1

0.8
0.6
0.4
0.2
z
0
-0.2 -10
-0.4 -5
-10 0 x
-5
0 5
y 5 10
10

CORRAL’S
VECTOR
CALCULUS
Michael Corral
and Anton Petrunin
Corral’s Vector Calculus

Michael Corral and Anton Petrunin

About the author:
Michael Corral is an Adjunct Faculty member of the Department of Mathematics at
Schoolcraft College. He received a B.A. in Mathematics from the University of California
at Berkeley, and received an M.A. in Mathematics and an M.S. in Industrial & Operations
Engineering from the University of Michigan.

This text was typeset in LATEX 2ε with the KOMA-Script bundle, using the GNU Emacs text
editor on a Fedora Linux system. The graphics were created using MetaPost, PGF, and
Gnuplot.

Copyright ©2016 Anton Petrunin.

Permission is granted to copy, distribute and/or modify this document under the terms of the
GNU Free Documentation License, Version 1.2 or any later version published by the Free
Software Foundation; with no Invariant Sections, no Front-Cover Texts, and no Back-Cover
Texts. A copy of the license is included in the section entitled “GNU Free Documentation
License”.
Preface
This book covers calculus in two and three variables. It is suitable for a one-semester course,
normally known as “Vector Calculus”, “Multivariable Calculus”, or simply “Calculus III”.
The prerequisites are the standard courses in single-variable calculus (also known as Cal-
culus I and II).
The exercises are divided into three categories: A, B and C. The A exercises are mostly
of a routine computational nature, the B exercises are slightly more involved, and the C
exercises usually require some effort or insight to solve. A crude way of describing A, B and
C would be “Easy”, “Moderate” and “Challenging”, respectively. However, many of the B
exercises are easy and not all the C exercises are difficult.
Answers and hints to most odd-numbered and some even-numbered exercises are provided
in Appendix A.
There are a few exercises that require the student to write a computer program, for ex-
ample, the Monte Carlo method for approximating multiple integrals, in Section 4.4. The
code samples in the text are in the Java programming language, hopefully with enough
comments so that the reader can figure out what is being done even without knowing Java.
Those exercises do not mandate the use of Java, so students are free to implement the solu-
tions using the language of their choice. While it would have been simple to use a scripting
language like Python, and perhaps even easier with a functional programming language
(such as Haskell or Scheme), Java was chosen due to its ubiquity, relatively clear syntax,
and easy availability for multiple platforms.
This book is released under the GNU Free Documentation License (GFDL), which allows
others to not only copy and distribute the book but also to modify it. For more details, see
the included copy of the GFDL. So that there is no ambiguity on this matter, anyone can
make as many copies of this book as desired and distribute it as desired, without needing a
permission.
This book can be downloaded at https://github.com/anton-petrunin/calc3book; the older,
original version by Michael Corral, can be also obtained from http://www.mecmath.net.

iii
Contents
Preface iii

1 Vectors in Euclidean Space 1

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Vector Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3 Dot Product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.4 Cross Product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.5 Lines and Planes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
1.6 Elementary surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
1.7 Curvilinear Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

2 Curves 56
2.1 Vector-Valued Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
2.2 Arc Length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
2.3 Curvature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

3 Functions of Several Variables 74

3.1 Functions of Two or Three Variables . . . . . . . . . . . . . . . . . . . . . . . . . . 74
3.2 Partial Derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
3.3 Tangent Plane to a Surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
3.4 Directional Derivatives and the Gradient . . . . . . . . . . . . . . . . . . . . . . . 87
3.5 Maxima and Minima . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
3.6 Numerical Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
3.7 Lagrange Multipliers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

4 Multiple Integrals 114

4.1 Double Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
4.2 Double Integrals Over a General Region . . . . . . . . . . . . . . . . . . . . . . . . 119
4.3 Triple Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
4.4 Numerical Approximation of Multiple Integrals . . . . . . . . . . . . . . . . . . . 130
4.5 Change of Variables in Multiple Integrals . . . . . . . . . . . . . . . . . . . . . . . 135
4.6 Application: Center of Mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
4.7 Application: Probability and Expected Value . . . . . . . . . . . . . . . . . . . . . 147

5 Line and Surface Integrals 155

5.1 Line Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155

iv
 Contents v

5.2 Properties of Line Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164

5.3 Green’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
5.4 Surface Integrals and the Divergence Theorem . . . . . . . . . . . . . . . . . . . . 179
5.5 Stokes’ Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
5.6 Gradient, Divergence, Curl and Laplacian . . . . . . . . . . . . . . . . . . . . . . . 202
5.7 Other coordinate systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207

Bibliography 213

Appendix A: Answers and Hints to Selected Exercises 215

GNU Free Documentation License 218

History 226

Index 227
1 Vectors in Euclidean Space
1.1 Introduction
In single-variable calculus, the functions that one encounters are functions of a variable
(usually x or t) that varies over some subset of the real number line (which we denote by R).
For such a function, say, y = f ( x), the graph of the function f consists of the points ( x, y) =
( x, f ( x)). These points lie in the Euclidean plane, which, in the Cartesian or rectangular
coordinate system, consists of all ordered pairs of real numbers (a, b). We use the word
“Euclidean” to denote a system in which all the usual rules of Euclidean geometry hold. We
denote the Euclidean plane by R2 ; the “2” represents the number of dimensions of the plane.
The Euclidean plane has two perpendicular coordinate axes: the x-axis and the y-axis.
In vector (or multivariable) calculus, we will deal with functions of two or three variables
(usually x, y or x, y, z, respectively). The graph of a function of two variables, say, z = f ( x, y),
lies in Euclidean space, which in the Cartesian coordinate system consists of all ordered
triples of real numbers (a, b, c). Since Euclidean space is 3-dimensional, we denote it by R3 .
The graph of f consists of the points ( x, y, z) = ( x, y, f ( x, y)). The 3-dimensional coordinate
system of Euclidean space can be represented on a flat surface, such as this page or a black-
board, only by giving the illusion of three dimensions, in the manner shown in Figure 1.1.1.
Euclidean space has three mutually perpendicular coordinate axes ( x, y and z), and three
mutually perpendicular coordinate planes: the x y-plane, yz-plane and xz-plane (see Figure
1.1.2).

z z
c
P(a, b, c)
yz-plane

b y xz-plane y
0 0
a x y-plane
x x

Figure 1.1.1 Figure 1.1.2

The coordinate system shown in Figure 1.1.1 is known as a right-handed coordinate

system, because it is possible, using the right hand, to point the index finger in the positive

1
2 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

direction of the x-axis, the middle finger in the positive direction of the y-axis, and the thumb
in the positive direction of the z-axis, as in Figure 1.1.3.

Figure 1.1.3 Right-handed coordinate system.

An equivalent way of defining a right-handed system is if you can point your thumb up-
wards in the positive z-axis direction while using the remaining four fingers to rotate the
x-axis towards the y-axis. Doing the same thing with the left hand is what defines a left-
handed coordinate system. Notice that switching the x- and y-axes in a right-handed
system results in a left-handed system, and that rotating either type of system does not
change its “handedness”. Throughout the book we will use a right-handed system.
For functions of three variables, the graphs exist in 4-dimensional space (R4 ), which we
can not see in our 3-dimensional space, let alone simulate in 2-dimensional space. So we can
only think of 4-dimensional space abstractly. For an entertaining discussion of this subject,
see the book by A BBOTT.1
So far, we have discussed the position of an object in 2-dimensional or 3-dimensional space.
But what about something such as the velocity of the object, or its acceleration? Or the
gravitational force acting on the object? These phenomena all seem to involve motion and
direction in some way. This is where the idea of a vector comes in.
You have already dealt with velocity and acceleration in single-variable calculus. For
example, for motion along a straight line, if y = f ( t) gives the displacement of an object after
time t, then d y/ dt = f ′ ( t) is the velocity of the object at time t. The derivative f ′ ( t) is just a
1 One thing you will learn is why a 4-dimensional creature would be able to reach inside an egg and remove the

yolk without cracking the shell!

 1.1 Introduction 3

number, which is positive if the object is moving in an agreed-upon “positive” direction, and
negative if it moves in the opposite of that direction. So you can think of that number, which
was called the velocity of the object, as having two components: a magnitude, indicated
by a nonnegative number, preceded by a direction, indicated by a plus or minus symbol
(representing motion in the positive direction or the negative direction, respectively); that
is, f ′ ( t) = ±a for some number a ≥ 0. Then a is the magnitude of the velocity (normally called
the speed of the object), and the ± represents the direction of the velocity (though the + is
usually omitted for the positive direction).
For motion along a straight line (which is a 1-dimensional space) the velocities are also
contained in that 1-dimensional space, since they are just numbers. For general motion
along a curve in 2- or 3-dimensional space, however, velocity will need to be represented by
a multidimensional object which should have both a magnitude and a direction. A geomet-
ric object which has those features is an arrow, which in elementary geometry is called a
“directed line segment”. This is the motivation for how we will define a vector.

Definition 1.1. A (nonzero) vector is a directed line segment drawn from a point P (called
its initial point) to a point Q (called its terminal point), with P and Q being distinct
−−→
points. The vector is denoted by PQ . Its magnitude is the length of the line segment,
° −−→ °
denoted by ° PQ °, and its direction is the same as that of the directed line segment. The
zero vector is just a point, and it is denoted by 0.

To indicate the direction of a vector, we draw an arrow from its initial point to its terminal
point. We will often denote a vector by a single bold-faced letter (for instance, v) and use
the terms “magnitude” and “length” interchangeably. Note that our definition could apply to
systems with any number of dimensions (see Figure 1.1.4 (a)–(c)).
z
y R
−−→S Q
−−→Q Q R
P

S
−−→
PQ

v P R y
x
0
−−→ −−→ 0 −−→
RS PQ RS v
P
S 0 R P Q x S x
(a) One dimension (b) Two dimensions (c) Three dimensions

Figure 1.1.4 Vectors in different dimensions.

A few things need to be noted about the zero vector. Our motivation for what a vector is
included the notions of magnitude and direction. What is the magnitude of the zero vector?
We define it to be zero; that is, k 0 k = 0. This agrees with the definition of the zero vector as
just a point, which has zero length. What about the direction of the zero vector? A single
point really has no well-defined direction. Notice that we were careful to only define the
4 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

direction of a nonzero vector, which is well-defined since the initial and terminal points are
distinct. Not everyone agrees on the direction of the zero vector. Some contend that the
zero vector has arbitrary direction, some say that it has indeterminate direction (that is, the
direction can not be determined), while others say that it has no direction. Our definition of
the zero vector, however, does not require it to have a direction, and we will leave it at that.2
Now that we know what a vector is, we need a way of determining when two vectors are
equal. This leads us to the following definition.

Definition 1.2. Two nonzero vectors are equal if they have the same magnitude and the
same direction. Any vector with zero magnitude is equal to the zero vector.

By this definition, vectors with the same magnitude and direction but with different initial
points wouldpbe equal. For example, in Figure 1.1.5 the vectors u, v and w all have the same
magnitude 5 (by the Pythagorean Theorem). And we see that u and w are parallel, since
they lie on lines having the same slope 21 , and they point in the same direction. So u = w,
even though they have different initial points. We also see that v is parallel to u but points
in the opposite direction. So u 6= v.
y

u
4
3

2 v
w
1
x
0 1 2 3 4

Figure 1.1.5

So we can see that there are an infinite number of vectors for a given magnitude and
direction, those vectors all being equal and differing only by their initial and terminal points.
Is there a single vector which we can choose to represent all those equal vectors? The answer
is yes, and is suggested by the vector w in Figure 1.1.5.
Unless otherwise indicated, when speaking of “the vector” with a given magnitude and
direction, we will mean the one whose initial point is at the origin of the coordinate
system.

Thinking of vectors as starting from the origin provides a way of dealing with vectors in
a standard way, since every coordinate system has an origin. But there will be times when
2 In the subject of linear algebra there is a more abstract way of defining a vector where the concept of “direction”

is not really used. See A NTON and R ORRES.

 1.1 Introduction 5

it is convenient to consider a different initial point for a vector (for example, when adding
vectors, which we will do in the next section).
Another advantage of using the origin as the initial point is that it provides an natural
correspondence between a vector and its terminal point.

Example 1.1. Let v be the vector in R3 whose initial point is at the origin and whose ter-
minal point is (3, 4, 5). Though the point (3, 4, 5) and the vector v are different objects, it is
convenient to write v = (3, 4, 5). When doing this, it is understood that the initial point of v
is at the origin (0, 0, 0) and the terminal point is (3, 4, 5).

z z

P(3, 4, 5) v = (3, 4, 5)

y y
0 0

x x
(a) The point (3,4,5) (b) The vector (3,4,5)

Figure 1.1.6 Correspondence between points and vectors.

Unless otherwise stated, when we refer to vectors as v = (a, b) in R2 or v = (a, b, c) in R3 ,

we mean vectors in Cartesian coordinates starting at the origin. Also, we will write the zero
vector 0 in R2 and R3 as (0, 0) and (0, 0, 0), respectively.
The point-vector correspondence provides a way to check if two vectors are equal, without
having to determine their magnitude and direction. Similar to seeing if two points are the
same, you are now seeing if the terminal points of vectors starting at the origin are the same.
For each vector, find the (unique!) vector it equals whose initial point is the origin. Then
compare the coordinates of the terminal points of these “new” vectors: if those coordinates
are the same, then the original vectors are equal. To get the “new” vectors starting at the
origin, you translate each vector to start at the origin by subtracting the coordinates of
the original initial point from the original terminal point. The resulting point will be the
terminal point of the “new” vector whose initial point is the origin. Do this for each original
vector then compare.
−−→ −−→
Example 1.2. Consider the vectors PQ and RS in R3 , where P = (2, 1, 5),Q = (3, 5, 7), R =
−−→ −−→
(1, −3, −2) and S = (2, 1, 0). Does PQ = RS ?
−−→
Solution: The vector PQ is equal to the vector v with initial point (0, 0, 0) and terminal point
Q − P = (3, 5, 7) − (2, 1, 5) = (3 − 2, 5 − 1, 7 − 5) = (1, 4, 2).
−−→
Similarly, RS is equal to the vector w with initial point (0, 0, 0) and terminal point S − R =
(2, 1, 0) − (1, −3, −2) = (2 − 1, 1 − (−3), 0 − (−2)) = (1, 4, 2).
6 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

−−→ −−→
So PQ = v = (1, 4, 2) and RS = w = (1, 4, 2).
−−→ −−→
∴ PQ = RS

z Q
−−→
PQ (3, 5, 7)
P
(2, 1, 5)
−−→
Translate PQ to v

v=w
(1, 4, 2)
0 y
−−→
Translate RS to w S
(2, 1, 0)
−−→S
R R
(1, −3, −2) x

Figure 1.1.7

Recall the distance formula for points in the Euclidean plane:

For points P = ( x1 , y1 ), Q = ( x2 , y2 ) in R2 , the distance d between P and Q is:

q
d= ( x2 − x1 )2 + ( y2 − y1 )2 . (1.1)

By this formula, we have the following result:

−−→
For a vector PQ in R2 with initial point P = ( x1 , y1 ) and terminal point
−−→
Q = ( x2 , y2 ), the magnitude of PQ is:
° −−→ ° q
° PQ ° = ( x2 − x1 )2 + ( y2 − y1 )2 . (1.2)

Finding the magnitude of a vector v = (a, b) in R2 is a special case of formula (1.2) with
P = (0, 0) and Q = (a, b) :

For a vector v = (a, b) in R2 , the magnitude of v is:

p
kvk = a2 + b 2 . (1.3)

To calculate the magnitude of vectors in R3 , we need a distance formula for points in

Euclidean space (we will postpone the proof until the next section):
 1.1 Introduction 7

Theorem 1.1. The distance d between points P = ( x1 , y1 , z1 ) and Q = ( x2 , y2 , z2 ) in R3 is:

q
d= ( x2 − x1 )2 + ( y2 − y1 )2 + ( z2 − z1 )2 . (1.4)

The proof will use the following result:

Theorem 1.2. For a vector v = (a, b, c) in R3 , the magnitude of v is:

p
kvk = a2 + b 2 + c 2 . (1.5)

Proof: There are four cases to consider: p p

Case 1: a = b = c = 0. Then v = 0, so k v k = 0 = 02 + 02 + 02 = a2 + b2 + c2 .
Case 2: exactly two of a, b, c are 0. Without loss of generality, we assume that a = b = 0 and
c 6= 0 (the other two possibilities are handled in a similar p manner).
p Then v =p(0, 0, c), which
is a vector of length | c | along the z-axis. So k v k = | c| = c = 0 + 02 + c2 = a2 + b2 + c2 .
2 2

Case 3: exactly one of a, b, c is 0. Without loss of generality, we assume that a = 0, b 6= 0

and c 6= 0 (the other two possibilities are handled in a similar manner). Then vp= (0, b, c),
which is a vector 2 2
p p in the yz-plane, so by the Pythagorean Theorem we have k v k = b + c =
2 2 2 2
0 +b +c = a +b +c . 2 2

Case 4: none of a, b, c are 0. Without loss of generality, we can as- z

Q(a, b, c)
sume that a, b, c are all positive (the other seven possibilities are
handled in a similar manner). Consider the points P = (0, 0, 0), v
c
Q = (a, b, c), R = (a, b, 0), and S = (a, 0, 0), as shown in Figure
1.1.8. Applying the Pythagorean Theorem to the right trian- y
gle △PSR gives |PR |2 = a2 + b2 . A second application of the P
a 0
Pythagorean Theorem, this time to the right triangle △PQR , S
q p b R
gives k v k = |PQ | = |PR |2 + |QR |2 = a2 + b2 + c2 . x
This proves the theorem. QED Figure 1.1.8

Example 1.3. Calculate the following:

−−→ 2
(a) The magnitude of the vector PQ in R with P = (−1, 2) and Q = (5, 5).
° −−→ ° p p p p
Solution: By formula (1.2), ° PQ ° = (5 − (−1))2 + (5 − 2)2 = 36 + 9 = 45 = 3 5.
2
(b) The magnitude of the vector v = (8 p, 3) in R . p
Solution: By formula (1.3), k v k = 82 + 32 = 73.
2
(c) The distance between the points P = (2, −1, 4)
pand Q = (4, 2, −3) in R .
Solution: By formula
p p (1.4), the distance d = (4 − 2) + (2 − (−1)) + (−3 − 4)2 =
2 2

4 + 9 + 49 = 62.
8 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

3
(d) The magnitude of the vector v = (5 p, 8, −2) in R . p p
Solution: By formula (1.5), k v k = 5 + 8 + (−2)2 = 25 + 64 + 4 = 93.
2 2

Exercises
A
1. Calculate the magnitudes of the following vectors:
(a) v = (2, −1); (b) v = (2, −1, 0); (c) v = (3, 2, −2); (d) v = (0, 0, 1); (e) v = (6, 4, −4).
−−→ −−→
2. For the points P = (1, −1, 1), Q = (2, −2, 2), R = (2, 0, 1), S = (3, −1, 2), does PQ = RS ?
−−→ −−→
3. For the points P = (0, 0, 0), Q = (1, 3, 2), R = (1, 0, 1), S = (2, 3, 4), does PQ = RS ?

B
4. Let v = (1, 0, 0) and w = (a, 0, 0) be vectors in R3 . Show that k w k = | a | k v k.

5. Let v = (a, b, c) and w = (3a, 3 b, 3 c) be vectors in R3 . Show that k w k = 3 k v k.

C
z
6. Though we will see a simple proof of Theorem 1.1
in the next section, it is possible to prove it using Q(x2 , y2 , z2 )
methods similar to those in the proof of Theorem P(x1 , y1 , z1 )
1.2. Prove the special case of Theorem 1.1 where the R(x2 , y2 , z1 )
points P = ( x1 , y1 , z1 ) and Q = ( x2 , y2 , z2 ) satisfy the fol-
lowing conditions: y
0
x2 > x1 > 0, y2 > y1 > 0, and z2 > z1 > 0. S(x1 , y1 , 0)
(Hint: Think of Case 4 in the proof of Theorem 1.2,
T(x2 , y2 , 0)
and consider Figure 1.1.9.) x U(x2 , y1 , 0)
Figure 1.1.9
 1.2 Vector Algebra 9

1.2 Vector Algebra

Now that we know what vectors are, we can start to perform some of the usual algebraic
operations on them including addition and subtraction. Before doing that, we will introduce
the notion of a scalar.
Definition 1.3. A scalar is a quantity that can be represented by a single number.
For our purposes, scalars will always be real numbers.3 Examples of scalar quantities are
mass, electric charge, and speed (not velocity).4 We can now define scalar multiplication of
a vector.
Definition 1.4. For a scalar k and a nonzero vector v, the scalar multiple of v by k,
denoted by kv, is the vector whose magnitude is | k | k v k, points in the same direction as v if
k > 0, points in the opposite direction as v if k < 0, and is the zero vector 0 if k = 0. For the
zero vector 0, we define k0 = 0 for any scalar k.

Two vectors v and w are parallel (denoted by v ∥ w) if one is a scalar multiple of the other.
You can think of scalar multiplication of a vector as stretching or shrinking the vector, and
as flipping the vector in the opposite direction if the scalar is a negative number (see Figure
1.2.1).

v 2v 3v 0.5v −v −2v

Figure 1.2.1

Recall that translating a nonzero vector means that the initial point of the vector is
changed but the magnitude and direction are preserved. We are now ready to define the
sum of two vectors.
Definition 1.5. The sum of vectors v and w, denoted by v + w, is obtained by translating
w so that its initial point is at the terminal point of v; the initial point of v + w is the initial
point of v, and its terminal point is the new terminal point of w.
3 The term scalar was invented by 19th century Irish mathematician, physicist and astronomer William Rowan

Hamilton, to convey the sense of something that could be represented by a point on a scale or graduated ruler.
The word vector comes from Latin, where it means “carrier”.
4 An alternate definition of scalars and vectors, used in physics, is that under certain types of coordinate trans-

formations (for example rotations), a quantity that is not affected is a scalar, while a quantity that is affected
(in a certain way) is a vector. See M ARION for details.
10 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

Intuitively, adding w to v means tacking on w to the end of v (see Figure 1.2.2).

v+w
w
w w
v v v

(a) Vectors v and w (b) Translate w to the end of v (c) The sum v + w

Figure 1.2.2 Adding vectors v and w.

Notice that our definition is valid for the zero vector (which is just a point, and hence can
be translated), and so we see that v + 0 = v = 0 + v for any vector v. In particular, 0 + 0 = 0.
Also, it is easy to see that v + (−v) = 0, as we would expect. In general, since the scalar
multiple −v = −1 v is a well-defined vector, we can define vector subtraction as follows:
v − w = v + (−w). See Figure 1.2.3.

v v
w v−w
−w −w
v

(a) Vectors v and w (b) Translate −w to the end of v (c) The difference v − w

Figure 1.2.3 Subtracting vectors v and w.

Figure 1.2.4 shows the use of “geometric proofs” of various laws of vector algebra, that is,
it uses laws from elementary geometry to prove statements about vectors. For example, (a)
shows that v + w = w + v for any vectors v, w. And (c) shows how you can think of v − w as
the vector that is tacked on to the end of w to add up to v.

v w
v+w
w+v v−w
w w v v−w w
v+w v−w
v −w v
(a) Add vectors (b) Subtract vectors (c) Combined add/subtract

Figure 1.2.4 “Geometric” vector algebra.

Notice that we have temporarily abandoned the practice of starting vectors at the origin.
In fact, we have not even mentioned coordinates in this section so far. Since we will deal
mostly with Cartesian coordinates in this book, the following two theorems are useful for
performing vector algebra on vectors in R2 and R3 starting at the origin.
 1.2 Vector Algebra 11

Theorem 1.3. Let v = (v1 , v2 ), w = (w1 , w2 ) be vectors in R2 , and let k be a scalar. Then
(a) kv = ( kv1 , kv2 );
(b) v + w = (v1 + w1 , v2 + w2 ).

Proof: (a) Without loss of generality, we assume that v1 , v2 > 0 (the other possibilities are
handled in a similar manner). If k = 0 then kv = 0v = 0 = (0, 0) = (0v1 , 0v2 ) = ( kv1 , kv2 ), which
is what we needed to show. If k 6= 0, then ( kv1 , kv2 ) lies on a line with slope kv 2 v2
kv1 = v1 , which
is the same as the slope of the line on which v (and hence kv) lies, and ( kv1 , kv2 ) points in
the same directionqon that line asq kv. Also, by formula q (1.3) the magnitude of ( kv1 , kv2 ) is
p
( kv1 )2 + ( kv2 )2 = k2 v12 + k2 v22 = k2 (v12 + v22 ) = | k | v12 + v22 = | k | k v k. So kv and ( kv1 , kv2 )
have the same magnitude and direction. This proves (a).

(b) Without loss of generality, we assume that y

v2 + w 2
v1 , v2 , w1 , w2 > 0 (the other possibilities are han- v
dled in a similar manner). From Figure 1.2.5, we w2 w w2
w v+w
see that when translating w to start at the end of v2
w1
v, the new terminal point of w is (v1 + w1 , v2 + w2 ), v x
so by the definition of v + w this must be the ter- 0 w1 v1 v1 + w 1
minal point of v + w. This proves (b). QED
Figure 1.2.5

Theorem 1.4. Let v = (v1 , v2 , v3 ), w = (w1 , w2 , w3 ) be vectors in R3 , let k be a scalar. Then

(a) kv = ( kv1 , kv2 , kv3 );
(b) v + w = (v1 + w1 , v2 + w2 , v3 + w3 ).

The following theorem summarizes the basic laws of vector algebra.

