrary.

2 X4 x5 1.
12. y = x + x + 24 + 15 + .. " Ixl < 5(2r + 1) ,r arbitrary.

3 4
13. y = 1 + 3x + ix - ix + ... ,
r
Ixl < 4[(r + 3) log (r + 1) + r] ,r < 1.
2 4 5
14. Y = 1 + (x - 1) + (x - 1) _ (x - 1) + ...
2 12 120 '
r
Ix - 11 < 5[(r + 1)2 log (r + 1) + r] ,r < 1.
570 SERIES METHODS Chapter 9

LESSON 40. Ordinary Points and Singularities of a Linear

Differential Equation. Method of Frobenius.

LESSON 40A. Ordinary Points and Singularities of a Linear Dif-

ferential Equation. As remarked previously, power series methods are
especially well suited for finding solutions of linear differential equations
with nonconstant coefficients. These methods, however, cannot be applied
indiscriminately. For example, if we tried to find a power series solution of
(40.1)
in the form
(40.11)

by the method of successive differentiations, we would run into trouble.

By (37.26), a2 = y"(0)/2!, and when x = 0, we see from (40.1) that y"
and, therefore, a2 do not exist. If we tried to use the method of undeter-
mined coefficients, we would find ao = 0, al = 0, a2 = 0, ....
The trouble arises because after division by x 2 in (40.1), the coefficient
of y' which becomes l/x is not analytic at x = 0, i.e., it does not have
a Maclaurin series expansion in powers of x. Hence the hypothesis of
Theorem 37.51 is not satisfied. We distinguish between points Xo which
satisfy the hypothesis of Theorem 37.51 and those which do not by
means of Definitions 40.2 and 40.22 which follow.
Definition 40.2. A point x = Xo is called an ordinary point of the
linear differential equation
(40.21) y(n) + Fn_1(x)y(n-O + ... + Fl(X)Y' + Fo(x)y = Q(x),
if each function F o, FI, ... , Fn- 1, and Q is analytic at x = Xo. (Remem-
ber this means each function has a Taylor series expansion in powers of
x - Xo valid in a neighborhood of xo.)
By Theorem 37.51, if x = Xo is an ordinary point, then (40.21) has a
solution which is also analytic at x = xo, i.e., the solution has a Taylor
series representation in powers of (x - xo) valid in a neighborhood of Xo.
Definition 40.22. A point x = Xo is called a singularity of (40.21),
if one or more of the functions Fo(x), .. " Fn_1(x), Q(x) is not analytic
at x = Xo.
By Definition 40.22, the point x = 0 is therefore a singularity of (40.1).
For the remainder of this lesson, we shall confine our attention to a.
second order linear equation
(40.23)
Lesson 40A ORDINARY POINT AND SINGULARITY OF LINEAR EQUATION 571

where FI and F2 are continuous functions of x on a common interval I.

Its singularities, if there are any, have been divided into two kinds, regular
singularities and irregular singularities.
Definition 40.24. If x = Xo is a singularity of (40.23) and if the
multiplication of Fl(x) by (x - xo) and of F 2(x) by (x - XO)2 result in
functions, each of which is analytic at x = Xo, then the point x = Xo is
called a regular singularity of (40.23).
Example 40.25. Show that x = 0 and x = 1 are regular singularities of

(a) (x - l)y" + -x1 y' - 2y = O.

Solution. Dividing (a) by (x - 1), we obtain

(b) y
"+ x(x 1- 1) y
, _ _2_ _ 0
x-I Y - .

Comparing (b) with (40.23), we see that

1 F 2 (x) = _ _2_.
(c) FI(X) = x(x - 1) , x-I
By Definition 40.22, x = 0 and x = 1 are singularities of (b).
Consider first the point x = O. Following the instructions in Definition
40.24, we multiply FI by (x - 0) and F2 by (x - 0)2. There results
respectively new functions 1/(x - 1) and -2x 2/(x - 1), each of which
is analytic at x = O. Their Taylor series expansions are in fact
1
(d) x-I = -(1 + x + x 2 + ... ),
- 2x21
x-
= 2(x 2 + x + + ...), Ixl < 1.
3
X4

Hence by Definition 40.24, x = 0 is a regular singularity of (b) and

therefore of (a).
Second we consider the point x = 1. Multiplication of FI by (x - 1)
and of F2 by (x - 1)2 give respectively

(e) x1 and -2(x - 1),

both of which are analytic at x = 1. Their Taylor series expansions are

+
respectively 1 - (x - 1) (x - 1)2 - (x - 1)3 and -2(x - 1). + ... ,
Hence by Definition 40.24, x = 1 is a regular singularity of (b) and there-
fore of (a).
Definition 40.26. If x = Xo is a singularity of (40.23) and if the
multiplication of FI(X) by (x - xo) and F 2(x) by (x - XO)2 result in
572 SERIES METHODS Chapter 9

functions one or both of which are not analytic at x = xo, then the point
x = Xo is called an irregular singularity of (40.23).
E%ample 40.27. Show that x = 0 and x = 1 are irregular singulari-
ties of
(a) (x - I)V' + 2x1 y' + 2y = O.
Solution. Division of (a) by (x - 1)2 gives

(b) Y"+ x2(x 1- 1)2

y'+ (x - 2 1)2 Y = 0.

By Definition (40.22), x = 0 and x = 1 are singularities of (b). Consider

first the point x = O. Here FI = 2( 1 )2' Multiplication of Fl by
x x-I
(x - 0) gives l/x(x - 1)2, which is not analytic at x = O. Hence by
Definition 40.26, x = 0 is an irregular singularity of (b) and therefore of
(a). Second, consider the point x = 1. Multiplication of F I by (x - 1)
gives l/x 2(x - 1) which is not analytic at x = 1. Hence x = 1 is an
irregular singularity of (b) and therefore of (a).

LESSON 40B. Solution of a Homogeneous Linear Differential

Equation About a Regular Singularity. Method of Frobenius. If
the linear equation

(40.3)

has an irregular singularity at x = xo, then the problem of finding a series

solution is too difficult for discussion here. If, however, (40.3) has a
regular singularity at x = xo, then we shall describe a method for finding
a series solution, valid in a neighborhood of Xo. It is known as the method
of Frobenius. The series solution which Frobenius obtained, namely

(40.31) y = (x - xo)m[ao + al(x - xo)

+ a2(x - XO)2 + a3(x - XO)3 + ...], ao 0,

is known as a Frobenius series.

Note that when m = 0 or a positive integer, the series becomes the
usual Taylor series. However, for negative values of m or for nonintegral
positive values of m, (40.31) is not a Taylor series. A Frobenius series,
therefore, includes the Taylor series as a special case.
Assume Xo is a regular singularity of (40.3). It therefore follows by
Definitions 40.22 and 40.24, that F I or F 2 or both are not analytic at
x = Xo, but that (x - XO)Fl and (x - xo)2F2 are. This means that FI
has (x - xo) in its denominator and/or F2 has (x - XO)2 in its denomina-
Lesson 40B REGULAR SINGULARITY. METHOD OF FROBENIUS 513

tor. Hence either Fl(x) = h(x)/(x - xo) or F 2(x) = J2(x)/(x - XO)2 or

both F 1 and F 2 have these respective fonns. In any event, multiplication
of (40.3) by (x - XO)2 will transfonn it into an equation of the fonn

(40.311) (x - xo)2y" + (x - xo)h(x)y' + J2(x)y = 0,

in which bothfl(x) andJ2(x) are now analytic at x = Xo.

The relevant theorem which will assure the existence of a Frobenius
series solution (40.31) of (40.3) is the follo"'ing.
Theorem 40.32. Let Xo be a regular singularity of (40.311). Then
(40.311) has at lea8t one Frobenius series solution of the form (40.31). It is
valid in the common interval of convergence of fl (x) and J2(x) of (40.311),
except perhaps for x = Xo, i.e., if each Taylor series expansion of flex) and
f2(x) is valid in the interval I: Ix - xol < r, then at least one Frobenius
series solution is also valid in I: Ix - xol < r except perhaps for x = Xo.
No loss in generality results if, in (40.311), we take Xo = 0 since by a
translation of axes we can always replace an expansion in powers of
(x - xo) by one in powers of x. With this understanding, we rewrite
(40.311) as

(40.33) x 2y" + xh(x)y' + J2(x)y = 0,

where the functions h (x) andJ2(x) are analytic at x = O. Hence each has
a Taylor series expansion in powers of x valid in a neighborhood of x = O.
Let
(40.34) hex) = bo + blx + b2x2 + ... ,
J2(x) = Co + ClX + C2X2 + ... ,

be their respective series expansions.

The Frobenius series (40.31) and its next two derivatives are, with
Xo = 0,
(40.35) Y = x"'(ao + alX + a2x2 + ... + anxn + ...)
= aox'" + alx",+l + a2x",+2 + ... + a,.x",+n + .. " ao 0,
y' = aomx",-l + al(m + l)x'" + a2(m + 2)x",H + ...
+ an(m + n)x",+n-l + ... ,
y" = aom(m - l)x",-2 + alm(m + I)X",-l
+ a2(m + 1)(m + 2)x'" + ...
+ an(m + n - l)(m + n)x",+n-2 + ....
The function y(x) will be a solution of (40.33) if it satisfies the equation.
Hence SUbstituting (40.35) in (40.33) and replacing at the same time the
574 SERIES METHODS Chapter 9

functions h(x) and f2(X) by their series expansions in (40.34), we obtain

(40.36) x2[aom(m - l)xm- + alm(m + l)x 1 + ...
2 m
-

+ an(m + n - 1)(m + n)xm+,.-2 + .. ·1

+ x(b o + blx + b2x2 +., .. )
X [aomxm - l + al (m + l)xm + ... + an(m + n)xm+n - l + .. ·1
+ (co + CIX + C2 X2 + ...)
X (aoXm + alxm + l + ... + anxm+,. + ...) == o.

Expanding (40.36) and collecting coefficients of like powers of x, there

results
(40.37) ao[m(m - 1) + bom + colxm
+ (al[(m + l)m + bo(m + 1) + col + ao[blm + Clj)Xm+l
+ (a2[(m + 2)(m + 1) + bo(m + 2) + col + al[bl(m + 1) + cil
+ ao[b 2m + C2j) xmH
+ (aa[(m + 3)(m + 2) + bo(m + 3) + col + a2[b l (m + 2) + cil
+ al[b 2(m + 1) + C2] + ao[bam + cal Jxm+3

+ (an[(m+ n)(m+ n - 1) + bo(m+ n) + col

+
+ an_l[bl(m n - 1) cd +
+ an_2[b 2 (m + n - 2) + C2] + ...
+ ao[bnm + Cn ] Jx m +,. + ... == o.
Equation (40.37) will be an identity in x, if each of the coefficients of i',
k = m, ... , m + n is zero. Since we have assumed ao 0, the first co-
efficient in (40.37) will be zero only if
(40.38) m(m - 1) + bom + Co = O.
This equation has been given a special name. It is called the indicial
equation. Since it is a quadratic equation in m, it has two roots. Let
us call these roots ml and m2. These roots may be:
1. Distinct, and their difference not equal to an integer.
2. Distinct, and their difference equal to an integer.
3. The same.
We shall consider each of these possibilities separately.

Case 1. The roots ml and m2 of the indicial equation (40.38) are

distinct and their difference is not an integer. Each of the roots
m = mt, and m = m2 of the indicial equation (40.38) will make the first
coefficient in (40.37) zero. We concentrate on the root mI' Substituting
it for m in the remaining coefficients in (40.37) and setting each equal to
Lesson 40B SOLUTION ABOUT A REGULAR SINGULARITY-CASE 1 575

zero, will enable us to solve each of these equations respectively for a},
a2, ... , an, ... in terms of ao. [Remember the b's and c's are constants
given by (40.34).] Since for this ?til and for this set of values of aI, ••• ,
an, each coefficient in (40.37) is zero, (40.37) is an identity in x. Hence
the substitution of this ml and this set of a's in the first equation of (40.35)
will make y(x) a solution of (40.33).
Following the same procedure outlined above, using the second root m2
we obtain a second set of values of a}, a2, ... , an, ... in terms of ao which
with m2 will make each coefficient in (40.37) zero. Hence the substitution
of this second set of values of the a's and m2 in the first equation of (40.35)
will make y a second solution of (40.33).
Comment 40.381. Not every equation of the form (40.33) has two
independent Frobenius series solutions. Some, as we shall show later,
have only one. If, however, (40.33) has two Frobenius series solutions,
then the following relevant theorem, stated without proof, supplements
Theorem 40.32.
Theorem 40.39. The two Frobenius series solutions of (40.33) are
linearly independent. Each solution is valid for every x in the common in-
terval of convergence of it (x) and f2(x) except perhaps for x = O.
Comment 40.391. Theorem 40.39 can also be stated as follows.
Each Frobenius series solution will converge for every x, except perhaps
for x = 0, in a circle in the complex x plane, whose center is at 0, and
whose radius extends at least to the next nearest singularity of (40.33),
i.e., each solution is valid at least for 0 < Ixl < a, where a is the nearest
singularity to O.
Example 40.392. Find the interval of convergence of the series solu-
tion of
(a) 2
x y" + 1 .; x2 y' - 3y = O.

Solution. Division of (a) by x 2 and application of Definitions 40.22

and 40.24 shows that x = 0 is a regular singularity of (a). Comparing
(a) with (40.33), we see that

(b) f()
I X = 1 +1 x 2 = 1- x
2
+ x4 - x
6
+"', Ixl < 1,
h(x) = -3, -00 < x < 00.

Hence, by Theorem (40.39), each Frobenius series solution of (a) is valid

for Ixl < 1, except perhaps for x = O.
Remark. You can verify that ±i are also regular singularities of
(a). Since ±i are singularities, it follows by Comment 40.391, that each
576 SERIES METHODS Chapter 9

Frobenius series solution will converge for every x, except perhaps for
x = 0, in a circle in the complex x plane of at least unit radius. This
fact may help make clear why the interval of convergence of the series
representation of 1/(1 + x 2 ) is Ixl < 1. Since the function is continuous
for all real x, it would seem that it ought to have a series representation
valid for all real x.
Example 40.393. Find a Frobenius series solution of

(a) xV' + x(x + !)y' - (x 2 + !)y = o.

Division by x 2 and application of Definitions 40.22 and
Solution.
40.24, shows that x = 0 is a regular singularity of (a). We therefore
seek a Frobenius series solution of the form (40.31), namely

(b)

If we wish, we can differentiate (b) twice to obtain y' and y", substitute
these values in (a) and then equate the coefficient of each like power of x
to zero. But since we already did this work in obtaining (40.37), we may
as well make use of this equation. To use it, we need to know the values
of the b's and c's in it. By (40.34), these are the coefficients in the series
representation of h (x) and hex). Comparing (a) with (40.33) we see that

(c)

both of which are already in series form. Hence comparing (c) with
(40.34), we have

(d) bo = !, b1 = 1, Co = -!, C1 = 0,

All remaining b's and c's are zero. With these values of bo and Co, the in-
dicial equation (40.38) becomes

(e) m(m - 1) + 1m - 1 = 0, 2m2 - m - 1 = 0,

whose roots are m = 1, m = -i. Since these roots are distinct and do
not differ by an integer, the method of this case is applicable. Following
the method outlined above, we substitute the root m = 1 in the re-
maining coefficients in (40.37) and solve each for a1, a2, ... , in terms of
ao. The most effective way to use (40.37) is to set the coefficient of xm+n
equal to zero. When you do this, keep in mind that m = 1, bo, blJ Co, ClJ C2
have the values in (d) and that b2 , ba, ... , Ca, C4, ••• are zero. Substituting
these values in the coefficient of xm +.. , we obtain

(f) 0,.[(1 + n)(1 + n - 1) + !(1 + n) - 1]

+ 0,._1[1(1 + n - 1) + 0] + 0,.-2[0 - 1] = o.
 Lenon 40B SOLUTION ABOUT A REGULAR SINGULARITy-CASE 1 577

Solving for an, there results

2
2n + 3n
(g) 2 an = -na..-l + an-2·
Fonnula (g) is a recursion formula. It will give the value of an for each
n 2. Before we can use it, therefore, we must find al. Setting the second
coefficient in (40.37) equal to zero, we have, with m = 1 and the values
of bo, Co, bl , CI as given in (d),

(h) at[(2)(I) + !(1 + 1) - !1 + ao[l + 01 = 0, al = -fao.

We can now use (g) and (h) to obtain, when
(i)
n = 2:
7a2 =- 2a l + ao = "54 ao + ao = "59 ao,
n=3:
27 27 2 41
"2 a3 = -3 a 2 + al = :- 35 ao - "5 ao = - 35 ao,
n=4:

22a, = -4aa + a2 =
328
945 ao + 359 ao = 571
945 ao,

Substituting the value m = 1, and the above values of the a's in (b), we
have
2 9 2 82 a 571, )
(j) Yt = aoX ( 1 - "5 x + 35 x - 945 x + 20,790 x - . . . ,

which are the first five tenns of one series solution of (a). To obtain a
second solution, we use the root m = -! and proceed as above. The re-
cursion fonnula, obtained from the coefficient of x m +n in (40.37) becomes,
with m = -! and the values given in (d),

(k) an[(n - !)(n - i) + !(n - !) + (-;)1

+ an_l[n - i1 + an_2[O - 11 = 0,
which simplifies to

(1)
2n2 - 3n
2 an = - (3)2 an-I +
n - an -2, n 2.

Ail before, we find at by setting the second coefficient in (40.37) equal to

zero. There results

(m) at[!( -i) + (!)(!) - !1 + ao( -i) = 0, at = -ao.

578 SERIES METHODS Chapter 9

Hence from (1) and (m), we obtain, when

(n) n = 2: a2 = -tal + ao = tao + ao = iao;

n = 3: faa = -ia2 + al = -fao - ao = -ljao, aa = -Mao;
n = 4: 10a4 = -faa + a2 = !tao + tao = Wao, a4 = mao.
Substituting m = -! and the values of the above set of
a's in (b), we
obtain for the first five terms of a second series solution of (a)

By Theorem 40.39, the two series solutions are linearly independent.

Hence by Theorem 19.3 and Comment 19.41, a general solution of (a) is
a linear combination of (j) and (0), namely

a
(p) Y = C1X (1 - 52 x + 35
9 2
x - 945 x
82 571
+ 20,790 4
x - ...
)

+ C2 X
-1/2 (
1- x + 23 x 2 13 a
- 18 x
119 4
+ 360 x -...
)
,

where ao has been incorporated into the arbitrary constants Cl and C2.
We observe from (c) that ft(x) and f2(X) are polynomials. Hence, by
Comment 37.53, their series representations are valid for all x. Therefore,
by Theorem 40.39, each Frobenius series solution is valid for all x except
perhaps for x = O. In this example the solution (j) is valid for all x. The
second solution (0), because of the presence of X- l / 2, is valid for all x
except x = O. Hence the general solution (p) is valid for 0 < Ixl < 00.
Note that for x < 0, the second series in (p) is imaginary. We can make
it real by choosing C2 = ci, where C is real.
Case 2. The roots ml> mz of the indicial equation (40.38) differ
by an integer. If the two roots of the indicial equation (40.38) differ by
a nonzero integer, we can write them as m and m N, where N is a +
positive integer. Since m +
N is a root of (40.38), it satisfies this equa-
tion. Hence,
(40.4) (m + N)(m + N - 1) + bo(m + N) + Co = O.

Now compare the left side of (40.4) with the coefficient of an in the last
term of (40.37). They both will be exactly the same if n is replaced by N.
This means that if we use the smaller root m in (40.37) to find a set of values
for the a's which will make the coefficient of each x" zero, we will be
stopped when we reach the term in which aN appears, since its coefficient
will be zero. Hence we cannot solve this equation for aN in terms of previ-
ous a's unless by accident, the remaining terms in the equation also add
to zero. In this case, the equation will be satisfied for any arbitrary value
Lelson 40B SOLUTION ABOUT A REGULAR SINGULARITY-CASE 2A 579

of aN. We can then continue to detennine values of succeeding a's, i.e., of

aN+l, aN+2, ••• in tenns of ao and aN.
If therefore the roots of the indicial equation (40.38) differ by a positive
integer N, two possibilities may occur. Each is considered separately in
Cases 2A and 2B below.

Case 2A. The coefficient of aN in (40.37) is zero and the remain-

ing terms in the coefficient of x m+N also add to zero. In this case the
larger root m + N will detennine, by (40.37), a set of values of the a's in
terms of ao; the smaller root m will determine two sets of values of the a's,
one in terms of ao and the other in terms of aN. However, the Frobenius
series solution obtained by the larger root in tenns of ao will not be linearly
independent of the one in terms of ao obtained by the smaller root. Hence
the smaller root alone will give, in this case, two independent solutions,
whose linear combination will be a general solution of (40.33). It will have
two arbitrary constants ao and aN. These solutions will be valid in the
same intervals given in Theorem 40.39.
Example 40.41. Find a general solution of

(a) xV, + xy' + (X2 - 212) Y = O.

[NOTE. This equation is known as the Bessel equation of index,. A

discussion of the general Bessel equation of index n will be found in Lesson
42.]
Solution. By Definitions 40.22 and 40.24, it can be verified that
x = 0 is a regular singularity of (a). Hence we seek a Frobenius series
solution of the fonn
(b)
Comparing (a) with (40.33), we see that
(c) h(x) = 1 and !2(x) = -t + X2,
both of which are already in series fonn. Hence comparing (c) with (40.34),
we have
(d) bo = 1.
Co = -t,
All remaining b's and c's are zero. Therefore the indicial equation (40.38)
becomes
(e) m(m - 1) +m - i = 0, m2 - != 0,
whose roots are m = , and m = -,. These roots differ by the integer 1.
Hence for this example N = 1. When m = - " which is the smaller
580 SERIES METHODS Chapter 9

root, we obtain, by setting the second coefficient in (40.37) equal to zero

(f) al[(i)(-i) +i - !l + ao(O) = 0, Oal + Oao = O.
The coefficient of al is zero, but the other tenu in its equation is also zero.
Hence this Case 2A is applicable. (NOTE. Since N = 1, aN = al. We
therefore could have anticipated from what we said above, that the co-
efficient of al would be zero. We could not of course have anticipated
that the coefficient of ao would also be zero.) Any values of ao and al will
satisfy the last equation in (f). The remaining a's, namely a2, aa, ...
obtained by equating each subsequent coefficient in (40.37) to zero, will
now be solvable in tenus of an arbitrary ao and an arbitrary al.
With m = - i, the recursion fonuula obtained by setting the co-
efficient of the general tenu xm +n in (40.37) equal to zero, becomes, with
the help of (d),
(g) a,,[(-i + n)(-!+ n) + (-i + n) - !l + an_I(O) + a,,_2(1) = 0,

which simplifies to
(h) (n 2 - n)an = -an -2, n 2.

From it we find (remember ao and al are now arbitrary) when

(i) n = 2: a2 = -iao, n = 3: aa = -iab
n = 4: a4 = --ha2 = -hao, n = 5: a5 = --kaa = ThaI,
n = 6: a6 = -:h;a4 = -..,hao, n = 7: a7 = --J.Ias =
Substituting m = -i and the values of the above a's in (b), it becomes

(j) y = aox-
I 2
/ (1 - x
2
+ X4 - + .. -) x
6

6
x
+ alx l /2 ( 1 - -
X2
6
+ -120 - --
X4
5040
+ . .. , )

which are the first seven tenus of a general series solution of (b). We ob-
serve from (c) that It (x) and !2(x) are polynomials. Hence, by Comment
37.53, their series representations are valid for all x. Therefore, by Theorem
40.39, each Frobenius series solution converges for all x, except perhaps
for x = O. Here the first series converges for all x except x = 0, the
second for all x. The general solution (j) converges for 0 < Ixl < 00.
Note that for x < 0, the first series is imaginary. We can make it real
by choosing ao = ci where c is real.
Case 2B. The coefficient of aN in (40.37) is zero, but the remain-
ing terms in the coefficient of x m +N do not add to zero. In this case
only the larger root m + N of the indicial equation (40.38) will determine
Leason 40B SOLUTION ABOUT A REGULAR SINGULARITY-CASE 2B 581

a set of values of the a's in terms of ao. There will therefore be only one
Frobenius series solution of (40.33).
Comment 40.5. A second independent solution of (40.33), it has
been proved, will be of the form
(40.51) Y2(X) = u(x) - bNY1(X) log x, x > 0,
where N is the positive integral difference between the roots of the indicial
equation (40.38), Yl is a Frobenius series solution of (40.33) obtained with
the larger root m + N, and u(x) is a Frobenius series of the form
(40.52)
In (40.52), m is the amalier of the two roots of (40.38). By substituting
(40.51), (40.52), and the necessary derivatives in (40.33), you will find
that Y2 will be a solution of (40.33), if (see Exercise 40, 14)
(40.53)
If, in (40.53), you now substitute u and its derivatives as determined by
(40.52), the Frobenius series solution Yl and its derivative, then the
resulting equation will enable you to find the values of the b's in (40.52).
We shall, however, not discuss this matter further since logarithmic
solutions, as these solutions are called, have limited applications. See
Exercise 40,14-17.
Example 40.531. Find a solution of
(a) xV' - x(2 - x)y' + (2 + x 2)y = O.
Solution. Division by x 2 and application of Definitions 40.22 and
40.24 shows that x = 0 is a regular singUlarity of (a). Hence we seek a
Frobenius series of the form
(b) y = xm(ao + alX + a2x2 + ...).
Comparing (a) with (40.33), we see that
(c) l1(x) = -2 +x and hex) = 2 + X2,
both of which already are in series form. Hence, comparing (c) with
(40.34), we find
(d) bo = -2, b1 = 1.
Co = 2, Cl = 0,
All the remaining b's and c's are zero. The indicial equation (40.38) be-
comes
(e) m(m - 1) - 2m +2 = 0, m2 - 3m +2 = 0,
whose roots are m = 1 and m = 2. These roots differ by an integer N = 1.
582 SERIES METHODS Chapter 9

Using the smaller root m = 1 and the values in (d), we obtain by setting
the second coefficient in (40.37) equal to zero,
(f) al (2 - 4 + 2) + ao(1 + 0) = 0,

The coefficient of aN = al is zero, but the coefficient of ao is not. Hence

this root m = 1 will not lead to a solution. As remarked at the outset
of this lesson, only the larger root in this case will give a solution.
Therefore using the larger root m = 2, we obtain by setting the second
coefficient in (40.37) equal to zero,

(g) al (3 . 2 - 2 . 3 + 2) + 2ao = 0, al = -ao.

The recursion formula, obtained by setting the coefficient of xm+" in

(40.37) equal to zero becomes, with the help of (d),

(h) an[(2 + n)(1 + n) - 2(2 + n) + 2] + (1 + n)an-l + an_2 = 0,

which simplifies to

(i) (n 2 + n)an = -en + l)a n-l - a,,_2, n 2.

By (g) and (i), when

(j) n = 2:

n = 3:

Substituting m = 2 and the above values of the a's in (b), we have

(k)

which are the first four terms of a Frobenius series solution of (a). By
(c), it and h are polynomials. Their series representations are there-
fore, by Comment 37.53, valid for all x. Hence, by Theorem 40.32, Yl is
also valid for all x.
By Comment 40.5, a second solution of (a) will have the form [here N,
the difference of the roots of (40.38), is one]
(I) Y2(X) = u(x) - b1Yl(X) log x, x > 0,

where u(x) is a series of the form

(m)

See Exercise 40, 15.

Lesson 40B SOLUTION ABOUT A REGULAR SINGULARITY-CASE 3 583

Case 3. Roots of indicial equation (40.38) equal. If the roots of the

indicial equation (40.38) are equal, it is evident that only one set of a's
and therefore only one Frobenius series solution of (40.33) can be obtained
from (40.37).
E:cample 40.6. Find a general solution of

(a) x 2 y" + xy' + x 2y = O.

(NOTE. This equation is known as Bessel's equation of index zero. A fuller

discussion of the general Bessel equation of index n will be found in Lesson
42.]
Solution. Division by x 2 and application of Definitions 40.22 and
40.24, shows that x = 0 is a regular singularity of (a). Hence we seek a
series solution of the form

(b)

Comparing (a) with (40.33), we see that

(c)

both of which are already in series form, valid, by Comment 37.53, for
all x. Comparing (c) with (40.34), we find

(d) bo = 1, Co = 0, Cl = 0,

All the remaining b's and c's are zero. The indicial equation (40.38)
therefore becomes
(e) m(m - 1) +m = 0, m2 = 0,

whose roots are m = 0 twice. Hence we can expect only one Frobenius
solution from these roots.
Using the root m = 0, we obtain from the second coefficient in (40.37),

(f) al(1 + 0) + ao(O) = 0, al = O.

Setting the coefficient of x m +n in (40.37) equal to zero, we obtain, with

the help of (d), the recursion formula

(g) an[(n)(n - 1) + (1)(n)] + an_l(O) + an -2(1) = 0,

which simplifies to

(h)
584 SERIES METHOOS Chapter 9

From (f) and (h), when

1
(i) n = 2: a2 = - 22 ao,
1
n = 3: a3 = - 3 2 al = 0,
1 1
n= 4: a4 = - 4 2 a2 = 2 2 • 4 2 ao,
1
n = 5: as = - 52 a3 = 0,
1 1
n = 6: as = - 6 2 a4 = - 2 2 . 42 . 6 2 ao·

Substituting in (b) these values of the a's and m = 0, we obtain the

Frobenius series solution
6 8
x
•
(J) YI = ao
(X2
1 - 22 + 22X4. 42 - 22 . 42 . 62 + 22 . 42x. 62 . 82 - )
. .. .

By Theorem 40.32 this series is valid for all x.

A second solution of (a) can be found by means of the substitution
(40.51), with N = O. (Remember N is the difference between the roots.)
This solution is the coefficient of C2 in (k) below. Hence the general solu-
tion of (a) is
X2 1 +! 4 1+ ! + i 6
(k) Y = CIYI + C2 ( 22 - 22 . 42 X + 22 . 42 . 6 2 X + ..•

2.;2 ...
1+!+"'+! )
2
+ (-1)"+1 2 x " + ... + yllogx , x > 0, '
where YI is given in (j).

EXERCISE 40
1. Determine the singularities of each of the following differential equations.
Also indicate whether they are regular or irregular singUlarities.
(a) (x - 1)3 x2y" - 2(x - l)xy' - 3y = O.
(b) (x - 1)2xV' +
2(x - l)xy' - Y = O.
(c) (x + 1)2y" +
xy' - (x - l)y = O.
Verify that the origin is a regular singularity of each of the equations
2-6 and that the roots of the indicial equation (40.38) do not differ by an
integer. Find, by the method of Frobenius, two independent solutions of
each equation and intervals of convergence.
2. x 2y" + x(x + i)y' + xy = O.
3. 2x 2y"+ +
3xy' (2x - l)y = O.
4. 2xy" + (x + l)y' + 3y = O.
5. 2x 2y" - xy' + (1 - x 2)y = O.
6. 2(x 2 + x 3 )y" - (x - 3x 2)y' + y - O.
Lesson 4O-Exercise 585

Verify that the origin is a regular singularity of each of the equations

7-9 and that the roots of the indicial equation (40.38) differ by an integer.
Each equation, however, has two independent Frobenius series solutions.
Find these solutions and intervals of convergence.
7. x 2y" - x 2y' + (x 2 - 2)y = O.
8. xV' + (1 + x 3 )xy' - Y = O.
9. x2y" + xy' + (x 2 - l)y = o.
Verify that the origin is a regular singularity of each of the equations
10 and 11, that the roots of the indicial equation (40.38) differ by an
integer, and that each equation has only one Frobenius series solution.
Find this solution and an interval of convergence.
10. x2y" + xy' + (x 2 - l)y = o.
n. x2y" + xy' + (x 2 - 4)y = O.
Verify that the origin is a regular singularity of each of the equations
12 and 13 and that there is only one root of the indicial equation (40.38).
Hence, there is only one Frobenius series solution. Find this solution and
an interval of convergence.
12. x2y" - 3xy' + 4(x +l)y = O.
13. xy" + (1 - x)y'+ !y = O.
14. Prove that the substitution in equation (40.33) of Y2(X), Y2'(X), Y2"(X) as
determined by (40.51) will yield (40.53). Hint. Use the fact that YI(X) is a
solution of (40.33).
15. In Example 40.531, we found one solution of (a), namely

YI(X) = x
2
(1 - x + - ;; - .. -).
By use of (40.51), (40.52), and (40.53), find a second solution of (a). Hint.
See (I) and (m) of the example for the form of the second solution Y2(X)
and of u(x). Substitute u, u', u", YI, YI' in (40.53). The functions/I and/2
are given in (c). Equating coefficients of like powers of x, will determine,
in terms of bo, those values of bl, b2, .•. needed in (I) and (m).
16. Follow the instructions given in problem 15 to find the second solution of (a)
of Example 40.6. It is given in (k). Hint. Each b with odd subscript is zero.
Each b with an even subscript has two in its numerator. Break up the two
+
into 1 1. Two fractions thus result. One gives bOYI, where YI is the solu-
tion (j); the other gives the coefficient of C2 in (k).
17. With the aid of the hint given in problem 15, find the general solution of
each of the following equations.
(a) x 2y" - 3xy' + 4(x + l)y = O. See problem 12 for solution YI(X).
(b) x 2y" - x(2 - 5x)y' + (x - 6x2)y = O.
18. The method of Frobenius may also be applied to find a particular solution of
the nonhomogeneous linear equation
(40.61) xV' +- x/I(x)y' + h(x)y = Q(x),
where/I(x), hex) are analytic at x = 0 and Q(x) is a function that can be
expressed as a Frobenius series.
(40.62)
586 SERIES METHODS Chapter 9

If neither root of the indicial equation (40.38) exceeds n - 1 by a positive

integer, then a trial solution yp(x) will be of the same form as Q(x), namely
yp(x) = x"(ao + +
at a2x 2+ ...). Substituting yp(x) , YP'(x) , 1//'(x) in
the original equation and equating coefficients of like powers of x, will de-
termine ao, at, a2, .. '. For each of the following equations find the roots
of the indicial equation, then verify that these roots do not exceed n - 1
by a positive integer, and finally find a particular solution.
(a) xy" + y' - 2xy = 4x + x 2. Hint. MUltiply the equation by x to
determine n, hex), and hex).
(b) 2x 2y" + xy' + X1/ == sin x+ 2X2 cos x. Hint. Replace sin x and cos x
by their Maclaurin series expansions.
19. A series solution, valid in a neighborhood of x = 0, may not be of much
practical use for large values of x since too many terms in the series may be
needed to obtain a desired degree of accuracy. In such cases, it is best to
make the substitution

1 1 du 1 2
(40.63) u =-,
x
x =-,
u dx= - x2 =-u,
dy dy du 2 dy
dx=dudx =-u du'
2 2
d y -2u du dy _ u2 d y du
dx 2 = dx du du 2 dx
2
= 2u3 dy + u" d y,
2
du du
and solve the resulting equation for y in powers of u. A solution y(u) for a
small value of u in the new equation will then correspond to a solution 1/(x)
for a large value of x in the original equation. Using the above substitutions,
find a Frobenius series solution, first in powers of u, then in powers of l/x,
of each of the follo\\ing.
+ +
(a) 2x 2 (x - l)y" x(3x l)y' - 2y = O.
(b) 2x3y" + + x 2 y' y = o.

(c) xV, + x (5 - ny' + (7

problem 18.
y == +! + :3' yp only. See

Also determine an interval of convergence in each case.

20. A linear differential equation F2(X)y" +
Ft(x)y' +
Fo(x)y = 0 is said to
have a regular singularity at co if the substitution in it of (40.61) results
in an equation which has a regular singularity at u = O. Show that each of
the following equations
(a) x"y" - 3x 2y' +
(1 - x)y == 0,
(b) the LetJendre equation (1 - x 2)y" - 2xy' k(k+ +
l)y = 0,
has a regular singularity at co. In the case of the Legendre equation, find
a Frobenius series solution first in powers of u, then in powers of l/x. Show
that the series is convergent for Ixl > 1.
Lell80n 4O-Exereiee 587

21. The following equation

(40.7) x(1 - x)y" + ['Y - (a + (3 + l)x)y' - a(3'Y = 0,
where a, (3, 'Y are constants, is known as Gauss's equation or as the hyper-
geometric equation. Verify that:
(a) x == 0 is a regular singularity of (40.7).
(b) The roots of the indicial equation (40.38) are m = 0 and m = 1 - 'Y.

(c) a,,+l = (n
(n +
+ mm +
+ a)(n + m + (3)
l)(n + m + 'Y) a", n = 0, 1, 2, ... ,
where ao, ai, a2, ••• , a", ••• are the coefficients in the Frobenius series
+ +
y = x"'(ao alX a2x 2 a3x3 + + ...),
ao '¢ O. Hint. Substitute yand
its derivatives directly in (40.7) and equate to zero the coefficient of x ..+n.
(d) Solutions, using first the root m = 0 and then the root m = 1 - 'Y,
with ao - 1, are respectively
2
(40.71) YI
= F( (3 'Y' ) = 1
a, , ,x
+ a(3'Y x + a(a +'Y('Y1)(3«(3
+ 1)+ 1) x21
+ ... + a(a + 1) ... (a + n - 1)(3«(3 + 1) ... «(3 + n - + ...
1) x"
'Y('Y + 1) .•• ('Y + n - 1) n!'
'Y F 0, -1, -2, -3""

Y2=X
1-'Y[1+ (a-'Y+ 1)«(3-'Y+l)
2-'Y x
+ .. ·]
== xl-'YF(a - + 1, (3 - 'Y + 1,2 - 'Y; x), 'Y '¢ 2,3,4" ...
'Y
The notation F(a - 'Y + 1, (3 - 'Y + 1, 2 - 'Y; x) means, replace in the
YI equation of (40.71), a by a - 'Y + 1, (3 by (3 - 'Y + 1, 'Y by 2 - 'Y.
(e) Both solutions are the same when'Y = 1. Hint. Note that when 'Y = 1,
the root m = 1 - 'Y == O.
(f) If 'Y is not an integer, the general solution of (40.7) is
(40.72)

convergent for Ixl < 1 except perhaps at x = O. If 'Y is an integer,

then only one of Yl or Y2 is a solution of (40.7).
The series on the right of YI in (40.71) is known as the hypergeometric
eeries. By Theorem 37.16, it defines, on the interval Ixl < 1, a fUnction,
called the hypergeometric function, which we have designated by
F(a,(3,'Y; x). If a or (3 is zero or a negative integer, the hypergeometric
series terminates and the hypergeometric function is a polynomial.
22. Using the results obtained in problem 21, find a series solution of each of
the following differential equations.
(a) x(1 - x)y" + (I - 2x)y' - iy = O. Hint. In (40.7), 'Y = I,
a + +
(3 1 = 2, a[:J = i. Therefore a = t, (3 = t, 'Y = i-
(b) x(1 - x)Y" +
(2 - 4x)y' -'- 2y = O. Hint. In (40.7), 'Y = 2,
a + +
(3 1 ... 4, a[:J = 2. Therefore a = 1, (3 = 2, 'Y = 2.
588 SERIES METHODS Chapter 9

23. Show that the substitution

x - Tl du 1
U = ---, -=---,
T2 - Tl dx T2 - Tl

in the equation

(40.74) (x - Tl)(X - T2)Y" + a (x - T3)Y' +aY = 0

will transform it into the Gauss equation (40.7), namely

2
(40.75) u(1 _ u) d y + (Tl - T3) _ u] dy - y = O.
du2 a (Tl - T2) a du a
A series solution of (40.75) is, therefore, see problem 21,
(40.76) y = cIF(a,fJ,'Yj u) + C2u1-'YF(a - 'Y + 1, fJ - 'Y + 1,2 - 'Y; u),
where

(40.77) 'Y = b(Tl - T3) ,

a(Tl - T2)
a + fJ + 1 =
a
afJ = cia.
Hence, by (40.73) and (40.76), a solution of (40.74) is

(40.78) y = qF (a,fJ,'Y; x -
T2 -
Tl)Tl

+ C2 (
X -
--
Tl)l-'Y (
F a - 'Y + 1, fJ - 'Y + 1, 2 - X -
'Yj - - .
Tl)

X - Tl) X - Tl
The notation F ( a,fJ,'Yj - - - means replace x in (40.71) by - --.
T2 - Tl T2 - Tl
24. With the aid of problem 23, find a series solution of each of the following.
(a) (x - 2)(x + l)y" + l(x + l)y' - fy
= O.
. b 7 c 3
Htnt. Tl = 2, T2 = -1, Ta = -1, - = -, - = - -, x = -3u
a 2 a 2
2. +
(b) (x - 2)(x - I)y" + 4xy' +
211 = O.
Hint. Tl = 2, T2 = 1, T3 = 0, a = 1, b = 4, c = 2, x = -u + 2.
25. Prove each of the following identities.

(a) F(I,fJ,fJj x) = 1 x· Hint. See answer to problem 22(b).

(b) F(a,fJ,fJj x) = (1 - x)-a.
(c) xF(I,I,2j -x) = log (1 x). +
(d) xF(!,!,ij x 2) = sin- 1 x. Hint. See answer to problem 22(a).
(e) F(!,-i.!j x2) = (1 - X2)1/2.
Leseon 4O-Answers 589

26. With the aid of problem 23, find a series solution of the following equation,
known as Tschebyscheft"s equation,
(40.79) (x - l)(x + l)y" + xY' - = O.
Hint Tl
a =. n, tJ
=
=
1 T2 =
'Y =
-1 T3
i. = 0 a
' = 1 b
, = 1
, C = -n 2 x
, = -2u + 1,
27. With the aid of (40.71) and the answer given in problem 26, find the
Tschebyscheff polynomials To(x), Tl(X), T2(X), T3(X), T,(x).

ANSWERS 40
1. (a) x = 0 regular singularity, x = 1 irregular singularity. (b) x = 0
irregular singularity, x = 1 regular singularity. (c) x = -1 irregular
singularity.
2 3
2. Yl = ctX 112 ( 1 - x + x2! - x3! + . ..) ,all x;
Y2 = c2(1 - 2x + tx2 - -h;x 3 + ...), all x.
3. Yl = CIx1l2(1 - !x + + ...), all x;
- mx3
Y2 = C2X-1(1 + 2x - + + ...), x F O.
2x2 tx 3

4. Yl = Cl (1 - 3x + 2x2 - + .. -) , all x;
3
7x llx
Y2 = C2X
112 (
1 - "6 + 21x2
40 - 80 + . ..) ,all x •.

5. Yl = ClX (1 + + :a: + .. -), all x;

Y2 = C2X
1l2
(1 + x
6
2
+ 168 + ...) ,all x.
6. Y = (ctX1l2 + c2x)(1 - x + x 2 - x 3 + ... ), Ixl < 1.
7. Yl = CIX- 1 (1 + tX + tX2 - ix' - ... ), x F 0;
Y2 = c2x2(1 + tX + kx 2 - nx3 - ••• ), all x.

8. Yl = CIX (1 - x
15
3
+ 180 - ...), all X''
Y2 = C2X -1 (1 + - ;: + .. -) , x FO.
9. Yl = CIX
1l3
[1 - 1 ! + 2(1 + + !) -...J' all x;

Y2 = C2X-
1/3
[1 - 1 ! + 2(1 _ _ !) -.. J x F O.

10. Y = CX[1 - + + .. J all x.

11. Y = cx
2
[1 _! (=)2 + _1 (=)' + ...J'
3 2 2(3)(4) 2
all x
.
12. y(x) = cx 2(1 - 4x + 4x 2 -lh + tx' 3 - ••• ), all x.
590 8BlUE8 METBOD,8 Chapter 9

13. Y = C (1 - - x
2
- x
3
+ .. -), all x.
15. Y = bo[x(1 +x - ix 2 +
Hx 3 - ••• ) - Yllog xl, x > o.
17. (a) Y = Yl + 2
C2[x (8x - 12x2 Wx3 + + ...) + Yllog xl, where Yl is
given in problem 12.
(b) Yl = Clx 3(1 - 4x + 9x 2 - .•. ),
Y2 = c2[(1 +
!x - ix 2 - ljx 3 + ...) + ljyl log xl.
2 3
2 (
18. (a) YP = x 1 9 + +"8 +
x x 2x
225
)
+ ... .
(b) yp = x (1 + 6
_ x
45
2
_ ••• ) .

3
19. (a) Y = Cl ( 1 + 2u + 37u2 + 45
112u
+ ...)
+ C2ul/2 (1 + 4u + 22u + 484u + ...) ,
2 3
3 15 315
2 7 112 )
Y = Cl (1 +; + 3x2 + 45x 3 + ...
22
+ C2X
-112 (
1 + 3x4 + 15x2 484
+ 315x )
3 + . .. .
The u series converges for lui < 1. Hence the x series converges for
Ixl > 1.
(b) Y = CI(I- -

l/2
+ C2x (1 - - 9;x 3 + .. .), x F- O.
(c) YP = (; + :2 + + .. .), x F- O.
k[ k(k - 1) -2 k(k - 1)(k - 2)(k - 3) -4
20. Yk(X) = aox 1 - 2(2k _ 1) x + 22. 2!(2k _ 1)(2k _ 3) x + .. .
+ (-I)nk(k - l)(k - 2)(k - 3) ... (k - 2n + 1) x-2.. + ...J
2 nn!(2k - 1)(2k - 3) ... (2k - 2n + 1)
+ alx -(k+l) [1 + (k + 1)(k + 2) -2
2(2k + 3) x

+ (k + 1)(k + 2)(k + 3)(k + 4) -4 + ...

2 2 . 21(2k + 3)(2k + 5) x
(k + 1)(k + 2)(k + 3)··· (k.+ 2n)
+ 2nn!(2k + 3)(2k + 5) ... (2k + 2n + 1) x
-2.. J
. .. .
22. (a) Y = clF(!,t,ti x) + C2X-1/2F(O,O,!i x)
2 3
= Cl ( 1
3x
+ 6X + 40 5x
+ 112 + ...) + C2X -1/2•
cIF(I,2,2i x) = cl(l + x + x + ...) = _Cl_.
2
(b) Yl =
1 - x
Lesson 41A THE LEGENDRE DIFFERENTIAL EQUATION 591

2
You can verify that Y2 = C2X-IF(I,2,2; x) = x(1 c_ x) is also a solution.
Here a = 1, fJ = 2, 'Y = 2. Note that in this special case, the second and
subsequent numerators and denominators of the solution Y2 of (40.71) are
both zero.
24. (a) The transformed equation is

+ (72 - 27) 3
2
d y dy
u(1 - u) du 2 u du + 2Y = o.

Here a = 3, fJ = -i, 'Y = I·

y = cI F(3,-i,1; u) + C2ul-7/2F(!,_3,_!; u)
y = clF (3,-i, I; 2-; x) + C2 (2 -; x)-5/2 F (i,-3,-I; 2-; x).
(b) y = cIF(I,2,8; 2 - x) + c2(2 - x)-7F(-6,-5,-6; 2 - x).

26. y,,(x) = clF (n,-n,i; 1 -; x)

+C2 (1 -; x) 112 F(n+ i, -n+ i,i; 1-; x).
27. To(x) = 1, TI(x) = x, T2(X) = 2x2 - 1, Ta(x) = 4xa - 3x,
T4(X) = 8x4 - 8x 2 1. +
LESSON 41. The Legendre Differential Equation. Legendre
Functions. Legendre Polynomials P,,(x).
Properties of Legendre Polynomials P,,(x).

LESSON 41A. The Legendre Differential Equation. The linear

differential equation
(41.1) (1 - x 2 )y" - 2xy' + k(k + l)y = 0,

where k is a real constant, occurs in many physical problems. It is known as

the Legendre equation, named after the French mathematician A. M.
Legendre (1752-1833). Because of its great practical importance, this
equation has been studied by many mathematicians and a considerable
amount of literature is available on the subject.
You can verify by Definition 40.2, that x = 0 is an ordinary point of
(41.1). This equation can therefore be solved by the power series method
outlined in Lesson 37B. Dividing (41.1) by 1 - x 2 and then comparing
the resulting equation with (37.5), we see that

(41.11) fo{x) = kik! = k{k + 1)(1 + x 2 + X4 + ...),

fl{x) = -
1 - x2
= -2x(1 + x 2 + x4 + ... ).
Both series are valid for Ixl < 1. Hence by Theorem 37.51, (41.1) has a
592 SERIES METHODS Chapter 9

series solution in powers of x which is also valid for Ixl < 1. We therefore
seek a Maclaurin series solution of the form

(41.12) y(x) = ao + alX + a2x2 + a3x3 + ...

+ a"x" + a,,+lx.. + 1 + a,.+2X .. +2 + ....

Successive differentiations of (41.12) give

(41.13) y'(x) = al +
2a2x +3a3x2 + ...
+ na,.x.. - 1
+ + + +
(n l)a.. +lx" (n 2)a,.+2X"H + ... ,
y"(X) = 2a2 +
2 . 3asx + ... + n(n - 1)a..x .. - 2
+ n(n + 1)a,,+lx 1 + (n + l)(n + 2)a,.+2x" + ....
n
-

As usual y will be a solution of (41.1), if the substitution in it of (41.12)

and (41.13) is an identity in x. Making these substitutions, we obtain

(41.14) (1 - X2)[2a2 + 2 . 3a3x + ... + n(n - 1)a"x..- 2

+ n(n+ 1)a.. +lx.. - 1 + (n + l)(n + 2)a,.+2X"· . ·1
- 2x[al + 2a2x + 3a3x2 + ... + na,.x .. - 1 •• ·1
+ k(k + l)[ao + alX + a2x2 + ... + a..x" + .. ·1 = o.

Expanding and collecting coefficients of like powers of x, we have

(41.15) [2a2 + k(k + 1)ao1 + [3!a3 - 2al + k(k + 1)a11x

+ [12a4 - 6a2 + k(k + 1)a21x2 + ...
+ [en + l)(n + 2)a"+2 - n(n + l)a,. + k(k + l)a"jx" = O.
Hence we must choose the set of a's so that
(41.16) k(k + l)ao = 0,
2 . 3a3 - 2al + k(k + l)al = 0,
3·4a4 - 2· 3a2 + k(k + 1)a2 = 0,
4·5a5 - 3· 4a3 + k(k + 1)a3 = 0,
5 . 6a6 - 4 . 5a4 + k(k + 1)a4 = 0,

(n + l)(n + 2)a"+2 - n(n + l)a.. + k(k + l)a" = O.

Therefore, by (41.16),

-k(k + 1)
(41.17) , a2 = 2 ao,

a3 =
2 - k(k
31
+ 1) al = -
(k - l)(k
31
+ 2) all
Leuon 4lB LEGENDRE POLYNOMIALS Pk(X) 593

6 - k(k + 1) (k - 2)(k + 3)
a4 = 12 a2 = - 12 a2
k(k + I)(k - 2)(k + 3)
= 41 ao,
12 - k(k + 1) (k - 3)(k + 4)
as = 20 a3 = - 20 a3

(k - I)(k + 2)(k - 3)(k + 4)

= 51 aI,

a6 =
20 - k(k
30
+ 1) a4 = -
(k - 4)(k
30
+ 5) a4
-k(k + I)(k - 2)(k + 3)(k - 4)(k + 5)
= 61 ao,

(-I)"k(k + I)(k - 2)(k + 3)(k - 4)(k + 5)··· (k + 2n - 1)

a2" = (2n) I ao,
(-I)"(k -I)(k+2)(k - 3)(k+4)(k - 5)(k+6)··· (k+2n)
a2"+1 = (2n + I)! al·
Substituting (41.17) in (41.12), we obtain

(41.18)

where ao and al are arbitrary, and

(41.2)
S,,(x) = [1 _k(k 21+ 1) x2 + k(k + I)(k 41- 2)(k + 3) X4 + ...
... (k+2n-I) 2"+ ... J
+ (-I)"k(k+ l)(k-2)(k+3)(k-4)
(2n) 1 x,
(41.21)
T,,(x) = [x _ (k - l)(k + 2) x3 + (k - Ij(k + 2)(k - 3)(k + 4) xS + ...
31 51
+ (-I)"(k - l)(k + 2)(k - 3)(k + 4)· .. (k + 2n) 2,,+1 + ...J.
(2n+l)1 x

LESSON 41B. Comments on the Solution (41.18) of the Legendre

Equation (41.1). Legendre Functions. Legendre Polynomials
P,,(:c). The following comments apply to the solution (41.18).
1. The solution (41.18) is valid for Ixl < 1. See remarks immediately
following (41.11).
2. When k is zero or a positive even integer, 2, 4, 6, ... , or a negative odd
integer, -I, -3"", the series in (41.2) terminates, i.e., it becomes a
594 SERIES METHODS Chapter 9

polynomial in x. For example, when k = 2, the third and all succeeding

terms in (41.2) are zero; when k = -1, the second and all succeeding
terms are zero.
3. When k is a positive odd integer, 1,3,5, ... , or a negative even integer,
-2, -4, ... , the series in (41.21) terminates, i.e., it becomes a poly-
nomial in x. For example, when k = 3, the third and all succeeding
terms in (41.21) are zero. When k = -2, the second and all succeeding
terms are zero.
4. If k is not an integer or zero, both series (41.2) and (41.21) are non-
terminating. By Theorem 37.16, therefore, each defines, for each k, a
continuous function on I: Ixl < 1. These functions are called Legendre
functions.
5. Since a Taylor series is a special case of a Frobenius series, the Legendre
functions (or a Legendre function and a polynomial) defined by the
series (41.2) and (41.21) are, by Theorem 40.39, linearly independent.
And since each is a solution of (41.1), it follows by Theorem 19.3 and
Comment 19.41, that y(x) of (41.18) is 'a general solution of (41.1) on
I: Ixl < 1.
Because of Comments 2 and 3 above, we can write polynomial solutions
of the Legendre differential equation (41.1) when k is zero or an integer.
All we need do is select the proper series (41.2) or (41.21) for the given k.
In Table 41.22, we have listed solutions of (41.1) for a few integer values
of k and for k = o.
Table 41.22

A Solution by
Uk= The Legendre Equation (41.1) Becomes
(41.2) or (41.21) Is

o (1 - x 2 )y" - 2xy' = 0 yo(x) = 1

(1 - x 2 )y" - 2xy' + 2y = 0 Yl(X) = x
2 (1 - x 2 )y" - 2xy' + 6y = 0 Y2(X) = 1 - 3x 2
3 (1 - x 2 )y" - 2xy' + 12y = 0 Y3(X) = x - !x3
4 (1 - x 2 )y" - 2xy' + 20y = 0 Y4(X) = 1 - 10x 2 + 3lx 4
If we solve the last equation in (41.16) for an +2, we obtain
[n(n + 1) - k(k + 1)]
(41.23) an+2 = + l)(n + 2) an,
(n n = 0, 1, 2, 3,

Since the numerator of (41.23) is equal to -(k - n)(k + n + 1), we can

write (41.23) as
(k - n)(k + +
n 1)
(41.24) an+2 = - (n+ 1)(n+ 2) an·
Lesson 41B LEGENDRE POLYNOMIALS Pk(x) 595

Therefore by (41.24), if k is an integer, and

2(2k - I)
(41.25) n = k - 2: ak = - k(k _ I) ak-2,
4(2k - 3)
n = k - 4: ak-2 = - (k _ 2) (k _ 3) ak-4,

Therefore by (41.25),
k(k - I)
(41.26) ak-2 = - 2(2k _ I) ak.

(k - 2)(k - 3) k(k - I)(k - 2)(k - 3)

ak-4 = - 4(2k _ 3) ak-2 = 2. 4(2k - 1)(2k - 3) ak,

Substituting (41.26) in (41.12), we obtain polynomial solutions Pk(X) of

Legendre's equation (41.1) of the form [write y(x) in descending powers
of x, namely y(x) = ... + + ak_4xk - 4 + ... J
(41.27)
k k(k - I) k-2 k(k - I)(k - 2)(k - 3) k-4
Pk(X) = ak [ x - 2(2k _ I) x + 2. 4(2k _ 1)(2k _ 3) x
_ k(k - l)(k - 2)(k - 3)(k - 4)(k - 5) X k- 6 + ...J,
2·4· 6(2k - 1)(2k - 3)(2k - 5)
k = 0, 1,2,"',
where ak is an arbitrary constant. Let us take for it the particular value
(41.3)
(2k)! 1·2·3·4 ... (2k - 1)(2k) 1·3·5 ... (2k - I)
ak=--=
2k(k!)2 k!2' 4·6· .. (2k)
= k!
.
(The reason for selecting this value of ak will be made clear in Lesson
4IC-A.) To distinguish between the polynomial solutions Pk(X) of (41.27)
and those we shall obtain by giving ak the value (41.3), we replace Pk
by Pk. Hence by (41.3) and (41.27)

(41 .31) P ( ) 1 . 3 . 5· .. (2k - I) [k k(k - 1) k-2

k X = kf x - 2(2k _ 1) x

+ k(k - I)(k - 2)(k - 3) k-4

2. 4(2k - I)(2k - 3) x
_ k(k - I)(k - 2)(k - 3)(k - 4)(k - 5) xk- 6 + ...
2·4· 6(2k - 1)(2k - 3)(2k - 5)
+ (-Itk(k - I)(k - 2) ... (k - 2n + 1) Xk- 2n + ...J,
2nn!(2k - I)(2k - 3) ... (2k - 2n + 1)
k = 0, 1,2, ....
596 SERIES METHODS Chapter 9

By definition, 81 = 1 ·2·3· .. 8. Hence,

(41.311) (8 + 1)(8 + 2) ... (8 + n)

1 . 2· 3· .. 8(8 + 1) ... (8 + n) (8 + n) 1
= 1 . 2· 3 ... 8 = 81 •

By (41.311)

(41.312) (28 + 2)(28 + 4) ... (28 + 2n)

= 2"(8 + 1) ... (8 + n) = 2" (8 +81 n)! .
By· (4.311) and (41.312),
(41.313) (28+ 1)(28 + 3) ... (28 + 2n - 1)
(28 + 1)(28 + 2)(28 + 3)(28 + 4) ... (28 + 2n - 1)(28 + 2n)
= (28 + 2) (28 + 4) . . . (28 + 2n)
(28 + 2n)!
(28)! 8!(28+2n)!
= 2"(8 + n)! = 2"(28)!(8 + n)!·
8!

In (41.313), let 8 = k - n. There results

(k - n) !(2k) I
(41.314) (2k - 2n + 1)(2k - 2n + 3) ... (2k - 1) = 2"(2k _ 2n) !kl
By the definition of k!,
kl
(41.315) k(k - l)(k - 2) ... (k - 2n + 1) = "(k-----;:-2n7':)
I

Substituting (41.315) and (41.314) in the coefficient of :J!<-2" of (41.31),

we obtain for the coefficient
(-I)"kl 2"(2k - 2n)lk! (-I)"(k!)2(2k - 2n)!
(41.316)
(k - 2n) I 2"n!(k - n) !(2k)! = n!(2k) !(k - 2n) !(k - n)! .

Since (41.316) is the coefficient of the general term of (41.31), we can write
(41.31), with 1·3·5··· (2k - 1)/kl replaced by its equivalent fac-
torial form as given in (41.3) after at, as

1 (-I)"(2k - 2n)! iH"

(41.317) P/c(x) = !=o 21: nICk - 2n) I(k - n)! x .

The symbol [k/2] means the largest integer in k/2. For example, if k = 3,
the largest integer in k/2 is one. Hence to evaluate Pa(x), replace k by 3
in (41.317). and sum the two terms obtained with n = 0 and n = 1.
If k = 4, ilie largest integer in k/2 is 2.
Lesson 41B LEGENDRE POLYNOMIALS Pk(X) 597

By (41.31) or (41.317), when

(41.32) k = 0: Po(x) = 1.
k = 1: P 1 (x) = x.

k = 2: P 2 (x) = 3
1; (X2 - D= (3x
2
- 1).

k = 3: P 3(x)
1.3.5(3
= ----ay-- x -
3X)
5" = 21(5x
3
- 3x).

k= 4: P 4(x) = 1· 7(X4 - 6;2 + ;5)

=8 (35x4 - 30x 2 + 3).
The polynomials in (41.32) are known as Legendre polynomials.