Theorem 1.5. For any vectors u, v, w, and scalars k, l , we have

(a) v + w = w + v Commutative Law;
(b) u + (v + w) = (u + v) + w Associative Law;
(c) v + 0 = v = 0 + v Additive Identity;
(d) v + (−v) = 0 Additive Inverse;
(e) k( l v) = ( kl )v Associative Law;
(f) k(v + w) = kv + kw Distributive Law;
(g) ( k + l )v = kv + l v Distributive Law.

Proof: (a) We already presented a geometric proof of this in Figure 1.2.4(a).

(b) To illustrate the difference between analytic proofs and geometric proofs in vector alge-
bra, we will present both types here. For the analytic proof, we will use vectors in R3 (the
proof for R2 is similar).
12 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

Let u = ( u 1 , u 2 , u 3 ), v = (v1 , v2 , v3 ), w = (w1 , w2 , w3 ) be vectors in R3 . Then

u + (v + w) = ( u 1 , u 2 , u 3 ) + ((v1 , v2 , v3 ) + (w1 , w2 , w3 ))
= ( u 1 , u 2 , u 3 ) + (v1 + w1 , v2 + w2 , v3 + w3 ) by Theorem 1.4(b)
= ( u 1 + (v1 + w1 ), u 2 + (v2 + w2 ), u 3 + (v3 + w3 )) by Theorem 1.4(b)
= (( u 1 + v1 ) + w1 , ( u 2 + v2 ) + w2 , ( u 3 + v3 ) + w3 ) by properties of real numbers
= ( u 1 + v1 , u 2 + v2 , u 3 + v3 ) + (w1 , w2 , w3 ) by Theorem 1.4(b)
= (u + v) + w

This completes the analytic proof of (b). Figure 1.2.6 provides the geometric proof.
u + (v + w) = (u + v) + w

v+w w
u u+v

v
Figure 1.2.6 Associative Law for vector addition

(c) We already discussed this on p.10.

(d) We already discussed this on p.10.
(e) We will prove this for a vector v = (v1 , v2 , v3 ) in R3 (the proof for R2 is similar):

k( l v) = k( lv1 , lv2 , lv3 ) by Theorem 1.4(a)

= ( klv1 , klv2 , klv3 ) by Theorem 1.4(a)
= ( kl )(v1 , v2 , v3 ) by Theorem 1.4(a)
= ( kl )v

(f) and (g): Left as exercises for the reader. QED

A unit vector is a vector with magnitude 1. Notice that for any nonzero vector° v, the °
vector k vv k is a unit vector which points in the same direction as v, since k v1 k > 0 and ° k vv k ° =
kvk
kvk = 1. Dividing a nonzero vector v by k v k is often called normalizing v.
There are specific unit vectors which we will often use, called the basis vectors:
i = (1, 0, 0), j = (0, 1, 0), and k = (0, 0, 1) in R3 ; i = (1, 0) and j = (0, 1) in R2 .
These are useful for several reasons: they are mutually perpendicular, since they lie on
distinct coordinate axes; they are all unit vectors: k i k = k j k = k k k = 1; every vector can
be written as a unique scalar combination of the basis vectors: v = (a, b) = a i + b j in R2 ,
 1.2 Vector Algebra 13

v = (a, b, c) = a i + b j + c k in R3 . See Figure 1.2.7.

z z
2 v = (a, b, c)

y y 1
2 v = (a, b)
ck
k
y y
1 bj i 0 j 0
1 2 ai
j x 1
x
0 2 bj
i 1 2 0 ai x x
(a) R2 (b) v = a i + b j (c) R3 (d) v = a i + b j + c k

Figure 1.2.7 Basis vectors in different dimensions.

When a vector v = (a, b, c) is written as v = a i + b j + c k, we say that v is in component

form, and that a, b, and c are the i, j, and k components, respectively, of v. We have:

v = v1 i + v2 j + v3 k, k a scalar =⇒ kv = kv1 i + kv2 j + kv3 k;

v = v1 i + v2 j + v3 k, w = w1 i + w2 j + w3 k =⇒ v + w = (v1 + w1 )i + (v2 + w2 )j + (v3 + w3 )k;
q
v = v1 i + v2 j + v3 k =⇒ k v k = v12 + v22 + v32 .

Example 1.4. Let v = (2, 1, −1) and w = (3, −4, 2) in R3 .

(a) Find v − w.
Solution: v − w = (2 − 3, 1 − (−4), −1 − 2) = (−1, 5, −3).

(b) Find 3v + 2w.

Solution: 3v + 2w = (6, 3, −3) + (6, −8, 4) = (12, −5, 1).

(c) Write v and w in component form.

Solution: v = 2 i + j − k, w = 3 i − 4 j + 2 k.

(d) Find the vector u such that u + v = w.

Solution: By Theorem 1.5, u = w − v = −(v − w) = −(−1, 5, −3) = (1, −5, 3), by part(a).

(e) Find the vector u such that u + v + w = 0.

Solution: By Theorem 1.5, u = −w − v = −(3, −4, 2) − (2, 1, −1) = (−5, 3, −1).

(f) Find the vector u such that 2u + i − 2 j = k.

Solution: 2u = −i + 2 j + k =⇒ u = − 12 i + j + 12 k.
v
(g) Find the unit vector kvk . ³ ´
v p 1 p2 , p1 , p
−1
Solution: kvk = (2, 1, −1) = .
22 +12 +(−1)2 6 6 6
14 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

We can now easily prove Theorem 1.1 from the previous section. The distance d between
two points P = ( x1 , y1 , z1 ) and Q = ( x2 , y2 , z2 ) in R3 is the same as the length of the vector w − v,
where the vectors v and w are defined as v = ( x1 , y1 , z1 ) and p w = ( x2 , y2 , z2 ) (see Figure 1.2.8).
So since w − v = ( x2 − x1 , y2 − y1 , z2 − z1 ), then d = k w − v k = ( x2 − x1 )2 + ( y2 − y1 )2 + ( z2 − z1 )2 by
Theorem 1.2.
z P(x1 , y1 , z1 )
w−v
v Q(x2 , y2 , z2 )

w y
0
x

Figure 1.2.8 Proof of Theorem 1.2: d = k w − v k.

Exercises
A
1. Let v = (−1, 5, −2) and w = (3, 1, 1).
° °
(a) Find v − w. (b) Find v + w. (c) Find v
kvk . (d) Find ° 21 (v − w) °.
°1 °
(e) Find ° (v + w) °. (f) Find −2 v + 4 w.
2 (g) Find v − 2 w.
(h) Find the vector u such that u + v + w = i.
(i) Find the vector u such that u + v + w = 2 j + k.
(j) Is there a scalar m such that m(v + 2 w) = k? If so, find it.
2. For the vectors v and w from Exercise 1, is k v − w k = k v k − k w k? If not, which quantity
is larger?

3. For the vectors v and w from Exercise 1, is k v + w k = k v k + k w k? If not, which quantity

is larger?
B
4. Prove Theorem 1.5(f) for R3 . 5. Prove Theorem 1.5(g) for R3 .

C
6. We know that every vector in R3 can be written as a scalar combination of the vectors i,
j, and k. Can every vector in R3 be written as a scalar combination of just i and j; that is
for any vector v in R3 , are there scalars m, n such that v = m i + n j? Justify your answer.
 1.3 Dot Product 15

1.3 Dot Product

You may have noticed that while we did define multiplication of a vector by a scalar in the
previous section on vector algebra, we did not define multiplication of a vector by a vector.
We will now see one type of multiplication of vectors, called the dot product.

Definition 1.6. Let v = (v1 , v2 , v3 ) and w = (w1 , w2 , w3 ) be vectors in R3 .

The dot product of v and w, denoted by v · w, is given by:

v · w = v1 w1 + v2 w2 + v3 w3 . (1.6)

Similarly, for vectors v = (v1 , v2 ) and w = (w1 , w2 ) in R2 , the dot product is:

v · w = v1 w1 + v2 w2 . (1.7)

Notice that the dot product of two vectors is a scalar, not a vector. So the associative law
that holds for multiplication of numbers and for addition of vectors (see Theorem 1.5(b),(e)),
does not hold for the dot product of vectors. Why? Because for vectors u, v, w, the dot
product u · v is a scalar, and so (u · v) · w is not defined since the left side of that dot product
(the part in parentheses) is a scalar and not a vector.
For vectors v = v1 i + v2 j + v3 k and w = w1 i + w2 j + w3 k in component form, the dot product
is still v · w = v1 w1 + v2 w2 + v3 w3 .
Also notice that we defined the dot product in an analytic way, that is, by referencing
vector coordinates. There is a geometric way of defining the dot product, which we will now
develop as a consequence of the analytic definition.

Definition 1.7. The angle between two nonzero vectors with the same initial point is the
smallest angle between them.

We do not define the angle between the zero vector and any other vector. Any two nonzero
vectors with the same initial point have two angles between them: θ and 360◦ − θ . We will
always choose the smallest nonnegative angle θ between them, so that 0◦ ≤ θ ≤ 180◦ . See
Figure 1.3.1.

θ 360◦ − θ
θ
θ
360◦ − θ 360◦ − θ

(a) 0◦ < θ < 180◦ (b) θ = 180◦ (c) θ = 0◦

Figure 1.3.1 Angle between vectors.

We can now take a more geometric view of the dot product by establishing a relationship
between the dot product of two vectors and the angle between them.
16 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

Theorem 1.6. Let v, w be nonzero vectors, and let θ be the angle between them. Then
v· w
cos θ = (1.8)
kvk kwk
We will prove the theorem, assuming that the notion of angle as well as the Law of Cosines
are known. In a more rigorous approach, one could define the angles between the vectors
using the statement of the theorem above.

Proof: We will prove the theorem for vectors in R3 (the proof for R2 is similar). Let v =
(v1 , v2 , v3 ) and w = (w1 , w2 , w3 ). By the Law of Cosines (see Figure 1.3.2), we have

k v − w k2 = k v k2 + k w k2 − 2 k v k k w k cos θ . (1.9)

(note that equation (1.9) holds even for the “degenerate” cases θ = 0◦ and 180◦ ).
z
v−w
v
θ
w y
0

x
Figure 1.3.2

Since v − w = (v1 − w1 , v2 − w2 , v3 − w3 ), expanding k v − w k2 in equation (1.9) gives

k v k2 + k w k2 − 2 k v k k w k cos θ = (v1 − w1 )2 + (v2 − w2 )2 + (v3 − w3 )2

= (v12 − 2v1 w1 + w12 ) + (v22 − 2v2 w2 + w22 ) + (v32 − 2v3 w3 + w32 )
= (v12 + v22 + v32 ) + (w12 + w22 + w32 ) − 2(v1 w1 + v2 w2 + v3 w3 )
= k v k2 + k w k2 − 2(v · w) , so
−2 k v k k w k cos θ = −2(v · w) , so since v 6= 0 and w 6= 0,
v· w
cos θ = . QED
kvk kwk

Example 1.5. Find the angle θ between the vectors v = (2, 1, −1) and w = (3, −4, 1).
p p
Solution: Since v · w = (2)(3) + (1)(−4) + (−1)(1) = 1, k v k = 6, and k w k = 26, then
v· w 1 1
cos θ = = p p = p ≈ 0.08 =⇒ θ ≈ 85.41◦ .
kvk kwk 6 26 2 39
Two nonzero vectors are perpendicular if the angle between them is 90◦ . Since cos 90◦ =
0, we have the following important corollary to Theorem 1.6:
 1.3 Dot Product 17

Corollary 1.7. Two nonzero vectors v and w are perpendicular if and only if v · w = 0.

We will write v ⊥ w to indicate that v and w are perpendicular.

Since 0 · w = 0, it is convenient to assume that zero vector 0 is perpendicular to any other
vector. So we can write 0 ⊥ w despite that the angle between 0 and w is undefined.
Since cos θ > 0 for 0◦ ≤ θ < 90◦ and cos θ < 0 for 90◦ < θ ≤ 180◦ , we also have:

Corollary 1.8. If θ is the angle between nonzero vectors v and w, then


◦ ◦
> 0 for 0 ≤ θ < 90 ,


v · w is 0 for θ = 90◦ ,


< 0 for 90◦ < θ ≤ 180◦ .

By Corollary 1.8, the dot product can be thought of as a way of telling if the angle be-
tween two vectors is acute, obtuse, or a right angle, depending on whether the dot product
is positive, negative, or zero, respectively. See Figure 1.3.3.
w w w
90◦ < θ ≤ 180◦
θ = 90◦
◦ ◦
0 ≤ θ < 90
v v v
(a) v · w > 0 (b) v · w < 0 (c) v · w = 0

Figure 1.3.3 Sign of the dot product & angle between vectors.

Example 1.6. Are the vectors v = (−1, 5, −2) and w = (3, 1, 1) perpendicular?
Solution: Yes, v ⊥ w since v · w = (−1)(3) + (5)(1) + (−2)(1) = 0.

The following theorem summarizes the basic properties of the dot product.

Theorem 1.9. For any vectors u, v, w, and scalar k, we have

(a) v · w = w · v Commutative Law;
(b) ( kv) · w = v · ( kw) = k(v · w) Associative Law;
(c) v · 0 = 0 = 0 · v;
(d) u · (v + w) = u · v + u · w Distributive Law;
(e) (u + v) · w = u · w + v · w Distributive Law;
(f) | v · w | ≤ k v k k w k Cauchy–Schwarz Inequality.5

5 Also known as the Cauchy–Schwarz–Buniakovski Inequality.

18 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

Proof: The proofs of parts (a)–(e) are straightforward applications of the definition of the
dot product, and are left to the reader as exercises. We will prove part (f).
(f) If either v = 0 or w = 0, then v · w = 0 by part (c), and so the inequality holds trivially. So
assume that v and w are nonzero vectors. Then by Theorem 1.6,

v · w = cos θ k v k k w k , so
| v · w | = | cos θ | k v k k w k , so
| v · w | ≤ k v k k w k since | cos θ | ≤ 1. QED

Using Theorem 1.9, we see that if u · v = 0 and u · w = 0, then

u · ( kv + l w) = k(u · v) + l (u · w) = k(0) + l (0) = 0

for all scalars k, l . Thus, we have the following fact:

If u ⊥ v and u ⊥ w, then u ⊥ ( kv + l w) for all scalars k, l .

For vectors v and w, the collection of all scalar combinations kv + l w is called the span
of v and w. If nonzero vectors v and w are parallel, then their span is a line; if they are
not parallel, then their span is a plane. So what we showed above is that a vector which is
perpendicular to two other vectors is also perpendicular to their span.
The dot product can be used to derive properties of the magnitudes of vectors, the most
important of which is the Triangle Inequality, as given in the following theorem:

Theorem 1.10. For any vectors v, w, we have

(a) k v k2 = v · v;
(b) k v + w k ≤ k v k + k w k Triangle Inequality;
(c) k v − w k ≥ k v k − k w k.

Proof: (a) Left as an exercise for the reader.

(b) By part (a) and Theorem 1.9, we have

k v + w k2 = (v + w) · (v + w) = v · v + v · w + w · v + w · w
= k v k2 + 2(v · w) + k w k2 , so since a ≤ | a | for any real number a, we have
≤ k v k2 + 2 | v · w | + k w k2 , so by Theorem 1.9(f) we have
≤ k v k2 + 2 k v k k w k + k w k2 = (k v k + k w k)2 and so
k v + w k ≤ k v k + k w k after taking square roots of both sides, which proves (b).

(c) Since v = w + (v − w), then k v k = k w + (v − w) k ≤ k w k+k v − w k by the Triangle Inequality,

so subtracting k w k from both sides gives k v k − k w k ≤ k v − w k. QED
 1.3 Dot Product 19

The Triangle Inequality gets its name from the fact that in any triangle, v+w
no one side is longer than the sum of the lengths of the other two sides (see
w
Figure 1.3.4). Another way of saying this is with the familiar statement “the v
shortest distance between two points is a straight line.”
Figure 1.3.4

Exercises
A
1. Let v = (5, 1, −2) and w = (4, −4, 3). Calculate v · w.

2. Let v = −3 i − 2 j − k and w = 6 i + 4 j + 2 k. Calculate v · w.

For Exercises 3–8, find the angle θ between the vectors v and w.

3. v = (5, 1, −2), w = (4, −4, 3); 4. v = (7, 2, −10), w = (2, 6, 4);

5. v = (2, 1, 4), w = (1, −2, 0); 6. v = (4, 2, −1), w = (8, 4, −2);

7. v = − i + 2 j + k, w = −3 i + 6 j + 3 k; 8. v = i, w = 3 i + 2 j + 4k.

9. Let v = (8, 4, 3) and w = (−2, 1, 4). Is v ⊥ w? Justify your answer.

10. Let v = (6, 0, 4) and w = (0, 2, −1). Is v ⊥ w? Justify your answer.

11. For v, w from Exercise 5, verify the Cauchy–Schwarz Inequality | v · w | ≤ k v k k w k.

12. For v, w from Exercise 6, verify the Cauchy–Schwarz Inequality | v · w | ≤ k v k k w k.

13. For v, w from Exercise 5, verify the Triangle Inequality k v + w k ≤ k v k + k w k.

14. For v, w from Exercise 6, verify the Triangle Inequality k v + w k ≤ k v k + k w k.

B
15. Prove Theorem 1.9(a). 16. Prove Theorem 1.9(b).

17. Prove Theorem 1.9(c). 18. Prove Theorem 1.9(d).

19. Prove Theorem 1.9(e). 20. Prove Theorem 1.10(a).

21. Prove or give a counterexample: If u · v = u · w, then v = w.

C
22. Prove or give a counterexample: If v · w = 0 for all v, then w = 0.

23. Prove or give a counterexample: If u · v = u · w for all u, then v = w.

24. Prove that |k v k − k w k| ≤ k v − w k for all v, w.

 20 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

v
25. For nonzero vectors v and w, the projection of v onto w (some-
times written as pro j w v) is the vector u along the same line L
as w whose terminal point is obtained by dropping a perpendic- L
ular line from the terminal point of v to L (see Figure 1.3.5). w
u
Show that
|v · w| Figure 1.3.5
kuk = .
kwk
(Hint: Consider the angle between v and w.)

26. Assume k v k = k w k. Show that (v + w) ⊥ (v − w).

27. Let α, β, and γ be the angles between a nonzero vector v in R3 and the vectors i, j, and k,
respectively. Show that cos2 α + cos2 β + cos2 γ = 1. (The angles α, β, γ are often called the
direction angles of v, and cos α, cos β, cos γ are called the direction cosines.)
 1.4 Cross Product 21

1.4 Cross Product

In Section 1.3 we defined the dot product, which gave a way of multiplying two vectors. The
resulting product, however, was a scalar, not a vector. In this section we will define a product
of two vectors that does result in another vector. This product, called the cross product, is
only defined for vectors in R3 . The definition may appear strange and lacking motivation,
but we will see the geometric basis for it shortly.

Definition 1.8. Let v = (v1 , v2 , v3 ) and w = (w1 , w2 , w3 ) be vectors in R3 . The cross product
of v and w, denoted by v × w, is the vector in R3 given by:

v × w = (v2 w3 − v3 w2 , v3 w1 − v1 w3 , v1 w2 − v2 w1 ). (1.10)

z
Example 1.7. Find i × j.
Solution: Since i = (1, 0, 0) and j = (0, 1, 0), then 1 k = i× j
y
i × j = ((0)(0) − (0)(1), (0)(0) − (1)(0), (1)(1) − (0)(0))
i 0 j 1
= (0, 0, 1) 1
x
= k.
Figure 1.4.1
Similarly it can be shown that j × k = i and k × i = j.

In the above example, the cross product of the given vectors was perpendicular to both
those vectors. It turns out that this will always be the case.

Theorem 1.11. If the cross product v × w of two nonzero vectors v and w is also a nonzero
vector, then it is perpendicular to both v and w.

Proof: We will show that (v × w) · v = 0:

(v × w) · v = (v2 w3 − v3 w2 , v3 w1 − v1 w3 , v1 w2 − v2 w1 ) · (v1 , v2 , v3 )
= v2 w3 v1 − v3 w2 v1 + v3 w1 v2 − v1 w3 v2 + v1 w2 v3 − v2 w1 v3
= v1 v2 w3 − v1 v2 w3 + w1 v2 v3 − w1 v2 v3 + v1 w2 v3 − v1 w2 v3
= 0 , after rearranging the terms.

∴ v × w ⊥ v by Corollary 1.7.
The proof that v × w ⊥ w is similar. QED

As a consequence of the above theorem and Theorem 1.9, we have the following:

Corollary 1.12. If the cross product v × w of two nonzero vectors v and w is also a nonzero
vector, then it is perpendicular to the span of v and w.
22 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

The span of any two nonzero, nonparallel vectors v, w in R3 is a plane P , so the above
corollary shows that v × w is perpendicular to that plane. As shown in Figure 1.4.2, there
are two possible directions for v × w, one the opposite of the other. The choice of direction
of v × w can be visualized using the right-hand rule, that is, the vectors v, w, v × w form
a right-handed system. Recall from Section 1.1 that this means that you can point your
thumb upwards in the direction of v × w while rotating v towards w with the remaining four
fingers.

z
v× w
v

θ y
0
w
P

x −v × w

Figure 1.4.2 Direction of v × w.

We will now derive a formula for the magnitude of v × w, for nonzero vectors v, w:

k v × w k2 = (v2 w3 − v3 w2 )2 + (v3 w1 − v1 w3 )2 + (v1 w2 − v2 w1 )2

= v22 w32 − 2v2 w2 v3 w3 + v32 w22 + v32 w12 − 2v1 w1 v3 w3 + v12 w32 + v12 w22 − 2v1 w1 v2 w2 + v22 w12
= v12 (w22 + w32 ) + v22 (w12 + w32 ) + v32 (w12 + w22 ) − 2(v1 w1 v2 w2 + v1 w1 v3 w3 + v2 w2 v3 w3 )

and now adding and subtracting v12 w12 , v22 w22 , and v32 w32 on the right side gives

= v12 (w12 + w22 + w32 ) + v22 (w12 + w22 + w32 ) + v32 (w12 + w22 + w32 )
− (v12 w12 + v22 w22 + v32 w32 + 2(v1 w1 v2 w2 + v1 w1 v3 w3 + v2 w2 v3 w3 ))
= (v12 + v22 + v32 )(w12 + w22 + w32 )
− ((v1 w1 )2 + (v2 w2 )2 + (v3 w3 )2 + 2(v1 w1 )(v2 w2 ) + 2(v1 w1 )(v3 w3 ) + 2(v2 w2 )(v3 w3 ))

so using (a + b + c)2 = a2 + b2 + c2 + 2ab + 2ac + 2 bc for the subtracted term gives

= (v12 + v22 + v32 )(w12 + w22 + w32 ) − (v1 w1 + v2 w2 + v3 w3 )2

= k v k2 k w k2 − (v · w)2
³ (v · w)2 ´
= k v k2 k w k2 1 − , since k v k > 0 and k w k > 0, so by Theorem 1.6
k v k2 k w k2
= k v k2 k w k2 (1 − cos2 θ ) , where θ is the angle between v and w, so
k v × w k2 = k v k2 k w k2 sin2 θ , and since 0◦ ≤ θ ≤ 180◦ , then sin θ ≥ 0, so we have:
 1.4 Cross Product 23

If θ is the angle between nonzero vectors v and w in R3 , then

k v × w k = k v k k w k sin θ . (1.11)

It may seem strange to bother with the above formula, when the magnitude of the cross
product can be calculated directly, like for any other vector. The formula is more useful for
its applications in geometry, as in the following example.

Example 1.8. Let △PQR and PQRS be a triangle and parallelogram, respectively, as shown
in Figure 1.4.3.
P S P S

h w h
θ θ
Q b R Q v R

Figure 1.4.3

Think of the triangle as existing in R3 , and identify the sides QR and QP with vectors v
and w, respectively, in R3 . Let θ be the angle between v and w. The area A PQR of △PQR is
1
2 bh, where b is the base of the triangle and h is the height. So we see that

b = k v k and h = k w k sin θ ,

1
A PQR = k v k k w k sin θ
2
1
= k v × w k.
2
So since the area A PQRS of the parallelogram PQRS is twice the area of the triangle △PQR ,
then
A PQRS = k v k k w k sin θ .

By the discussion in Example 1.8, we have proved the following theorem:

Theorem 1.13. Area of triangles and parallelograms

(a) The area A of a triangle with adjacent sides v, w (as vectors in R3 ) is:

1
A= k v × w k;
2

(b) The area A of a parallelogram with adjacent sides v, w (as vectors in R3 ) is:

A = k v × w k.
24 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

It may seem at first glance that since the formulas derived in Example 1.8 were for the
adjacent sides QP and QR only, then the more general statements in Theorem 1.13 that the
formulas hold for any adjacent sides are not justified. We would get a different formula for
the area if we had picked PQ and PR as the adjacent sides, but it can be shown (see Exercise
26) that the different formulas would yield the same value, so the choice of adjacent sides
indeed does not matter, and Theorem 1.13 is valid.
Theorem 1.13 makes it simpler to calculate the area of a triangle in 3-dimensional space
than by using traditional geometric methods.

Example 1.9. Calculate the area of the triangle △PQR , where P = (2, 4, −7), Q = (3, 7, 18),
and R = (−5, 12, 8).
−−→ −−→ z
Solution: Let v = PQ and w = PR , as in Figure 1.4.4. Then Q(3, 7, 18)

v = (3, 7, 18) − (2, 4, −7) = (1, 3, 25) R(−5, 12, 8)

v
and w y
0
w = (−5, 12, 8) − (2, 4, −7) = (−7, 8, 15),
x
P(2, 4, −7)
so the area A of the triangle △PQR is
Figure 1.4.4
1 1
A = k v × w k = k (1, 3, 25) × (−7, 8, 15) k
2 2
1° °
= ° ((3)(15) − (25)(8), (25)(−7) − (1)(15), (1)(8) − (3)(−7)) °
2
1° °
= ° (−155, −190, 29) °
2
1p 1p
= (−155)2 + (−190)2 + 292 = 60966.
2 2
A ≈ 123.46.

Example 1.10. Calculate the area of the parallelogram PQRS , where P = (1, 1), Q = (2, 3),
R = (5, 4), and S = (4, 2).
 1.4 Cross Product 25

−−→ −−→ y
Solution: Let v = SP and w = SR , as in Figure 1.4.5. Then
R
4
v = (1, 1) − (4, 2) = (−3, −1) and w = (5, 4) − (4, 2) = (1, 2). Q
3 w

2
But these are vectors in R2 , and the cross product is only de- S
fined for vectors in R3 . However, R2 can be thought of as the 1 v
P
subset of R3 such that the z-coordinate is always 0. So we can x
write v = (−3, −1, 0) and w = (1, 2, 0). Then the area A of the 0 1 2 3 4 5
parallelogram PQRS is Figure 1.4.5

° °
A = k v × w k = ° (−3, −1, 0) × (1, 2, 0) °
° °
= ° ((−1)(0) − (0)(2), (0)(1) − (−3)(0), (−3)(2) − (−1)(1)) °
° °
= ° (0, 0, −5) °.
A = 5.

The following theorem summarizes the basic properties of the cross product.

Theorem 1.14. For any vectors u, v, w in R3 , and scalar k, we have

(a) v × w = −w × v Anticommutative Law
(b) u × (v + w) = u × v + u × w Distributive Law
(c) (u + v) × w = u × w + v × w Distributive Law
(d) ( kv) × w = v × ( kw) = k(v × w) Associative Law
(e) v × 0 = 0 = 0 × v
(f) v × v = 0
(g) v × w = 0 if and only if v ∥ w

Proof: The proofs of properties (b)–(f) are straightforward. We will prove parts (a) and (g)
and leave the rest to the reader as exercises.
(a) By the definition of the cross product and scalar multipli- z
v× w
cation, we have: v

v × w = (v2 w3 − v3 w2 , v3 w1 − v1 w3 , v1 w2 − v2 w1 ) y

= −(v3 w2 − v2 w3 , v1 w3 − v3 w1 , v2 w1 − v1 w2 ) 0
w
= −(w2 v3 − w3 v2 , w3 v1 − w1 v3 , w1 v2 − w2 v1 )
= −w × v x w× v
Figure 1.4.6
26 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

Note that this says that v × w and w × v have the same mag-
nitude but opposite direction (see Figure 1.4.6).

(g) If either v or w is 0 then v × w = 0 by part (e), and either v = 0 = 0w or w = 0 = 0v, so v

and w are scalar multiples, in which case they are parallel.
If both v and w are nonzero, and θ is the angle between them, then by formula (1.11),
v × w = 0 if and only if k v k k w k sin θ = 0, which is true if and only if sin θ = 0 (since k v k > 0
and k w k > 0). So since 0◦ ≤ θ ≤ 180◦ , then sin θ = 0 if and only if θ = 0◦ or 180◦ . But the
angle between v and w is 0◦ or 180◦ if and only if v ∥ w. QED

Example 1.11. Adding to Example 1.7, we have

i× j = k j × k = i, k× i = j
j × i = −k, k × j = −i, i × k = −j ,
i × i = j × j = k × k = 0.

Recall that a parallelepiped is a 3-dimensional solid with 6 faces, all of which are parallel-
ograms.6

Example 1.12. Volume of a parallelepiped: Let the vectors u, v, w in R3 represent adjacent

sides of a parallelepiped P , with u, v, w forming a right-handed system, as in Figure 1.4.7.
Show that the volume of P is the scalar triple product u · (v × w).
v× w
Solution: Recall that the volume vol(P ) of a par-
allelepiped P is the area A of the base parallel-
ogram times the height h. By Theorem 1.13(b),
the area A of the base parallelogram is k v × w k. u
And we can see that since v × w is perpendicular h
θ
to the base parallelogram determined by v and w
w, then the height h is k u k cos θ , where θ is the
angle between u and v × w. By Theorem 1.6 we
v
know that

u · (v × w) Figure 1.4.7 Parallelepiped P

cos θ = .
kuk kv × wk

6 An equivalent definition of a parallelepiped is: the collection of all scalar combinations k v + k v + k v of

1 1 2 2 3 3
some vectors v1 , v2 , v3 in R3 , where 0 ≤ k 1 , k 2 , k 3 ≤ 1.
 1.4 Cross Product 27

Hence,

vol(P ) = A h
k u k u · (v × w)
= kv × wk
kuk kv × wk
= u · (v × w).

In Example 1.12 the height h of the parallelepiped is k u k cos θ , and not −k u k cos θ , because
the vector u is on the same side of the base parallelogram’s plane as the vector v × w (so that
cos θ > 0). Since the volume is the same no matter which base and height we use, then
repeating the same steps using the base determined by u and v (since w is on the same
side of that base’s plane as u × v), the volume is w · (u × v). Repeating this with the base
determined by w and u, we have the following result:

For any vectors u, v, w in R3 ,

u · (v × w) = w · (u × v) = v · (w × u). (1.12)
(Note that the equalities hold trivially if any of the vectors are 0.)
Since v × w = −w × v for any vectors v, w in R3 , then picking the wrong order for the three
adjacent sides in the scalar triple product in formula (1.12) will give you the negative of the
volume of the parallelepiped. So taking the absolute value of the scalar triple product for
any order of the three adjacent sides will always give the volume:

Theorem 1.15. If vectors u, v, w in R3 represent any three adjacent sides of a paral-

lelepiped, then the volume of the parallelepiped is | u · (v × w) |.

Another type of triple product is the vector triple product u × (v × w). The proof of the
following theorem is left as an exercise for the reader:

Theorem 1.16. For any vectors u, v, w in R3 ,

u × (v × w) = (u · w)v − (u · v)w. (1.13)

An examination of the formula in Theorem 1.16 gives some idea of the geometry of the
vector triple product. By the right side of formula (1.13), we see that u × (v × w) is a scalar
combination of v and w, and hence lies in the plane containing v and w (that is, the vectors
u × (v × w), v and w are coplanar). This makes sense since, by Theorem 1.11, u × (v × w) is
perpendicular to both u and v × w. In particular, being perpendicular to v × w means that
u × (v × w) lies in the plane containing v and w, since that plane is itself perpendicular to
v × w. But then how is u × (v × w) also perpendicular to u, which could be any vector? The
following example may help to see how this works.
28 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

Example 1.13. Find u × (v × w) for u = (1, 2, 4), v = (2, 2, 0), w = (1, 3, 0).
Solution: Since u · v = 6 and u · w = 7, then
u × (v × w) = (u · w)v − (u · v)w
= 7 (2, 2, 0) − 6 (1, 3, 0) = (14, 14, 0) − (6, 18, 0)
= (8, −4, 0).

Note that v and w lie in the x y-plane, and that u × (v × w) also lies in that plane. Also,
u × (v × w) is perpendicular to both u and v × w = (0, 0, 4) (see Figure 1.4.8).
z
v×w u

y
0 w
v

u × (v × w)
x

Figure 1.4.8

For vectors v = v1 i + v2 j + v3 k and w = w1 i + w2 j + w3 k in component form, the cross product

is written as: v × w = (v2 w3 − v3 w2 )i + (v3 w1 − v1 w3 )j + (v1 w2 − v2 w1 )k. It is often easier to use the
component form for the cross product, because it can be represented as a determinant. We
will not go too deeply into the theory of determinants7 ; we will just cover what is essential
for our purposes.
A 2 × 2 matrix is an array of two rows and two columns of scalars, written as
· ¸ µ ¶
a b a b
or
c d c d
where a, b, c, d are scalars. The determinant of such a matrix, written as
¯ ¯ · ¸
¯a b ¯
¯ ¯ or det a b ,
¯ c d¯ c d
is the scalar defined by the following formula:
¯ ¯
¯a b ¯
¯ c d ¯ = ad − bc.
¯ ¯

It may help to remember this formula as being the product of the scalars on the downward
diagonal minus the product of the scalars on the upward diagonal.

7 See A NTON and R ORRES for a fuller development.

 1.4 Cross Product 29

Example 1.14. ¯ ¯
¯1 2¯
¯3 4¯ = (1)(4) − (2)(3) = 4 − 6 = −2.
¯ ¯

A 3 × 3 matrix is an array of three rows and three columns of scalars, written as

   
a1 a2 a3 a1 a2 a3
b1 b2 b3  or b1 b2 b3  ,
c1 c2 c3 c1 c2 c3

and its determinant is given by the formula:

¯ ¯
¯a 1 a2 a 3 ¯¯ ¯ ¯ ¯ ¯ ¯ ¯
¯ ¯b2 b 3 ¯¯ ¯b1 b 3 ¯¯ ¯b1 b 2 ¯¯
¯b1 b2 b 3 ¯ = a 1 ¯¯
¯ − a 2 ¯¯ + a 3 ¯¯ . (1.14)
¯
¯c c2 c3 ¯ c1 c3 ¯ c1 c2 ¯
1 c2 c3 ¯

One way to remember the above formula is the following: multiply each scalar in the first
row by the determinant of the 2 × 2 matrix that remains after removing the row and column
that contain that scalar, then sum those products up, putting alternating plus and minus
signs in front of each (starting with a plus).

Example 1.15.
¯ ¯
¯ 1 0 2 ¯¯ ¯ ¯ ¯ ¯ ¯ ¯
¯ 4 −1 3 ¯ = 1 ¯ −1 3 ¯ − 0 ¯ 4 3 ¯ + 2 ¯ 4 −1
¯ ¯ ¯ ¯ ¯ ¯ ¯
¯ = 1(−2 − 0) − 0(8 − 3) + 2(0 + 1) = 0.
¯ ¯ ¯ 0 2 ¯ ¯ 1 2 ¯ ¯ 1 0 ¯
¯ 1 0 2 ¯

We defined the determinant as a scalar, derived from algebraic operations on scalar entries
in a matrix. However, if we put three vectors in the first row of a 3 × 3 matrix, then the
definition still makes sense, since we would be performing scalar multiplication on those
three vectors (they would be multiplied by the 2×2 scalar determinants as before). This gives
us a determinant that is now a vector, and lets us write the cross product of v = v1 i + v2 j + v3 k
and w = w1 i + w2 j + w3 k as a determinant:
¯ ¯
¯i j k ¯¯ ¯¯ ¯ ¯ ¯ ¯ ¯
¯ v2 v3 ¯¯ ¯ v1 v3 ¯¯ ¯ v1 v2 ¯¯
v × w = ¯¯ v1 v2 v3 ¯¯ = ¯¯ i − ¯¯ j + ¯¯ k
¯w w2 w3 ¯ w1 w3 ¯ w1 w2 ¯
1 w2 w3 ¯
= (v2 w3 − v3 w2 )i + (v3 w1 − v1 w3 )j + (v1 w2 − v2 w1 )k .

Example 1.16. Let v = 4 i − j + 3 k and w = i + 2 k. Then

¯ ¯
¯ i j k ¯¯ ¯¯ ¯ ¯ ¯ ¯ ¯
¯ −1 3 ¯¯ ¯ j + ¯ 4 −1
¯ 4 3 ¯ ¯ ¯
v × w = ¯¯ 4 −1 3 ¯¯ = ¯¯ i − ¯ ¯ k = −2 i − 5 j + k .
¯ 1 0 2 ¯ ¯ 1 2 ¯ ¯ 1 0 ¯
0 2 ¯
30 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

The scalar triple product can also be written as a determinant. In fact, by Example 1.12,
the following theorem provides an alternate definition of the determinant of a 3 × 3 matrix
as the volume of a parallelepiped whose adjacent sides are the rows of the matrix and form
a right-handed system (a left-handed system would give the negative volume).

Theorem 1.17. For any vectors u = ( u 1 , u 2 , u 3 ), v = (v1 , v2 , v3 ), w = (w1 , w2 , w3 ) in R3 :

¯ ¯
¯ u1 u2 u 3 ¯¯
¯
u · (v × w) = ¯¯ v1 v2 v3 ¯¯ . (1.15)
¯w w2 w3 ¯
1

Example 1.17. Find the volume of the parallelepiped with adjacent sides u = (2, 1, 3), v =
(−1, 3, 2), w = (1, 1, −2) (see Figure 1.4.9).
Solution: By Theorem 1.15, the volume vol(P ) of the parallelepiped z
P is the absolute value of the scalar triple product of the three
adjacent sides (in any order). By Theorem 1.17, v
¯ ¯ u
¯ 2 1 3 ¯¯ y
¯
u · (v × w) = ¯¯ −1 3 2 ¯¯ 0
¯ 1 1 −2 ¯ x w
¯ ¯ ¯ ¯ ¯ ¯
¯ 3 2 ¯ − 1 ¯ −1 2 ¯ + 3 ¯ −1 3 ¯
¯ ¯ ¯ ¯ ¯
= 2 ¯¯ Figure 1.4.9 P
1 −2 ¯ ¯ 1 −2 ¯ ¯ 1 1 ¯

= 2(−8) − 1(0) + 3(−4) = −28, so

vol(P ) = |−28 | = 28.

Interchanging the dot and cross products can be useful in proving vector identities:
¯ ¯
¯u · w u · z¯
Example 1.18. Prove: (u × v) · (w × z) = ¯
¯ ¯ for all vectors u, v, w, z in R3 .
v · w v · z¯
Solution: Let x = u × v. Then

(u × v) · (w × z) = x · (w × z)
= w · (z × x) (by formula (1.12))
= w · (z × (u × v))
= w · ((z · v)u − (z · u)v) (by Theorem 1.16)
= (z · v)(w · u) − (z · u)(w · v)
= (u · w)(v · z) − (u · z)(v · w) (by commutativity of the dot product).
¯ ¯
¯u · w u · z¯
=¯¯ ¯.
v · w v · z¯
 1.4 Cross Product 31

Exercises
A
For Exercises 1–6, calculate v × w.

1. v = (5, 1, −2), w = (4, −4, 3); 2. v = (7, 2, −10), w = (2, 6, 4);

3. v = (2, 1, 4), w = (1, −2, 0); 4. v = (1, 3, 2), w = (7, 2, −10);

5. v = − i + 2 j + k, w = −3 i + 6 j + 3 k; 6. v = i, w = 3 i + 2 j + 4k.

For Exercises 7–8, calculate the area of the triangle △PQR .

7. P = (5, 1, −2), Q = (4, −4, 3), R = (2, 4, 0); 8. P = (4, 0, 2), Q = (2, 1, 5), R = (−1, 0, −1).

For Exercises 9–10, calculate the area of the parallelogram PQRS .

9. P = (2, 1, 3), Q = (1, 4, 5), R = (2, 5, 3), S = (3, 2, 1);

10. P = (−2, −2), Q = (1, 4), R = (6, 6), S = (3, 0).

For Exercises 11–12, find the volume of the parallelepiped with adjacent sides u, v, w.

11. u = (1, 1, 3), v = (2, 1, 4), w = (5, 1, −2); 12. u = (1, 3, 2), v = (7, 2, −10), w = (1, 0, 1).

For Exercises 13–14, calculate u · (v × w) and u × (v × w).

13. u = (1, 1, 1), v = (3, 0, 2), w = (2, 2, 2); 14. u = (1, 0, 2), v = (−1, 0, 3), w = (2, 0, −2).

15. Calculate (u × v) · (w × z) for u = (1, 1, 1), v = (3, 0, 2), w = (2, 2, 2), z = (2, 1, 4).

B
16. If v and w are unit vectors in R3 , under what condition(s) would v × w also be a unit
vector in R3 ? Justify your answer.

17. Show that if v × w = 0 for all w in R3 , then v = 0.

18. Prove Theorem 1.14(b). 19. Prove Theorem 1.14(c).

20. Prove Theorem 1.14(d). 21. Prove Theorem 1.14(e).

22. Prove Theorem 1.14(f). 23. Prove Theorem 1.16.

24. Prove Theorem 1.17. (Hint: Expand both sides of the equation.)

25. Prove the following for all vectors v, w in R3 :

(a) k v × w k2 + | v · w |2 = k v k2 k w k2
32 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

(b) If v · w = 0 and v × w = 0, then v = 0 or w = 0.

C
26. Prove that in Example 1.8 the formula for the area of the triangle △PQR yields the
same value no matter which two adjacent sides are chosen. To do this, show that

1 1
k u × (−w) k = k v × w k,
2 2
−−→ −−→ −−→ −−→
where u = PR , −w = PQ , and v = QR , w = QP as before. Similarly, show that

1 1
k (−u) × (−v) k = k v × w k,
2 2
−−→ −−→
where −u = RP and −v = RQ .

27. Assume that the vector equation a × x = b in R3 , with unknown x and a 6= 0 has a
solution. Show that:
(a) a · b = 0.
b× a
(b) x = + ka is a solution to the equation, for any scalar k.
k a k2
28. Prove the Jacobi identity:

u × (v × w) + v × (w × u) + w × (u × v) = 0.

29. Show that u, v, w lie in the same plane in R3 if and only if u · (v × w) = 0.

30. For all vectors u, v, w, z in R3 , show that

(u × v) × (w × z) = (z · (u × v))w − (w · (u × v))z

and that

(u × v) × (w × z) = (u · (w × z))v − (v · (w × z))u

Why do both equations make sense geometrically?

31. Describe geometrically the set of points with position vector x satisfying the equation

(v × x) × x = v

for given vector v 6= 0

 1.5 Lines and Planes 33

1.5 Lines and Planes

Now that we know how to perform some operations on vectors, we can start to deal with
some familiar geometric objects, like lines and planes, in the language of vectors. As you
will see, using vectors makes it easier to study objects in 3-dimensional Euclidean space.
We will first consider lines.

Line through a point, parallel to a vector

Let P = ( x0 , y0 , z0 ) be a point in R3 , let v = (a, b, c) be a nonzero vector, and let L be the line
through P which is parallel to v (see Figure 1.5.1).
z
tv L
P(x0 , y0 , z0 )
r + tv
r t>0
r + tv y
t<0 v
0

Figure 1.5.1

Let r = ( x0 , y0 , z0 ) be the vector pointing from the origin to P . Since multiplying the vector
v by a scalar t lengthens or shrinks v while preserving its direction if t > 0, and reversing
its direction if t < 0, then we see from Figure 1.5.1 that every point on the line L can be
obtained by adding the vector tv to the vector r for some scalar t. That is, as t varies over all
real numbers, the vector r + tv will point to every point on L. We can summarize the vector
representation of L as follows:

For a point P = ( x0 , y0 , z0 ) and nonzero vector v in R3 , the line L through P parallel to v

is given by
r + tv, for − ∞ < t < ∞, (1.16)
where r = ( x0 , y0 , z0 ) is the vector pointing to P .

Note that we used the correspondence between a vector and its terminal point. Since
v = (a, b, c), then the terminal point of the vector r + tv is ( x0 + at, y0 + bt, z0 + ct). We then get
the parametric representation of L with the parameter t:

For a point P = ( x0 , y0 , z0 ) and nonzero vector v = (a, b, c) in R3 , the line L through P

parallel to v consists of all points ( x, y, z) given by

x = x0 + at, y = y0 + bt, z = z0 + ct, for − ∞ < t < ∞. (1.17)

Note that in both representations we get the point P on L by letting t = 0.

34 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

In formula (1.17), if a 6= 0, then we can solve for the parameter t: t = ( x − x0 )/a. We can also
solve for t in terms of y and in terms of z if neither b nor c, respectively, is zero: t = ( y − y0 )/ b
and t = ( z − z0 )/ c. These three values all equal the same value t, so we can write the following
system of equalities, called the symmetric representation of L:
For a point P = ( x0 , y0 , z0 ) and vector v = (a, b, c) in R3 with a, b and c all nonzero, the line
L through P parallel to v consists of all points ( x, y, z) given by the equations
x − x0 y − y0 z − z0
= = . (1.18)
a b c

What if, say, a = 0 in the above scenario? We can not divide by z

zero, but we do know that x = x0 + at, and so x = x0 + 0 t = x0 . Then the L
symmetric representation of L would be:
y
y − y0 z − z0
x = x0 , = . (1.19) 0
b c x = x0
x0
Note that this says that the line L lies in the plane x = x0 , which is
parallel to the yz-plane (see Figure 1.5.2). Similar equations can be x
derived for the cases when b = 0 or c = 0. Figure 1.5.2
You may have noticed that the vector representation of L in formula (1.16) is more compact
than the parametric and symmetric formulas. That is an advantage of using vector notation.
Technically, though, the vector representation gives us the vectors whose terminal points
make up the line L, not just L itself. So you have to remember to identify the vectors r + tv
with their terminal points. On the other hand, the parametric representation always gives
just the points on L and nothing else.

Example 1.19. Write the line L through the point P = (2, 3, 5) and parallel to the vector
v = (4, −1, 6), in the following forms: (a) vector, (b) parametric, (c) symmetric. Lastly: (d) find
two points on L distinct from P .
Solution: (a) Let r = (2, 3, 5). Then by formula (1.16), L is given by:

r + tv = (2, 3, 5) + t(4, −1, 6), for − ∞ < t < ∞.

(b) L consists of the points ( x, y, z) such that

x = 2 + 4 t, y = 3 − t, z = 5 + 6 t, for − ∞ < t < ∞.

(c) L consists of the points ( x, y, z) such that

x−2 y−3 z−5

= = .
4 −1 6
(d) Letting t = 1 and t = 2 in part(b) yields the points (6, 2, 11) and (10, 1, 17) on L.
 1.5 Lines and Planes 35

Line through two points

Let P1 = ( x1 , y1 , z1 ) and P2 = ( x2 , y2 , z2 ) be distinct points z
P2 (x2 , y2 , z2 )
in R3 , and let L be the line through P1 and P2 . Let r1 = P1 (x1 , y1 , z1 ) r2 − r1
( x1 , y1 , z1 ) and r2 = ( x2 , y2 , z2 ) be the vectors pointing to P1
r2
and P2 , respectively. Then as we can see from Figure L r1 r1 + t(r2 − r1 )
1.5.3, r2 − r1 is the vector from P1 to P2 . So if we multiply y
the vector r2 − r1 by a scalar t and add it to the vector 0
r1 , we will get the entire line L as t varies over all real
x
numbers. The following is a summary of the vector, para-
metric, and symmetric forms for the line L: Figure 1.5.3

Let P1 = ( x1 , y1 , z1 ), P2 = ( x2 , y2 , z2 ) be distinct points in R3 , and let r1 = ( x1 , y1 , z1 ), r2 =

( x2 , y2 , z2 ). Then the line L through P1 and P2 has the following representations:
Vector:
r1 + t(r2 − r1 ) , for − ∞ < t < ∞. (1.20)
Parametric:

x = x1 + ( x2 − x1 ) t, y = y1 + ( y2 − y1 ) t, z = z1 + ( z2 − z1 ) t, for − ∞ < t < ∞. (1.21)

Symmetric:
x − x1 y − y1 z − z1
= = (if x1 6= x2 , y1 6= y2 , and z1 6= z2 ). (1.22)
x2 − x1 y2 − y1 z2 − z1

Example 1.20. Write the line L through the points P1 = (−3, 1, −4) and P2 = (4, 4, −6) in
parametric form.
Solution: By formula (1.21), L consists of the points ( x, y, z) such that

x = −3 + 7 t, y = 1 + 3 t, z = −4 − 2 t, for − ∞ < t < ∞.

Distance from a point to a line

Let L be a line in R3 in vector form as r + tv (for −∞ < t < ∞), P
and let P be a point not on L. The distance d from P to L is the w d
length of the line segment from P to L which is perpendicular to L L
θ
(see Figure 1.5.4). Pick a point Q on L, and let w be the vector from Q v
Q to P . If θ is the angle between w and v, then d = k w k sin θ . So
Figure 1.5.4
since k v × w k = k v k k w k sin θ and v 6= 0, then:

kv × wk
d= . (1.23)
kvk
In other words, d is the hight of the parallelogram with adjacent sides v and w. Since its
area is k v × w k and its base k v k, we get the expression (1.23).
36 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

Example 1.21. Find the distance d from the point P = (1, 1, 1) to the line L in Example 1.20.
Solution: From Example 1.20, we see that we can represent L in vector form as: r + tv, for
−−→
r = (−3, 1, −4) and v = (7, 3, −2). Since the point Q = (−3, 1, −4) is on L, then for w = QP =
(1, 1, 1) − (−3, 1, −4) = (4, 0, 5), we have:
¯ ¯
¯ i j k ¯ ¯ ¯ ¯ ¯ ¯ ¯
¯ ¯ ¯ 3 −2 ¯ ¯ 7 −2 ¯ ¯ 7 3 ¯
v × w = ¯ 7 3 −2 ¯ = ¯
¯ ¯ ¯ ¯ i−¯
¯ ¯ j+¯¯ ¯ k = 15 i − 43 j − 12 k , so
¯ 4 0 ¯ 0 5 ¯ 4 5 ¯ 4 0 ¯
5
° ° p p
k v × w k ° 15 i − 43 j − 12 k ° 152 + (−43)2 + (−12)2 2218
d= = ° ° = p = p ≈ 5.98.
kvk ° (7, 3, −2) ° 2 2
7 + 3 + (−2) 2 62

Two lines
It is clear that two lines L 1 and L 2 , represented in vector form as r1 + sv1 and r2 + tv2 ,
respectively, are parallel (denoted as L 1 ∥ L 2 ) if v1 and v2 are parallel. Also, L 1 and L 2 are
perpendicular (denoted as L 1 ⊥ L 2 ) if v1 and v2 are perpendicular.
In 2-dimensional space, two lines are either identical, parallel, or they z
intersect. In 3-dimensional space, there is an additional possibility: two L1
lines can be skew, that is, they do not intersect but they are not parallel.
L2 y
However, even though they are not parallel, skew lines are on parallel
planes (see Figure 1.5.5). 0
3 x
To determine whether two lines in R intersect, it is often easier to use
the parametric representation of the lines. In this case, you should use dif- Figure 1.5.5
ferent parameter variables (usually s and t) for the lines, since the values of the parameters
may not be the same at the point of intersection. Setting the two ( x, y, z) triples equal will
result in a system of 3 equations in 2 unknowns ( s and t).