If in the recursion formula (41.24), we let n = -k - 3, -k - 5, etc.,

we obtain, in place of (41.27),

(41.33) qk (X)
_
- a-k-l
[
x
-k-l + (k 2(2k
+ l)(k
+ 3)+ 2) x -k-3
+ (k + 1)(k + 2)(k + 3)(k + 4) -k-5 + ...J.
2· 4(2k + 3)(2k + 5) x

Since a-k-l is an arbitrary constant we may take for it the particular value
2k(k!)2
(41.34) a_k-l = (2k + I)!
k!2 . 4 . 6· .. (2k) k!
= 1·2·3··· (2k)(2k 1) + = 1·3·5··· (2k + 1) ,
k -1, -2,···.
The functions
k!
(41.35) Qk(X) = 1.3.5 ... (2k + 1)
-k-l + (k + 1)(k + 2) -k-3
X[
x 2(2k + 3) x
+ (k + l)(k + 2)(k + 3)(k + 4) -k-5 + ...J
2 . 4(2k + 3)(2k + 5) x ,
k -1, -2,"',

obtained from (41.33) by replacing a-k-l by its value in (41.34) are

known as Legendre functions of the second kind.
598 SERIES METHODS Chapter 9

LESSON 4-Ie. Properties of Legendre Polynomials P,,(x). The

Legendre polynomials are important primarily in the interval I: Ixl 1.
The graphs of the first five Legendre polynomials in the interval 0 x 1
are shown in Fig. 41.4.

Figure 41.4

It has been proved that a Legendre polynomial of degree n has exactly

n distinct real zeros in the interval -1 x 1. If we had extended the
graphs shown in Fig. 41.4 to include the interval -1 x 0, we would
have found that Po(x) has no zeros, P1(x) has one zero, P 2(x) has two
zeros, Pa(x) has three zeros, and P 4(X) has four zeros in the interval
-1 x 1.
In the remainder of this lesson we shall develop a number of properties
of Legendre polynomials which are valid in the interval I: Ixl 1.
A. The Legendre polynomials Pn(x) are the coefficients of t n in
the Maclaurin series expansion of (I - 2xt +
t 2)-1/2, i.e.,

(41.41) (1 - 2xt + t2)-1/2 = Po(x) + P 1(x)t + P 2(X)t 2 + ...

+ Pn(x)t n + ... ,
if Ixl I, ItI < I.
Proof. We shall merely give the formal steps of the proof, without
attempting to justify each of the steps used. A rigorous proof is beyond
the scope of this text. The Maclaurin series expansion of (1 + U)k, called
Lesl!On 4Ie PROPERTIES OF LEGENDRE POLYNOMIALS" Pk(:C) 599

a binomial series, is

(a) (1 + U)k = 1 + ku + k(k 1) u 2 + ...

+ k(k - 1) ••• (k - n + 1) un + ...
n! '
valid for those values of u and k for which the series converges. If k = -t,
the series converges for lui < 1. Let us now write the left-hand member
of (41.41) as [1 +
t(t - 2X)]-1/2. Hence by (a), if It 2 - 2xtl < 1,

(b) [1 + t(t - 2X)]-1/2 = 1+ t(2x - t) + 212 t 2 (2x - t)2 + .. .

+ 1 . 3 . 5· .. (2n - 1) tn(2x _ t)n + .. .
2nn!

= 1 + 21 (2xt - t 2) + 222
3 t 2(4x 2 - 4xt + t 2) + .. .
+ 1 . 3 . 5· .. (2n - 1) tn(2x _ tt + ... .
2nn!

From the last expression on the right of (b), we see that the coefficients
of to, t, and t 2 are respectively

(c) 1, 21 (2x) = x, - -21+ -222

(3 4 x2) = -
1( 2 )
2 3x-l .

A comparison of (c) with (41.32) shows that these coefficients are respec-
tively Po, PI, P 2 • The coefficient of the general term tn is the slIm of the
coefficients of tn in the last term on the right of (b) and of coefficients of tn
in preceding terms. Hence the total coefficient of tn is
1· 3 . 5 ... (2n - 1) n 1 . 3 . 5 ... (2n - 3) (n - 1) n-2
(d) 2nn! (2x) - 2n len _ I)! 11 (2x)
1 . 3 . 5 ... (2n - 5) (n - 2)(n - 3) n-4
+ 2n 2(n _ 2)! 2! (2x) - ....

The last term in (d) can be written as

1 . 3 . 5· .. (2n - 5)(2n - 3)(2n - 1) (n)(n - 1)
(e)
2n 2 (2n - 3)(2n - 1) (n)(n - 1)(n - 2)1
(n - 2)(n - 3) 2n-4 n-4
X 21 x
= 1·3·5 .. · (2n - 1) [n(n - 1)(n - 2)(n - 3) xn- 4] .
n! _ 2· 4(2n - 1)(2n - 3)
600 SERIES METHODS Chapter 9

Hence (d) becomes, with its last tenu replaced by the expression on the
right of (e),

(f) 1· 3 . 5 ... (2n - 1)

n!
X [x -
n n(n - 1) n-2 + n(n - l)(n - 2)(n - 3) n-4 + J
2(2n _ 1) x 2. 4(2n _ 1)(2n _ 3) x ...•

A comparison of (f), which is the coefficient Pn(x) of tn of (41.41), with

the polynomial P/c(x) of (41.31) shows that with k replaced by n, the first
three tenus of each are alike. If we had used more tenus of the binomial
series (d), we would have obtained additional tenus in (f) in agreement
with (41.31). Note also the similarity of the coefficient in (f) with the value
of a/c as given in (41.3). This value was in fact given to a/c in order to ob-
tain the identity (41.41).
B. Values of Pn(O) , Pn(I), and Pn( -I). If x = 1, the left side of
(41.41) simplifies to
(a) (1 - 2t + t2)-1/2 = (1 - t)-I.

Its series expansion is

(b) 1+t + t 2 + t3 + ... + tn + .. " ItI < 1.

Comparing the coefficients of this series with the coefficients of the series
on the right of (41.41), we see that, with x = 1,

(41.42) P o(l) = 1,

If x = 0, the left side of (41.41) becomes

(c)

whose series expansion is

(d)
1- W+!. it' - ... + (-It 2·4·
1· 3··· (2n -
.. (2n)
1) t 2n + . ", ItI < 1.
A comparison of this series with the right side of (41.41) shows that, with
x = 0,
(41.43) PI(O) = 0, P 3 (0) = 0, .. " P 2n - I (O) = O.

(41.44) Po(O) = 1, P 4(0) = i· i, ... ,

P (O)
2n
= (_ I)n I . 3 ... (2n - I) .
2·4··· (2n)
Lellon 41C PROPERTIES OF LEGENDRE POLYNOMIALS Pk(x) WI

If X = -1, the left side of (41.41) becomes

(e) (1 + 2t + t 2 )-1/2 = (1 + t)-l = 1 - t + t2 - t3 + t" - ... ,

It I < 1.
Comparing (e) with the right side of (41.41), we find, with x = -1,
(41.45) P o(-I) = 1, P1(-I) = -1, P 2 (-I) = 1,
P a(-I) = -1,···, Pn(-I) = (-It.
c. Recursion formula for Pn(x). Differentiating (41.41) with respect
to t, we obtain
(a)
-!(1 - 2xt + t2)-3/ 2(-2x + 2t) = PI + 2P2t + ... + nP"tn- 1 + ....
Multiplying (a) by 1 - 2xt + t 2, there results
(b) (x - t)(1 - 2xt + t 2 )-1/2 = (1 - 2xt + t 2 )
X [PI + + 2P 2 t + ... + (n - I)P"_lt,,-2 + nP"t,,-l
+ (n + I)P"+lt" + ...].
By (41.41), we can write (b) as
(c) x[P o + PIt + ... + P"t" + ...]- [Pot + P 1t 2 + ... + P"_lt"+ ...]
= [PI + 2P2t + ... + (n + I)P"+lt" + ...]
- 2x[P 1t + ... + nP"t" + ...]
+ [Plt 2 + ... + (n - I)Pn _ 1t" + ...].
Equating the coefficient of t" on both sides of the equal sign, we obtain
(d) xPn - Pn- 1 = (n + I)P,,+l - 2xnPn + (n - 1)P"-1,
which simplifies to the recursion formula
(41.46) (n + I)P"+I(x) - (2n + l)xPn(x) + nP,,_l(x) = 0,
n = 1,2,···.
It is easily verified, by (41.32), that
(41.461) = 0,
and, by (41.46), that when
(41.47) n = 1: 2P 2 (x) - 3xP 1 (x) + Po(x) = 0,
n = 2: 3P a(x) - 5xP 2 (x) + 2P 1 (x) = 0,

etc. We leave it to you as an exercise to verify, by means of (41.32),

that the formulas in (41.47) are indeed true equations.
602 SERIES METHODS Chapter 9

D. Rodrigue's Formula. Another compact formula for expressing

the Legendre polynomials P ,,(x) is

(41.48)

It is known as Rodrigue's formula.

Proof of the Formula. The binomial series expansion of (x - I)"
can be written as

(a) (x - I t = E" (_I)" k!(n

, k)! xn-k.
Hence
(b)

Taking n successive derivatives of X 2n - 2k., we obtain for the nth derivative,

(c) fxn (X 2n - 2k ) = (2n - 2k)(2n - 2k - 1) ... (n - 2k + I)xn- 2k,

2k;:;; n,
=0, n < 2k.
By definition of the factorial function, (c) can be written as

(d) ( 2n-2k) _ (2n - 2k)! n-2k 2k

dxn x - (n _ 2k)! x , ;:;; n,
= 0, n < 2k.
Hence by (b) and (d),

(x2 _ I)n =
(e)
dxn 1::0 (-It k!(n n!- (2n - 2k)! Xn-2k
k)! (n - 2k)! .

Therefore by (41.48) and (e),

(f) Pn(x) E
=
[,,/21 n 1 n! (2n - 2k)! n-2k
(-1) n!2n k!(n - k)! (n _ 2k)! x

=E
[n/21 n (2n - 2k)! n-2k
(-1) 2nk!(n - k)!(n - 2k)! x .

A comparison of (f) with (41.317) shows they are alike. (Interchange n

and k in either equation.)
E. Orthogonal Property of Legendre Polynomials.
Definition 41.5. A set of functions Jr, /Z, ... , is said to be orthog-
onal on an interval I: a ;:;; x ;:;; b, if, for every two distinct functions
Lesson 41C PROPERTIES OF LEGENDRE POLYNOMIALS Pk(x) 603

of the set,

(41.51)

We shall now prove that the set of Legendre polynomials is orthogonal

on the interval I: -1 x 1, i.e., we shall prove

(41.511)
f l
-1
Pm{X)Pn{x) dx = 0, m n.

Proof. Each Legendre polynomial Pm{x) satisfies the Legendre equa-

tion (41.1). Therefore

(a) (I - x 2)Pm"{x) - 2xPm'{x) + m{m + I}pm{x) = 0,

which can be written as

d
(b) dx [(I - x 2)Pm'{x)] + m{m + I)Pm{x) = O.

Multiplying (b) by Pn{x), n m, and integrating between the limits -1,

1, we obtain
(c)

Pn{x) tx 2
[(I - x )Pm'{x)] dx + m{m + 1) Pn{x)Pm{x) dx = O.

The first integral in (c) can be integrated by parts in accordance with

the formula
(d) f u dv = uv - f v duo

In the first integral of (c), let u = P n{x) and dv equal the rest of the in-
tegrand. Then by (d), this integral becomes

2
(e) Pn[{1 - - (l - x )Pm'Pn'dx.

You can easily verify that the first term in (e) is zero. Therefore, by (e),
(c) simplifies to

(f) _fl -1
(I - x 2)Pm'Pn'dx + m{m + l)fl
-1
PnPmdx = O.

If we had started with Pn{x) in (a) instead of with Pm{x) and followed
the steps outlined above, we would have obtained the same result as in (f)
604 SERIES METHODS Chapter 9

with m and n interchanged. Hence also

2
(g) - (1 - x )P"'P,,,' dx + n(n + 1) P",P" dx = O.

Subtracting (g) from (f), there results, with m n,

(h) (m 2 +m - n 2 - n)jl P",(x)P,,(x) dx = O.

-1

+
Since m 2 m - n 2 - n = (m - n)(m n + + 1) and m n, we
may divide (h) by this term to obtain (41.511).
F. Other Integral Properties of Legendre Polynomials. The first
integral property we shall prove is,

(41.52) R".(x)P.. (x) dx = 0,

where R",(x) is a polynomial of degree m less than n.

Proof of (41.52). Let

(a) R",(x) = ao + alx + a2x2 + ... + am_lX",-l + amx"'.

The polynomial R",(x) can always be written as a linear combination of
the Legendre polynomials Po, PI! ... , P"" i.e., we can always write

(b) R",(x) = ao + alX + a2x2 + ... + am_lxm- l + amx'"

= CoP o + ClP l + ... + C",-lP",-l + C",P",.
The truth of this statement can be demonstrated as follows. P", is the
only polynomial on the right of (b) which has a term in x"'. Hence we can
choose Cm so that the coefficient of the x'" term of P", is equal to a",.
Next we sum the two coefficients of X",-l in P m- l and Pm, and give Cm- l
a value so that C",-l times this sum equals a"'-I! etc.
In (41.52) replace R",(x) by its equal last expression in (b). We thus
obtain

(c) Cofl Pop.. dx

-1
+ Clfl -1
PIP.. dx + ... + Cmfl -1
P",P"dx, m < n.

By (41.511), each term in (c) is zero. Hence (41.52) follows.

The second integral property we shall prove is
l

(41.53) j -1
2
[P.. (x)] dx = 2n
2
+ 1.
Lesson 41-Exercise 605

Proof of (41.53). Squaring both sides of (41.41), we obtain

(a) (1 - 2xt + t2)-1 = [t n-O

Pn(x)t]2 .

Integration of (a) with respect to x between the limits -1 and 1 gives

(b) /1-1
(1 _ 2xt + t 2)-1 dx = /1 [t-1 n_O
Pn(x)tn]2 dx.

When we square the integrand on the right of (b) and perform the inte-
gration, the only terms, by (41.511), which are not zero are those in which
the subscripts of P n(x) are the same. Hence (b) simplifies to

(c) /1
-1
(1 _ 2xt + t 2)-1 dx = t
n=O
t2n /
-1
1
[Pn(X)]2 dx.

Integration of the left side of (c) results in (remember t is a constant in

this integration and log (1 + t) = t - !t 2 + i-t 3 - ••• )
(d)
1
- 2t log (1 - 2xt +t )
2 11
-1 = - 2t log
1 (1 - t)2
(1 + t)2 = t1 log 11 + t
- t
3 S 7
t t t
=t2 [ t+3"+5+7+'" ]

t2 t 4 2n
= 2 [1 + 3" + 5" + ... + 2nt + 1 + .. ,] , It I < 1.
Substituting the last expression of (d) for the left side of (c), we have
1
ao 2t2n ao
2n
(e) L
n-O
2n + 1 = L
11.-0
t /
-1 [Pn(X)]2 dx.

Equating coefficients of t 2n in (e), we obtain for each n,

1
2 2
(f)
/
-1 [Pn(x)] dx = 2n + 1.

EXERCISE 41
I. Find the general solution of each of the following Legendre equations.
(a) (1 - x 2 )y" - 2xy' + 2y = O. (b) (1 - x 2 )y" - 2xy' + 6y = O.
2. Find a polynomial solution of the Legendre equation (1 - x 2 )y" - 2xy'
+
30y = O.
3. Find the Legendre polynomials Ps(x), P6(X), P7(X).
606 SERIES METHODS Chapter 9

4. Find the general solution of (1 - X2 )y" - 2xy' fy = O. +

5. Verify by direct substitution in (41.32) that
Po(l) = 1, PI (1) = 1, P2(1) = 1, Pa(l) = 1, P 4(1) = 1;
Po(-I) = 1, P I (-I) = -1, P2(-I) = 1, Pa(-I) = -1, P4(-I) = 1;
Po(O) = 1, PI(O) = 0, P2(0) = -i, Pa(O) = Q, P4(0) = i.
6. Verify, by use of (41.32), the accuracy of the equations in (41.47).
7. Express xa as a linear combination of Legendre polynomials.
8. Express x4 +
2x 3 +
2x2 - X - 3 as a linear combination of Legendre
polynomials.
9. Assume that a functionf(x), defined in the interval (-1,1), can be repre-
sented by a series of Legendre polynomials, i.e., assume
(a)
Show that if this series, after mUltiplication by Pk(X), can be integrated
term by term, then the coefficients co, CI, ••• , are given by

Ck
2k+
= -2-
1/ 1

-1 f(X)Pk(X) !lx.

Hint.Multiply (a) by Pk(X), integrate from -1 to 1, then use (41.511)

and (41.53).
10. Prove that P2,,(-X) = P2,,(X); P2"+1(-X) = -P2,,+1(X).
ll. Prove each of the following identities:
(a) xP,,'(x) - P,,_I'(X) = nP,,(x). Hint. Let
u = (1 - 2xt + t2)-1I2 = Po(x) + Pl(X)t + ... + P,,(x)t" + ....
First show (x - t)(ou/ox) is equal to t(ou/ot). Then perform the indi-
cated operations and equate coefficients of t".
(b) P,,+1'(x) - (n +
I)P,,(x) = xP,,'(x). Hint. Differentiate (41.46), then
substitute for P ,,-I' (x) its value as given in (a) of this problem.
(c) (x 2 - I)P,,'(x) = nxP,,(x) - nP,,_I(x). Hint. Multiply (a) of this
problem by x; replace n +
1 in (b) by n, then take the difference between
the two resulting equations.
12. Verify that the formula in problem l1(a) is valid for P2, Pa, P4.
13. Show that the hypergeometric series F (k + 1, -k, 1; 1 -; x) is a solution
of the Legendre equation (41.1). Hint. Write (41.1) as (x - l)(x l)y" +
+ 2xy' - k(k+ l)y = O. Then follow the procedure given in Exercise 40.23.
Here TI = 1, T2 = -1, Ta = 0, a = 1, b = 2, c = -k(k + 1), x =
-2u +
1. The transformed Gauss equation (40.7) becomes
2
d y d
u(1 - u) du 2 + (1 - 2u) + k(k + l)y = 0,

whose solution is F(k +

1, -k, 1; u).
14. Verify each of the following.
Lesson 41-Exercise 607

15. The following differential equation

(41.6) y" - 2xy' + 2ky = 0, k real,

is known as the Hermite equation. Show that its solution, valid for all x, is
(41.61)
2222 42
3 6]
Yk(X) = ao [ I - 2! kx + 4! k(k - 2)x - 6! k(k - 2)(k - 4)x + ...
2
+ al [ x - 3i (k - I)x
3
+ 225! (k - I)(k - 3)x
5

- (k - I)(k - 3)(k - 5)x7 + ...]

n
= ao [ I + L:
'" (-lf2
(2n)!
k(k - 2)(k - 4) + ... + (k - 2n + 2)x 2n]
+ al [x + t (-1)"2" (k - l)(k - 3)··· (k - 2n + I)X2n +l ] .
,,-1 (2n +
I)!

16. Find polynomial solutions of (41.6) for k = 0, 1,2, .. " 7. Hint. Note that
for k = 0 or a positive integer, one of the series in (41.61) terminates.
Ans. yo(x) = ao, Yl(X) = alX, Y2(X) = ao(l - 2x 2),
Y3(X) = aleX - 3 ), ix Y4(X) = ao(l - 4x2 ix4), +
Y5(X) = aleX - tx 3 +
b 5), Y6(X) = ao(1 - 6x2 4x4 - + b 6),
+
Y7(X) = aleX - 2x 3 !x 5 - 'Ihx 7).
17. The following polynomials, known as Hermite polynomials,
Ho(x) = I,
Hl(X) = 2x,
H2(X) = 4X2 - 2,
H3(X) = 8x3 - 12x,
H 4(X) = 16x4 - 48x2 12, +
H 5(X) = 32x 5 - 160x3 120x, +
H 6(X) = 64x6 - 480x 4 +
720x 2 - 120,
H 7(X) = 128x 7 - 1344 5 +
3360x 3 - I 680x,

can be obtained from the solutions given in problem 16, by choosing appro-
priate values for ao and al. Show that these Hermite polynomials can also
be obtained from the formula

n z2 dn _z2
(41.62) H,,(x) = (-I) e -d e , n = 0, 1,2, ....
x"
18. In solving (41.6), the following recursion formula results,
2(n - k)
(a)
an+2 = (n + I)(n + 2) an·
608 SERIES METHODS Chapter 9

Following exactly the steps from (41.24) on, obtain polynomial solutions
of the Hermite equation (41.6), of the form (i.e., in descending powers of x)

(
b) h( ) _
k X - ak
[k _ k(k -
x 22
1)
x
k-2 + k(k - 1)(k -
2!
2)(k -
k
3) x - '
24
k(k - l)(k - 2) ... (k - 5) xk-O
- 3! y+'"
k
(-1)"k(k - 1)··· (k - 2n+ 1) X - 2,,]
+ n!
[k/2J (-I)"k! x k- 2"

= ak L:O nICk _
,,-
2n)! k = 0, 1, 2, ....

In (b), ak is an arbitrary constant. If we take for it the value 2k, we obtain

the Hermite polynomials
[k/2J (-I)"k
(c) Hk(X) = L: I(k _ 2' )1 (2x)k-2",
,,-0 n. n.
l
k = 0,1,2,'"

Verify that the above formula (c) also gives the Hermite polynomials shown
in problem 17 above.
19. In the summation in (c) of problem 18 above, and in (41.317), n cannot be
larger than [k!2]. Why?
20. Show that the coefficient of t" in the series expansion of e2",-12 in powers of
tis H,,(x)!n! Hint. e2",-12 '"' e2",e-12. Write the series expansion of each
term of this product, multiply both series, then show the coefficient of t"
is H,,(x)!n!.

ANSWERS 41

1•. (a) y = ao (1 - (1 t 1) x 2 + (1 + 1)(1 2)(1 + 3) x' + .. -) + alX.

(b) Y = ao (1 - 3x 2) + al ( x - (2 - 1)(2
3! + 2) x 3
+ (2 - 1)(2 + 2)(2 - 3)(2 + 4) x 5 + ...) .
5!
2. Y = x - ljx3 .y..xs. +
3. P5(X) = i(63x s - 70x3 + 15x);
Po(x) = h(231x O - + 105x 2 - 5);
315x 4
P7(X) = h(429x 7 - + 315x3 - 35x).
693x 5

4. Y = ao (1 _ t(t + 1) x 2 + t(! + l)(t - 2)(t + 3) x4 + ...)

2! 4!
(! - l)(t + 2) 3
+ al (x - 3! x
_ (t - 1)(t + 2)(t - 3)(! + 4) x5 + ...) .
5!
7. x 3 = !P3(X) + !PI(X).
Leason 42A THE BESSEL DIFFERENTIAL EQUATION 609

8. -AP4(X) + !P3(X) + !?-P2(X) + lPl(x) - HPo(x).

19. For n > [k/2], k - 2n is a negative integer and (k - 2n)! is undefined for
negative integers.

LESSON 42. The Bessel Differential Equation. Bessel Function

of the First Kind JI«%)' Differential Equations
Leading to a Bessel Equation. Properties of JI«%)'

LESSON 42A. The Bessel Differential Equation. Another differen-

tial equation which arises frequently in physical problems and about
which much has been written is known as the Bessel equation, named
in honor of the German mathematician F. W. Bessel (1784-1846). It is
(42.1)
where k is a positive constant or zero. You can verify by definitions 40.22
and 40.24 that x = 0 is a regular singularity of (42.1). Hence we seek a
Frobenius series solution of the form
(42.11)
Comparing (42.1) with (40.33), we see that
(42.12)
both of which are already in series form. Comparing (42.12) with (40.34),
we have
(42.13) bo = 1, Cl = 0, C2 = 1.
All remaining b's and c's are zero. The indicial equation (40.38) therefore
becomes
(42.14)
whose roots are ± k. As we shall discover, when k is not 0 or an integer,
there will be two Frobenius series solutions of (42.1). If k is zero or an
integer, there will be only one Frobenius series solution of (42.1). The
second independent solution will, by Comment 40.5, be a logarithmic
solution and have the form (40.51).
Using the root m = k and the values in (42.13) we obtain, by setting
the second coefficient in (40.37) equal to zero,
(42.15) al[(k + l)k + (k + 1) - k 2] + Oao = 0, (2k + l)al = O.
Since k is a positive constant or zero, the second equation in (42.15) will
be an id!lntity only if
(42.151) al = O.
610 SERIES METHODS Chapter 9

The recursion formula, obtained by setting the coefficient of xm+n in

(40.37) equal to zero, becomes with m = k and with the help of (42.13),
(42.16) an[(k + n)(k + n - 1) + (k + n) - k 2] + + an-2(1) = 0, °
which simplifies to
n n > 2
(42.17) an = - n 2 a+-22nk' =.
Since, by (42.151), al = 0, we see from (42.17) that
(42.18) aa = a5 = a7 = ... = a2n+l = 0, n = 0, 1,2, ....
For even values of n, we have, by (42.17),
(42.19)
ao (t)2
a2 = - 4 + 4k = - 1 + k ao,
a2 -1 -m 2 m4
a4 = - 16 + 8k 8(2 + k) (1 + k) ao = 2!(2 + k)(l + k) ao,

a6 = - 36
a4
+ 12k = - 2. 12(3
m 4

+ k)(2 + k)(l + k) ao
_(!)6
3!(3 + k)(2 + k)(1 + k) ao,

a2n = n!(1 + k)(2 + k) ... (n + k) ao·

Substituting in (42.11) the root m = k and the above values of the a's,
we obtain

(42.2) Yk = aoX
k
[
1- 1 ! k + 2!(1 + :)(2 + k)
1
- 3!(1 + k)(2 + k)(3 + k) + ...
+ k) ... (n + k) (x)2n
+ n!(1 + k)(2(-It ]
2 + . .. ,
which is one solution of (42.1), valid for k 0.
Similarly, by following the above procedure for the root m = -k,
k 0, we obtain [equivalent to replacing k by -k in (42.2)]

[1 - 1
4
k 1
(42.21) Y_k = aox- k 2 + 2!(1 _ k )(2 _ k)

- 3!(1 - k)(21_ k)(3 - k) + ...

(-1)" (x)2n ]
+ n!(1 - k)(2 - k) ... (n - k) 2 + . .. ,
Lell!lOn 42B BESSEL FuNt'TIONS OF THE FIRST KIND h(x) 6Il

which is a second solution of (42.1), valid for k :;<!i 1, 2, 3, .. '. (For

these values of k, one of the denominators in (24.21) is zero.) Since the
functions II and 12 of (42.12) are polynomials, by Comment 37.53, their
series representations are valid for all x. Hence by Theorem 40.39 the
series solutions, (42.2) and (42.21), converge for all x except perhaps at
x = 0.
Note that the above results support the statement we made after
(42.14). If k = 0, the two solutions (42.2) and (42.21) are, as you can
verify, identical, and there is therefore only one Frobenius series solution
of (42.1). And if k = 1, 2, 3, ... , there is again only one Frobenius
series solution of (42.1), namely (42.2). For all other positive values of
k, Yk, and Y-k are, by Theorem 40.39, two linearly independent solutions
of (42.1). Hence if k :;<!i 0, I, 2, 3, "', the general solution of (42.1),
by Theorem 19.3 and Comment 19.41, is

(42.22) y(x) = CIYk + C2Y-k, k:;<!i 0, 1, 2, 3, ... ,

convergent for all x :;<!i 0, where Yk and Y-k are defined by (42.2) and
(42.21) respectively.

LESSON 42B. Bessel Functions of the First Kind J,,(x).

In (42.2) ao is an arbitrary constant. Let us choose for it the value

1
(42.3) ao = 2k k!' k O.

(We showed in Lesson 27E, that kl is defined for all values of k except
negative integers.) The functions resulting from (42.2) when ao is given
the value (42.3) are designated by Jk(X) and are called Bessel functions
of the first kind of index k. Hence, with k!(k + 1) replaced by its
equal (k + I)!, we obtain, by (42.3) and (42.2),

+1 3)! (x)
6
- 3(k 2 + ...
(-I)"
+n!(k+n)!
(x)2 2" ]
+ ... ,
valid for all k 0.
If, in (42.31), we replace k by -k [equivalent to choosing ao =
612 SERIES METHODS Chapter 9

in (42.21)], we obtain
(42.32)
2

J_Ic(X) = ( 2x)-J; [ (-k)1

1
(1 -
1
k)1
(x)
2 + 21(2 1 (x)'
- k)1 2 + ...
+ nl(n -
(-1)"
k) 1
(x)2 2"
+ . ..] ,
valid for all k :;o<!i 1, 2, 3, ....
We showed in Lesson 42A, that if k is zero or a positive integer, then
Ylc of (42.2) is the only Frobenius series solution of the Bessel equation
(42.1). However, if k is not zero or an integer, then the functions Ylc and
Y-k of (42.2) and (42.21) are two linearly independent solutions of (42.1).
Since J k(X) and J _Ic(X) defined in (42.31) and (42.32) are these same
functions Yk and Y-k multiplied by an appropriate constant, it follows
that if k is not zero or an integer, J and J Ic -Ic
are linearly independent
solutions of (42.1), each valid for all x except, perhaps for x = 0. Hence
we can assert, by Theorem 19.3 and Comment 19.41, that
(42.33)

is also a general solution of (42.1), valid for all x :;o<!i 0, where J k(X) and
J _Ic(X) are defined by (42.31) and (42.32).
If however k = 0, 1,2,3, ... , then (42.31) is the only Frobenius solu-
tion of (42.1). In this case, a second solution of (42.1) will have the
logarithmic form shown in (40.51). The one we shall give below· is desig-
nated by -N,.(x) and is called a Bessel function of the second kind
of index k. It can be shown that if, in (42.1), we make the substitution

Y2(X) = u(x) - JJ;(x) log x, x > 0,

it becomes (see Exercise 42,2)

x 2 u"+ xu' + (x 2 - k2)U = 2xJ ,.'(x),

where
u(x) = x-k(bo + b1x + b2x2 + ...), k > 0.
A second solution of (42.1), if k is a positive integer, is then
(42.34)
1 k-l (k - n - 1)1 (x)2"-" 1 GO (-I)"
-NJ;(x) = 21:
.. -0
nl 2 + 2 ..1:
-0
n!(n+ k)1

X[H.. + (1 + 4+ ... + n! k)]

- J,,(x) log x, x > 0,

'"There are other forms of the Bessel function of the second kind.
Lesson 42B BESSEL FUNCTIONS OF THE FIRST KIND Jk(X) 613

where
Hn = 1 + 21 + 31 + ... + n'
1
if n F- 0,
= 0, if n = 0.
Hence for positive integer k, the general solution of (42.1) is
(42.35)
Remark. A second solution, when k = 0, can be found at the end
of Example 40.6. See (k), page 584.
Note that thus far we have no solution J _k(X) when k is a positive integer.
The solution J _k(X) of (42.32) is undefined for k = 1, 2, 3, .. '. To fill
in this gap, we define J _k(X) , k = 0, 1,2,3, ... , by the relation
(42.36).
The justification for this definition is outlined below.
Justification for (42.36). By (42.31),

(42.37) Jk(x) = E ..
n!(k
(-It
+ n)1 2
(x)2n+k
' k = 0, 1,2,3, ....

By (42.32), we write formally

(42.371) J _k(X) = E
.. (-It
nl(n - k)!
(x)2n-k
2 ' k = 0, 1, 2, 3, ... ,

even though the series is not defined for n < k. [For these values of n,
(n - k) is a negative integer and as we showed in Lesson 27E, (n - k)!
is undefined for negative integers.] We also showed in Lesson 27E that
(k - I)! = r(k) -+ 00 as k -+ 0, -1, -2", '. Since a negative integer
factorial appears in the denominator of each term of the series (42.371)
for which n < k, we are led to define each term, for which n = 0, 1,
2, ... , k - 1, to be zero. Hence we can write (42.371) as

(42.38) J_k(X) = L..

n-II
(-It
n!(n _ k)!
(x)2n-k
2 ,k = 0,1,2,3,···.

Let p = n - k. Hence when n = k, p = 0. This means we can start

°
our summation in (42.38) with p = instead of with n = k, if at the same
time we substitute p + k for its equal n. Therefore (42.38) can be written
as
.. (_I)p+k (X)2P+k
(42.39) J_,,(x) = Lo
(p+k)!(p+k-k)! 2
P-
,k=0,1,2, .. ·.
614 SERIES METHODS Chapter 9

Since the summation in (42.39) is over p, we may remove (-I)k from in-
side the summation. It then becomes, after simplification,

(42.4) J_k(x) = (-1) fa

k "" (-I)P (X)2P+k
+
prep k)! 2" ,k = 0,1,2,···.

Comparing the summation in (42.37) with that in (42.4), we see that

both are the same. Hence (42.36) follows.

The most frequently encountered Bessel functions of the first kind are
Jo(x) and J 1(x). By (42.31), these are
x2 1 X4 1 x6
(42.41) Jo(x) = 1 - 22 + (2!)2 24 - (3!)2 2 6 + ...
(-1)" x2n
+ (n!)2 2 2" + ... ,

The graphs of these two functions are shown in Fig. 42.43.

Figure 42.43

Tables of values exist for Bessel functions of the first kind J k(X), just
as they do for sin x, log x or eX. Thus if you had to evaluate J 1 (2), you
could use (42.42) with x = 2, or look up its value in a table. Its value is
0.5767. Therefore by (42.36), J _1(2) = -0.5767. If you had to evaluate
Les80n 42C EQUATIONS WHICH LEAD TO A BESSEL EQUATION 615

J -1/2(3), you could use (42.32) with k = i, x = 3 or look up its value

in a table. Its value is -0.4560.

LESSON 42C. Differential Equations Which Lead to a Bessel

Equation. We give below examples of differential equations whose
solutions can be obtained by transforming each into a Bessel equation
by means of a suitable substitution.
E%ample 42.5. Find a solution of

(a) u" + (1 + 1 U = 0, kreal.

Solution. The substitution,

(b) u = 1 2
X / y,

U' = iX- 1/ 2y +
X 1/ 2y',
U" = -iX- 3/ 2y +
X- 1/ 2y' + Xl/2y",
will transform (a) into the Bessel equation

(c)

Since y = J k(X) is a solution of (c), it follows by the first equation in (b)

that
(d)
is a solution of (a).
E%ample 42.51. Use the result obtained in Example 42.5 to find a
solution of
(a) u" + (i - 4!2) u = O.
Solution. If in (a) of Example 42.5 above, we choose k = 1, it
becomes the (a) of this example. Hence by (d) above, a solution of (a) is

(b)

E%ample 42.52. Find a solution of

(a)

Solution. Here two substitutions will be needed. The first substitu-

tion is
(b)
616 SERIES METHODS Chapter 9

Substituting in (a) the value of u as given in (b) and the value of u" as
given in the last equation of (b) of Example 42.5, and then multiplying
the result by X 3 / 2 , we obtain
(c) xV' + xy' + (x" - !)y = o.
The second substitution is
x2 2
(d) W = 2"' x = (2w) 11 ,

dw = x = (2w) 1/2.
dx
With the help of (d), we obtain

(e) dy = dy dw = (2W)1/2 dy,

dx dw dx dw
2 2
d y = (2W)1/2 d y dw +
(2W)-1/2 dw dy
dx 2 dw 2 dx dx dw
d 2y dy
= (2w) dw2 +
dw'

Substituting (d) and (e) in (c), it becomes

(f) (2W)2 + 2w ;! + ;! +
2w [(2W)2 - !]y = 0,

which simplifies to

(g) w
2
+w ;! + 2
(w - n)y = O.

Since (g) is the Bessel equation (42.1) with k = !, its solution is

(h)

Replacing w by its value in (d) and then substituting the resulting value
of y in (b), we obtain
(i)
which is a solution of (a).
Comment 42.53. The method used to solve equation (a) of Example
42.52 can be applied to the more general equation

(42.54)

The first substitution is the same as (b) of Example 42.52, namely

(a)
Lesson 42C EQUATIONS WHICH LEAD TO A BESSEL EQUATION 617

The second substitution, however, is more complicated than the one in

(d) of Example 42.52. It is

(b) W
2Vb
= m + 2 vxm+ 2 •
The substitution of (a) in (42.54) and of (b) in the resulting equation, will
yield the Bessel equation

2dw + (2 +1)
2
(c) W
d y
2
dy
+ w dw w - (m 2)2 y = o.
Its solution is
(d) y = J I/(m+2)(W).
Replacing w by its value in (b) and then substituting the resulting value
of y in (a), we obtain
(42.55) U = x
1/2
J 1 f(m+2) (2Vb+ v.r=i:9)
m 2 xm+
2
,

which is a solution of (42.54). Note that when m = 2, b = 1, (42.54)

reduces to (a) of Example 42.52, and (42.55) reduces to its solution (i).
E%ample 42.56. Find a solution of

(a) u" + 9xu = o.

Solution. If in (42.54) we let b = 9, m = 1, it reduces to (a) above.
Hence a solution of (a), by (42.55) is

(b)

Comment 42.57. The method used to solve equation (a) of Example

42.52 can be applied to a still more general equation than (42.54), namely

(42.58) x 2 u" + (1 - 2a)xu' + (b 2c2x 2c + a 2 - k 2c 2 )U = o.

The first substitution
(a) u = x"y,
will change (42.58) into the equation
d2 y dy
(b) x2 -
dx 2
+ X -dx + (b 2x 2C - k 2 )C 2 y = o.
Verify it. The second substitution

(c) w = bx c,
618 SERIES METHODS Chapter 9

will change (b) into the Bessel equation

d2 d
(d) w2 -.JL2 + w -1L + (w 2 - k2)y = O.
dw dw
Its solution is
(e)

Hence by (a), (e), and (c), a solution of (42.58) is

(42.59)

Comment 42.6. If, in (42.58), we make the substitutions

m+2 1
(a) a =;, c=-2-' k=--,
m+2
and replace b2 by 4bj(m + 2)2, it reduces to
(b)

A solution of (b), by (42.59), is then

(c) u = X 1/ 2 J 1/(m+2) 2Vb

(- - V xm+ 2
) •
m+2
Note that the second equation in (b) is now.the same as (42.54). Note
too that their respective solutions (c) and (42.55) are, as they should be,
also the same. Hence (42.54) is only a special case of (42.58). By assigning
different values to a, b, c, and k in (42.58), we can thus obtain many
different equations whose solutions will be given by (42.59).

E:cample 42.61. Find a solution of

2
(a) x2 u" + xu' + (x - k4 ) U - 0
- .

Solution. If in (42.58), we let

(b) a = 0, b = 2, c= ;,
it reduces to (a). Hence a solution of (a) by (42.59) is

(c)

Comment 42.62. If k is not zero or a positive integer, then, by (c)

above and (42.33),
(d) u(x) = C1J k(2vx) +
is a general solution of (a) of Example 42.61. If k is a positive integer,
Lesson 42D PROPERTIES OF BESSEL FUNCTIONS OF FIRST KIND Jk(X) 619

then by (42.35), a general solution of (a) is

(e)
where Nk(x) is given by (42.34).

Example 42.63. Find a solution of

(a) x 2u" + xu' - (x 2 + k 2)u = O.

Solution. If in (42.58), we let

(b) a = 0, c = 1,

it reduces to (a). Hence a solution of (a) by (42.59) is

(c)

Comment 42.64. The function obtained by multiplying (c) of Exam-

ple 42.63 above, by the constant i- k is also a solution of (a) of this
example. It is usually written as h(x). Hence

(42.65)

The function I,,(x) of (42.65) is called a modified Bessel function of

the first kind.

Comment 42.66. We list one more equation which can, by a suitable

substitution, be changed into a Bessel equation. It is

(42.67) x 2u" + x(1 - 2x tan x)u' - (x tan x + k2)U = O.

The substitution u = y/cos x will transform (42.67) into the Bessel

equation
d2 d
(a) x2 J 2
dx
+ x ..Jl.
dx
+ (x 2 - k2)y = O.

Since a solution of (a) is y = h(x), a solution of (42.67) is

1
(42.68) u = --Jk(X).
cos x

LESSON 42D. Properties of Bessel Functions of the First Kind

],,(x).
A. Properties of the Zeros of Bessel Functions ],,(x). We list
below, without proof, a number of properties of the zeros of Bessel func-
tions J k(X) of the first kind.
620 SERIES METHODS Chapter 9

1. If Xl and X2 are two zeros of Jk(X), then in the interval I: Xl < X < X2,
there is a zero of Jk_l(X) and h,+I(X),
2. The Bessel function Jo(x) has a zero in each interval of length 1r.
3. Each Bessel function J/r,(x), k = 1, 2, ... , has an infinite number of
real positive zeros in the interval I: 0 < X < 00.
4. If k > !, then the difference between two consecutive zeros of Jk(X)
is greater than 1r.
5. If k > !, then the difference between two consecutive zeros of J/r,(x)
approaches 1r, as X approaches infinity.
6. The first positive zero of J /r,(x) is greater than k.
7. A Bessel function J/r,(x) has only real zeros.
B. Integral Property of Bessel Functions J".(x).
Theorem 42.7. Let r}, r2," ., be distinct positive zeros oj a Bessel
junction J/r,(x), where k is a fixed real number. Then

(42.71) i l

xJ,.(rix)J,,(rjx) dx = 0, if r, ¢ rj,
= ![J/c'(r,)] 2, if r, = rio
Proof. In Example 42.5, we proved that u(x) = Xl/2J/c(X) is a solution
of u" + (1 + 1 U = O. In a similar manner, it can be shown
that
(a) Ul(X) = x l /2J/c(r.-x) is a solution of u" + (r;2 + 1 U = O.

U2(X) = x l /2J/c(r,-x) is a solution of u" + (rl + 1 U = O.

[Or you can obtain these solutions by letting a = !, b = r, c = 1 in

(42.58). It will then reduce to x 2u" + (r2x2 + i - k 2)u = O. Division
by X2 will give it the form of the differential equations in (a). A solution
by (42.59) will then be
(b) u = x l /2J/c(rx).] .
Hence, by (a),
(c) Ul
" + (r; 2 + 1 -4X24k2) Ul = 0,

U2 "+(ri 2+1-4k2)
4X2 U2 = 0.
Multiplying the first equation in (c) by U2, the second by -Ul and adding
the two resulting equation, we obtain
(d)
Lesson 42D PROPERTIES OF BESSEL FUNCTIONS OF FIRST KIND Jk(x) 621

°
Integrating (d) between the limits and x, and recognizing that the left
side of (d) is (d/dx)(u2ul' - UIU2'), there results

(e) [U2Ul' - UIU2']O = (rl - r,2) fox UlU2 dx.

By (a) and its derivatives,

(f) Ul (0) = 0, U2(0) = 0,

Ul' = r,x1/2Jk'(rix) + !x- 1/ 2J k(r,x),
U2' = rjx l / 2Jk'(rjx) + !x- 1/ 2J k(rjx).

Substituting (a) and (f) in (e), and then simplifying the result, we obtain

(g) r
x[riJk(rjx)Jk'(r,x) - rjh(rix)J/c'(rjx)] = (r/ - ri 2 ) 10 xh(rix)J/c(rjx) dx.

If x = 1, i.e., if the interval (O,x) is the interval (0,1), then (g) becomes

(h) riJ/c(rj)h'(r,) - rjJ/c(ri)J/c'(rj) = (r/ - r/) fol xJk(rix)J/c(rjx) dx.

By hypotheses ri and rj are zeros of J/c(x). Therefore, J/c(ri) = 0,
J/c(rj) = °and the left side of (h) vanishes. Hence if ri, rj, are two distinct
zeros of J/c(x), we can divide (h) by r/ - ri 2 to obtain the first equation
in (42.71).
We now prove the second equation in (42.71). Differentiating the equa-
tion in (g) with respect to rj, we obtain

(i) x[rixJ/c'(rjx)Jk'(rix) - Jk(r,-x)h/(rix) - rixJk(rix)J/c"(rix)]

= 2ri1'" xJk(r,x)J/c(rix) dx
o
+ (rl- r/) dd 1'"
0
xJ/c(rix)J/c(riX) dx.

If ri = rj, then (i) simplifies to

(j) x[xr,J/c,2(r,x) - J/c(r,x)h.'(rix) - rixJ/c(rix)J/c"(rix)]

= 2r,fo'" xJ,,2(rix) dx.

And if the interval is (0,1) so that x = 1, then (j) becomes [remember ri

is a zero of J ,,(x); therefore J /c(r i) = 0]

(k) riJ/c,2(ri) = 2rifol xJ/c2(rix) dx.

622 SERIES METHODS Chapter 9

Hence

(1) i 1
xJ k2(r,-X) dx = llk,2(ri),

which is the second equation in (42.71).

Comment 42.72. Because of the first equality in (42.71), the set of
functions Jk(rlx), J,,(r2x), "', is said to be orthogonal on I: 0 ;:;; x ;:;; 1
with respect to the weight function x. Compare with Definition 41.5
for an orthogonal set of functions.

EXERCISE 42
1. Find general solutions of each of the following Bessel equations.
(a) x 2 y" +
xy' +
(x 2 - l)y = O.
(b) xV' + +
xy' (x 2 - 4)y = O.
(c) x 2 y" + +
xy' (x 2 - !)y = O.
(d) x 2 y" + +
xy' (x 2 - l)y = o.
2. Prove that the substitution in (42.1) of Y2(X) = Uk(X) - Jk(X) log x,
x > 0, transforms the equation into X2Uk" XUk' + +
(x 2 - k2)Uk =
2xJk '(x). Hint. Make use of the fact that hex) is a solution of (42.1).
3. Verify, by direct substitution in (42.31), that Jl(2) = 0.5767; J-1I2(3) =
-0.4560. Hint. Factor out (k)! and use fact that (-1)1 = Vi. Also
verify that Jo(0.3) = 0.9776; Jt(0.2) = 0.0995.
4. Prove that (a) of Example 42.5 is transformed into (c) by the substitution
(b).
5. Verify the accuracy of (c) of Example 42.52.
6. Verify the accuracy of (a) of Comment 42.66.
Find a solution of each of the following equations 7-14, by using the
appropriate formula given in Lesson 42C.
7. y" + 9x 2 y = O. Hint. In (42.54), b = 9, m = 2.

8. y" + (1 + I:X 2 ) y = O.
9. y" + (1 - 9:2) y = O.
10. y" + 4x 3 y = O.
n. xV' + xy' + (x - l)y = O.
12. x 2 y" + xy' - (x 2 + 4)y = O.
13. x 2 y" - xy' + x 2 y = O.
14. x y" + "2 y' + (x - n)y = O.
2 X 4. II

15. By assigning various values to a, b, c, and k in (42.58), obtain at least three

different differential equations and their solutions.
16. Show that the substitution U = 2&'12, eZ = u 2 j4, will transform the equa-
tion y" + (ez - m2 )y = 0 into the Bessel equation u 2 dy +
+ U du
u
Lesson 42-Exercise 623

(u 2 - 4m2)y = O. Since a solution of the second equation is y = J2 ... (U) ,

a solution of the original equation is Vex) = J2m(2e"/2). Hint. Follow the
procedure used in making the second substitution after (c) of Example 42.52.
17. With the help of problem 16, find a solution of each of the following differ-
ential equations.
(a) y" + (e z - 9)y = O. (b) y" + (e z - !)y = o.
(c) y" + (e z - i)y = O.
18. Assume that a function I(x), defined on the interval (0,1), can be repre-
sented by a series of Bessel functions, Le., assume
(a)
where TO, TI, T2, ••• , are the distinct, positive zeros of hex) and k is a fixed
real number. Show that the coefficients Co, ct, C2, ••• I are given by

[Jk'(T;)]2c; = 21oIXI(X)h(T;X) dx.

Hint. MUltiply (a) by Xh(T;X), integrate from 0 to 1, then use (42.71).

19. Prove each of the following identities.
d
(a) dx [xkh(x») = xkJk_I(X). Hint. Multiply (42.37) by Xk and then take
its derivative.

(b) [X-kJk(X)j = -X-kJk+l(X). See hint in (a).

(c) Jk'(X) +
kX-1Jk(X) = Jk-I(X). Hint. Carry out the differentiation
in (a) and then divide by Xk.
(d) Jk'(X) - kX-1Jk(X) = -Jk+l(X). Apply hint in (c) to (b). Divide
by x- k •
(e) Jk-I(X) - Jk+l(X) = 2N(x). Hint. Add (c) and (d).
2k
(f) h-I(X) + Jk+I(X) = - Jk(X).
x
d
(g) dx Jo(x) = -J1(x). Hint. Set k = 0 in (a) above, and then make
use of (42.36).
d2
(h) dx2 [hex») "" ![Jk-2(X) - 2Jk(X) + Jk+2(X)]. Hint. Differentiate (e).
Then use (e) again to find Jk+I'(X) and Jk_l'(X). Substitute these
values in the J k " (x) equation.

(i) J1I2(X) "" /2 sinx.

\j-n Hint. W! = "';;/2.

(j) LI/2(X) "" /2 cos x.

\j-n Hint. (-!)! = ..;:i.
20. Show that the coefficient of t" in the series expansion of e(z/2)[1-(l/tl] in powers
of t is J ,,(x). Hint. e(z/2)[1-(l/I)] = ezI/2e-z/21. Write the series expansion
of each term of this product, multiply both series, then show that the co-
efficient of t" = J ,,(x).
624 SERIES METHODS Chapter 9

ANSWERS 42
1. Substitute: (a) k = 1 in (42.2) and (42.34); (b) k = 2 in (42.2) and (42.34);
(c) k = i in (42.2) and (42.21); (d) k = ! in (42.2) and (42.21).
1l2 2
7. Y = x J1I4(ix ). 11. y = J2(2Vx).
8. y = xl/2 J1I4(X). 12. y = J2(ix).
9. Y = Xli 2151 6(X). 13. y = xJI(X).
10. Y = x1I2J1Is(ixS/2). 14. Y = x 114 J1I2 (x2)
"'2 .
zI2 zI2 zI2
17. (a) y = J 6(2e ). (b) y = Jl(2e ). (c) y = J2/3(2e ).

LESSON 43. The Laguerre Differential Equation. Laguerre

Polynomials L,,(x). Properties of L,,(x).

LESSON 43A. The Laguerre Differential Equation and Its Solu-

tion. The differential equation

(43.1) xy" + (1 - x)y' + ky = 0, k real,

is called the Laguerre equation, after E. Laguerre (1834-1866). It is of

interest only when k is an integer and the interval is x O. You can verify
by Definitions 40.22 and 40.24 that x = 0 is a regular singularity of (43.1).
Hence we seek a Frobenius series solution of the fonn
(43.11)

Multiplying (43.1) by x and comparing the resulting equation with

(40.33), we find that

(43.12) hex) = 1 - x, hex) = kx,

both of which are already in series fonn and valid, by Comment 37.53,
for all x. Hence, by Theorem 40.32, a Frobenius series solution of (43.1)
will be valid for all x, except perhaps at x = O. Comparing (43.12)
with (40.34), we find
(43.13) bo = 1, Co = 0,
All remaining b's and c's are zero. The indicial equation (40.38) therefore
becomes
(43.14) m2 - m +m = 0, m 2 = 0,
whose roots are m = 0 twice. We can therefore expect only one Frobenius
series solution of (43.1).
Lesson 43B THE LAGUERRE POLYNOMIAL Lk(X) 625

Using the root m = 0 and the values in (43.13), we obtain, by setting

the second coefficient in (40.37) equal to zero,

(43.15)

The recursion formula, obtained by setting the coefficient of xm +n in

(40.37) equal to zero, becomes with m = 0 and the help of (43.13),

(43.16) an[n(n - 1) + n] + an_If-en - 1) + k] = 0,

which simplifies to
(n - 1) - k
(43.17) an = n2 an-I·
By (43.17) and (43.15),
1 - k -k(1 - k) k(k - 1)
(43.18) a2 = al = 22 ao = 22 ao,
2 - k -k(k - l)(k - 2)
a3 = a2 = 22 . 3 2 ao,
3 - k k(k - 1)(k - 2)(k - 3)
a4 = a3 = 22 . 3 2 . 42 ao,

(-l)nk(k - l)(k - 2)··· (k - n + 1)

an = 22 . 32 . 42 ... n 2 ao,

(-ltk! ao n = 0 1 2 ...
(n!)2(k - n)!' , " .

Substituting m = 0 and the above values of the a's in (43.11), we obtain

(43.19) -
y" (X) - ao
(1 _ k
x
+ k(k 22- 1) 2 _ k(k - 1)(k - 2) 3
X 22 . 3 2 X
+ k(k - l)(k - 2)(k - 3) X4 + ...
2 2 .3 2 .4 2
(-ltk! " )
+ (n!)2(k _ n)! x +... ,

which is a series solution of (43.1), valid for all x. A second solution will
have the logarithmic form shown in (40.51).

LESSON 43B. The Laguerre Polynomial L,,(x). If k = 0, 1, 2,

3, "', the series (43.19) terminates. The resulting polynomials, with
ao = k!, are known as Laguerre polynomials and are designated by
626 SERIES METHODS Chapter 9

Lk(X), Hence, by (43.19), with ao = k!,

(43.2) Lo(x) = 1,
Ll(X) = 1- x,
L 2(x) = 2 - 4x+ X2,
L 3(x) = 6 - 18x+ 9X2 - X3,
L 4(x) = 24 - 96x + 72x2 - 16x 3 + X4.

L k (x) -- (k,)2
.
(-It
f::o (n!)2(k _ n)! x .
10

We may place (k!)2 outside the summation sign, since the summation is
over n. Graphs of the first four Laguerre polynomials are shown in
Fig. 43.21.
y
16

7 X

Figure 43 •.21

It has been proved that a Laguerre polynomial of degree n has exactly

n real zeros in the interval 0 < x < 00. Note in Fig. 43.21 that Lo(x) has
no zeros, L 1 (x) has one zero, L 2(x) has two zeros, and L 3 (x) has three
zeros in the interval x > o.
We give below two equations which can be transformed into Laguerre
equations by a proper substitution.
1. The substitution

(43.22)
Lesson 43C PROPERTIES OF LAGUERRE POLYNOMIALS Lk(x) 627

in the equation
(43.23) xu" + (1 + x)u' + (k + l)u = 0
will transform it into the Laguerre equation (43.1). Since y = Lk(X) is a
solution of (43.1), it follows by (43.22) that
(43.24)
is a solution of (43.23).
2. Similarly, the substitution
(43.25)
in the equation
(43.26) u" + (_1_2 + 2k + 1_ .!) u = 0'
4x 2x 4
will transform it into the Laguerre equation (43.1). Since y = Lk(x) is a
solution of (43.1), it follows by (43.25) that
(43.27)
is a solution of (43.26).

LESSON 4-3C. Some Properties of Laguerre Polynomials L,,(x).

A. Analog of Rodrigue's Formula for the Legendre Polynomial.

(43.3) Ln(x) = eZ :;n (xne- Z).

Proof. We shall show first that the formula is valid for the first three
Laguerre polynomials. By (43.3),
(a) Lo(x) = eZxOe-Z = 1,
L 1 (X) = eZ dx
d (-Z)
xe = eZ(e-z - xe -Z) = 1 - x,

L 2 (x) = eZ d:22 (x 2e-Z) = eZ d: (2xe- Z - x 2e-Z)

= eZ(2e- Z _ 4xe-z + x 2e-Z)
= 2 - 4x + X2.
La(x) = eZ d:33 (x 3e-Z) = eZ d:22 (3x 2e- Z - x 3e-Z)
=eZ d (6 xe -z - 62-",+
xe xe3-"')
dx
= e"'(6e-Z - 18xe- z +
9x 2e-z - x 3e- Z)
= 6 - 18x + 9x 2 - x3 •
You can verify that each of the formulas in (a) agrees with those in (43.2).
628 SERIES METHODS Chapter 9

The proof of (43.3) for the general case follows. First verify that

(43.31) [f(x)g(x)](") = t;" k!(n

, k)! f(k)(x)y(n-k)(x),
i.e., the coefficients of the nth derivative of f(x)g(x) are the same as the
coefficients in the binomial expansion of (x + 1)n. In (43.31), letf(x) = x"
and g(x) = e- X • It then becomes
-'"
_u_ ( n -X) '"
= 4.J n. ( n)(k)( -x)(n-k)
(43.32)
dx n x e k'(
10=0 • n _ k)'. x e .
When
(43.33) n = 1: d: (xn) = nxn - 1
,

2
d
n = 2: dx 2 (xn) = n(n - l)x.. - 2,

n = 3: d:: (xn) = n(n - l)(n - 2)x,,-a,

n = k:
dx k
(xn) = n(n - l)(n - 2) ... (n - k + l)x,,-k
n! n-k
(n - k)! x .
Also
(43.34)

Substituting in (43.32) the last equality of (43.33), the equality (43.34)

and then multiplying the result hy e"', we have

(43.35)
dn X
eX dx" (x"e- ) = (n!)2 t; n (1),,-10 k
_ k)!]2 x.. - .

You can verify that the right side of (43.35) also will give the first four
Laguerre polynomials in (43.2). It is, in fact, another form of writing the
polynomial solutions of the Laguerre equation. We shall now show that
the right side of (43.35) and of the last equation of (43.2) are equivalent.
In the summation of (43.35), let p = n - k. Therefore k = n - p
and when k = 0, p = n and when k = n, p = O. Hence the right side
of (43.35) can be written as

(43.36) (n.1)2 4.J ( _(-l)P P

}'( 1)2 x .
P_" n p. p.

Therefore, by (43.36), with n = k,

,)2 (-l)P k
(43.37) (k . (k _ p)!(p!)2 x .
P="
Lesson 43C PROPERTIES OF LAGUERRE POLYNOMIALS Lk{x) 629

The summation in (43.37) is now the same as the right side of Lk(x) of
(43.2).

B. Integral Property of Laguerre Polynomials.

Theorem 43.4. Let Lo(x), li1(x), L 2(x), . ", be Laguerre polynomial

solutions of (43.1). Then

(43.41) i'" e-XLm(x)L.. (x) dx = 0, if m ¢ n.

Proof. Let Um and Un be two solutions of (43.26). Therefore

(a) Um
" + (_1
4x2
+ 2m2x+ 1_ !)
4 Um
= 0,
U.. " + (1
4x 2 + 2n + 1- 4
1) Un _0
- .

Multiplying the first by Un, the second by -Um and adding the resulting
equations, we obtain
(b) unUm" - umun" = (-n--xm)
- UmUn·

Integrating between the limits 0, 00, and recognizing that the left side of
(b) is (d/dx)(Unu m' - umun')', there results

(c) lim [Unum' - Umun'li = (n - m)j'" ! UmU n dx, m ¢ n.

h_", 0 X

By (43.27), solutions of (a) and their respective derivatives are,

(d) Um = e- 2x 1 / 2 Lm,
X
/

Um' = -!e-x/ 2X1/ 2Lm + !e-x/2x-l/2Lm + e-x/2Xl/2Lm',

Un = e- x / 2X 1/ 2Ln,

Un' = -!e-x/ 2X 1/ 2Ln + !e-x/2x-l/2Ln + e- x/ 2x 1/2L ..'.

Inserting the above values in (c), we obtain

(e) lim [e-XxLn(x)Lm'(X) - e-XxLm(x)Ln' (x)li

h-too CIO

= (n - m) fo e-XLm(x)Ln(x) dx.

The left side of (e) is zero when x = 0, and, by (27.113)(d), it approaches

zero for positive h as h --+ 00. By hypothesis m ¢ n. Therefore (e)
simplifies to
(f)
630 SERIES METHODS Chapter 9

Because of (f), the set of Laguerre polynomials is said to be orthogonal

on I: 0 x < 00 with respect to the weight function e-"'. Compare with
Definition 41.5 for an orthogonal set of functions.

EXERCISE 43
1. Find a series solution of each of the following differential equations.
(a) xy" + (1 - x)y' + !Y = O. (b) xy" + (1 - x)y' + 1.2y = O.

2. Verify that the substitution (43.22) in (43.23) gives the Laguerre equation
(43.1).
3. Verify that the substitution (43.25) in (43.26) gives the Laguerre equation
(43.1).
4. Use formula (43.3) to find L4(X).
5. Find a series solution of each of the following.
(a) xy" + +
(1 x)y' + y = O. Hint. See (43.23).
(b) xy" + +
(1 x)y' + 2y = O. (c) xy" (1 + + x)y' + iY = O.
6. Find a series solution of each of the following.

(a) y" + (_1_ + 2x - !) y = O. Hint. See (43.26).

n n
4x 2 4

(b) y" + (4!2 + - y = o. (c) y" + +; - y = O.

7. Assume that a function/(x), defined on the interval (0, co), can be represented
by a series of Laguerre polynomials, i.e., assume
(a) lex) = coLo(x) + qLl(X) + c2L2(X) + ....
Show that the coefficients Co, ct, C2, ••• are given by

10'" e-ZLk(x)/(x) dx
10'" e-"'[Lk(x»)2 dx
Hint. Multiply (a) by e-ZLk(x), integrate from 0 to co, then use (43.41).

ANSWERS 43
1. (a) replace k by 1 in (43.19). (b) replace k by 1.2 in (43.19).
5. (a) y = e-zLo(x) = e- Z. (b) y = e-zLl(x) = e-Z(l - x).
(c) y = e- z L1I2(X).
6. (a) y = e-z/2x1l2Lo(x) = e-z/2x 1l2.
(b) y = e-z/2x1l2Ll(X) = e- z/2 x1l2(1 - x).
(c) y = e- z/2 X1I2L1I2 (x).
 Chapter 10

Numerical Methods

Introduction. In many practical problems involving differential equa-

tions, what is often wanted is a table of values of a solution y = y(x),
satisfying given initial conditions, for a limited range of values of x near
the initial point Xo. For example, we may want values of y(x) when
x = Xo + h, Xo + 2h, Xo + 3h, etc., where h is 0.05 or 0.1 or 0.2, etc.
Even when a solution y = y(x) of a differential equation can be written
in terms of elementary functions, it may at times be easier to obtain this
limited table of values by the numerical methods we shall describe in this
chapter, rather than from the analytic solution itself. This statement is
especially true when the solution is an implicit one. As we have remarked
on numerous occasions, implicit solutions are usually such complicated
expressions that it is almost impossible to find the needed function g(x)
which it implicitly defines, or to calculate values of y for given values of x.
Moreover, in a great many problems, a solution of a differential equation
cannot be expressed in terms of elementary functions for the very good
reason that the differential equation does not have any such solution. For
example, the equation y' = 1/yx3 + 1 does not have a solution in terms
of elementary functions.
By a numerical solution of a differential equation, we shall mean a
table of values such that for each x there is a corresponding value of y(x).
In this sense, even an explicit solution in terms of an elementary function,
such as y = sin x or in terms of a nonelementary function such as y =- J o(x),
is a numerical solution. For each x, we can look in a table and find a value
of sin x or of the Bessel function J o(x).
You will soon discover that the work involved in computing a table of
values even when a moderate degree of accuracy is needed is laborious
and tedious. However, with the current increased availability of high-
speed computing machines, it is possible to have them make many burden-
some calculations for you. But it will still be necessary for you to know
how and what to feed these machines.
In the following lessons of this chapter, we shall explain various methods
by which a numerical solution of a differential equation can be obtained.
631
632 NUMERICAL METHODS Chapter 10

To keep the arithmetical calculations within reasonable bounds, and also

to be able to check the accuracy of our results, we have selected simple
differential equations for our examples, ones which can be solved explicitly
in terms of elementary functions. You must keep in mind, however, that
we are using them only to illustrate a method. These same methods can
be employed to find numerical solutions of more complicated equations.
We illustrate the methods we shall develop for finding a numerical solu-
tion of a first order differential equation by applying them to an equation
of the form y' = f(x,y) for which y(xo) = Yo. We assume in our discus-
sion that a unique particular solution of this equation, satisfying the given
initial condition, exists. (For criteria which will give a sufficient condition
for the existence of this unique particular solution, see Theorem 58.5. If
the criteria of the existence theorem are too difficult to apply, a practical
man will usually know from his experience and from the nature of the
physical problem which gave rise to the differential equation whether a
solution exists.)
The methods we shall develop for finding a numerical solution of a dif-
ferential equation have been divided into three categories. In one cate-
gory, we include those methods which need only the given equation
y' = f(x,y) and the initial condition y(xo) = Yo in order to start the con-
struction of a table of values of y for given values of x. They are therefore
called appropriately starting methods. In a second category, we include
those methods which need more values of y than only the initial condition
y(xo) = Yo before they can be used. These methods are therefore called
appropriately continuing methods since they can be used to continue
the construction of the table only after the needed preliminary values
have been obtained by starting methods. In a third category we include
those methods whose only purpose is to correct values of y obtained by
starting and continuing methods. These methods are therefore called
appropriately corrector methods.

LESSON 44. Starting Method. Polygonal Approximation.

In this lesson we shall show, by a method called the polygonal method,

how to start the construction of a table of approximate values of y(xo + h),
y(xo + 2h), ... , where h is a constant and y(x) is the unique particular
solution of

(44.1) y' = f(x,y)

satisfying the initial condition

(44.11) y(xo) = .Yo·

We proceed as follows (see Fig. 44.12). By (44.1) and (44.11), we determine

Lesson 44 STARTING METHOD. POLYGONAL ApPROXIMATION 633

y(Xo + 2h)=y(xl + h)=

y(xo+h)= y(X2)=Y2
y(x\)=y\

Xo

Figure 44.12

the value of y' at (xo,Yo). The equation of the tangent to the integral
curve y(x) at the point (xo,Yo) is therefore
(44.13) y - y(xo) = (x - xo)y'(xo).

This tangent line will intersect the line x = Xo +h in a point whose

ordinate is [in (44.13), replace x by Xo + h],
(44.14) Y(Xo + h) = y(xo) + y'(xo)h.
In our table, we can now record the value of y(xo + h) obtained by
(44.14). It is an approximation to the actual value of y(xo h). The +
error in this computation is shown as EI in Fig. 44.12.
For convenience we write (XI,YI) for the point [xo h, y(xo +
h)]. At +
[xI,yd we repeat the above procedure as if (Xt,YI) were actually on the
integral curve. We find the equation of a line through (XI,YI) having a
slope obtained by (44.1), with x = Xl, Y = YI. If we call this slope
Y'(XI)' then the equation of this line is

(44.15)

Its intersection with the line x = Xo + 2h == Xl + h is

(44.16)
In our table, we can therefore now record the approximate value of
y(xo+ 2h). The error in the computation is shown as E2 in Fig. 44.12.
Continuing in this manner, we find
(44.17) Y(Xo + 3h) == Y(X2 + h) = Y(X2) + y'(x2)h,
634 NUMERICAL METHODS Chapter 10

which is an approximate value of y(xo + 3h). And in general, we find

(44.18)

which is an approximate value of y[xo + (n + l)h].

E%ample 44.2. By means of the polygonal method, find approximate
values when x = 0.1, 0.2, 0.3 of the particular solution of the differential
equation
(a) y' = x2 + y,
for· which y(O) = 1. Take h = 0.1 and h = 0.05.
Solution. Comparing the initial condition with (44.11), we see that
Xo = 0, Yo = 1. By (44.18), with h = 0.1 and Xn taking on the values
0, 0.1, 0.2, we obtain
(b) y(O + 0.1) = + y'(O)(O.I),
y(O.I) = y(O)
(c) y(O.1 + 0.1) = y(0.2) = y(O.I) + y'(O.I)(O.I),
(d) y(0.2 + 0.1) = y(0.3) = y(0.2) + y'(0.2)(0.1).

By (a) and the initial conditions, y'(O) = 0 + 1 = 1. Therefore (b)

becomes
(e) y(O.I) = 1 + 1(0.1) = 1.1.

By (a), when x = 0.1, y = 1.1, we find y'(O.I) = (0.1)2 + 1.1 = 1.11.

Therefore (c) becomes
(f) y(0.2) = 1.1 + 1.11 (0.1) = 1.211.

With x = O.?, y = 1.211, we find by (a), y'(0.2) = (0.2)2 + 1.211 =

1.251, and by (d)
(g) y(0.3) = 1.211 + 1.251(0.1) = 1.336.

In numerical solutions, it is usually desirable to construct a table in

which all relevant computations are systematically recorded. For the
above example, the table has the appearance of Table 44.21. The actual

Table 44.21

Xn = y(x n) = y'(x n ) = hy'(x n ) y(x n + h) = Actual Values of

Y(Xn) =

0.0 1.0 1.000 0.1 1.1 1.000

0.1 1.1 1.110 0.111 1.211 1.106
0.2 1.211 1.251 0.125 1.336 1.224
0.3 1.336 1.360
Lesson 44 STARTING METHOD. POLYGONAL ApPROXIMATION 635

solution of (a) satisfying y(O) = 1 is y = 3e% - x 2 - 2x - 2. By it we

obtained the figures in the last column of Table 44.21.
With h = 0.05, our table of values becomes, by (a) and (44.18), Table
44.22.
Table 44.22

Actual Values of
x,. y(x,,) y'(x,.) hy'(x n ) y(x n + h) y(x,.)

0.0 1.0000 1.0000 0.0500 1.0500 1.0000

0.05 1.0500 1.0525 0.0526 1.1026 1.0513
0.1 1.1026 1.1126 0.0556 1.1582 1.1055
0.15 1.1582 1.1807 0.0590 1.2172 1.1630
0.2 1.2172 1.2572 0.0629 1.2801 1.2242
0.25 1.2801 1.3426 0.0671 1.3472 1.2896
0.3 1.3472 1.3596

1.1

1.0

o 0.05 0.10 0.15 0.20 0.25 0.30

Figure 44.23

A graph of the actual solution and of the polygonal approximations are

shown in Fig. 44.23.
636 NUMERICAL METHODS Chapter 10

General Comment on Errors in a Numerical Computation. The

question of detennining the errors in a table of numerical values is an
extremely complex one. There are in general four types of errors.
1. Arithmetical errors made by the individual or due to the misbehavior
of a calculator.
2. Rounding of! errors due to stopping with a certain decimal place.
3. Formula errors due to the use of an approximating fonnula to obtain a
numerical answer.
4. Cumulative errors. At each step in a lengthy process an error occurs
that is carried along to the next stage.

We shall assume that the error due to the rounding off of a decimal can
be compensated for by retaining a sufficient number of decimal places at
each step, so that the accuracy desired in the last tabulated value of
y(xo + nh) will not be affected even in the most unfavorable circum-
stances, as for example when we have to drop the 49 in 0.3265349 in order
to round off the decimal to five places. If, however, too many steps are
required to reach y(xo +
nh), it may not always be possible to attain this
desired objective, but then the loss in rounding off at one step may be
offset by a gain at another. As regards the other types of errors, we shall
comment on each of them at the appropriate time.

Comment 44.3. Comment on Error in Polygonal Method.

1. In this method, we start with a point and a slope that agree with the
solution of (44.1) satisfying (44.11). But since the straight line drawn at
this point (xo,Yo) may not be the actual integral curve y = y(x), a formula
error is introduced in the first step. It is represented by El in Fig. 44.12.
At each successive point (X2,Y2), (xa,Ya), ... , at least two errors are intro-
duced, a starting error and a formula error, so that the cumulative error
may soon become large. Hence this method is useful only if too great
accuracy is not required or if h is very small.
2. Comparing (44.14) with the Taylor series fonnula, see (37.35),

y(xo + h) = y(xo) + Y'(xo)h + h2 + ...

+ y"(xo) hn + y,,+I(X) hn+1
n! (n+1)!'

we see that (44.14) contains the first two terms of a Taylor series. Its
remainder or error term is therefore, see (37.36),

(44.31)
Lesson 44 STARTING METHOD. POLYGONAL ApPROXIMATION 637

where X is a value of x in the interval under consideration of width h. Let

(44.32) E(xo + h) = error in computing y(xo + h) by (44.14).

Then, by (44.31),
(44.33)

where c = y"(X)/2!. Let us now divide the h interval in half and com-
pute y(xo + h) in two steps. If we have some basis for believing that for
a small h, y"(X) changes rather slowly in this h interval so that the varia-
tion in the value of y"(X) in each h/2 interval is negligible, then we com-
mit a small error in using the same c of (44.33) for each half interval.
Therefore, by (44.33), the error in computing y(xo + h/2) for a half
interval h/2 is approximately
2 2
(44.34) E ( Xo + 2h) = c (h) ch = !E(xo + h).
2 = T

This means that the error in y (XO + is approximately equal to one-

fourth the error of y(xo + h). Hence the error in computing in two steps
the value of y(xo + h), which we write as y [(xo + will have

an inherited error in the starting value of y (xo + plus its own for-
mula error. Since each error equals one-fourth the error in y(xo + h), the
total error in y [(xo + + i.e., the error in computing y(xo + h),
in two steps, is approximately equal to one-half the error in y(xo + h)
computed in one step. Hence

(44.35) E[ (xo + = !E(xo + h),

approximately. Let Y(xo + h) be the actual value of the solution. Then
(44.36) Y(xo + h) - y(xo + h) = E(xo + h),
Y(xo + h) - y[(xo = E[(X O +
Subtracting the second equation in (44.36) from the first, we obtain
(44.37)

y[(xo - y(xo + h) = E(xo + h) - E[(Xo +

638 NUMERICAL METHODS Chapter 10

Substituting (44.35) in the right side of (44.37), the following equations

result.

(44.38) (a) E[(XO = y[(XO - + y(xo h),

(b) !E(xo + h) = y[(xo - +

y(xo h).

The first formula says that the error in the value of y(xo + h) computed
in two steps is equal to the difference in values of y(xo + h) computed in
two steps and in one step.
Formulas (44.38) give us a means of estimating errors at each step.
Their accuracy does not depend on a knowledge of the size of y"(X),
which we do not know, but only on the variation of y"(X) over a small
interval. We have assumed this va.riation to be negligible, an assumption
which is not unreasonable if y"(x) changes slowly over h. For example,
from Tables 44.22 and 44.21, with Xo = 0, h = 0.1,

(a) y [(xo + + = y(0.05 +0.05) = y(O.l) = 1.1026,

y(xo + h) = y(O.l) = 1.1.
Therefore by (44.38)(a),
(b) E(0.05 + 0.05) = 1.1026 - 1.1 = 0.0026,
which is the approximate error of y(O.l) computed in two steps. The
actual error by Table 44.22 is 1.1055 - 1.1026 = 0.0029.
3. A check on errors, which practical people frequently use, and which
seems to work, is to make all calculations over again with an h half the
size of the original one. If the results obtained with the smaller h agree
with those obtained with the larger h to k decimal places, after being
properly rounded off, then it is assumed that their common numerical
value has k decimal place accuracy. For example, by Tables 44.21 and
44.22, y(0.3) computed with h = 0.1 and with h = 0.05 agree to one
decimal place. Hence we assume that y(0.3) = 1.3 has one-decimal
accuracy.
4. This method does not give a check on arithmetical errors. However,
if there is little agreement between the values obtained by using hand
h/2, all arithmetical computations should be checked. If the arithmetic
is correct, h should be reduced.
Comment 44.39. We have headed this lesson "Starting Method." It
can also be used as a continuing one. For instance, we can, in Example
44.2, use this method to find y(O.4), y(0.5), etc. However its low order of
accuracy makes it a poor continuing method. In the next and succeeding
lessons, we shall present better ones.
Lesson 44-Exercise 639

EXERCISE 44
1. Add to Table 44.21, values of y when x = 0.4 and 0.5. Also add actual
values obtained from the solution y = 3eX - x 2 - 2x - 2.
2. Add to Table 44.22, values of y when x = 0.35, 0.4, 0.45, 0.5. Also add actual
values obtained from the solution y = 3ex - x 2 - 2x - 2.
3. Reconstruct Table 44.21 by applying error formulas (44.38) to correct the
entries at each step before proceeding to the next one. For example, by
(44.38)(b), with Xo = 0, h = 0.1,
E(O.l) = 2[y(0.05 + 0.05) - y(O.l)] = 2(1.1026 - 1.1) = 0.0052.
Hence the corrected value of y(O.l) = 1.1 +
0.0052 = 1.1052. Record
this new value of y(O.l) in your reconstructed Table 44.21. Starting with this
corrected value of y(O.l) = 1.1052, compute y(0.2) in one step and two steps.
Then apply error formulas (44.38) to correct y(0.2), etc. Compare with
previous results and with actual values of the solution.
4. Find approximate values when x = 0.05, 0.1, 0.15, 0.2 of the particular
solution of the equation y' = X +y2 for which yeO) = 1. Take h = 0.05.
5. In problem 4, compute y(O.I) and y(0.2) with h = 0.1. Compare with the
value of y(0.2) obtained in 4. In the absence of a solution, how many deci-
mal place accuracy could you assume in the value of y(0.2)? Hint. See
Comment 44.3-3.
6. By use of error formulas (44.38), correct the value of y(O.l) obtained in
problems 4 and 5. Using this corrected figure, proceed to find y(0.2) in two
steps and in one step. Then correct y(0.2).
7. Find approximate values when x = 1.1, 1.2, 1.3 of the particUlar solution of
+
the differential equation y' = x 2 y2 for which y(l) = 1. Take h = 0.1.
What is the approximate formula error in y(Ll)? [Hint. Calculate y(1.1)
in two steps and then use error formula (44.38).] In the absence of a solu-
tion or error formula, how could you estimate the error in y(1.3)?
8. Find an approximate value when x = 1, of the particular solution of the
equation y' = 1/(1 +
x 2 ) for which yeO) = O. Use h = 0.2 and proceed
as follows. Find y(0.2) in one step and in two steps. Correct y(0.2) by means
of (44.38). Then find y(0.4) in one step and in two steps. Correct y(0.4) by
means of (44.38). Continue in this way until you reach y(l). Show how this
value of y(l) can be used to approximate 'If'. Compare with actual value of 'If'.
Hint. The solution of y' = 1/(1 + x 2) for which yeO) = 0 is y = Arc tan x.
Therefore y(l) = 'If'/4 so that 'If' = 4y(1).
9. Follow the procedure outlined in problem 8 to find an approximate value
when x = 1, of the particular solution of the equation y' = y for which
yeO) = 1. Show how this value of y(l) can be used to approximate e.
Compare with actual value of e. Hint. The solution of y' = y for which
yeO) = 1 is y = eX. Therefore y(l) = e.
10. Find an approximate value when x = 2. of the particular solution of the
equation y' = l/x for which y(l) = O. Take h = 0.2 and follow the pro-
cedure outlined in problem 8. Show how this value of y(2) can be used to
approximate log 2. Hint. The solution of y' = l/x for which y(l) = 0 is
y = log x. Therefore y(2) = log 2.
n. By formulas (44.14) and (44.31), we have
y(xo +
h) = y(xo) +
hy'(XO) E,+
2
where E = y"(X)h /2 and X is a value of x in the interval (xo,xo h). +
E is the formula error due to stopping with the y'(xo) term in a Taylor series.
640 NUMERICAL METHODS Chapter 10

If we knew which value of x to choose for X in this interval, we would know

the exact value of E. But we don't. However, if M is the maximum value
of ly"(x)1 in the interval (xo,xo + h), then lEI Mh 2 /2. We can thus
establish an upper bound of the error, in an interval of width h, in using
formula (44.14) to compute y(xo + h). Call this error El and call M 1 the
maximum value of y"(x) in this interval. Designate by E2, Ea, ... , En,
the errors in each additional interval of width h, and M 2, ••• , M n the
maximum value of Iy"(x) I in each such additional interval. Then the upper
bound of the total error E for all intervals is

(44.4)

Let M be the largest of the numbers M l , M 2 , " ' , Mn. Then, by (44.4),
2
h
(44.41) IEI::5:
-
-nM
2 '

where M is the maximum value of Iy"(x) I in the interval (xo, Xo + nh).

(a) Use formula (44.41) to compute the upper bound of the error in the
value of log 2 if in problem 10 we had used h = 0.2 without corrections.
Hint. h = 0.2, n = 5, M = max. ly"l = max.I-1/x 2 1 = t in the
interval (1,2).
(b) What is the largest value of h that can be used to insure that the upper
bound of the error in the computation of log 2 in problem 10 is 0.005,
if no corrections were made? Hint. By (44.41), we want an h such that
h2 nM/2 0.005. Remember hn = 1.

ANSWERS 44
1. y(O.4) = 1.471, y(0.5) = 1.634. Actual values: 1.51547, 1.69616.
2. y(0.35) = 1.4191, y(O.4) = 1.4962, y(0.45) = 1.5790, y(0.5) = 1.6681.
Actual values 1.43470, 1.51547, 1.60244, 1.69616.
3. y(0.2) = 1.2237, y(0.3) = 1.3557, y(O.4) = 1.5107, y(0.5) = 1.6902.
4. y(0.05) = 1.0500, y(O.l) = 1.1076, y(0.15) = 1.1739, y(0.2) = 1.2503.
5. y(O.l) = 1.100, y(0.2) = 1.231.
Can assume only zero decimal place accuracy if rounded off to one decimal.
6. y(O.l) = 1.1152, y[(O.l + 0.05) +
0.05] = 1.2598, y(O.l +
0.1) = 1.2496,
y(0.2) = 1.2700.
7. y(1.1) = 1.2000, y(1.2) = 1.4650, y(1.3) = 1.8236, E(1.1) = 0.0312.
Would need to make all calculations over again with h = 0.05, see Com-
ment 44.3-3.
8. y(0.2) = 0.1980, y(O.4) = 0.3815, y(0.6) = 0.5415, y(0.8) = 0.6756,
y(l) = 0.7860. Actual value: 11' = 3.14159.
9. y(0.2) = 1.2200, y(O.4) = 1.4884, y(0.6) = 1.8158, y(0.8) = 2.2152,
y(l) = 2.7026. Actual value: e = 2.71828.
10 y(1.2) = 0.1818, y(1.4) = 0.3355, y(1.6) = 0.4688, y(1.8) = 0.5864, y(2) =
0.6917., Actual value: log 2 = 0.6931.
n. (a) lEI 0.1. (b) h = 0.Q1.
Lesson 45 AN IMPROVEMENT OF THE POLYGONAL METHOD 641

LESSON 45. An Improvement of the Polygonal

Starting Method.
Let y(x) be a particular solution of
(45.1) y' = J(x,y)

satisfying the initial condition

(45.11) y(xo) = Yo·

In the polygonal method, we found the approximation, see (44.14),

(45.12) y(xo + h) = y(xo) + y'(xo)h.

It is possible to improve this estimate using a method similar to the one
of the previous lesson. In this previous lesson, we found y(xo h) of +
(45.12) by drawing, at (xo,Yo), a tangent line to the integral curve y(x),
and determining its intersection with the line x = Xo +
h. It is marked
R in Fig. 45.13. The error in y(xo +
h) is shown as E 1 • The point P in

FR[xo + h, y(Xo + h)]=

R[xo + h, y(xo) + y'(xo)h]

Xo xo+ E.2

Figure 45.13

the figure is the mid-point of the segment of this tangent line between Q
and R. The coordinates of P were obtained by using the mid-point for-
mula of analytic geometry. Substituting the coordinates of P in (45.1),
we find
(45.14) y(xo) + y'(xo) ;] .
642 NUMERICAL METHODS Chapter 10

The equation of the line through (xo,Yo) with slope (45.14) is

(45.15) y = y(xo) + (x - xo)y' (xo +

= y(xo) + (x - xo)f[Xo + y(xo) + y'(xo)
Its intersection with the line x = Xo + h is
(45.16) y(xo + h) = y(xo) + h![XO + Yo + y'(xo)].

We shall now prove that the right side of (45.16) is equivalent to the
first three tenus of a Taylor series instead of only the first two as in (45.12).
Hence the error in the approximate value of y(xo + h) of (45.16), shown
as E2 in Fig. 45.13, usually will be smaller than the error E 1 ••
Proof. By (38.12) and (38.13), a Taylor series expansion of a function
of two variables is, with x = Xo + a, y = Yo + b,
(a) !(xo + a, Yo + b) = !(xo,Yo) + a +b + ....
Therefore with a = b= y'(xo),

h h '( )] h a!(xo,Yo)
(b) ! [ Xo + 2"' Yo + 2" y xo = !(xo,Yo) + 2" ax

+ y'(xo) a!(x;;yo) + ....

Substituting (b) in (45.16), and replacing !(xo,Yo) by its equal y'(xo) of
(45.1), we obtain
(c) y(xo + h) = y(xo) + hy'(xo)
+ h2 [a!(XO,Yo) + y'(x ) a!(XO,Yo)] + ....
2 ax ° ay
By differentiation of (45.1)-we assume the derivatives exist-we obtain

(d)
" _ a!(x,y) + a!(x,y) dy .
y - ax ay dx

-By (37.36), the error or remainder term El = Y"<;l) h2 if the series includes the

first two terms of a Taylor series; the error term E2 = h3 if the series includes

the first three terms. Hence E2 < El if

"'(X )
h3 < Y"(X ) h2, i.e., if y"'(X2 )
3
< Ii, y"(X1 )·
For a small h, 3/h is large.
Leason 45-Exercise

Hence
(e) y"(xo) = of(x;;yo) + y'(xo).

Therefore by (e), we can write (c) as

2
h
(f) y(xo + h) = y(xo) + hy'(xo) + "2 y"(xo) + ....

A comparison of (f) with the Taylor series (37.27) shows, with x = xo + h,

that the first three terms of each are the same.
NOTE. Use of this method is therefore permissible only if of(x,y) and
. h
of(x,y) .
- - extst at (xo,yo).
oy
Example 45.2. Find, by the method of this lesson, an approximate
value, when x = 0.1, of the particular solution of

(a) y' = x2 +y
for which y(O) = 1.
Solution. Comparing the initial condition with (45.11), we see that
xo = 0, Yo = 1. Therefore by (a), y'(O) = 0 + 1 = 1. Hence with
xo = 0, h = 0.1, (45.16) becomes
(b) y(O.I) = 1 + 0.1f(0.05, 1 + 0.05).
Here f(x,y) = x2 + y. Therefore
(c) f(0.05, 1.05) = (0.05)2 + 1.05 = 1.0525.

Substituting (c) in (b), we obtain

(d) y(O.I) = 1 + 0.1053 = 1.1053.

The actual value of y(O.I) is 1.1055. In the polygonal method using two
steps, we obtained a value of 1.1026. Note the greater accuracy of the
above method.
EXERCISE 45
Apply the method of this lesson to solve the problems which follow.
1. Starting with y(O.!) = 1.1053, compute y(0.2) and y(0.3) of Example 45.2.
Take h = 0.1. Compare with the values found in Lesson 44 and in Exer-
cise 44,3.
2. Find y(O.!) of Example 45.2 in two steps, Le., first find y(0.05) and then
y(0.05 + 0.05). Compare with actual value of y(O.!) = 1.1055. Do the
same for y(0.2).
644 NUMERICAL METHODS Chapter 10

3. Following the steps in Comment 44.3-2, develop error formulas, comparable

to those in (44.38), for the numerical method of this lesson. Hint. Since
formula (45.16) is equivalent to the first three terms of a Taylor series, the
remainder or error term, by (37.36), is ylll(X)h3 /3!, where X is a value of
x in the interval under consideration of width h. Ans.

+ + = Hy [(x o n
+ + - + h)} ,
n
(45.3) E [(xo y(xo

E(xo + h) = {y [ (xo + + - y(xo + h)} .

4. Find approximate values when x = 0.05, 0.1, 0.15, 0.2 of the particular
lIolution of the differential equation y' = X + y, for which yeO) = 1. Take
h = 0.05.
5. In problem 4, compute y(O.l) and y(0.2) with h = 0.1. Compare with the
value of y(0.2) obtained in 4. In the absence of a solution in terms of ele-
mentary function or an error formula, how many decimal place accuracy
could you assume in your value of y(0.2). Hint. See Comment 44.3-3
Solve the equation and compare with actual value of y(0.2).
6. Find approximate values when x = 1.1, 1.2, 1.3 of the particular solution
+
of the equation y' = x 2 y2 for which y(l) = 1. Take h = 0.1. What
is the approximate formula error in y(1.1)? Hint. Calculate y(1.1) in two
steps and then apply error formula (45.3).
7. Find approximate values when x = 0.2, 0.4, 0.6, 0.8, 1 of the particular
solution of the equation y' = 1/(1 +
x 2) .for which yeO) = O. Take h = 0.2.
Use this value of y(l) to approximate 11'. Hint. See Exercise 44,8. Compare
results.
8. Find an approximate value when x = 1 of the particular solution of the
differential equation y' = y for which yeO) = 1. Take h = 0.1. Use the
value of y(l) to approximate e. Hint. See Exercise 44,9. Compare results.
9. Find an approximate value when x = 2 of the particular solution of the
equation y' = l/x for which y(l) = O. Take h = 0.25. Use this value of
y(2) to approximate log 2. Hint. See Exercise 44,10. Compare results.
10. By formulas (45.16) and the error term as given in problem 3, we have

y(xo + h) = y(xo) + hf[xo + Yo + yl(XO)] + E,

where E = y"I(X)h3 /3! Following the procedure outlined in Exercise 44,11,
show that the upper bound of the error E in the interval (xo, Xo nh) is +
3
h
(45.31) lEI 3! nM,

where M is the maximum value of yl//(x) in the interval (xo,xo + nh).

(a) Use formula (45.31) to compute the upper bound of the error in the
value of log 2 as found in problem 9 above. Hint. h = 0.25, n = 4,
M = max Iyllli = max 1:31 = 2 in interval (1,2).
(b) What is. the largest value of h which can be used to insure that the upper
bound of the error in the computation of log 2 is less than 0.005? Hint.
By (45.31) we want an h such that h3 nM/3! < 0.005 and remember
nh = 1.
Les80n 46 STARTING METHOD--TAYLOR SERIES 645

ANSWERS 45
1. y(0.2) = 1.2237, y(0.3) = 1.3586. .
2. y(0.05) = 1.0513, y(0.05 +
0.05) = 1.1055, y(0.15) = 1.1630,
y(0.15 +
0.05) = 1.2242.
4. y(0.05) = 1.0525, y(0.1) = 1.1103, y(0.15) = 1.1736, y(0.2) = 1.2427.
5. y(0.1) = 1.11, y(0.2) = 1.2421. Can assume two decimal place accuracy.
Actual value: y(0.2) = 1.2428.
6. y(1.1) = 1.2313, y(1.2) = 1.5506, y(1.3) = 2.0106, y(1.05) = 1.1077,
y(1.05 +
0.05) = 1.2334; E(1.1) = 0.0028.
7. y(0.2) = 0.1980, y(O.4) = 0.3815, y(0.6) = 0.5415, y(0.8) = 0.6757,
y(1) = 0.7862.
8. y(0.1) = 1.105, y(0.2) = 1.2210, y(0.3) = 1.3492, y(O.4) = 1.4909,
y(0.5) = 1.6474, y(0.6) = 1.8204, y(0.7) = 2.0115, y(0.8) = 2.2227,
y(0.9) = 2.4561, y(1) = 2.7140.
9. y(1.25) = 0.2222, y(1.5) = 0.4040, y(1.75) = 0.5578, y(2) = 0.6911.
10. (a) lEI 0.0209. (b) h = 0.12 to two decimal places.

LESSON 46. Starting Method-Taylor Series.

In Lesson 44, we found a numerical solution of the differential equation

(46.1) y' = f(x,y),
for which
(46.11)

by a method which is equivalent to using a Taylor series to terms of the

first order; in Lesson 45 by a method which is equivalent to using a Taylor
series to terms of the second order. These methods suggest that greater
accuracy may be achieved if, for a starting method, we use a Taylor series
to terms of order greater than two. There are, however, two practical
difficulties to the use of a Taylor series.
1. The function f(x,y) may not have a Taylor series expansion over the
interval in which a solution is desired. For example, if y' = f(x,y) =
+
Vx y2, then y" and higher derivatives do not exist at x = O.
2. If f{x,y) has a Taylor series expansion, it may be extremely difficult to
obtain the derivatives needed in formula (37.27). For example, try
taking a few derivatives of f{x,y) = yx 3y + xy3.
If these two difficulties are not present, then a Taylor series is indeed
a good starting method. By (37.27), a Taylor series has the form

(46.12) y(xo + h) = y(xo) + Y'(xo)h + h2 + h3

(4){ )
+y h4 + ....
646 NUMERICAL METHODS Chapter 10

It says in effect that if one knows the values of the function y(x) and its
derivatives at a point x = xo, then one can find the value of the function for
a neighboring point h units away. By direct substitution in (46.12), we can
therefore find y(xo + 0.1), y(xo + 0.2), y(xo + 0.3), etc., provided the
series converges for these values of x.

LESSON 46A. Numerical Solution of y' = f(x,y) by Direct Sub-

stitution in a Taylor Series. We illustrate this method by meansof
an example.
Example 46.2. Find approximate values when x = 0.1, 0.2, 0.3, 0.4
of a particular solution of the differential equation
(a) y' = x2 + y,
for which yeO) = 1.
Solution. From (a) we obtain

(b) y' = x2 + y, y" = 2x + y', y'll = 2 + y", y(4) = y'll.

Hence when x = 0 and y = 1, we find from (b)

(c) y'(O) = 1, y"(O) = 0 + 1 = 1, y"'(O) = 2 + 1 = 3, y(4)(O) = 3.

By (46.12), with Xo = 0, we have

(d) y(h) = yeO) + y'(O)h + y"(O)

2!
h 2 + y'''(O) h 3 + y(4)(0) h4 + ...
3! 41 .
Substituting in (d), the initial condition and the values found in (c), we
obtain
(e)

By direct substitution in (e), we have, using only terms to h4 /8,

(f) y(O.I) = 1 + 0.1 + !(0.1)2 + -1(0.1)3 + f(0.1)4 = 1.1055125.
y(0.2) = 1 + 0.2 + -1(0.2)2 + -1(0.2)3 + t(0.2)4 = 1.2242000.
y(0.3) = 1 + 0.3 + -1(0.3)2 + 1(0.3)3 + f(0.3)4 = 1.3595125.
y(O.4) = 1 + 0.4 + -1(0.4)2 + 1(0.4)3 + f(0.4)4 = 1.5152000.

LESSON 46B. Numerical Solution of y' = f(x,y) by the "Creeping

up" Process. By the direct substitution method, more and more terms
of the series must be included, as h increases, in order to maintain a de-
sired degree of accuracy. These terms, if the derivatives of f(x,y) are
complicated functions, may be difficult to obtain. A more accurate
Leason 46B SOLUTION BY "CREEPING UP" PROCESS 647

method of using a Tayior series with the same number of terms is to keep
hfixed and "creep up" to the value of y(xo + nh) in successive steps. We
shall demonstrate by the example below how this method works. You
will soon discover that the greater accuracy is purchased at a price-more
labor.
Example 46.21. Solve the problem of Example 46.2 by the "creeping
up" process.
Solution. Using our basic equation (46.12), with h = 0.1 and Xo
equal successively 0,0.1,0.2,0.3, we obtain
(a) yeO + 0.1) = y(O.l)
= yeO) + y'(O)(O.I) + (0.1)2 + (0.1)3

+ (0.1)4 + ... ,
y(O.1 + 0.1) = y(0.2)
= y(O.I) + y'(O.I)(O.I) + (0.1)2

+ (0.1)3 + (0.1)4 + ... ,

y(0.2 + 0.1) = y(0.3)
= y(0.2) + y'(0.2)(0.1) + (0.1)2

+ (0.1)3 + (0.1)4 + ... ,

y(0.3 + 0.1) = y(OA)
= y(0.3) + y'(0.3)(0.1) + (0.1)2

+ (0.1)3 + (0.1)4 + ....

The first equation in (a) is the same as (d) of Example 46.2 with h = 0.1.
Hence by (f) of that example

(b) y(O.I) = 1.1055125.

By (b) of Example 46.2, the values of the derivatives of y(x) when x = 0.1
and y = 1.1055125 are

(c) y'(O.l) = (0.1)2 + 1.1055125 = 1.1155125,

y"(O.I) = 0.2 + 1.i155125 = 1.3155125,
y'"(O.l) = 2 + 1.3155125 = 3.3155125,
y(4)(0.1) = 3.3155125.
648 NUMERICAL METHODS Chapter 10

Substituting (b) and (c) in the second line of (a) gives

(d) y(0.2) = 1.1055125 + (0.1)(1.1155125) + (1.3155125)

+ (3.3155125)

+ 0.:1 (3.3155125) = 1.2242077.

When x = 0.2 and y(0.2) = 1.2242077, we find from (b) of Example 46.2,

(e) y'(0.2) = 1.2642077,

y"(0.2) = 1.6642077,
y"'(0.2) = 3.6642077,
y<4l(0.2) = 3.6642077.

Substituting (d) and (e) in the third line of (a), it becomes

(f) y(0.3) = 1.2242077 + (0.1)(1.2642077) + (0.005)(1.6642077)
+ (3.6642077)

+ (3.6642077) = 1.3595755.

And when x = 0.3, y(0.3) = 1.3595755, we find from (b), Example 46.2,

(g) y'(0.3) = 1.4495755,

y"(0.3) = 2.0495755,
y"'(0.3) = 4.0495755,
y(4)(0.3) = 4.0495755.

Substituting (f) and (g) in the fourth line of (a), we obtain

(h) y(OA) = 1.3595755 + (0.1)(1.4495755) + (0.005)(2.0495755)
+ (4.0495755)

+ (4.0495755) = 1.5154727.

Comment 46.3. The particular solution of (a) of Example 46.2 for

which y(O) = 1 is

(i) y(x) = 3e'" - x2 - 2x - 2.

We can therefore compare the actual values of y with those obtained by

Lesson 46B SOLUTION BY "CREEPING UP" PROCESS 649

the direct substitution method and by the creeping up process, see Table
46.31. An examination of the table discloses that the direct substitution

Table 46.31

Direct Substitution Creeping Up Process

Actual Value
See (f) of See (d), (f), (h) of
Obtained from (i)
Example 46.2 Example 46.21

y(O.I) 1.1055125 1.1055125 1.1055128

y(0.2) 1.2242000 1.2242077 1.2242083
y(0.3) 1.3595125 1.3595755 1.3595764
y(OA) 1.5152000 1.5154727 1.5154741

method has given six decimal accuracy for y(O.I), four for y(0.2), and
three for y(0.3) and y(0.4), after being rounded off to these respective
number of places. On the other hand, the creeping up method has given
six decimal accuracy for y(O.I), y(0.2), y(0.3), and five decimal accuracy
for y(O.4).
Comment 46.4. Comment on Error in Taylor Series Method.
1. By Theorem 37.34, the remainder or error term of a Taylor series is

_ y(n+ll(x) n+l
(46.41) E - (n + I)! h ,

where X is a value of x in the interval under consideration of width h.

In Example 46.2, we stopped with y(4l(x). Therefore by (46.41), the limit
of error in using h = 0.4 is, with n = 4,

(46.42) lEI M = 0.OOOO85M,

where M is the maximum value of ly(Sl(x)1 in the interval (0,0.4) of width

0.4. In the creeping up process, the limit of error in each calculation is,
by (46.41), with h = 0.1,
(46.43) lEI (05?5 M.

The total upper limit of error for four steps due to formula only, i.e., the
limit of error in an interval of width 0.4 by the creeping up process, ex-
cluding cumulative errors, is therefore,

(46.44) E 4 M = O.OOOOOO333M.
650 NUMERICAL METHODS Chapter 10

A comparison of (46.44) with (46.42) shows approximately how much

more accurate the creeping up method is.
2. Let
(46.5) E(xo + h) = the error in computing y(xo + h) by use of formula
(46.12) including terms to h4.
Then by (46.41), with n = 4,
(46.51)
where c = y(Sl(X)/5! Let us divide the h interval in half and compute
y(xo + h) in two steps. If we have some basis for believing that y(Sl(X)
changes rather slowly in this small h interval, so that the variation in the
value of y(Sl(X) in each h/2 interval is negligible, then we commit a small
error by using the same c of (46.51) for each half interval. Therefore, by
(46.51), the error in computing y(xo +
h/2) for a half interval h/2 is
approximately

(46.52) ( + 2h) = (h)S

E Xo 2 = 32chS= 32E(xo + h).
c
1 1

This means that the error in y (xo + is approximately equal to one

thirty-second the error in y(xo + h). Hence the error in computing in
two steps the value of y(xo +h), which we write as y [(xo I

will have an inherited error in the starting value of y (xo + plus its
own formula error. Since each error equals one thirty-second the error
+
in y(xo + h), the total error in y(xo h) computed in two steps, i.e., the
error in y [ (x o + + , will be approximately equal to one-sixteenth
the error in y(xo + h) computed in one step. Hence

(46.53) E [(xo + + = 116E(xo + h),

approximately. Let Y(xo + h) be the actual value of the solution. Then
(46.54) Y(xo + h) - y(xo + h) = E(xo + h),

Y(xo +h) - y[(xo = E[(XO +

Subtracting the second equation in (46.54) from the first, we obtain
(46.55)

y [ (xo + + - y(xo + h) = E(xo + h) - E[ (xo + + .

LeallOn 46B SOLUTION By "CREEPING UP" PROCESS 651

Substituting (46.53) in the right side of (46.55), the following equations

result.

(46.56) y [(XO + + - y(xo + h)

and
= 16E[(XO + - E[(XO +
y[(XO + - y(xo + h) = E(xo + h) - isE(xo + h).
Simplification of (46.56) gives

(46.57) E[(XO + = 115 {Y[(XO - y(xo + h)},

and

E(xo + h) = {Y[(XO -r - y(xo + h)}'

The first formula says that the error in the value of y(xo + h) computed
in two steps is equal to one-fifteenth the difference in values of y(xo + h)
computed in two steps and in one step.
Formulas (46.57) give us a means of. estimating errors at each step.
Their accuracy does not depend on a knowledge of the size of y(5) (X)
which we do not know, but only on the variation of y(S)(X)/5! over a
small interval. We have assumed this variation to be negligible, an
assumption which is not unreasonable if y(lS)(x) changes slowly over h.
For example, from the column headed "creeping up" process in Table
46.31, we have with Xo = 0, h = 0.2,

(a) y [(xo + = y(O.1 + 0.1) = y(0.2) = 1.2242077,

and from the direct substitution column,

(b) y(xo + h) = y(0.2) = 1.2242000.

Hence by the first equation in (46.57),

(c) E[(XO + = E(O.1 + 0.1) = 1\(1.2242077 - 1.2242(0)

= 0.0000005,
which is the approximate error of y(0.2) computed in two steps. The
actual error is 1.2242083 - 1.2242077 = 0.0000006.
3. A check on errors which practical people frequently use, and which
seems to work, is to make all calculations over again with an h half the
652 NUMERICAL METHODS Chapter 10

size of the original one. If the results obtained with the smaller h agree
with those obtained with the larger h to k decimal places, after being
properly rounded off, then it is assumed that the common numerical
value has k decimal place accuracy. This method is, of course, meaningful
only if the creeping up process is used.
4. This method does not contain a check on arithmetical errors. How-
ever, if there is little agreement between the values obtained by using h
and h/2, all arithmetical computations should be checked. If arithmetical
computations are correct, h should be reduced.
Comment 46.58. We have headed this lesson "Starting Formula-
Taylor Series." It is evident that it may also be used as a continuing
method to find y(0.5), y(0.6), etc. There is, however, a practical objec-
tion to its continued use. To insure a desired degree of accuracy, more
and more terms of the series will be needed as the distance from the
initial point Xo increases. This means we must evaluate higher and higher
order derivatives. In many practical cases, these derivatives, as remarked
earlier, may become extremely complicated or may be even so right from
the start. Hence, we must seek other starting and continuing methods
which do not depend on our ability to obtain derivatives. If, however, it
is not difficult to obtain many derivatives, then a Taylor series is not only
a good starting method but is also a good continuing method. It can be
used as long as the x values remain within the interval of convergence and
as long as you are reasonably sure that you have enough terms in the
series ta give you values that are accurate to the desired number of decimal
places. A high-speed calculating machine can, for example, add with ease
100 terms of a series.

EXERCISE 46
Apply the methods of this lesson to solve the problems which follow.
Use Taylor series to terms of the fourth order.
1. Use the direct substitution method to find approximate values when x = 0.1,
0.2, 0.3, 0.4 of a particular solution of the differential equation y' = x + y
for which yeO) = 1. Take h = 0.1. Solve the equation and compare your
results with actual values.
2. (a) Solve problem 1 by using the creeping up method. In the absence of a
solution in terms of elementary functions or of an error formula, how
could you determine the accuracy of y(O.4)?
(b) Why could not 'one use error formula (46.41) to determine the upper
bound of the error in y(O.4), just as we did in Exercises 44,11 and 45,10
to find an upper bound of the error in the numerical value of y(2)?
Ans. Here y' is a function of x and y; so also is y(n+l)(x). For exam-
+ + +
ple, y(4) = ylll = y" = 1 y' = 1 x y. Hence one cannot determine
the maximum value of y(n+l)(x) in an interval of width h without
knowing y(x). But y(x) is the solution we seek and do not know. In the
previous problems y' was a function only of x.
Lesson 47 STARTING METHOo-RUNGE-KUTrA FORMULAS 653

3. Start again with the equation y' = x + y for which yeO) = 1. Using the
values of y(0.2) found in one step in 1 and in two steps in 2, apply formula
(46.57) to correct y(0.2). (NOTE. Formula (46.57) is based on using terms in
a Taylor series to order four.) Starting with the corrected figure of y(0.2),
compute y(O.4) in one step and in two steps. Correct y(O.4) by means of
(46.57). Compare results with those obtained in 1 and 2 and with actual
values of the solution.
4. Error formulas (46.57) are based on stopping with the fourth order term in a
Taylor series. Find comparable error formulas if one stopped with a third
order term, with a fifth order term.
5. Using the direct substitution method, find approximate values when x = 0.1,
0.2, 0.3 of a particular solution of the equation y' = -xv for which yeO) = 1.
6. Solve problem 5, using the creeping up method. How could you determine
the accuracy of y(0.3)? Hint. See answer to problem 2(a) above.
7. Can the direct substitution method be used in 5 to find approximate values
of y(O.4), y(0.5), y(0.6), y(0.7), y(0.8), ••• ?

ANSWERS 46
1. 1.1103417, 1.2428000, 1.3996750, 1.5834667. Actual values: 1.1103418,
1.2428055, 1.3997176, 1.5836494.
2. (a) 1.1103417, 1.2428052, 1.3997171, 1.5836486. Compute y(O.4) by the
creeping up method with h = 0.05. Then make use of Comment 46.4-3.
3. Corrected values: y(0.2) = 1.2428055, y(O.4) = 1.5836493.
4. Third order term:

E[(xo + + - = Hy [(xo+ + - y(xo + h)} ,

E(xo +h) = Hy [(xo - y(xo +h)}'
Fifth order term:

= ;1 - y(xo+ h)} ,
E(xo + h) = {y[(x o - y(xo + h)}'

5. 0.99501, 0.98020, 0.95601.

6. 0.99501, 0.98019, 0.95598.
7. Read Theorem 38.14. The maximum interval of convergence given by the
theorem is Ixl < !. Hence, in the absence of additional information, the
series found in 5 cannot be used for Ixl > !.

LESSON 47. Starting Method-Runge-Kutta Formulas.

The starting method we shall develop in this lesson for finding a nu-
merical solution of
(47.1) y' = f(x,y)
654 NUMERICAL METHODS Chapter 10

satisfying the initial condition

(47.11) y(xo) = Yo,

was developed by the two people after whom it is named, Runge (1856-
1927) and Kutta (1867-1944). It has an advantage over the Taylor series
method in that it does not use derivatives. As mentioned in the previous
lesson, if the function f(x,y) is very complicated, it may be extremely
difficult to obtain its second derivative, let alone its third and fourth.
Because of this fact and its high degree of accuracy, it is in practice prob-
ably the most widely used starting method.
By (46.12),
"( ) ",( ) 3
(47.12) y(xo +
h) = y(xo) y'(xo)h+ Y h2 +
Y h +
(4)( )
+y h4 + ....
Differentiating (47.1), we obtain

(47.13) y"(x)
a f(x,y) + aya f(x,y)y'(x).
= ax
Substituting (47.1) and (47.13) in (47.12), and stopping with the h 2 term,
we have
(47.14) y(xo + h) = y(xo) + f(xo,yo)h
+ [af(xo,Yo)
ax
+ ay 0, 0 2
2
af(xo,Yo) f(x y)] h •

Our aim will be to rewrite the right side of (47.14) so that it will have
the form
(47.15) y(xo + h)
= y(xo) + Ahf(xo,yo) + Bhf[xo + Ch, Yo + Dhf(xo,yo»),
and then find values of A, B, C, D so that the right side of (47.15) will
actually equal the right side of (47.14).
By (38.12) and (38.13), the Taylor series expansion of a function of two
variables about a point (xo,yo), is, with x = Xo + a, y = Yo + b,
(47.16)
f(xo + a, Yo + b) = f(xo,yo) +a +b + ....
In (47.16) let a = Ch and b = Dhf(xo,yo). There results

(47.17) f[xo + Ch, yo + Dhf(xo,Yo») = f(xo,Yo)

+ Ch af(xo,Yo)
ax
+ Dh/(x 0, y)0 af(xo,Yo)
'J ay
+ ....
Lesson 47 STARTING METHOo-RUNGE-KUTTA FORMULAS 655

Substituting (47.17) in (47.15) and simplifying the result, we obtain

(47.18) y(xo + h) = Y(XO) + (A + B)hf(xo,yo)
+ Bh2 [ C + Df(xo,Yo) .

Comparing (47.18) with (47.14), we see that both right sides will be alike
if, for example, *
(47.19) A +B = 1, B =!, C = 1, D = 1,
from which we obtain
(47.2) A =!, B =!, C = 1, D=1.
Substituting these values in (47.15), we have
(47.21) y(xo + h) = y(xo) + !hf(xo,yo) + !hf[xo + h, Yo + hf(xo,yo)].
Formula (47.21) can be written in the simpler form
(47.3)
where
(47.31) UI = hf(xo,yo) ,
U2 = hf(xo + h, Yo + UI)'
If the first four terms of the series (47.12) are used, then the Runge-
Kutta formula becomes
(47.32) y(xo + h) = y(xo) + i(VI + 4V2 + va),
where
(47.33) VI = hf(xo,yo) = Ur,
V2 = hf(xo + !h, Yo + !VI),
Va = hf(xo + h, Yo + 2V2 - VI).

And if the first five terms of the series (47.12) are used, then the
Runge-Kutta formula becomes
(47.34)
where
(47.35) WI = hf(xo,yo) = Vb

W2 = + !h, Yo + !WI) =
hf(xo t'2,
Wa = hf(xo + !h, Yo + !W2),
W4 = hf(xo + h, Yo + wa)·

"There are other possible choices of A, B, C, D, but thess are the simplest.
656 NUMERICAL METHODS Chapter 10

Since (47.3), (47.32), and (47.34) were obtained by including the second,
third, and fourth order terms respectively of a Taylor series, these for-
mulas have been called respectively the second, third, and fourth
order forms.
As in Taylor series, two methods are available for finding y(xo + 0.1),
y(xo + 0.2), y(xo + 0.3) by the Runge-Kutta method, one by direct sub-
stitution in formula (47.3) or (47.32) or (47.34), the other by creeping up
to y(xo + 0.2) and then to y(xo + 0.3) by keeping h = 0.1 fixed and using
anyone of the formulas in succession.
Example 47.36. Use the fourth order form and the creeping up proc-
ess to find an approximate value when x = 0.2 of the particular solution of
(a) y' = x2 + y,
for which yeO) = 1.
Solution. Comparing the initial condition with (47.11), we see that
Xo = 0, Yo = y(xo) = 1. Hence, by (47.34), with Xo = 0, h = 0.1,
(b)

By (47.35), with Xo = 0, Yo = y(xo) = 1, h = 0.1 and/(x,y) = x 2 + y,

we have

WI = O.lj(O,I) = 0.1(1) = 0.1,

W2 = 0.1/(0.05, 1 + 0.05) = 0.1[(0.05)2 + 1.05] = 0.10525,
Wa = 0.1/(0.05, 1 + 0.052625) = d.l(0.05 2 + 1.052625)
= 0.1055125,
W4 = 0.1/(0.1, 1 + 0.1055125) = 0.1(0.1 2 + 1.1055125)
= 0.11155125.

Hence, by the above values, the initial condition and (b), an approximate
value of y(O.I) is

(c) y(O.I) = 1 + 1(0.1 + 0.2105 + 0.211025 + 0.11155125) = 1.1055127.

With Xo = 0.1 and h = 0.1, (47.34) now becomes
(d) y(0.2) = y(O.I) + l(wl + 2W2 + 2wa + W4).
By (47.35), with Xo = 0.1, Yo == y(xo) = y(O.I) = 1.1055127 and/(x,y) =
x 2 + y,
WI = O.lj(O.I, 1.1055127) = 0.1(0.1 2 + 1.1055127) = 0.11155127,
W2 = 0.1/(0.15, 1.1055127 + 0.0557756)
= 0.1(0.15 2 + 1.1612883) = 0.1183788,
Lesson 47 STARTING METHOD-RUNGE-KUTI'A FORMULAS 657

Wa = 0.lj(0.15, 1.1055127 + 0.0591894)

= 0.1(0.15 2 + 1.1647021) = 0.1187202,
W4 = 0.11(0.2, 1.1055127 + 0.1187202) = 0.1264233.
Hence by (c) and (d), an approximation of y(0.2) is
(e) y(0.2)
= 1.1055127 +*(0.1115513 +0.2367576 + 0.2374404 + 0.1264233)
= 1.1055127 +0.1186954 = 1.2242081.
Comment 47.37. If we had found y(0.2) in one step, formula (47.34)
would have become, with h = 0.2, Xo = 0, y(xo) = Yo = 1,
(f)
where
(g) WI = 0.21(0,1) = 0.2(1) = 0.2,
W2 = 0.21(0.1, 1 + 0.1) = 0.2[(0.1)2 + 1.1] = 0.222,
Wa = 0.21(0.1,1 + 0.111) = 0.2[(0.1)2 + 1.111] = 0.2242,
W4 = 0.21(0.2,1 + 0.2242) = 0.2[(0.2)2 + 1.2242] = 0.25284.

Hence by (f) and (g)

(h) y(0.2) = 1 + *(0.2 + 0.444 + 0.4484 + 0.25284)

= 1 + *(1.34524) = 1.2242067.

Comment 47.4. Comment on Error in Runge-Kutta Method.

1. It has been proved that the error for an interval of width h, in using
the Runge-Kutta fourth order form, is

(47.41) E(xo + h) = Ch 5 ,

where C is a constant. Although the constant C in (47.41) is not the same

as the c in (46.51), if we start with (47.41) and follow the steps after
(46.51), we arrive at the same conclusion (46.57). Hence by (46.57),

(47.42) E[(XO = 115 {y[(xo - y(xo + h)},

E(xo + h) = {y[(xo + - y(xo + h)}'

In Example 47.36, we found, with Xo = 0, h = 0.2, [see (e)]

(a) y [(xo + + = y[O.l + 0.1) = y(0.2) = 1.2242081,

658 NUMERICAL METHODS Chapter 10

and, in Comment 47.37,

(b) y(xo + h) = y(0.2) = 1.2242067.
Therefore by (47.42), with Xo = 0, h = 0.2,

(c) E[ (XO = E(O.l + 0.1) = n{1.2242081 - 1.2242067)

= 0.0000001,
which is the approximate error in y(0.2) obtained in two steps. The actual
error in y(0.2) obtained in two steps is 0.0000002, see Table 46.31.
2. All other remarks on errors in Comment 46.4 carryover without
change to the Runge-Kutta method.
Comment 47.5. This method could also be used as a continuing one.
The objection to its use for this purpose lies in the large number of com-
putations which must be made at each step. We shall later give a con-
tinuing method which is less time consuming and just as accurate.

EXERCISE 47

In solving the following problems, use the fourth order Runge-Kutta

method.
1. Prove error formulas (47.42).
2. Using direct substitution, find approximate values when x = 0.1 and 0.2
of a particular solution of the equation y' = x +
y for which yeO) = 1.
Take h = 0.1. Compare results with actual values.
3. Solve problem 2 by using the creeping up method. In the absence of a solu-
tion in terms of elementary functions or of an error formula, how many
decimal place accuracy could you assume in the value of y(0.2)? Compare
with actual value.
4. Using the values of y(0.2) found in one step in 2 and in two steps in 3, apply
formula (47.42) to approximate the error in y(O.l +
0.1). Compare with
actual error.
5. Using direct substitution, find approximate values when x = 0.1 and 0.2
+
of a particular solution of the equation y' = x 2 y2 for which yeO) = 1.
6. Solve problem 5, using the creeping up method.
7. Using the values of y(0.2) found in 5 and 6, determine, by means of formula
(47.42), the approximate error in y(O.l + 0.1).

ANSWERS 47
2. y(O.I) = 1.1103417, y(0.2) = 1.2428000. Actual values: 1.1103418,
1.2428055.
3. y(0.2) = 1.2428052. Can assume five decimal accuracy, see Comment 46.4-3.
4. E(O.1 + 0.1) = 0.0000003. Actual error = 0.0000003.
5. y(O.I) = 1.1114629, y(0.2) = 1.2529908.
6. y(0.15) = 1.1776978, y(0.2) = 1.2530162.
7. E(O.I + 0.1) = 0.0000017.
Lesson 48A FINITE DIFFERENCES 659

LESSON 48. Finite Differences. Interpolation.

Before we can develop a continuing method which is just as accurate
as Runge-Kutta but less time consuming, we shall need additional infor-
mation. We momentarily digress, therefore, to develop this needed
information.

LESSON 48A. Finite Differences.

Definition 48.1. The first difference of a function f(x), written as
!J.f(x) (read "delta f of x") is defined as
(48.11) !J.f(x) = f(x + h) - f(x),
where h is a fixed constant. It is the difference in values of the function
for two neighboring values of x, h units apart.
The second difference !J.2f(x) is defined as the difference of the first dif-
ference of f(x) for two neighboring values of x, h units apart. By definition,
therefore,
(48.12) !J.2f(x) = !J.[!J.f(x)] = !J.[f(x + h) - f(x)] = !J.f(x + h) - !J.f(x).
Similarly the third difference !J.3f(x) is defined as the difference of the
second difference of f(x) for two neighboring values of x, h units apart.
Therefore
(48.13)
And in general, we define
(48.14)
With the aid of the above definitions, we shall now explain how to
construct a table of differences !J.f(x), !J.2f(x), .•. , !J.nf(x).
Example 48.15. If
(a) f(x) = eX

and h = 0.1, construct a table of differences of eX for values of x from

o to 0.5.
Solution. (See Table 48.16). Since h = 0.1, we shall need values of
f(x) = eX that, beginning with x = 0, are 0.1 unit apart. In the first
column of our table, we therefore write x = 0, 0.1, 0.2, 0.3, 0.4, 0.5. In
the second column we write the values of f(x) = eX when x = 0, 0.1,
... , 0.5. * By (48.11) !J.f(x) is the difference in the values of f(x) for two
·Various texts are available that contain tables of values of i', e-z , log:l:, sin:l:,
COS:l:, etc. Excellent tables of values of these and other functions, correct to twelve
and more decimal places, have been published under the sponsorship of the National
Bureau of Standards, Mathematical tables project.
660 NUMERICAL METHODS Chapter 10

Table 48.16

f(x) = Ilf(x) = 1l2f(x) = 1l3f(x) = 1l4f(x) = 1l5f(x) =

x = eX = Ile'" = 1l 2e'" = 1l3e", = 1l4e", = 1l5e", =

0 1.00000 0.10517 0.01106 0.00117 0.00010 0.00007

0.1 1.10517 0.11623 0.01223 0.00127 0.00017
0.2 1.22140 0.12846 0.01350 0.00144
0.3 1.34986 0.14196 0.01494
0.4 1.49182 0.15690
0.5 1.64872

neighboring values of x, 0.1 unit apart. Hence we obtain the third column
of the table by taking the differences of the second column. For example,
by (48.11), lleO. 2 = eO. 3 - eO. 2 = 1.34986 - 1.22140 = 0.12846. Anal-
ogously 1l 2f(x), by (48.12), is the difference of Ilf(x) for two neighboring
values of x, 0.1 unit apart. Therefore we obtain the fourth column by
taking differences of the third column. For example, by (48.12), 1l 2eO. 2 =
1l(lleo. 2) = lleO. 3 - lleO. 2 = 0.14196 - 0.12846 = 0.01350. The elements
of the fifth column 1l 3ez are the differences of the fourth column, etc.
Comment 48.17. If you want the fifth difference of f(x), i.e., 1l 5f(x),
when x = a, you must know f(a),J(a + h), ... ,f(a + 5h). And in general
if you want the mth difference of f(x) when x = a, you must know f(a) ,
f(a + h), ... ,f(a + mh). In the above example, a = 0, and we therefore
had to know f(O), f(O.l), ... , f(0.5) in order to evaluate the fifth differ-
ence of eZ , i.e., 1l5ez, when x = O.
Comment 48.18. Call Yo, Yl, ... , Yn the values of f(x) corresponding
to f(xo) , f(xo + h), ... , f(xo + nh). By Definition 48.11,

(48.181) Ilyo = Yl - Yo·

By (48.12), (48.181) and Definition 48.11,

(48.182) 1l 2yo = 1l[IlYol = Il(Yl - Yo) = IlYl - Ilyo

= Y2 - Yl - Yl +Yo = Y2 - 2Yl +Yo·

By (48.13), (48.181) and (48.182),

(48.183) 1l 3yo = +
1l[Il 2Yol = Il(Y2 - 2Yl Yo) = IlY2 - 21lYl Ilyo +
= Y3 - Y2 - 2Y2 + 2Yl + Yl - Yo = Y3 - 3Y2 + 3Yl - Yo·
A comparison of the coefficients on the right sides of (48.182) and (48.183)
with the respective coefficients in the expansion of (x - 1)2 and (x - 1)3
Lesson 48B POLYNOMIAL INTERPOLATION 661

shows that they are alike. This leads one to suspect that the fourth dif-
ference of Yo is
(48.184)

You can verify that this suspicion is indeed correct. In fact it can be
proved by induction, using an argument similar to the one given in Lesson
49A, that the coefficients in the expansion of t:"nyo are the same as the
coefficients in the expansion of (x - 1)n. Therefore

n (-I)kn! *
(48.19) t:" Yo = 6 k!(n - k)! Yn-k,

where it is understood that t:,,°yo = Yo. Hence, in terms of I(x), (48.19)

becomes
.. (-I)n!
(48.19n t:" f(xo) = 6 k!(n _ k)! I[xo + (n - k)hJ.