Example 1.22. Find the point of intersection (if any) of the following lines:
x+1 y−2 z−1 y−8 z+3
= = and x+3 = = .
3 2 −1 −3 2
Solution: First we write the lines in parametric form, with parameters s and t:

x = −1 + 3 s, y = 2 + 2 s, z = 1−s and x = −3 + t, y = 8 − 3 t, z = −3 + 2 t.

The lines intersect when (−1 + 3 s, 2 + 2 s, 1 − s) = (−3 + t, 8 − 3 t, −3 + 2 t) for some s, t:

−1 + 3 s = −3 + t : ⇒ t = 2 + 3 s,
2 + 2 s = 8 − 3 t : ⇒ 2 + 2 s = 8 − 3(2 + 3 s) = 2 − 9 s ⇒ 2 s = −9 s ⇒ s = 0 ⇒ t = 2 + 3(0) = 2,
1 − s = −3 + 2 t : 1 − 0 = −3 + 2(2) ⇒ 1 = 1. X (Note that we had to check this.)

Letting s = 0 in the equations for the first line, or letting t = 2 in the equations for the second
line, gives the point of intersection (−1, 2, 1).
 1.5 Lines and Planes 37

Plane through a point, perpendicular to a vector

Let P be a plane in R3 , and suppose it contains a point P0 = ( x0 , y0 , z0 ). Let n = (a, b, c) be
a nonzero vector which is perpendicular to the plane P . Such a vector is called a normal
vector (or just a normal) to the plane. Now let ( x, y, z) be any point in the plane P . Then
the vector r = ( x − x0 , y − y0 , z − z0 ) lies in the plane P (see Figure 1.5.6). So if r 6= 0, then r ⊥ n
and hence n · r = 0. And if r = 0 then we still have n · r = 0.
n

r
(x, y, z) (x0 , y0 , z0 )

Figure 1.5.6 The plane P.

Conversely, if ( x, y, z) is any point in R3 such that r = ( x − x0 , y − y0 , z − z0 ) 6= 0 and n · r = 0,

then r ⊥ n and so ( x, y, z) lies in P . This proves the following theorem:

Theorem 1.18. Let P be a plane in R3 , let ( x0 , y0 , z0 ) be a point in P , and let n = (a, b, c) be a

nonzero vector which is perpendicular to P . Then P consists of the points ( x, y, z) satisfying
the vector equation:
n · r = 0, (1.24)
where r = ( x − x0 , y − y0 , z − z0 ), or equivalently:

a( x − x0 ) + b( y − y0 ) + c( z − z0 ) = 0. (1.25)

The above equation is called the point-normal form of the plane P .

Example 1.23. Find the equation of the plane P containing the point (−3, 1, 3) and perpen-
dicular to the vector n = (2, 4, 8).
Solution: By formula (1.25), the plane P consists of all points ( x, y, z) such that:

2( x + 3) + 4( y − 1) + 8( z − 3) = 0.

If we multiply out the terms in formula (1.25) and combine the constant terms, we get an
equation of the plane in normal form:

ax + b y + cz + d = 0. (1.26)

For example, the normal form of the plane in Example 1.23 is 2 x + 4 y + 8 z − 22 = 0.

38 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

Plane containing three noncollinear points

In 2-dimensional and 3-dimensional space, two points determine a line. Two points do not
determine a plane in R3 . In fact, three collinear points (that is, all three on the same line) do
not determine a plane; an infinite number of planes would contain the line on which those
three points lie. However, three noncollinear points do determine a plane. For if Q , R and
−−→ −−→
S are noncollinear points in R3 , then QR and QS are nonzero vectors which are not parallel
−−→ −−→ −−→
(by noncollinearity), and so their cross product QR × QS is perpendicular to both QR and
−−→ −−→ −−→
QS . So QR and QS (and hence Q , R and S ) lie in the plane through the point Q with normal
−−→ −−→
vector n = QR × QS (see Figure 1.5.7).
−−→ −−→
n = QR × QS

−−→
QR
Q
R −−→
S QS

Figure 1.5.7 Noncollinear points Q, R, S.

Example 1.24. Find the equation of the plane P containing the points (2, 1, 3), (1, −1, 2) and
(3, 2, 1).
−−→
Solution: Let Q = (2, 1, 3), R = (1, −1, 2) and S = (3, 2, 1). Then for the vectors QR = (−1, −2, −1)
−−→
and QS = (1, 1, −2), the plane P has a normal vector
−−→ −−→
n = QR × QS = (−1, −2, −1) × (1, 1, −2) = (5, −3, 1).

So using formula (1.25) with the point Q (we could also use R or S ), the plane P consists of
all points ( x, y, z) such that:

5( x − 2) − 3( y − 1) + ( z − 3) = 0,

or in normal form,

5 x − 3 y + z − 10 = 0.

We mentioned earlier that skew lines in R3 lie on separate, parallel planes. So two skew
lines do not determine a plane. But two (nonidentical) lines which either intersect or are
parallel do determine a plane. In both cases, to find the equation of the plane that contains
those two lines, simply pick from the two lines a total of three noncollinear points (one point
from one line and two points from the other), then use the technique above, as in Example
1.24, to write the equation. We will leave examples of this as exercises for the reader.
 1.5 Lines and Planes 39

Distance between a point and a plane

The distance between a point in R3 and a plane is the length of the line segment from
that point to the plane which is perpendicular to the plane. The following theorem gives a
formula for that distance.
Theorem 1.19. Let Q = ( x0 , y0 , z0 ) be a point in R3 , and let P be a plane with normal form
ax + b y + cz + d = 0 that does not contain Q . Then the distance D from Q to P is:

| ax0 + b y0 + cz0 + d |
D= p . (1.27)
a2 + b 2 + c 2

Proof: Let R = ( x, y, z) be any point in the plane P (so that ax + b y + cz + d = 0) and let
−−→
r = RQ = ( x0 − x, y0 − y, z0 − z). Then r 6= 0 since Q does not lie in P . From the normal form
equation for P , we know that n = (a, b, c) is a normal vector for P . Now, any plane divides
R3 into two disjoint parts. Assume that n points toward the side of P where the point Q
is located. Place n so that its initial point is at R , and let θ be the angle between r and
n. Then 0◦ < θ < 90◦ , so cos θ > 0. Thus, the distance D is cos θ k r k = | cos θ | k r k (see Figure
1.5.8).
n Q

r D
D
θ
P
R

Figure 1.5.8
n· r
By Theorem 1.6 in Section 1.3, we know that cos θ = , so
knk krk

|n · r| | n · r | | a( x0 − x) + b( y0 − y) + c( z0 − z) |
D = | cos θ | k r k = krk = = p
knk krk knk a2 + b 2 + c 2
| ax0 + b y0 + cz0 − (ax + b y + cz) | | ax0 + b y0 + cz0 − (− d ) | | ax0 + b y0 + cz0 + d |
= p = p = p .
a2 + b 2 + c 2 a2 + b 2 + c 2 a2 + b 2 + c 2
If n points away from the side of P where the point Q is located, then 90◦ < θ < 180◦ and
so cos θ < 0. The distance D is then | cos θ | k r k, and thus repeating the same argument as
above still gives the same result. QED

Example 1.25. Find the distance D from (2, 4, −5) to the plane from Example 1.24.
Solution: Recall that the plane is given by 5 x − 3 y + z − 10 = 0. So
| 5(2) − 3(4) + 1(−5) − 10 | |−17 | 17
D= p = p = p ≈ 2.87.
52 + (−3)2 + 12 35 35
40 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

Line of intersection of two planes

Note that two planes are parallel if they have normal vectors that
are parallel, and the planes are perpendicular if their normal vectors
are perpendicular.
L
Suppose that two planes P1 and P2 with normal vectors n1 and n2 ,
respectively, intersect in a line L (see Figure 1.5.9). Since n1 × n2 ⊥ n1 ,
then n1 × n2 is parallel to the plane P1 . Likewise, n1 × n2 ⊥ n2 means that Figure 1.5.9
n1 × n2 is also parallel to P2 . Thus, n1 × n2 is parallel to the intersection
of P1 and P2 , which is L. Thus, we can write L in the following vector form:

L : r + t(n1 × n2 ) , for − ∞ < t < ∞ (1.28)

where r is any vector pointing to a point belonging to both planes. To find a point in both
planes, find a common solution ( x, y, z) to the two normal form equations of the planes. This
can often be made easier by setting one of the coordinate variables to zero, which leaves you
to solve two equations in just two unknowns.

Example 1.26. Find the line of intersection L of the planes 5 x − 3 y + z − 10 = 0 and 2 x + 4 y −

z + 3 = 0.
Solution: The plane 5 x − 3 y + z − 10 = 0 has normal vector n1 = (5, −3, 1) and the plane 2 x +
4 y − z + 3 = 0 has normal vector n2 = (2, 4, −1). Since n1 and n2 are not scalar multiples, then
the two planes are not parallel and hence will intersect. A point ( x, y, z) on both planes will
satisfy the following system of two equations in three unknowns:

5 x − 3 y + z − 10 = 0,
2 x + 4 y − z + 3 = 0.

Set x = 0 (why is that a good choice?). Then the above equations are reduced to:

−3 y + z − 10 = 0,
4 y − z + 3 = 0.

The second equation gives z = 4 y + 3, substituting that into the first equation gives y = 7.
Then z = 31, and so the point (0, 7, 31) is on L. Since n1 × n2 = (−1, 7, 26), then L is given by:

r + t(n1 × n2 ) = (0, 7, 31) + t(−1, 7, 26), for − ∞ < t < ∞

or in parametric form:

x = − t, y = 7 + 7 t, z = 31 + 26 t, for − ∞ < t < ∞

Projections
 1.5 Lines and Planes 41

Assume we need to find the orthogonal projection S of the given point Q with the position
vector q to the line L given by parametric equation r + tv.
Note that S is the point of intersection of line L and the plane P thru Q perpendicular to
v. This plane P is given by the equation (x − q) · v = 0 with unknown x.
Since S belongs to L, its position vector is r + tv some t. Since it lies on the plane, we get

(r + tv − q) · v = 0.

Solving for t, we get

(q − r) · v
t= .
v·v
Therefore,
(q − r) · v
r+ v
v·v
is the position vector of the projection of P on L.
Note that S is also the projecction of point R with the position vector r to the plane P .
Therefore, the same formula can be used to find the projection of to the plane.

Example 1.27. Find the projections of the point Q = (1, 1, 1) to the line x = 1 + 4 t, y = 2 + 5 t, z =
3 + 6 t.
Solution: The vector form of the parametric equation is (1, 2, 3) + t(4, 5, 6). Applying the
formula above, we get

(1 − 1)4 + (1 − 2)5 + (1 − 3)6 17

(1, 2, 3) + (4, 5, 6) = (1, 2, 3) + (4, 5, 6)
42 + 52 + 62 77
145 239 333
= ( 77 , 77 , 77 )
≈ (1.88, 3.1, 4.32)

is the position vector of the projection.

Exercises
A
For Exercises 1–4, write the line L through the point P and parallel to the vector v in the
following forms: (a) vector, (b) parametric, and (c) symmetric.

1. P = (2, 3, −2), v = (5, 4, −3); 2. P = (3, −1, 2), v = (2, 8, 1);

3. P = (2, 1, 3), v = (1, 0, 1); 4. P = (0, 0, 0), v = (7, 2, −10).

For Exercises 5–6, write the line L through the points P1 and P2 in parametric form.

5. P1 = (1, −2, −3), P2 = (3, 5, 5); 6. P1 = (4, 1, 5), P2 = (−2, 1, 3).

For Exercises 7–8, (a) find the distance d from the point P to the line L (b) find the orthogonal
projection of P to L
42 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

7. P = (1, −1, −1), L : x = −2 − 2 t, y = 4 t, z = 7 + t;

8. P = (0, 0, 0), L : x = 3 + 2 t, y = 4 + 3 t, z = 5 + 4 t.

For Exercises 9–10, find the point of intersection (if any) of the given lines.

9. x = 7 + 3 s, y = −4 − 3 s, z = −7 − 5 s and x = 1 + 6 t, y = 2 + t, z = 3 − 2 t;
x−6 x − 11 y − 14 z + 9
10. = y+3 = z and = = .
4 3 −6 2
For Exercises 11–12, write the normal form of the plane P containing the point Q and per-
pendicular to the vector n.

11. Q = (5, 1, −2), n = (4, −4, 3); 12. Q = (6, −2, 0), n = (2, 6, 4).

For Exercises 13–14, write the normal form of the plane containing the given points.

13. (1, 0, 3), (1, 2, −1), (6, 1, 6); 14. (−3, 1, −3), (4, −4, 3), (0, 0, 1).

15. Write the normal form of the plane containing the lines from Exercise 9.

16. Write the normal form of the plane containing the lines from Exercise 10.

For Exercises 17–18, (a) find the distance D from the point Q to the plane P and (b) find the
projection of Q to the plane P

17. Q = (4, 1, 2), P : 3 x − y − 5 z + 8 = 0; 18. Q = (0, 2, 0), P : −5 x + 2 y − 7 z + 1 = 0.

For Exercises 19–20, find the line of intersection (if any) of the given planes.

19. x + 3 y + 2 z − 6 = 0, 2 x − y + z + 2 = 0; 20. 3 x + y − 5 z = 0, x + 2 y + z + 4 = 0.

B
x−6
21. Find the point(s) of intersection (if any) of the line = y + 3 = z with the plane
4
x + 3 y + 2 z − 6 = 0. (Hint: Put the equations of the line into the equation of the plane.)

22. Explain why the following formula

−−→ −−→ −−→
|P A · (PQ × AB)|
−−→ −−→
|PQ × AB|

gives the distance between the skew lines AB and PQ .

 1.6 Elementary surfaces 43

1.6 Elementary surfaces

In the previous section we discussed planes in Euclidean space. A plane is an example of
a surface, which we will define informally8 as the solution set of the equation F ( x, y, z) = 0
in R3 , for some real-valued function F . For example, a plane given by ax + b y + cz + d = 0
is the solution set of F ( x, y, z) = 0 for the function F ( x, y, z) = ax + b y + cz + d . Surfaces are
2-dimensional. The plane is the simplest surface, since it is “flat”. In this section we will
look at some surfaces that are more complex, the most important of which are the sphere
and the cylinder.

Definition 1.9. A sphere S is the set of all points ( x, y, z) in R3 which are a fixed distance r
(called the radius) from a fixed point P0 = ( x0 , y0 , z0 ) (called the center of the sphere):

S = { ( x, y, z) : ( x − x0 )2 + ( y − y0 )2 + ( z − z0 )2 = r 2 }. (1.29)

Using vector notation, this can be written in the equivalent form:

S = { x : k x − x0 k = r }, (1.30)

where x = ( x, y, z) and x0 = ( x0 , y0 , z0 ) are vectors.

Figure 1.6.1 illustrates the vectorial approach to spheres.

z
z k x − x0 k = r
kxk = r
x x
x − x0

y ( x0 , y0 , z0 )
0
x0

y
0
x
x
(a) radius r , center (0, 0, 0) (b) radius r , center ( x0 , y0 , z0 )

Figure 1.6.1 Spheres in R3 .

Note in Figure 1.6.1(a) that the intersection of the sphere with the x y-plane is a circle of
radius r (that is, a great circle, given by x2 + y2 = r 2 as a subset of R2 ). Similarly for the
intersections with the xz-plane and the yz-plane. In general, a plane intersects a sphere
either at a single point or in a circle.
8 See O’N EILL for a deeper and more rigorous discussion of surfaces.
44 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

Example 1.28. Find the intersection of the sphere x2 + y2 + z2 = 169 with the plane z = 12.
z
p The sphere is centered at the origin and has radius
Solution:
13 = 169, so it does intersect the plane z = 12. Putting z = 12
z = 12 into the equation of the sphere gives
x2 + y2 + 122 = 169, y
0
x2 + y2 = 169 − 144 = 25 = 52
which is a circle of radius 5 centered at (0, 0, 12), parallel to
x
the x y-plane (see Figure 1.6.2).
Figure 1.6.2
If the equation in formula (1.29) is multiplied out, we get an equation of the form:
x2 + y2 + z2 + ax + b y + cz + d = 0 (1.31)
for some constants a, b, c and d . Conversely, an equation of this form may describe a sphere,
which can be determined by completing the square for the x, y and z variables.
Note that the equation (1.31) could be written as
k x k 2 + v · x + d = 0,
where x = ( x, y, z) and v = (a, b, c).
Example 1.29. Is 2 x2 + 2 y2 + 2 z2 − 8 x + 4 y − 16 z + 10 = 0 the equation of a sphere?
Solution: Dividing both sides of the equation by 2 gives
x 2 + y 2 + z 2 − 4 x + 2 y − 8 z + 5 = 0,
( x2 − 4 x + 4) + ( y2 + 2 y + 1) + ( z2 − 8 z + 16) + 5 − 4 − 1 − 16 = 0,
( x − 2)2 + ( y + 1)2 + ( z − 4)2 = 16
which is a sphere of radius 4 centered at (2, −1, 4).
Example 1.30. Find the points(s) of intersection (if any) of the sphere from Example 1.29
and the line x = 3 + t, y = 1 + 2 t, z = 3 − t.
Solution: Put the equations of the line into the equation of the sphere, which was ( x − 2)2 +
( y + 1)2 + ( z − 4)2 = 16, and solve for t:
(3 + t − 2)2 + (1 + 2 t + 1)2 + (3 − t − 4)2 = 16,
( t + 1)2 + (2 t + 2)2 + (− t − 1)2 = 16,
6 t2 + 12 t − 10 = 0.
4
The quadratic formula gives the solutions t = −1 ± p . Putting those two values into the
6
equations of the line gives the following two points of intersection:
µ ¶ µ ¶
4 8 4 4 8 4
2 + p , −1 + p , 4 − p and 2 − p , −1 − p , 4 + p .
6 6 6 6 6 6
 1.6 Elementary surfaces 45

If two spheres intersect, they do so either at a single point or in a circle.

Example 1.31. Find the intersection (if any) of the spheres x2 + y2 + z2 = 25 and x2 + y2 + ( z −
2)2 = 16.
Solution: For any point ( x, y, z) on both spheres, we see that

x2 + y2 + z2 = 25 ⇒ x2 + y2 = 25 − z2 , and
x2 + y2 + ( z − 2)2 = 16 ⇒ x2 + y2 = 16 − ( z − 2)2 , so
16 − ( z − 2)2 = 25 − z2 ⇒ 4z − 4 = 9 ⇒ z = 13/4
⇒ x2 + y2 = 25 − (13/4)2 = 231/16.
p
231 231
∴ The intersection is the circle x2 + y2 = 16 in the plane z = 13/4. It has radius 4 ≈ 3.8
and centered at (0, 0, 13
4 ).

The cylinders that we will consider are right circular cylinders. These are cylinders ob-
tained by moving a line L along a circle C in R3 in a way so that L is always perpendicular
to the plane containing C . We will only consider the cases where the plane containing C is
parallel to one of the three coordinate planes (see Figure 1.6.3).

z z

r
z

r y
y r
y 0
0
0
x x x
(a) x2 + y2 = r 2 , any z (b) x2 + z2 = r 2 , any y (c) y2 + z2 = r 2 , any x

Figure 1.6.3 Cylinders in R3 .

For example, the equation of a cylinder whose base circle C lies in the x y-plane and is
centered at (a, b, 0) and has radius r is

( x − a)2 + ( y − b)2 = r 2 , (1.32)

where the value of the z coordinate is unrestricted. Similar equations can be written when
the base circle lies in one of the other coordinate planes. A plane intersects a right circular
cylinder in a circle, ellipse, or one or two lines, depending on whether that plane is parallel,
oblique9 , or perpendicular, respectively, to the plane containing C . The intersection of a
surface with a plane is called the trace of the surface.
9 That is, at an angle strictly between 0◦ and 90◦ .
46 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

The equations of spheres and cylinders are examples of second-degree equations in R3 ;

that is, equations of the form

Ax2 + B y2 + Cz2 + Dx y + Exz + F yz + Gx + H y + I z + J = 0 (1.33)

for some constants A , B, . . . , J . If the above equation is not that of a sphere, cylinder, plane,
line or point, then the resulting surface is called a quadric surface.
One type of quadric surface is the ellipsoid, given z
c
by an equation of the form:

x 2 y2 z 2
+ + = 1. (1.34) y
a2 b 2 c 2 0 b
a
In the case where a = b = c, this is just a sphere.
In general, an ellipsoid is egg-shaped (think of an
ellipse rotated around its major axis). Its traces in x
the coordinate planes are ellipses. Figure 1.6.4 Ellipsoid

Two other types of quadric surfaces are the hyperboloid of one sheet, given by an
equation of the form:
x 2 y2 z 2
+ − =1 (1.35)
a2 b 2 c 2
and the hyperboloid of two sheets, whose equation has the form:

x 2 y2 z 2
− − = 1. (1.36)
a2 b 2 c 2

z z

y y
0 0

x x

Figure 1.6.5 Hyperboloid of one sheet. Figure 1.6.6 Hyperboloid of two sheets.
 1.6 Elementary surfaces 47

For the hyperboloid of one sheet, the trace in any plane parallel to the x y-plane is an
ellipse. The traces in the planes parallel to the xz- or yz-planes are hyperbolas (see Figure
1.6.5), except for the special cases x = ±a and y = ± b; in those planes the traces are pairs of
intersecting lines (see Exercise 8).
For the hyperboloid of two sheets, the trace in any plane parallel to the x y- or xz-plane is
a hyperbola (see Figure 1.6.6). There is no trace in the yz-plane. In any plane parallel to the
yz-plane for which | x | > | a |, the trace is an ellipse.
z
The elliptic paraboloid is another type of quadric surface,
whose equation has the form:

x 2 y2 z
+ = . (1.37)
a2 b 2 c

The traces in planes parallel to the x y-plane are ellipses, though

in the x y-plane itself the trace is a single point. The traces in
planes parallel to the xz- or yz-planes are parabolas. Figure y
1.6.7 shows the case where c > 0. When c < 0 the surface is 0
turned downward. In the case where a = b, the surface is called x
a paraboloid of revolution, which is often used as a reflecting sur- Figure 1.6.7 Paraboloid
face in vehicle headlights.10
A more complicated quadric surface is the hyperbolic paraboloid, given by:

x 2 y2 z
− = . (1.38)
a2 b 2 c

The hyperbolic paraboloid can be tricky to draw; using graphing software on a computer
can make it easier. For example, Figure 1.6.8 was created using the free Gnuplot package.
It shows the graph of the hyperbolic paraboloid z = y2 − x2 , which is the special case where
a = b = 1 and c = −1 in equation (1.38). The mesh lines on the surface are the traces in
planes parallel to the coordinate planes. So we see that the traces in planes parallel to the
xz-plane are parabolas pointing upward, while the traces in planes parallel to the yz-plane
are parabolas pointing downward. Also, notice that the traces in planes parallel to the x y-
plane are hyperbolas, though in the x y-plane itself the trace is a pair of intersecting lines
through the origin. This is true in general when c < 0 in equation (1.38). When c > 0, the
surface would be similar to that in Figure 1.6.8, only rotated 90◦ around the z-axis and the
nature of the traces in planes parallel to the xz- or yz-planes would be reversed.

10 For a discussion of this see pp. 157–158 in H ECHT.

48 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

100

50

z 0
-10
-50
-5
-100
-10
0
-5 x
0 5
y 5
10 10

Figure 1.6.8 Hyperbolic paraboloid.

The last type of quadric surface that we will consider is the z

elliptic cone, which has an equation of the form:

x 2 y2 z 2
+ − = 0. (1.39)
a2 b 2 c 2
y
The traces in planes parallel to the x y-plane are ellipses, ex-
0
cept in the x y-plane itself where the trace is a single point.
The traces in planes parallel to the xz- or yz-planes are hyper-
bolas, except in the xz- and yz-planes themselves where the
traces are pairs of intersecting lines. x
Notice that every point on the elliptic cone is on a line which
lies entirely on the surface; in Figure 1.6.9 these lines all go Figure 1.6.9 Elliptic cone
through the origin. This makes the elliptic cone an example of
a ruled surface. The cylinder is also a ruled surface.
What may not be as obvious is that both the hyperboloid of one sheet and the hyperbolic
paraboloid are ruled surfaces. In fact, on both surfaces there are two lines through each
point on the surface (see Exercises 11–12). Such surfaces are called doubly ruled surfaces,
and the pairs of lines are called a regulus.
 1.6 Elementary surfaces 49

It is clear that for each of the six types of quadric surfaces that we discussed, the surface
can be translated away from the origin (say, by replacing x2 by ( x − x0 )2 in its equation). It can
be proved11 that every quadric surface can be translated and/or rotated so that its equation
matches one of the six types that we described.
For example, z = kx y is a case of equation (1.33) with “mixed” variables; namely D 6= 0,
so that we get an x y term. This equation does not match any of the types we considered.
However, by rotating the x-pand y-axes by p 45◦ in the x y-plane by means of the coordinate
′ ′
transformation x = ( x − y )/ 2, y = ( x + y )/ 2, z = z′ , then z = kx y becomes the hyperbolic
′ ′

paraboloid z′ = k( x′ )2 − k( y′ )2 in the ( x′ , y′ , z′ ) coordinate system.