LESSON 48B. Polynomial Interpolation. Let us assume that the

only information we have about a function is its values when x = xo,
Xo + h, Xo + 2h, ... , Xo + mh, and that we wish to find a value of the
+
function when Xo = Xo dh, d ¢ 0, 1, 2, .... This situation occurs,
for example, when we need the value of sin 0.5245 and the table we con-
sult gives values of sin 0.5236, sin 0.5265, sin 0.5294, etc., i.e., it gives
values over intervals of width h = 0.0029. The usual procedure in this
case is to add to the value of sin 0.5236, n
of the difference in values be-
tween sin 0.5236 and sin 0.5265. When we do this, we are implicitly
assuming that the graph of the sine function between these two points is
approximated by a straight line. We are thus in effect using a linear
interpolation to approximate the sine function between two points of its
graph.
A better method of interpolating the value of sin 0.5245 from the
tabular values is to approximate the sine function by a polynomial/(x)
whose graph coincides with the sine function at more than the two points
sin 0.5236 and sin 0.5265. If it coincides with three values of the sine
function, we get a second degree polynomial to approximate sin x; if it
coincides with m + 1 values, we get a polynomial of degree m to approxi-
mate sin x. This approximating polynomial I(x) will then give, when
x = 0.5245, a more accurate value of sin 0.5245 than will the linear inter-
polation method described above. Two questions now arise. Given a set
of values of a function I(x) at xo, XII ••• ,xm • Does a polynomial exist
that can take on these values at the given points, and if there is such a

*The fraction nl/[kl(n - k)IJ is also frequently written as

662 NUMERICAL METHODS Chapter 10

polynomial, is it unique? The answer to both questions is given by the

following existence and uniqueness theorem which we state without proof.
Theorem 48.2. Let f(xo) , f(Xl), ... , f(xm), be m +
1 distinct values of
a function f(x). Then there is one and only one polynomial F(x) of degree
less than or equal to m which coincides with these m +
1 values of f(x).
Definition 48.21. The unique polynomial F(x) of degree less than or
equal to m, whose values coincide with m + 1 distinct points of a func-
tion f(x) , is called a polynomial interpolating function ofJ(x).
Comment 48.22. We shall be interested only in a polynomial inter-
polating function, although functions other than polynomials may be used
for interpolation purposes.
Comment 48.23. Very few of you have used nonlinear interpolation
to approximate the value of a function f(x) that is not given in tables,
from those which are. To use this method, it is necessary to find an inter-
polating polynomial F(x) which agrees with f(x) at three or more points
and is of degree 2. It would be troublesome indeed, if each time we
wished to use polynomial interpolation, we had to find the approximating
function F(x). We are indebted to Newton (1642-1727) for formulas by
which we can readily make polynomial interpolations whenever the
abscissa points, for which the functional values are known, are evenly
spaced. These formulas are derived in the next lesson.

EXERCISE 48
1. Iff(x) = log x and h = 0.1, construct a table of differences of log x for values
of x from 1 to 1.5.
2. If f(8) = sin (8) and h = 5°, construct a table of differences of sin 8 for
values of 8 from 10° to 35°.

ANSWERS 48
1.

f(x) = t:.f(x) = t:. 2f(x) = t:. 3f(x) = t:. 4f(x) = t:. 5f(x) =
x =
logx = t:.log x = t:. 2 logx = t:. 3 logx = t:.4logx = t:. 5 log x =

1.1 0.09531 0.08701 -0.00697 0.00104 -0.00022 0.00004

1.2 0.18232 0.08004 -0.00593 0.00082 -0.00018
1.3 0.26236 0.07411 -0.00511 0.00064
1.4 0.33647 0.06900 -0.00447
1.5 0.40547 0.06453
1.6 0.47000
Leason 49A NEWTON'S (FORWARD) INTERPOLATION FORMULA 663

2.
f:.2f(l}) = f:.3f(l}) = f:.4f(l}) = f:.5f(l}) =
I} = f(l}) = f:.f(l}) =
sin I} = f:.sin I} = f:.l!sin I} = f:.3 sin I} = f:.4 sin I} = f:.5 sin I} =

100 0.17365 0.08517 -0.00197 -0.00063 0.00001 0.00003

150 0.25882 0.08320 -0.00260 -0.00062 0.00004
200 0.34202 0.08060 -0.00322 -0.00058
25 0 0.42262 0.07738 -0.00380
300 0.50000 0.07358
350 0.57358

LESSON 49. Newton's Interpolation Formulas.

LESSON 49A. Newton's (Forward) Interpolation Formula. We

assume thatf(x) is a function which is defined on an interval (xo, Xo + mh),
and that its values have been given only for the m + 1 distinct abscissa
points: xo, Xo + h, Xo + 2h, ... ,Xo + mh, as happens, for example, when
we consult a table of values of the function or when these isolated, distinct
values have been found experimentally, Fig. 49.1. Our object will be to
{(xo +2h)
f(xo + h) {(Xo +mh)

xo+h xo+2h xo+mh

Figure 49.1

find a polynomial F(x), of degree that agrees with f(x) at these

m + 1 points. By Theorem 48.2, we know that there is such a polynomial
F(x) and that it is unique. By Definition 48.21, this unique polynomial
F(x) is an interpolating function of f(x), by which we mean that it will
give an approximate value of the function f(x) for an x which does not
coincide with the points xo, Xo + h, •.. , Xo + mho To obtain a formula
for F(x), we proceed as follows.
By Definition 48.1
!:if(xo) = f(xo + h) - f(xo).
Solving for f(xo + h), we have
(49.11) f(xo + h) = f(xo) + f:.f(xo).
Applying f:. to both sides of (49.11) gives
(a) Ilf(xo + h) = f:.f(xo) + f:.2f(xo).
664 NUMERICAL METHODS Chapter 10

Again by Definition 48.1, we obtain

(b) !If(xo + h) = f(xo + 2h) - f(xo + h).
Since the left sides of (a) and (b) are the same, we can equate their right
sides. Hence
(c) f(xo + 2h) = f(xo + h) + !If(xo) + !l2f(xo).
By replacing in (c) the value of f(xo + h) as given in (49.11), we have
finally
(49.12)
Applying!l to both sides of (49.12) gives
(d) !If(xo + 2h) = !If(xo) + 2!l2f(xo) + !l3f(xo).
By Definition 48.1,
(e) N(xo + 2h) = f(xo + 3h) - f(xo + 2h).
Hence equating the right sides of (d) and (e), we obtain
(f) f(xo+ 3h) = f(xo + 2h) + !If(xo) + 2!l2f(xo) + !l3f(xo).
Replacing in (f) the value of f(xo + 2h) as given in (49.12), we have
finally
(49.13) f(xo + 3h) = f(xo) + 3!lf(xo) + 3!l2f(xo) + !l3f(xo).
If you will compare the constant coefficients in the final expressions for
f(xo +h), f(xo +
2h), f(xo +
3h) as given in (49.11), (49.12), and (49.13),
with the coefficients in the respective expansions of (x 1), (x 1)2, + +
(x +
1)3, you will discover that they are the same. This observation
leads one to suspect that the coefficients in the expansion of f(xo + nh)
will be the same as those in the expansion of (x +
l)n. We shall now
prove by induction that this suspicion is indeed correct.
Proof. We assume that the coefficients in the expansion of
f[xo +(k - l)h] are the same as the coefficients in the expansion of
(x +
l)k-l. That is we assume

(a) f[xo + (k - l)h] = f(xo) + (k - 1) !If(xo)

+ (k - - 2) !l2f(xo)

+ (k - 1) (k -;,2)(k - 3) !l3f(xo) + ... + !lk-1f(xo)

is a valid equation. We must then prove, by Comment 24.1, that the
coefficients in the expansion of f(xo + kh) are the same as the coefficients
Lesson 49A NEWTON'S (FORWARD) INTERPOLATION FORMULA 665

in the expansion of (x + 1)". Applying to both sides of (a), we obtain

(b) + (k - 1)h] = d!(xo) + (k - 1)

+ (k - - 2) + ... +
Adding the left sides of (a) and (b) there results, by (48.11),
(c) f[xo + (k - 1)h] + + (k
- l)hJ
= f[xo + (k - l)hJ f(xo + + kh) - f[xo + (k - l)h]
= f(xo + kh).
The addition of the right sides of (a) and (b) gives

(d) f(xo) +k + (k - l)(k - :r) + 2(k - 1) 2f(xo)

+ (k - l)(k - 2)(k - + 3(k - l)(k - 2) + ...

(k - 1)(k - 2) ... (k - m) + m(k - l)(k - 2) ... (k - m + 1)
+ m!
X + ... +
which simplifies to
(e) f(xo) +k + k(k 1) + k(k - - 2)

+ ... + k(k - l)(k - 2) ., .. (k - m + 1) + ... +

m.
Since (c) is the sum of the left sides of (a) and (b), and (e) is the sum of
their right sides, we can equate these two to obtain

(f) f(xo + kh) = f(xo) + k + k(k 2!- 1) 2

.:l 'f(xo)

+ k(k - - 2) .:l3f(xo) + ...

+ k(k - 1)(k - 2) ., .. (k - m + 1) .:lmf(xo) + ... + .:lkf(xo).

m.
The coefficients in the expression on the right side are the same as those
in the expansion of (x + I? Hence our suspicion is proved.
We therefore can now write, by (f),
(49.14) f(xo + nh) = f(xo) + n .:lJ(xo) + n(n 2!- 1) 2
.:l 'f(xo)

+ n(n - - 2) .:l3f(xo) + ...

+ n(n - l)(n - 2) ... (n - m + 1) .:lm'!(xo),

m.,
n = 0,1,2,··· ,m.
666 NUMERICAL METHODS Chapter 10

[Note that when n = m, the last tenu simplifies to For each

n = 1, 2, ... , m, the right side of (49.14) gives a fonuula for computing
the respective ordinates f(xo) , f(xo + h), "', f(xo + nh) in tenus of
f(xo) and the differences of f(xo). *
Since the right side of (49.14) is meaningful even when n is not zero or
an integer, we can let
(49.15) F(x) = f(xo) + n + n(n 2!- 1) 2f(xo)
+ n(n - - 2) + ...
- m + 1)
+ n(n - l)(n - 2) ... (n
m.I
m
'!(xo) ,

where n is any number between 0 and m, and where

x - Xo
(49.16) x = Xo + nh, n = --h-'
Since the right side of (49.15) is the same as the right side of (49.14), we
know that
(49.161) F(xo + nh) = f(xo + nh), n = 0, 1,2, ... , m.

Substituting the second equality of (49.16) in (49.15), we obtain

(49.17) F(x) = f(xo) + -N(xo)

h - (x - xo)
2f(xo)
+ 2!h2 (x - xo)(x - Xo - h)

+ 3Ih3 (x - xo)(x - xo - h)(x - Xo - 2h) + ...

+ m!hm
(x - xo)(x - Xo - h)

X (x - xo - 2h) ... (x - Xo - [m - l)h).

By (49.17), we see that F(x) is a polynomial of degree

We have thus proved, by (49.161) and (49.17), that:
1. The function F(x) defined by (49.15) or (49.17) agrees with the func-
tion f(x) at the m +
1 points for which x = xo, xo h, Xo 2h, + +
"', Xo + mh,
2. F(x) is a polynomial of degree m.

Therefore, by Theorem 48.2 and Definition 48.21, F(x) is the unique

*Note the resemblance between this formula and the Taylor series formula (37.27)
by which one can compute a value of f(xo + nh) in terms of f(Xo) and the derivatives
of f(x) at x = Xo.
Lesson 49A NEWTON'S (FORWARD) INTERPOLATION FORMULA 667

polynomial interpolating function of the given function f(x). This means

that if we wish to find an approximate value of f(x), when

(49.18) x = Xo + nh, n = x - x o , n real,

h

we can use the function F(x) defined by either of the two formulas (49.15)
or (49.17). These formulas are known as Newton's (forward) inter-
polation formulas.

E%ample 49.19. Use Newton's (forward) interpolation formulas

(49.15) and (49.17) and the following table of values

(a) eO. l = 1.105 17,

eO.
2
= 1.221 40,
eO. 3 = 1.349 86,
eO
A
= 1.49182,
eO. 5 = 1.648 72,

l4
to find an approximate value of eO. •

Solution. By use of (49.15). Here f(x) = eX, Xo = 0.1 and h = 0.1.

Since we have been given five equally spaced values of the function eX, by
Comment 48.17, we shall be able to find only the fourth difference, namely
.1 4f(xo}, in formula (49.15) or (49.17). We shall therefore have to stop with
m = 4. And since we seek the value of eo. 14 , the x in these formulas is
0.14 - 0.1 .
0.14. Therefore by (49.18), n = = 0.4. Substituting the
0.1
above values in (49.15), it becomes
(b) F(0.14) = f(O.I) + (0.4) .1f(O.I) + (0.4)(02f - 1) .12f(0.1)
+ (0.4)(0.4 - 2) .13f(0.1)

+ (0.4)(0.4 - 1)(0.if - 2)(0.4 - 3) .14f(0.1).

The needed differences have already been calculated in Table 48.16 in the
row beginning with 0.1. Therefore (b) becomes

(c) F(0.14) = 1.105 17 + (0.11623)(0.4) - (0.01223)(0.12)

+ (0.001 27)(0.064) - (0.000 17)(0.0416)
= 1.15027,

which is the approximate value of eO 14.

668 NUMERICAL METHODS Chapter 10

Solution. By use of (.t,JJ.17). In this formula x - Xo = 0.14 -

0.1 = 0.04. As before h = 0.1, m = 4. Substituting these values in
(49.17), it becomes
0.04
(d) F(0.14) = f(O.l) + 0.1 .::l/(0.1) + .::l2f(0.1)
2!(0.1)2 (0.04)(0.04 - 0.1)

+ .::l3f(0.1)
3!(0.1)3 (0.04)(0.04 - 0.1)(0.04 - 0.2)

+ .::l4f(0.1)
4!(0.1)4 (0.04)(0.04 - 0.1)(0.04 - 0.2)(0.04 - 0.3),

which is equivalent to (b) above.

Comment 49.191. The actual value of eO. l4 to five decimals is
1.15027, the same as that obtained in (c) above by using a fourth degree
polynomial interpolating function. If you had used linear interpolation,
i.e., obtained the value of eO. l4 by adding to eO. l two-fifths of the dif-
ference in values between eO. l and eO. 2 , you would have obtained 1.15166
which is correct to only two decimal places.

LESSON 49B. Newton's (Backward) Interpolation Formula. New-

ton also gave a formula by which it is possible to interpolate backward
instead of forward. Let V, which is an upside down delta, called "del,"
be defined by
(49.2) Vf(x) = f(x) - f(x - h).

As before we assume we have been given values of I(x) only for those
points whose abscissas are Xo, Xo - h, Xo - 2h, ... , Xo - mh, Fig. 49.21.

f(xo - 2h)
f(xo -mh)

f[xo-(m-l)h]
f(xo -h)

Xo - mh Xo - (m -l)h Xo -2h Xo

Figure 49.21

Let F(x), of degree ;§; m, be a polynomial interpolating function of f(x).

By following the procedure outlined in Lesson 49A, you will obtain, in
place of (49.15),
Lesson 49B NEWTON'S (BACKWARD) INTERPOLATION FORMULA 669

(49.22) F(x) = f(xo) - nVf(xo) + n(n -2! 1) 2

V f(xo)

_ n(n - - 2) V 3f(xo) + ...

+ n(n - l)(n - 2) ., .. (n - m + 1) (-l) mVmf(xo),
m.
where
Xo - x x - Xo
(49.23) x = Xo - nh, n = --h- = - -h--' n real.

We have left the proof to you as an exercise.

If, in (49.22), we replace n by its value as given in (49.23), we obtain
the alternate form
Vf(xo) V 2f(xo)
(49.24) F(x) = f(xo) + - h - (x - xo) + 2!h2 (x - xo)(x - Xo + h)
3
V f(xo)
+ 31h3 (x - xo)(x - xo+ h) (x - Xo + 2h) + ...
Vmf(xo)
+ m!hm (x - xo)(x - Xo + h)
X (x - Xo + 2h) ... (x - Xo + [m - 11k).

Formulas (49.22) and (49.24) are known as Newton's (backward) inter-

polation formulas.

Example 49.25. Use the values of fiE given in (a) of Example 49.19,
and Newton's backward interpolation formula to find an approximate
value of eO. 46 •

Solution. As in the previous Example 49.19, f(x) = eX, h = 0.1 and

since we have been given five evenly spaced values of eX, the m in (49.22) is
0.5 - 0.46
four. Here, however, Xo = 0.5, x = 0.46 and by (49.23), n = =
0.1
0.4. Substituting these values in (49.22), we obtain

(a) F(0.46) = f(0.5) - 0.4Vf(0.5) + 0.4(;0.6) V2f(0.5)

_ 0.4( -0:1( -1.6) V3f(0.5) + (0.4)( 1.6)( -2.6) V4f(0.5).

The needed differences are given in Table 49.26. They are slightly dif-
ferent in this case from those in Table 48.16. Here Xo = 0.5. We must
therefore find differences of f(0.5) instead of f(O.I), and remember
VeO. 5 = eO. 5 - e°.4, Ve°.4 = eO. 4 - eO. 3 , etc. Substituting in (a) the
670 NUMERICAL METHODS Chapter 10

Table 49.26

f(x) = Vf(x) V2f(x) V3f(x) = V4f(x) =

X=
eZ
= Ve" = V2e" = V3e" = V4e" =

0.1 1.10517
0.2 1.22140 0.11623
0.3 1.34986 0.12846 0.01223
0.4 1.49182 0.14196 0.01350 0.00127
0.5 1.64872 0.15690 0.01494 0.00144 0.00017

figures shown in the row beginning with 0.5 of Table 49.26, we obtain

(b) F(0.46) = 1.64872 - 0.4(0.15690) - (0.12)(0.01494)

- (0.064)(0.00144) - (0.0416)(0.00017)
= 1.64872 - 0.06465
= 1.58407,
which is the approximate value of eO. 46 • Its actual value to five places is
1.58407 which agrees with (b) to five places. By a linear interpolation,
you would have obtained 1.58596.

LESSON 49C. The Error in Polynomial Interpolation. When we

use a polynomial F(x) of degree as an interpolating function for f(x) ,
we know only that both agree at those points whose abscissas are Xo,
Xo + h, "', Xo + mho For other values of x, the difference between
F(x) and f(x) may be very small or very large depending on what the
graph of f(x) looks like for these intermediate values.
The following error formula (49.31), which we state without proof, is
based on the assumption that f(x) has a continuous derivative of order
m + 1 in an interval I: a x b, which contains the points Xo, Xo + It-;
••• , Xo + mho Let

(49.3) E(x) = f(x) - F(x) ,

where E(x) is the error function or the remainder function. It is the

difference between the value of the function f(x) and the interpolating
function F(x). Then

(49.31) E(x) = (x - xo)(x - Xl) ..• (X - x",) f(m+l)(X)

(m +
I)! '
where xo, Xl = Xo + h, X2 = Xo + 2h,
'.', xm = Xo + mh are the
values of X for which f(x) and F(x) agree, and X is a number between the
Lesson 49-Exercise 671

smallest and largest values of xo, Xl! ••• , Xm and x. Unfortunately the
formula does not tell us how to choose X. However, if we can determine
the minimum and maximum values of If(m+l) (x) 1 in the given interval I,
then It<m+l) (X) 1 must be between them. Therefore by (49.31),

(x - xo)(x - Xl) ••• I

(X - xm) [min If(m+I)(x)1l < IE(x}l
(49.32)
I (m+l)! =
<
=
I(x - xo)(x - xd ... (x - xm )
(m +1)!
I[max If(m+l)(x)11 , xm .
. I

Formula (49.32) is also valid if Xl = Xo - h, X2 = Xo - 2h, "',

Xm = Xo - mho Hence (40.32) is valid for both Newton's forward and
backward interpolation formulas.

Example 49.33. Find the maximum error incurred in using Newton's

forward interpolation formula in Example 49.19.

Solution. In Example 49.19, h = 0.1, Xo = 0.1, x = O.14,!(x) = e",

+ +
m = 4. Hence Xl = Xo h = 0.2, X2 = Xo 2h = 0.3, Xa = Xo 3h = 0.3, +
x. = Xo +4h = 0.4. Since m = 4, we must, in (49.31), stop with x - x •.
Hence t<m+l)e X = f(5)e x = eX. In the interval (0.1,0.5), eX has the largest
value when X = 0.5. Therefore by (49.32),

(0.14 - 0.1)(0.14 - 0.2)(0.14 - 0.3)

X (0.14 - 0.4)(0.14 - 0.5) 0.5
(a) IE(x) 1 5! e
< 0.0000005.

EXERCISE 49
1. Use Newton's (forward) interpolation formulas (49.15) and (49.17) and the
following table of values, eO•a = 1.34986, eO .• = 1.49182, eO. 5 = 1.64872,
eO. 6 = 1.82212, eO. 7 = 2.01375, to find an approximate value of eO•a5 •
Compare with actual value and value obtained by a linear interpolation.
2. Use (49.15) and (49.17) and the following table of values, log 1.1 = 0.09531,
log 1.2 = 0.18232, log 1.3 = 0.26236, log 1.4 = 0.33647, log 1.5 = 0.40547,
log 1.6 = 0.47000, to approximate log 1.12. Compare with actual value and
value obtained by a linear interpolation. Hint. The needed table of differ-
ences is given in answer to Exercise 48,1.
3. The following temperatures were recorded at hourly intervals: 6 A.M., 2°;
7 A.M., 12°; 8 A.M., 17°; 9 A.M., 20°; 10 A.M., 22°. What was the approximate
temperature at 6 :30 A.M.?
4. Prove formula (49.22). Hint. Follow the procedure used to obtain (49.15).
5. Use Newton's (backward) interpolation formulas (49.22) and (49.24) and
the table of values given in 1 and 2 to approximate respectively (a) eO. 65 ,
(b) log 1.53. Compare with actual values and with those obtained by a
linear interpolation.
672 NUMERICAL METHODS Chapter 10

ANSWERS 49
1. eO•a5 = 1.41906. Actual value: 1.41907. Linear interpolation: 1.42084.
2. log 1.12 = 0.11333. Actual value: 0.11333. Linear interpolation 0.11271.
3. Approximately 7.9°.
5. (a) eO•65 = 1.91554. Actual value 1.91554. Linear interpolation 1.91794.
(b) log 1.53 = 0.42527. Actual value 0.42527. Linear interpolation: 0.42483.

LESSON 50. Approximation Formulas Including Simpson's Rule

and Weddle's Rule.
We are now ready to continue our study of the problem of finding a
numerical solution of
(50.1) y' = f(x,y)
satisfying the initial condition
(50.11) y(xo) = Yo.
Let y(x) be a solution of (50.1) fulfilling (50.11). Then, by (50.1),
(50.12) y' = f[x,y(x)],
where f is now a function of x alone. Integration of (50.12) and insertion
of the initial condition gives

(50.13) dy = i:xo f[x,y(x)] dx, y = Yo + Jxo

(" f[x,y(x)] dx.
If x = Xo + nh, (50.13) becomes, with Yo replaced by its equal y(xo) of
(50.11),
x o+nh
(50.14) y(xo + nh) = y(xo) + l
Xo
f[x,y(x)] dx.

Since y(x) is the unknown solution of (50.1), the function j[x,y(x)] is

also unknown. Hence we cannot possibly hope to perform the integration
in (50.14). However, we can use a polynomial interpolating function
F(x) for f[x,y(x)] just as we did in the previous lesson for f(x).
Let F(x) be such a polynomial interpolating function, agreeing with
j[x,y(x)] at the m + 1 points whose abscissas are Xo, Xo + h, Xo + 2h,
••• , Xo + mho Therefore, by (49.17), with f(xo) replaced by f[xo,y(xo)]
which, by (50.12), is equal to y'(xo),

(50.15) F( ) -
X -
[1 + x -
h
Xo ,1 + (x - xo)(x - Xo - h) ,12
2!h2
Xo - h)(x - Xo - 2h) a
+ (x - xo)(x -
3!ha ,1 + ...
+ (x - xo)(x - Xo - h)(x - Xo - 2h) ... (x - Xo - [m - l]h)
m!hm ,1
m] y,(xo),
Lesson 50 ApPROXIMATION FORMULAS. SIMPSON. WEDDLE. 673

where for convenience in writing we have placed y'(xo) outside the brackets.
Comment 50.16. If we stop with the term in (50.15), then m = 2
and F(x) becomes a polynomial of degree less than or equal to two agreeing
with f[x,y(x)] = y'(x) at the three points whose abscissas are Xo, Xo + h,
Xo + 2h, see 1 after (49.17). If we stop with the term, then m = 3
and F(x) becomes a polynomial of degree less than or equal to three that
agrees with f[x,y(x)] = y'(x) at the four points whose abscissas are
Xo, •.• , Xo + 3h, etc.
By (50.15), an approximation to the integral in (50.14) is, therefore,
"'0+"A
(50.2)
1 "'0
F(x) dx

= f",o+nA[1 +x - Xo + (x - xo)(x, -; Xo - h) + ...

1"'0 h 2.h

{
(x - xo)(x - Xo - h)(x - Xo - 2h) ... }
(x - Xo - [m - l]h)
1 d
+ m!hm
Am
y Xo
'( )
X.

To simplify integrating the right side of (50.2), we make the substitution

= x - Xo. Therefore du = dx, 1.£ = 0 when x = Xo and 1.£ = nh when
1.£
x = Xo + nh. Hence we can replace (50.2) by
"'0+nA
(50.21)
, "A
1"'0
F(x) dx

= f ...
10 h 2!h2 3!h 3
1.£(1.£ - h)(u - 2h) ... (1.£ - [m - l]h) m] ,
+ m!hm y (Xo) du

i + !!: + 1.£ 2 - + 1.£ 3 - + 2h 21.£

nA ( 2
= 1 hu 2 3hu 3
o h 2!h2 3!h3
1.£4 - +
6hu 3 llh 2u 2 -
3
6h u 4
+ 4!h4

1.£5 - lOhu 4 + 35h 2u 3 - 50h 3u 2 + 24h 4u 5

+ 5!h 5
2 4 3 3
6
u - 15hu
5
+
85h u - 225h u }
{ + 274h 4 u 2 - 120h 5u 6
+ 6!h 6

+ .. -) y'(XO) dUo
674 NUMERICAL METHODS- Chapter 10

Integration of (50.21) gives

"o+nh
(50.22)
l"0
F(x) dx
2 2
=
3 3 2 3
h +! n h A+ _1_ (n h _ n h ) A2
[n h 2 2h2 a 2

+ _1_ (n4h4 _ +
n3h4 n2h4) A3
6h3 4
5 5 45 3 5
_1_ (n h _ 3n h + lln h _ 3 2h S) A4
+ 24h4 5 2 3 n
6 6 4 6 3 6
+ _1_
120h5 6 n 4
+
(n h _ 2 5h6 35n h _ 50n h
3
12 2h6) A5
n
+
6 7 47
_1_ (n7h7 _ 5n h
+ 720h6
+
17 5h7 _ 225n h
7 2 n 4

+ 274;3 h7 _ 60n2h7) A6 + ... y'(xo). J

Simplification of (50.22) results in
"o+nh
(50.23)
l"

: nh[1
F(x) dx

+!!:A+! (n2 _!!:) A2 +! (n3 _ n2+n) A3

2 2 a 2 6 4

24 5 2
3
_3n +l1n2
3
-an)A 4

5 3 2
+ _1_ (n _ 2n 4+ 35n _ 50n + 12n) A5
120 6 4 3
5 3 2
+ _1_ (n6 _ 5n + 17n4 _ 225n + 274n _ 60n) A6
720 7 2 4 3

+ .. oJ Y'(xo)o
If n = 1 in (50.23), we obtain

(50.24) /,,:O+h F(x) dx = h[l +! A- nA2 + nA3 - NoA4

+ rto-A 5 - s8\0A
6
+ oJ y'(xo)o
0 0

Therefore, by (50.14) and (50.24), an approximation to y(xo + h) in

Lesson 50 ApPROXIMATION FORMULAS. SIMPSON. WEDDLE. 675

tenns of forward differences of y'(xo) is

(50.25) y(xo + h) = y(xo) + H,

where H is the expression on the right of (50.24).
If in (50.23), we stop with the ..:l tenn, then by Comment 50.16, the poly-
nomial interpolating function F(x) is of degree less than or equal to one,
agreeing with J[x,y(x)] at the two points for which x = Xo and Xo + h.
With n = 1 and using only tenns including .:l of (50.23), we obtain, with
the help of (48.11),
"'0+ 11
(50.3)
1 "'0
F(x) dx =
=
h[l + i.:l] y'(xo)
h{y'(xo) + ![Y'(xo + h) - y'(xo)]}.

= 2"h [y'(xo) + y'(xo + h)].

Therefore, by (50.14) and (50.3), an approximation of y(xo + h) is
h
(50.31) y(xo + h) = y(xo) + 2 [Y'(xo) + y'(xo + h),].

Formula (50.31) is known as the trapezoidal rule for approximating a

numerical solution of (50.1) satisfying (50.11).
If in (50.23), we stop with the .:l2 term, then by Comment 50.16, F(x)
is the polynomial interpolation function of f[x,y(x)] of degree less than or
equal to two, agreeing with f at the three points for which x = Xo, Xo + h,
Xo + 2h. With n = 2, and using only tenns including .:l2 of (50.23), we
obtain, with the help of Comment 48.18,
"'0+ 211
(50.32)
1 "'0
F(x) dx = 2h[1 +.:l + i(t - 1).:l2] y'(xo)
= 2h[y'(xo) + .:ly'(xo) + i.:lV(xo)]
= 2h(y'(xo) + y'(xo + h) - y'(xo)
+t[y'(xo + 2h) - 2y'(xo + h) + y'(xo)]l
h
= 3" [y'(xo) + 4y'(xo + h) + y'(xo + 2h)].

Therefore by (50.14) and (50.32), an approximation to y(xo + 2h) is

h
(50.33) y(xo + 2h) = y(xo) + 3" [Y'(xo) + 4y'(xo + h) + y' (xo + 2h)].

Fonnula (50.33) is known as Simpson's rule for approximating a nu-

merical solution of (50.1) satisfying (50.11).
676 NUMERICAL METHODS Chapter 10

If in (50.23), we stop with the d 3 term, then by Comment 50.16, the

polynomial interpolation function F(x) is of degree less than or equal to
three, agreeing with f[x,y(x)] at the four points for which x = Xo, Xo + h,
Xo + 2h, Xo + 3h. With n = 4, * and using only terms including d 3 of
(50.23), we obtain, with the help of Comment 48.18,
r"'O+4h
(50.34) 1"'0 F(x) dx
= 4h[1 + 2d + HY- + !(16 - 16 + 4)d 3]y'(Xo)
-2) d 2
= 4h[y'(xo) + 2dy'(Xo) + !dV(Xo) + idV(xo)]
= 4h/Y'(xo) + 2[y'(xo + h) - y'(xo)]
+ ![y'(xo + 2h) - 2y'(xo + h) + y'(xo)]
+ i[y'(xo + 3h) - 3y'(xo + 2h)
+ 3y'(xo + h) - y'(xo)]}
4h
= "3 [2y'(xo + h) - y'(xo + 2h) + 2y'(xo + 3h)].
Therefore, by (50.14) and (50.34), an approximation of y(xo + 4h) is
(50.35) y{xo + 4h) = y(xo)
4h
+ "3 [2Y'(xo + h) - y'(xo + 2h) + 2y'(xo + 3h)].
Proceeding as we did above, we can obtain approximating polynomials
F(x) of still higher degrees, agreeing with f[x,y(x)] at more and more
points. The only other two approximating formulas, however, that will
interest us are those in which F(x) is of degree five and six; see Exercise
50,2 for the case when F(x) is of degree four. For a fifth degree poly-
nomial interpolating function, (50.23) will give
"'O+6h
(50.36) j "'0
F(x) dx

5h
= 288 [19y'(xo) + 75y'(xo + h) + 50y'(xo + 2h) + 5Oy'(xo + 3h)
+ 75y'(xo + 4h) + 19y'(xo + 5h)].
Therefore, by (50.14), an approximation of y(xo + 5h) is
(50.37) y(xo + 5h) = y(xo) + K,
where K is the expression on the right of (50.36).
"The reason for taking n = 4 instead of n = 3 will become apparent when we use
the resulting Formula (50.35). For the case when n = 3, see Exercise 50,1.
Lesson 50 ApPROXIMATION FORMULAS. SIMPSON. WEpDLE. 677

For a sixth degree interpolating function, (50.23) will give

",0+6h
(50.38)
1 "'0
F(x) dx
3h
= 420 [41y'(xo) + 216y'(xo + h) + 27y'(xo + 2h)
+ 272y'(xo + 3h) + 27y'(xo + 4h)
+ 216y'(xo + 5h) + 41y'(xo + 6h)).
Hence, by (50.14), an approximation of y(xo + 6h) is
(50.39) y(xo + 6h) = y(xo) + G,
where G is the expression on the right of (50.38).
Formula (50.39) is usually replaced by the following formula,
(50.4) +
y(xo 6h) = y(xo)
3h
+ 420 [42y'(xo) + 21Oy'(xo + h) + 42y'(xo + 2h)
+ 252y'(xo + 3h) + 42y'(xo + 4h) + 21Oy'(xo + 5h) + 42y'(xo + 6h)).
Note that it adds a little to some terms in (50.38) and subtracts a little
from others, so that its overall accuracy is very close to that given by
(50.38). Its advantage over (50.38) lies in its being reducible to the
simpler form
(50.41) y(xo + 6h) = y(xo)
3h
+ 10 (y'(xo) + 5y'(xo + h) + y'(xo + 2h) +6y'(xo +3h)
+ y'(xo + 4h) + 5y'(xo + 5h) + y'(xo + 2h)).
This last formula is known as Weddle's rule.
We obtained (50.23) by using Newton's forward interpolation formula
(49.17). As an exercise, start with Newton's backward interpolation for-
mula (49.24) and show that

(50.5)
("'o+nh
J"'0 F(x) dx =
[
nh 1 + V +
(2)
+ V
2

+n2+n) V3 + + 3;3 + l;n2 +3n) V4

3 2
+-1 (n5 35n - +
120 -6+ 2 n 4 +
4
5On
3- + 12n V
5
+ ...] y'(xo) .
For convenient reference we collect the approximation formulas de-
veloped in this lesson.
 CI\
-I
co

Z
q
Interpolating Polynomial Approximate Numerical Solution of (50.1) Satisfying (50.11) Reference J;C
t'l

h
(50.6) a+ bx y(xo + h) = y(xo) + "2 [Y'(xo) + y'(xo + h)] (50.31)
n
:.-t"'
(trapezoidal rule)
t'l

h
Sl0
(50.61) a+bx+cx
2
y(xo + 2h) = y(xo) + 3 [y'(xo) + 4y'(xo + h) + y'(xo + 2h)] (50.33) 0
rfl
(Simpson's rule)

4h
(50.62) Third degree polynomial y(xo + 4h) = y(xo) + 3" [2y'(xo + h) - y'(xo + 2h) + 2y'(xo + 3h)] (50.35)

5h
(50.63) Fifth degree polynomial y(xo + 5h) = y(xo) + 288 [19y'(xo) + 75y'(xo + h) + 50y'(xo + 2h) (50.37)

+ 50y'(xo + 3h) + 75y'(xo + 4h) + 19y'(xo + 5h)]

(50.64) Weddle's rule y(xo + 6h) = y(xo) + 3h

10 [y'(xo) + 5y'(xo + h) + y'(xo + 2h) (50.41)

+ 6y'(xo + 3h) + y'(xo + 4h) + 5y'(xo + 5h) + y'(xo + 6h)] t:r

III
...
'1:1
.....
It

=
Lesson 50 ApPROXIMATION FORMULAS. SIMPSON. WEDDLE. 679

We also collect below for convenience, the error term associated with
each of the above formulas. *

Interpolating Polynomial Error

3
(50.7) a + bx (trapezoidal rule) - !!:.... y"'(X)
12
5
(50.71) a + bx + cx2 (Simpson's rule) _!!:.... y(5i (X)
90

(50.72) Third degree polynomial 28 h 5 (S)(X)

90 y

(50.73) Fifth degree polynomial (7)(X)

12,096 y
9
h7 (7) 9h (9)
(50.74) Weddle's rule - 140 y (Xl) - 1400 y (X2)

In all the above error formulas, X is a value of x in the interval

(xo, Xo+ + nh in the respective formulas
nh), corresponding to the Xo
(50.6) to (50.64).
Comment on Approximation Formulas (50.6) to (50.64). We
assume in the discussion which follows that y(x) is a solution of
(50.8) y' = f(x,y)
satisfying the initial condition
(50.81)
Hence when x = Xo, y' by (50.8), has the value
(a) y'(xo) = j[xo,y(xo)].
Since y(xo) is the initial condition given in (50.81), we can calculate y'(xo)
of (a). Again by (50.8), when x = Xo + h,
(b) y'(xo + h) = f[xo + h, y(xo + h)].
To determine y'(xo + h) of (b), we must therefore know the value of
y(xo + h). But we cannot know y(xo + h) unless we know y(x). And
since y(x) is the very solution we seek, we cannot hope to evaluate
y'(xo + h) by use of (b). It therefore follows that we cannot, for example,
use formula (50.6) unless we have other means available by which we can
estimate a value of the y'(xo + h) term in it.

·William Edmund Milne, Numerical Calculu8, Princeton University Press, Princeton,

N.J., 1949; Numerical Solutions of Differential Equations, John Wiley & Sons, New York,
1953.
680 NUMERICAL METHODS Chapter 10

All the remaining formulas (50.61) to (50.64) have a similar drawback.

Formula (50.61), for example, can be used only after we have been able
to obtain, in addition to y(xo), estimated values of y(xo + h) and
y(xo + 2h). With them, we can then, by (50.8), estimate values of the
terms
y'(xo + h) = f[xo + h, y(xo + h)]
and y'(xo + 2h) = f[xo + 2h, y(xo + 2h)]
that appear in the formula. The last formula (50.64) requires six pre-
liminary estimates in addition to y(xo), before it can be used, namely
estimates of y(xo + h), ... , y(xo + 6h). By (50.8), we can then estimate
values of y'(xo + h), ... , y'(xo + 6h).
With the exception of (50.62), there is another unusual feature about
all these formulas. Consider for example formula (50.61). After the
needed two preliminary values of y'(xo + h) and y'(xo + 2h) have been
estimated by finding approximate values of y(xo + h) and y(xo + 2h) by
other means, all (50.61) will do is give a value of y(xo + 2h) all over
again. Why then is the formula necessary at all? If we have approximated
y(xo + 2h) by some other method, why calculate y(xo + 2h) once more?
It turns out that formula (50.61) is actually a corrector formula, i.e., re-
peated application of the formula will improve an estimated approxima-
tion of y(xo + 2h) computed by a less accurate formula than itself (as
determined by their respective error terms). We shall clarify this point at
the time we use the formula. It is therefore called appropriately a corrector
formula. If you will carefully examine all the other formulas in this list,
you will find that, with the exception of (50.62), all are corrector formulas.
All they will do is correct the last estimated approximation of y(xo + nh)
[needed to approximate y'(xo + nh) in the formula] obtained by other less
accurate methods.
Formula (50.62) on the other hand is a continuing formula. With
it, we can evaluate y(xo + 4h) provided we have obtained estimated values
of y(xo) and y'(xo + h), y'(xo + 2h), y'(xo + 3h).
It is evident, therefore, that we cannot start to use any of the formulas
(50.6) to (50.64), whether to correct or to continue, unless we have a cer-
tain number of preliminary estimates. It is for this reason that the for-
mulas developed in previous lessons, since they do not require preliminary
estimates for their use, have been called appropriately starting formulas.
There are, therefore, as mentioned in the introduction to this chapter,
three types of formulas in numerical methods.
1. Starting formulas.
2. Continuing formulas.
3. Corrector formulas.
We have already developed various starting formulas. In the next lesson,
we shall describe a simple continuing and corrector combination formula ..
Leason 50-Exercise 681

EXERCISE 50
1. In (50.23), take n = 3 and stop with the term. The polynomial in-
terpolating function F(x) is therefore of degree less than or equal to three,
agreeing withf[x,y(x)] of (50.1) at four points. Prove that
ZO+3h
3h
"'0 F(x) dx = 8" [y'(xo) + 3y'(xo + h) + 3y'(xo + 2h) + y'(xo + 3h)].
/.

This formula is known as the three-eighths rule.

2. In (50.23), take n = 4 and stop with the term. The polynomial inter-
polating function F(x) is therefore of degree less than or equal to four, agree-
ing with f[x,y(x)] of (50.1) at five points. Prove that
"'0+4h
2h
(50.82) "'0 F(x) dx = 45 [7y'(xo) + 32y'(xo + h) + 12y'(xo + 2h)
/.
+ 32y'(xo + 3h) + 7y'(xo + 4h)].
3. Prove (50.36) by taking n = 5 and stopping with the term in (50.23).
4. Prove (50.38) by taking n = 6 and stopping with the term in (50.23).
5. Prove (50.5). Hint. Start with Newton's backward interpolation formula
(49.24) and follow the procedure used in the text to arrive at (50.23).
6. Prove that, if n = 1, (50.5) reduces to
"'O+h
(50.83) /.
F(x) dx
"'0 + tV + -fiv2 + iv3 + illV4 + Mv5 + .. .]y'(xo).
= h[1
Hence an approximation to y(xo + nh) of (50.14), with n = 1, in terms of
backward differences, is
(50.84) y(xo + h) = y(xo) + k,
where k is the expression on the right (50.83). The method which makes use
of formula (50.84) is known as Adams' method of approximating a nu-
merical solution of y' = f(x,y) satisfying y(xo) = Yo. To calculate V"y'(xo) ,
you must know m +
1 evenly spaced values of y'(x), namely y'(xo),
y'(xo - h), ••.• ,y'(xo - mh), see Lesson 49B. These values can be
obtained from the given differential equation y' = f(x,y) only after corre-
sponding values of y(xo), ..• , y(xo - mh) have been found by starting
formulas. Hence (50.84) is a continuing formula.
By means of Adams' method, find an approximate value, when x = 0.5,
of the particular solution of the differential equation y' = x 2 +
y for which
yeO) = 1. Take h = 0.1 and stop with the V4 term in (50.84). Hint. By
(50.84), with Xo = 0.4 and h = 0.1,
(50.85) y(0.5) = y(O.4) + O.I[y'(O.4) + tVy'(O.4) + -fiV2y'(0.4)
+ iV3y'(0.4) + H!V4y'(0.4)].
To calculate V4y'(O.4), you must know y'(O.4), y'(0.3), y'(0.2), y'(O.I), y'(O).
From the given equation, when x = 0, y = 1, y'(O) = 0 +
1 = 1. Use
Table 46.31, creeping up process column, to find the other needed values
of y'.* Then construct a table of differences of y'(0.4). Head the first
column y' = x 2 +
y and enter the values of y'(O), y'(O.I), y'(0.2), y'(0.3),

"The actual values will, of course, give a more accurate result for y(O.S). However,
we llo88ume these actual values are unknown.
682 NUMERICAL METHODS Chapter 10

y'(O.4). Head the second column Vy' and enter in it the differences of the
firstcolumn-rememberVy'(O.l) = y'(O.l) - y'(O.O), Vy'(0.2) = y'(0.2) -
y'(O.l). Head the third column V2y' and enter in it differences of the second
column. Head the fourth column V3 y' and enter in it differences of the third
column. Head the last column V4y' and enter in it differences of the fourth
column. The last row should then contain all the differences needed in the
above formula (50.85).
7. In (50.1), let !(x,y) be a function only of x so that y' = lex) for which
y(xo) = Yo. Hence, by (50.14),
%o+nh
(50.9) y(xo + nh) = y(xo) + f.
%0
lex) dx.

Let F(x) be the interpolating function for lex), agreeing with lex) at the
n +
1 points whose abscissas are Xo, Xo h, Xo +
2h, •• " Xo nh, + +
f[xo +(n -l)h] =
F[xo+(n-l)h]

f(xo +2h)= f(xo+nh)=

F(xo + 2h) F(xo+nh)

Xo xo+h xo+2h xo+(n -l)h xo+nh

Figure 50.91

Fig. 50.91. Let the graph of F(x) be the straight line joining the end points
of these ordinates. Then an approximation of y(xo +
nh) of (50.9) is
%o+nh
(50.92) y(xo + nh) = y(xo) + f.
%0
F(x) dx,

where F(x) is the function whose graph consists of these straigftt lines.
By (50.6) and (50.7), with y' replaced by its equal !(x) , we have

(50.93) y(xo + h) = y(xo) + 2h [!(xo) + !(xo + h)J + E,

where E = -h 3f"(X)/12, and X is a value of x in the interval (xo,xo + h).
h
The term 2 [!(Xo) + !(xo + h») may also be looked at as the area of a
trapezoid of width h and heights !(xo), !(xo +
h), Fig. 50.91; hence the
name trapezoidal rule.
(a) By adding the areas of each trapezoid in Fig. 50.91, show that
%o+nh
(50.94)
f.%0 F(x) dx = {!(xo) + 2!(xo + h) + 2!(xo + 2h) + ...
+ 2f[xo + (n - l)h] + I(xo + nh)} . ..
Hence an approximation to (50.9) is
(50.95) y(xo + nh) = y(xo) + H,
where H is the right side of (50.94).
 Lesson 50-Exercise 683

(b) Following the procedure outlined in Exercise 44,11, show that the
upper bound of the error in formula (50.95) due to using F(x) to ap-
proximate f(x) in (50.9), is
3
(50.96) lEI
where M is the maximum value of I!"(x) I in the interval (xo, Xo nh). +
8. (a) Use formula (50.95), with n = 4, h = t, to obtain an approximate
value, when x = 2, of the particular solution of the equation y' = l/x
for which y(l) = O. Show how this value of y(2) can be used to ap-
proximate log 2. Hint. Here Xo = 1, y(xo) = 0, nh = 1. The solution
of y' = l/x for which y(l) = 0 is y = log x. Therefore y(2) = log 2.
(b) Use (50.96) to find an upper bound to the error in the result obtained
in (a). Compare with actual error. Hint. n = 4, h = t, f(x) = l/x,
M = max. I!"(x) I = max. 12/x3 1 = 2 in interval (1,2).
(c) What is the largest value of h which can be used to insure that the error
E in the computation of log 2 is less than 0.0005? Into how many parts
would it be necessary to divide the interval (1,2)?
9. In (50.1), let f(x,y) be a function only of x, so that y' = f(x) for which
y(xo) = Yo. Hence, by (50.14),
",o+nh

(50.97) y(xo + nh) = y(xo) + f.

"'0
f(x) dx.

Let F(x) be the interpolating function for f(x), agreeing with f(x) at the
n + 1 points whose abscissas are x, Xo t h, Xo + 2h, ••• , Xo + nh, where
n is even. Let the graph of F(x) consist of the parabolic arc connecting the
three ordinates f(xo), f(xo + h), f(xo + 2h), solid curve in Fig. 50.91, plus
the parabolic arc connecting the three ordinates f(xo + 2h), f(xo + 3h),
f(xo + 4h), etc. Then an approximation to y(xo + nh) of (50.97) is
",o+nh
(50.98) y(xo + nh) = y(xo) + f.
"'0
F(x) dx,

where F(x) is the function whose graph consists of these parabolic arcs.
By (50.61) and (50.71), with y' replaced by its equal f(x) , we have

(50.99) y(xo + 2h) = y(xo) + 'h3 [f(xo) + 4f(xo + h) + f(xo + 2h)] + E,

where E = -h 5J<4) (X)/90, and X is a value of x in the intervA.l (xo,xo + 2h).
The term [f(xo) + 4f(xo + h) + f(xo + 2h)], it has been proved, see a
calculus text, is the area under a parabolic arc joining the ends of three
ordinates f(xo) , f(xo +
h), f(xo 2h). +
(a) By adding the areas under each parabolic arc in Fig. 50.91, show that
",o+nh
" (50.991)
f."'0 F(x) dx = {f(xo) + 4f(xo + h) + 2f(xo + 2h)
+ 4f(xo + 3h) + 2f(xo + 4h) + ...
+ 2f[xo + (n - 2)h]
+ 4![xo + (n - l)hJ + f(xo + nh) \.
684 NUMERICAL METHODS Chapter 10

Hence an approximation to (50.97) is

(50.992) y(xo + nh) = y(xo) + k,
where k is the right side of (50.991). Formula (50.992) is known as
Simpson's rule.
(b) Following the procedure outlined in Exercise 44,11 [note that here the
+
interval is (xo, Xo 2h) instead of (xo, Xo +
h)], show that the upper
bound of the error in formula (50.992) due to using F(x) to approximate
I(x) in (50.97), is

(50.993)

where M is the maximum value of 1J<4)(x)1 in the interval (xo,xo + nh).

10. Use formulas (50.992) and (50.993), with n = 4, h = ! to answer questions
(a), (b), (c) of problem 8. Hint. Here M = max. 1/(4)(x) I = max.124/x 51 =
24 in interval (1,2).

ANSWERS 50
6. y(0.5) = 1.6961610.
8. (a) y(2) = 0.69702. Actual value: log 2 = 0.69315. (b) lEI
< 0.01042.
Actual error = 0.00387. (c) h = 0.05 to two decimal places. Therefore
we need to divide the interval (1,2) into twenty parts.
10. (a) y(0.2) = 0.69325. (b) lEI < 0.00052. Actual error = 0.00010.
(c) h < 0.24. Since n must be even, we need to divide the interval (1,2)
into six parts. Compare with 8(c).

LESSON 51. Milne's Method of Finding an Approximate

Numerical Solution of y' = f(%,y).
The method we are about to describe for finding a numerical solution
of the differential equation y' = f(x,y) satisfying the initial condition
y(xo) = Yo is probably the one most widely used today. It is simple in
form and has a relatively high degree of accuracy. The method uses con-
tinuing formula (50.62) to estimate or predict a value of y(xo + 4h) and
then employs Simpson's formula (50.61) to correct it. For convenience
we rewrite these formulas below, using a subscript p for the formula we
shall use as a predictor or estimator and a subscript c for the one we shall
use as a corrector.
(51.1) yp(xo + 4h) = y(xo)
4h
+ "3 [2y'(xo + h) - y'(xo + 2h) + 2y'(xo + 3h)],

(51.11) Yc(xo + 4h) = y(xo + 2h)

+ 3h + 2h) + 4y'(xo + 3h) + YP'(xo + 4h)].

When these two formulas are used in combination, they are known col-
lectively as Milne's method. In order to use (51.1) we must know
Lesson 51 MILNE'S METHOD ApPLIED TO y' = f(x,y) 685

y'(xo + h), y'(xo + 2h), y'(xo + 3h). We can determine these values
from the given differential equation y' = f[x,y(x)] only after we know the
corresponding values of y(xo + h), y(xo + 2h), y(xo + 3h). Hence we
must use starting formulas in order to obtain these needed preliminary
estimates. Formula (51.1) will then predict or estimate a value of
y(xo + 4h); formula (51.11) will .correct this estimate. The new value of
y(xo + 4h) thus found can in its turn be used as a new estimated value
and (51.11) used over again to correct it. It has been proved that, if the
original estimate is not too far away from the true value and if h is suffi-
ciently small, the repeated use of (51.11) will give a sequence of values of
y(xo + 4h) which will eventually converge.
Example 51.12. Use Milne's method to find an approximate value
when x = 0.5 of the particular solution of the differential equation

(a) y' = x2 + y,
for which yeO) = 1.
Solution. In Example 46.21, we found by Taylor series methods,

(b) y(O.I) = 1.1055125, y'(O.I) = 1.1155125;

y(0.2) = 1.2242077, y'(0.2) = 1.2642077;
y(0.3) = 1.3595755, y'(0.3) = 1.4495755.

By (51.1) and (51.11) with Xo = 0, h = 0.1,

(c) yp(O.4) = yeO) + 034 [2y'(0.1) - y'(0.2) + 2y'(0.3)],

yc(0.4) = y(0.2) + 031 [y'(0.2) + 4y'(0.3) + YP'(O.4)].
Substituting the values of (b) and the initial condition yeO) = 1, in the
first equation of (c), we obtain

(d) yp(O.4) = 1 + 034 - 1.2642077 + 2(1.4495755)]

1.5154624.
By (a) and (d),
yP'(O.4) = 0.16 + 1.5154624 = 1.6754624.

Hence the second equation of (c) becomes

(e) Yc(O.4) = 1.2242077+ 031 [1.2642077 + 4(1.4495755)

+ 1.6754624] = 1.2242077 + 0.2912657 = 1.5154734.
686 NUMERICAL METHODS Chapter 10

We now make use of (51.11) once more to see whether it will correct the
value in (e). Using the value in (e) as a new estimate, we have
(f) Yp'(Oo4) = 004 2 + 1.5154734 = 1.6754734.

Our corrector formula (c) becomes, with the help of (b) and (f),

(g) yc(004) = 1.2242077 + 031 [1.2642077 + 4(1.4495755)

+ 1.6754734] = 1.5154738.
You can verify that a third application of (51.11), using the value in (g)
as a new estimate, will not change this value. Since our corrector will
make no more corrections, we accept this value of y(004).
From here on we continue to make use of our predictor and corrector
formulas, using the first to estimate an approximation, the second to cor-
rect this estimate; the corrector formula being used repeatedly until no
further correction results. By (51.1), with h = 0.1 and Xo = 0.1,

(h) yp(0.5) = y(O.I) + 034 [2y'(0.2) - y'(0.3) + 2y'(004)].

By (a) and (g)
(i) y'(0.4) = 0.16 + 1.5154738 = 1.6754738.

Therefore by (b) and (i), (h) becomes

(j) yp(0.5) = 1.1055125 + 034 [2(1.2642077) - (1.4495755)

+ 2(1.6754738)] = 1.6961508,
r
which is an estimated value of y(0.5). Using this value in (a), we obtain
(k) yP'(0.5) = 0.25 + 1.6961508 = 1.9461508.

By (51.11), with h = 0.1, Xo = 0.1,

(I) yc(0.5) = y(0.3) + 3h [y'(0.3) + 4y'(004) + yp'(0.5)].

By (b), (i), and (k), (I) becomes

(m) yc(0.5) = 1.3595755 + 031 [1.4495755 + 4(1.6754738)

+ 1.9461508] = 1.6961629.

Using the value y(0.5) = 1.6961629 as a ne"w estimate, we find by (a),

(n) yp'(0.5) = 0.25 + 1.6961629 = 1.9461629,
Lesson 51 MILNE'S METHOD ApPLIED TO y' = f(x,y) 687

and by (51.11)
(0) Yc(0.5) = 1.3595755 + °31 [1.4495755 + 4(1.6754738)
+ 1.9461629] = 1.6961633.
You can verify that further use .of (51.11) will not change the value in
(0). We therefore accept this estimated value of y(0.5).
Comment 51.2. In this manner, alternately using predictor formula
(51.1) and corrector formula (51.11) as many times as needed, you can
obtain y(0.6), y(0.7), y(0.8), etc. It will be found desirable to construct a
table in which to record all relevant calculations as they are found. Its
form is given in Table 51.22 below. The letter D which appears in it is
defined as
(51.21) D(xo + 4h) = Yc(xo + 4h) - yp(xo + 4h),
where YP is the estimate computed by (51.1) and Yo is the first correction
of this YP computed by (51.11). Its important purpose will be explained
later. We have left it to you as an exercise to complete the line beginning
with 0.6.
Table 51.22

yp of Yc of D of Actual
x yp' yo'
(51.1) (51.11) (51.21) Value

0.4 1.5154624 1.5154734 1.6754624 1.6754734 0.0000110

1.5154738 1.6754738 1.5154741
0.5 1.6961508 1.6961629 1.9461508 1.9461629 0.0000121
1.6961633 1.9461633 1.6961638
0.6 1.9063564

Comment 51.23. Formula (51.1) is a continuing formula. However,

it is rarely, if ever, used for this purpose since, as we shall show below, it is
considerably less accurate than corrector formula (51.11). By (50.72), the
error term Ep of (51.1) is

(51.24)

By (50.71) the error term E. of (51.11) is

(51.25)

If we may assume that h is sufficiently small so that the variation in value

of y(S)(X I ) and y(S)(X 2 ) in an interval of width 4h [Xl is a value of x in
688 NUMERICAL METHODS Chapter 10

the interval (xo, Xo + 4h); X 2 is a value of x in the interval (xo + 2h,

Xo + 4h)] is negligible, then we see from error formulas (51.24) and (51.25)
that the corrector formula at each single step is approximately 28 times
as accurate as the predictor formula.
Comment 51.3. Comment on Error in Milne's Method.
1. Call Ep(xo +
4h) the error in computing yp(xo +
4h) by using for-
mula (51.1); call Ec(xo +
4h) the error in computing Ye(xo 4h) by +
using formula (51.11) to correct this predicted value of Yp(xo + 4h).
Let Y(xo + 4h) be the actual value of y(xo +
4h). Then

(51.31) Ec(xo + 4h) = Y(xo + 4h) - Yc(xo + 4h),

Ep(xo + 4h) = Y(xo + 4h) - Yp(xo + 4h).
Subtracting the first equation from the second, we have
(51.32) Ye(Xo + 4h) - Yp(xo + 4h) = Ep(xo + 4h) - Ee(xo + 4h).
If we may assume that h is sufficiently small so that the variation between
yeo) (X 1) of error formula (51.24) and yes) (X 2) of error formula (51.25) is
negligible, then we commit a small error by using y(S)(X) as their approxi-
mate common value. Hence subtracting (51.25) from (51.24) and re-
placing y(S)(X 1 ) and y(S)(X 2 ) by y(S\X), we obtain

(51.33)
By (51.21), the left side of (51.32) is D(xo + 4h). Hence by (51.32),
(51.33), and (51.25), we have
(51.34) D(xo + 4h) = flhDylI(X) = -29Ee(xo + 4h).
Therefore by (51.34),
(51.35) Ee(xo + 4h) = D(xo + 4h)
29
Formula (51.35) tells us that the approximate formula error in the first
corrected value of Yc(xo + 4h) is the difference between the first
corrector and predictor values of y(xo + 4h).
2. The column headed D in Table 51.22 thus serves a very useful pur-
pose. Dividing it by 29 will give an estimate of the formula error in the
first corrected value of Ye over one h step. If therefore we want lEI <
0.000005, then ID/291 must be <0.000005 or IDI must be <0.000145.
As long as IDI remains less than this figure, we assume that the error in
our first corrected esticate Ye, over each h step, is less than 0.000005.
3. As long as D/29 is less than the desired accuracy, we may continue
to use the same h interval. As soon as D/29 becomes greater than desired
accuracy, we must reduce h. Conversely, if D/29 is considerably smaller
Lesson 51-Exercise 689

than desired accuracy, we can safely increase h. The method of reducing

and increasing h in the course of an extended computation is discussed in
the next Lesson 52.
4. The corrector formula may be used repeatedly at each step, until
there i<; no difference between two successive values of y(xo +
4h). How-
ever, as the difference between predictor and first corrector values in-
creases, you will find that you must use the corrector more and more
times at each step. When this happens, even though D/29 may still be
less than desired accuracy, it will usually be found best to reduce h.
5. The column headed D serves another useful purpose. It controls in
Borne measure arithmetical accuracy at each step. Whenever an entry in
D shows a sudden change from a definite behavior pattern, the preceding
and current calculations should be checked.
6. If predictor and corrector values agree to k decimal places after being
properly rounded off, then k decimal accuracy is assumed.