That is, the equation
z = kx y (1.40)

describes a hyperbolic paraboloid as in equation (1.38), but rotated 45◦ in the x y-plane.
Exercises
A
For Exercises 1–4, determine if the given equation describes a sphere. If so, find its radius
and center.

1. x2 + y2 + z2 − 4 x − 6 y − 10 z + 37 = 0; 2. x2 + y2 + z2 + 2 x − 2 y − 8 z + 19 = 0;

3. 2 x2 + 2 y2 + 2 z2 + 4 x + 4 y + 4 z − 44 = 0; 4. x2 + y2 − z2 + 12 x + 2 y − 4 z + 32 = 0.

5. Find the point(s) of intersection of the sphere ( x − 3)2 + ( y + 1)2 + ( z − 3)2 = 9 and the line
x = −1 + 2 t , y = −2 − 3 t , z = 3 + t .

6. Find the intersection of the spheres x2 + y2 + z2 = 9 and ( x − 4)2 + ( y + 2)2 + ( z − 4)2 = 9.

7. Find the intersection of the sphere x2 + y2 + z2 = 9 and the cylinder x2 + y2 = 4.

x2 y2 z2
8. Find the trace of the hyperboloid of one sheet a2
+ b2 − c2
= 1 in the plane x = a, and the
trace in the plane y = b.

x2 y2 z
9. Find the trace of the hyperbolic paraboloid a2
− b2 = c in the x y-plane.

10. It can be shown that any four noncoplanar points (that is, points that do not lie in the
same plane) determine a sphere.12 Find the equation of the sphere that passes through
the points (0, 0, 0), (0, 0, 2), (1, −4, 3) and (0, −1, 3). (Hint: Equation (1.31))
11 See Ch. 7 in P OGORELOV.
12 See W ELCHONS and K RICKENBERGER, p. 160, for a proof.
 50 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

11. Show that the hyperboloid of one sheet is a doubly ruled surface; that is, each point on
the surface is on two lines lying entirely on the surface. (Hint: Write equation (1.35) as
x2 2 y2
a2
− zc2 = 1 − b2 , factor each side. Recall that two planes intersect in a line.)

12. Show that the hyperbolic paraboloid is a doubly ruled surface. (Hint: Exercise 11)
z
13. Let S be the sphere with radius 1 centered at (0, 0, 1), (0, 0, 2)
and let S ∗ be S without the “north pole” point (0, 0, 2). Let
(a, b, c) be an arbitrary point on S ∗ . Then the line passing (a, b, c)
S
through (0, 0, 2) and (a, b, c) intersects the x y-plane at some 1
point ( x, y, 0), as in Figure 1.6.10. Find this point ( x, y, 0) in y
terms of a, b and c. (x, y, 0) 0
(Note: Every point in the x y-plane can be matched with a x
point on S ∗ , and vice versa, in this manner. This method is Figure 1.6.10
called stereographic projection, which essentially identifies
all of R2 with a “punctured” sphere.)

14. Given two points P and Q in the space consider the set of points X such that the distance
from X to P is twice larger than the distance from X to Q . Show that this set is a sphere.
Find its radius and center if P = (1, 2, 3) and Q = (2, 4, 5).

15. Show that the equidistant set from a plane and a point not on the plane is formed by a
elliptic paraboloid. (Hint:Use the coordinate system with the given pane as the x y-plane.)
 1.7 Curvilinear Coordinates 51

1.7 Curvilinear Coordinates

The Cartesian coordinates of a point ( x, y, z) are determined by z (x, y, z)

following straight paths starting from the origin: first along the
z
x-axis, then parallel to the y-axis, then parallel to the z-axis, as y
in Figure 1.7.1. In curvilinear coordinate systems, these paths can 0
x
be curved. The two types of curvilinear coordinates which we will
consider are cylindrical and spherical coordinates. Instead of ref- y
x
erencing a point in terms of sides of a rectangular parallelepiped,
Figure 1.7.1
as with Cartesian coordinates, we will think of the point as ly-
ing on a cylinder or sphere. Cylindrical coordinates are often used when there is symmetry
around the z-axis; spherical coordinates are useful when there is symmetry about the origin.
If a problem is given in curvilinear coordinates, the typical solution consist of (1) convert-
ing the data in Cartesian coordinates, (2) solving it in the Cartesian coordinates and (3)
converting the results back to the original curvilinear coordinates if necessarily. Unless you
know what you are doing, we suggest to follow this procedure.
Let P = ( x, y, z) be a point in Cartesian coordinates in R3 . Then the cylindrical coordi-
nates ( r, θ , z) and the spherical coordinates (ρ , θ , φ) of P ( x, y, z) are defined as follows:13

Cylindrical coordinates ( r, θ , z): z P(x, y, z)

q
x = r cos θ , r= x 2 + y2 , z
¡ y¢ y
0
y = r sin θ , θ = tan−1 x , r
x θ
z = z, z = z,
y P0 (x, y, 0)
x
where 0 ≤ θ ≤ π if y ≥ 0 and π < θ < 2π if y < 0.
Figure 1.7.2
Cylindrical coordinates

Spherical coordinates (ρ , θ , φ): z P(x, y, z)

q ρ
x = ρ sin φ cos θ , ρ= x 2 + y2 + z 2 , z
φ
¡ ¢ y
−1 y 0
y = ρ sin φ sin θ , θ = tan x ,
³ ´ x θ
−1 p z
z = ρ cos φ, φ = cos ,
x2 + y2 + z2 y P0 (x, y, 0)
x
where 0 ≤ θ ≤ π if y ≥ 0 and π < θ < 2π if y < 0. Figure 1.7.3
Spherical coordinates
Both θ and φ are measured in radians. Note that r ≥ 0, 0 ≤ θ < 2π, ρ ≥ 0 and 0 ≤ φ ≤ π.
Also, θ is undefined when ( x, y) = (0, 0), and φ is undefined when ( x, y, z) = (0, 0, 0).
13 This “standard” definition of spherical coordinates used by mathematicians results in a left-handed system.

For this reason, physicists usually switch the definitions of θ and φ to make (ρ , θ , φ) a right-handed system.
52 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

Assume P0 = ( x, y, 0) is the projection of P upon the x y-plane and ( r, θ , z) are cylindrical

coordinates of P = ( x, y, z). Then ( r, θ ) are the polar coordinates of P0 (see Figure 1.7.2).
In the spherical coordinates ρ is length of the line segment from the origin to P , and φ be
the angle between that line segment and the positive z-axis (see Figure 1.7.3). The angle φ
is called the zenith angle.

Example 1.32. Convert the point (−2, −2, 1) from Cartesian coordinates to (a) cylindrical
and (b) spherical coordinates.
p p ¡ −2 ¢ 5π
Solution: (a) r = (−2)2 + (−2)2 = 2 2, θ = tan−1 − −1
2 = tan (1) = 4 , since y = −2 < 0.
¡ p 5π ¢
∴ (r, θ, z) = 2 2, 4 , 1 .
p p ¡ ¢
(b) ρ = (−2)2 + (−2)2 + 12 = 9 = 3, φ = cos−1 13 ≈ 1.23 radians.
¡ ¢
∴ (ρ , θ, φ) = 3, 54π , 1.23 .

For cylindrical coordinates ( r, θ , z), and constants r 0 , θ0 and z0 , we see from Figure 1.7.4
that the surface r = r 0 is a cylinder of radius r 0 centered along the z-axis, the surface θ = θ0
is a half-plane emanating from the z-axis, and the surface z = z0 is a plane parallel to the
x y-plane.
z z z
r0 z0

y y y
0
0 0

θ0
x x x
(a) r = r 0 (b) θ = θ0 (c) z = z0

Figure 1.7.4 Cylindrical coordinate surfaces.

For spherical coordinates (ρ , θ , φ), and constants ρ 0 , θ0 and φ0 , we see from Figure 1.7.5
that the surface ρ = ρ 0 is a sphere of radius ρ 0 centered at the origin, the surface θ = θ0 is a
half-plane emanating from the z-axis, and the surface φ = φ0 is a circular cone whose vertex
is at the origin.
Figures 1.7.4(a) and 1.7.5(a) show how these coordinate systems got their names.
Sometimes the equation of a surface in Cartesian coordinates can be transformed into a
simpler equation in some other coordinate system, as in the following example.

Example 1.33. Write the equation of the cylinder x2 + y2 = 4 in cylindrical coordinates.

p
Solution: Since r = x2 + y2 , then the equation in cylindrical coordinates is r = 2.

Using spherical coordinates to write the equation of a sphere does not necessarily make
the equation simpler, if the sphere is not centered at the origin.
 1.7 Curvilinear Coordinates 53

z
z z

ρ0
φ0
y y
0
0
y
θ0 0
x x x
(a) ρ = ρ 0 (b) θ = θ0 (c) φ = φ0

Figure 1.7.5 Spherical coordinate surfaces.

Example 1.34. Write the equation ( x − 2)2 + ( y − 1)2 + z2 = 9 in spherical coordinates.

Solution: Multiplying the equation out gives

x2 + y2 + z2 − 4 x − 2 y + 5 = 9 , so we get
ρ 2 − 4ρ sin φ cos θ − 2ρ sin φ sin θ − 4 = 0 , or
ρ 2 − 2 sin φ (2 cos θ − sin θ ) ρ − 4 = 0 after combining terms.

Note that this actually makes it more difficult to figure out what the surface is, as opposed
to the Cartesian equation where you could immediately identify the surface as a sphere of
radius 3 centered at (2, 1, 0).

Example 1.35. Describe the surface given by θ = z in cylindrical coordinates.

Solution: This surface is called a helicoid. As the (vertical) z coordinate increases, so does
the angle θ , while the radius r is unrestricted. So this sweeps out a (ruled!) surface shaped
like a spiral staircase, where the spiral has an infinite radius. Figure 1.7.6 shows a section
of this surface restricted to 0 ≤ z ≤ 4π and 0 ≤ r ≤ 2.
Exercises
A
For Exercises 1–4, find the (a) cylindrical and (b) spherical coordinates of the point whose
Cartesian coordinates are given.
p p p p
1. (2, 2 3, −1); 2. (−5, 5, 6); 3. ( 21, − 7, 0); 4. (0, 2, 2).

For Exercises 5–7, write the given equation in (a) cylindrical and (b) spherical coordinates.

5. x2 + y2 + z2 = 25; 6. x2 + y2 = 2 y; 7. x2 + y2 + 9 z2 = 36.

B
8. Describe the intersection of the surfaces whose equations in spherical coordinates are
θ = π2 and φ = π4 .
54 CHAPTER 1. VECTORS IN EUCLIDEAN SPACE

14
12
10
8
z 6
4
-2
2 -1.5
-1
0-2 -0.5
-1.5 -1 0
-0.5 0.5 x
0 0.5 1
y 1 1.5 1.5
2 2

Figure 1.7.6 Helicoid θ = z.

9. Show that for a 6= 0, the equation ρ = 2a sin φ cos θ in spherical coordinates describes a
sphere centered at (a, 0, 0) with radius | a |.

C
10. Let P = (a, θ , φ) be a point in spherical coordinates, with a > 0 and 0 < φ < π. Then P
lies on the sphere ρ = a. Since 0 < φ < π, the line segment from the origin to P can be
extended to intersect the cylinder given by r = a (in cylindrical coordinates). Find the
cylindrical coordinates of that point of intersection.

11. Let P1 and P2 be points whose spherical coordinates are (ρ 1 , θ1 , φ1 ) and (ρ 2 , θ2 , φ2 ), respec-
tively. Let v1 be the vector from the origin to P1 , and let v2 be the vector from the origin
to P2 . For the angle γ between v1 and v2 , show that

cos γ = cos φ1 cos φ2 + sin φ1 sin φ2 cos( θ2 − θ1 ).

This formula is used in electrodynamics to prove the addition theorem for spherical har-
monics, which provides a general expression for the electrostatic potential at a point due
to a unit charge. See pp. 100–102 in J ACKSON.

12. Show that the distance d between the points P1 and P2 with cylindrical coordinates
 1.7 Curvilinear Coordinates 55

( r 1 , θ1 , z1 ) and ( r 2 , θ2 , z2 ), respectively, is
q
d= r 21 + r 22 − 2 r 1 r 2 cos( θ2 − θ1 ) + ( z2 − z1 )2 .

13. Show that the distance d between the points P1 and P2 with spherical coordinates
(ρ 1 , θ1 , φ1 ) and (ρ 2 , θ2 , φ2 ), respectively, is
q
d= ρ 21 + ρ 22 − 2ρ 1 ρ 2 [sin φ1 sin φ2 cos( θ2 − θ1 ) + cos φ1 cos φ2 ] .
2 Curves

2.1 Vector-Valued Functions

Now that we are familiar with vectors and their operations, we can begin discussing func-
tions whose values are vectors.

Definition 2.1. A vector-valued function of a real variable is a rule that associates a

vector f( t) with a real number t, where t is in R or its interval (called the domain of f). We
write f : D → R3 to denote that f is a mapping of D into R3 .

For example, f( t) = ti + t2 j + t3 k is a vector-valued function in R3 , defined for all real num-

bers t. We would write f : R → R3 . At t = 1 the value of the function is the vector i + j + k,
which in Cartesian coordinates has the terminal point (1, 1, 1).
A vector-valued function of a real variable can be written in component form as

f( t) = f 1 ( t)i + f 2 ( t)j + f 3 ( t)k or f( t) = ( f 1 ( t), f 2 ( t), f 3 ( t))

for some real-valued functions f 1 ( t), f 2 ( t), f 3 ( t), called the component functions of f. The first
form is often used when emphasizing that f( t) is a vector, and the second form is useful when
considering just the terminal points of the vectors.
z
3
Example 2.1. Define f : R → R by f( t) = (cos t, sin t, t).
This is a parametric equation of a helix (see Figure 1.8.1). As the
value of t increases, the terminal points of f( t) is spiraling upward.
y
For each t, the x- and y-coordinates of f( t) are x = cos t and y = sin t, f(2π) 0
so
f(0)
x2 + y2 = cos2 t + sin2 t = 1.
x
Thus, f( t) lies on the surface of the right circular cylinder x2 + y2 = 1 Figure 2.1.1
for any t.

Since each of the three component functions are real-valued, it will sometimes be the case
that results from single-variable calculus can simply be applied to each of the component
functions to yield a similar result for the vector-valued function. However, there are times
when such generalizations do not hold (see Exercise 13). The concept of a limit, though, can
be extended naturally to vector-valued functions, as in the following definition.

56
 2.1 Vector-Valued Functions 57

Definition 2.2. Let f( t) be a vector-valued function, let a be a real number and let c be a
vector. Then we say that the limit of f( t) as t approaches a equals c, written as lim f( t) = c,
t→ a
if lim k f( t) − c k = 0.
t→ a
Equivalently, if f( t) = ( f 1 ( t), f 2 ( t), f 3 ( t)), then
³ ´
lim f( t) = lim f 1 ( t), lim f 2 ( t), lim f 3 ( t) ,
t→ a t→ a t→ a t→ a

provided that all three limits on the right side exist.

The above definition shows that continuity and the derivative of vector-valued functions
can also be defined in terms of its component functions.

Definition 2.3. Let f( t) = ( f 1 ( t), f 2 ( t), f 3 ( t)) be a vector-valued function, and let a be a real
number in its domain. Then f( t) is continuous at a if lim f( t) = f(a). Equivalently, f( t) is
t→ a
continuous at a if and only if f 1 ( t), f 2 ( t), and f 3 ( t) are continuous at a.
df
The derivative of f( t) at a, denoted by f′ (a) or (a), is the limit
dt
f(a + h) − f(a)
f′ (a) = lim
h→0 h

if that limit exists. Equivalently, f′ (a) = ( f 1′ (a), f 2′ (a), f 3′ (a)), if the component derivatives exist.
We say that f( t) is differentiable at a if f′ (a) exists.

A real-valued function whose first derivative is continuous is called continuously differ-

entiable (or a C 1 function), and a function whose derivatives of all orders are continuous is
called smooth (or a C ∞ function). All the functions we will consider will be smooth.
Continuous vector valued functions are also called curves; in this
case the vector f( t) is usually regarded as its terminal point. A regular
curve f( t) is one whose derivative f′ ( t) is never the zero vector.
For example consider the plane curve f( t) = ( t2 , t3 ); it is so called
semicubical parabola shown on the picture. The curve has smooth
components but it is not regular since f′ ( t) = (2 t, 3 t2 ) vanish at t = 0. In
fact this curve does not look “smooth” at t = 0; it has so called cusp at
this point.
Recall that the derivative of a real-valued function of a single vari-
able is a real number, representing the slope of the tangent line to the
graph of the function at a point. Similarly, the derivative of a vector-
valued function is a tangent vector to the curve in space which the
function represents, and it lies on the tangent line to the curve (see
Figure 2.1.2).

Example 2.2. Let f( t) = (cos t, sin t, t). Then f′ ( t) = (− sin t, cos t, 1) for all t. The tangent line
58 CHAPTER 2. CURVES

z
f′ (a)
f(a L
+
f(a) h)
−
f(a f( t)
)
f(a + h) y
0
x

Figure 2.1.2 Tangent vector f′ (a) and tangent line L = f(a) + sf′ (a).

L to the curve at f(2π) = (1, 0, 2π) is L = f(2π) + s f′ (2π) = (1, 0, 2π) + s(0, 1, 1), or in parametric
form: x = 1, y = s, z = 2π + s for −∞ < s < ∞.

A scalar function is a real-valued function. Note that if u( t) is a scalar function and

f( t) is a vector-valued function, then their product, defined by ( u f)( t) = u( t) f( t) for all t, is a
vector-valued function (since the product of a scalar with a vector is a vector).
The basic properties of derivatives of vector-valued functions are summarized in the fol-
lowing theorem.

Theorem 2.1. Let f( t) and g( t) be differentiable vector-valued functions, let u( t) be a

differentiable scalar function, let k be a scalar, and let c be a constant vector. Then

d
(a) (c) = 0;
dt
d df
(b) ( kf) = k ;
dt dt
d df dg
(c) (f + g) = + ;
dt dt dt
d df dg
(d) (f − g) = − ;
dt dt dt
d du df
(e) ( u f) = f+u ;
dt dt dt
d df dg
(f) (f · g) = · g + f· ;
dt dt dt
d df dg
(g) (f × g) = ×g+f× .
dt dt dt

Proof: The proofs of parts (a)–(e) follow easily by differentiating the component functions
and using the rules for derivatives from single-variable calculus. We will prove part (f),
and leave the proof of part (g) as an exercise for the reader.
 2.1 Vector-Valued Functions 59

(f) Write f( t) = ( f 1 ( t), f 2 ( t), f 3 ( t)) and g( t) = ( g 1 ( t), g 2 ( t), g 3 ( t)), where the component functions
f 1 ( t), f 2 ( t), f 3 ( t), g 1 ( t), g 2 ( t), g 3 ( t) are all differentiable real-valued functions. Then

d d
(f( t) · g( t)) = ( f 1 ( t) g 1 ( t) + f 2 ( t) g 2 ( t) + f 3 ( t) g 3 ( t))
dt dt
d d d
= ( f 1 ( t) g 1 ( t)) + ( f 2 ( t) g 2 ( t)) + ( f 3 ( t) g 3 ( t))
dt dt dt
d f1 d g1 d f2 d g2 d f3 d g3
= ( t) g 1 ( t) + f 1 ( t) ( t) + ( t) g 2 ( t) + f 2 ( t) ( t) + ( t) g 3 ( t) + f 3 ( t) ( t)
dt dt dt dt dt dt
³df d f2 d f3 ´
1
= ( t), ( t), ( t) · ( g 1 ( t), g 2 ( t), g 3 ( t))
dt dt dt
³dg d g2 d g3 ´
1
+ ( f 1 ( t), f 2 ( t), f 3 ( t)) · ( t), ( t), ( t)
dt dt dt
df dg
= ( t) · g( t) + f( t) · ( t) for all t. QED
dt dt

Example 2.3. Suppose f( t) is differentiable. Find the derivative of k f( t) k.

Solution: Since k f( t) k is a real-valued function of t, then by the Chain Rule for real-valued
d d
functions, we know that k f( t) k2 = 2k f( t) k k f( t) k.
dt dt
d d
But k f( t) k2 = f( t) · f( t), so k f( t) k2 = (f( t) · f( t)). Hence, we have
dt dt
d d
2k f( t) k k f( t) k = (f( t) · f( t)) = f′ ( t) · f( t) + f( t) · f′ ( t) by Theorem 2.1(f), so
dt dt
= 2f′ ( t) · f( t) , so if k f( t) k 6= 0 then
d f′ ( t) · f( t)
k f( t) k = .
dt k f( t) k

d
We know that k f( t) k is constant if and only if k f( t) k = 0 for all t. Also, f( t) ⊥ f′ ( t) if and
dt
only if f′ ( t) · f( t) = 0. Thus, the above example shows this important fact:

If k f( t) k 6= 0, then k f( t) k is constant if and only if f( t) ⊥ f′ ( t) for all t.

This means that if a curve lies completely on a sphere (or circle) centered at the origin, then
the tangent vector f′ ( t) is always perpendicular to the position vector f( t).

³ cos t sin t −at ´

Example 2.4. The spherical spiral f( t) = p ,p ,p , for a 6= 0.
1 + a2 t2 1 + a2 t2 1 + a2 t2
Figure 2.1.3 shows the graph of the curve when a = 0.2. In the exercises, the reader will be
asked to show that this curve lies on the sphere x2 + y2 + z2 = 1 and to verify directly that
f′ ( t) · f( t) = 0 for all t.
60 CHAPTER 2. CURVES

1
0.8
0.6
0.4
z 0.2
0
-0.2
-0.4
-0.6 -1
-0.8 -0.8
-0.6
-1-1 -0.4
-0.2
-0.8 -0.6 0
-0.4 -0.2 0.2 x
0 0.4
0.2 0.4 0.6
y 0.6 0.8 0.8
1 1

Figure 2.1.3 Spherical spiral with a = 0.2.

Just as in single-variable calculus, higher-order derivatives of vector-valued functions are

obtained by repeatedly differentiating the (first) derivative of the function:

d ′ d ′′ dnf d ³ d n −1 f ´
f′′ ( t) = f ( t) , f′′′ ( t) = f ( t) , ... , = (for n = 2, 3, 4, . . .).
dt dt dt n dt dt n−1
We can use vector-valued functions to represent physical quantities, such as velocity, ac-
celeration, force, momentum, etc. For example, let the real variable t represent time elapsed
from some initial time ( t = 0), and suppose that an object of constant mass m is subjected
to some force so that it moves in space, with its position ( x, y, z) at time t a function of
t. That is, x = x( t), y = y( t), z = z( t) for some real-valued functions x( t), y( t), z( t). Call
r( t) = ( x( t), y( t), z( t)) the position vector of the object. We can define various physical quan-
tities associated with the object as follows:1

position: r( t) = ( x( t), y( t), z( t));

dr
velocity: v( t) = ṙ( t) = r′ ( t) =
dt
= ( x′ ( t), y′ ( t), z′ ( t));
1 We will often use the older dot notation for derivatives when physics is involved.
 2.1 Vector-Valued Functions 61

dv
acceleration: a( t) = v̇( t) = v′ ( t) =
dt
d2r
= r̈( t) = r′′ ( t) = 2
dt
′′ ′′ ′′
= ( x ( t), y ( t), z ( t));
momentum: p( t) = mv( t);
dp
force: F( t) = ṗ( t) = p′ ( t) = (Newton’s Second Law of Motion).
dt
The magnitude k v( t) k of the velocity vector is called the speed of the object. Note that since
the mass m is a constant, the force equation becomes the familiar F( t) = ma( t).
Example 2.5. Let r( t) = (5 cos t, 3 sin t, 4 sin t) be the position vector of an object at time t ≥ 0.
Find its (a) velocity and (b) acceleration vectors.
Solution: (a) v( t) = ṙ( t) = (−5 sin t, 3 cos t, 4 cos t).
(b) a( t) = v̇( t) = (−5 cos t, −3 sin t, −4 sin t).
p
Note that k r( t) k = 25 cos2 t + 25 sin2 t = 5 for all t, so by Example 2.3 we know that r( t) ·
ṙ( t) = 0 for all t (which we can verify from part (a)). In fact, k v( t) k = 5 for all t also. And not
only does r( t) lie on the sphere of radius 5 centered at the origin, but perhaps not so obvious
is that it lies completely within a circle of radius 5 centered at the origin. Also, note that
a( t) = −r( t). It turns out (see Exercise 16) that whenever an object moves in a circle with
constant speed, the acceleration vector will point towards the center of the circle.