EXERCISE 51
1. Complete line 0.6 in Table 51.22.
2. Replace (b) of Example 5Ll2, by the actual values of y(O.l), y(0.2), y(0.3) as
given in Table 46.31. Following the method of this lesson, compute yp(O.4),
y.(O.4), and D(O.4). Then use (51.35) to correct y.(O.4). Add this corrected
value of y(O.4) to your table. Now proceed to compute yp(O.5), y.(O.5), and
D(0.5). Use (51.35) to correct y.(0.5). Add this corrected value of y(0.5) to
your table. Calculate yp(0.6), y.(0.6), and D(0.6). Correct yc(0.6). Compare
these values with those obtained previously and with actual values given in
Table 51.22.
3. Using the six preliminary results obtained in 2, compute yw(0.6) by means of
Weddle's rule (50.64). In the absence of a solution, how many decimal place
accuracy could you assume in y(0.6)? Hint. See 3 after "3. Cumulative
Errors," Lesson 52A.
4. Find numerical approximations when x = 0.4, 0.5, 0.6 of the particular solu-
tion of the differential equation y' = X+ y for which yeO) = 1. Take h =
0.1. Follow the method employed in Example 5Ll2, using first the predictor
formula, then the corrector formula as often as is necessary until no further
correction results. For preliminary values, take y(O.l) = 1.1103418, y(0.2) =
1.2428055, y(0.3) = 1.3997176. Solve the equation and compare your results
with actual values.
5. Using the six preliminary results obtained in 4, compute yw(O.6) by means of
Weddle's rule (50.64). In the absence of a solution, how many decimal place
accuracy could you assume in y(0.6)? See Hint in 3.
6. Find numerical approximations when x = 1.4, 1.5 of the particular solution
of the differential equation y' = xy for which y(l) = 1. Take h = 0.1.
Follow the method used in Example 51.12, employing first the predictor and
then the corrector formula as often as necessary. For preliminary values
take yeLl) = Ll1071, y(1.2) = 1.24608, y(1.3) = 1.41199. Solve the equa-
tion and compare results with actual values.
7. Using the results obtained in problem 6, compute y(1.5) by means of corrector
formula (50.63). Here Xo = 1, y(xo) = 1, h = 0.1. In the absence of a solu-
tion, how many decimal place accuracy could you assume? See Hint in 3.
690 NUMERICAL METHODS Chapter 10

ANSWERS 51
1. yp(0.6) = 1.9063415, y.(0.6) = 1.9063561, D(0.6) = 0.0000146.
2. Yp(O.4) = 1.5154627, y.(O.4) = 1.5154742, D(O.4) = 0.0000115.
y(O.4) = 1.5154738, yp(0.5) = 1.6961512, y.(0.5) = 1.6961638,
D(0.5) = 0.0000126, y(0.5) = 1.6961634, yp(0.6) = 1.9063424,
y.(0.6) = 1.9063561, D(0.6) = 0.0000137, y(0.6) = 1.9063556.
3. yw(0.6) = 1.9063563. Can assume six decimal place accuracy if rounded
off to six decimals.
4. y.(O.4) = 1.5836497, y.(0.5) = 1.7974429, y.(0.6) = 2.0442384. Actual
values: 1.5836494, 1.7974425, 2.0442376.
5. yw(0.6) = 2.0442377. Can assume six decimal place accuracy.
6. yc(1.4) = 1.61609, y.(1.5) = 1.86826. Actual values: 1.61607, 1.86825.
7. y(1.5) = 1.86825. Can assume four decimal place accuracy if rounded off to
four decimals.

LESSON 52. General Comments. Selecting h. Reducing h.

Summary and an Example.

LESSON 52A. Comment on Errors.

1. Formula Errors. With each approximating formula we have also

given a companion error term which measures the magnitude of the error.
Since these error formulas cannot usually be used, we have also suggested
practical means by which the magnitude of error can be estimated. Unless
a numerical method has associated with it a useful error formula, the
method is of little value.
2. Rounding off Errors. In a practical problem, the number of steps
needed will usually be known, also the accuracy desired. It will thus be
possible to estimate the number of decimal places that should be used at
the start in order to offset rounding off errors under the most unfavorable
circumstances. For example, suppose you want four decimal accuracy
and expect to use eight steps. If you round off to six decimals, then the
maximum possible absolute value of the error due only to rounding off,
i.e., due to dropping the seventh and later decimals, is, after eight steps,
0.0000005 X 8 = 0.000004. It must be remembered, however, that the
use of a formula may considerably magnify, at each step, the effect of the
rounding off error.
3. Cumulative Errors. At each step of a building up process, two errors
occur. One, because we start off with an inherited error, and two, because
we are using an approximation formula. Since each succeeding step de-
pends on the previous estimate, it will be unusual indeed if we obtain
increasing accuracy as we proceed. It may happen in a rare case that one
error may be offset by a succeeding one. In the usual case, it will not
happen, and accuracy will decrease at each step.
Although formulas exist which will give the upper limit of error at the
end of each step due both to formula and cumulative errors, they are not
Lesson 52B CHOOSING THE SIZE OF h 691

easy to use. We give below three practical suggestions for estimating

accuracy, ones which have been found adequate in most cases.

1. Start with many more decimals than you need.

2. Make all calculations over again with an h equal to one half its pre-
vious value. If the new final result agrees with the previous one to k
decimal places, after being properly rounded off, then k decimal place
accuracy is assumed.
3. Apply corrector formula (50.63) after five preliminary estimates have
been obtained or Weddle's corrector (50.64) after six steps. If the re-
sult obtained by these corrector formulas agrees with the last estimate
used in the formula to k decimal places, after being properly rounded
off, then k decimal place accuracy is assumed. Weddle's rule, in par-
ticular, is simple in structure and is extremely good at discovering
errors. If there is little agreement between the last estimated value
and its corrected value, then either an arithmetical error has been
made or h is too large.

A final word of caution. In most problems, the practical approach to

errors as outlined above, will within reasonable certainty, assure you of a
result which is correct to k decimal places. However, only a formula
which gives the upper bound of the error due to rounding off, formula and
cumulative errors, can give, with certainty, the magnitude of the error in
a numerical computation.

LESSON 52B. Choosing the Size of h. You may have been saying
to yourself, "How does one know what size h to select at the start?"
We used h = 0.1 in our examples, but what made us pick 0.1 instead of
0.2 or 0.05 or 0.3? If it is possible to use an error formula for a given
method, then it is also possible to determine h so that the error due to the
approximating formula will remain within the desired limits. Frequently
a knowledge of the problem plus practical experience will determine a
starting value of h.
In the absence of a useful error formula or practical information, then
all you can do is to start with an h which seems reasonable. Say you decide
to start with h = 0.3 and to use the Runge-Kutta method to get your
first approximations. Calculate y(0.3) in one step and then in two steps,
i.e., calculate y(0.3) and y(0.15+ 0.15). Formula (47.42) will then give
you an approximate value of the magnitude of the error. It states that
the approximate error in y(0.15 + 0.15) is one-fifteenth the difference
between y(0.15 + 0.15) and y(0.3). If one-fifteenth this difference is
greater than the desired error, you must reduce h; if it is reasonably less
than the desired error, you may retain h; if it is very much less than the
desired error, you may increase h.
692 NUMERICAL METHODS Chapter 10

Let us assume h = 0.3 is satisfactory. Starting with the value of y(0.3)

found in two steps, we then proceed to find y(0.6) and y(0.9). The initial
condition plus these three preliminary estimates, y(0.3), y(0.6) and y(0.9),
will enable us to switch to Milne's method. From here on, we begin to
watch the size of the column D in our table. By (51.21), it is the differ-
ence between the predictor value computed by using (51.1) and the first
corrected value computed by using (51.11). Dividing D by 29 will keep
us posted as to the approximate magnitude of the error in the first Yo
figure in our table. When it becomes larger than the desired error, we
must reduce h. This brings up the question of how to reduce h in the
course of an extended computation; also how to increase h.

LESSON 52C. Reducing and Increasing h. As we proceed step by

step, an h which is satisfactory in early stages may become too large in
later ones. Suppose, for example, we are satisfied that the values of
y(O.I), y(0.2), y(0.3) found by a starting method have the desired accuracy.
We switch to the Milne method and determine that the values of y(O.4),
y(0.5), y(0.6), y(0.7) still have the desired accuracy. However to obtain
y(0.8), we find it necessary to reduce h to 0.05. The next value we must
find is, therefore, y(0.75). To use Milne's predictor formula (51.1) with
h = 0.05, xo = 0.55, we must know y(0.55), y(0.6), y(0.65), y(0.7). We
already know y(0.6) and y(0.7). How do we find y(0.55) and y(0.65)
without the necessity of starting from the beginning all over again with
h = 0.05?
One way of obtaining y(0.55) and y(0.65) is by use of Newton's back-
ward interpolation formula (49.22). Since we know eight evenly spaced
values of y(x), 0.1 unit apart, namely yeO), y(O.I), ... , y(0.6), y(0.7), we
can use terms in this formula to V7 y(0.7). By (49.23), with Xo = 0.7
x = 0.65 and h = 0.1, we find n = (7 - 0.65)/0.1 = 1. With n = 1
Yl

Yo

Ya

xo-3h xo-2h xo-h Xo- Xo

-3h h
-2h -h -"2 0

Figure 52.1
Lesson 52C REDUCING AND INCREASING h 693

and the needed values in the formula obtained by constructing a table of

differences of y(0.7), we can determine y(0.65). Similarly, by taking
x = 0.55 so that n = I, we can approximate y(0.55).
A second and perhaps easier way to obtain these values is by use of a for-
mula which has approximately the same order of accuracy as Simpson's
rule. Let the third degree polynomial F(x) = a + bx + cx 2 + dx 3 be an
interpolating function for Vex). Since F(x) is a third degree polynomial, we
may assume by Theorem 48.2, that it agrees with Vex) at four points whose
abscissa values are h units apart. Call these four abscissas xo, Xo - h,
Xo - 2h, Xo - 3h, and their respective ordmates Yo, VI, Y2, Y3, Fig. 52.1.
For convenience in calculation, we take the origin at xo, so that the coordi-
nates of the points at which F(x) and Vex) agree are (O,Yo), (-h,YI),
(-2h,Y2), (-3h,Y3). Since each of these points satisfies the equation

(52.11) F(x) = a + bx + cx 2 + dx a,
we have
(52.12) Yo = a,
YI = a - hb + h 2c - h 3d,
Y2 = a - 2hb + 4h 2c - 8h ad,
Ya = a - qhb + 9h 2c - 27h 3d.
Solving (52.12), for a, b, c, d, we obtain
(52.13) a = Yo,
-
b-
11yo - 18YI + 9Y2 - 2Ya
6h '
2yo - 5YI + 4Y2 - Ya
c = 2h2 '
d = Yo - 3YI + 3Y2 - Ya .
6ha
Substituting (52.13) in (52.11) will give the equation of the interpolating
function F(x) for Vex). Therefore when x = -h/2, F( -h/2) will give the
approximate value of y( -h/2). Making the substitution (52.13) in
(52.11) and replacing x by -h/2, we have

(52.14) F (- = Yo _ llyo - 18YI + 9Y2 - 2Ya

2 12
+ 2yo - 5YI + 4Y2 - Ya _ Yo - 3YI + 3Y2 - Ya
8 48
_ 5yo+ 15YI - 5Y2+Y3
- 16
which is an approximation of y( -h/2). Hence in terms of our original
694 NUMERICAL METHODS Chapter 10

abscissas, see Fig. 52.1, we obtain the approximation formula,

(52.15) y (xO - =
M5y(xo) + 15y(xo - h) - 5y(xo - 2h) + y(xo - 3h)].
Comment 52.16. To double h is a simple matter. If you decide that
the h you have been using can be safely doubled, all you need do is to
take every other preceding estimate. For example, if you have been using
h = 0.1 and find h = 0.2 will be satisfactory, you need use only the pre-
vious estimates, y(xo + 0.2), y(xo + 0.4), etc.

LESSON 52D. Summary and an Illustrative Example. We have

given various starting, continuing, predictor, and corrector formulas.
Which ones you should choose in a particular problem will be determined
by the degree of accuracy desired, the relative difficulty of the method,
and the amount of labor involved. If you have many occasions to use
numerical methods, experience will be your best guide.
It is the usual practice to start with the Runge-Kutta method. After
the needed preliminary estimates have been obtained, it is then customary
to switch to the Milne method. In solving problem 52.2 below, we shall
assume that we know nothing about its solution. Hence we shall have to
depend for a determination of the approximate accuracy of our result on
the suggestions made in Lesson 52A.
E:cample 52.2. Find an approximate value when x = 0.6 of the par-
ticular solution of the differential equation
(a) y' = x2 + y,
for which yeO) = 1. Assume we wish the error to be less than 0.00005.
Solution. To bring more methods into the discussion, we shall staJ;t
with Taylor series instead of with the usual Runge-Kutta formulas, then
switch to Runge-Kutta and end finally with Milne.
First we must choose an h. We decide to try h = 0.1, and proceed to
calculate y(O.I) in two steps and in one step in order to see how much
agreement there is in the two results. Using our basic equation (46.12)
with h = 0.05 and Xo equal successively to 0 and 0.05, we obtain
(b) yeO + 0.05) = y(0.05) = + y'(0)(0.05)
yeO)
+ +
(0.05)2 (0.05)3 + (0.05)\

y(0.05 + 0.05) = y(O.I) = y(0.05) + y'(0.05)(0.05)

+ (0.05)2 + (0.05)3 + (0.05)4.
Leason 52D SUMYARY AND AN ILLUSTRATIVE EXAMPLE 695

The derivatives of (a) to order four are

(c) y' = x2 + y, y" = 2x + y', y'" = 2 + y", y(4) = y"'.

Hence when x = 0 and yeO) = 1, we find from (c)

(d) y'(O) = 1, y"(O) = 1, y"'(O) = 3,

Substituting these values in the first equation of (b), we obtain

(e) y(0.05) = 1 + 0.05 + 1(0.05)2 + !(0.05)3 + 1(0.05)4

= 1.05 + 0.00125 + 0.0000625 + 0.0000008 = 1.0513133.

With x = 0.05, y(0.05) = 1.0513133, we find from (c)

(f) y'(0.05) = 1.0538133, y"(0.05) = 1.1538133,

y"'(0.05) = 3.1538133, y(4l(0.05) = 3.1538133.

Substituting (e) and (f) in the second equation of (b), we obtain

(g) y(0.05 + 0.05)

= 1.0513133 + 1.0538133(0.05) + ;(1.1538133)(0.05)2
+ ,(3.1538133)(0.05)3 + i4(3.1538133)(0.05)4
= 1.0513133 + 0.05269067 + 0.00144227 + 0.00006570
+ 0.00000082 = 1.1055]28.
In Example 46.2, we found the value of y(O.I) in one By (f) of
that example, y(O.I) = 1.1055125. We now note that the value of y(O.I)
obtained in one step agrees, when rounded off, to six decimal places with
the value of y(O.I) obtained in two steps. We note also by (46.57), that

(h) E(O.I) = H[y(0.5 + 0.5) - y(O.I)] = H(0.0000003) = 0.00000032.

Since this error is sufficiently smaller than the desired one, and because
y(O.I) and y(O.05 + 0.05) agree to six decimals, we accept h = 0.1 as a
proper starting value.
Our next task is to determine how many decimal places to carry. Since
h = 0.1 and we want y(0.6), there will be, if h does not need to be re-
duced, a total of six steps. If therefore we carry seven decimal places,
dropping the eighth, the absolute value of the error due only to rounding
off, in the most unfavorable circumstances, is less than 6(0.00000005) =
0.0000003, which does not affect the sixth place. Since we want our error
to be <0.00005, we decide that seven decimals will give us a sufficient
margin of safety. We are now ready to calculate y(0.2). With Xo = 0.1
696 NUMERICAL METHODS Chapter 10

and h = 0.1, our basic equation (46.12) becomes

(i) y(0.2) = y(O.l) + y'(O.l) (0.1) + (0.1)2

+ (0.1)3 + (0.1)4 + ....

Although the value of y(O.l) = 1.1055128 found in two steps with h =
0.05 is more accurate than y(O.l) = 1.1055125 found in one step with
h = 0.1, we shall use the last figure because of the adoption of a 0.1
interval. We shall then be able to use those error fonnulas that are based
on h = 0.1.
By (c), with x = 0.1, y = 1.1055125.

(j) y'(O.l) = 1.1155125, y"(O.l) = 1.3155125,

ylll(O.l) = 3.3155125, y(4l(0.1) = 3.3155125.

Substituting the above values in (i), we obtain

(k) y(0.2) = 1.1055125 + (0.1)(1.1155125) + (1.3155125)

+ (3.3155125) + (3.3155125) = 1.2242077.

Before proceeding to find y(0.3) we shall check the accuracy of y(0.2) =

1.2242077 as given in (k). By (46.57),

(1) E(O.l + 0.1) = ts-[y(O.l + 0.1) - y(0.2)].

By (k)
(m) y(O.l + 0.1) = 1.2242077.
By (f) of Example 46.2
(n) y(0.2) = 1.2242000.
Hence
E(O.l + 0.1) = *(1.2242077 - 1.2242(0) = 0.0000005,

which is the approximate error in y(0.2) of (k) found in two steps. Since
this error is still much less than the desired one of 0.00005, we continue
to use h = 0.1.
To find y(0.3), we switch to the Runge-Kutta method. By (47.34),
with Xo = 0.2, h = 0.1, y(0.2) = 1.2242077,

(0) y(0.3) = 1.2242077 + !<Wl + 2W2 + 2W3 + W4)'

By (47.35), with h = 0.1, Xo = 0.2, Yo = y(xo) = y(0.2) = 1.2242077,
Lesson 52D SUMMARY AND AN ILLUSTRATIVE EXAMPLE 697

I(x,y) = x2 + y,
(p) WI = 0.11(0.2, 1.2242077) = 0.1(0.2 2 + 1.2242077)
= 0.12642077,
W2 = 0.lj(0.25, 1.2242077 +
0.0632104)
= 0.1(0.25 2 +1.2874181) = 0.1349918,
Wa = 0.lj(0.25, 1.2242077 +
0.0674959)
= 0.1(0.25 2 +1.2917036) = 0.1354204,
W4 = 0.11(0.3, 1.2242077 + 0.1354204)
= 0.1(0.3 2 + 1.3596281) = 0.1449628.
Substituting (p) in (0), we have

(q) y(0.3)
= 1.2242077 + i(0.1264208 + 0.2699836 + 0.2708408 + 0.1449628'
= 1.2242077 + 0.1353680
= 1.3595757.

We are now ready to proceed with Milne's method. For convenience, we

collect the results thus far obtained.

(52.21) yeO) = 1.0000000, y'(O) = 1.0000000;

y(O.l) = 1.1055125, y'(O.l) = 1.1155125;
y(0.2) = 1.2242077, y'(0.2) = 1.2642077 [by (a) and (k)];
y(0.3) = 1.3595757, y'(O.3) = 1.4495757 [by (a) and (q)].

The Milne formulas are, by (51.1) and (51.11),

(52.22) yp(xo + 4h) = y(xo)
4h
+ "3 [2y'(xo + h) - y'(xo + 2h) + 2y'(xo + 3h)],
(52.23) yc(xo + 4h) = y(xo+ 2h)
+ 3"h [y'(xo + 2h) + 4y'(xo + 3h) + YP "(xo +4h)].
Using (52.22) and the values in (52.21), we have with Xo = 0, h = 0.1,

(r) yp(OA) = yeO) + 034 [2y'(0.1) - y'(0.2) + 2y'(0.3)]

= 1.0 + 034 [2(1.1155125) - 1.2642077 + 2(1.4495757)]
= 1.5154625.
698 NUMERICAL METHODS Chapter 10

By (a) and (r),

(s) yP'(O.4) = 1.6754625.
By (52.23) with Xo = 0, h = 0.1,

(t) Yc(O.4) = y(0.2)+ 0;/ [y'(0.2) + 4y'(0.3) + yP'(O.4)

= 1.2242077 + 0 1 [1.2642077 + 4(1.4495757) + 1.6754625]
3
= 1.5154735.
And by (a) and (t),
(u) yc'(O.4) = 1.6754735.

Using the value in (u) as a new estimated value in Simpson's formula

(52.23), we have, with the help of (52.21),

(v) Yc(O.4) = 1.2242077 + 031 [1.2642077 + 4(1.4495757) + 1.6754735]

= 1.5154738.
A third application of Simpson's formula will not change the value m
(v). By (a) and (v),
(w) yc'(O.4) = 1.6754738.

In Table 52.24, we start to record the Milne values, keeping a careful

watch on the column headed D, which by (51.21), is the difference between

Table 52.24

yp of Yc of D=
x
(52.22) (52.23)
YP' yo'
first Yc - YP
ID/291 <
0.4 1.5154625 1.5154735 1.6754625 1.6754735 0.0000110 0.0000004
1.5154738 1.6754738
0.5 1.6961508 1.6961631 1.9461508 1.9461631 0.0000123 0.0000005
1.6961635 1.9461635
0.55 1.7972722 1.7972764 2.0997722 2.0997764 0.0000042 0.0000002
1.7972765 2.0997765
0.6 1.9063602 1.9063573 2.2663602 2.2663573 -0.0000029 0.0000001

the predictor value and the first corrector value. And, by (51.35), the
error in the firo;t corrector value for one step is approximately -D/29.
Since, for y(O.4), ID/291 is less than the desired error of 0.00005, we con-
tinue with h = 0.1.
Lesson 52D SUMMARY AND AN ILLUSTRATIVE EXAMPLE 699

By (52.22) with h = 0.1, Xo = 0.1,

(x) yp(0.5) = y(O.I) + 034 [2y'(0.2) - y'(0.3) + 2y'(0.4)]

= 1.1055125 + 034 [2(1.2642077) - 1.4495757 + 2(1.6754738)]
= 1.6961508.
By (a) and (x)
(aa) yP'(0.5) = 1.9461508.

Therefore by (52.23) with h = 0.1, Xo = 0.1,

(bb) yc(0.5) = y(0.3) + 031 [y'(0.3) + 4y'(0.4) + yP'(0.5)]

1.3595757 + 0 1 [1.4495757 + 4(1.6754738) + 1.9461508]
=
3
= 1.6961631.
A second application of Simpson's corrector will change the value in (bb) to
(cc) yc(0.5) = 1.6961635.

A third application of (52.22) will not change the value in (cc). With
this value of yc(0.5), we find from (a), yc'(0.5) = 1.9461635. Although
ID/291 is still much less than desired accuracy, and we may safely continue
to use h = 0.1, we shall calculate y(0.6) by reducing h to 0.05 in order to
demonstrate how to use formula (52.15).
With h now equal to 0.05, we cannot use the predictor formula (52.22)
to find yp(0.55) unless we know four preceding values of y, 0.05 unit
apart, namely y(0.5), y(0.45), y(O.4) , and y(0.35). We already know
y(0.5) and y(0.4). To find y(0.45) and y(0.35), we make use of formula
(52.15). With h = 0.1, the formula becomes
(dd) y(0.45)
= y(0.5 - 0.05)
= ts[5y(0.5) + 15y(0.4) - 5y(0.3) + y(0.2)]
= ts[5(1.6961635) + 15(1.5154738) - 5(1.3595757) + 1.2242077]
= 1.6024534,
y(0.35)
= y(O.4 - 0.05)
= ts[5y(0.4) + 15y(0.3) - 5y(0.2) + y(O.I)]
= fs[5(1.5154738) + 15(1.3595757) - 5(1.2242077) + 1.1055125]
= 1.4347174.
700 NUMERICAL METHODS Chapter 10

By (a) and (dd),

(ee) y'(0045) = (0045)2 + 1.6024534 = 1.8049534,
We now have the needed preliminary values to use the predictor formula
(52.22) to estimate yp(0.55). With h = 0.05, Xo = 0.35, it becomes
(ff) yp(0.55)
= + 0.20)
yp(0.35

= y(0.35) + 4(0 05) [2y'(004) - y'(0045) + 2y'(0.5)]

3
= 1.4347174 + 032 [2(1.6754738) - 1.8049534 + 2(1.9461635)]
= 1.7972721.

By (a) and (ff)

(gg) YP'(0.55) = (0.55)2 + 1.7972721 = 2.0997721.
The corrector formula (52.23), with h = 0.05, Xo = 0.35, now becomes
(hh) yc(0.55)
= yc(0.35 + 0.20)
= y(0045) + [y'(0045) + 4y'(0.5) + yP'(0.55)]
= 1.6024534 + [1.8049534 + 4(1.9461635) + 2.0997721]
= 1.7972764.
A second application of the corrector formula changes the value in (hh)
to
(ii) yc(0.55) = 1.7972765.

With this value of yc(0.55), we find from (a), yc'(0.55) = 2.0997765.

Returning to the predictor formula (52.22) with h = 0.05, Xo = 004,
(jj) yp(0.6)
= yp(004 + 0.2)
= y(0.4) + 4(0 05) [2y'(0045) - y'(0.5) + 2y'(0.55)]
3
= 1.5154738 + 032 [2(1.8049534) - 1.9461635 + 2(2.0997765)]
= 1.9063602.
By (a) and (jj)
(kk) y/(0.6) = (0.6)2 + 1.9063602 = 2.2663602.
Lesson 52-Exercise 701

By (52.23) with h = 0.05, Xo = 0.4,

(11) Yc(0.6) = + 0.2)
y(O.4

= y(0.5) + [y'(0.5) + 4y'(0.55) + yp'(0.6)]

= 1.6961635 + [1.9461635 + 4(2.0997765) + 2.2663602]

= 1.9063573.
A second application of formula (52.23) does not change the value in (ll).
Hence, by (a),

(mm) yo' (0.6) = 2.2663573.

As a final check on the overall accuracy of our result, we make use of

Weddle's corrector formula (50.64). With h = 0.1, Xo = 0, it becomes
(we use the subscript w for Weddle)

(nn) Y1O(0.6) = yeO) + + 5y'(0.1) + y'(0.2) + 6y'(0.3)

[y'(O)
+ y'(O.4) + 5y'(0.5) + y'(0.6)]
= 1 + 0.03[1 + 5(1.1155125) + 1.2642077 + 6(1.4495757)
+ 1.6754738 + 5(1.9461635) + 2.2663573]
= 1.9063562,
which is a closer approximation to the true value of y(0.6) than is the one
in (11). Since the difference between Yc(0.6) and Y1O(0.6) is not significant,
we assume their common value 1.90636 rounded off to five decimals is
accurate to five places and its error is therefore < 0.000005. [The actual
value of y(0.6) is 1.9063564 so that the error in the value of y(0.6) in (nn)
is 0.0000002. Rounded off to five decimal places, the actual value of
y(0.6) is 1.90636, a figure which agrees with our final result rounded off
to five decimal places.]

EXERCISE 52
1. In Example 52.2, we found y(O.l) = 1.1055125 and y(0.05 + 0.05) =
1.1055128. Use formula (46.57) to correct y(O.l). Using this corrected figure of
y(O.l), compute by series methods y(0.2) and y(O.l +
0.1), i.e., compute the
value of y(0.2) in one step and in two steps. The value of y(0.2) = 1.2242000,
computed in one step can be found in (n) of this example. Use formula (46.57),
with Xo = 0, h = 0.2, to correct y(0.2). Starting with this corrected value
of y(0.2), switch to the Runge-Kutta method, fourth order form, and com-
pute y(O.2 +0.1) and y[(0.2 +
0.05) +
0.05], i.e., compute y(0.3) in one step
and in two steps. Apply formula (47042) to correct y(0.3). Replace (52.21)
by these new corrected figures just obtained. Use Milne's method to find
yp(004), yc(004), D(004). Apply formula (51.35) to correct yc(004) and add this
702 NUMERICAL METHODS Chapter 10

corrected figure to your table. Compute yp(0.5), y.(0.5), D(0.5). Correct

y.(0.5) by (51.35). Do the same for yc(0.6). Finally apply Weddle's rule
(50.64) to compute y",(0.6). How many decimal place accuracy do you now
have? Compare all results with previous figures and with actual values.
2. Use the method of this lesson to find an approximate value when x = 0.6 of
the particular solution of y' = x + y for which y(O) = 1. Assume you know
nothing about the solution and that you wish the error to be less than 0.00005.
Use Taylor series for the first two approximations, Runge-Kutta for the third
approximation. Then switch to Milne's method. In the course of your com-
putations reduce h by one-half even though it may not be essential. Use
formula (50.63) to check the accuracy of y(0.5) and Weddle's rule (50.64)
to check the accuracy of y(0.6). In the absence of a solution, how many
decimal place accuracy could you assume in y(0.5), in y(0.6)? See 3 after
"3. Cumulative Errors," Lesson 52A. Solve the equation and compare your
results with actual value of y(0.5) and y(0.6).
3. Follow the instructions in 2 to find an approximate value when x = 1.6,
of the particular solution of y' = xy for which y(l) = 1.

ANSWERS 52
1. y(O.l) = 1.1055128, y(0.2) = 1.2242085, y(0.3) = 1.3595771,
yp(0.4) = 1.5154628, yp(0.5) = 1.6961512, yp(0.6) = 1.9063429,
y.(OA) = 1.5154745, y.(0.5) = 1.6961646, yc(0.6) = 1.9063565,
D(OA) = 0.0000117, D(0.5) = 0.0000134, D(0.6) = 0.0000136,
y(OA) = 1.5154741, y(0.5) = 1.6961641, y(0.6) = 1.9063560,
y",(0.6) = 1.9063566. Can assume five decimal place accuracy. Actual value:
y(0.6) = 1.9063564.
2. y = 2e z - x - I . Actual values: y(0.5) = 1.7974425,
y(0.6) = 2.0442376.
3. y = e(z2- 1 )/2. Actual value: y(1.6) = 2.1814723.

LESSON 53. Numerical Methods Applied to a System

of Two First Order Equations.
We consider a system of two first order equations
dx
(53.1) dt = f(t,x,y),
dy
dt = g(t,x,y) ,
for which
(53.11) y(t o) = Yo·
In Lesson 39, Example 39.17, we showed how to find a series solution of
a system such as (53.1). Hence we can use this method to obtain starting
approximations, provided the derivatives can be obtained without exces-
sive difficulty. If they cannot, we can use the following Runge-Kutta
fourth order formulas.
(53.12) x(t o + h) = x(t o) + i(Vl + 2V2 + 2va + V4),
y(to + h) = y(t o) + t(Wl + 2W2 + 2wa + W4),
Lesson 53 SYSTEM OF Two FIRST ORDER EQUATIONS 703

where
(53.13) VI = hf(to,xo,yo) ,
WI = hg(to,xo,yo) ,
V2 = hf(to + !h, Xo + !VI, Yo + !WI),
W2 = hg(to + !h, Xo + !VI, Yo + !wd,
Va = hf(t o + !h, Xo + !V2, Yo + !W2),
Wa = hg(to + !h, Xo + !V2, Yo + !W2),
v, = hf(to + h, Xo + Va, Yo + wa),
W4 = hg(to + h, Xo + Va, Yo + wa).
The Milne predictor and corrector formulas for the system (53.1) are
4h
(53.14) Xp(to + 4h) = x(t o) +g[2x'(t o +h) - x'(to + 2h) + 2x'(to + 3h)],
4h
yp(t o + 4h) = y(t o) + g [2y'(t o + h) - y'(to + 2h) + 2y'(to + 3h)],

(53.15) xc(t o + 4h)

= x(to + 2h) + [x' (to + 2h) + 4x'(to + 3h) + xP'(to + 4h)],
yc(to + 4h)
= y(to + 2h) + [x' (to + 2h) + 4y'(t o + 3h) + YP'(to + 4h)].
All comments and formulas in regard to errors, made in connection
with the numerical solution of the equation y' = f(x,y), apply to each
function making up the solution of a system. The fifth degree formula
(50.63) or Weddle's formula (50.64) can be used to check the overall
accuracy of the values of x(t) and of yet).
E:cample 53.16. Find approximate values when x = 0.4 and y = 0.4
of a particular solution of the system
(a) x'(t) = ty,
y'(t) = xy,
for which x(O) = 1 and yeO) = 1.
Solution. To bring into the discussion all the above formulas, we be-
gin by using Taylor series. * The series solution of (a) has already been
·By (f) of Example 39.17, the interval of convergence is ItI < 0.0625. We, therefore,
cannot use Taylor series to find x(O.1), y(O.I), x(0.2), y(0.2) unless we can establish a
larger interval of convergence. The chances are that the series solution does actually
have such a larger interval since the theorem gives only a minimum interval. But
unless we can prove both series converge for t = 0.1, t = 0.2, we would need to use
Runge-Kutta or some other starting method in place of Taylor series. We have used
Taylor series only for illustrative purposes.
704 NUMERICAL METHODS Chapter 10

found in Example 39.17. It is [see (k)]

t2 t3 t4
(b) x(t) = 1 + 2" + '3 + 8" + ... ,
2 3 4
yet) = 1+t + 2"t + '3t + 7t24 + ....
Hence, by (b), with t = 0.1 and 0.2 respectively, we obtain

(c) x(O.l) = 1 + (0;)2 + (0;)3 + (0;)4 + ... = 1.0053,

y(O.l) = 1 + 0.1 + (0;)2 + (0;)3 + + ... = 1.1054,

x(0.2) = 1 + (0;)2 + (0;)3 + (0:)4 + .. , = 1.0229,

y(0.2) = 1 + (0.2) + (0;)2 + (0;)3 + + ... = 1.2231,

Remark. We have used direct substitution to find x(0.2) and y(0.2).

It would have been more accurate, but have involved much more labor,
if we had used the creeping up method.

To find x(0.3) and y(0.3), we switch to the Runge-Kutta method.

With to = 0.2, h = 0.1, x(0.2) = 1.0229, y(0.2) = 1.2231, (53.12)
becomes
(d) x(0.3) = 1.0229 + l(vi + 2V2 + 2V3 + V4),
y(0.3) = 1.2231 + l(Wl + 2W2 + 2W3 + W4).

With h = 0.1, to = 0.2, Xo = x(t o) = x(0.2) = 1.0229, Yo = y(to) =

y(0.2) = 1.2231, f(t,x,y) = ty and g(t,x,y) = xy, (53.13) becomes

(e) VI = 0.lf(0.2, 1.0229, 1.2231) = 0.1 (0.2)(1.2231) = 0.02446,

WI = 0.lg(0.2, 1.0229, 1.2231) = 0.1(1.0229)(1.2231) = 0.1251,
V2 = 0.lf(0.25, 1.0229 + 0.0122, 1.2231 + 0.0626)
= 0.1(0.25)(1.2857) = 0.0321,
W2 = 0.lg(0.25, 1.0229 + 0.0122, 1.2231 + 0.0626)
= 0.1(1.0351)(1.2857) = 0.1331,
V3 = 0.lf(0.25, 1.0229 + 0.0161, 1.2231 + 0.0666)
= 0.1(0.25)(1.2897) = 0.0322,

W3 = 0.lg(0.25, 1.0229 + 0.0161, 1.2231 + 0.0666)

= 0.1(1.0390)(1.2897) = 0.1340,
Lesson 53 SYSTEM OF Two FIRST ORDER EQUATIONS 705

V4 = 0.11(0.3,1.0229 + 0.0322, 1.2231 + 0.1340)

=.0.1(0.3)(1.3571) '= 0.0407,
W4 = 0.lg(0.3, 1.0229 + 0.0322, 1.2231 + 0.1340)
= 0.1(1.0551)(1.3571) = 0.1432.

Hence (d) becomes

(f) x(0.3) = 1.0229 + t[O.0245 + 2(0.0321) + 2(0.0322) + 0.0407]

= 1.0552,
y(0.3) = 1.2231 + t[O.1251 + 2(0.1331) + 2(0.1340) + 0.1432]
= 1.3568.

Since we now have the needed number of preliminary estimates, we switch

to Milne's method to find x(004) and y(O.4). Before we can use formulas
(53.14) and (53.15), however, we must know x'(O.I), x'(0.2), x'(0.3) and
y'(O.I), y'(0.2), y'(0.3). We obtain these values by use of (a) and the
values of x(O.I), x(0.2), x(0.3), y(O.I), y(0.2), y(0.3) as found in (c) and
(f) above. Therefore

(g) x'(O.l) = O.ly(O.I) = (0.1)(1.1054) = 0.1105,

x'(0.2) = 0.2y(0.2) = (0.2)(1.2231) = 0.2446,
x'(0.3) = 0.3y(0.3) = (0.3)(1.3568) = 0.4070,
y'(O.l) = x(O.l)y(O.l) = (1.0053)(1.1054) = 1.1113,
y'(0.2) = x(0.2)y(0.2) = (1.0229)(1.2231) = 1.2511,
y'(0.3) = x(0.3)y(0.3) = (1.0552)(1.3568) = 1.4317.

Hence with to = 0, h = 0.1, and with the initial conditions x(O) = 1,

yeO) = 1, we can write (53.14) as

(h) xp(004) = 1 + 034 [2x'(0.1) - x'(0.2) + 2x'(0.3)]

= 1 + 034 [2(0.1105) - 0.2446 + 2(004070)] = 1.1054,

yp(004) = 1 + 034 [2y'(0.1) - y'(0.2) + 2y'(0.3)]

= 1 + 034 [2(1.1113) - 1.2511 + 2(1.4317)] = 1.5113.

By (a) and (h), we obtain

(i) xP'(Oo4) = Oo4y(004) = 004(1.5113) = 0.6045,

yP'(Oo4) = x(004)y(Oo4) = (1.1054)(1.5113) = 1.6706.
706 NUMERICAL METHODS Chapter 10

The corrector formulas (53.15) are therefore

(j) x c(0.4) = x(0.2) + °31 [x'(0.2) + 4x'(0.3) + xp'(OA)]

1.0229 + 031 [0.2446 + 4(004070) + 0.6045]
1.1055,
yc(OA) = y(0.2) + °31 [y'(0.2) + 4y'(0.3) + Yp'(OA)]
1.2231 + °31 [1.2511 + 4(1.4317) + 1.6706]
1.5114.

EXERCISE 53
1. Using the values of x(O.4) and y(O.4), given in (j) of Example 53.16, as new
predictor values, use corrector formulas (53.15) to see if they will correct these
results. If they do, repeat the procrss until two successive values of x(O.4)
and y(O.4) agree. Then compute x(0.5) and y(0.5) by means of the predictor
formulas and the repeated use of the corrector formulas. With five values of
x(t) known in addition to the initial condition, use corrector formula (50.63)
to check the accuracy of x(0.5). Do the same for y(0.5). How many deci-
mal place accuracy can you assume in x(0.5), y(0.5)? Hint. See 3 after
"3. Cumulative Error8," Lesson 52A.
2. Find approximate values when x = 0.1, 0.2"", 0.5, Y = 0.1, 0.2,"',
0.5 of the particular solution of the first order linear system
x'(t) = x - y, y'(t) = -4x + y,
for which x(O) = 1, yeO) = 1. Take h = 0.1. Use Taylor series to the fourth
order, direct substitution method, to calculate x(O.I), y(O.I), x(0.2), y(0.2);
Runge-Kutta method to calculate x(0.3), y(0.3); Milne's method to calculate
x(O.4), y(O.4) , x(0.5), y(0.5). Finally apply corrector formula (50.63) to
compute x(0.5), y.(0.5). How many place accuracy can you assume in x(0.5),
y(0.5)? Now solve the system and compare results.
3. In problem 2, compute x(0.2), y(0.2) by the creeping up method. Then apply
(46.57) to correct x(0.2), y(0.2). With these corrected values of x(0.2), y(0.2),
+
use Runge-Kutta method to compute x[(0.2+ 0.05) 0.05], x(0.2+ 0.1). Do
the same for y(0.3). Correct x(0.3), y(0.3) by means of (47.42). Compute
x p (O.4), x c(O.4), D(O.4). Correct x c(O.4) by means of (51.35). Do the same for
Yc(O.4). Compute x p (0.5), x c(0.5), D(0.5). Correct Do the same for
yc(0.5). Finally apply corrector formula (50.63) to compute x(0.5), y(0.5).
How many decimal place accuracy can you assume in your results? Compare
with results obtained in 2 and with actual solution.

ANSWERS 53
1. x(O.4) = 1.1055, y(O.4) = 1.5114; x(0.5) = 1.1776, y(0.5) = 1.6938.
Corrector (50.63): x(0.5) = 1.1776, y(0.5) = 1.6939. Can assume four and
three decimal place accuracy.
Lesson 54 SECOND ORDER EQUATION 707

2. x(t) values: 1.01609,1.06940, 1.17026, 1.33246, 1.57490. yet) values: 0.68234,

0.31740, -0.11807, -0.65395, -1.33021. Corrector values (50.63): x(0.5) =
1.57504, y(0.5) = -1.33050. Can assume three and two decimal place
accuracy respectively, if we round off to this number of decimals.
Solution: x = He 31 +3e- I ), y = !(-e31 3e- I ). +
Actual values: x(0.5) = 1.5753203, y(0.5) = -1.3310485.
3. x(0.2) = 1.06958, y(0.2) = 0.31705;
x(0.3) = 1.17051, y(0.3) = -0.11855;
x (0.4) = 1.33274, y(O.4) = -0.65450;
x(0.5) = 1.57525, y(0.5) = -1.33091.
Corrector values (50.63): x(0.5) = 1.57530, y(0.5) = -1.33101.

LESSON 54. Numerical Solution of a Second Order

Differential Equation.
In the proof of Theorem 62.22, we show how a second order differential
equation y" = f(x,y,y') can be reduced to a system of two first order
equations. A numerical solution of this equation can therefore be found
by the method of Lesson 53. However, it is also possible to find a numerical
solution of such an equation without the necessity of reducing it to a
system. We illustrate the method by an example.
Example 54.1. Find an approximate value when x = 004 of a par-
ticular solution of the equation

(a) y" = 2x + 2y - y',

for which y(O) = 1, y'(O) = 1.

Solution. In order to bring into the discussion Taylor series, Runge-
Kutta and Milne methods, we shall find y(O.l) and y(0.2) by Taylor series
method, y(0.3) by Runge-Kutta's method and y(004) by Milne's method.
By Theorem 37.2, if

then

From (a), and by differentiation of (a), we obtain

(b) y" = 2x + 2y - y', y'" = 2 + 2y' - y", y(4) = 2y" _ y"'.

By the initial conditions, x = 0, y = 1, y' = 1. Substituting these

values in (b), there results
(c) y"(O) = 1, y'''(O) = 2 +2 - 1 = 3, y(4)(0) = :t - 3 = -1.
708 NUMERICAL METHODS Chapter 10

Substituting the initial conditions and (c) in (54.11) it becomes, with

Xo = 0,
h2 h3 h4
(d) y(h) = 1 + h + -2 + -2 - -24 + ... .

As in the first order case, two methods are available to us to find y(O.I)
and y(0.2). By direct substitution, we obtain from (d),
(e) y(O.I) = 1 + 0.1 + !(0.1)2 + !(0.1)3 - -h(0.1)4 = 1.105496,
y(0.2) = 1 + 0.2 + !(0.2)2 + !(0.2)3 - -h(0.2)4 = 1.223933.

The creeping up method involves much more arithmetic, but insures

greater accuracy. By our basic equations (54.11) and (54.12), with h = 0.1
and Xo equal successively 0.0, 0.1, we find

(f) yeO + 0.1) = y(O.l)

= yeO) + y'(O)(O.l) + (0.1)2

+ (0.1)3 + (0.1)4 + .. "

y'(O + 0.1) = y'(O.l)
= y'(O) + y"(O)(O.l) + (0.1)2

+ (0.1)3 + ... ,
y(O.l + 0.1) = y(0.2)
= y(O.I) + y'(O.l)(O.I) + (0.1)2

+ (0.1)3 + (0.1)4 + ... ,

y'(O.1 + 0.1) = y'(O.2)
= y'(O.l) + y"(O.I)(O.l) + (0.1)2

+ (0.1)3 + ....

Substituting the initial conditions and (c) in the first two lines of (f), we
obtain

(g) y(O.I) = 1+ (1)(0.1) + !(0.I)2 + i(0.1)3 - -h(0.1)4 = 1.105496,

y'(O.I) = 1+ (1)(0.1) + - t(0.1)3 = 1.114833.

Therefore with x = 0.1, y(O.I) = 1.105496, y'(O.I) = 1.114833, we find

Lesson 54 SECOND ORDER EQUATION 709

by (b),

(h) y"(O.l) = 2(0.1) + 2(1.105496) - 1.114833 = 1.296159,

y"'(O.l) = 2 + 2(1.114833) - 1.296159 = 2.933507,
y(4l(0.1) = 2(1.296159} - 2.933507 = -0.341189.

Substituting (g) and (h) in the third and fourth lines of (f), there results

(i) y(0.2) = 1.105496 + (1.114833)(0.1) + 1.29;159 (0.1)2

+ 2.93:507 (0.1)3 _ (0.1)4 = 1.223948,

y'(0.2) = 1.114833 + (1.296159)(0.1) + 2.93;507 (0.1)2

_ (0.1)3 = 1.259060.

We now switch to the Runge-Kutta method to find y(0.3). The fourth

order form for the second order equation y" = f(x,y,y'5 for which y(xo) =
Yo, y'(xo) = YI is

(54.13) y(xo + h) = y(xo) + l(Vl + 2V2 + 2V3 + V4),

y'(xo + h) = y'(xo) + l(WI + 2W2 + 2wa + W4),
where, with Yo' = y'(xo),
(54.14) VI = hyo',
WI = h/(xo,Yo,Yo'),
V2 = h(yo'+ !WI),
W2 = hf(xo + !h, Yo + !v!, Yo' + !WI),
V3 = h(yo' + !W2),
W3 = hf(xo + !h, Yo + !V2, Yo' + !w2),
V4 = h(yo' + wa),
W4 = hf(xo + h, Yo + V3, Yo' + W3).

Therefore with Xo = 0.2, h = 0.1, y(;to) = y(0.2) = 1.223948, y'(0.2) =

1.259060, (54.13) becomes

(j) y(0.3) = y(0.2 + 0.1) = 1.223948 + l(vi + 2V2 + 2V3 + V4),

y'(0.3) = y'(0.2 + 0.1) = 1.259060 + i(Wl + 2W2 + 2W3 + W4).

With h = 0.1, Yo' = y'(xo) = y'(0.2) = 1.259060, Yo = y(xo) = y(0.2) =

710 NUMERICAL METHODS Chapter 10

1.223948, and l(x,Y,y') = 2x + 2y - y', (54.14) becomes

(k) VI = 0.1(1.259060) = 0.125906,

WI = 0.11(0.2, 1.223948, 1.259060)
= + 2(1.223948) - 1.259060] = 0.158836,
0.1[2(0.2)
V2 0.1(1.259060 + 0.0794418) = 0.1338502,
=
W2 = 0.11(0.25, 1.223948 + 0.062953, 1.259060 + 0.079418)
= 0.1(0.5 + 2.573802 - 1.338502) = 0.173530,
Va = 0.1(1.259060 + 0.086765) = 0.1345825,
Wa = 0.11(0.25, 1.223948 + 0.066925, 1.259060 + 0.086765)
= 0.1(0.5 + 2.581746 - 1.345825) = 0.173592,
V4 = 0.1(1.259060 + 0.173592) = 0.1432652,

W4 = 0.lj(0.3, 1.223948 + 0.1345825, 1.259060 + 0.173592)

= 0.1(0.6 + 2.717061 - 1.432652) = 0.188441.
Substituting (k) in (j), there results
(1) y(0.3) = + -1(0.125906 + 0.267700 + 0.269165 + 0.143265)
1.223948
= + t(0.806036) = 1.358287,
1.223948
y'(0.3) = 1.259060 + -1(0.158836 + 0.347060 + 0.347184 + 0.188441)
= 1.259060 + ,(1.041521) = 1.432647.
We summarize the results thus far obtained
(m) yeO) = 1.000000, y'(O) = 1.000000;
y(O.I) = 1.105496, y'(O.l) = 1.114833;
y(0.2) = 1.223948, y'(0.2) = 1.259060;
y(0.3) = 1.358287, y'(0.3) = 1.432647.

The Milne predictor formulas for the second order equation y" =
= Yo, y'(xo) = YI are
I(x,y,y') for which y(xo)

(54.15) (a) YP'(xo+ 4h) = y'(xo)

+ [2y"(xo + h) - y"(xo + 2h) + 2y"(xo + 3h)],
(b) yp(xo + 4h) = y(xo + 2h)

+ 3h [Y'(xo + 2h) + 4y'(xo + 3h) + YP'(xo + 4h)].

The corrector formulas are
(54.16) (a) yc'(xo + 4h) = y'(xo + 2h)
i
+ [Y"(xo + 2h) + 4y"(xo + 3h) + YP"(xo + 4h)],
Lelson 54 SECOND ORDER EQUATION 711

(b) Yc(xo + 4h) = + 2h)

y(xo
h
+"3 [y'(xo + 2h) + 4y'(xo + 3h) + yc'(xo + 4h)].
As in the first order case, formulas (54.16) can be used again and again
until two successive values of Yt(xo + 4h) agree.
By (54.15) and (54.16), with Xo = 0, h = 0.1,

(n) Yp'(Oo4) = y'(O) + 034 [2y"(0.1) - y"(0.2) + 2y"(0.3)],

Yp(004) = y(0.2) + 031 [y'(0.2) + 4y'(0.3) + YP'(Oo4)],
yc'(Oo4) = y'(0.2) + 031 [y"(0.2) + 4y"(0.3) + yp"(Oo4)],
yc(004) = y(0.2) + 031 [y'(0.2) + 4y'(0.3) + yc'(Oo4)].
Before we can use these formulas, we must know the value of y" when
x = 0.1,0.2, 0.3. By (a) and (m), these needed values are
(0) y"(O.I) = 2(0.1) + 2(1.105496) - 1.114833 = 1.296159,
y"(0.2) = 2(0.2) + 2(1.223948) - 1.259060 = 1.588836,
y"(0.3) = 2(0.3) + 2(1.358287) - 1.432647 = 1.883927.
Substituting in (n), the initial conditions and the values in (m) and (0),
we obtain
(p) Yp'(Oo4) = 1 + 034 [2(1.296159) - 1.588836 + 2(1.883927)]
= 1.636178,
yp(004) = 1.223948 + 031 [1.259060 + 4(1.432647) + 1.636178]
= 1.511476.
To use the third formula in (n), we need to know yp"(Oo4). With x = 004
and yp(004), yp'(Oo4) having the values in (p), we obtain from (a),
(q) yp"(Oo4) = 2(004) + 2(1.511476) - 1.636178 = 2.186774.

Hence the last two formulas in (n) become

(r) y.'(Oo4) = 1.259060 + 031 [1.588836 + 4(1.883927) + 2.186774]

= 1.636104,
yc(004) = 1.223948 + 031 [1.259060 + 4(1.432647) + 1.636104]
= 1.511473.
712 NUMERICAL METHODS Chapter 10

EXERCISE 54
1. Using the value of y(OA), given in (r) of Example 54.1, as a new predictor
value, apply corrector formulas (54.16) repeatedly until two successive values
of y(OA) agree. Then find y(0.5) by means of the predictor formulas and re-
peated use of the corrector formulas. Do the same for y(0.6). Finally apply
Weddle's rule (50.64) to compute y .. (0.6). How many decimal place accuracy
can you assume in y(0.6)? See 3 after "3. Cumulative Errors," Lesson 52A.
Solve (a) of Example 54.1 and compare your results with actual values.
2. Apply the method of this lesson to find an approximate value when x = 0.5
of a particular solution of the equation y" = x 2y + +
y' for which yeO) = 1,
y'(O) = 1. Apply corrector formula (50.63) to evaluate y(0.5). How many
decimal place accuracy can you assume? Solve the equation and compare
your results with actual values.
3. The Adams method, see Exercise 50,6, can also be used as a continuing formula
for finding a numerical solution of a second order equation. The needed
formulas are
(54.2) (a) y(xo + h) = y(xo)
+ h[y'(xo) + !Vy'(xo) + "hV2y'(xo) + iV3y'(xo)].
(b) y'(xo + h) = y'(xo)
+ h[y"(xo) + !Vy"(xo) + -hV2y"(xo) + iV3y"(xo)].
Use these formulas to find approximate values of y(OA) and y'(OA) of the
particular solution of the equation y" = 2x +
2y - y' for which yeO) = 1,
y'(O) = 1. Hint. With Xo = 0.3, h = 0.1, you must know y'(O), y'(O.l),
y'(0.2), y'(0.3) j y"(O), y"(O.I), y"(0.2), y"(0.3) in order to set up the needed
tables of differences. You will find these values in Example 54.1.
4. Use Milne's method to find an approximate value of y(OA) of the particular
solution of the equation y'" = y" +
xy' +
2y for which yeO) = y'(O)
y" (0) = 1. The necessary Milne's formulas are

yP"(OA) = y"(O) + 034 [2y"'(0.1) - y"'(0.2) + 2y'''(0.3)],

yp'(OA) = y'(0.2) + 031 [y"(0.2) + 4y"(0.3) + yP"(OA)],
yc(OA) = y(0.2) + 031 [y'(0.2) + 4y'(0.3) + YP'(0.4)].
Obtain the needed preliminary values by means of Taylor series, direct
substitution.

ANSWERS 54
1. y(OA) = 1.511473, y(0.5) = 1.686538, y(0.6) = 1.886643. y,.(0.6) = 1.886640.
Can assume five decimal place accuracy. Solution: y = iez - ie-2z - X - !.
Actual value: y(0.6) = 1.886666.

2. y = 43e2z - 2x + 4·
1
Actual value: y(0.5) = 2.038711.
3. y(OA) 1.51145, y'(OA) = 1.63612.
4. y(0.4) = 1.520.
Lesson 55 PERTURBATION METHOD. FIRST ORDER EQUATION. 713

LESSON 55. Perturbation Method. First Order Equation.

In physical problems, we frequently encounter a differential equation,
as for example, the differential equation
(55.1) y' + y2 = 0, y(1) = 1,
which has been disturbed by a small effect, so that (55.1) has to be modi-
fied to read
(55.2) y' + y2 = EX, y(l) = 1,
where E is small. It then becomes necessary to determine by how much
the solution of (55.1) has been altered because of the presence of the dis-
turbing function EX. We refer to this change in the solution as a per-
turbation.
A precise perturbation theory is extremely difficult. In this lesson, we
shall aim to give only a rough outline of a method by which this problem
can be handled. Call Yo(x) a solution of (55.1) satisfying y(l) = 1, and
denote the solution of (55.2) by
(55.3) y(X) = Yo(x) + p(x)
where p(x) is the perturbation. We next expand y(x) in a series in powers
of E, so that
(55.4) y(x) = Yo(x) + EYI (x) + E2Y2(X) + E3Y3(X) + ....
Comparing (55.3) with (55.4), we see that
(55.5) p(x) = EYI(X) + E2Y2(X) + E3Y3(X) + .. '.
The first term EYI (x) is called the first order perturbation; the second
term E2Y2(X) is called the second order perturbation, etc.
Substituting (55.4) in (55.2), we obtain
(55.6) Yo' + EYI' + E2Y2' + E3y3' + ...
+ (Yo + EYI + E2Y2 + E3 Y3 + ...)2 = EX.

Carrying out the indicated multiplication, then collecting coefficients of

like powers of E, we have
(55.7) (Yo' + Y02) + (YI' + 2YOYl)E
+ (Y2' + 2YoY2 + Y12)E2 + (.... ')E 3 + ... = EX.

Next we equate like powers of E. There results

(55.8) Yo'+ Y02 = 0,
YI' + 2YOYl = X,
2
Y2' + 2YOY2 + Y1 = 0,
714 NUMERICAL METHODS Chapter 10

By solving each equation of (55.8) in succession, we can thus determine

the functions Yl (x), Y2(X), •.• in (55.4). Each of these functions, how-
ever, must satisfy an initial condition. Since the initial condition asso-
ciated with the original equation (55.1) is y(l) = 1, and since Yo is a
solution of (55.1) so that Yo(1) = 1, this initial condition will be satisfied
if, in (55.4), we assume

(55.81) Yo(l) = 1, Y2(1) = 0, .. '.

We illustrate the details of the above method by solving Example 55.9

below. In practice the first and second order perturbation terms of (55.5)
are usually sufficient.

Example 55.9. Find the first and second order perturbation terms in
the solution of
(a) y' + y2 = 0, for which y(l) = 1,

due to the presence of a disturbing function EX, where E is small.

Solution. Because of the disturbing function EX, (a) must be modified

to read
(b) Y' + y2 = EX,

for which y(l) = 1. Following the procedure outlined above we let, see
(55.4) and (55.81),

(c)

with initial conditions

(d) Yo(1) = 1, Yl(l) = 0, Y2(l) = 0, ....

Substituting (c) in (b), we obtain, see (55.7),

(e) (Yo' + Y02) + (Yl' + 2YOYl)E + (Y2' + 2YOY2 + Y12)E2 + ... = EX.

Equating coefficients of like powers of EO, E, E2, we obtain from (e) the
system of equations

(f) Yo' + Y02 = 0, Yl' + 2YOYI = X,

A solution of the first equation of (f), satisfying the initial condition

Yo(l) = 1 of (d), is
1
(g) Yo = x'
Lesson 56 PERTURBATION METHOD. SECOND ORDER EQUATIONS. 715

Substituting (g) in the second equation of (f), we obtain

(h) Yl ,+2
-XY1 = X.

A solution of (h) satisfying Yl(l) = 0 of (d) is

(i) Yl = !4 (X2 - X2 1-) .

Substituting (g) and (i) in the third equation of (f), we obtain

(j) Y2 ,2
+- x Y2 = - -16 1(4 x - 2 + -1) .
X4

A solution of (j) satisfying Y2(1) = 0 of (d), is

(k) Y2 = - - 2; - ;3) - 2{x 2.

Substituting (g), (i), (k) in (c), we obtain

(1) Y = -x1 + -4E ( x 2 - -1)

x2
- - E2 ( 3x 5 -
336
14x - -213
x
+ -32)
x2
.
The solution of (a) satisfying y(l) = 1, i.e., its solution if there were no
disturbing function EX present, is 1/x. Because of the disturbing function
EX, the first and second order perturbation terms are, respectively, the
second and third terms in (1).

EXERCISE 55
1. Find the first and second order perturbation terms in the solution of y' +
y2 = 0, for which y(l) 1, due to the presence of a disturbing function EX 2,
where E is small.

ANSWERS 55

1. First order: i (X3 - :i) ; second order: - (2x7 - 9X2 - +

LESSON 56. Perturbation Method. Second Order Equation.
In Lesson 55, we outlined a method of determining the perturbation of
a solution of a first order differential equation due to a small disturbance.
We shall now apply this method to determine the perturbation of a solu-
tion of a second order equation. Consider the differential equation

(56.01) y" +y = 0,
716 NUMERICAL METHODS Chapter 10

for which
(56.02) yeO) = 0, y'(O) =' 1.

Note that (56.01) is the differential equation of simple harmonic motion.

Let the disturbing function be -2E(y,)2, where E is small. Therefore
(56.01) becomes
(56.1) Y" +Y = -2E(y,)2
for which
(56.11) yeO) = 0, y'(O) = 1.