Recall from Section 1.5 that if r1 , r2 are position vectors to distinct points then r1 + t(r2 − r1 )
represents a line through those two points as t varies over all real numbers. That vector
sum can be written as (1 − t)r1 + tr2 . So the function l( t) = (1 − t)r1 + tr2 is a line through
the terminal points of r1 and r2 , and when t is restricted to the interval [0, 1] it is the line
segment between the points, with l(0) = r1 and l(1) = r2 .
In general, a function of the form f( t) = (a 1 t + b 1 , a 2 t + b 2 , a 3 t + b 3 ) represents a line in R3 . A
function of the form f( t) = (a 1 t2 + b 1 t + c 1 , a 2 t2 + b 2 t + c 2 , a 3 t2 + b 3 t + c 3 ) represents a (possibly
degenerate) parabola in R3 .
Example 2.6. Bézier curves are used in Computer Aided Design to approximate the shape of
a polygonal path in space (called the Bézier polygon or control polygon). For instance, given
three points (or position vectors) b0 , b1 , b2 in R3 , define

b10 ( t) = (1 − t)b0 + tb1 ,

b11 ( t) = (1 − t)b1 + tb2 ,
b20 ( t) = (1 − t)b10 ( t) + tb11 ( t)
= (1 − t)2 b0 + 2 t(1 − t)b1 + t2 b2

for all real t. For t in the interval [0, 1], we see that b10 ( t) is the line segment between b0 and
b1 , and b11 ( t) is the line segment between b1 and b2 . The function b20 ( t) is the Bézier curve
62 CHAPTER 2. CURVES

for the points b0 , b1 , b2 . Note from the last formula that the curve is a parabola that goes
through b0 (when t = 0) and b2 (when t = 1).
As an example, let b0 = (0, 0, 0), b1 = (1, 2, 3), and b2 = (4, 5, 2). Then the explicit formula for
the Bézier curve is b20 ( t) = (2 t + 2 t2 , 4 t + t2 , 6 t − 4 t2 ), as shown in Figure 2.1.4, where the line
segments are b10 ( t) and b11 ( t), and the curve is b20 ( t).

(1, 2, 3)

2.5

2
(4, 5, 2)
1.5
z
1 (0, 0, 0)

0
0.5 0.5
1
1.5
2 x
0 0 2.5
1 2 3
3 4 3.5
5 4
y

Figure 2.1.4 Bézier curve approximation for three points.

In general, the polygonal path determined by n ≥ 3 noncollinear points in R3 can be used

to define the Bézier curve recursively by a process called repeated linear interpolation. This
curve will be a vector-valued function whose components are polynomials of degree n − 1,
and its formula is given by de Casteljau’s algorithm.2 In the exercises, the reader will be
given the algorithm for the case of n = 4 points and asked to write the explicit formula for
the Bézier curve for the four points shown in Figure 2.1.5.

Example 2.7. The pedal curve is traced by the orthogonal projection of a fixed point P on
the tangent lines of a given curve f( t).

Write a parametric expression h( t) for the pedal curve for the unit circle f( t) = (cos( t), sin t)
and the point P = (1, 0), so its position vector is i. (This curve is called cardioid.)
2 See pp. 27–30 in FARIN .
 2.1 Vector-Valued Functions 63

Denote by w( t) the projection of v( t) = i − f( t) to the tangent line at

f( t), so h( t) = f( t) + w( t).
The velocity vector f′ ( t) = (− sin t, cos t) is parallel to the tangent line
at f( t).
Note that kf′ ( t)k = 1 for any t. Therefore the vector w( t) can be
found useing the following formula (compare to Example 1.27 and Ex- P
ercise 25, on page 20)

w( t) = (f′ ( t) · v( t)) f′ ( t)
= (f′ ( t) · (i − f( t))) f′ ( t)
= (sin2 t, − sin t cos t).

and

h( t) = f( t) + w( t)
= (cos t + sin2 t, sin t − sin t cos t).

Exercises
A
For Exercises 1–4, calculate f′ ( t) and find the tangent line at f(0).
2
1. f( t) = ( t + 1, t2 + 1, t3 + 1); 2. f( t) = ( e t + 1, e2 t + 1, e t + 1);

3. f( t) = (cos 2 t, sin 2 t, t); 4. f( t) = (sin 2 t, 2 sin2 t, 2 cos t).

For Exercises 5–6, find the velocity v( t) and acceleration a( t) of an object with the given
position vector r( t).

5. r( t) = ( t, t − sin t, 1 − cos t); 6. r( t) = (3 cos t, 2 sin t, 1).

B
³ cos t sin t −at ´
7. Let f( t) = p ,p ,p , with a 6= 0.
1 + a2 t2 1 + a2 t2 1 + a2 t2
(a) Show that k f( t) k = 1 for all t.
(b) Show directly that f′ ( t) · f( t) = 0 for all t.

8. If f′ ( t) = 0 for all t in some interval (a, b), show that f( t) is a constant vector in (a, b).

9. For a constant vector c 6= 0, the function f( t) = tc represents a line parallel to c.

(a) What kind of curve does g( t) = t3 c represent? Explain.
(b) What kind of curve does h( t) = e t c represent? Explain.
64 CHAPTER 2. CURVES

(c) Compare f′ (0) and g′ (0). Given your answer to part (a), how do you explain the differ-
ence in the two derivatives?

10. Show that

d³ df ´ d2f
f× =f × 2.
dt dt dt
11. Let a particle of (constant) mass m have position vector r( t), velocity v( t), acceleration
a( t) and momentum p( t) at time t. The angular momentum L( t) of the particle with
respect to the origin at time t is defined as L( t) = r( t) × p( t). If F( t) is the force acting on
the particle at time t, then define the torque N( t) acting on the particle with respect to
the origin as N( t) = r( t) × F( t). Show that L′ ( t) = N( t).
d df ³ dg ´ ³ dh ´
12. Show that (f · (g × h)) = · (g × h) + f · × h + f· g × .
dt dt dt dt
13. The Mean Value Theorem does not hold for vector-valued functions: Show that for f( t) =
(cos t, sin t, t), there is no t in the interval (0, 2π) such that
f(2π) − f(0)
f′ ( t) = .
2π − 0

14. Wrie a parametric equation for the pedal curve to f( t) = ( t, t2 , t3 ) with respect to the
origin.
C
15. The Bézier curve b30 ( t) for four noncollinear points b0 , b1 , b2 , b3 in R3 is defined by the
following algorithm (going from the left column to the right):

b10 ( t) = (1 − t)b0 + tb1 ,

b11 ( t) = (1 − t)b1 + tb2 , b20 ( t) = (1 − t)b10 ( t) + tb11 ( t),
b12 ( t) = (1 − t)b2 + tb3 . b21 ( t) = (1 − t)b11 ( t) + tb12 ( t), b30 ( t) = (1 − t)b20 ( t) + tb21 ( t).

(a) Show that b30 ( t) = (1 − t)3 b0 + 3 t(1 − t)2 b1 + 3 t2 (1 − t)b2 + t3 b3 .

(b) Write the explicit formula (as in Example 2.6) for the Bézier curve for the points
b0 = (0, 0, 0), b1 = (0, 1, 1), b2 = (2, 3, 0), b3 = (4, 5, 2).

16. Let r( t) be the position vector for a particle moving in R3 , v( t) be its velocity and a( t) be
its acceleration. Show that
d
(r × (v × r)) = k r k2 a + (r · v)v − (k v k2 + r · a)r.
dt

17. Let r( t) be the position vector in R3 for a particle that moves with constant speed c > 0 in
a circle of radius a > 0 centered at the origin in the x y-plane. Show that its acceleration
a( t) points in the opposite direction as r( t) for all t. (Hint: Use Example 2.3 to show that
r( t) ⊥ v( t) and a( t) ⊥ v( t), and hence a( t) ∥ r( t).)
 2.1 Vector-Valued Functions 65

18. Prove Theorem 2.1(g).

19. Show that there is no plane which is tangent3 to the curve f( t) = ( t, t2 , t3 ) at two distinct
points.

2
(4, 5, 2)
(0, 1, 1)
1.5

1
z
(0, 0, 0)
0.5
0
0.5
(2, 3, 0) 1
1.5
2 x
0 0 2.5
1 2 3
3 4 3.5
5 4
y

Figure 2.1.5 Bézier curve approximation for four points.

3 A plane is called tangent to a curve f( t) at point f( t ) it it contains the tangent line at f( t ).

0 0
66 CHAPTER 2. CURVES

2.2 Arc Length

Definition 2.4. Let f( t) = ( x( t), y( t), z( t)) be a curve in R3 whose domain includes the interval
[a, b]. Suppose that in the interval (a, b) the first derivative of each component function x( t),
y( t) and z( t) exists and is continuous. Then the arc length L of the curve from t = a to t = b
is
wb wb q
′
L = k f ( t) k dt = x ′ ( t)2 + y ′ ( t)2 + z ′ ( t)2 dt. (2.1)
a a

If f( t) = ( x( t), y( t), z( t)) is the position vector of an object moving in R3 then its speed at time
t is k f′ ( t) k, that is the magnitude of the velocity vector. Therefore it seems natural to define
the distance s traveled by as the definite integral of its speed in the time interval (2.1).
Example 2.8. Find the length L of the helix f( t) = (cos t, sin t, t) from t = 0 to t = 2π.
Solution: By formula (2.1), we have

w2πp w2πp w2πp

2
L= (− sin t)2 + (cos t)2 + 12 dt = sin t + cos2 t + 1 dt = 2 dt
0 0 0
p p
= 2(2π − 0) = 2 2π.

Notice that the set traced out by the curve f( t) = (cos t, sin t, t) from Example 2.8 is also
traced out by the function g( t) = (cos 2 t, sin 2 t, 2 t). For example, over the interval [0, π], g( t)
traces out the same section of the curve as f( t) does over the interval [0, 2π]. Intuitively,
this says that g( t) traces the curve twice as fast as f( t). This makes sense since, viewing the
functions as position p vectors and their
p derivatives as velocity vectors, the speeds of f( t) and
′ ′
g( t) are k f ( t) k = 2 and k g ( t) k = 2 2, respectively. We say that g( t) is a reparametrization
of curve f( t).

Definition 2.5. Let f( t) be a smooth curve in R3 defined on an interval [a, b], and let
α : [ c, d ] → [a, b] be a smooth one-to-one mapping of an interval [ c, d ] onto [a, b]. Then the
function g : [ c, d ] → R3 defined by g( s) = f(α( s)) is a reparametrization of f( t) with param-
eter s. If the derivative of α does not vanish, we say that the reparametrization is regular
and g( s) is equivalent to f( t).

s t f( t)
α f
[ c, d ] [a, b] R3

g( s) = f(α( s)) = f( t)

Note that the differentiability of g( s) follows from a version of the Chain Rule for vector-
valued functions (the proof is left as an exercise):
 2.2 Arc Length 67

Theorem 2.2. Chain Rule: If f( t) is a differentiable vector-valued function of t, and t = α( s)

is a differentiable scalar function of s, then g( s) = f(α( s)) is a differentiable vector-valued
function of s, and

d g d f dt
= or equivalently g′ ( s) = f′ (α( s)) α′ ( s) (2.2)
ds dt ds
for any s where the composite function f(α( s)) is defined.

Example 2.9. The following are all regular reparametrizations of one curve:
f( t) = (cos t, sin t, t) for t in [0, 2π],
g( s) = (cos 2 s, sin 2 s, 2 s) for s in [0, π],
h( s) = (cos 2π s, sin 2π s, 2π s) for s in [0, 1].
To see that g( s) is regular reparametrization of f( t), define α : [0, π] → [0, 2π] by α( s) = 2 s.
Then α is smooth, one-to-one, maps [0, π] onto [0, 2π], and is strictly increasing (since α ′ ( s) =
2 > 0 for all s). Likewise, defining α : [0, 1] → [0, 2π] by α( s) = 2π s shows that h( s) is regular
reparametrization of f( t).

A curve can be reparametrized, with different speeds, so which one is the best to use? In
some situations the arc length parametrization can be useful. The idea behind this is to
replace the parameter t, for any given smooth parametrization f( t) defined on [a, b], by the
parameter s given by
wt
s = s( t) = k f′ ( u) k du. (2.3)
a
In terms of motion along a curve, s is the distance traveled along the curve after time t
has elapsed. So the new parameter will be distance instead of time. There is a natural
correspondence between s and t: from a starting point on the curve, the distance traveled
along the curve (in one direction) is uniquely determined by the amount of time elapsed, and
vice versa.
Since s is the arc length of the curve over the interval [a, t] for each t in [a, b], then it is a
function of t. By the Fundamental Theorem of Calculus, its derivative is

d w ′
t
′ ds
s ( t) = = k f ( u) k du = k f′ ( t) k for all t in [a, b].
dt dt a

Since f( t) is smooth, then k f′ ( t) k > 0 for all t in [a, b]. Thus s ′ ( t) > 0 and hence s( t) is strictly
increasing on the interval [a, b]. Recall that this means that s is a one-to-one mapping of the
interval [a, b] onto the interval [ s(a), s( b)]. But we see that
wa wb
s( a) = k f′ ( u) k du = 0 and s( b ) = k f′ ( u) k du = L = arc length from t = a to t = b.
a a
68 CHAPTER 2. CURVES

So the function s : [a, b] → [0, L] is a one-to-one, differentiable α( s )

mapping onto the interval [0, L]. From single-variable calculus, s t
we know that this means that there exists an inverse function [0, L] [a, b]
α : [0, L] → [a, b] that is differentiable and the inverse of s : [a, b] →
s( t)
[0, L]. That is, for each t in [a, b] there is a unique s in [0, L] such
that s = s( t) and t = α( s). And we know that the derivative of α is Figure 2.2.1 t = α(s)

1 1
α ′ ( s) = = .
s ′ (α( s)) k f′ (α( s)) k

So define the arc length parametrization g : [0, L] → R3 by

g( s) = f(α( s)) for all s in [0, L].

Then g( s) is smooth, by the Chain Rule. In fact, g( s) has unit speed:

g′ ( s) = f′ (α( s)) α ′ ( s) by the Chain Rule, so

1
= f′ (α( s)) ′ , so
k f (α( s)) k
k g′ ( s) k = 1 for all s in [0, L].

So the arc length parametrization traverses the curve at a “normal” rate.

Inr practice, parametrizing a curve f( t) by arc length requires you to evaluate the integral
t
s = a k f′ ( u) k du explicitly as a function of t, so that you could then solve for t in terms of s.
If that can be done, you would then substitute the expression for t in terms of s (which we
called α( s)) into the formula for f( t) to get g( s) = f(α( s)).

Example 2.10. Parametrize the helix f( t) = (cos t, sin t, t), for t in [0, 2π], by arc length.
Solution: By Example 2.8 and formula (2.3), we have

wt wt p p
′
s= k f ( u) k du = 2 du = 2 t for all t in [0, 2π].
0 0

s
So we can solve for t in terms of s: t = α( s) = p .
2
³ s s s ´ p
∴ g(s) = cos p , sin p , p for all s in [0, 2 2π]. Note that k g′ (s) k = 1.
2 2 2

Exercises
A
For Exercises 1–3, calculate the arc length of f( t) over the given interval.

1. f( t) = (3 cos 2 t, 3 sin 2 t, 3 t) on [0, π/2];

 2.2 Arc Length 69

p
2. f( t) = (( t2 + 1) cos t, ( t2 + 1) sin t, 2 2 t) on [0, 1];

3. f( t) = (2 cos 3 t, 2 sin 3 t, 2 t3/2 ) on [0, 1].

4. Parametrize the curve from Exercise 1 by arc length.

5. Parametrize the curve from Exercise 3 by arc length.

B
6. Assume that g( s) is a regular reparametrization of f( t). Show that both curves have the
same length.

7. Let f( t) be a differentiable curve such that f( t) 6= 0 for all t. Show that

µ ¶
d f( t) f( t) × (f′ ( t) × f( t))
° ° = .
dt ° f( t) ° k f( t) k3

8. Show that the arc length L of a curve whose spherical coordinates are ρ = ρ ( t), θ = θ ( t)
and φ = φ( t) for t in an interval [a, b] is

wb q
L= ρ ′ ( t)2 + (ρ ( t)2 sin2 φ( t)) θ ′ ( t)2 + ρ ( t)2 φ ′ ( t)2 dt.
a

(Hint: Convert the data in Cartesian coordinates.)

9. Let f( t) be a smooth curve. The pedal curve of f( t) is traced by the orthogonal projections
of the origin on the tangent lines to f. Write a parametric equation for the pedal curve
h( t) for the given smooth curve f( t).

C
10. Assume that the trajectory of the back wheel of an ideal bicycle is given by smooth plane
curve b( t), here t denotes time. We assume that in the ideal bicycle the distance from back
wheel and front wheel is fixed, let us denote it by R and the back wheel always moves in
the direction to the front wheel.

(a) Write an expression for the trajectory of the front wheel f( t).
(b) Show that the speed of the back wheel can not exceed the speed of the front wheel.
70 CHAPTER 2. CURVES

2.3 Curvature
In the field of mathematics known as differential geometry4 special attention is given to the
parametrization-independent constructions. For example, depending on the parametriza-
tion, the velocity vector of the curve at given point can be multiplied by a scalar, so it
is not parametrization-independent; on the other hand the tangent line at given point is
parametrization-independent — although it is defined using parametrization the resulting
line is the same.
An other example is so called osculating plane. Given a smooth regular curve f, its oscu-
lating plane at f( t) is the plane passing thru f( t) and containing the velocity vector f′ ( t) and
the acceleration f′′ ( t). The osculating plane is defined if f′ ( t) is not parallel to f′′ ( t). Note that
in this case the cross product f′ ( t) × f′′ ( t) is perpendicular to the osculating plane. Therefore
the equation of the osculating plane at f( t) can be written as

(x − f( t)) · (f′ ( t) × f′′ ( t)) = 0

with the unknown x.

Example 2.11. Let us show that osculating plane does at given point does not depend on the
parametrization. That is, if g( s) = f(α( s)) is a regular reparametrization then the plane thru
g( s) and containing the velocity vector g′ ( s) and the acceleration g′′ ( s) is the same as the
plane thru f( t) and containing the velocity vector f′ ( t) and the acceleration f′′ ( t) for t = α( s).
Since f( t) = g( s), we only need to show that f′ ( t) × f′′ ( t) ∥ g′ ( s) × g′′ ( s).
By chain rule
g′ ( s) = f′ (α( s))α′ ( s)

and by chain rule again

g′′ ( s) = f′′ (α( s))α′ ( s)2 + f′ (α( s))α′′ ( s).

Since f′ × f′ = 0 and t = α( s) we get

g′ ( s) × g′′ ( s) = f′ ( t)α′ ( s) × (f′′ ( t)α′ ( s)2 + f′ ( t)α′′ ( s))

= α′ ( s)3 f′ ( t) × f′′ ( t).

Since the reparametrization is regular, α′ ( s) 6= 0. Therefore f′ ( t) × f′′ ( t) ∥ g′ ( s) × g′′ ( s) as re-

qured.

Yet an other example is so called curvature. Assume a smooth regular curve g has arc
length parametrization. Note that if g parametrize a straight line then g′ ( s) is a constant
unit vector and therefore g′′ ( s) = 0 at all points. Therefore the value κ( s) = k g′′ ( s) k can be
used to measure how fast the curve deviates from the straight line. The value κ( s) and the
vector g′′ ( s) are called curvature and curvature vector of the curve g at the point g( s).
4 See O’N EILL for an introduction to elementary differential geometry.
 2.3 Curvature 71

If κ( s) 6= 0 then the value R ( s) = κ(1s) is called curvature radius of g at the point g( s). It
is called this way since the best approximation of the curve g at the point g( s) by a circle,
so called osculating circle, has radius R ( s). This circle is lying in the osculating plane, its
center lies in the direction of curvature vector g′′ ( s) from g( s) on the distance R ( s). If κ( s) = 0
then the osculating circle degenerates to the tangent line.

The osculating circle to the sinusoid at two points.

Assume you want to find the curvature of the given curve using the definition above. Then
you first have to find the arc length parametrization and then apply the formula above at the
given point. Finding this parametrization often leads to an integral that is either difficult or
impossible to evaluate explicitly. The simple integral in Example 2.10 is the exception, not
the norm. In general, arc length parametrizations are more useful for theoretical purposes
than for practical computations.5
The following theorem provides a direct way to calculate the curvature, without passing
to the reparametrization. Exercises 9 guides you through similar calculations.

Theorem 2.3. The curvature κ of a smooth curve f at the point f( t) can be found using the
following formula:

k f′′ ( t) × f′ ( t) k
κ= . (2.4)
k f′ ( t) k3

Proof: Let g( s) be the arc length parametrization of f( t); in particular k g′ ( s) k = 1 for any s.
As above, we assume t = α( s) and therefore g( s) = f(α( s)) and α′ ( s) = k f′1( t) k .

5 For example, the usual parametrizations of Bézier curves, which we discussed in Section 1.8, are polynomial

functions in R3 . This makes their computation relatively simple, which, in Computer-aided design, is desir-
able. But their arc length parametrizations are not only not polynomials, they are in fact usually impossible to
calculate at all.
72 CHAPTER 2. CURVES

Applying Chain and Product Rules, we get

g′′ ( s) = f (α( s))′′

¡ ¢′
= f′ (α( s))α′ ( s)
= f′′ (α( s))(α′ ( s))2 + f′ (α( s))α′′ ( s)
f′′ ( t)
= + f′ ( t)α′′ ( s),
k f′ ( t) k2

Since f′ ( t) × f′ ( t) = 0, we get
µ ′′ ¶
f′′ ( t) × f′ ( t) f ( t) ′ ′′ f′ ( t)
= + f ( t )α ( s ) ×
k f′ ( t) k3 k f′ ( t) k2 k f′ ( t) k
′′ ′
= g ( s) × g ( s).

Since k g′ ( s) k = 1, we have

0 = (g′ ( s) · g′ ( s))′ = 2 g′′ ( s) · g′ ( s).

That is, g′′ ( s) ⊥ g′ ( s) for any s. Since k g′ ( s) k = 1, we can continue

k f′′ ( t) × f′ ( t) k
= k g′′ ( s) × g′ ( s) k
k f′ ( t) k3
= k g′′ ( s) k k g′ ( s) k
= κ.

QED

Exercises
A
For Exercises 1–4, find the tangent line, the osculating plane and the curvature at each point
of the curve f( t).

1. f( t) = (cos t, sin t, t); 2. f( t) = ( t, t2 , t3 );

3. f( t) = ( t sin t, t cos t); 4. f( t) = ( e t sin t, e t cos t).

B
5. Let f( t) be a smooth regular curve and g( s) = f(α( s)) be its regular reparametrization.
Show that the osculating plane of f at f( t) coinsides with the osculating plane of g at g( s)
if t = α( s).
 2.3 Curvature 73

6. Let f( t) be a smooth regular curve; in particular, f′ ( t) 6= 0 for all t. Then we can define the
unit tangent vector T by
f′ ( t)
T( t) = ′ .
k f ( t) k
(a) Show that
f′ ( t) × (f′′ ( t) × f′ ( t))
T′ ( t ) = .
k f′ ( t) k3

(b) Use this formula to get an other proof of Theorem 2.3.

7. Let g( s) be a smooth curve with arc length parametrization and κ( s) be its curvature.
Show that
g′′′ ( s) · g′ ( s) = −κ( s)2 .

8. Let g be a smooth plane curve with arc length parametrization. The curve

h( s) = g( s) − sg′ ( s)

is called involute of g( s).

(a) Show that
k h′ ( s) k = s κ( s)
where κ( s) is curvature of g at g( s)
1
(b) Show that the curvature of h at h( s) equals to s for s > 0. (Hint: Use Exercise 7.)

C
9. Let f( t) be a smooth curve in the plane. Assume its curvature κ( t) is increasing in t. Show
that the curve has no self-intersections; that is, if t 0 6= t 1 then f( t 0 ) 6= f( t 1 ). (Hint: Write an
expression for the center and radius of the osculating circles and use it to show that they
do not intersect each other.
3 Functions of Several Variables
3.1 Functions of Two or Three Variables
In Section 1.8 we discussed vector-valued functions of a single real variable. We will now
examine real-valued functions of a point (or vector) in R2 or R3 . For the most part these
functions will be defined on sets of points in R2 , but there will be times when we will use
points in R3 , and there will also be times when it will be convenient to think of the points as
vectors (or terminal points of vectors).
A real-valued function f defined on a subset D of R2 is a rule that assigns to each point
( x, y) in D a real number f ( x, y). The largest possible set D in R2 on which f is defined is
called the domain of f , and the range of f is the set of all real numbers f ( x, y) as ( x, y)
varies over the domain D . A similar definition holds for functions f ( x, y, z) defined on points
( x, y, z) in R3 .

Example 3.1. The domain of the function

f ( x, y) = x y

is all of R2 , and the range of f is all of R.

Example 3.2. The domain of the function

1
f ( x, y) =
x− y

is all of R2 except the points ( x, y) for which x = y. That is, the domain is the set D = {( x, y) :
x 6= y}. The range of f is all real numbers except 0.

Example 3.3. The domain of the function

q
f ( x, y) = 1 − x 2 − y2

is the set D = {( x, y) : x2 + y2 ≤ 1}, since the quantity inside the square root is nonnegative if
and only if 1 − ( x2 + y2 ) ≥ 0. We see that D consists of all points on and inside the unit circle
in R2 (D is sometimes called the closed unit disk). The range of f is the interval [0, 1] in R.

74
 3.1 Functions of Two or Three Variables 75

Example 3.4. The domain of the function

f ( x, y, z) = e x+ y− z

is all of R3 , and the range of f is all positive real numbers.

A function f ( x, y) defined in R2 is often written as z = f ( x, y), as was mentioned in Section

1.1, so that the graph of f ( x, y) is the set {( x, y, z) : z = f ( x, y)} in R3 . So we see that this
graph is a surface in R3 , since it satisfies an equation of the form F ( x, y, z) = 0 (namely,
F ( x, y, z) = f ( x, y) − z). The traces of this surface in the planes z = c, where c varies over R,
are called the level curves of the function. Equivalently, the level curves are the solution
sets of the equations f ( x, y) = c, for c in R. Level curves are often projected onto the x y-plane
to give an idea of the various “elevation” levels of the surface (as is done in topography).