We wish to find the first and second order perturbation terms of a solution
of (56.01) satisfying (56.02) resulting from the presence of a disturbing
function -2E(y')2.
Call Yo(x) a solution of (56.1) satisfying (56.11). Let
(56.12)
therefore

In order that (56.12) may satisfy the initial conditions (56.02), we assume
(56.13) Yo(O) = 0, Yl(O) = 0, Y2(0) = 0, ... .
Yo'(O) = 1, Yl'(O) = 0, Y2'(0) = 0, ... .
Substituting (56.12) in (56.1), we obtain, using only terms to E2,

(56.14) Yo" + EYl" + E2Y2" + Yo + EYI + E2Y2

= -2E[(Yo,)2 + E2(Yl,)2 + E'(Y2,)2 + 2EYo'Yl'
+ 2E 2Yo'Y2' + 2E 3YI'Y2').
Collecting coefficients of like powers of E, we have

(56.15) (Yo" + Yo) + (YI': + YI)E + (Y2" + Y2)E 2

= -2(Yo,)2E - 4YO'YI'E 2.
Equating coefficients of like powers of EO, E, E2, we obtain from (56.15),

(56.16) Yo" + Yo = 0, YI" + YI = -2(Yo,)2, Y2" + Y2 = -4Yo'YI'.

A solution of the first equation of (56.16) satisfying Yo(O) = 0, Yo'(O)
= 1 is
(56.2) Yo = sin x, Yo' = cosx.
Substituting (56.2) in the second equation of (56.16), we obtain
Lesson 56 PERTURBATION METHOD. SECOND ORDER EQUATIONS. 717

(56.21) Yl" + Yl = -2 cos 2 x.

A general solution of (56.21) is

(56.22) Yl = Cl sin x + C2 cos X - t sin 2 x - i cos 2 x,

Yl' = Cl cos X - C2 sin x - t sin x cos x.

A particular solution of (56.22) satisfying Yl(O) = 0, Yl'(O) = 0 is

(56.23) Yl =i cos x - t sin 2 x - i cos 2 x,

Yl' = -isinx - Isinxcosx + tsinxcosx
= -i sin x - t sin x cos x.
Substituting in the third equation of (56.16), the values of Yo' and Yl' as
given in (56.2) and (56.23), we obtain

(56.24) Y2" + Y2 = -4( -i sin x cos x - t sin x cos 2 x)

= I sin x cos x + ¥ sin x cos 2 x
+
= I sin x cos x ¥ sin x - ¥ sin a x.
The complementary function of (56.24) is

(56.25) Yc = Cl sin x + C2 cos X.

A particular solution
(56.251) of Y2" + Y2 = I sin x cos x, is Y2 = -J sin x cos x,
of Y2" + Y2 = ¥ sin x, is Y2 = -Ix cos x,
of Y2" + Y2 = ¥ sin a x, by Example 21.32,
is Y2 = t sin 3x·- 2x cos x.

Hence a general solution of (56.24), by (56.25) and (56.251), is

(56.26) Y2 = Cl sin x + C2 cos X - J sin x cos x - ix cos x - t sin 3x,

Y2' = Cl cos X - C2 sin x + J(sin 2 x - cos 2 x)
+ i(x sin x - cos x) - ! cos 3x.

By (56.13), Y2(0) = 0, Y2'(0) = O. Inserting these values in (56.26),

we find
(56.27) Cl = J+i + ! = H-
Therefore by (56.26) and (56.27), a particular solution of (56.24) satisfying
Y2(0) = 0, Y2'(0) = 0 is

(56.28) Y2 = Hsinx - Jsinxcosx - ix cos x - tsin3x.

718 NUMERICAL METHODS Chapter 10

Substituting, in (56.12), the values of Yo, Yl, and Y2 as found in (56.2),

(56.23), and (56.28), we obtain
(56.29) Y = sin x + E(i cos x - t sin 2 x
i cos 2 x)
-
+ 2
E (Hsinx - Jsinxcosx - ix cos x -lsin3x).
The solution of (56.01) satisfying (56.11), i.e., its solution if there were no
disturbing function present, is sin x. Because of the disturbance function
-2E(y')2, the first and second order perturbation tenus are, respectively,
the second and third tenus in (56.29).

EXERCISE 56
1. Find the first and second order perturbation terms in the solution of V" -
V = 0 for which v(O) = 0, V' (0) = 2 due to the presence of a disturbing
function EY/, where E is small.

ANSWERS 56
.
1. Flrst order: E
('"2 - -")
ze ze
-2- ;

2
[e
secondorder:E 8(-I+z+z)+8(I+z-z)'
2 e 2 ]
 Chapter II

Existence and Uniqueness Theorem

for the First Order Differential
Equation y' = /(x,y). Picard's Method.
Envelopes. Clairaut Equation.

Introductory Remarks. As we have repeatedly emphasized, differ-

ential equations whose solutions can be expressed explicitly or implicitly
in terms of elementary functions are relatively few in number. Even a
first order differential equation

(57.1) y' = f(x,y)

will usually not have an elementary solution. In these cases, it is desirable

to have theorems which will answer the following questions for us.
1. Does (57.1) have a I-parameter family of solutions? See Examples
4.21 and 4.22 for differential equations which have no solutions; Exam-
ple 4.2 for one which has only one solution; (4.652) or Example 5.3
for one which has two I-parameter family of solutions.
2. If (57.1) has a I-parameter family of solutions, is it a general solution
as we defined this term in Definition 4.7, i.e., does it contain every
particular solution? See the first example in Lesson 4C for a I-para-
meter family which does not contain every particular solution.
3. Is there a particular solution of (57.1) valid on some interval and satis-
fying a given initial condition y(xo) = Yo?
4. Is a particular solution satisfying an initial condition y(xo) = Yo
unique? See (b) of Example 5.3 where the point (0,1) lies on an infinite
number of particular solutions.
Fortunately there are theorems which will give us, under appropriate
hypotheses, the answers to these questions. A theorem which answers
questions 1, 2, and 3, i.e., one which tells us whether a solution exists is
719
720 EXISTENCE THEOREM: y' = f(x,y). PICARD. CUlHAUT. Chapter 11

called an existence theorem.. A theorem which answers question 4, i.e.,

one which tells us whether a solution is unique is called a uniqueness
theorem.
It should be emphasized that an existence and uniqueness theorem only
guarantees or assures the existence and uniqueness of a solution. It will
not tell you whether the solution can or cannot be expressed in terms of
elementary functions, or help you to find the solution. For example, the
particular solution of the differential equation
(57.11)

for which y(xo)

(57.12)
= Yo is

y(X) = Yo + 'i"
"'0
",2
e- dx.

It has been proved that this integral cannot be expressed in terms of ele-
mentary functions. However, the existence and uniqueness theorem,
which we shall state later, will not help you to discover this fact. All the
theorem will tell you, is that since the function e-",2 and the point (xo,Yo)
satisfy its hypotheses, an unique particular solution satisfying (57.11) and
the initial condition exists.
There are several ways of proving the existence and uniqueness theorem
for the first order differential equation y' = f(x,y), satisfying the condi-
tion y(xo) = Yo. The one we shall use is dependent on a method which is
known as Picard's approximation method, named after the French
mathematician, Charles Picard (1856-1941). Hence before we can
prove the theorem, we shall first need to explain Picard's method.

LESSON 57. Picard's Method of Successive Approximations.

We assume for the moment that a unique particular solution of the

differential equation
(57.2) y' = f(x,y),
satisfying the initial condition

(57.21) y(xo) = Yo,

exists. Let y(x) be the required particular solution. Then by (57.2),
y' = f[x,y(x)], where f[x,y(x)] is now a function only of x. Integrating
this equation between the limits Xo and x and noting, by (57.21), that
when x = Xo, Y = Yo, we obtain

(57.22)
J("lIo dy = j",o
('" f[t,y(t)] dt, y(x) = Yo + j",o
f'" j[t,y(t)] dt.
Lesson 57 PICARD'S METHOD 721

In employing numerical methods to approximate the particular solution

y(x) of (57.22), we used a polynomial interpolating function in place of the
integrand,J[x,y(x)]. In Picard's method we also use the idea of an approxi-
mation but of a totally different kind. In this method, we obtain a sequence
of functions Yo(x), Yl (x), ... , Yn(X), each of which satisfies the initial con-
dition (57.21). The existence and uniqueness Theorem 58.5 which follows
will then give the conditions that f(x,y) of (57.2) must fulfill in order
that an interval about Xo exist, on which, as n 00, this sequence of
functions approach the particular solution y(x) of (57.22). Furthermore,
each function in the sequence is an approximation of the particular solu-
tion y(x): a later one, in general, being a better approximation than a
preceding one. Hence the name successive approximations.
We illustrate the method by means of examples. We concentrate for the
moment only on the mechanics of the method without considering the size
of the interval about Xo for which the sequence of functions thus obtained
converges to the particular solution y(x) of (57.22). (For the meaning of
the convergence of a sequence of functions, see Definition 58.1 and
Example 58.13.)
The first approximation of a solution of (57.2) satisfying (57.21) is called
Yo(x). The function Yo(x) may, as we shall show later, be any arbitrary
continuous function defined in a neighborhood of Xo. In the absence of
additional information, it is usually taken to be the constant function
(57.23) Yo(x) = Yo,

where Yo is the initial value given in (57.21). It is evident that this ap-
proximation to the solution is not a very satisfactory one. It is the equa-
tion of a straight line parallel to the x axis and Yo units from it.
The subsequent members of the sequence of approximating solutions of
(57.2) satisfying (57.21) are called Yl (x), Y2(X), ..• , Yn(X), ••. , and are
obtained in the following manner.

(57.24) Yl(X) = Yo + 1'"

"'0
f[x,yo(X)] dx,

Y2(X) = Yo + 1'"
"'0
f[X,Yl(X)] dx,

Ya(X) = Yo + 1'"
"'0
f[X'Y2(X)] dx,

Yn(x) = Yo + 1'" f[X,Yn-l(X)] dx,

"',.
where Xo and Yo are given in (57.21). (Remember f[x,yo(x)] means replace
yin f(x,y) by Yo (x) ; f[X,Yl(X)] means replace Y in f(x,y) by Yl (x), etc.)
722 EXISTENCE THEOREM: y' = f(x,y). PICARD. CLAIRAUT. Chapter 11

Comment 57.241. For different permissible starting approximations

Yo(x), different sequences Yo(x), Yl (x), .•• , Yn(X) will result. However,
each will have the property that for an x in an interval about xo,
lim Yn(X) = y(X) ,
n-->oo

where y(x) is the solution of (57.2) satisfying (57.21). The rapidity with
which the sequence of approximations will converge to the solution y(x)
will depend on how closely the starting solution Yo(x) approximates the
actual solution y(x); the closer the approximation, the quicker the con-
vergence.
Example 57.25. Find the first four Picard approximations if
(a) y' = xy,
and yeO) = 1.
Solution. Comparing (a) and the initial condition with (57.2) and
(57.21), we see that !(x,y) = xy, Xo = 0, Yo = 1. By (57.23), our first
approximation is therefore Yo(x) = 1. The succeeding approximations,
by (57.24), are

(b) Yl(X) = 1+ i" !(x,Yo) dx = 1+ i" x dx = 1 + ;2,

Y2(X) = + i"
1 !(X,Yl) dx

= 1+ i" x (1 + dx1+ ;2 + ,
=

Ya(x) = 1+ i" !(X,Y2) dx = 1+ i" (1 + ;2 + x dx,

Comment 57.251. If we solve (a) by the method of separation of

variables, we obtain the particular solution.y = e"Z/2 whose series expan-
sion is

Note that each succeeding function in the sequence Yo, Yl, Y2, Ya, ..• is a
closer approximation to the actual solution than is the previous one.
Example 57.26. Find the first three Picard approximations if
(a) y' = x2 - Y
and y(l) = 2.
Lesson 57 PICARD'S METHOD 723

Solution. Comparing (a) and the initial condition with (57.2) and
(57.21), we see that f(x,y) = x 2 - y, Xo = 1, Yo = 2. By (57.23) our
first approximation is therefore Yo(x) = 2. The succeeding approximations
are, by (57.24),
(b) YI(X) = 2 + f(x,Yo) dx

= 2 + f '"
I
2
(x - 2) dx
X3
= "3 - 2x
11
+ 3'

Y2(X) = 2 + f(X,YI) dx

= 2 + f"'(
I X2 - "3
X3
+ 2x - 11) dx,
3"
x3 X4 2 11 53
= "3 - 12 +x - 3" x + 12 .
Picard's Method Applied to a Systelll of Two First Order Equa-
tions. Picard's method of successive approximations can also be applied
to a system of first order equations. We illustrate the method for the pair
of first order equations
dx dy
(57.3) dt = ft(t,x,y), dt = !2(t,x,y),
satisfying the initial conditions.

(57.31) y(to) = Yo.

The first approximations of the solution of the system (57.3) satisfying
(57.31) are called xo(t) and yo(t). In the absence of additional information,
they are usually taken to be the constant functions
(57.32) yo(t) = Yo,
where Xo and Yo are given in (57.31). The subsequent approximations are

(57.33) 1. Xl (t) = Xo t ft[t,xo(t),yo(t)] dt,

+ Jlo

YI (t) = Yo t !2[t,xo(t),yo(t)] dt.

+ Jlo
724 EXISTENCE THEOREM: y' = f(x,y). PICARD. CLAIRAUT. Chapter II

n. Xn(t) = Xo t h[t,Xn_l (t),Yn-l (t)] dt,

+ JtO

Yn(t) = Yo t !2[t,Xn-l(t),Yn-l(t)] dt.

+ JtO
Picard's method thus yields two sequences of functions, xo, Xl, ••• ,
x n, ... , and Yo, Yt, ... , Yn, ... , each satisfying the appropriate initial
condition in (57.31). The existence Theorem 62.12 which follows, will then
give the conditions which hand h of (57.3) must fulfill in order that an
interval I about to exist, on which, as n -+ 00, the first sequence approach
a limiting function x(t) and the second sequence approach a limiting func-
tion y(t). This pair of functions is, on I, the unique solution of (57.3)
satisfying (57.31). For different permissible starting approximations, dif-
ferent sequences will result but each will have the property that on I,
lim Xn(t) = x(t) and lim Yn(t) = y(t).
n--+oo

The rapidity with which each sequence will converge to its limiting func-
tion will depend on how close the starting approximations are to the
actual solutions.
E:cample 57.34. Find the first three Picard approximations if the first
order system is
dx dy
(a) dt = t + X, dt = t - X,

and x(O) = 1, y(O) = -1.

Solution. Comparing (a) and the jnitial conditions with (57.3) and
(57.31), we see that 11 (t,x,y) = t + X, !2(t,x,y) = t - X, to = 0, Xo = 1,
Yo = -1. By (57.32) our first approximations are xo(t) = 1, yo(t) = -1.
By (57.33), the succeeding approximations are

(b) 1. Xl(t) = 1+ it h(t,xo,yo) dt = 1+ i' (t + 1) dt

t2
= 1 +"2 + t,

Yl(t) = -1 + it !2(t,xo,Yo) dt = -1 + fot (t - 1) dt

2
-1 +"2t -to
Lesson 57 PICARD'S METHOD 725

2. X2(t) = 1+ fot It(t,X}'Yl) dt = 1+ L (t + 1 + + t) dt

2 t3
= 1 + t + t + 3!'

Y2(t) = -1 + fot h(t,Xl,yd dt

= -1 + fot (t - 1- - t) dt

t3
= -1 - t - 3r'

Example 57.35. Find the first four Picard approximations if the first
order system is
dx dy
(a) dt = ty, dt = xy,
and x(O) = 1, y(O) = 1.
Solution. Comparing (a) and the initial conditions with (57.3) and
(57.31), we see that It (t,x,y) = ty,j2(t,x,y) = xy, to = 0, Xo = 1, Yo = 1.
By (57.32), our first approximations are xo(t) = 1, yo(t) = 1. By (57.33),
the succeeding approximations are

(b) 1. Xl(t) = 1+ fot It (t,xo,yo) dt = 1+ fot t dt = 1+

Yl (t) = 1+ fot h(t,xo,yo) dt = 1+ fot dt = 1 + t.

2. X2(t) = 1+ fot It (t,X},Yl) dt = 1+ fot (t + t 2) dt

t2 t3
= 1+2"+3'

Y2(t) = 1+ fot h(t,X}'Yl) dt = 1+ fot (1 + t+ i dt

t2 t3 t4
= l+t+2"+6'+8"
726 EXISTENCE THEOREM: y' = f(x,y). PICARD. CLAIRAUT. Chapter 11

EXERCISE 57

Find the first k Picard approximations after yo(x), of the particular

solution of each of the following equations 1-5, where k is the number
shown alongside each equation.
1. y' = x - y, y(O) = 1, k = 4.
2. y' +
= x 2 y, y(l) = 3, k = 3.
3. y' = x +
y2, y(O) = 0, k = 3.
4. y' = 1 +
xy, y(l) = 2, k = 3.
5. y' = e" +
y, y(O) = 0, k = 4.

Find the first k Picard approximations after xo(t), yo(t), of the particular
solution of each of the following systems, where k is the number shown
alongside each system.
dx
6. dt = t + y, dydt = t -
2
x , x(O) = 2, y(O) = 1, k = 3.

dx
7. dt = t + y 2, dydt = x - t, x(O) = 0, y(O) = 1, k = 3.

ax dy ,
8. dt = xt, dt = x - e ,x(O) = 1, y(O) = -1, k = 3.

9. Picard's method of successive approximations can also be applied to a system

of first order equations greater than two. For the system of three first order
equations: dx/dt = !I(t,x,y,z), dy/dt = /2(t,x,y,z), dz/dt = h(t,x,y,z) for
which x(to) = Xo, y(t(j) = Yo, z(to) = Zo, the successive approximations are

(57.4) O. xo(t) = Xo, yo(t) = Yo, zo(t) = 21 0.

1. Xl (t) = Xo + ft !I(t,xo,YO,Zo) dt,

10

Yl (t) = Yo + f' h(t,xo,yo,zo) dt,

'0
,
211 (t) = Zo + fto h(t,xo,YO,zo) dt.
1

2. X2(t) = Xo +f !I (t,xl,YlZl,) dt,

'0
,
Y2(t) = Yo + f'0
h(t,Xl,Yl,Zl) dt,
Lesson 57-Exereise 727

Z2(t) = Zo +f 10
I

!a(t,Xl,YI, Zl) dt.

+f
1

n. x,,(t) = xo /!(t,Xn-l,Y,,-l,Zn-l) dt,

10

Use (57.4) to find the Picard approximations X3, Y3, Z3 of the particular
solution of the system
dx 2 dy dz
dt =Y , dt = x + Z, dt = Z - Y,

for which x(O) = 1, y(O) = 0, z(O) = 1.

ANSWERS 57

2 5 S 11
X X X x
3. Y3 = "2 + 20 + 160 + 4400 .
7 5 2 i i x6
4. Y3 = 20 + x + 24 x +"3 + 15 + 24'
3
z X 2
5. Y4 = 4e - 6" - x - 3x - 4.

6. X3 = 2 +t - it 2 - it 3 - !t4 - ic;t 5 - Tht 6,

Y3 = 1 - 4t - it 2 + + it 3 nt4 - 5 Ut +
nt6 - mt7.
2 4 3 4 6
t t l t t t
7. X3 = t + "2 + 12 + 252 ' Y3 = 1 + 3' - 12 + 120 .
2 4
t t t
6
t3 t
5
t
7
1
8. X3 = 1 + "2 + 8' + 48 ' Y3 = t + "6 + 40 + 252 - e.
3 4 5 2 3 4 2
4t t t t3
9. X3 = 1 + 3' + "2 + 20' Y3 = 2t + "2t - "6t + '3'
t
Z3 = 1
t
+ t - "2 - 3' .
728 ExISTENCE THEOREM: y' = f(x,y). PICARD. CUIRAUT. Chapter 11

LESSON 58. An Existence and Uniqueness Theorem for the

First Order Differential Equation y' = f(x,y)
Satisfying y(xo) = Yo.
For a clearer understanding of the proof of the existence and uniqueness
Theorem 58.5 which follows, it will be necessary to know the meaning of
the convergence of a sequence of functions, the meaning of the uniform
convergence of a sequence of functions, the meaning of a Lipschitz condi-
tion, and to be acquainted with certain theorems from analysis. Hence,
before beginning the proof, we shall briefly discuss these topics and list the
needed theorems.

LESSON 58A. Convergence and Uniform Convergence of a Se-

quence of Functions. Definition of a Continuous Function.
Definition 58.1. A sequence of functions
(58.11) !I (x), f2(X), .•• ,fn(x), ••• ,
each defined on a common set S is said to converge to a function f(x)
on S, if for each x in S and for each fixed E > 0, no matter how small,
there is an N such that
(58.12) Ifn(x) - f(x)1 < E, when n > N.
In words the definition says the following. Pick any x you wish in this
set S and choose any positive number E as small as you like. For this x
calculate !I (x), f2(X), fa(x), .•• and f(x). If the sequence of functions
!I (x), !2(x), ••• ,fn(x), ..• converges to f(x), then according to the defini-
tion you must eventually reach a function fN+l(X) in the sequence, such
that IfN+l(X) - f(x)1 is less than this chosen positive E. Further every
function in the sequence after fN +1 (x) must also differ fromf(x) in absolute
value by an amount less than E.
Example 58.13. Show that the sequence of functions
1 1 1
!I (x) = 1 + x ' = + 2x ' fa(x) = -- , ...
(a) !2(x) 1 1 + 3x '
1
fn(x) = - - , ...
1 + nx '
converges to the function f(x) = 0 on I: 0 < x 1.
Solution. Let x be any number in the interval I: 0 < x 1, and
let E be any positive number. Here fn(x) of (58.12) is 1/(1 nx) and +
f(x) = O. Hence by Definition 58.1, we must show that there exists an
N such that
(b) 11 nx - 0 I < E, when n > N.
Lesson 58A CONVERGENCE. UNIFORM CONVERGENCE. CONTINUITY. 729

The inequality (b) is equivalent to

1 1
(c) -E < 1 + nx, E
l<nx , n >x! (E! - l ) '

If, therefore,
(d)

then (b) will hold when n > N. [NOTE. l/E - 1 is negative, take
N = 0.] For example, if we choose x = ! and E = 0.02, then by (d),
N = 2(50 - 1) = 98. It should therefore follow that the value of each
1 1
+
function after the 98th, namely,j99<!) = 1 99/2 ' hoo(x) = 1 50' ... +
should be less than E = 0.02. You can easily verify that each such func-
tion is indeed less than 0.02; remember, with a fixed numerator, the value
of a fraction decreases as the denominator increases. And if we choose
x = 1/100, then by (d), N = 100(50 - 1) = 4900. We would therefore
need to go to the 4901 th function in the sequence before coming to the first
one whose value differs from zero by less than E = 0.02. And if we pick
x = 1/1,000,000, then by (d), N = 1,000,000(50 - 1) = 49,000,000, i.e.,
we shall need to go to the 49,000,00Ist function in the sequence before
coming to the first one whose value differs from zero by less than 0.02.
It is evident from (d) and the above examples, that for each fixed posi-
tive E, and for each different x, a different N will be required in the sequence
to insure the validity of (58.12), i.e., N depends on both the values of
x and E. This dependency of N on both x and E is common to a great many
sequences. On the other hand, there are certain sequences where N does
not depend on x but only on the value of E, i.e., for a fixed E > 0, no matter
how small, it may be possible to find an N such that (58.12) will be valid
for every x in I when n > N. In other words, we do not, in this case, need
to hunt for a different N for each different x. We find an N once and for
all which will hold for every x in I. In this event, we say the sequence of
functions (58.11) converges uniformly on I to the functionf(x).
Definition 58.14. A sequence of functions (58.11), each defined on a
common set S is said to converge uniformly on S to a function f(x), if
for a fixed positive E, no matter how small, there is an N such that
(58.15) Ifn(x) - f(x) I < E when n > N,

for every x in S.
We emphasize once more, that here fora fixed E > 0, we pick an N only
once. For each x in S, the absolute value of the difference of each of the
functions fN +1 (x), fN +2(X), ... and f(x) will be less than E.
730 EXISTENCE THEOREM: y' = J(X,y). PICARD. CLAIRAUT. Chapter 11

Example 58.16. Show that the sequence of functions

1 1
(a) fr(x) = 1 !x' h(x) = 1 + 2x' fa(x) = - - , ...
1 +
3x '
1
fn(x) = 1 + nx' ... ,

converges uniformly to the function f(x) = 0 on I: 1 x.

Solution. Note that this sequence of functions is the same as that of
Example 58.13, but that the interval is different. Here we must show, by
Definition 58.14, that for a given E > 0,

(b) 11 : nx - 0 I < E, when n > N,

for every x in I: 1 x < 00. By (c) of the previous example,

(c)

For a fixed E > 0 and for an x in I: 1 x, the expression on the right of

(c) will have its largest value whcn x = 1. And when x = 1, (c) becomes
1
n > - - 1. If therefore
E

(d)

then (b) will hold for each x in I when n > N. For example, if E = 0.02,
then by (d), N = 49. You can verify that each functional value in the
1 1
+
sequence after the 49th, namely, f50(1) = 1 50' f51(1) = 1 51' ... +
is less than 0.02. Hence for x > 1, each functional value in the sequence
after the 49th is surely less than 0.02. Our given sequence of functions,
therefore, converges uniformly on I: 1 x < 00 to the function fix) = O.
Definition 58.17. A functionf(x) is continuous at a point x = a
if f( a) exists and if
(58.171) limf(x) = f(o,).

Definition 58.172. A function f(x) is continuous on, or in, an

interval I if it is continuous at every point of I.
Definition 58.18. A functionf(x,y) is continuous at a point (a,b),
if f(a,b) exists and
(58.181) lim f(x,y) = f(a,b).
x.lI---ta,b
Lesson 58B LIPSCHITZ CONDITION. THEOREMS FROM ANALYSIS. 731

Definition 58.19. A function f(x,y) is continuous on, or in, a

region S if it is continuous at every point of S.

LESSON 58B. Lipschitz Condition. Theorems from Analysis.

(The theorems from analysis which we shall need have been stated without
proof. Their proofs can be found in advanced calculus text books.)
Definition 58.2. If f(x,y) is a function of x and y in a region S such
that, for every two p )ints (x,y) and (x,Y) in S,
(58.21) If(x,y) - f(x,Y)1 Nly - yl.
where N is a positive constant, then f(x,y) is said to satisfy a Lipschitz
condition in S (see Fig. 58.57).
Theorem 58.22. Law of the mean. See Fig. 58.57. If f(x,y) is a
function of x and y that has a continuous partial derivative with respect to y
in a region S, then for each x there exists a number Y such that

(58.23) f(x,y) - f(x,y) = f(x Y)

y - y ay"
where (x,y) and (x,ll) are any two points in Sand Y lies between y and y.

Remark. The notation .!...- f(x, Y) means the value of the partial deriv-
ay
ative of the function f(x,y) with respect to y when y = Y.
Comment 58.24. If f(x,y) has a continuous partial derivative with
respect to y in a region_S.Jtnd if this partial derivative as a function of the
two variables X,y is bounded in S, then f(x,y) satisfies a Lipschitz condi-
tion. The proof proceeds as follows. Since af(x,y)/ay is bounded in S,
there exists a constant N such that

(a) 1 f(x,y) 1 N,
for every point (x,y) in S. Let (x,y) and (x,y) be any two points in S.
Then by Theorem 58.22 there exists a number Y between y and 'ii, such that

(b) f(x,y) - f(x,Y) 1= 1 f(x Y) I·

I y-y ay'
Since (x, Y) is a point in S, we can substitute the inequality (a) for the right
side of (b) and thus obtain (58.21).
Theorem 58.25. If f(x) is a Riemann-integrable function on I: a
x b, then

(58.26)
732 EXISTENCE THEOREM: y' = f(x,y). PICARD. CLAIRAUT. Chapter 11

Theorem 58.3. If f(x) is a continuous function on the interval I:

a x b and if
(58.31) F(x) = L" f(t) dt, a x b,

then F(x) is continuous on I, and

(58.32) F'(x) = f(x), a < x < b.
Theorem 58.4. If a sequence of continuous functionsh (x), hex), ... ,
fn(x), .... each defined on a common interval I, converges uniformly on
I, to a function f(x), thenf(x) is continuous on I.
Comment 58.41. The preceding theorem is not true if, on I, the
sequence of functions merely converges to f(x), but not uniformly. For
instance, the sequence of continuous functions of Example 58.13 converges
on I: 0 x 1, but not uniformly. The sequence converges on 0 <
x 1 to the function f(x) = O. But when x = 0, each function in the
sequence has the value one. Hence on the interval I: 0 x 1, the
limiting function of the sequence is

(a) f(x) = {O, 0 < x 1,

1, x = O.

The function f(x) is therefore discontinuous on 1.

Theorem 58.42. If a sequence of continuous functions h (x), f2(X), ... ,
f .. (x), ... , each defined on a common interval I, converges uniformly on
I to a function f(x) , then

(58.43) lim
n-+QO Jt" fn(x) dx = Jt" lim fn(x) dx = Jt" f(x) dx,
zo Zo %0

where the interval (xo,x) is contained in I.

Definition 58.44. A series of functions

hex) + hex) + ... + fn(x) + .. "

each defined on a common interval I, is said to converge uniformly on I
to a function f(x) , if the sequence of partial sums FI(x), F 2(x), ... , Fn(x),
where
Fn(x) = hex) + ... + fn(x) ,
converges uniformly on I to f(x).
Theorem 58.45. If each function fleX), hex), ... ,fn(x) is defined and
bounded on a common interval I, i.e., if
(58.46) If;(x) I M i, i= 1,2,"',n,"',
Lesson S8C PROOF OF EXISTENCE THEOREM: y' = f(x,y) 733

and if the infinite series of positive terms

(58.47) Ml + M2 + ... + Mn + '"

converges, then the series

(58.48) hex) + !2(x) + ... + fn(x) + ...

converges uniformly on I to a function f(x).

Comment 58.481. By the above theorem, if each If;(x) I ;;;i! M i , and

converges, then for a given E > 0, there is a positive N such that

[hex) + !2(x) + ... + fn(x) - f(x) I < E, when n > N,

for every x in I.
E:cample 58.49. Show that the series of functions,
QO 1
(a) hex) = 2i + x' I: 0 ;;;i! x ;;;i! 1,
" t=l

converges uniformly on I.
Solution. Take Mi = 1/2i. Therefore
1 1 1
= M 1 + M 2 + M 3 + ... = 2 + 22 + 23 + ....
QO

(b) Mi
i=l

By (a), and for each x such that 0 ;;;i! x ;;;i! 1,

(c)

The last series on the right of (c) is a geometric series which converges to
one. Hence the first series of functions, which is the given series (a), by
Theorem 58.45, converges uniformly on I to a function f(x).

LESSON S8C. Proof of the Existence and Uniqueness Theorem

for the First Order Differential Equation y' = f(:c,y). In Theorem
38.14, we gave a sufficient condition for the existence of a power series
solution of y' = f(x,y) satisfying an initial condition y(xo) = Yo. Com-
pare it with the theorem we now state, which gives a sufficient condition
for the existence and uniqueness of a solution of y' = f(x,y) for which
y(xo) = Yo·
734 EXISTENCE THEOREM: y' = f(x,y). PICARD. CLAIRAUT. Chapter 11

Theorem 58.5. See Fig. 58.57. Let f(x,y) be a bounded, continuous

function of x and y in a region S of the xy plane and let (xo,Yo) be a point
of S. In S, let the function f satisfy the Lipschitz condition (58.21), namely
(58.51) If(x,y) - f(x,1]) I Nly - 1]1,
for every two points (x,y) and (x,1]) in S.
Then an interval
(58.52) 10: Ix - xol < h, 'h > 0,
exists on which there is one and only one continuous function y(x), with a
continuous derivative on 10 , satisfying the differential equation
(58.53) y' = f(x,y)
and the initial condition
(58.54) y(xo) = yo·
Remark. The above theorem is a special case of the more general
Theorem 62.12 on systems of first order equations.
Proof. The proof of the theorem will be based on Picard's methods of
successive approximations. By (57.23) and (57.24), these approximating
functions are
(58.55) Yo (x) = Yo, '"
Yl(X) = Yo + j",o
( f[t,yo(t)) dt,
Y2(X) = Yo + j",o
('" f[t,Yl(t)) dt,

Yn(X) = Yo + jxo
('" f[t,Y,,-l (t)) dt,
where f(x,y) is the continuous function of (58.53), Yo is the constant of
(58.54) and the interval (xo,x) is contained in S. Because the proof is
long, we have divided it into four parts. NOTE. We shall use I: Ix - xol
h to denote the closed interval (58.52).
A. First we shall show how to determine the interval 10 of (58.52).
B. Second we shall prove that on I each of the functions Yo(x), Yl (x),
... , Yn(X) of (58.55) is continuous and its graph lies in a rectangle R
contained in the given region S.
C. Third we shall prove that this sequence of functions Yo, Yl, •.• , Yn of
(58.55) converges uniformly on I to a function y(x) which, on 10 , is a
solution of (58.53) satisfying (58.54). We thus establish the existence
of a particular solution y(x).
Lesson sse PROOF OF EXISTENCE THEOREM: y' = f(x,y) 735

D. Finally we shall prove that this solution y(x) is, on 1 0 , the unique par-
ticular solution of (58.53) satisfying (58.54).
A. Determining an interval I. Since the given function f(x,y) is, by
hypothesis, bounded in S, there exists a positive constant M such that
(58.56) If(x,y) I < M,
for every point (x,y) in S. The point (xo,Yo) of (58.54) is by hypothesis a
point of a region S. (See Fig. 58.57 and keep referring to it.) Hence we

(xoY)

I(x,y) - (x,.Y)1

rT1Y-YI
1: Ix-xol
I' x

Figure 58.57

can find a positive number h such that the rectangle of dimensions

Ix - xol h, Iy - Yol < Mh which has this point (xo,Yo) at its center,
and where M is the constant in (58.56), lies entirely in S. Call this rectangle
R. Therefore all points in R (since they are also in S) whose x coordinates
are in the interval
(58.58) I: Xo - h x Xo + h, Ix - xol h,
satisf:' (58.51) and (58.56). This interval I, without its endpoints, is the
interval 10 of (58.52) referred to in the theorem.
Proof of B. Each function Yo(x), Yl (x), ... , Yn(X) of (58.55) is, on the
interval I of (58.58), continuous and its graph lies in R. Consider the first
736 EXISTENCE THEOREM: y' = f(x,y). PICARD. CLAIRAUT. Chapter II

function Yo(x) = Yo. Its graph is a straight line parallel to the x axis
and Yo units distant from it. Hence for all x in I, the graph of Yo(x) = Yo
lies in R (keep referring to Fig. 58.57). To complete the proof, we use the
inductive method of reasoning described in Lesson 24A. We assume the
graph of the function Yn-l(X) lies in R and must then show that the graph
of the function Yn(X) is also in R.
Hence we assume that for all x in Iof (58.58), the graph of Yn-l (x) lies
in R. This means we assume that for each x in I, Yn-l(X) will give a value
of Y which is in R. Therefore [X,Yn_l(X)] is a point of R. Hence by (58.56)
(a)
By the last equation of (58.55), by (58.26), (a) and (58.58), in the order
listed, we obtain

(b) Iy,,(x) - Yol = I

i>[t'Y"-l (t») dtl Ii: If[t,Yn_l (t»)1 dtl

< Mli: dtl = Mix - xol Mh.

Equation (b) says that for each x in I, the distance from Yn(x) to Yo <
Mh (see Fig. 58.57). Therefore the graph of Yn(x) lies in R.
We have thus proved that for each x in Iof (58.58), [X,Yk(X»), k = 0,
1, 2, ... , n, ... is a point of R contained in S. Hence, by the hypothesis
of the theorem, each integrand f[X,Yk(X»), k = 0, 1, 2,"', n,'" of
(58.55) is, in R, and therefore on I, a continuous function. It therefore
follows by Theorem 58.3, that each function Yl(X), "', Yn(x), ... of
(58.55) is also a continuous function on I. And sip.ce Yo(x) = Yo, a con-
stant, it too is a continuous function on I.
Proof of C. There exists a particular solution of (58.53) satisfying
(58.54). By the second equation in (58.55) and Theorem 58.25,

(c) IYl(X) - Yo(x)1 = li>[t,Yo(t») dtl Ii: If[t,Yo(t)]1 dtl·

In B, we proved that for each x in Iof (58.58), the point [x,Yo(x)] is in R.

Therefore by (58.56) and (58.58), (c) becomes

(d) IYl(X) - Yo(x)1 < Mli: dtl = Mix - xol Mh.

Subtracting the second equation of (58.55) from the third, we obtain

(e) IY2 - Yll = Ii: {f[t,Yl(t») - f[t,Yo(t)]} dtl

Ii: If[t,Yl (t») - f[t,Yo(t)] I dtl·

Lesson sac PROOF OF EXISTENCE THEOREM: y' = f(x,y) 737

For each x in I, both points [X,Yl(X)] and [x,Yo(x)] are, by B, in R.

Hence by (58.51), we can write (e) as

(f) IY2 - Yll Nli: IYl(t) - Yo(t) Idtl·

Replacing the integrand in (f) by its value Mit - xol as given in (d), we
obtain with the help of (58.58),

l2
(g) IY2 - Yll < MNli: It - xol dtl = MN Ix -;to MN .

Repeating the above procedure, we find, with the help of (58.55), (58.51),
(g), and (58.58), in the order listed,

(h) IY3 - Y21 If[t,Y2(t)] - f[t,Yl (t)]1 dtl

Yl(t)1 dtl
Nil: IY2(t) -
2 3
< MN211'" It - ,xol dtl = MN 2 Ix -,xol
"'0 2. 3.
3
< MN 2 h •
3!
And in general, it can be shown by the same type of inductive argument
we used in B, that for every x in I and every n,

(i) IYn(x) - Yn-l (x) I < MNn-1 Ix - xol"

n!
n
< MN n- 1 h = M (Nh)" .
n! N n!
By (i)
M Nh
(j) IYl - Yol < NIT'
M (Nh)2
IY2 - Yll < N

M (Nh)"
IYn - Yn-ll < N -n.
,-'

The sum of Yo and the positive terms on the right side of (j) is
738 EXISTENC}; THEOREM: y' = f(x,y). PICARD. CLAIRAUT. Chapter II

a series which, by (37.42), converges to Yo + (e Nh - 1). Hence, by

Theorem 58.45, the series

(k) Yo + (YI - Yo) + (Y2 - YI) + ... + (Yn - Yn-I),

which is the sum of Yo and the functions on the left side of (j), converges
uniformly on I to a function y(x). But the sum (k) is Yn. We have thus
proved that on the interval I of (58.58),
(1) lim Yn(x) = y(x)
n ...... oo

uniformly, i.e., the sequence of continuous functions Yo(x), YI (x), .•. ,

Yn(X), defined in (58.55), converges uniformly on I to a function y(x). By
Theorem 58.4, therefore, this function y(x) is continuous on I. Moreover,
it follows by Definition 58.14 since the graph of Yn(x) is in R, that the
graph of y(x) must also be in R.
Since for every x in I, [x,y(x») and [x,Yn(x») are points of R, we have,
by (58.51),

(m) If[x,Yn(x») - f[x,y(x»)1 NIYn(x) - y(x)l·

By (I), the sequence Yn(X) -+ y(x) uniformly on I. Therefore, by Defini-

tion 58.14, there exists an index P such that IYn(x) - y(x) I < fiN, when
n > P for every x in I. Hence, by (m),

If[x,Yn(X») - f[x,y(x)ll < f, when n > P,

for every x in I. Therefore, by Definition 58.14, the sequence of con-
tinuous functionsf[x,yo(x)}, f[X,YI(X»), ... ,f[x,Yn(x»), ... , converges uni-
formly on I to f[x,y(x»), i.e.,
lim f[x,Yn(x») = f[x,y(x»)
n-->oo

uniformly. Therefore, by Theorem 58.4, f[x,y(x») is a continuous function

on [.
Hence by (I), the last equation of (58.55), Theorem 58.42, and the equa-
tion immediately above, in the order listed,

(n) y(x) = lim Yn(x)

n---+oo
= Yo + lim r f[t,Yn-l(t») dt
n---+CIO }%O

= Yo + 1" lim f[t,Yn_l(t») dt

:eo n--Joo

= Yo r f[t,y(t») dt.
+ }"O
Lesson 58C PROOF OF EXISTENCE THEOREM: y' = !(x,Y) 739

We showed above thatf[x,y(x)] is a continuous function on I: Ix - xol h.

It therefore follows by (n) and Theorem 58.3, that
y'(x) = f[x,y(x)], 10: Ix - xol < h.

By (n), we see also that

1
"'0

Y(Xo) = Yo + "'0
f[t,y(t)] dt = Yo +0 = Yo·

We have thus proved the existence on 10 of a particular solution of

(58.53) satisfying (58.54). It is the limiting function y(x) of (1). We must
still prove that this particular solution y(x) is unique.
Proof of D. The function y(x) of (1) is the unique particular solution of
(58.53) satisfying (58.54) Assume g(x) is another particular solution of
(58.53) satisfying (58.54). Therefore we assume

(0) g'(x) = f[x,g(x)],

for which g(xo) = Yo. By integrating (0) and inserting the initial condi-
tions g(xo) = Yo, we obtain

g(x) = Yo + )"'0
['" f[t,g(t)] dt.
Subtracting (n) from the equation immediately above, we have

(p) Ig(x) - y(x) I = Ii: {f[t,g(t)] - f[t,y(t)]} dtl

Ii: If[t,g(t)] - f[t,y(t)]1 dtl·

We proved in B, that for every x in Iof (58.58), the graph of each function
Yo, Yl, .•. , Yn of (58.55) lies in a rectangle R contained in S. We proved
in C that this sequence of functions converges unifonnly on I, to a continu-
ous function y(x) whose graph also lies in R. Similarly, it can be shown
that g(x) is a continuous function on I whose graph lies in a rectangle
contained in S. Therefore for each x in I of (58.58), the points [x,y(x)]
and [x,g(x)] are in S. We may therefore apply (58.51) to the last expression
on the right side of (p). We thus obtain

(q) Ig(x) - y(x) I Nli: I(g(t) - y(t») I dtl·

For every x in I, the graphs of g(x) and y(x) are in bounded rectangles
contained in S. Call D the maximum value of Ig(x) - y(x)1 for x in I.
740 EXISTENCE THEOREM: y' = f(x,y). PICARD. CLAlRAUT. Chapter 11

Then by (q)
(r) Ig(x) - y(x) I DN 1.( dtl = DNlx - xol·
Substituting (r) for the integrand in (q), we obtain

(s) Ig(x) - y(x) I DN2Ii: It - xol dtl = 2

DN Ix -;tol2 .
Again substituting (s) for the integrand in (q), we obtain

(t) Ig(x) - y(x)1 DN 3 11'" It - 2.,XOI2 dtl

"'0
= DN3 IX - ,X
3.
OI3
•

Continuing in this manner, we find, with the help of (58.58),

(u) I (x) - (x) I DN" Ix - xol" DNnh" = D (Nh)" .

g y - n! - n! n!

In the last term of (u), D is a constant and (Nh)nln! is, by (37.42), the
general term of the series expansion of eN" which converges for all Nh.
Its nth term, therefore, approaches zero as n -+ 00. Hence D(Nh) " Inl
approaches zero as n approaches 00. Since Ig(x) - y(x) I can be made
less than any number no matter how small, it follows that

(v) g(x) - y(x) = 0, g(x) = y(x).

We have thus proved finally that the particular solution y(x) of (1) IS
unique. The assumed second solution g(x) is the same as y(x).
Comment 58.59. It is now evident why in Picard's method the first
starting approximation yo(x) need not be the line yo(x) = Yo, where Yo is
the initial condition given in (58.54). It can be any continuous function
through the point (xo,Yo) whose graph is in R. Each such starting function
will determine a sequence of functions whose limiting function is a par-
ticular solution of the given equation satisfying the given initial condition,
valid on the interval Iof (58.52). By the theorem there can be only one
such particular solution.
Comment 58.6. Let xo be a fixed value of x. Then each point (xo,c)
of S, where c is an arbitrary constant, determines a unique solution y(x)
of (58.53). Hence the solution y(x) is a function not only of x but also of
the ordinate c, i.e., y = Y(x,c). There is therefore a I-parameter family of
solutions of. (58.53) cutting across each line x = Xo.
Comment 58.61. Theorem 58.5 gives only a sufficient condition for
the existence and uniqueness of a particular solution of a differential equa-
tion y' = f(x,y) for which y(xo) = Yo. It is not a necessary condition.
Lesson sse PROOF OF EXISTENCE THEOREM: y' = f(x,y) 741

The sufficient condition means that if a region S contains only points which
fulfill (58.51) and (58.56), then the conclusion of the theorem must follow,
i.e., each point of S lies on one and only one particular solution of (58.53).
That the condition is not necessary implies its conclusion may still be true
for points in S which do not fulfill (58.51) and (58.56), i.e., these points
may still lie on one and only one particular solution.
Comment 58.62. Theorem 58.5 serves another useful purpose. By
Definition 5.4, an ordinary point of a first order equation y = f(x,y) lies
on one and only one integral curve. Hence every point (xo,Yo) of a region S
that fulfills (58.51) and (58.56) must be an ordinary point since, by the
theorem, each such point lies on one and only one integral curve. Assume
that S contains only ordinary points and that we are able to write a
I-parameter family of solutions of a differential equation explicitly or im-
plicitly in terms of elementary functions. If for each point (xo,Yo) of S,
the I-parameter family yields a particular solution through it-and by the
theorem there is only one such solution-then this family, by Definition 4.7,
must be a general solution since it contains every particular solution of the
differential equation. In this special case, therefore, the theorem makes
us aware whether our n-parameter family of solutions is a general one or
not.
E%ample 58.63. Obtain four Picard approximations if
(a) y' = 1 + y2
and yeO) = o. Find an interval for which the sequence of Picard approxi-
mations will converge to the actual solution.
Solution. Following the method outlined in Lesson 57B, we find
(b) Yo(x) = 0,

Yl(X) = L" dx = x,

Y2(X) = foX (1 + x 2) dx = x + ,
x

1
3
Y3(X) = ° 1 + (X3)2
x + "3 dx = x
x
+ "3 + 152 x 5 + 631 x 7 .
Heref(x,y) of Theorem 58.5 is 1 + y2 and 8f(x,y)
8y
= 2y. Therefore in any
bounded region S, no matter how large, f(x,y) is continuous and satisfies
(58.56). Its partial derivative is also continuous and bounded in S. Hence,
by Comment 58.24, f(x,y) satisfies (58.51). Let us take for S the region
-5 < y < 5 (see Fig. 58.64). Then the M of (58.56) is found as follows.
(c) If(x,y) I = 11 + y21 < 1 + 25 = 26 = M.
142 EXISTENCt; THEOREM: y' = f(x,y). PICARD. CLAIRAUT. Chapter 11

We must now choose h, according to A of the theorem, so that the rec-

tangle R of dimensions (here Xo = 0 and Yo = 0) Ixl h, IYI < 26h
which has the point (0,0) at its center lies in S. Remember S is now
bounded by the lines Y = 5 and Y = -5. If therefore we choose h =

y=5

0,0l.L-_-'
x
s
y= -5

Figure 58.64

0.19, then the rectangle R of dimensions Ixl 0.19, Iyl < 4.94 which has
(0,0) at its center will lie in S. The interval I of (58.58) and the rectangle
R are shown in Fig. 58.64. Hence the sequence of functions Yo, Yb ... , Yn
will converge on the interval -0.19 < x < 0.19, to a function y(x) which,
on this interval, is the actual solution of (a) satisfying y(O) = O.
Comment 58.65. The actual solution of (a) satisfying y(O) = 0 is

(d) y = tan x.
Its Maclaurin series is
x3 2x 5 17x 7 62x 9
(e) tan x = x + 3" + 15 + 815 + 2835 + ...
which converges for Ixl < 7r/2. A comparison of (e) with Y3(X) of (b)
shows that the first three terms of each series are the same. If we had
obtained additional Picard approximations, subsequent functions in the
sequence would contain more and more terms in agreement with (e) and
would approach the actual solution tan x. Note, however, how much
smaller the interval I, obtained from the theorem, is than the interval
(-7r/2, 7r/2) for which the series (e) is valid. In the absence of a solution,
we would know by the theorem only, that for Ixl < 0.19, the sequence of
approximating functions Yo(x), Yl(X), "', Yn(X) converges to a function
y(x) which is the particular solution of (a) satisfying y(O) = O. We would
thus know, for example, that Yn(O.l) is an approximation to tan (0.1).
Howcver, we could not know from the thNlrf'm whether Yo(0.5),
Lesson 58-Exercise 743

Yl (0.5), ... , Yn(0.5) converges to the actual solution tan (0.5) unless we
could find a larger h. The theorem therefore does not always give us a maxi-
mum interval of convergence.

EXERCISE 58
1. Show that the sequence of functions x, x 2, x 3 , ••• , xn converges on I:
o x < 1, to the functionf(x) = 0, but not uniformly. What is the limiting
function of this sequence if I is the interval 0 x I? Is the limiting
function continuous?
.
2. Show that the sequence of functlOns 1 +x x' 1 +2x 2x ' ... , 1 +nx nx converges
on I: 0 < x < 00 to the functionf(x) = 1, but not uniformly. What is the
limiting function of this sequence if I is the interval 0 x < oo? Is this
limiting function continuous?
3. Shpw that the sequence of functions l/x, 1/2x, 1/3x,···, l/nx converges
uniformly on I: 1 x < 00 to the continuous function f(x) = O.
4. Show that the series

whose sequence of partial sums is

x2 x4 (x2)n
1 + x 2 ' 1 + x4 ' ... , 1 + (x 2 )n'
converges to the function
f(x) = 0, Ixl < 1,
=!, Ixl = 1,
= 1, Ixl > 1.
5. In comment 58.59, we stated that the first Picard approximation need not
be the function Yo (x) = Yo, where Yo is the given initial value. In Example
58.63, our first approximation was Yo (x) = o. Show that yo(x) = x would
have been a better first approximation. Solve the problem starting with this
new first approximation.
6. Solve the problem in Exercise 57,1, starting with the approximation yo(x) =
1 - x. See 5 above. Is it a better or poorer approximation?
7. Find an interval of convergence for each sequence of Picard approximations
obtained in Exercise 57,1, Exercise 57,2, Exercise 57,5.

ANSWERS 58
1. f(x) = 0, 0 x <
1; f(x) = 1, x = 1. Limiting function is discontinuous.
2. f(x) = 1,0 < x <
00 ;f(x) = 0, x = O. Limiting function is discontinuous.

6. Better approximation. Actual solution is y = 2e-Z x-I = 1 - x + +

x2 --+-
x3
3
x4
12
__
x5
60
... .
744 EXISTENCE THEOREM: y' = f(x,y). PICARD. CLAIRAUT. Chapter II

7. (1) If(x,y) I +
Ixl IYI < ltf. Any values of x and yare permissible tor
region S. Choose h so that rectangle of dimensions Ixl h, Iy - 11 < ltfh
which has (0,1) at its center lies in S. The interval is Ixl < h.
(2) If(x,y) I +
Ix 1 Iyl < ltf. Any values of x and yare permissible for
2
region S. Choose h so that rectangle of dimensions Ix - 11 h, Iy - 31 <
JIJ h which has (1,3) at its center lies in S. The interval is Ix - 11 < h.

LESSON 59. The Ordinary and Singular Points of a First Order

Differential Equation y' = f(x,y).
In Lesson 5, we introduced the concept of an ordinary point and of a
singular point of a first order equation
(59.1) y' = I(x,y).

For convenience we repeat these Definitions 5.4 and 5.41.

Definition 59.11. An ordinary point of a first order differential equa-
tion (59.1) is a point in the plane which lies on one and only one member
of a I-parameter family of solutions, Le., it lies on one and only one integral
curve.
Definition 59.12. A singular point of the first order differential equa-
tion (59.1) is a point in the plane which meets the following two require-
ments.
1. It lies on none or more than one integral curve of (59.1).
2. If a circle of arbitrarily small radius is drawn about this point, there is
at least one ordinary point in its interior.
Comment 59.13. By Definition 59.11 and Comment 58.61, all points
(x,y) of a region S, in which/(x,y) satisfies (58.51) and (58.56), are ordinary
points of (59.1). They lie on one and only one integral curve of its family
of solutions. If the region S contains points which do not fulfill (58.51)
or (58.56), then by Definition 59.12 and Comment 58.61, they may be singu-
lar. Hence in hunting for singular points, we need only examine those
which fail to fulfill (58.51) or (58.56). There is certainly no need to look
for them among those which do.
Example 59.2. Determine whether the differential equation
(a) y' = 2x

has singular points.

Solution. Here I(x,y) of Theorem 58.5 is 2x and al(x,y)ay = O.
Therefore in any bounded region S, I(x,y) is continuous and satisfies
(58.56). Its partial derivative with respect to y is also continuous and
bounded in S. Therefore by Comment 58.24, I(x,y) satisfies (58.51).
Hence, by Comment 59.13, each point of S is an ordinary point. There
are no singular points.
Lesson 59 ORDINARY AND SINGULAR POINTS OF y' = f(x,y) 745

Remark. The solution of (a) is

(b) y = x2 + c,
which is a family of parabolas with its vertices on the y axis. For each
point (xo,Yo) in the plane, (b) will give an unique particular solution of (a)
satisfying the point (xo,Yo). Hence, by Comment 58.62, we now also know
that (b) is a general solution of (a).
Example 59.21. Determine whether the differential equation

(a) y' = -VI=Y2, -1;;;; y ;;;; 1,

has singular points.

Solution. Since -v'1=Y2 is defined only for values of y for which
-1 ;;;; Y ;;;; 1, we confine our attention to a region which is contained
between these lines. Here f(x,y) of Theorem 58.5 is --v'1=Y2 and
iJf(x,y)jiJy = Let us take for S the region defined by
-1 < y < 1, Ixl < A, where A is a positive number, Fig. 59.22. Let
y
r--------------f--------------7Y=1
x -A
(0,0)
x

Figure 59.22

(xo,Yo) be a point of S. Since it is an interior point of S, it can be sur-

rounded by a rectangle contained in S. In this rectangle, f(x,y) is con-
tinuous and satisfies (58.56). Its partial derivative with respect to y is
also continuous and bounded. Therefore, by Comment 58.24, f(x,y) satis-
fies (58.51). Hence each point in this rectangle, by Comment 59.13, is an
ordinary point. Since each point of S can be similarly surrounded by a
rectangle, all points of S are ordinary points.
If we enlarge S to include the lines y = 1 and y = -1, then for these
values of y, (58.51) is not satisfied. [In (58.51), take '!J = 1. Then as
y -+ 1, I( -VI=1J2 +O)j(y - 1)1 -+ 00.] Hence the points on these
lines may, by Comment 59.13, be singular points of (a). That they are
in fact singular points can be seen from the solutions of (a). These are
(b) y = cos (x + c)
746 EXISTENCE THEOREM: y' = f(x,y). PICARD. CLAlRAUT. Chapter 11

where c is the usual arbitrary constant, and the functions

(c) y = ±1.

Each point on the line y = 1, i.e., each point (xo,l) lies on the integral
curves y = cos (x - xo) and y = 1; each point on the line y = -1, i.e.,
each point (xo, -1), lies on the curves y = cos (x + 7r - xo), and y =
-1. For example, the point (0,1) lies on the integral curves y = cos x
and y = 1, the point (-2,1) lies on the integral curves y = cos (x + 2)
+
and y = 1, the point (3, -1) lies on the integral curves y = cos (x 7r - 3)
and y = -1. Hence, by Definition 59.12, the points on the lines y = 1
and. y = -1 are singular points.
Comment 59.23. If the region S in the above example excludes the
lines y = ±1, then each point of S, as remarked above is an ordinary
point. Hence by Definition 59.11, there is one and only one particular
solution through each point of S. Since for each point of S the family (b)
yields a particular solution through it, we now know, by Comment 58.62,
that the I-parameter family (b) is, in S, a general solution of (a). If, how-
ever, S includes the lines y = ±1, then in the absence of further infonna-
tion, we cannot know whether (b) is a general solution of (a), i.e., whether
it contains every particular solution. For since (x,±I) may be singular
points, there mayor may not, by Definition 59.12, be solutions through
these points. Actually, as we saw above, y = ±1 are particular solutions
of (a) not obtainable from the family (b). Hence for this enlarged region S,
(b) is not the general solution of (a).
Example 59.24. Determine whether the differential equation

(a) y' = (y + 1)2/y, y F 0,

has singular points.

Solution. Heref(x,y) of Theorem 58.5 is (y + 1)2/y and iJf(x,y)/iJy =
(y2 _ 1)/y2. Let us take for S the region defined by -B y < 0,
Ixl < A, where A, and B > 1, are positive numbers, Fig. 59.241. Let
(xo,Yo) be a point of S. Since it is an interior point of S, it can be sur-
rounded by a rectangle contained in S. In this rectangle f(x,y) is continu-
ous and satisfies (58.56). Its partial derivative with respect to y is also
continuous and bounded. Therefore, by Comment 58.24, f(x,y) satis-
fies (58.51). Hence, each point in this rectangle, by Comment 59.13, is an
ordinary point. Since each point of S can be similarly surrounded by a
rectangle, all points of S are ordinary points. Therefore, by Definition
59.11, each point in S lies on one and only one integral curve of (a).
The n-parameter family of solutions of (a) is

(b) y +1 1 + log Iy + 11 = x + c, y F -1.

Lesson 60 ENVELOPES 747

We now note that (b) excludes particular solutions of (a) that lie on points
whose y coordinate is -1. Hence we now also know that (b) cannot be
the general solution of (a) and that there must be an additional solution
(or solutions) of (a) that lie on the points (x,-I). This particular solution

r-__________

x A"

1
B
y

Figure 59.241

is, in fact, the line y + I = O. This solution together with (b) contain
every solution of (a). There are no others. Note the role played by the
existence theorem in warning us that (b) cannot be the general solution
of (a).
EXERCISE 59

Determine the singular points, if there are any, of each of the following
equations.
1. y' = x - y.
2. y' = 2y - 1, x F- O.
x
3. yy' Iyl 1,y F- O.
4. y' = Vii, y O.
5. y' = _1__ , x - y F- O.
x-y

ANSWERS 59
1. None. 2. None in each half plane x > 0, x < O.
3. Lines y = ± 1. 4. Line y = O. 5. None in each of the two regions
divided by the line x - y = o.

LESSON 60. Enve]opes.

In this text, we have described various techniques for finding a 1-
parameter family of solutions f(x,y,c) = 0, for special types of first order
differential equations of the form F (x,y, ::) = O. If this I-parameter
748 EXISTENCE THEOREM: y' = J(X,y). PICARD. CLAIRAUT. Chapter 11

family of solutions is not a general one, then the problem of finding par-
ticular solutions, if there are any, that are not obtainable from the family
is usually a complex and difficult one. However, there is one case where
a standard method exists for finding such particular solutions. The method
we shall describe is closely associated with the notion of an "envolope" of
a family of curves. Hence we shall first discuss the meaning of an envelope
of a family of curves.

LESSON 6OA. Envelopes of a Family of Curves. Consider the

family of circles (x - C)2 + y2 = 1, with centers at (c,O) and radius 1,
Fig. 60.01. Each circle is tangent to the lines y = ±1. Since these two

Figure 60.01

lines may be looked at as enclosing the family of circles, and since the
word envelope means a wrapper for enclosing something, each has been
called an envelope of the given family of circles. More generally, we de-
fine the envelope of a family of curves as follows.
Definition 60.1. Let f(x,y,c) = °
be a given family of curves, Fig.
60.2. A curve C will be called an envelope of the family if the following
two properties hold.
1. At each point of the envelope there is a unique member of the family
tangent to it.