Example 3.5. The graph of the function

p
sin x2 + y2
f ( x, y) = p
x 2 + y2

is shown below. Note that the level curves (shown both on the surface and projected onto the
x y-plane) are groups of concentric circles.

You may be wondering what happens to the function in Example 3.5 at the point ( x, y) =
(0, 0), since both the numerator and denominator are 0 at that point. The function is not
defined at (0, 0), but the limit of the function exists (and equals 1) as ( x, y) approaches (0, 0).
We will now state explicitly what is meant by the limit of a function of two variables.

Definition 3.1. Let (a, b) be a point in R2 , and let f ( x, y) be a real-valued function defined
on some set containing (a, b) (but not necessarily defined at (a, b) itself). Then we say that
the limit of f ( x, y) equals L as ( x, y) approaches (a, b), written as

lim f ( x, y) = L , (3.1)
( x,y)→(a,b)

if given any ǫ > 0, there exists a δ > 0 such that

q
| f ( x, y) − L | < ǫ whenever 0 < ( x − a)2 + ( y − b)2 < δ.

A similar definition can be made for functions of three variables. The idea behind the
above definition is that the values of f ( x, y) can get arbitrarily close to L (that is, within ǫ
of L) if we pick ( x, y) sufficiently close to (a, b) (that is, inside a circle centered at (a, b) with
some sufficiently small radius δ).
If you recall the “epsilon-delta” proofs of limits of real-valued functions of a single variable,
you may remember how awkward they can be, and how they can usually only be done easily
76 CHAPTER 3. FUNCTIONS OF SEVERAL VARIABLES

1
0.8
0.6
0.4
0.2
z
0
-0.2 -10
-5
-0.4
-10 0
-5 x
0 5
y 5
10 10

p
sin x2 + y2
Figure 3.1.1 The function f (x, y) = p 2 2 .
x +y

for simple functions. In general, the multivariable cases are at least equally awkward to go
through, so we will not bother with such proofs. Instead, we will simply state that when the
function f ( x, y) is given by a single formula and is defined at the point (a, b) (for example,
is not some indeterminate form like 0/0) then you can just substitute ( x, y) = (a, b) into the
formula for f ( x, y) to find the limit.

Example 3.6.
xy (1)(2) 2
lim = =
( x,y)→(1,2) x2 + y2 2
1 +2 2 5
xy
since f ( x, y) = x2 + y2
is properly defined at the point (1, 2).

The major difference between limits in one variable and limits in two or more variables
has to do with how a point is approached. In the single-variable case, the statement “ x → a”
means that x gets closer to the value a from two possible directions along the real number
line (see Figure 3.1.2(a)). In two dimensions, however, ( x, y) can approach a point (a, b) along
an infinite number of paths (see Figure 3.1.2(b)).

Example 3.7.
xy
lim does not exist
( x,y)→(0,0) x2 + y2
 3.1 Functions of Two or Three Variables 77

x x (a, b)
x
0 a x 0
(a) x → a in R (b) ( x, y) → (a, b) in R2

Figure 3.1.2 “Approaching” a point in different dimensions.

Note that we can not simply substitute ( x, y) = (0, 0) into the function, since doing so gives an
indeterminate form 0/0. To show that the limit does not exist, we will show that the function
approaches different values as ( x, y) approaches (0, 0) along different paths in R2 . To see this,
suppose that ( x, y) → (0, 0) along the positive x-axis, so that y = 0 along that path. Then

xy x0
f ( x, y) = = =0
x 2 + y 2 x 2 + 02

along that path (since x > 0 in the denominator). But if ( x, y) → (0, 0) along the straight line
y = x through the origin, for x > 0, then we see that

xy x2 1
f ( x, y) = 2 2
= 2 2
= ,
x +y x +x 2

which means that f ( x, y) approaches different values as ( x, y) → (0, 0) along different paths.
Hence the limit does not exist.

Limits of real-valued multivariable functions obey the same algebraic rules as in the
single-variable case, as shown in the following theorem, which we state without proof.

Theorem 3.1. Suppose that lim f ( x, y) and lim g( x, y) both exist, and that k is
( x,y)→(a,b) ( x,y)→(a,b)
some scalar. Then:
h i h i
(a) lim [ f ( x, y) ± g( x, y)] = lim f ( x, y) ± lim g( x, y) ;
( x,y)→(a,b) ( x,y)→(a,b) ( x,y)→(a,b)
h i
(b) lim k f ( x, y) = k lim f ( x, y) ;
( x,y)→(a,b) ( x,y)→(a,b)
h ih i
(c) lim [ f ( x, y) g( x, y)] = lim f ( x, y) lim g( x, y) ;
( x,y)→(a,b) ( x,y)→(a,b) ( x,y)→(a,b)

lim f ( x, y)
f ( x, y) ( x,y)→(a,b)
(d) lim = if lim g( x, y) 6= 0;
( x,y)→(a,b) g( x, y) lim g( x, y) ( x,y)→(a,b)
( x,y)→(a,b)

(e) If | f ( x, y) − L | ≤ g( x, y) for all ( x, y) and if lim g( x, y) = 0, then lim f ( x, y) = L.

( x,y)→(a,b) ( x,y)→(a,b)
78 CHAPTER 3. FUNCTIONS OF SEVERAL VARIABLES

Note that in part (e), it suffices to have | f ( x, y) − L | ≤ g( x, y) for all ( x, y) “sufficiently close”
to (a, b) (but excluding (a, b) itself).
Example 3.8. Show that
y4
lim = 0.
( x,y)→(0,0) x 2 + y2
Since substituting ( x, y) = (0, 0) into the function gives the indeterminate form 0/0, we need
an alternate method for evaluating this limit. We will use Theorem 3.1(e). First, notice that
¡p ¢4 ¡p ¢4 ¡p ¢4
y4 = y2 and so 0 ≤ y4 ≤ x2 + y2 for all ( x, y). But x2 + y2 = ( x2 + y2 )2 . Thus, for
all ( x, y) 6= (0, 0) we have
¯ ¯
¯ y4 ¯ ( x2 + y2 )2 2 2
¯ x2 + y2 ¯ ≤ x2 + y2 = x + y → 0 as ( x, y) → (0, 0).
¯ ¯

y4
Therefore, lim = 0.
( x,y)→(0,0) x2 + y2

Continuity can be defined similarly as in the single-variable case.

Definition 3.2. A real-valued function f ( x, y) with domain D in R2 is continuous at the

point (a, b) in D if lim f ( x, y) = f (a, b). We say that f ( x, y) is a continuous function if
( x,y)→(a,b)
it is continuous at every point in its domain D .

Unless indicated otherwise, you can assume that all the functions we deal with are con-
tinuous. In fact, we can modify the function from Example 3.8 so that it is continuous on all
of R2 .

Example 3.9. Define a function f ( x, y) on all of R2 as follows:


0
 if ( x, y) = (0, 0)
f ( x, y) = 4
y

 2 if ( x, y) 6= (0, 0)
x + y2
Then f ( x, y) is well-defined for all ( x, y) in R2 (that is, there are no indeterminate forms for
any ( x, y)), and we see that
b4
lim f ( x, y) = = f (a, b) for (a, b) 6= (0, 0).
( x,y)→(a,b) a2 + b 2
So since
lim f ( x, y) = 0 = f (0, 0) by Example 3.8,
( x,y)→(0,0)

then f ( x, y) is continuous on all of R2 .

Exercises
A
For Exercises 1–6, state the domain and range of the given function.
 3.1 Functions of Two or Three Variables 79

1
1. f ( x, y) = x2 + y2 − 1; 2. f ( x, y) = ;
x 2 + y2
p x2 + 1
3. f ( x, y) = x2 + y2 − 4; 4. f ( x, y) = ;
y
p
5. f ( x, y, z) = sin( x yz); 6. f ( x, y, z) = ( x − 1)( yz − 1).

For Exercises 7–18, evaluate the given limit.

7. lim cos( x y); 8. lim ex y;

( x,y)→(0,0) ( x,y)→(0,0)

x 2 − y2 x y2
9. lim ; 10. lim ;
( x,y)→(0,0) x 2 + y2 ( x,y)→(0,0) x2 + y4

x 2 − 2 x y + y2 x y2
11. lim ; 12. lim ;
( x,y)→(1,−1) x− y ( x,y)→(0,0) x 2 + y2
2 2
x −y x 2 − 2 x y + y2
13. lim ; 14. lim ;
( x,y)→(1,1) x− y ( x,y)→(0,0) x− y
µ ¶
y4 sin( x y) 1
15. lim ; 16. lim ( x2 + y2 ) cos ;
( x,y)→(0,0) x 2 + y2 ( x,y)→(0,0) xy
x µ ¶
17. lim ; 1
( x,y)→(0,0) y 18. lim cos .
( x,y)→(0,0) xy

B
1 2 2 2
−( x + y )/2σ
19. Show that f ( x, y) = 2πσ 2 e , for σ > 0, is constant on the circle of radius r > 0
centered at the origin. This function is called a Gaussian blur, and is used as a filter in
image processing software to produce a “blurred” effect.

20. Suppose that f ( x, y) ≤ f ( y, x) for all ( x, y) in R2 . Show that f ( x, y) = f ( y, x) for all ( x, y) in

R2 .
p
21. Use the substitution r = x2 + y2 to show that
p
sin x2 + y2
lim p =1 .
( x,y)→(0,0) x 2 + y2

(Hint: You will need to use L’Hôpital’s Rule for single-variable limits.)

C
22. Prove Theorem 3.1(a) in the case of addition. (Hint: Use Definition 3.1.)

23. Prove Theorem 3.1(b).

80 CHAPTER 3. FUNCTIONS OF SEVERAL VARIABLES

3.2 Partial Derivatives

Now that we have an idea of what functions of several variables are, and what a limit of
such a function is, we can start to develop an idea of a derivative of a function of two or more
variables. We will start with the notion of a partial derivative.

Definition 3.3. Let f ( x, y) be a real-valued function with domain D in R2 , and let (a, b) be
a point in D . Then the partial derivative of f at (a, b) with respect to x, denoted by
∂f
(a, b), is defined as
∂x
∂f f (a + h, b) − f (a, b)
(a, b) = lim (3.2)
∂x h→0 h
∂f
and the partial derivative of f at (a, b) with respect to y, denoted by (a, b), is defined
∂y
as
∂f f (a, b + h) − f (a, b)
(a, b) = lim . (3.3)
∂y h→0 h
Note: The symbol ∂ is pronounced “del”.1

Recall that the derivative of a function f ( x) can be interpreted as the rate of change of
that function in the (positive) x direction. From the definitions above, we can see that the
partial derivative of a function f ( x, y) with respect to x or y is the rate of change of f ( x, y) in
the (positive) x or y direction, respectively. What this means is that the partial derivative of
a function f ( x, y) with respect to x can be calculated by treating the y variable as a constant,
and then simply differentiating f ( x, y) as if it were a function of x alone, using the usual
rules from single-variable calculus. Likewise, the partial derivative of f ( x, y) with respect to
y is obtained by treating the x variable as a constant and then differentiating f ( x, y) as if it
were a function of y alone.

∂f ∂f
Example 3.10. Find ( x, y) and ( x, y) for the function f ( x, y) = x2 y + y3 .
∂x ∂y
Solution: Treating y as a constant and differentiating f ( x, y) with respect to x gives
∂f
( x, y) = 2 x y
∂x
and treating x as a constant and differentiating f ( x, y) with respect to y gives
∂f
( x, y) = x2 + 3 y2 .
∂y
∂f ∂f ∂f ∂f
We will often simply write and instead of ( x, y) and ( x, y).
∂x ∂y ∂x ∂y

1 It is not a Greek letter. The symbol was first used by the mathematicians A. Clairaut and L. Euler around

1740, to distinguish it from the letter d used for the “usual” derivative.
 3.2 Partial Derivatives 81

∂f ∂f sin( x y2 )
Example 3.11. Find and for the function f ( x, y) = .
∂x ∂y x2 + 1
Solution: Treating y as a constant and differentiating f ( x, y) with respect to x gives

∂f ( x2 + 1)( y2 cos( x y2 )) − (2 x) sin( x y2 )

=
∂x ( x2 + 1)2

and treating x as a constant and differentiating f ( x, y) with respect to y gives

∂f 2 x y cos( x y2 )
= .
∂y x2 + 1

∂f ∂f
Since both and are themselves functions of x and y, we can take their partial
∂x ∂y
derivatives with respect to x and y. This yields the higher-order partial derivatives:

∂2 f ∂ ³∂f ´ ∂2 f ∂ ³∂f ´
= , = ,
∂ x2 ∂ x ∂ x ∂ y2 ∂ y ∂ y
∂2 f ∂ ³∂f ´ ∂2 f ∂ ³∂f ´
= , = ,
∂ y ∂x ∂ y ∂x ∂x ∂ y ∂x ∂ y
∂3 f ∂ ³ ∂2 f ´ ∂3 f ∂ ³ ∂2 f ´
= , = ,
∂ x3 ∂ x ∂ x2 ∂ y3 ∂ y ∂ y2
∂3 f ∂ ³ ∂2 f ´ ∂3 f ∂ ³ ∂2 f ´
= , = ,
∂ y ∂ x2 ∂ y ∂ x2 ∂ x ∂ y2 ∂ x ∂ y2
∂3 f ∂ ³ ∂2 f ´ ∂3 f ∂ ³ ∂2 f ´
= , = ,
∂ y2 ∂ x ∂ y ∂ y ∂ x ∂ x2 ∂ y ∂ x ∂ x ∂ y
∂3 f ∂ ³ ∂2 f ´ ∂3 f ∂ ³ ∂2 f ´
= , = ,
∂x ∂ y ∂x ∂x ∂ y ∂x ∂ y ∂x ∂ y ∂ y ∂x ∂ y
..
.

∂f ∂f ∂2 f ∂2 f ∂2 f ∂2 f
Example 3.12. Find the partial derivatives , , 2
, 2
, and for the
∂x ∂y ∂x ∂y ∂ y ∂x ∂x ∂ y
2
function f ( x, y) = e x y + x y3 .
82 CHAPTER 3. FUNCTIONS OF SEVERAL VARIABLES

Solution: Proceeding as before, we have

∂f 2 ∂f 2
= 2 x ye x y
+ y3 , = x2 e x y
+ 3 x y2 ,
∂x ∂y
∂2 f ∂ 2 ∂2 f ∂ 2
= (2 x ye x y + y3 ) = ( x 2 e x y + 3 x y2 )
∂ x2 ∂x ∂ y2 ∂y
2 2 2
= 2 ye x y
+ 4 x 2 y2 e x y , = x4 e x y
+ 6 x y,

∂2 f ∂ 2 ∂2 f ∂ 2
= (2 x ye x y + y3 ) = ( x 2 e x y + 3 x y2 )
∂ y ∂x ∂y ∂x ∂ y ∂x
2 2 2 2
= 2 xe x y
+ 2 x3 ye x y
+ 3 y2 , = 2 xe x y
+ 2 x3 ye x y
+ 3 y2 .

Higher-order partial derivatives that are taken with respect to different variables, such
∂2 f ∂2 f
as ∂ y ∂ x and ∂ x ∂ y , are called mixed partial derivatives. Notice in the above example that
∂2 f ∂2 f ∂2 f
∂ y ∂x
= ∂x ∂ y
. It turns that this will usually be the case. Specifically, whenever both ∂ y ∂x
and
∂2 f 2
are continuous at a point (a, b), then they are equal at that point. All the functions
∂x ∂ y
we will deal with will have continuous partial derivatives of all orders, so you can assume in
the remainder of the text that

∂2 f ∂2 f
= for all ( x, y) in the domain of f .
∂ y ∂x ∂x ∂ y

In other words, it doesn’t matter in which order you take partial derivatives. This applies
even to mixed partial derivatives of order 3 or higher.
The notation for partial derivatives varies. All of the following are equivalent:

∂f
: f x ( x, y) , f 1 ( x, y) , D x ( x, y) , D 1 ( x, y) ;
∂x
∂f
: f y ( x, y) , f 2 ( x, y) , D y ( x, y) , D 2 ( x, y) ;
∂y
∂2 f
: f xx ( x, y) , f 11 ( x, y) , D xx ( x, y) , D 11 ( x, y) ;
∂ x2
∂2 f
: f yy ( x, y) , f 22 ( x, y) , D yy ( x, y) , D 22 ( x, y) ;
∂ y2
∂2 f
: f x y ( x, y) , f 12 ( x, y) , D x y ( x, y) , D 12 ( x, y) ;
∂ y ∂x
∂2 f
: f yx ( x, y) , f 21 ( x, y) , D yx ( x, y) , D 21 ( x, y) .
∂x ∂ y
2 See pp. 214–216 in T AYLOR and M ANN for a proof.
 3.2 Partial Derivatives 83

Exercises
A
∂f ∂f
For Exercises 1–16, find ∂x
and ∂y
.

1. f ( x, y) = x2 + y2 ; 2. f ( x, y) = cos( x + y);
p
3. f ( x, y) = x2 + y + 4; x+1
4. f ( x, y) = ;
y+1
5. f ( x, y) = e x y + x y; 6. f ( x, y) = x2 − y2 + 6 x y + 4 x − 8 y + 2;

7. f ( x, y) = x4 ; 8. f ( x, y) = x + 2 y;
p
9. f ( x, y) = x2 + y2 ; 10. f ( x, y) = sin( x + y);
p
11. f ( x, y) = 3 x2 + y + 4; xy+1
12. f ( x, y) = ;
x+ y
2
+ y2 )
13. f ( x, y) = e−( x ; 14. f ( x, y) = ln( x y);

15. f ( x, y) = sin( x y); 16. f ( x, y) = tan( x + y).

∂2 f ∂2 f ∂2 f
For Exercises 17–26, find ,
∂ x2 ∂ y2
and ∂ y ∂x
(use Exercises 1–8, 14, 15).

17. f ( x, y) = x2 + y2 ; 18. f ( x, y) = cos( x + y);

p
19. f ( x, y) = x2 + y + 4; x+1
20. f ( x, y) = ;
y+1
21. f ( x, y) = e x y + x y; 22. f ( x, y) = x2 − y2 + 6 x y + 4 x − 8 y + 2;

23. f ( x, y) = x4 ; 24. f ( x, y) = x + 2 y;

25. f ( x, y) = ln( x y); 26. f ( x, y) = sin( x y).

B
27. Show that the function f ( x, y) = sin( x + y) + cos( x − y) satisfies the wave equation

∂2 f ∂2 f
− =0 .
∂ x2 ∂ y2

The wave equation is an example of a partial differential equation.

28. Let u and v be twice-differentiable functions of a single variable, and let c 6= 0 be a con-
stant. Show that f ( x, y) = u( x + c y) + v( x − c y) is a solution of the general one-dimensional
wave equation3
∂2 f 1 ∂2 f
− =0 .
∂ x 2 c 2 ∂ y2

3 Conversely, it turns out that any solution must be of this form. See Ch. 1 in W EINBERGER.
84 CHAPTER 3. FUNCTIONS OF SEVERAL VARIABLES

3.3 Tangent Plane to a Surface

∂f ∂f
In the previous section we mentioned that the partial derivatives ∂ x and ∂ y can be thought
of as the rate of change of a function z = f ( x, y) in the positive x and y directions, respectively.
dy
Recall that the derivative dx of a function y = f ( x) has a geometric meaning, namely as the
slope of the tangent line to the graph of f at the point ( x, f ( x)) in R2 . There is a similar
∂f ∂f
geometric meaning to the partial derivatives ∂ x and ∂ y of a function z = f ( x, y): given a
point (a, b) in the domain D of f ( x, y), the trace of the surface described by z = f ( x, y) in the
plane y = b is a curve in R3 through the point (a, b, f (a, b)), and the slope of the tangent line
∂f ∂f
L x to that curve at that point is ∂ x (a, b). Similarly, ∂ y (a, b) is the slope of the tangent line
L y to the trace of the surface z = f ( x, y) in the plane x = a (see Figure 3.3.1).

z z ∂f
z = f (x, y) slope = ∂y
(a, b)
∂f
(a, b, f (a, b)) slope = ∂x
(a, b) z = f (x, y)
(a, b, f (a, b))

Ly
Lx
b y y
0 0
a
(a, b) (a, b)
x x
D D
(a) Tangent line L x in the plane y = b (b) Tangent line L y in the plane x = a

Figure 3.3.1 Partial derivatives as slopes.

dy
Since the derivative dx of a function y = f ( x) is used to find the tangent line to the graph
of f (which is a curve in R2 ), you might expect that partial derivatives can be used to define
a tangent plane to the graph of a surface z = f ( x, y). This indeed turns out to be the case.
First, we need a definition of a tangent plane. The intuitive idea is that a tangent plane “just
touches” a surface at a point. The formal definition mimics the intuitive notion of a tangent
line to a curve.

Definition 3.4. Let z = f ( x, y) be the equation of a surface S in R3 , and let P = (a, b, c) be

a point on S . Let T be a plane which contains the point P , and let Q = ( x, y, z) represent a
−−→
generic point on the surface S . If the (acute) angle between the vector PQ and the plane
T approaches zero as the point Q approaches P along the surface S , then we call T the
tangent plane to S at P .

Note that since two lines in R3 determine a plane, then the two tangent lines to the surface
z = f ( x, y) in the x and y directions described in Figure 3.3.1 are contained in the tangent
plane at that point, if the tangent plane exists at that point. The existence of those two
 3.3 Tangent Plane to a Surface 85

tangent lines does not by itself guarantee the existence of the tangent plane. It is possible
that if we take the trace of the surface in the plane x − y = 0 (which makes a 45◦ angle with
the positive x-axis), the resulting curve in that plane may have a tangent line which is not
in the plane determined by the other two tangent lines, or it may not have a tangent line
∂f ∂f
at all at that point. Luckily, it turns out4 that if ∂ x and ∂ y exist in a region around a point
(a, b) and are continuous at (a, b) then the tangent plane to the surface z = f ( x, y) will exist
at the point (a, b, f (a, b)). In this text, those conditions will always hold.
Suppose that we want an equation of the tangent plane T z
to the surface z = f ( x, y) at a point (a, b, f (a, b)). Let L x and z = f (x, y)
L y be the tangent lines to the traces of the surface in the
(a, b, f (a, b)) L
planes y = b and x = a, respectively (as in Figure 2.3.2), and y
suppose that the conditions for T to exist do hold. Then the Lx
equation for T is T y
0
A ( x − a) + B( y − b) + C ( z − f (a, b)) = 0 (3.4) x
Figure 3.3.2 Tangent plane
where n = ( A, B, C ) is a normal vector to the plane T . Since
T contains the lines L x and L y , then all we need are vectors vx and v y that are parallel to L x
and L y , respectively, and then let n = vx × v y .
∂f
Since the slope of L x is ∂ x (a, b), then the vector vx = (1, 0, ∂ x (a, b)) is
∂f z
∂f
vx = (1, 0, ∂ x (a, b))
parallel to L x (since vx lies in the xz-plane and lies in a line with slope
∂f
∂x
(a,b) ∂f
1 =∂x
(a, b). See Figure 2.3.3). Similarly, the vector
∂f
∂f (a, b)
vy = (0, 1, ∂ y (a, b)) is parallel to L y . Hence, the vector ∂x
x
¯ ¯ 0
¯ i j k ¯¯ 1
¯
∂f
(a, b) ¯ = − ∂∂ xf (a, b) i − ∂∂ fy (a, b) j + k
¯ ¯
n = vx × v y = ¯ 1 0 ∂x Figure 3.3.3
¯ ∂f ¯
¯ 0 1 (a, b) ¯
∂y

is normal to the plane T . Thus the equation of T is

∂f ∂f
− ∂ x (a, b) ( x − a) − ∂ y (a, b) ( y − b) + z − f (a, b) = 0 . (3.5)

Multiplying both sides by −1, we have the following result:

The equation of the tangent plane to the surface z = f ( x, y) at the point (a, b, f (a, b)) is
∂f ∂f
∂x
(a, b) ( x − a) + ∂ y (a, b) ( y − b) − z + f (a, b) = 0 (3.6)

Example 3.13. Find the equation of the tangent plane to the surface z = x2 + y2 at the point
(1, 2, 5).
4 See T AYLOR and M ANN, § 6.4.
86 CHAPTER 3. FUNCTIONS OF SEVERAL VARIABLES

∂f ∂f
Solution: For the function f ( x, y) = x2 + y2 , we have ∂x
= 2 x and ∂y
= 2 y, so the equation of
the tangent plane at the point (1, 2, 5) is

2(1)( x − 1) + 2(2)( y − 2) − z + 5 = 0 , or
2x + 4 y − z − 5 = 0 .

In a similar fashion, it can be shown that if a surface is defined implicitly by an equation

of the form F ( x, y, z) = 0, then the tangent plane to the surface at a point (a, b, c) is given by
the equation
∂F
∂x
(a, b, c) ( x − a) + ∂∂Fy (a, b, c) ( y − b) + ∂∂Fz (a, b, c) ( z − c) = 0 . (3.7)
Note that formula (3.6) is the special case of formula (3.7) where F ( x, y, z) = f ( x, y) − z.

Example 3.14. Find the equation of the tangent plane to the surface x2 + y2 + z2 = 9 at the
point (2, 2, −1).
∂F ∂F ∂F
Solution: For the function F ( x, y, z) = x2 + y2 + z2 − 9, we have ∂x
= 2 x, ∂y
= 2 y, and ∂z
= 2 z,
so the equation of the tangent plane at (2, 2, −1) is

2(2)( x − 2) + 2(2)( y − 2) + 2(−1)( z + 1) = 0 , or

2x + 2 y − z − 9 = 0 .