-----"Members of family f(x,y,c)=O

Figure 60.2
Lesson 60A ENVELOPES OF A FAMILY OF CURVES 749

2. Every member of the family is tangent to the envelope at a distinct

point of the envelope.
Definition 60.21. A functionf(x,y,c) is said to be twice differentiable
if all its first and second partial derivatives exist.
We now state without proof* a theorem which gives a sufficient condition
for the existence of an envelope of a family f(x,y,c) = O.
Theorem 60.22. If f(x,y,c) is a twice differentiable function defined for
a set of values of x,y,c, and if, for this set of values,

(60.23) f(x,y,c) = 0, af(x,y,c) = 0

ac '
and
af af
(60.24)
ax ay a 2f
0, a 2c 0,
a 2f a 2f
axac ayac

then the family of curves f(x,y,c) = 0 has an envelope whose parametric

equations are given by (60.28).
Assume that f(x,y,c) is a function which satisfies the hypothesis of
Theorem 60.22 so that we can be sure the family of curves

(60.25) f(x,y,c) = 0

has an envelope. In place of the difficult proof of the above theorem

which we have omitted, we shall now describe a formal method of finding
this envelope. At the same time, we shall discover how the equations in
(60.23) originated. No attempt will be made to justify each step rigorously.
For each value of c, we obtain a member of the family (60.25). Hence if
c = Co and c = Co + where 0 is small, then
(60.26) f(x,y,co) = 0 and f(x,y,co + = 0
are two neighboring members of the family (60.25). By Definition 60.1,
each of these two curves is tangent to the envelope of the family (60.25).
Let us call PI, P 2 their respective points of tangency, Fig. 60.29. The
two curves themselves will usually intersect a point, called P in the
figure, which is not too far distant from PI and P 2 • Since the coordinates
of the point P satisfy each of the equations of (60.26), it will also satisfy
the equation
(60.27) f(x,y,co + - f(x,y,co) = o.
*A proof can be found in William F. Osgood, Advanced Calculus.
750 EXISTENCE THEOREM: y' = f(x,y). PICARD. CLAIRAUT. Chapter II

(The point P lies in each curve; its coordinates will therefore make each
term zero.) Since dC ¢ 0, we may divide (60.27) by it to obtain

(60.28)
f(x,y,co + dC) - f(x,y,co) = 0
dC .

Figure 60.29

By the definition of a derivative, which we assumed exist,s,

(60.3)
. f(x,y,co
11m
+ dC) - f(x,y,co) iJf(x,y,c)
= ,
<1c-+O dC iJc
evaluated at c = co.
Therefore as dC - 0:
. ) iJf(x,y,c)
(a) The expreSSlOn on the left of (60.28 approaches .
iJc
(b) The points P 2 and P approach Pl.
(c) The curve f(x,y,co + dC) = 0 approaches the curve f(x,y,co) = 0
so that both are tangent to the envelope at Pl.

Since Co is an arbitrary value of c, we have thus proved that if a family

of curves has an envelope, then each point of the envelope, by (c) and (a)
above and (60.28), must satisfy the two equations

(60.31) f(x,y,c) = 0,
iJf(x,y,c) = 0
iJc .

These two equations may be regarded as the parametric equations of an

envelope of the family (60.25). Note that they are the same as (60.23).
Comment 60.32. It may not always be possible to eliminate the
parameter c between the two equations in (60.31), but if it can be elimi-
nated, then the resulting Cartesian equation y = g(x) is called the elimi-
nant of (60.31). An eliminant, however, may introduce additional loci
Lesson 60A ENVEWPES OF A FAMILY OF CURVES 751

which are not part of the envelope. For example,

(a) x = V2Tc, y = V2=C
are the parametric equations of that part of a circle of radius two that lies
in the first quadrant (since x and yare positive). Their eliminant is x 2 +
y2 = 4, which is the entire circle.
Comment 60.33. Not every family has an envelope. For example,
the family of concentric circles x 2 + y2 = e2 has no envelope.
Comment 60.34. Theorem 60.22 gives only a sufficient condition for
the existence of an envelope, not a necessary one. This means that if a
family satisfies the hypotheses of Theorem 60.22, it must have an en-
velope; if it does not, it mayor may not have an envelope.
Example 60.35. Find the envelopes, if there are any, of the family of
curves
(a) y = cos (x + e).
oj(x,y,e) . (
Solution. Herej(x,y,e) = y- cos (x + e) and = sm x + e).
oe
The determinant (60.24) is therefore

(b) t = -cos (x + e).

Also o2j/oe 2 = cos (x + e). If x + e 7r/2, then cos (x + c) O.
Hence, by Theorem 60.22, the sets of values, excluding x + e = 7r/2, that
satisfy the parametric equations
(c) y - cos (x + c) = 0,
sin (x + c) = 0,
are envelopes of the family (a). Writing the first equation of (c) as
cos (x + c) = y, then squaring both equations of (c) and adding, we obtain
the eliminant,
(d) y = ±1.
The two curves y = 1 and y = -1 are the envelopes of the family (a).
Comment 60.36. If we write (a) as
(e) Arc cos y = x + e,
then j(x,y,e) = Arc cos y - x - e and oj/oe = -1. Equations (60.23)
therefore become
(f) Arc cos y - x - e = 0,
-1 = o.
752 EXISTENCE THEOREM: y' = f(x,y). PICARD. CLAIRAUT. Chapter II

Since -1 ;t6 0, the parametric equations are not satisfied. Yet we showed
above that (a) has two envelopes, y = ±1. The trouble is that now
a 2j/ac 2 = 0 for all x,y, so that one of the hypotheses of Theorem 60.22
is not fulfilled. The fact that the family (e) has an envelope shows that
the theorem gives only a sufficient condition, not a necessary one.
Example 60.37. Find the envelopes, if there are any, of the family of
circles
(a) (x - C)2 + y2 = 1,
where c is a parameter.
Solution. Here j(x,y,c) = (x - C)2 + y2 - 1, aj/ac = -2(x - c).
The determinant (60.24) is therefore
2
(b) /2(X - c) 2Y / = 4y ;t6 0 if y ;t6 0 and a { = 2 ;t6 O.
-2 0 k
Hence by Theorem 60.22, the sets of values, excluding y = 0, which satisfy
the parametric equations
(c) (x - C)2 + y2 - 1 = 0,
-2(x - c) = 0, x - c = 0,
are envelopes of the family (a). Substituting the second equation of (c) in
the first, we obtain the eliminant
(d) y2 = 1, y = ±1.
Therefore by Theorem 60.21, y = 1 and y = -1 are envelopes of (a).
See Fig. 60.01.
Comment 60.38. A trouble similar to that mentioned in comment
60.36, occurs if we write (a) as
(e) c= x ± v'1=Y2.
Equations (60.23) therefore become
(f) c- x ± v'1=Y2 = 0,
1 = O.
Since 1 ;t6 0, (f) cannot be satisfied. But here a2j/ac 2 = 0 for all x,y,c,
so that one of the hypotheses of Theorem 60.22 is not fulfilled. The fact
that the family (e) has an envelope shows again that the theorem gives
only a sufficient condition, not a necessary one.
Example 60.39. Find the envelopes if there are any of the family
(a)
Lesson 60A ENVELOPES OF A FAMILY OF CURVES 753

Solution. You can verify that (60.24) is satisfied if y O. The

parametric equations (60.23) are:
(b) y2 _ 2cx + c2 = 0
-2x + 2c = 0, x = c.
Substituting the second equation of (b) in the first, we obtain the elimi-
nants
(c) y2 - 2X2 + x 2 = 0, y2 - x 2 = 0, X - Y = 0, x + y = O.
Hence x = y and x = -y, y 0, are envelopes of (a). See Fig. 60.4.

Figure 60.4

Example 60.41. Find the envelopes, if there are any, of the family of
curves
(a) (x - C)2 = 3y2 _ y3.
Solution. Herej(x,y,c) = (x - C)2 - 3y2 + y3, iJj/iJc = -2(x - c).
The determinant (60.24) is therefore

(b) 12 (X; c) -6Y : 3y21 0, if y 0 and 2; :;{ = 2 O.

By Theorem 60.22, sets of values, excluding y = 0 and 2, that satisfy the

parametric equations
(c) (x - C)2 - 3y2 + y3 = 0,
2(x - c) = 0, (x - c) = 0
754 EXISTENCE THEOREM: y' = f(x,y). PICARD. CLAIRAUT. Chapter 11

are envelopes of (a). Substituting the second equation in the first, we

obtain
(d) y = 0, y = 3.
The eliminant y = 0 is excluded by (b). However, since Theorem 60.22
gives only a sufficient condition, we cannot assert without further study
that y = 0 is not an envelope. (Actually it is not, see Example 60.53.)
However, y = 3 is an envelope.
Example 60.42. Find the envelopes, if there are any, of the family of
curves
(a) ceY=x-y-l.
Solution. Here f(x,y,c) = ceY - x + y + 1, af/ac = eY. The deter-
minant (60.24) is therefore
(b)
1- ::u + 11 = -e Y
0, but ;;{ = O.

Hence one of the hypotheses of Theorem 60.22, namely a2f / ac 2 0, is not

satisfied. The family (a), therefore, mayor may not have envelopes.
(It actually has none, see Example 60.54.)

LESSON 6OB. Envelopes of a I-Parameter Family of Solutions.

Letf(x,y,c) be a I-parameter family of solutions of the differential equation
y' = f(x,y). The family mayor may not have an envelope. If it has, then
by Definition 60.1, at each point of the envelope, there is a member of
the family of solutions that is tangent to it. Hence the envelope, at each
of its points, has the same slope y' as an integral curve. It therefore follows
that each envelope of the family of solutions, excluding those points of the
envelope where the tangent is vertical, also satisfies the given differential
equation y' = f(x,y). Hence an envelope of a family of solutions of y' =
f(x,y) is also a solution of this differential equation.
Unless an envelope is coincident with an integral curve of a family, it is,
by Definition 59.12, a locus of singular points. For each point of the en-
velope lies on two integral curves: the envelope itself and an integral curve
of the family. Hence an envelope is a particular solutioh of a differential
equation y' = f(x,y) that cannot be obtained from a I-parameter family
of solutions by assigning a value to the arbitrary constant. The converse,
however, need not be true. A particular solution not obtainable from a
I-parameter family of solutions is not necessarily an envelope. If, there-
fore; we can find an envelope of a I-parameter family of solutions of a
differential equation y' = f(x,y), we shall at the same time have succeeded
in finding a particular solution of the equation, not obtainable from the
family.
Lesson 60B ENVELOPES OF A I-PARAMETER FAMILY OF SOLUTIONS 755

Comment 60.43. By Theorem 60.22, we know that if f(x,y,c) = 0

is a twice differentiable function and satisfies (60.24) over a set of values
of X,y,c, then the parametric equations of its envelope are given by

(60.44) f(x,y,c) = 0, of(x,y,c) = 0

oc .

If the parameter c can be eliminated between these two equations, the

resulting equation y = g(x), which we called the eliminant, gives an en-
velope. However, as noted in Comment 60.32, the eliminant may introduce
extraneous loci. It is essential, therefore, to verify by actual substitution
in a given differential equation whether each factor of the eliminant satisfies
the equation. If it does, it is a particular solution not obtainable from the
family. See Example 60.53 below for an eliminant which is not a solution.
Example 60.5. Find, by means of envelopes, particular solutions, if
there are any, of the differential equation
(a) y' = vr=yz,
not obtainable from the family of solutions
(b) y = cos (x + c)
Solution. In Example 60.3.5, we found that y = 1 and y = -1 are
envelopes of the family (b). You can verify by direction substitution that
each is a solution of the given differential equation (a). Hence each en-
velope is a particular solution of (a) not obtainable from (b).
Example 60.51. Find by means of envelopes, particular solutions, if
there are any, of the differential equation

(a) y' -
- - V1=Y2 ,
y
not obtainable from the family of solutions
(b)
Solution. In Example 60.37, we found that y = 1 and y = -1 are
envelopes of (b). Since each function satisfies (a), each is a particular
solution of (a), not obtainable from the family (b).
Example 60.52. Find, by means of envelopes, particular solutions, if
there are any, of the differential equation

(a) ,_ x
Y -
+ vx 2 - y2
,
Y
not obtainable from the family of solutions
(b)
756 EXISTENCE THEOREM: y' = f(x,y). PICARD. CLAIRAUT. Chapter 11

Solution. In Example 60.39, we found that y = ± x are envelopes

of (b). Since these functions satisfy (a), they are particular solutions of
(a), not obtainable from the family (b).
Example 60.53. Find, by means of envelopes, particular solutions, if
there are any, of the differential equation
(a) 9(y,)2(2 - y)2 = 4(3 - y),

not obtainable from the family of solutions

(b)

Solution. In Example 60.41, we found two eliminants, namely y = 3

and y = o. The first satisfies (a), and is therefore a particular solution
of (a), not obtainable from (b). The second y = 0, however, does not
satisfy (a). Hence, we now also know that y = 0 is not an envelope of (b).
(In example 60.41, we remarked that in the absence of further information,
we could not know whether or not y = 0 is an envelope.)
Example 60.54. Find, by means of envelopes, particular solutions, if
there are any, of the differential equation
, 1
(a) y =--,
x-y

not obtainable from the family of solutions

(b) ce7l=x-y-1.

Solution. In Example 60.42, we found that (60.24) of Theorem 60.22

was not satisfied. Hence we did not know whether or not the family (b)
had an envelope. However, if we apply the existence Theorem 58.5 to the
function f(x,y) = 1/(x - y) of (a), we find that the only possible singular
points lie on the line x - y = O. Since the function y = x does not
satisfy (a), it cannot be an envelope of the family (b). In this example,
therefore, a particular solution not obtainable from (b), if there is one,
cannot be found by means of envelopes.

EXERCISE 60
Find the envelopes, if there are any, of each of the following family of
curves.
1. y = x + c. 5. y =cx + aV1+C2, y > 0
2. Y = x + f(x + d. 6. (y + x - d = 4xy.
3. y = (x - c)3. 7. y = sin [(x - c)2).
4. ylf3 = x-c.
Lesson 61 THE CLAIRAUT EQUATION 757

Find, by means of envelopes, particular solutions, if there are any, of

each of the following differential equations, not obtainable from the given
family of solutions.
8. y' = +
1, y = x + Hx + d, x y x + 1.
9. y' = 3l'3, Y = (x - c)3.
10. y = xy' + 3Vl + (y')2, y = cx + 3-v'i+C2, y > 0, x 2 < 9.
11. y' = 2V (1 - y2) VArc sin y, x = VArc sin y + c.
ANSWERS 60
1. None. 6. xy = O.
2. y = x. 7. y = ±1.
3. y = O. 8. y = x.
4. y = O. See problem 3 and read 9. y = O.
Comments 60.36, 60.38. 10. y = v9 - x 2.
5. y = v'9"=X"2. 11. y = 0, y = ± 1.

LESSON 61. The Clairaut Equation.

The simplest type of family of curves is a family of straight lines. From

analytic geometry we know that
(61.1) y = mx + b,
represents such a family. The slope of each line of the family is m, and
its y intercept is b. The family (61.1) is thus a 2-parameter family. We can
form a I-parameter family from it by requiring that b be a function of the
slope m, i.e., that b = f(m). Hence (61.1) becomes
(61.12) y = mx + f(m).
To find a differential equation whose I-parameter family of solutions is
the family of lines (61.12), we proceed as we did in Lesson 4B. Differenti-
ation of (61.12) gives y' = m. Substituting this value of m in (61.12), we
obtain the differential equation
(61.13) y = y'x + f(y').
Equation (61.13) is called Clairaut's equation. *
Comment 61.14. If we start with the differential equation (61.13),
its solution (61.12) is easily found. Replace y' by the parameter m (or, if
you prefer, bye). For example, the solution of the Clairaut equation
y = y'x + (y,)2 is the family of nonvertical lines y = mx + m 2 or
y = ex + e 2 •

·Named after the French mathematician, Alex Claude Clairaut (1713-1765).

758 EXISTENCE THEOREM: y' = f(x,y). PICARD. CLAIRAUT. Chapter 11

As we pointed out in Lesson 60B, if the family (61.12) has an envel-

ope, then this envelope is a particular solution of (61.13) not obtainable
from the family of solutions (61.12). By Theorem 60.22, a sufficient
condition, among others, for the existence of an envelope of the family
(61.12) is that points of the envelope satisfy the parametric equations
(61.15) y - ex - fee) = 0,
x + f'(e) = O.
The first equation in (61.15) has the same form as (61.12). It therefore
satisfies Clairaut's equation (61.13). The eliminant of (61.15) also satisfies
this first equation. (A solution of two simultaneous equations satisfies
each equation.) Hence, if (61.15) yields an eliminant, then this eliminant
by Comment 60.43, may be a particular solution of (61.13). It is if it
satisfies (61.13).
Example 61.2. Find a I-parameter family of solutions of the Clairaut
equation
(a) y = y'x + (y')2.
Also investigate for envelopes of the family of solutions.
Solution. First we note by (61.13), that (a) is a Clairaut equation
with fey') = (y,)2. Hence by Comment 61.14, its I-parameter family of
solutions is
(b)
You can verify that (b) satisfies (60.24). Therefore, by Theorem 60.22, a
set of values that satisfies the parametric equations (60.23),
(c) y - ex - e2 = 0,
x + 2e = 0, e = -x/2,

is an envelope of (b). Substituting the second equation of (c) in the first,

we obtain the eliminant
x2 X4 x2
(d) y = - '2 + '4 = - '4'
which is an envelope of the family of nonvertical lines (b). Since this
function (d) satisfies (a), it is a particular solution of (a) not obtainable
from the family (b).
Example 61.21. Find a I-parameter family of solutions of the Clairaut
equation
(a) y = y'x + logy'.
Also investigate for envelopes of the family of solutions.
Lesson 61 THE CLAlRAUT EQUATION 759

Solution. First we note by (61.13) that (a) is a Clairaut equation

withf(y') = log y'. Hence by Comment 61.14,
(b) y = ex + loge,
is a family of solutions of (a). You can verify that (b) satisfies (60.24).
Therefore, by Theorem 60.22, the set of values that satisfies the para-
metric equations (60.23).,
(c) y - ex - log e = °
1 1
x+-e = 0, x = e
is an envelope of (b). Substituting the second equation of (c) in the
first, we obtain
(d) y +1- log (-X-I) = 0, y + 1 + log (-x) = 0, x < 0,
which is an envelope of the family of nonvertical lines (b). Since this
function (d) satisfies (a), it is a particular solution of (a) not obtainable
from the family (b).
Example 61.3. A source of light or sound which strikes a curve in the
same plane with it, is reflected in a fixed direction. Find the equation of
this curve.

Reflected direction
Q

x
0(0,0) x

Figure 61.31

Solution. (See Fig. 61.31.) We place the origin at the source of the
light or sound, and let the x axis be parallel to the fixed reflected direction.
Let y = y(x) be the equation of the curve we seek. We assume y(x) is
defined and differentiable on an interval I.
760 EXISTENCE THEOREM: y' = f(x,y). PICARD. CLAIRAUT. Chapter 11

Call:
(a) i the angle the incident ray OP makes with the nonnal to the required
curve,
r the angle the reflected ray PQ makes with the nonnal to the re-
quired curve,
y' the slope of the tangent to the required curve y = y(x) at P(x,y),
ml the slope of the normal to the required curve at P(x,y) , (=-I/y'),
m2 the slope of OP (=y/x).

Bya law of physics

(b) tan i = tan r,
and by a fonnula of analytic geometry

-x - yy'
(c)
xy' - y

From Fig. 61.31, we see that tan (180° - r) = -1/y'. Hence, tan r =
l/y'. Therefore by (b), tan i = l/y'. Substituting this value of tan i in
(c), we obtain
1 -x - yy'
(d) y = 2xy' + y(y,)2.
17= xy'-y'

The equation is almost like a Clairaut equation, but not quite. If, however,
we multiply it by y, we obtain
(e) y2 = 2xyy' + y2(y')2.
Now let
(f) u' = 2yy', (y,)2 = (U')2.
4y2
Substituting (f) in (e), we have
(g)

which is a Clairaut equation in u. Hence its I-parameter family of solu-

tions is
c2
(h) u=cx+_·
4
Substituting in (h) the value of u as given in (f), we obtain the family of
parabolas
(i)
Lesson 61-Exercise 761

Geometric Problems Giving Rise to a Clairaut Equation. In Les-

sons 13 and 36D, we solved geometric problems which gave rise, respec-
tively, to a first order equation and to a special type of second order
equation. A curve whose tangents have properties which are independent
of the point at which the tangent is drawn will lead to a Clairaut equa-
tion. Since the solution of a Clairaut equation is a family of straight
lines, and since these straight lines will have the properties of the tangent
lines, the envelope of the family will give the curve we seek.
Example 61.4. Every tangent to a curve has the property that the
sum of its intercepts has a constant value k. Find the curve.
Solution. By Exercise 13,I(a) and l(b), the x and y intercepts of a
tangent line are respectively x - y/y' and y - xy'. Hence the curve must
satisfy the condition
(a) x - 1L
y' +y-xy'=k , xy' - y + yy' - X(y,)2 = ky',
ky'
y(y' - 1) = xy'(Y' - 1) + ky', y=xY'+y'_I'

The last equation in (a) is, by (61.13), a Clairaut equation. Hence its
solution, by Comment 61.14, is
kc
(b)
c - 1'
y=cx+-- xc 2 - (x +y - k)c +y = 0,

which is a family of straight lines. You can verify that (b) satisfies (60.24).
Therefore, by Theorem 60.22, the c-eliminant of
(c) xc 2 - (x + y - k)c + y = 0
x+y-k
2xc - (x +y - k) = 0, c = 2x '

may be an envelope of (b). Substituting the second equation of (c) in the

first, we obtain the eliminant

(d) (x +y - k)2 _ (x +y - k)2 + = 0

4x 2x y,
(x +y - k)2 - 4xy = 0, x +y - k = ± 2xI/2yI/2,

Since it satisfies (a), it is the required solution.

EXERCISE 61
Find a I-parameter family of solutions of each of the following Clairaut
equations 1-6. Also investigate for envelopes.
762 EXISTENCE THEOREM: y' = f(x,y). PICARD. CLAIRAUT. Chapter 11

I. y = xy' _ (y,)2. 4. Y = xy' _ (y,)3.

2. y = xy' + +
[1 (y,)2]. 5. y = xy' + v'17"1"""'
. +,---;"(Y"')C"::2.
3. y = xy' _ (y')2/3. 6.y=xy'-e".
7. Solve the equation y = 3xy' +
6y2(y')2. Hint. Multiply by y2. Then make
the substitution u = y3, du/dx = 3y2y' to obtain the Clairaut equation

In problems 8-13, every tangent to a curve has the property indicated.

Find the curve.
8. The sum of its intercepts has a constant value 9.
9. The product of its intercepts has a constant value k.
10. The length of the segment intercepted by the coordinate axes has a constant
value k 2 • Hint. Set the square root of the sum of the squares of its intercepts
equal to k 2 •
11. Its distance from the origin has a constant value k. Hint. First show that
thc equation of a tangent line at a point (x,y) on the required curve is given
by y'X - Y + y - xy' = 0, where (X,Y) is a point on the tangent line.
Then use the fact that the distance of a point (A,B) to the line aX bY + +
e = 0 is ± (aA.+ + bB e)/v'(i2.f02. Here A = 0, B = 0.
12. The sum of its distances from the points (a,O) and (-a,O) has a constant
value k. In the distance formula from a point to a line, as given in 11, take
the positive square root.
13. The segment intercepted by the coordinate axes forms with these axes a right
triangle whose area has a constant value k 2 •

In each of problems 14-16, find the Clairaut equation whose family of

solutions of straight lines has the envelope indicated. Hint. First find x
and y in terms of y' and then solve (61.13) for f(y').
14. y2 = 2x. IS. 2y = x 2 • 16. xy = -k.

ANSWERS 61
I. 2
y = ex - e , y = x /4. 2
2. y = ex+ 1 + +
e2, x 2 4y = 4.
3. y = ex - e , 27x 2y
2/3 + 4 = 0.
4. Y = ex - e3, 27y2 = 4x 3.
5. y = >0.
6. y = ex - ee, y = x log x-x.
7. y3 = ex + ic 2 • 12. 4(x 2 +
y2) = k 2.
8. Xl/2 ± yll2 = ±3. 13. 2xy = ± k2.
1
9. 4xy = k. 14.y=y'x+_·
2y'
, (y,)2
10. +
x 2/3 y2/3 = k4/3 . IS. y = y x - - - .
2
11. x +l
2 2
= k • 16. y = y'x ± zVki/.
 Chapter 12

Existence and Uniqueness Theorems

for a System of First Order
Differential Equations and for Linear
and Nonlinear Differential Equations
of Order Greater Than One. Wronskians.

LESSON 62. An Existence and Uniqueness Theorem. for a

System. of n First Order Differential Equations
and for a Nonlinear Differential Equation of
Order Greater Than One.

LESSON 62A. The Existence and Uniqueness Theorem. for a Sys-

tem. of n First Order Differential Equations. In Theorem 39.12 we
gave a sufficient condition for the existence of a power series solution of a
system of n first order equations

(62.1)
dYl =
dt h(t,YIJY2,' .. ,Yn),

dY2 =
dt f2(t,YIJY2, ... , Yn),

dXt = fn(t,YIJY2, .•• , Yn),

satisfying the initial conditions,
(62.11)
Compare it with the theorem we now state, which gives a sufficient condi-
tion for the existence and uniqueness of a solution of (62.1) satisfying
(62.11).
763
764 OTHER EXISTENCE THEOREMS. WRONSKIAN. Chapter 12

Theorem 62.12. Let the functions fl' f2' ... ,fn of the first order system
(62.1) be continuous in a region S defined by
(62.13) It - tol ko, IYI - ad kb
IY2 - a21 k2' ... , IYn - ani kn.
In S, let each function satisfy a Lipschitz condition
(62.14) Ih(t,YI,Y2,"', Yn) - h(t,'01,'02, .. " '010)1
N(IYI - 'OIl + IY2 - '021 + ... + IYn - '0101), i = 1,2, ... n,
where (t,YbY2, ... , Yn) and (t,'OI,'02, ... , 'On) are any two points in S.
Then an interval 10: It - tol < h, h > 0, exists in which there is one and
only one set of continuous functions Yl (t), Y2(t), ... , Yn(t) with continuous
derivatives in 10 satisfying the given system (62.1) and the initial conditions
(62.11).
NOTE. The theorem does not always give a maximum interval of con-
vergence; read Comment 58.65.
Remark. When n = 1, the system (62.1) reduces to a first order
equation YI' = h(t,YI). Theorem 58.5, which we proved previously for
this first order equation, is thus a special case of the above more general
theorem.
The proof of Theorem 62.12, which we have omitted, although more
complicated than the proof of Theorem 58.5, follows its pattern exactly.
We find the interval 10 of Theorem 62.12, for example, in practically the
same manner we found the interval 10 of Theorem 58.5. Since the region S
consists of closed intervals and the functions h, f2, ... , fn are each con-
tinuous in S, each is bounded in S. Hence there is a positive number M
such that
(62.15) Ifi(t,YI,Y2, ... ,Yn)1 < M, i = 1, 2, ... , n,
for every point (t,YI,Y2, ... , Yn) in S. By (62.13), (to,al, ... , an) is an
interior point of S. It is therefore possible to pick an h such that the
n + 1 dimensional rectangle R,
(62.16) It - tol h ko, IYI - ad < Mh kl' ..• ,
IYn - ani < Mit kn,
which has this point (to,al,a2, ... , an) at its center, lies entirely in S. The
M is given by (62.15); the k's are given in (62.13).
We then find sets of sequences of Picard approximations
YII, Y12, ... , Yin,
Y21, Y22, ... , Y2n,

Ynl, Yn2, ... , Ynn,

Lesson 62B EXISTENCE THEOREM: NONLINEAR EQUATION OF ORDER n 765

in a manner similar to that given in (57.33) for a system of two first order
equations. The graph of each function in each approximating set, it is
then proved, lies in the n +
1 dimensional rectangle R, and each set of
sequences of approximating functions converges uniformly on I: It - tol
h to the respective set of functions YI(t), Y2(t), ... , Yn(t), which on 10
is a solution of (62.1) satisfying (62.11). Finally we show that this set of
functions Y!' Y2, ... , Yn is unique.
Example 62.17. Find an interval, for which the approximating func-
tions obtained in Example 57.35 converge to the actual solution.
Solution. In this Example fl (t,x,y) = ty, f2(t,X,y) = xy. Comparing
the initial conditions of this example with (62.11) we see that to = 0,
al = 1, a2 = 1. Since fl and 12 are continuous for all t,x,y, we may take
for S any bounded region, say one defined by It I 5 = k o, Ix - 11
10 = k l , Iy - 11 15 = k 2 • Hence Ixl 11, lyl 16 and
(a) Ifd 80, 1121 176 = M.

We must now choose h according to (62.16) so that

(b) It I h 5, Ix - 11 < 176h 10, Iy - 11 < 176h 15.

If h = 0.056, all three inequalities in (b) will be satisfied. The Picard ap-
proximations obtained in Example 57.35, will therefore converge to the
actual solutions x(t), yet) for at least It I < 0.056.

LESSON 62B. Existence and Uniqueness Theoren_ for a Non-

linear Differential Equation of Order n. In Theorem 39.32, we gave a
sufficient condition for the existence of a power series solution of the nth
order nonlinear differential equation
n
d
-.Jl. = f(x y y' y" ... y(n-I))
(62.2) dxn ' , " ,
satisfying the initial conditions
(62.21) y(xo) = ao, y"(xo) = a2, ... ,
y(n-I)(xo) = an-I'
Compare it with the theorem we now state which gives a sufficient condi-
tion for the existence and uniqueness of a solution of (62.2) satisfying
(62.21).
Theorem 62.22. Let the function f in the nth order differential equation
(62.2) be a continuous function of its arguments x,y,y', ... , y{n-ll in a
region S defined by
(62.23) Ix - xol k o, Iy - aol kb Iy' - all k2,
Iy" - a21 k a, ... , Iy(n-I) - an_II kn,
766 OTHER EXISTENCE THEOREMS. WRONSKIAN. Chapter 12

and let f satisfy a Lipschitz condition,

(62.24) If(x,YI,YI', ... , Yl (n-ll) - f(x,Y2,Y2', ... , Y2 (n-O) I

N(IYI - Y21 + IYI' - Y2'I + IYI" - Y2"1 + ...
+ IYI(n-O - Y2(n-0I),
where (X,Yl,YI', ... , YI (n-O) and (X,Y2,Y2', ... , Y2 (n-O) are any two points
in S.
Then an interval 10: Ix - xol < h, h > 0, exists in which there is one
and only one continuous function y(x) with a continuous derivative of order n
satisfying (62.2) and the initial conditions (62.21).
Proof. Assume that y(x) is a solution of (62.2). We now define new
functions YI(X), Y2(X), ... , Yn(X) by the following relations
(62.25) y(x) = YI(X),
'y'(x) = YI'(X) = Y2(X),
y"(x) = YI"(X) = Y2'(X) = Y3(X),

Differentiating the function in the last line of (65.25), we obtain

(62.251) y(nl(X) = YI(nl(x) = Y2(n-ll(x) = ... = Yn_I"(X) = Yn'(x).

By (65.251) and (62.25), we can therefore write (62.2) as

(62.26)

Hence, because of (62.25) and (62.26), the set of functions Yl,Y2, ... , Yn
satisfies the of first order differential equations

(62.27) YI'(X) = Y2(X),

Y2'(X) = Y3(X),
Y3'(X) = Y4(X) ,

Yn-I'(X) = Yn(x),
Yn'(x) = f(x,Yl,Y2,Y3, ... , Yn).
By (62.25), YI (xo) = y(xo) , Y2(XO) = y'(xo), .. " Yn(xo) = y(n-O(xo).
We can therefore replace the initial conditions (62.21) by the conditions

(62.28) YI (xo) = ao, Y2(XO) = at, Y3(XO) = a2, ..• ,

Yn(Xo) = an_I·
Lesson 62B EXISTENCE THEOREM: NONLINEAR EQUATION OF ORDER n 767

Conversely, if we start with the system (62.27) and the initial conditions
(62.28), and define
(62.281) y(x) = YI (x),

then the relations in (62.25), (62.251), and (62.26) hold; the last equation
of (62.27) becomes identical with (62.2), and the initial conditions (62.28)
become (62.21). Hence if there exists a set of functions YlIY2, •. " Y..
which is a solution of the system (62.27) satisfying (62.28), then, by
(62.281), the function Yl (x) of this set is the solution y(x) of (62.2) satis-
fying (62.21).
You can verify that the system (62.27) with initial conditions (62.28)
satisfies the hypotheses of Theorem 62.12. Hence by the theorem, an
interval 10 about Xo exists on which there is one and only one set of par-
ticular solutions Yl, Y2, " ' , Y.. satisfying (62.27) and (62.28). Since Yl(X)
exists and is unique, it follows, by (62.281), that its equal y(x) exists and
is the unique solution of (62.2) satisfying (62.21).
Comment 62.282. We wish to emphasize that it is always possible,
by means of (62.25) and (62.26), to replace a single differential equation of
order n > 1 by an equivalent system of n first order equations.

E%ample 62.29. Set up a system of first order equations equivalent to

the third order nonlinear equation

(a) Y'/(X) = 3xy' _ y2y"

for which
(b) yeO) = 1, y'(O) = -1, y"(O) = 2.
Solution. Let y(x) be a solution of (a). As suggested in (62.25), we
define
(c) y(x) = YI (x),
y'(x) = YI'(X) = Y2(X),
y"(x) = Yl"(X) = Y2'(X) = Y3(X).

Differentiating the last equation in (c), we obtain

(d) ylIJ(x) = YI"'(X) = Y2"(X) = Ya'(x).

By (d) and (c), we can write (a) as

(e)

Hence, by (c) and (e), we can replace the single equation (a) by the system
768 OTHER EXISTENCE THEOREMS. WRONSKIAN. Chapter 12

of first order equations

(f) Yl'(X) = Y2(X),
Y2'(X) = Ya (X) ,
Ya'(X) = 3XY2 - Y1 2 Ya,
and the initial conditions (b) by
(g) Yl(O) = yeO) = 1, Y2(O) = y'(O) = -1, Ya(O) = y"(O) = 2.
The function Yl(X) of the solution of the system (f) satisfying (g) will be
the solution y(x) of (a) satisfying (b).

LESSON 62C. Existence and Uniqueness Theorem for a System

of n Linear First Order Equations. In Theorem 62.12, no restrictions
were placed on the degree of the dependent variables in the functions
fl' f2' ... , fn of (62.1). In Theorem 62.3 below, each of the dependent
variables Y!' Y2, ... , Yn appears linearly, i.e., each of these variables has
exponent one.
Theorem 62.3. Given a system of n linear first order equations
dYl
(62.31) dt = fll(t)YI + II2(t)Y2 + ... + IIn(t)Yn + QI(t),
dY2
dt = hl(t)YI + h2(t)Y2 + ... + hn(t)Yn + Q2(t),

where all functions fii and Qi, i = 1, 2, ... , n, j = 1, 2, ... n, are con-
tinuous on a common interval I. Then there is, on 10 , i.e., I without its end
points, one and only one set of continuous functions YI (t), Y2(t), ... , Yn(t)
with continuous derivatives, satisfying the system (62.31) and the initial
conditions
(62.32)
where to is a point in 10 ,
Remark. In Theorem 39.22, we gave a sufficient conditio!} for the
existence of a power series solution of (62.31) satisfying (62.32). Compare
it with the above theorem which gives a sufficient condition for the exist-
ence and uniqueness of a solution of (62.31) satisfying (62.32).
Proof of Theorem. The proof of the theorem will consist in showing
that the continuity requirement of the functions hi on I is equivalent to
the requirement that each function satisfy a Lipschitz condition (62.14)
Lesson 62C EXISTENCE TUEOREM: LINEAR SYSTEM 769

in I, provided I is finite and closed. If I is not finite and closed, then it is

sufficient, as we shall show, that each function satisfy this Lipschitz con-
dition on a slightly smaller closed subinterval I' contained in I. And since
all functions on the right of (62.31) are, by hypothesis, continuous on this
common, closed interval lor I', the hypotheses of Theorem 62.12 will be
satisfied. Hence its conclusion will follow. Let
(a) h(t,YI,Y2,' .. ,Yn) = fll(t)YI + fdt)Y2 + ... + fln(t)Yn + QI(t),
h(t,YbY2, ..• ,Yn) = f21(t)YI + h2(t)Y2 + ... + hn(t)Yn + Q2(t),

We assume the interval I of Theorem 62.3 is finite and includes its end
points. If it is not, we take a finite closed subinterval I' of I. Hence in I',
whether it is a subinterval of I or the whole interval I, each functionf;i(t)
and Qi(t) of (62.31) is, by hypothesis, continuous. All are therefore bounded
in I'. This means that the absolute value of each function for all t in I'
is less than some positive number. Let N be the largest of these numbers.
Then for each function fii(t) of (62.31) and for every tin I',
(b) Ifii(t) I N, i = 1,2, ... ,n, j = 1,2, ... , n.

Let JI1, Y2, ... , Yn and '1h, '02, ... , 'On be two sets of values of the de-
pendent variables. Therefore by (a)
(c) f;(t,YI,Y2,"', Yn) = fil (t)YI + f;2(t)Y2 + ... + hn(t)Yn + Qi(t),
#t,'OI,'02, •.• , 'On) = fiI (t)'OI, + h2(t)'02 + ... + hn(t)'!Jn + Qi(t),
for each i = 1,2, ... ,n. Subtracting the second equation in (c) from the
first, we obtain
(d) lJ;(t,YbY2,"', Yn) - h(t,'OI,'02, ... , '!In)1
= Ifil(t)(YI - '01) + f;2(t)(Y2 - '02) + ... + f;n(t)(Yn - 'On)l,
i = 1,2, "', n.
Since la - bl lal + Ibl, and since, by (b), Ifiil N, we obtain from (d)

(e) lJ;(t,YI,"', Yn) - fi(t,'OI,' . " 'On) I Ifil(t)IIYI - YII

+ Ih2(t)IIY2 - '021 + ... + If;nCt)IIYn - 'On I
N(IYI - 'OIl + IY2 - '021 + ... + IYn - 'Onl}·
If you will compare (e) with the Lipschitz condition (62.14), you will find
that both are alike. Hence we have proved that for the special system of
linear equations (62.31), the continuity requirement of the functions!;i in I'
is equivalent to the Lipschitz condition (62.14) of Theorem 62.12.
770 OTHER EXISTENCE THEOREMS. WRONSKIAN. Chapter 12

If the interval I of the theorem is finite and closed, then I = I'. If it is

not, then since it is always possible to surround each t in I by a finite,
closed subinterval l' contained in I, in which the theorem is valid, it
follows that the theorem is valid on I. In this last case I = 1 o.

EXERCISE 62
The following problems refer to the solutions of the systems in Exercise
57. For each find an interval for which the approximating solution con-
verges.
1. Problem 6. 2. Problem 7. 3. Problem 8. 4. Problem 9.
Set up a system of first order equations with appropriate initial condi-
tions equivalent to each of the following equations.
5. y"(x) = 3x 2y' - y2, yeO) = 0, y'(O) = 1.
6. y'''(X) = 2X(y')2 - 3yy" +
xv, yeO) = 1, y'(O) = -1, y"(O) = 2.

ANSWERS 62
1. In Theorem 62.12, h = t + y,h = t - x 2, to = 0, al = 2, a2 = 1. Scan
be any bounded region: ItI ko, Ix - 21 kl, Iy - 11 k2, where ko,
kI, k2 are arbitrary. M is the larger of Ml and M2 in Ihl +
ItI Iyl < Ml,
\t
121 l + Ix 1< M2. Choose h so that ItI h ko, Ix - 21 < Mh kI,
2

/y - 1 < Mh k2.
2. See answer to problem 1 for method.
3. This system is a first order linear one. See Theorem 62.3: fu(t) = t'/21(t) = 1,
Q2(t) = -ee. Each function is continuous for all t. Therefore the approximat-
ing functions converge for every t.
4. In Theorem 62.12, h = y2, 12 = x+ z, fa = z - Yi to = 0, al = 1,
a2 = 0, aa = 1. S can be any bounded region: ItI ko, Ix - 11 kl,
Iyl k2, Iz - 11 ka, where ko, kl' ka are arbitrary. M is the larger of M 1,
M2, Ma in Ihl < y2 < Ml, 1121 Ix l +
Izl < M2, lfal Izl Iyl < Ma. +
Choose h so that ItI h ko, x-II < Mh kl. IYI < Mh k2,
Iz - 11 < Mh / ka. 2
5. y1/(x) = Y2, Y2 (X) = 3X Y2 - Y12; Yl(O) = 0, Y2(0) = 1.
6. yI/(x) = Y2, Y2 / (X) = Ya, ya/(x) = 2XY2 2 - 3Ylya +
XYl; Yl(O) = 1,
Y2(0) = -1, Ya(O) = 2.

LESSON 63. Determinants. Wronskians.

LESSON 63A. A Brief Introduction to the Theory of Deter-

minants. In this lesson, we shall outline only those essentials of the
theory of determinants that we shall need for our purposes. We have al-
ready given the definition of a 2 X 2 and a 3 X 3 determinant in
Comment 31.35 and (31.72). For convenience we recopy them here.
(63.1) al b11 = a 1b2 - a2 b 1,

Ia2 b2
Lesson 63A DETERMINANTS 771

(63.11)
al b1 Cl
a2 b2 C2 = a1b2Ca + a2baCl + aablC2 - aab2Cl - a2blCa - a1baC2.
aa ba Ca

We observe that each tenn in a detenninant, as written on the right,

contains one and only one element from each row and each column. Hence
we conclude:

Theorem 63.12. If each element in a row or in a column of a deter-

minant is zero, then the determinant is zero.

Consider a system of two equations in two unknowns.

(63.2) alx + b1y = CI,

a2 x + b2 y = C2·
Multiplying the first equation by b2 , the second by -b 1, adding the two
and then solving for x, we obtain

(63.21)

By (63.1), we can write (63.21) as

(63.22)

Note that the elements of the denominator determinant are the coefficients
of x and Y in (63.2); the elements of the numerator determinant are ob-
tained by replacing the coefficients of x by the constants Cl and C2 and
recopying the coefficients of y. Similarly you will find, if you solve (63.2)
for y, that

(63.23)
Y = -i-1-a=-1--;"b-=-:+-I '

a2 b2

where the numerator detenninant is obtained by replacing the coefficients

of y by Cl and C2, and the denominator is the same as that in (63.22).
772 OTHER EXISTENCE THEOREMS. WRONSKIAN. Chapter 12

Comment 63.24. From (63.22) and (63.23), we infer the following:

1. If I 0, the system of equations (63.2) has one and only one

solution.
2. If the determinant in 1 equals zero and the numerators of (63.22) and
(63.23) are 0, the system has no solutions.
3. If the determinant in 1 equals zero and the numerators of (63.22) and
(63.23) also equal zero, the system has an infinite number of solutions.
Each of these three possibilities is illustrated in the examples below.
Example 63.25. Solve the system
(a) 2x + 3y = 1,
x - y = 4.
Solution. By (63.22) and (63.23),

-1 - 12 13
(b) x =
-2 - 3
=-,
5
y=

Hence (a) has an unique solution. Note that the denominator determinant
0.
Example 63.26. Solve the system
(a) 2x + 3y = 1,
4x + 6y = 3.

Solution. By (63.22) and (63.23),

-3
(b) x= =0'

Hence there are no solutions for x and y. Note that the denominator de-
terminant is zero but the numerator determinants are not. [The lines in
(a) are actually parallel lines.]
Example 63.27. Solve the system
(a) 2x + ay = 1,
4x + 6y = 2.
Lesson 63A DETERMINANTS 773

Solution. By (63.22) and (63.23)

(b)
o
x= = 0' Y=

Note that both numerator and denominator determinants are zero. As

you can easily verify, the system (a) has an infinite number of solutions.
Give x any value you wish in the first equation of (a). Solve it for y. You
will find that these values of x and Y will satisfy the second equation of (a).
[The lines in (a) are in fact coincident.)
Comment 63.24- is also applicable to a system of n equations in n unknowns.
For n = 3, the system becomes
(63.3) alX + bly + ClZ = dl,
a2x + b2y + C2Z = d2,
aaX + baY + CaZ = da.
Its solution is
(63.31)
d l bl Cl al dl Cl al bl d l
d 2 b2 C2 a2 d2 C2 a2 b2 d 2
d a ba Ca aa da Ca aa ba d a
X= Y= Z=
al bl Cl al bl Cl al bl Cl I
a2 b2 C2 a2 b2 C2 a2 b2
C21
aa ba Ca aa ba Ca aa ba Ca

The system will have an unique solution for x, y, and Z if the denominator
0; no solution if the denominator equals zero and at least one numerator
is not equal to zero; an infinite number of solutions if denominator and all
numerators are equal to zero.
Our primary purpose in introducing determinants is to enable you to
prove the following two important theorems. Although for convenience, we
have stated the theorems for three unknowns, they are applicable to a system
of n equations in n unknowns.
Theorem 63.4. The system of equations

(63.41) alX + bly + ClZ = db

a2x + b2y + C2Z = d2,
aa x + baY + caz = da,
774 OTHER EXISTENCE THEOREMS. WRONSKIAN. Chapter 12

has one and only one solution for x, y, z if the determinant formed by their
coefficients is not equal to zero. It has none or an infinite number of solutions
if this determinant equals zero.
Theorem 63.42. The system of equations

(63.43) alX + b1y + CIZ = 0,

a2X + b2y + C2Z = 0,
aaX + baY + C3Z = 0,

always has the trivial solution x = 0, Y = 0, z = 0. It has an infinite num-

ber of nontrivial solutions, i.e., it has an infinite number of sets of values of
x, y, and z satisfying (62.4-3), where at least one of x, y, z is not zero, if and
only if the determinant formed by their coefficients is zero.

LESSON 63B. Wronskians. In Lesson 19A, when we discussed the

linear dependence and independence of a set of functions, we stated that
we would in time produce theorems by which their linear dependence or
independence could be determined. Since these theorems are closely asso-
ciated with the concept of a Wronskian, we shall first discuss this subject
below.

Definition 63.5. The Wronskian* of a set of functions It(x), f2(x),

..• , fn(x), each of which possesses derivatives of order n - 1, is defined
to be the following determinant.

(63.51) flex) 12 (x) · .. fn(x)

H(x) H(x) · .. fn'(x)
It" (x) 12" (x) · .. fn"(x)
· ......
It (n-l\x)h(n-l)(x) •.. fn (n-ll(x)

Note that the elements of the deterniinant are the given set of functions
and their derivatives through order n - 1. We denote the Wronskian
(63.51) by
(63.52) W(ft,!2, ... fn; x).

If n = 2, the Wronskian of It and f2 is, by (63.52) and (63.51),

(63.53) W(ft,/2; x) = litft' 12'12 I= ftN - hf{ .

·Pronounced Vronskian and named for Hoene Wronski, a Polish mathematician.

Lesson 63B WRONSKIANS 775

Definition 63.54. A function f(x) is said to vanish identically on

an interval I: a x b, or to be identically zero on I, if for every
x in I, f(x) = O. We indicate this fact by writingf(x) == O.
Theorem 63.55. If a set of functions ft, 12, ... , fn, each of which
possesses a derivative of order n - 1, is linearly dependent on an interval
I: a x b, then its Wronskian vanishes identically on I.
Proof. Since the set of given functions ft, 12, ... , fn is linearly de-
pendent on I, there exist, by Definition 19.1, constants Cr, C2, ••• , Cn not
all zero such that

(a) cdl + Cd2 + ... + cnfn = 0,

for every x in I. By differentiating (a) (n - 1) times, we obtain

(b) ftCl + hC2 + ... + fncn = 0,

ft'Cl + !2'C2 + ... + fn'c n = 0,

The system (b) is a set of n equations, each equal to zero for every x in I.
The c's in each equation are not all zero. There is therefore a set of c's
not all zero that satisfies each equation in (b). Hence this set of c's is a
solution of the system of equations (b). It therefore follows by Theorem
63.42, that the determinant
(c) fn
fn'

which is the Wronskian of the given set of functions, is zero.

Comment 63.56. Theorem 63.55 gives a necessary condition for the

dependence of a set of functions. It says that if a set of functions is de-
pendent on I, then its Wronskian is identically zero on I. Unfortunately
it is not a sufficient condition. If the Wronskian of the set of functions is
identically zero on I, one cannot conclude that the set is linearly dependent
on I. It need not be. Here is an example. Let fl (x) = x 2 , !2(x) = xJxJ,
Fig. 63.57. Let the interval I include x = O. By (63.53), the Wronskian
of the functions is
(a)
X2 xJxJ I= x 3 JxJ' +x 2
JxJ - 2x 2 JxJ.
I2x xJxJ' + JxJ
776 OTHER EXISTENCE THEOREMS. WRONSKIAN. Chapter 12

When x > 0, Ixl = x. Therefore lxi' = x' = 1. The right side of (a)
thus becomes x 3 + x 3 - 2x 3 = O. When x = 0, Ixl = x = 0, .and again
the right side of (a) is zero. When x < 0, Ixl = -x and therefore
lxi' = -x' = -1. The right side of (a) now becomes x 3 ( -1) - x 3 +
2x 3 = O. We have thus shown that the Wronskian of the two given func-

Figure 63.57

tions is identically zero on an interval I which includes x = O. However,

we shall now show that the two functions are linearly independent on I.
Form with them the linear combination CIX 2 + c2xlxl = O. If x = 1,
then Cl + C2 = 0, and if x = -1, then Cl - C2 = O. The only solution
of this pair of equations is Cl = 0, C2 = O. Hence by Definition 19.1, the
functions x 2 and xlxl are linearly independent.
On the other hand, if the Wronskian of a set of functions is not equal to
zero for only one x in an interval I, then the set, by Theorem 63.55, must
be linearly independent on I. In short, no conclusion can be drawn as
regards thc linear dependence or independence of a set of functions on I,
from the fact that its Wronskian is identically zero on I. But if the
Wronskian of the set is not equal to zero for only one value of x in I, then
the set must be linearly independent on I.
In Exercise 63,5, you will find a necessary and sufficient condition for
the linear independence of a set of functions on an interval.

EXERCISE 63
1. Determine, for each of the following systems, whether it has one, none, or
infinitely many solutions.
(a) x + 4y = 1, (d) 2y = 0,
2x - y = 3. 2x+ 4y = O.
(b) x + y = 5, (e) -x +y = 0,
2x+ 2y = ID. x - y = O.
(c) 2x + y = -2, (f) 5x + y = 0,
-2x - y = 3. x - 3y = O.
Lesson 63-Exercise 777

2. Set up the Wronskian of each of the following set of functions.

(a) 1, eX. (b) 1, x. (c) x, eX. (d) x, xe X. (e) x, xe', 3x.
(f) sin x, sin 2x. (g) sin x, cos x. (h) ax, bx +
c, c O.
(i) ea" ebx, a b. (j) x - cos 2x, 2x - 2 cos 2x. (k) 1, X, x 2.
(I) x, eX, xe x, 2e x. (m) 1, e', e2x . (n) x, e', xe X. (0) 1, sin 2 x,
2
cos x, (p) sin x, sin x - ! sin 3x.
3

3. Determine, by Comment 63.56, which of the sets in 2 are linearly independent

over a suitable interval. For those sets whose Wronskian is zero, use Defini-
tion 19.1 to determine which are dependent and which are independent.
4. Prove that any two nonconstant functions that differ by a constant are
linearly independent.
S. Theorem 63.55 gives only a necessary condition for the linear independence
of a set of functions. The following theorem gives a necessary and sufficient
condition for the linear independence of a set of functions. A set of function8
/lex), h(x),"', fn(x), each continuous on an interval I: a x b, i8
linearly dependent on I, if and only if the determinant
b
f. /l2 dx f.b /lh dx f.b /lfn dx

b
f.b h2dx f.b hfndx = 0
G= f. h/l dx
. . .. ...... . . . . .......

f.b fnll dx f.b fnh dx f.b fn 2 dx

on I. The determinant G is called the Grammian of the set of functions.

Use this theorem to prove that
(a) x, 2x, 3x is a linearly dependent set on I: -1 x 1,
(b) eX, e2x is a linearly independent set on I: -1 x 1,
(c) x, Ixl is a linearly independent set on I: -1 x 1.
6. Prove Theorem 63.4.
7. Prove Theorem 63.42.

ANSWERS 63
1. (a) one. (b) Infinitely many. (c) None. (d) Infinitely many. (e) Infinitely
many. (f) x = 0, y = O.

2. (a) I :: I = eX. (d) I:

xe
xe X
X

+ e%
I= X
2 X
e.

(e) x xe% 3x
(b) I :I 1.
xe% + e% 3 = O.

0 xe% + 2e% 0
(f) ISin x sin2x I -2 sin 3x.
cos x 2 cos 2x
778 OTHER EXISTENCE THEOREMS. WRONSKIAN. Chapter 12

(g) ISinx cos.x I -1.

cosx -sm x

(h) I:x :x + C I= -ac.

az bz
(i) Ie e I = (b - a)e<a+b)%.
ae4Z bebz
(j) I x - cos 2x 2x - 2 cos 2x I= O. (0) O.

1+2sin2x 2(1+2sin2x)
2
(k) 1 x x (p) O.

o 2x = 2.
002
3. Linearly independent sets are (a), (b), (c), (d), (f), (g), (h), (i), (k), (m), (n).
Linearly dependent sets are (e), (j), (I), (0), (p)-see Example 21.32.

LESSON 64. Theorems About Wronskians and the Linear

Independence of a Set of Solutions of a
Homogeneous Linear Differential Equation.

In the previous lesson, we proved that if a set of functions is linearly

dependent on I, then their Wronskian is identically zero on I. We also
showed by an example, that the converse need not be true. If, however,
the set of functions are n solutions of a homogeneous linear differential
equation of order n, namely of

(64.1) fn(x)y(n) + fn_l(x)y(n-l> + ... + JI(x)y' + fo(x)y = 0,

where fo(x) , fl (x), ... , fn(x) are each continuous functions on an interval
I, and fn(x) r6 0 when x is in I, then, as we shall show (see Comment
64.15 below) the vanishing of their Wronskian is a necessary and sUfficient
condition for the linear dependence of the set on I.
Theorem 64.11. If each function Yb Y2, ••• , Yn is a solution of (64.1)
on an interval I: a x b, then the Wronskian of this set of functions is
'lither identically zero on I, or it is not zero for any x in I.
Proof. We shall prove the theorem only for n = 2. Call Yl,Y2 the
two solutions of (64.1) when n = 2. Since each function satisfies a linear
equation of the second order, their first and second derivatives exist.
By (63.53),
(a)
Lesson 64 LINEAR INDEPENDENCE OF A SET OF SOLUTIONS 779
Differentiating both sides of (a) with respect to x, we obtain

By hypothesis, each of the functions Yl and Y2 is a solution of (64.1) with

n = 2. Hence,
(c) !2(X)Yl" + fl(X)Yl' + fO(X)Yl = 0,
h(X)Y2" + h(X)Y2' + fO(X)Y2 = 0.

Since we assumed f2 r6 0, we may divide each equation in (c) by it to

obtain
(d)

Substituting these values in (b) gives, with the help of (63.53),

(e) W'(Yl,Y2; x) = -Yl t fOY2) + Y2 t fOYl)

- hfl (
YIY2' - Y2Yl ') it
= - h W( YIY2; x.
)
By (e), therefore
(f) W' + itf2 W = °
'

an equation which is linear in W. Its solution is

(64.12)

where F(x) = - f (fl dx/h). Since, by (18.86), the exponential function

eF(x) is never zero, we conclude from (64.12) that, if c = 0, then W =
°
for every x in I; if c r6 0, then W r6 for every x in I.
°
Remark. Formula (64.12) is known as Abel's formula.
Theorem 64.13. If each function Yl, Y2, •.. , Yn is a solution of (64.1)
in an interval I: a x b, and if

(64.14) W(YbY2, •.• , Yn; x) = 0,

at a point x = Xo in I, then the set of functions is linearly dependent on I.

Proof. We shall prove the theorem only for n = 2. By hypothesis,
there is a point Xo in I, for which
(a)

where Yl and Y2 are two solutions of (64.1) with n = 2. Hence, by (63.53)

780 OTHER EXISTENCE THEOREMS. WRONSKIAN. Chapter 12

and (a),
(b) Yl (XO) Y2(XO) I = o.
IYl'(XO) Y2'(XO)

Using the elements of the detenninant (b) as coefficients, we fonn the two
equations
(c) Yl (XO)Cl + Y2(XO)C2 = 0,
Yl'(XO)Cl + Y2'(XO)C2 = o.
Since the detenninant whose elements are the coefficients of Cl and C2 of (c)
is, by (b), zero, it follows, by Theorem 63.42, that there is an infinite num-
ber of pairs of nontrivial solutions of (c) for Cl and C2. With anyone such
pair, fonn the linear combination

(d)

where Yl and Y2 are the two given solutions. Therefore, by Theorem 19.3,
Ya(x) is also a solution of (64.1) with n = 2. By (d) and (c), Ya(xo) =
CIYl(XO) +
C2Y2(XO) = 0 and Ya'(xo) = CIYl'(XO) C2Y2'(XO) = o.+
We have thus proved that Ya(x) of (d) is a solution of (64.1), with
n = 2, i.e., it is a solution of

(e) !2(X)Y" + fl(X)Y' + fo(x)y = 0,

satisfying the conditions

(f) Ya(Xo) = 0, Ya'(Xo) = O.

But you can easily verify that y(x) == 0 is also a solution of (e) and that
this solution satisfies y(xo) == 0, y'(xo) = O. By the uniqueness Theorem
19.2 (which we still have not proved but shall soon, see Theorem 65.2)
the two solutions Ya(x) of (d) and y(x) = 0 are identical. Hence by (d)

(g)

We remarked after (c), that Cl and C2 are nontrivial solutions of (c) and
hence not both are zero. Therefore by Definition 19.1, the functions Yl
and Y2 are linearly dependent.
Comment 64.141. If the Wronskian of a set of solutions of (64.1) is
zero at a point Xo in I, then it is easy to show, by Theorem 64.13, that the
Wronskian is identically zero on I. We prove the statement, which is
equivalent to Theorem 64.11, for n = 2. Let Yll Y2 be two solutions
(64.1). Since W = 0 at a point Xo in I, the solutions, by Theorem 64.13
Lesson 64 LINEAR INDEPENDENCE OF A SET OF SOLUTIONS 781

are linearly dependent. Therefore

(a)
on I, where CI and C2 are not both zero. Differentiation of (a) gives
(b)
Since CI and C2 are not both zero, it follows by Theorem 63.42, that
(c)

But the determinant in (c) is, by (63.53), the Wronskian of YI and Y2.
NOTE. Although the above proof is much shorter than the one in
Theorem 64.11, the latter proof was given in order to obtain (64.12),
an equation which will be needed later.
Comment 64.15. Theorems 63.55 and 64.13 together give a necessary
and sufficient condition for the linear dependence of a set of functions that
are solutions of the linear differential equation (64.1). The first says that if
the set is dependent on I, then its Wronskian is == 0 on I. The second
says that if its Wronskian is zero for only one x in I (and therefore by
Theorem 64.11 it is == 0 on I) then the set is linearly dependent on [.
Hence you now have a test to determine whether a set of n solutions of a
homogeneous linear differential equation (64.1) is linearly dependent or
independent. If its Wronskian is zero for one x in I, the set is linearly
dependent; if its Wronskian is not equal to zero for one x in I, the set is
linearly independent.
Example 64.2. Show that if ml, m2, ma are distinct, then the set of
functions
(a)
which are three solutions of a third order homogeneous linear equation,
are linearly independent on I: - 00 < x < 00.
Solution. By (63.51), the Wronskian of the given functions is
(b)
m1
mle % m2 em2z maemaz
m2 2em2z ma 2emaz

By (63.11), (b) is equal to

(c) e(ml+m2+ma)Z[(m2ma2 -- mam22) - (mlma 2 - mam12)

+ 2
(mlm2 -
2
m2 m I )],
782 OTHER EXISTENCE THEOREMS. WRONSKIAN. Chapter 12

which can be written as

Since we have assumed mb m2, m3 are distinct and since, by (18.86), the
exponential function e(m 1 +m2+ ma)x is never zero, it follows that (d) cannot
be zero for any x. Hence the Wronskian (b) is not zero for any x. By
Comment 64.15, the set of solutions in (a) is therefore linearly independent
on I: - 00 < x < 00.

Remark. The above proof can be extended to show that if mlJ

m2, .. " mn are distinct, then the set of n functions Yl = em1x, Y2 =
em2x , ••• , Yn = emnx , each of which is a solution of (64.1), is linearly inde-
pendent on I: -00 < x < 00.

Example 64.21. Show that the set of functions

(a)

which are the solutions of y" - 2ay' + a2y = 0, are linearly independent
on I: - 00 < x < 00.

Solution. By (63.53), the Wronskian of the given functions is

(b)

Since e2lu: 0 for every x, the given functions are independent on I:

-00 <x < 00.

Example 64.22. Show that the second solution Y2(X) of (23.28) ob-
tained by the reduction of order method described in Lesson 23B, is inde-
pendent of the solution Yl (x).

Solution. Let F(x) = !l(X)/f2(X) and

f
e-fF(x)dx
(a) g(x) = 2 dx.
Yl

Then, by (23.28), Y2 = Ylg and, by (63.53), the Wronskian of the two

solutions is
(b) W(YI,Y2; x) = IYI YIg. I
Yl' Ylg' YI'g +
= Yl2g' + YIYl'g - Yl'Ylg
2
= Y1 g'.
Lesson 65 EXISTENCE THEOREM: LINEAR EQUATION, ORDER n 783

By (a) and Theorem 58.3

(c)

SUbstituting (c) in the right of (b), we obtain

(d)

Since the exponential function is never zero, it follows, by Comment 64.15,

that the given functions are linearly independent solutions on any interval
I over which the functions YI and Y2 are defined.

EXERCISE 64
1. Show that the set of functions 1, x, x 2 which are three solutions of y'" = 0
are linearly independent.
2. Show that each of the functions YI = sin x - ! sin 3x and Y2 = sin 3 x, is a
solution of y" +
(tan x - 2 cot x)y' = 0, but that y = CIYI +
C2Y2 is not
the general solution of the differential equation. Hint. Show that the two
functions are not linearly independent; see Exercise 63.3(p).
3. Prove that Y = CI sin x + C2 cos x is a general solution of Y" +
Y = O.
Hint. Show that each function satisfies the equation and that the two func-
tions are linearly independent.
4. Prove that Y = (CI + c2x)e" is a general solution of y" - 2y' +
y = O.
5. Prove that y = Cle" + C2e 2"+ c3xe 2" is a general solution of y'" - 5y" +
8y' - 4y = O.

LESSON 65. Existence and Uniqueness Theorem for the Linear

Differential Equation of Order n.
We are ready at last to prove the existence and uniqueness Theorem 19.2
for the linear differential equation of order n. Since on the interval I in
which we shall be interested, fn(x) -:;t= 0, we can divide (19.21) by fn(x) to
obtain a linear differential equation

(65.1) y(n)(x) + fn_l(x)y(n-l) + ... + fx(x)y' + fo(x)y = Q(x).

Theorem 65.2. If the coefficients fo(x), fx(x), ... , fn-t(x) and Q(x) in
the linear differential equation (65.1) are each continuous functions of x on a
common interval I, then for each point Xo in I and for each set of constants
ao, aI, ... , an-lJ there is one and only one function y(x) that satisfies (65.1)
and the initial conditions

(65.21) y(xo) = ao, y'(Xo ) = alJ ••• , y (n-l)(Xo ) = an-to

Remark 65.22. In Theorem 37.51, we gave a sufficient condition for

the existence of a power series solution of (65.1) satisfying (65.21). Compare
784 OTHER EXISTENCE THEOREMS. WRONSKIAN. Chapter 12

it with the above theorem, which gives a sufficient condition for the exist-
ence and uniqueness of a solution of (65.1) satisfying (65.21).
Proof of the Theorem. Assume y(x) is a solution of (65.1). We now
define new functions YI(X), Y2(X), ••• , Yn(X) by the following relations.
(65.23) y(X) = YI(X),
y'(x) = 1Il'(x) = Y2(X),
y"(x) = YI"(X) == Y2'(X) = Y3(X),

y(n-ll(X) = YI (n-ll(x) = Y2(n-2l(x) = Y3(n-3l(x)

= ... = Yn.!...I'(X) = Yn(X).

Differentiating the function in the last line of (65.23), we obtain

(65.231) y(nl(x) = YI (nl(x) = Y2(n-1l(X) = ... = Yn-I"(X) = Yn'(X).

By (65.231) and (65.23), we can therefore write (65.1) as

(65.24) Yn'(x) + fn-I(X)Yn(X) + ... + h(X)Y2(X) + fO(X)YI(X) = Q(x).

Solving (65.24) for Yn'(X), we obtain

(65.25) Yn'(X) = -fn-IYn - ••• - flY2 - fOYI + Q.

Hence because of (65.23) and (65.25), the set of functions Yl, Y2, ••. , Yn,
satisfies the system of linear first order equations
(65.26)
YI'(X) = Y2(X),
Y2'(X) = Y3(X),

Yn-I'(X) = Yn(X),
y,/(x) = -fn-I(X)Yn(X) - .•• - h(X)Y2(X) - fO(X)YI(X) + Q(x).
By (65.23), YI(XO) = y(xo), Y2(XO) = y'(xo), •. " Yn(Xo) = y(n-ll(xo).
We can therefore replace the initial conditions (65.21) by the conditions

Conversely if we start with the system of equations (65.26) and the

initial conditions (65.27), and define

(65.28) y(X) = YI(X),

then the relations in (65.23), (65.231), and (65.24) hold; the last equation
of (65.26) becomes identical with (65.1), and the initial conditions (65.27)
Leuon65 EXISTENCE THEOREM: LINEAR EQUATION, ORDER n 785

become (65.21). Hence if there exists a set of functions YI(X), Y2(X), •.• ,
Yn(X) that is a solution of the system (65.26) satisfying (65.27), then, by
(65.28), the function YI(X) of this set is the solution y(x) of (65.1) satisfy-
ing (65.21).
All t.he functions fo(x), flex), •.. ,fn-l(x), Q(x) in the last equation of
(65.26), are, by assumption, continuous on I. The functions Y2, Ys, •.. ,Yn
are also continuous since their derivatives exist. Therefore the hypotheses
of Theorem 62.3 are satisfied. Hence, by the theorem, there is, on I, or
10 if I is closed, one and only one set of particular solutions YI (x), Y2(X),
•. " Yn(X) that satisfies (65.26) and (65.27) where Xo is a point in 10.
And since YI (x) exists and is unique, it follows that its equal y(x) also
exists and is the unique solution of (65.1) and (65.21).
Example 65.3. Set up a system of linear first order equations for the
second order linear equation
(a) Y" - 3y' + 2y = x
and show that the solution YI (x) of the resulting system is the same as the
solution y(x) of (a).
Solution. Let y(x) be a solution of (a). As suggested in (65.23), we
define
(b) y(x) = YI(X),
y'(x) = YI'(X) = Y2(X).
Differentiating the last equation of (b), we obtain
(c) y"(x) = Yl"(X) = Y2'(X).

By (c) and (b), we can write (a) as

(d) Y2'(X) - 3Y2(X) + 2YI(X) = x.

Hence, by (b) and (d), we can replace the single equation (a) by the sys-
tem of linear first order equations
(e) YI'(X) = Y2(X),
Y2'(X) = 3Y2 - 2YI + X.
In operator notation we can write (e) as
(f) DYI - Y2 = 0,
2Yl + (D - 3)Y2 = X.
Multiplying the first equation in (f) by D - 3 and adding it to the second,
we obtain
(g) (D2 - 3D + 2)YI = x,
786 OTHER EXISTENCE THEOREMS. WRONSKIAN. Chapter 12

Comparing the second equation in (g) with (a), we see that the solution
Yl(X) of the system (f) will be the same as the solution y(x) of (a).
In Theorem 19.3, we stated that a homogeneous linear differential equa-
tion of order n has n linearly independent solutions, but proved this part
of the theorem only for the ca.se when the coefficients in the equation are
constants. We now prove the statement for nonconstant coefficients.

Theorem 65.4. If fo(x) , !I (x), ... ,fn(x) are each continuous functions
of x on a common interval I, then the homogeneous linear differential equation
(65.41) yen) + fn_l(x)y(n-ll + ... + fl(x)y' + fo(x)y = 0,
has n linearly independent solutions Yb Y2, ... , Yn.
Proof. We consider first a special set of solutions, gl(X), g2(X), •.. ,
gn(x) of (65.41), each satisfying respectively the initial conditions

(a) gl(XO) = 1, gl'(XO) = 0, gl "()

Xo = 0, " ' , gl (n-ll(Xo ) -- 0,
g2(XO) = 0, g2'(XO) = 1, g2 "()
Xo = 0 , " ' , g2 (n-l)( Xo ) -- 0,
g3(XO) = 0, g3'(XO) = 0, g3"(XO) = 1, ... , ga(n-ll(xo) = 0,

gn (Xo) = 0 , gn ' (Xo) = 0 , gn "()

Xo = 0, " ' , gn (n-ll()
Xo = 1,

where Xo is a point in I. By Theorem 65.2, each function gl, g2, ... , gn

exists. We form a linear c(')mbination of this special set of solutions, set it
equal to zero, and take its successive derivatives. We thus obtain
(b) Clgl (x)+ C2g2(X) + ... + cngn(X) = 0,
Clgl'(X) + C2g2'(X) + ... + cngn'(x) = 0,
Clgl"(X) + C2g2"(X) + ... + cngn"(x) = 0,

Clgl(n-ll(X) + c2g2(n-l)(x) + ... + cngn(n-l)(x) = o.

Let x = Xo. Then the first equation in (b), by the first column of values in
(a), simplifies to Cl = 0; the second equation in (b), by the second column
of values in (a), simplifies to C2 = 0, ... , the last equation in (b), by the
last column of values in (a), simplifies to Cn = O. We have thus shown that
each constant ClJ C2, ... , Cn in (b) and in particular each such constant in
the first equation of (b) is zero for Xo in I. Therefore by Definition 19.1,
the set of functions glJ g2, ... , gn is, on I, a linearly independent set.
Since each function is, by assumption, a solution of (65.41), they form
collectively a set of n linearly independent solutions of (65.41).
Let y(x) be a solution of (65.41), and let Xo be a point in I. By Theorem
19.3, the function
(c) h(x) = y(XO)gl(X) + y'(XO)U2(X) + ... + y(n-ll(xo)gn(X),
Lesson 65 EXISTENCE THEOREM: LINEAR EQUATION, ORDER n 787

is also a solution of (65.41) since it is a linear combination of n independent

solutions. [Remember the coefficients in (c) are constants.] Taking suc-
cessive derivatives of (c), we obtain

(d)
h'(x) = y(xo)gl'(x)+ y'(XO)g2'(X) + ... + y(n-O(Xo)gn'(X),
h"(x) = Y(XO)gl"(X) + y'(XO)g2"(X) + ... + y(n-ll(xo)gn"(x),

Let x = Xo. Then (c), by the first column of values in (a), simplifies to
h(xo) = y(xo); the first equation in (d), by the second column of values
in (a), simplifies to h'(xo) = y'(xo), •.. , the last equation in (d), by the
last column of values in (a), simplifies to hn-I(xo) = y(n-l)(xo). We have
thus shown that the two solutions hex) and y(x) of (65.41), and their first
n - 1 derivatives are equal when x = Xo. Hence, by the uniqueness
Theorem 65.2, the two solutions hex) and y(x) are identical. We can there-
fore replace hex) by y(x) in (c) to obtain

where Xo is a point in I. Since y(x) is an arbitrary solution of (65.41), it

follows that every solution of (65.41) can be expressed as a linear combi-
nation of the special set of n linearly independent solutions gl, g2, ••• , gn.
Let YI(X), Y2(X), •.. , Yn(X) be n solutions of (65.41), and let Xo be a
point in I. Therefore bY,(e),

(f) YI(X) = Yl(XO)gl(X) + Yl'(XO)(l2(X) + ... + Yl(n-O(Xo)(ln(x),

Y2(X) = Y2(XO)(ll(X) + Y2'(XO)(l2(X) + ... + Y2(n-1l(Xo)(ln(X),

are each valid equations. The set of functions (11, (12, ••• , (In are linearly
independent. Hence the system (f) can be solved for (II, (12, ••• , (In in
terms of Yb Y2, .•. ,Yn. This means, by Theorem 63.4, that the determi-
nant formed by the coefficients of the (I'S is not zero. Therefore,

(g) Yl(XO) Yl'(XO) YI (n-l)(xo)

Y2(XO) Y2'(XO) Y2(n-l)(xo) '¢ O.

A theorem of determinants now permits us to interchange rows and

788 OTHER EXISTENCE THEOREMS. WRONSKIAN. Chapter 12

columns to obtain from (g) the equivalent determinant

(h)
o.
YI (n-ll(XO) Y2 (n-I'(XO) •.• Yn (n-ll(xo)

By Definition 63.5, the determinant (h) is the Wronskian of a set of solu-

tions Yt, Y2, ... , Yn of (65.41) evaluated at x = Xo, where Xo is a point
in I. And since this Wronskian is not zero for one x in I, the set of solu-
tions, by Comment 64.15, is linearly independent on I. We have thus
proved the existence of a set of n linearly independent solutions of (65.41).

Theorem 65.5. If Yt, Y2, ... , Yn are n linearly independent solutions

of (65.4-1), then
(65.51) Yc = CIYI + C2Y2 + ... + CnYn
is a general solution of (65.4-1), i.e., every solution of (65.4-1) can be obtained
from (65.51) by a proper choice of the constants Ct, C2, ... , Cn.
Proof. We prove the theorem for n = 2. Let Yl1 Y2 be two linearly
independent solutions of
(a) y" + h(x)y' + fo(x)y = o.
Therefore, by Theorem 19.3,
(b)

is also a solution of (a). Assume Ya is a solution of (a) not obtainable from

(b). In the proof of Theorem 64.11, we showed that for each two solutions
of (a), equation (64.12) is valid. Therefore, by (64.12),

(c) W(Yt,Y2; x) = C12eF(xl,

W(Yt,Ya; x) = ClaeF(xl,

W(Y2,Ya; x) = C2aeF(x>,

where C12, CIa and C2a are constants. Multiplying the first equation in
(c) by Ya, the second by Y2, the third by Yt, and adding the resulting
equations, we obtain

(d) YaW(Yt,Y2; x) + Y2W(YI,Ya; x) + y IW(Y2,Ya; x)

= (YaC12 + Y2Cla + YIC2a)e F (xl.
By (63.53), the left side of (d) is
(e) Ya(YIY2' - Y2YI') + Y2(YIYa' - YaYI') + YI(Y2Ya' - YaY2'),
Lesson 65 EXISTENCE THEOREM: LINEAR EQUATION, ORDER n 789

which reduces to zero. Hence the right side of (d) must also be zero. And
since, by (18.86), eF(x) :;6 0 for every x, it follows that

(f)

The solutions Yb Y2 are, by assumption, linearly independent. Therefore,

by Comment 64.15, the first Wronskian in (c) is not zero, i.e., Cl2 :;6 O.
Hence dividing (f) by C12, we obtain
(g)
Cl3 C23
Ya = --Y2 - - Y I ,
Cl2 Cl2

from which we see that Ya is obtainable from (b) by a proper choice of the
constants CI and C2 in (b). Hence our assumption that Ya is not obtain-
able from (b) is false.

Theorem 6.5.6. Let Yb Y2, ..• , Yn be n linearly independent solutions

of the homogeneous linear differential equation (65.41), and let
(65.61) Yc = CIYI + C2Y2 + ... + CnYn
be its complementary function. Let YP be a particular solution of the non-
homogeneous linear differential equation (65.1).
Then
(65.62) Y = Yc + yp
is a general solution of (65.1), i.e., (65.62) includes every solution of (65.1).
Proof. By Theorem 65.2, at least one particular solution YP of (65.1)
exists. Assume g is a solution of (65.1) not obtainable from (65.62). Since
YP and g are each solutions of (65.1), we have
(a) Yp(n.) + fn_l(x)Yp(n-I> + ... + fo(x)yp = Q(x),
g(n) + fn_I(X)g<n-l) + ... + fo(x)g = Q(x).
Subtracting the first equation from the second, we obtain
(b) (g - Yp)(n) + fn-l(x)(g - Yp)(n-l) + ... + fo(x)(g - Yp) = O.
Hence Y = g - YP is a solution of the homogeneous linear equation
(65.41). Therefore, by Theorem 65.5, (g - Yp) can be obtained from
(65.61) by a proper choice of the constants Cll C2, ... ,Cn. Therefore,
(c) g - YP = alYI + a2Y2 + ... + anYn,
g = YP + (aIYI + a2Y2 + ... + anYn).
We have thus shown that the solution g can be obtained from (65.62).
Hence our assumption that g is not obtainable from (65.62) is false.
790 OTHER EXISTENCE THEOREMS. WRONSKIAN. Chapter 12

All statements made in Theorems 19.2, 19.3, and Comment 19.41, in

connection with a linear differential equation of order n, have now been
proved.

EXERCISE 65
1. Set up an equivalent system of linear first order equations for each of the
following. Then show that solution Yl(X) of the resulting system will be
identical with the solution y(x) of the given equation.
(a) Y" - Y' +
3y = x2 + 1. (c) y" + +2y' 3y = tan x.
(b) y" +3y' - 2y = eX + x. +
(d) y'" - 5y" 8y' - 4y = O.
2. Set up an equivalent system of linear first order equations for each of the
following.
(a) y" - 4y'+ 2y = x + x2, y(O) = 1, y'(O) = -1.
(b) y'" + 3y" - y' + y = eX, y(O) = 1, y'(O) = 0, y"(O) = 1.
(c) y(4) - 3y'" + y" + 5y' - 6y = 0, y(O) = 0, y'(O) = 1, y"(O) = 2,
y"'(O) = 3.

ANSWERS 65
1. (a) Yl' (x) = Y2, (c) Yl'(X) = Y2,
Y2'(X) = + +
Y2 - 3Yl x 2 1. Y2'(X) = -2Y2 - 3Yl + tan x.
(b) Yl'(X) = Y2, (d) Yl'(X) = Y2,
Y2' (x) = -3Y2 + + +
2Yl eX x. Y2'(X) = Y3,
Y3'(X) = 5Y3 - 8Y2+ 4Yl.
2. (a) Yl'(X) = Y2,
Y2'(X) = 4Y2 - 2Yl + +
X x 2, Yl(O) = 1, Y2(0) = -1.
(b) Yl'(X) = Y2,
Y2'(X) = Y3,
Y3'(X) = -3Y3 +
Y2 - Yl +
eX, Yl(O) = 1, Y2(0) = 0, Y3(0) = 1.
(c) Yl'(X) = Y2, Y3'(X) = Y4,
Y2'(X) = Y3, Y4'(X) = 3Y4 - Y3 - 5Y2 +
6Yl,
Yl(O) = 0, Y2(0) = 1, Y3(0) = 2, Y4(0) = 3.
 Bibliography

Intended for students who wish to advance beyond the material of this text.
Bessel Functions-G. N . Watson. A Treatise on the Theory of Bessel
Functions. New York: Macmillan Co., 1944.
Celestial Mechanics-F. R. Moulton. An Introduction to Celestial
Mechanics. New York: Macmillan Co., 1914. (Dover, 1970)
General Theory-E. L. Ince. Ordinary Differential Equations. London:
Longmans, Green & Co., 1927. (Dover, 1956)
Laplace Transforms-R. V. Churchill. Modern Operational Mathematics
in Engineering. New York: McGraw-Hill, 1944.
Legendre Functions-E. W. Hobson. The Theory of Spherical and
Ellipsoidal Harmonics. Cambridge, England: Cambridge University
Press, 1931.
Mechanics-W. F. Osgood. Mechanics. New York: Macmillan Co., 1946.
Numerical Methods-F. B. Hildebrand. Introduction to Numerical
Analgsis. New York: McGraw-Hill, 1956.
Perturbation Theory-F. R. Moulton. An Introduction to Celestial
Mechanics. New York: Macmillan Co., 1914. (Dover, 1970)

791
Index

Note: A boldfaced number refers to a problem in the exercises.

Abel's formula, 779 Beams, (cqntinued)

Absolute value of a complex number, 198 condition of continuity of slope of, 387,
Acceleration, definition of, 138 388
due to gravity, 140 elastic curve of, 384
in polar coordinates, 461 modulus of elasticity of, 385
in rectangular coordinates, 460 neutral surface of, 384
Accretion problems, 122-124 simple, 384
Adams' method, y' = f(x,y), 681 simply supported, 384
second order equation, 712-3 Young's modulus of, 385
Air resistance, 141 Beats, 346
Airplane problems, see Pursuit curves Bending of beams, see Beams
Algebra, of complex numbers, 200-201 Bending moment, 384
fundamental theorem of, 197, 198 Bernoulli equation, 95-96
of operators, 255-259 Bessel equation, 579, 583, 609-622
Amplification ratio, 363 equations leading to a, 615-619
Amplitude, 317 of index 1/2, 579
damped,351 of index zero, 583
modulated, 345 Bessel functions, 611-622
slowly varying, 345 of the first kind, Jk(X), 611-6lfi
Analog computation, 375-376 Jo(x), J 1 (x), 614
Analytic function, 537, 548 J_k(x),613
on an interval, 537 modified Bessel functions of the first
Analytic geometry, review of, 62 kind,619
Angular momentum, 472 properties of Bessel functions of the
Apogee, 490 first kind, Jk(x), 619-622: integral,
Approximations, to e, 639-9, 644-8 620-622; orthogonal, with respect to
to integrals, see Numerical methods weight function x, 622; zeros, 619-620
to log 2, 639-10, 644-9 of the second kind, -Nk (x),612
to pi (11"), 639-8, 644-7 Beta function, 306
by Picard method, see Picard's method Binomial series, 599
of successive approximations Biological problem, 447-451
by polynomial interpolation, see Poly- Body faIling in water, 144
nomial interpolation Bounded region, 14
Arbitrary constant!!, 129, 130 Bridge, over gorge, 362
Arc length of a curve, 111 troops crossing a, 362
Arc tan x, expansion of, 533, 537 Buoyancy, 335
Archimedes' principle, 335
Argument (Arg) of a complex number, 199 Cable, hanging, 507-512
Arithmetical errors, 636 suspension, 507-514
Associative law for operators, of addition, Calorie, 185
256 Cantilever beam, 384
of multiplication, 257 Capacitance, 370
Atmospheric pressure, 186-188 Capacitor, 369
Axis of a beam, 383 in series, 455-3
Carbon-14 test, 5
Backward differences (v, del), 668-670 Catenary, 510
Beams, axis of, 383 Cave, prehistoric, 2
bending of, 383-389 Center of attraction, 470
bending moment of, 384 Central force, 380, 470
cantilever, 384 See also, Particle in motion in space
condition of ('ontinuity of curve of, 387 subject to a central force
795
796 INDEX

Centrifugal force, 380 Continuing formulas, Adams, 681, 712-3

Centripetal force, 380 improvement of polygonal method,
Chain around a cylinder, 188-189
Chain sliding from a table, 190 polygonal, see Polygonal method
Characteristic equation, 212-220, 444 Runge-Kutta, see Runge-Kutta method
definition of, 212 Taylor series, 64H52
roots of: real and distinct, 213-214; real third degree interpolating polynomial,
but some multiple, 214-217; imagi- 676
nary, 217-220 Continuing methods, 632
Charcoal problem, 2 Continuous function, 732
Charge, electric, 370 on an interval, 730
Circuit, closed electric, 369 at a point, 730
See also, Electric circuits in a region, 731
Clairaut equation, 757-760 Convergence, of improper integral, 292,
geometric problems giving rise to, 761 294
parabolic reflector, 759-760 interval of, 532
Closed electtic circuit, 369 of power series, 531-532
Coefficient, of friction, 160 ratio test for, 532
of inductance, 370 of a sequence of functions, 728, 732
of resistance, 348 uniform: of a sequence of functions,
of sliding biction, 160 729, 732; of a series of functions,
of static friction, 160 732-733
Combination, linear, 205 Coordinates of a point, 6
Commutative law for operators, of addi- Corrector formulas, 680
tion, 256 fifth degree polynomial, 676
of multiplication, 258 Simpson, 675
Complementary function, 210 sixth degree polynomial, 677
Complete elliptic integral of the first kind, three-eights rule, 681
333 trapezoidal, 675
Complex electric circuits, 451-455 Weddle's rule, 677
Complex functions, 201-203 Corrector methods, 632
exponential, 201 Cosine, hyperbolic, 203
hyperbolic, 203 Cosine x, cosine z, expansions of, 201
trigonometric, 201 Coupled springs, 440-443
Complex numbers, 197-201 Critically damped, 350
absolute value of, 198 Cumulative errors, 636, 690
algebra of, 200-201 Current, electric, 369, 370
argument (Arg) of, 199 induced, 185
conjugate of, 198 steady state, 372
definition of, 197 transient, 372
imaginary part of, 197 Curvature, radius of, 528
polar form of, 199 Curve, elastic, 384
real part of, 197 Curves, envelopes of a family of, 748-
rectangular form of, 199 754
Complex variables, 201-203 of pursuit, 168-175,523-525
solution by use of, 230 Cycloid, 335
Compound interest, 128-5, 6
Computation, analog, 375-376 D, differential operator, 251
Condition of continuity, of a curve, 387 V- n , inverse operator, 270
of a slope, 387, 388 Damped, amplitude, 351
Conductivity, thermal, 185 critically, 350
Conic, equation in polar coordinates, frequency, 351
483 harmonic motion, 348-353
Conjugate of a complex number, 198 motion, 347-353, 359-364; definition,
Conservation, of angular momentum, 347; forced, 359-364; free (damped
472 harmonic), 348-353
of energy, 329, 475 period,351
Conservative field, 473 periodic, 351
Construction of a table of Laplace trans- Damping factor, 351
forms, 302-306, 309-311 Decomposition problems, 131-132
 INDEX 797

Decrement, logarithmic, 356 Differential equations, (continued)

Degenerate systems, 413-415 ordinary, 20
Dependence, see Linear dependence of partial,20
functions particular solution of, 35
Determinant, definition of, 399, 770, 771 recognizable exact, 80-82
Grammian, 777 Riccati, 97, 247-22,23,24
of a system of equations, 400 with separable variables, 51-55
theory of, 770-774 singular, point of, 43: solution of, 34
Wronskian, 774 solution of a, definition, 22
Diagram, lineal element, 39 systems of, see System of equations
Differences, backward, 668-670 Tschebyscheff, 589
finite, 659-661 variables separable, 51-55
forward, 659-668 Dilution problems, 122-124
tables of, 660, 662, 663, 670 Dirac Hunction, 344
Differential, exact, 72 Direction field, 39
operator, 251 construction of a, 38-41
total,72 isoclines of a, 40-41
Differential, of a function, 47-51 Distributive law of multiplication for
of one independent variable, 48 operators, 258
of two independent variables, 50 Divergence of improper integrals, 292
Differential equations, Bernoulli, 95-96 Domain of independent variable, 9, 11
Bessel, 579, 583, 609-622
Clairaut, 757-760 e, an approximation to, 639-9, 644-8
definition of ordinary, 20 eX, eZ, expansion of, 201
Euler, 247 Elastic curve, 384
exact, 70-78, 248-26: definition of, 73; Elastic helical springs, 324-326
recognizable, 80-82 parallel, 330-13
existence of solutions of, see Existence in series, 330-14
and uniqueness theorems Elasticity, modulus of, 385
explicit solution of a, 22 Electric circuits, complex, 451-455
finding a, from its n-parameter family, first order, 184-185,377-7, 10
31-33 simple, 369-375
finding an integrating factor, 84-90, Electromotive force, 369
94-95, 248: for first order, 84-90; induced, 185
for linear first order, 94-95; for linear Element, line, 39
second order, 248 lineal, 39
of first order, see First order equations of a set, 6
Gauss's, 587 Elementary functions, 17
general solution of, 28-37: definition, 35 Eliminant, 750
Hermite, 607 Ellipse, equation in polar coordinates, 483
with homogeneous coefficients, 57-60 Elliptic integral of the first kind, complete,
homogeneous in x, 505 333
hypergeometric, 589 incomplete, 334
implicit solution of, 24 Energy, inverse square law, 487-488
initial conditions, 36 kinetic, 329, 475
integral curve of, definition, 39 law of conservation of, 329, 475
integrating factor of, see Integrating potential, 329, 473
factors Envelopes of a family, of curves, 747-754
Laguerre, 624-630 definition of, 748
Legendre, 586, 591-605, 606-13 eliminant of, 750
with linear coefficients, 63-68 of solutions, 754-757
linear: of first order, 91-95; of higher Equation, indicial, 574
order, see Linear differential equations Equilibrium position, 317
miscellaneous first order, 101-103 Equivalent triangular system, 405, 406,
multiplicity of solutions of, 28 416
n-parameter family of solutions of, 30 Error, in fifth degree polynomial, 679
nonlinear, see Nonlinear equations function, 670
numerical solutions of, see Numerical in an improvement of polygonal
methods method, 644-3
order of, 21 in Milne method, 688-689
798 INDEX

Error, (continued) First order equations, (continued)

in Newton's interpolation formulas, exact, 70--78
670-671 existence and uniqueness theorem for,
in polygonal method, 636--638 134-743
in polynomial interpolation, 670--671, with homogeneous coefficients, 57--60
679 integrating factors of, 84-90, 94-95
in Runge-Kutta, 657-658 linear, 91-95
in Simpson's rule, 679--684 with linear coefficients, 63-68
in Taylor series, 537, 649--652, 653--4 miscellaneous, 101-103
in third degree polynomial, 679 perturbation theory, 713-715
in trapezoidal rule, 679, 683 problems giving rise to, see Problems,
in Weddle's rule, 679 first order equations
Errors, arithmetical, 636 recognizable exact, 80--82
cumulative, 636, 690 Riccati, 97, 247-22, 23, 24
formula, 636, 690 with separable variables, 51-55
general comment on, 636, 690--691, 703 solution of y' = f(x,y): by numerical
rounding off, 636, 690 methods, see Numerical methods; by
Escape velocity, 148 Picard's method, 720--723; by series
Euler equation, 247 method, 548-553
Exact differential, 72 First order, processes, 137
Exact differential equation, 70--78, systems, see System of first order
248-25 equations
definition of, 73, 248-25 Flow through an orifice, 183
necessary and sufficient condition for, Force, central, see Particle in motion in
73,248-25 space subject to a central force
recognizable, 80--81 centrifugal, 380
solution of, 73, 76 centripetal, 380
Existence and uniqueness theorems, 720 damping, 349, 350, 351
first order equation y' = f(x,y), 734-743 electromotive, 369
linear equation of order n, 783-786 field, 473
nonlinear equation of order n, 765-767 field of, 473
system: of n first order equations, 763- frictional, 160
765; of n linear first order equations, function, 473
768-770 of gravity, 140
Explicit solution, 21-23 impressed (forcing function), 338, 360
definition of, 22 induced electromotive, 185
Exponential complex function, 201 intermittent, 344
Exponential shift. theorem, for inverse inversely proportional: to cube of dis-
operators, 277 tance, 521-522; to square of distance,
for polynomial operators, 260 481-488, 494
moment of, 381
F-2 region of atmosphere, 155 proportional to distance, 476-479
Factorial function (n!), 306-308, 596 Forced motion, with damping, 359-364
Factors, integrating, see Integrating undamped, 338-342
factors Forcing function, 338
of polynomial operators, 258 Formula errors, 636, 679
Falling bodies, see Vertical motion Forward differences, (6, delta), 659-
Faltung theorem, 303 668
Family, of curves, envelopes of, 748-754 Fourth degree interpolating polynomial,
n-parameter, of solutions, 30, 31-33: 681-2
envelopes of, 754-756 Fractions, partial, 283-284
Field, of force, 473 of inverse operators, 287, 289
conservative, 473 Free motion, damped, 347-353
direction, 39 undamped, 313-329
slope, 39 See also, Simple harmonic motion
Fifth degree interpolating polynomial, Frequency, damped, 351
676, 678, 679 impressed, 339
Finite differences, 659-661 natural (undamped), 318, 319
First order equations, Bernoulli, 95-96 normal,443
Clairaut, 757-760 resonance, 373, 443
 INDEX 799

Frequency, (continued) Fundamental theorem, of algebra, 197,198

undamped resonant 341 of calculus, 70
Friction, coefficient or' 160
sliding, 160 ' Gamma function, 306-309
static, 160 definition of, 307
Frictional force, 160 Gauss's equation, 587
General solution, of a differential equation
Frobenius, method, 572; see also, Solution
28-37 '
about a regular singularity
series, 572 definition of a, 35
of a first order linear equation, 93
f(a), definition of, 11
f(b,y), meaning of, 18-10 of a homogeneous linear equation 210
788 ' ,
f(x,a), definition of, 13
of a nonhomogeneous linear equation
Function of one independent variable
6-11, 14-17, 48 ' 210,789 '
definition of, 6, 9 Geometric problems leading to Clairaut
equation, 761 '
definition of f(a), 11
differential of a, 48 first order equation, 107-111
elementary, 17 special types of second order equations
528-530 '
implicit, 14-17
range of, 9 Grammian, 777
Function of two independent variables, Graphical solutions, 38-44, 424-438
11-14, 18,50,57-58 Gravitation, Newton's universal law of
491 '
definition of, 11
definition: of f(x,a), 13; of f(b y) 18- Gravitational constant, 139 491
10 ' , Gravity, force of, 140 '
differential of a, 50 specific, 153
domain of definition of a 11 Growth problems, 131-132

range of, 11
n:
homogeneous, of order 57,58 Halley's comet, 492
Hanging cable, 507-514
Functions, analytic, 537: on an interval
motion, damped, 348-353
537 '
SImple, see Simple harmonic motion
Bessel, see Bessel functions undamped, see Simple harmonic motion
beta, 306 Harmonic oscillators, 323-329, 377
complementary, 210
elastic helical spring, 324-326
complex, see Complex functions simple pendulum, 327-329
continuous, 730, 731: on an interval
Heat, specific, 130
,730; at a point, 730; in a region, 73i
steady state flow of, 185-186
DIrac 8, 344 Helical spring, see Elastic helical springs
elementary, 17
Hermite, equation, 607
elliptic, 333, 334 polynomials, 607
error, 670
Homogeneous order of a, 57, 58
factorial (n!), 306-308, 596
Homogeneous hnear equationd see Linear
force, 473
differential equations '
forcing, 338, 360
Homogeneous in x, equation, 505
gamma, 306-309
Hooke's law, 324
homogeneous, of order n, 57, 58 Horizontal motion, 160-162
hypergeometric, 587
Hyperbolic, cosine, 203
Legendre, 594: of second kind Qk(X)
5m ' , sine, 203
tangent, 203
linear dependence of, see Linear
Hypergeometric, equation 587
dependence of functions function, 587 '
linear independence of see Linear
series, 587
independence of
orthogonal, 602 i (current), 369, 370
polynomial interpolating 662 i (imaginary unit), 197
remainder, 670 ' Identically zero, meaning of 775
sequence of, 728, 729 732 Imaginary number, part of a complex
series of, 732-733 ' number, 197
unit impulse, 344 pure, 197
vector point, 473 Impedance of a circuit, 372
800 INDEX

Implicit function, 14-17 Inverse operators, 268-282

definition of, 16 definition of, 269, 270
Implicit solution of an equation, 24-27 exponential shift theorem for, 277
definition of, 24 meaning of, 269-271
Impressed, frequency, 339 partial fraction expansion of, 287
force (forcing function), 338 series expansion of, 272-277
Improper integrals, convergence of, 292, solution of linear equation by, 272-282,
294 288-291
divergence of, 292 Inverse square law, 481-488, 491, 494
Improvement of polygonal method, Irregular singularity, 572
641-643 Isoclines of a direction field, 40-41
Impulse function, unit, 344 Isogonal trajectory, 115-117
Impulsive response of a system, 344 definition of, 115
Inclined motion, 164-166 Isothermal surface, 185
Incomplete elliptic integral of the first
kind,334 Kepler's laws, 492
Independence of functions, see Linear Kinetic energy, 329, 475
independence of functions Kirchhoff's, first law, 451
Independent solutions of linear equation, second law, 370
number of, 208, 786 Kutta, see Runge-Kutta formulas
Indicial equation, 574
Induced, current, 185
electromotive force, 185 Laguerre equation, 624-630
Inductance, coefficient of, 370 Laguerre polynomials, Lk(X), 625-630
mutual,454 properties of, 627-630: analogue of
Induction, mathematical, 250 Rodrigue's formula, 627-629; inte-
Inductor, 369 gral property, 629-630; orthogonal
Inertia, 381 with respect to weight function e-",
moment of, 381 630
rotational, 381 Laplace transforms, construction of table
Infinite series, see Series of, 302-306, 309-311
Initial conditions, 33-37 definition of, 294
definition of, 36 Faltung theorem, 303
number of, 36 inverse, 296
Input of a system, 341, 360, 372 properties of, 295-296
Integrable combinations, 81 solution of a linear equation by,
Integral, elliptic of the first kind, 296-302
complete, 333 solution of a system by, 418-420
incomplete, 334 tables of, 306, 310
Integral, improper, 292, 294 Law of conservation, of angular momen-
convergence of, 292 tum, 472
divergence of, 292 of energy, 329, 475
Integral, Riemann, 70 Law of the mean, 731
Integral curve, definition of, 39 of universal gravitation, 491
Integrals, approximations to, see Legendre equation, 586, 591-605, 606-13
Numerical methods comment on solution of, 593-594
Integrating factors, 82-90 functions, 594: of the second kind,
definition of, 82 Qk(X), 597
of first order equation, 84-90, 94-95 Legendre polynomials, Pk(X), 597-605
of second order linear equation, 248 properties of, 598-605: coefficients in
Interest problems, 126-127 binomial series expansion, 598-600;
Intermittent force, 344 orthogonal property, 602-604; other
Interpolation, see polynomial inter- integral properties, 604-605; recur-
polation sion formula, 601; Rodrigue's for-
Interval,6 mula, 602; values of Pn(O), P n(I),
Interval of convergence, 532 P n (-I),60Q-601
Inverse Laplace transform, definition of, Libby, Dr. Willard F., 5
296 Line elements, 39
linear property of, 296 Lineal element diagram, 39
 INDEX 801

Lineal elements, 39 Linear independence, of functions, 205,

Linear coefficients, 62--69 775-Comment 63.56,777-5
Linear combinations, definition of, 205 definition of, 205
Linear dependence, of functions, 205, of solutions, 781-Comment 64.15
775-Comment 63.56,777-5 Linear property of, inverse operators, 296
definition of, 205 Laplace transformation, 295
of solutions, 781-Comment 64.15 polynomial operators, 253
Linear differential equations, Bessel, see Linearization of first order systems, 424-
Bessel equation 438
characteristic equation of, 212-220: defi- Lipschitz condition, 731, 734, 764, 766
nition, 212; roots imaginary, 217-220; Liquid, flowing through an orifice, 183
roots real and distinct, 213-214; roots rotating in a cylinder, 193-194
real but some multiple, 214-217 Logarithmic decrement, 356
complementary function of, 210
definition of, 92, 196 Maclaurin series, 535
Euler, 247 Magnification ratio, 363
exact, 248 Mass, variable, 191-193
existence theorem for, see Existence and Mathematical induction, 250
uniqueness theorems Mean, law of the, 731
first order, 91-95: definition of, 92; Method of, Frobenius, 572; see also, Solu-
general solution of, 93; integrating tion about a regular singularity
factor for, 94; solution of, 92 reduction of order, 242-246 ;
form of solution of, 211-220 undetermined coefficients, 221-230
fundamental theorems for, 207, 208 variation of parameters, 233-240
Gauss's, 587 Milne method, comment on error in, 688-
general solution of, 93, 210, 788 689
Hermite, 607 for second order equation, 710-711
homogeneous, with constant coeffi- for system of two first order equations,
cients, 211-220: definition of, 196; 703
with nonconstant coefficients, 241- for third order equation, 712--4
246 for y' = f(x,y), 684--689
hypergeometric, 587 Modulated amplitude, 345
integrating factors for, 94, 248 Modulus, of elasticity, 385
Laguerre, 624--630 Young's, 385
Legendre, see Legendre equation Moment, bending, 384
n linearly independent solutions of, 208, of force, 381
786 of inertia, 381
nonhomogeneous, 221-246: with con- Momentum, angular, 472
stant coefficients, 221-240; definition Motion, of a complex system, 189-191
of, 196; with nonconstant coefficients, damped: forced, 359-364; free (damped
236-237,241-246 harmonic),348-353
ordinary point of, 570 forced: damped, 359-364; undamped,
reduction to a system, 784-785 338-342
singularity of, 570: irregular, 572; regu- free: damped (damped harmonic), 348-
lar,571 353; undamped, see Simple harmonic
solution of, by: complex variables, 230- motion
231; inverse operators, 272-282, 288- horizontal, 160-162
291; Laplace transforms, 296-302; inclined, 164-166
method of Frobenius, 572-584; par- of a particle: on a circle, 316-317; in
tial fraction expansion of inverse space, see Particle in motion in space
operators, 288-291; polynomial op- subject to a central force; on a
erators, 262-265; power series, 537- straight line, 138-166, 314-316
546; reduction of order, 242-246; un- period of, 318, 333, 351, 477
determined coefficients, 221-230; va- planetary, 491-492
riation of parameters, 233-240 of a projectile, 463-465
systems of, see System entries simple harmonic, see Simple harmonic
Tschebyscheff, 589 motion
uniqueness theorem for, see Existence stable, 339
and uniqueness theorems steady state, 361
802 INDEX

Motion, (continued) Newton's interpolation formulas, 663-

transient, 361 671: backward, 669; forward, 667
undamped: forced, 338-342; free, see perturbation theory, 713-718: first order
Simple harmonic motion equation, 713-715; second order equa-
unstable, 340 tion, 716-718
vertical, see Vertical motion Picard's, see Picard's method of succes-
Multiplicity of solutions, 28-31 sive approximations
Mutual inductance, 454 polygonal, 632-638
polynomial interpolation, see Poly-
Natural (undamped) frequency, 318, 319 nomial interpolation
Neutral surface, 384 reducing and increasing h, 692-694
Newton's, first law of motion, 138 Runge-Kutta, see Runge-Kutta for-
interpolation formulas, 663-671: back- mulas
ward, 669; error in, 670, 671; for- for a second order equation, 707-711,
ward,667 712-3, 715-718
law of universal gravitation, 491 series, 645-652
proof of inverse square law, 494-495 Simpson's rule, 677, 678, 684
second law'pf motion, 138, 459 sixth degree polynomial, 677
Nonhomogeneous linear equation, defini- starting methods, 632
tion of, 196 See also, Starting methods
Nonlinear equations, existence and summary and an example, 694-701
uniqueness theorem for, 765-767 for a system of: three first order equa-
numerical solution of, see Numerical tions, 726-2; two first order equa-
methods tions, 702-706, 723-726
reduction to a system, 766-767 Taylor series, 645-652
series solution of, 562-567 third degree polynomial, 676, 678
See also, Special types of second for a third order equation, 712-4
order equations three-eighths rule, 681
Normal, coordinates, 444 trapezoidal rule, 675, 678, 682
frequencies, 443 Weddle's rule, 677, 678, 691, 703
n-parameter family of solutions, definition
of,30 Oceanic pressure, 186-188
finding a differential equation from an, Operator, differential, 251
31-33 See also, Inverse operators; Poly-
Number, complex, 197 nomial operators
pure imaginary, 197 Order, of a differential equation, 21
real, 197 of a homogeneous function, 57, 58
Numerical methods, 631-718 Ordinary differential equation, definition
Adams', 681, 712-3 of, 20
choosing size of h, 691-692 Ordinary point, of y' = f(x,y), 43, 744-747
continuing, 632 of a linear differential equation, 570
corrector, 632 Orifice, flow through an, 183
decreasing h, 692-694 Orthogonal, definition of, functions, 602
difference tables, 660, 662, 663, 670 property: of Bessel functions, 622; of
errors, see Error; Errors Laguerre polynomials, 630; of
fifth degree polynomial, 676, 678, 691, Legendre polynomials, 602-604
703 Orthogonal trajectory, in polar coordi-
finite differences, 659-661 nates, 118
for first order equation y' = f(x,y), 632- in rectangular coordinates, 117
658, 681-6, 684-688, 690-701, 713- Oscillator, harmonic, 323-329, 377
715 Output of a system, 341, 360, 372
fourth degree polynomial, 681-2 Overdamped, 349
illustrative example and summary, 694-
701 Parabolic reflector, 759-760
improvement of polygonal method, 641- Parameters, number of, 30
643 variation of, 233-240
increasing and reducing h, 692-694 Parasite problem, 447-451
Milne, see Milne method Partial differential equation, meaning of, 20
 INDEX 803

Partial fraction expansion, 283-284 Point, ordinary, see Ordinary point

of inverse operators, 287, 289 singular, 43, 744-747
Particle, moving, in a circle, 316-317 Polar form of a complex number, 199
in a plane, see Particle in motion in Polygonal method, 632-638
space subject to a central force comment on errors in, 636-638
in rotating tube, 380 an improvement of, 641-643
on a straight line, 138-166, 314-316 Polynomial interpolation, 661-684
Particle in motion in space subject to a Adams', 681, 712-3
central force, 47()--496, 521-522 error in, 670-671, 679
force inversely proportional to cube of fifth degree polynomial, 676, 678, 679
distance, 521-522 function, 662
force inversely proportional to square Newton's: (backward) formulas, 669;
of distance, 481-488, 494: determin- (forward) formulas, 667
ing constants of integration, 484-486: Simpson's rule, 675, 678, 679, 684
energy considerations, 487-488; plan- sixth degree polynomial, 677
etary motion, 491-492 third degree polynomial, 676, 678, 679
force proportional to distance, 476-479 three-eighths rule, 681
Kepler's laws, 492 trapezoidal rule, 675, 678, 679, 682
law of conservation: of angular momen- Weddle's, 677, 678, 679
tum, 471-472; of energy, 475 Polynomial operators, 251-265
moves in plane, proof that particle, algebraic properties of, 255-259
47()--471 associative law: of addition, 256;
period of, 477, 492 of multiplication, 257
satisfies law of conservation: of angular commutative law: of addition, 256;
momentum, 471-472; of energy, 475 of multiplication, 258
special central force problem, 521-522 definition of, 251
sweeps out equal areas in equal times, distributive law of multiplication, 258
proof that particle, 472 exponential shift theorem for, 260:
Particular solution of a differential equa- corollaries to, 261
tion, 33-37 factoring of, 258
definition of, 35 linear property of, 253
Pendulum, simple, 327-329 of order n, 251
actual period of, 333 P(D + a), definition, 259
amplitude of, 328 P(D)y, definition, 252
period independent of amplitude, product of h(x) by, 256
334-34 product of two, 257
weight of wire not negligible, 331-29 solution: of linear equations by, 262-
Perigee, 490 265; of systems of linear equations
Period, 318, 477 by, 398-417
actual, of a simple pendulum, 333 sum of two, 255
damped,351 Polynomials, Hermite, 607
Perturbation method, 713-718 interpolation by, see Polynomial inter-
first order equation, 713-715 polation
second order equation, 715-718 Laguerre, 625-630
Phase and phase angle, 318 597-605
Picard's method of successive approxima- Tschebyscheff, 589
tions, 720-726 Potential, 473
for first order equations, 720-723 energy, 329, 473
for a system: of three first order equa- Power series, see Series; Solution by series
tions, 726-9; of two first order equa- methods
tions, 723-726 Pressure, atmospheric and oceanic, 186-
Plane analytic geometry, a review of, 188
62-63 Principal value of 8, 199
Plane motion of a particle, see Particle in Principle of superposition, 211, 254
motion in space subject to a central Problem involving a centrifugal force, 380
force Problems, first order equations, 107-195,
Plane motion of a projectile, 463-465 377-7,10
Planetary motion, 491-492 accretion, 122-124
804 INDEX

Problems, first order equations, (continued) Problems, linear second order equations,
atmospheric pressure, 186-188 (continued)
body falling in water, 144 simple harmonic motion, 313-329
chain around a cylinder, 188-189 See also, Simple harmonic motion
decomposition, 131-132 straight line, particle moving on, 314-
dilution, 122-124 316
electric circuit, 184-185, 377-7, 10 twisting bodies, 383
first order processes, 137 undamped motion, 313-342: examples
flow of heat, steady state, 185-186 of, 323-329; forced, 338-342; free,
flow through an orifice, 183-184 313-321
geometric, 107-111 Problems, special types of second order
growth, 131-132 nonlinear equations, 506-530
horizontal motion, 160-162 central force, 521-522
inclined motion, 164-166 geometric, 528-530
interest, 126-127 pursuit, 523-525
isogonal trajectory, 115-117 special central force, 521-522
moon, 151--6,7,8; 156-38,39 suspension cable, 507-514
motion of a complex system, 189- Problems, system of equations, 440-492
191 biological, 447--451
oceanic pressure, 186, 187 central force, see Particle in motion in
orthogonal trajectories, 117-120 space subject to a central force
pursuit curves, 168-175 electrical, 451-455
raindrop, 143 mechanical, 440-443
relative pursuit curves, 177-182 particle in motion in space, see Particle
rocket motion, 191-193 in motion in space subject to a central
rope around a cylinder, 188-189 force
rotation of a liquid in a cylinder, plane motion: of a projectile, 463-465;
193-194 of a particle, see Particle in motion in
second order processes, 134-137 space subject to a central force
steady state flow of heat, 185-186 planetary motion, 491-492
straight line motion, 138-166 projectile in plane, 463-465
temperature, 129-130 Product of two operators, 257
variable mass, 191-193 Projectile, motion of a, 463-465
vertical motion, 139-151 P.roof by mathematical induction, 250
Problems, linear second order equations, Properties of, Bessel function of first kind,
313-389 619-622
bending of beams, 383-389 Laguerre polynomials, 627-630
circle, particle moving on circumference Laplace transforms, 295-296
of,316-317 Legendre polynomials, 598-605
damped harmonic motion, 347-353: polynomial operators, see Polynomial
definition of, 347 operators
damped motion: forced, 359-364; free, Pure imaginary number, 197
347-353 Pursuit curves, 168-175, 523-525
elastic helical spring, 324-326: in paral- relative, 177-182
lel, 330-13; in series, 330-14
electric circuit, simple, 369-375 Radioactive material, 2
forced motion with damping, 359-364 Radius of curvature, 528
forced undamped motion, 338-342 Raindrop problem, 143
free damped motion, 347-353 Range of a function, 9, 11
free undamped motion, 313-329 Ratio test for convergence, 532
See also, Simple harmonic motion Real number system, 197
particle moving: on circumference of a Real part of a complex number, 197
circle, 316-317; on a straight line, Rectangular form of a complex number,
314-316 199
pendulum, simple, see Pendulum Recursion formula, 577
problem involving a centrifugal force, for Bessel equation, 610
380 for Hermite polynomial, 607-18
rolling bodies, 381-383 for Laguerre equation, 625
simple electric circuit, 369-375 for Legendre polynomial, 601
 INDEX 805

Reduction of order method, 241-246 Series, (continUed)

Region, bounded, 14 differentiation of a, 534
F-2 of atmosphere, 155 for e= and e', 201
simply connected, 71 equality of, 534
Regular singularity, 571 expansion of inverse operators by,
at 00,586 272-277
Relative pursuit curve, 177-182 Frobenius, 572: solution of homo-
Remainder function, 670 geneous linear equation by, 572-584;
Resistance, coefficient of, 348 solution of nonhomogeneous linear
electric, 370 equation, 585-18
Resistors, 369 of functions, 732-733: converges uni-
in series, 454-2 formly, 732, 733
Resonance, 373, 443 hypergeometers, 587
undamped, 341 interval of convergence of, 532
Resonant frequency, undamped, 341 Maclaurin, 535
Response of a system, impulsive, 344 partial sums of, 732
Riccati equation, 97, 247 power, 531
Riemann-integrable, 70, 731 ratio test for convergence of, 532
Riemann integral, 70 for sin x and sin z, 201
Rocket motion, 191-193 solutions by, 8ee Solution by series
Rodrigue's formula, 602 methods
Rolling bodies, 381-383 for tan x, 742
Roots, in conjugate pairs, 204-9 Taylor, 535: with remainder, 537
imaginary, 217-220 Set, elements of a, 6
of indicial equation, 574-584 meaning of a, 5-6
real and distinct, 213-214 Simple beam, 384
real but some multiple, 214-217 Simple electric circuit, 369-375
Rope around a cylinder, 188-189 Simple harmonic motion, 313-329
Rotating tube, particle in a, 380 amplitude, 317
Rotation of a liquid, 193-194 definition of, 314
Rotational inertia, 381 description of the motion, 315, 317
Rounding off errors, 636, 690 elastic helical spring, 324-326
Runge-Kutta formulas, comment on equilibrium position, 317
error in, 657-658 examples of bodies moving in, 323-329
for second order equation, 709 frequency, natural (undamped), 318,
for system of two first order equations, 319
702-703 harmonic oscillators, 323-329
for y' = f(x,y), 653-658 natural (undamped) frequency, 318, 319
particle moving: on circumference of a
Satellites, 488-491, 496 circle, 316-317; on a straight line,
Second order equations, 500-504, 707-711, 314-316
712-1,3 pendulum, 327-329
numerlcal solution of, 707-711: by period,318
Adams' method, 712-3; error in, phase and phase angle, 318
712-1; by Milne's method, 710-711; simple pendulum, 327-329
by Runge-Kutta method, 709 spring, helical, 324-326
special types of nonlinear, 500-504 Simply, connected region, 71
Second order processes, problems, 134- supported beam, 384
137 Simpson's rule, 675, 678, 679, 684
Separable variables, 52 Simultaneous equations, see entries under
Separated variables, 52 System
Sequence of functions, convergence of a, Sine, hyperbolic, 203
728,732 Sine x, sin z, expansions of, 201
uniform convergence of a, 729, 732 Singular point of y' - f(x,y), 43, 744-747
Series, binomial, 599 Singular solution, 34
convergence of, 531, 532 Singularity of a linear equation, 570
for cos x and cos z, 201 irregular, 572
defines a function, 533 regular, 571
determination of coefficients of a, 534 regular at co, 586
806 INDEX

Sixth degree interpolating polynomial, 677 Solutions of differential equations,

Sliding friction, 160 (continued)
Slope field, 39 by polynomial operators, 262-265
Slowly varying amplitude, 345 by power series, 537-553, 562-567
Solution about a regular singularity, 572- by reduction of order, 242-246
584, 585, 586 by separation of variables, 51-55
of a homogeneous linear equation, where by series, 537-553, 562-567
roots of indicial equation: differ by singular, 34
an integer, 578-582; do not differ by steady state, 361, 372
an integer, 574-578; equal, 583-584 by substitution and other means, 101
at <Xl, 586 by successive approximations, 720-726
of a nonhomogeneous linear equation, transient, 361, 372
585-18 by undetermined coefficients, 221-230
Solution by series methods, 537-584 uniqueness of, see Existence and unique-
by Frobenius series, see Solution about ness theorems
a regular singularity by variation of parameters, 233-240
by Taylor series: of a linear equation, Wronskian of, 781-Comment 64.15
537-546; of a nonlinear equation of Solve a differential equation, meaning of,
order n, 562-567 i of a system of first 22,81
order equations. 555-559 i of a system Special central force problem, 521-522
of linear first order equations, 559- Special types of second order equations,
562; of y' = f(x,y), 548-553 500-504
Solution of a system of equations, 394-420 x absent, 503-504
of a first order linear system, 396-397 x and y' absent, 500-501
of a first urder system, 394-396 y absent, 502-503
of a linear system with constant coeffi- Specific, gravity, 153
cients, 398-420: by Laplace trans- heat, 130
forms, 418-420; by polynomial opera- weight, 512
tors, 398-417 Speed, of escape, 148
by numerical methods, 702-706 terminal, 142
by Picard's method, 723-726, 726-9 Spring, helical, see Elastic helical springs
by series method, 555-562 Spring constant, 324
See also, each heading under Sputnik,491-7
System Stable motion, 339
Solutions of differential equations, by Starting formula, 680
choice of method, 99 Starting methods, 632
by complex variables, 230-231 polygonal method, see Polygonal
definition of, 22 method: an improvement of, 641-643
existence of, see ExiRtence and unique- Runge-Kutta, see Runge-Kutta
ness theorems formulas
explicit, 21-23 Taylor series, 645-652
general, 28-37: definition of, 35 Static friction, 160
implicit, 24-27 Steady state, current, 372
by integrating factors, 82-90, 94 flow of heat, 185-186
by inverse operators, 272-282, 288-291 motion, 361
by Laplace transforms, 296-302 solution, 361, 372
by lineal element diagram, 39 Stiffness coefficient of a spring, 324
linear combination of, 208 Stiffness constant, torsional, 383
by method of Frobenius, 572-584 Straight line motion, 138-166,314-316
multiplicity of, 28-31 Subnormal, 112-1
number of independent, 208, 786 Substitution, solving an equation by, 101
by numerical methods, 631-718 Subtangent, 112-1
by operators, 262-265, 272-282, 288-291 Successive approximations, see Picard's
n-parameter family of, 30, 31-33 method of successive approximations
by partial fraction expansion of inverse Sum of two operators, 256
operators, 288-291 Superposition principle, 211, 254
particular, 35 Surface, isothermal, 185
by Picard's method of successive neutral, 384
approximations, 720-726 Suspension cable, 507-514
 INDEX 807

System of equations, meaning of a solution Taylor series, (continued)

of a, 393 solution by, 645-652: comment on error
See also, System of first order equa- in, 649-652, 653--4; creeping up
tions; System of linear equations process, 646-647; direct substitution,
with constant coefficients; Sys- 646; See also, Solution by series
tem of linear first order equations Tchebycheff, see Tschebyscheff's
System of first order equations, definition Temperature problems, 129
of a, 394 Terminal velocity, 142
errors in numerical solution of, 703 Thermal conductivity, 185
existence and uniqueness theorem for, Third degree interpolating polynomial,
763-765 676, 678, 679
linearization of, 424-438 Three-eighths rule, 681
Milne formula for a, 703 Time constant, 351
Picard's method of solution of a, 723- Torque, 381
726,726-9 Torsional, resistance constant, 499
problems giving rise to 8., see Problems, stiffness constant, 383, 498
system of equations Total differential, 72
Runge-Kutta formula for a, 702-703 Tractrix, 176- 6
series method of solution of a, 555-562 Trajectories, 115-120
solution of a: by numerical methods, isogonal, 115-117: definition of, 115
702-706; by Picard's method, 723- orthogonal, 117-120: definition of, 117;
726, 726-9; by series methods, 555- in polar coordinates, 118; in rectangu-
562 lar coordinates, 117
special types of second order equations Transform, Laplace, see Laplace trans-
giving risl! to a, 500-504 forms
System of linear equations with constant Transient, current, 372
coefficients, definition of a, 398 motion, 361
degenerate, 413-445 solution, 361, 372
determinant of, 399, 400 Trapezoidal rule, 675, 678, 679, 682
equivalent triangular, 405-413 Triangular system, equivalent, 405, 406,
general solution of a, 398, 399 416
problems giving rise to a, see Problems, Trigonometric functions of complex
system of equations numbers, 201
solution of a: by Laplace transforms, Tschebyscheff's, equation, 589
418-420; by operators, 398-417 polynomials, 589
of three equations, 415-420 Twice differentiable function, definition
System of linear first order equations, of,749
396-397 Twisting bodies, 383
definition of a, 396
existence and uniqueness theorem for a, Undamped frequency, 318-319
768-770 Undamped motion, 313-342
problems leading to a, see Problems, examples of, 323-329
system of equations free, see Simple harmonic motion
solution of a: lly numerical methods, forced, 338-342
702-706; by Picard's method, 723- Undamped, resonance, 341
726, 726-9; by series methods, 555- resonant frequency, 341
562 Underdamped,351
See also, System of first order equa- Undetermined coefficients, 221-230
tions; System of linear equations Uniform convergence, of a sequence of
with constant coefficients functions, 729, 732
of a series of functions, 732-733
Table, chain sliding from, 190 Uniqueness theorems, see Existence and
Tables, of differences, 660, 662, 663, 670 uniqueness theorems
of Laplace transforms, 306, 310 Unit impulse function, 344
Tangent, hyperbolic, 203 Universal gravitation, Newton's law of,
Tan x, expansion of, 742 491
Taylor series, 535 Unstable motion, 340
with remainder, 537
reviewof,531-537 Vanish identically, definition of, 775
808 INDEX

Variable mass, 191-193 Voltage drop, 369

Variables, separable, 52
separated, 52 Weddle's rule, 677, 678, 679, 691, 703
Variation of parameters, 233-240 Weight, specific, 512
Varying amplitude, slowly, 345 Wronskian, Abel's formula for, 779
Vector, point-function, 473 definition of, 774
quantities, 164, 459 determinant of, 774
Velocity, of escape, 148 determining linear dependence and inde-
formulas in polar coordinates, 461 pendence by, 775-Comment 63.56
terminal, 142 determining linear dependence and inde-
Vertical motion, 139-151 pendence of solutions by, 781-Com-
body near earth's surface: air resistance, ment 64.15
142-145; no resistance, 140-141 theorems about, 778-780
body far from earth's surface, 146-151
Volt, 369 Young's modulus, 385