Exercises
A
For Exercises 1–6, find the equation of the tangent plane to the surface z = f ( x, y) at the
point P .

1. f ( x, y) = x2 + y3 , P = (1, 1, 2); 2. f ( x, y) = x y, P = (1, −1, −1);

3. f ( x, y) = x2 y, P = (−1, 1, 1); 4. f ( x, y) = xe y , P = (1, 0, 1);

p
5. f ( x, y) = x + 2 y, P = (2, 1, 4); 6. f ( x, y) = x2 + y2 , P = (3, 4, 5).

For Exercises 7–10, find the equation of the tangent plane to the given surface at the point
P.
³ p ´
2 y2 z2
7. x4 + 9 + 16 = 1, P = 1, 2, 2 311 ; 8. x2 + y2 + z2 = 9, P = (0, 0, 3);
p
9. x2 + y2 − z2 = 0, P = (3, 4, 5); 10. x2 + y2 = 4, P = ( 3, 1, 0).

B
11. Find the angles between the curve f( t) = ( t, t2 , t3 ) and the surface x6 + y3 + z2 = 3 at their
intersections.
 3.4 Directional Derivatives and the Gradient 87

3.4 Directional Derivatives and the Gradient

∂f ∂f
For a function z = f ( x, y), we learned that the partial derivatives ∂ x and ∂ y represent the
(instantaneous) rate of change of f in the positive x and y directions, respectively. What
about other directions? It turns out that we can find the rate of change in any direction
using a more general type of derivative called a directional derivative.

Definition 3.5. Let f ( x, y) be a real-valued function with domain D in R2 , and let (a, b) be a
point in D . Let v be a vector in R2 . Then the directional derivative of f at (a, b) in the
direction of v, denoted by D v f (a, b), is defined as

f ((a, b) + hv) − f (a, b)

D v f (a, b) = lim . (3.8)
h→0 h

Notice in the definition that we seem to be treating the point (a, b) as a vector, since we
are adding the vector hv to it. But this is just the usual idea of identifying vectors with their
terminal points, which the reader should be used to by now. If we were to write the vector v
as v = (v1 , v2 ), then
f (a + hv1 , b + hv2 ) − f (a, b)
D v f (a, b) = lim . (3.9)
h→0 h
∂f ∂f
From this we can immediately recognize that the partial derivatives ∂ x and ∂ y are special
cases of the directional derivative with v = i = (1, 0) and v = j = (0, 1), respectively. That is,
∂f ∂f
∂x
= D i f and ∂ y = D j f .
∂f ∂f
If f ( x, y) has continuous partial derivatives ∂ x and ∂ y (which will always be the case in
this text), then there is a simple formula for the directional derivative:

Theorem 3.2. Let f ( x, y) be a real-valued function with domain D in R2 such that the partial
∂f ∂f
derivatives ∂ x and ∂ y exist and are continuous in D . Let (a, b) be a point in D . Then

∂f ∂f
D v f (a, b) = v1 (a, b) + v2 (a, b) . (3.10)
∂x ∂y

for any vector v = (v1 , v2 ) in R2

∂f
Proof: Note that if v = i = (1, 0) then the above formula reduces to D v f (a, b) = ∂x
(a, b),
∂f
which we know is true since D i f = ∂x
, as we noted earlier. Similarly, for v = j = (0, 1) the
∂f ∂f
formula reduces to D v f (a, b) = ∂y
(a, b),
which is true since D j f = ∂y
. Fix such a vector
v = (v1 , v2 ) and fix a number h 6= 0. Then

f (a + hv1 , b + hv2 ) − f (a, b) = f (a + hv1 , b + hv2 ) − f (a + hv1 , b) + f (a + hv1 , b) − f (a, b) . (3.11)

88 CHAPTER 3. FUNCTIONS OF SEVERAL VARIABLES

Since g(α) = f (a + hv1 , y + α hv2 ) is a real-valued function, we can apply the Mean Value
Theorem from single-variable calculus on the interval [0, 1]. It provides a number 0 < α < 1
such that
g(1) − g(0)
g ′ (α) =
1−0
= f (a + hv1 , b + hv2 ) − f (a + hv1 , b).

By chain rule
∂f
g ′ (α) = (a + hv1 , b + α hv2 ) hv2 .
∂y
Therefore,
∂f
f (a + hv1 , b + hv2 ) − f (a + hv1 , b) = hv2 (a + hv1 , b + α hv2 ).
∂x
By a similar argument, there exists a number 0 < β < 1 such that

∂f
f (a + hv1 , b) − f (a, b) = hv1 (a + β hv1 , b) .
∂x
Thus, by equation (3.11), we have
∂f ∂f
f (a + hv1 , b + hv2 ) − f (a, b) hv2 ∂ y (a + hv1 , b + α hv2 ) + hv1 ∂ x (a + β hv1 , b)
=
h h
∂f ∂f
= v2 (a + hv1 , b + α hv2 ) + v1 (a + β hv1 , b)
∂y ∂x

so by formula (3.9) we have

f (a + hv1 , b + hv2 ) − f (a, b)

D v f (a, b) = lim
h→0 h
· ¸
∂f ∂f
= lim v2 (a + hv1 , b + α hv2 ) + v1 (a + β hv1 , b)
h→0 ∂y ∂x

∂f ∂f ∂f ∂f
= v2 (a, b) + v1 (a, b) by the continuity of and , so
∂y ∂x ∂x ∂y

∂f ∂f
D v f (a, b) = v1 (a, b) + v2 (a, b)
∂x ∂y

after reversing the order of summation. QED

Along the same lines one can prove the following generalization of the chain rule.
 3.4 Directional Derivatives and the Gradient 89

Theorem 3.3. Let f ( x, y) be a real-valued function with domain D in R2 such that the
∂f ∂f
partial derivatives ∂ x and ∂ y exist and are continuous in D and h( t) = ( h 1 ( t), h 2 ( t)) be a
smooth function with values in D . Then

f (h( t))′ = D h′ (t) f (h( t))

∂f ∂f (3.12)
= h′1 ( t) ( h 1 ( t), h 2 ( t)) + h′2 ( t) ( h 1 ( t), h 2 ( t)).
∂x ∂y

³ ´
∂f ∂f
Note that D v f (a, b) = v · ∂x
(a, b), ∂ y (a, b) . The second vector has a special name:

Definition 3.6. For a real-valued function f ( x, y), the gradient of f , denoted by ∇ f , is the
vector ³∂f ∂f ´
∇f = , (3.13)
∂x ∂ y
in R2 . For a real-valued function f ( x, y, z), the gradient is the vector
³∂f ∂f ∂f ´
∇f = , , (3.14)
∂x ∂ y ∂z

in R3 . The symbol ∇ is pronounced “del” or “nabla”.5

Corollary 3.4. In the assumptions of the theorems 3.2 and 3.3 we have

(a) D v f = v · ∇ f ;

(b) f (h( t))′ = h′ ( t) · ∇ f

2 3
´ directional derivative of f ( x, y) = x y + x y at the point (1, 2) in the
Example 3.15. ³Find the
direction of v = p1 , p1 .
2 2

Solution: We see that ∇ f = ( y2 + 3 x2 y, 2 x y + x3 ), so

³ ´
D v f (1, 2) = v · ∇ f (1, 2) = p1 , p1 · (22 + 3(1)2 (2), 2(1)(2) + 13 ) = 15
p
2 2 2

∂f ∂f
A real-valued function z = f ( x, y) whose partial derivatives ∂ x and ∂ y exist and are con-
tinuous is called continuously differentiable. Assume that f ( x, y) is such a function and that
∇ f 6= 0. Let c be a real number in the range of f and let v be a vector in R2 which is tangent
to the level curve f ( x, y) = c (see Figure 3.4.1).
5 Sometimes the notation grad( f ) is used instead of ∇ f .
90 CHAPTER 3. FUNCTIONS OF SEVERAL VARIABLES

y
v ∇f

f ( x, y) = c

x
0

Figure 3.4.1

The value of f ( x, y) is constant along a level curve, so since v is a tangent vector to this
curve, then the rate of change of f in the direction of v is 0; that is, D v f = 0. But we know
that D v f = v · ∇ f . In other words, ∇ f ⊥ v, which means that ∇ f is normal to the level curve.
In general, for any unit vector v in R2 , we have D v f = k∇ f k cos θ , where θ is the angle
between v and ∇ f . At a fixed point ( x, y) the length k∇ f k is fixed, and the value of D v f then
varies as θ varies. The largest value that D v f can take is when cos θ = 1 (θ = 0◦ ), while the
smallest value occurs when cos θ = −1 (θ = 180◦ ). In other words, the value of the function
f increases the fastest in the direction of ∇ f (since θ = 0◦ in that case), and the value of
f decreases the fastest in the direction of −∇ f (since θ = 180◦ in that case). We have thus
proved the following theorem:

Theorem 3.5. Let f ( x, y) be a continuously differentiable real-valued function, with ∇ f 6= 0.

Then:

(a) The gradient ∇ f is normal to any level curve f ( x, y) = c.

(b) The value of f ( x, y) increases the fastest in the direction of ∇ f .

(c) The value of f ( x, y) decreases the fastest in the direction of −∇ f .

Example 3.16. In which direction does the function f ( x, y) = x y2 + x3 y increase the fastest
from the point (1, 2)? In which direction does it decrease the fastest?
Solution: Since ∇ f = (³y2 + 3 x´2 y, 2 x y + x3 ), then ∇ f (1, 2) = (10, 5) 6= 0. A unit vector
³ in´ that
∇f 2 1 2 1
direction is v = k∇ f k = p , p . Thus, f increases the fastest in the direction of p , p and
5 5 5 5
³ ´
− 2 − 1
decreases the fastest in the direction of p , p .
5 5
 3.4 Directional Derivatives and the Gradient 91

Though we proved Theorem 3.5 for functions of two variables, a similar argument can
be used to show that it also applies to functions of three or more variables. Likewise, the
directional derivative in the three-dimensional case can also be defined by the formula D v f =
v· ∇f .
Example 3.17. The temperature T of a solid is given by the function
T ( x, y, z) = e− x + e−2 y + e4 z ,
where x, y, z are space coordinates relative to the center of the solid. In which direction from
the point (1, 1, 1) will the temperature decrease the fastest?
Solution: Since ∇ f = (− e− x , −2 e−2 y , 4 e4 z ), then the temperature will decrease the fastest in
the direction of −∇ f (1, 1, 1) = ( e−1 , 2 e−2 , −4 e4 ).
Exercises
A
For Exercises 1–10, compute the gradient ∇ f .
1
1. f ( x, y) = x2 + y2 − 1; 2. f ( x, y) = ;
x 2 + y2
p
3. f ( x, y) = x2 + y2 + 4; 4. f ( x, y) = x2 e y ;

5. f ( x, y) = ln( x y); 6. f ( x, y) = 2 x + 5 y;

7. f ( x, y, z) = sin( x yz); 8. f ( x, y, z) = x2 e yz ;
p
9. f ( x, y, z) = x2 + y2 + z2 ; 10. f ( x, y, z) = x2 + y2 + z2 .

For ³Exercises
´ 11–14, find the directional derivative of f at the point P in the direction of
v= p1 , p1 .
2 2
1
11. f ( x, y) = x2 + y2 − 1, P = (1, 1); 12. f ( x, y) = , P = (1, 1);
x 2 + y2
p
13. f ( x, y) = x2 + y2 + 4, P = (1, 1); 14. f ( x, y) = x2 e y , P = (1, 1).

For ³Exercises 15–16,

´ find the directional derivative of f at the point P in the direction of
1 1 1
v= p ,p ,p .
3 3 3
15. f ( x, y, z) = sin( x yz), P = (1, 1, 1); 16. f ( x, y, z) = x2 e yz , P = (1, 1, 1).

17. Repeat Example 2.16 at the point (2, 3).

18. Repeat Example 2.17 at the point (3, 1, 2).

B
For Exercises 19–26, let f ( x, y) and g( x, y) be continuously differentiable real-valued func-
tions, let c be a constant, and let v be a unit vector in R2 . Show that:
92 CHAPTER 3. FUNCTIONS OF SEVERAL VARIABLES

19. ∇( c f ) = c ∇ f ; 20. ∇( f + g) = ∇ f + ∇ g;
g ∇f − f ∇g
21. ∇( f g) = f ∇ g + g ∇ f ; 22. ∇( f / g) = if g( x, y) 6= 0;
g2
23. D −v f = −D v f ; 24. D v ( c f ) = c D v f ;

25. D v ( f + g) = D v f + D v g; 26. D v ( f g) = f D v g + g D v f .
p
27. The function r ( x, y) = x2 + y2 is the length of the position vector r = x i + y j for each
point ( x, y) in R2 . Show that ∇ r = 1r r when ( x, y) 6= (0, 0), and that ∇( r 2 ) = 2 r.

C
28. Let g( x) and f ( x, y) be smooth function such that

f ( x, g( x)) = 0.

Show that
∂f ∂f
∂x
( x, g( x)) + g′ ( x) ∂ y ( x, g( x)) = 0.
(Hint: Apply Theorem 3.3 for the curve h( t) = ( t, g( t)).)
 3.5 Maxima and Minima 93

3.5 Maxima and Minima

The gradient can be used to find extreme points of real-valued functions of several variables,
that is, points where the function has a local maximum or local minimum. We will consider
only functions of two variables; functions of three or more variables require methods using
linear algebra.

Definition 3.7. Let f ( x, y) be a real-valued function, and let ( x0 , y0 ) be a point in the domain
of f . We say that f has a local maximum at ( x0 , y0 ) if f ( x, y) ≤ f ( x0 , y0 ) for all ( x, y) inside
some disk of positive radius centered at ( x0 , y0 ); that is, there is some sufficiently small r > 0
such that f ( x, y) ≤ f ( x0 , y0 ) for all ( x, y) for which ( x − x0 )2 + ( y − y0 )2 < r 2 .
Likewise, we say that f has a local minimum at ( x0 , y0 ) if f ( x, y) ≥ f ( x0 , y0 ) for all ( x, y)
inside some disk of positive radius centered at ( x0 , y0 ).
If f ( x, y) ≤ f ( x0 , y0 ) for all ( x, y) in the domain of f , then f has a global maximum at
( x0 , y0 ). If f ( x, y) ≥ f ( x0 , y0 ) for all ( x, y) in the domain of f , then f has a global minimum
at ( x0 , y0 ).

Suppose that ( x0 , y0 ) is a local maximum point for f ( x, y), and that the first-order partial
derivatives of f exist at ( x0 , y0 ). We know that f ( x0 , y0 ) is the largest value of f ( x, y) as
( x, y) goes in all directions from the point ( x0 , y0 ), in some sufficiently small disk centered at
( x0 , y0 ). In particular, f ( x0 , y0 ) is the largest value of f in the x direction (around the point
( x0 , y0 )), that is, the single-variable function g( x) = f ( x, b) has a local maximum at x = a. So
∂f ∂f
we know that g ′ (a) = 0. Since g ′ ( x) = ∂ x ( x, b), then ∂ x ( x0 , y0 ) = 0. Similarly, f ( x0 , y0 ) is the
∂f
largest value of f near ( x0 , y0 ) in the y direction and so (x , y )
∂y 0 0
= 0. We thus have the
following theorem:

∂f ∂f
Theorem 3.6. Let f ( x, y) be a real-valued function such that both ∂ x ( x0 , y0 ) and ∂ y ( x0 , y0 )
exist. Then a necessary condition for f ( x, y) to have a local maximum or minimum at ( x0 , y0 )
is that ∇ f ( x0 , y0 ) = 0.

Note: Theorem 3.6 can be extended to apply to functions of three or more variables.
A point ( x0 , y0 ) where ∇ f ( x0 , y0 ) = 0 is called a critical point for the function f ( x, y).
So given a function f ( x, y), to find the critical points of f you have to solve the equations
∂f ∂f
∂x
( x, y) = 0 and ∂ y ( x, y) = 0 simultaneously for ( x, y). Similar to the single-variable case, the
necessary condition that ∇ f ( x0 , y0 ) = 0 is not always sufficient to guarantee that a critical
point is a local maximum or minimum.

∂f
Example 3.18. The function f ( x, y) = x y has a critical point at (0, 0): ∂x
= y = 0 ⇒ y = 0, and
∂f
∂y
= x = 0 ⇒ x = 0, so (0, 0) is the only critical point. But clearly f does not have a local
maximum or minimum at (0, 0) since any disk around (0, 0) contains points ( x, y) where the
values of x and y have the same sign (so that f ( x, y) = x y > 0 = f (0, 0)) and different signs (so
that f ( x, y) = x y < 0 = f (0, 0)). In fact, along the path y = x in R2 , f ( x, y) = x2 , which has a
94 CHAPTER 3. FUNCTIONS OF SEVERAL VARIABLES

local minimum at (0, 0), while along the path y = − x we have f ( x, y) = − x2 , which has a local
maximum at (0, 0). So (0, 0) is an example of a saddle point; that is, it is a local maximum
in one direction and a local minimum in another direction. The graph of f ( x, y) is shown in
Figure 3.5.1, which is a hyperbolic paraboloid.

100

50

0
z

-50

-10
-100-10 -5
-5 0 x
0 5 5
y 10 10

Figure 3.5.1 f (x, y) = x y, saddle point at (0, 0).

From the course of single-variable calculus, you may remember second derivative test.
If f ′ ( x0 ) = 0 and f ′′ (0) > 0 then the real-to-real function f has a local minimum at x0 .

In order to explain the multi-variable analog of this test, let us introduce second direc-
tional derivative. Fix a vector v ∈ R2 and a smooth function of two variables f ( x, y). The
directional derivative h( x, y) = D v f ( x, y) is an other smooth function of two variables, so we
can take its directional derivative again D v h( x, y); it is called second directional derivative
and denoted as D v2 f ( x, y).
If D v f ( x0 , y0 ) = 0 and D v2 f ( x0 , y0 ) > 0 for any vector v 6= 0 then the the smooth function
f ( x, y) of two variables, has a local minimum at ( x0 , y0 ).

In this form the second derivative test is not useful since it requires to check inequality
D v2 f ( x0 , y0 ) > 0 for infinite number of vectors v. Let us try to remove this weak point.
 3.5 Maxima and Minima 95

Note that if v = (a, b) then

∂f ∂f
D v f ( x0 , y0 ) = a ∂ x + b ∂ y .

and

∂f ∂f
D v2 f ( x0 , y0 ) = D v (a ∂ x + b ∂ y )
∂2 f ∂2 f ∂2 f ∂2 f
= a2 ∂ x2 + ab ∂ x∂ y + ab ∂ y∂ x + b2 ∂ y2
∂2 f ∂2 f ∂2 f
= a2 ∂ x2 + 2ab ∂ y∂ x + b2 ∂ y2 ,

∂2 f ∂2 f
the last equality holds since the function f is smooth and, therefore, ∂ x∂ y
= ∂ y∂ x
Therefore, the condition D v2 f ( x0 , y0 ) > 0 for any v 6= 0 means that

∂2 f ∂2 f ∂2 f
a2 ∂ x2 + 2ab ∂ y∂ x + b2 ∂ y2 > 0

for any pair of real numbers (a, b) at least one of which is not zero.
Analyzing the last inequality for all possible pairs (a, b) leads to the following theorem
which is true analog of second derivative test for smooth functions of two variables; it gives
sufficient conditions for a critical point to be a local maximum or minimum of a smooth
function (that is, a function whose partial derivatives of all orders exist and are continuous).
The theorem will not be proved here.6

Theorem 3.7. Let f ( x, y) be a smooth real-valued function, with a critical point at ( x0 , y0 )

(that is, ∇ f ( x0 , y0 ) = 0). Define

∂2 f ∂2 f ³ ∂2 f ´2
D= (x , y )
2 0 0
(x , y ) −
2 0 0
( x0 , y0 )
∂x ∂y ∂ y ∂x

Then
∂2 f
(a) if D > 0 and ( x , y ) > 0,
∂ x2 0 0
then f has a local minimum at ( x0 , y0 )

∂2 f
(b) if D > 0 and ( x , y ) < 0,
∂ x2 0 0
then f has a local maximum at ( x0 , y0 )

(c) if D < 0, then f has neither a local minimum nor a local maximum at ( x0 , y0 )

(d) if D = 0, then the test fails.

If condition (c) holds, then ( x0 , y0 ) is a saddle point; that is, the second directional deriva-
tive D v2 f ( x0 , y0 ) can be positive and negative for different vectors v.
6 See T AYLOR and M ANN, § 7.6.
96 CHAPTER 3. FUNCTIONS OF SEVERAL VARIABLES

∂2 f ∂2 f
Recall that the assumption that f ( x, y) is smooth implies that ∂ y ∂x
= ∂x ∂ y
. Therefore
¯ 2 ¯
¯ ∂ f ∂2 f ¯
¯
¯ ( x0 , y0 ) ( x0 , y0 )¯¯
¯ ∂ x2 ∂ y ∂x ¯
D=¯ 2 ¯.
¯ ∂ f ∂2 f ¯
¯ ∂ x ∂ y ( x0 , y0 ) ( x0 , y0 ) ¯¯
¯
∂y 2

³ ´2
∂2 f ∂2 f ∂2 f ∂2 f ∂2 f
Also, if D > 0 then (x , y )
∂ x2 0 0 ∂ y2 0 0
( x , y ) = D + ∂ y ∂ x ( x0 , y0 ) > 0, and so ( x , y ) and ∂ y2 ( x0 , y0 )
∂ x2 0 0
have the same sign. This means that in parts (a) and (b) of the theorem one can replace
∂2 f ∂2 f
( x , y ) by ∂ y2 ( x0 , y0 ) if desired.
∂ x2 0 0

Example 3.19. Find all local maxima and minima of f ( x, y) = x2 + x y + y2 − 3 x.

Solution: First find the critical points; that is, solve ∇ f = 0. Since
∂f ∂f
= 2x + y − 3 and = x + 2y
∂x ∂y

then the critical points ( x, y) are the common solutions of the equations

2x + y − 3 = 0
x + 2y =0

which has the unique solution ( x, y) = (2, −1). So (2, −1) is the only critical point.
To use Theorem 3.7, we need the second-order partial derivatives:

∂2 f ∂2 f ∂2 f
=2 , =2 , =1
∂ x2 ∂ y2 ∂ y ∂x

and so
∂2 f ∂2 f ³ ∂2 f ´2
D = 2
(2, −1) 2
(2, −1) − (2, −1) = (2)(2) − 12 = 3 > 0
∂x ∂y ∂ y ∂x
2
∂ f
and ∂ x2
(2, −1) = 2 > 0. Thus, (2, −1) is a local minimum.

Example 3.20. Find all local maxima and minima of f ( x, y) = x y − x3 − y2 .

Solution: First find the critical points; that is, solve ∇ f = 0. Since
∂f ∂f
= y − 3 x2 and = x − 2y
∂x ∂y

then the critical points ( x, y) are the common solutions of the equations

y − 3 x2 = 0
x − 2y = 0
 3.5 Maxima and Minima 97

The first equation yields y = 3 x2 , substituting that into the second equation yields x −6 x2 = 0,
¡ ¢2 1
which has the solutions x = 0 and x = 61 . So x = 0 ⇒ y = 3(0) = 0 and x = 61 ⇒ y = 3 61 = 12 .
¡1 1 ¢
So the critical points are ( x, y) = (0, 0) and ( x, y) = 6 , 12 .
To use Theorem 3.7, we need the second-order partial derivatives:

∂2 f ∂2 f ∂2 f
= −6 x , = −2 , =1
∂ x2 ∂ y2 ∂ y ∂x

So
∂2 f ∂2 f ³ ∂2 f ´2
D = 2
(0, 0) 2
(0, 0) − (0, 0) = (−6(0))(−2) − 12 = −1 < 0
∂x ∂y ∂ y ∂x
and thus (0, 0) is a saddle point. Also,

∂2 f ¡ 1 ¢ ∂2 f ¡ 1 ¢ ³ ∂2 f ¡ 1 1 ¢´2 ¡1¢
D = , 1 , 1 − , 12 = (−6 )(−2) − 12 = 1 > 0
∂ x2 6 12 ∂ y2 6 12 ∂ y ∂x 6 6

∂2 f ¡ 1 1 ¢ ¡1 1
¢
and ,
∂ x2 6 12
= −1 < 0. Thus, 6 , 12 is a local maximum.

Example 3.21. Find all local maxima and minima of f ( x, y) = ( x − 2)4 + ( x − 2 y)2 .
Solution: First find the critical points; that is, solve ∇ f = 0. Since

∂f ∂f
= 4( x − 2)3 + 2( x − 2 y) and = −4( x − 2 y)
∂x ∂y

then the critical points ( x, y) are the common solutions of the equations

4( x − 2)3 + 2( x − 2 y) = 0
−4( x − 2 y) = 0

The second equation yields x = 2 y, substituting that into the first equation yields 4(2 y − 2)3 =
0, which has the solution y = 1, and so x = 2(1) = 2. Thus, (2, 1) is the only critical point.
To use Theorem 3.7, we need the second-order partial derivatives:

∂2 f ∂2 f ∂2 f
= 12( x − 2)2 + 2 , =8 , = −4
∂ x2 ∂ y2 ∂ y ∂x

So
∂2 f ∂2 f ³ ∂2 f ´2
D = 2
(2, 1) 2
(2, 1) − (2, 1) = (2)(8) − (−4)2 = 0
∂x ∂y ∂ y ∂x
and so the test fails. What can be done in this situation? Sometimes it is possible to examine
the function to see directly the nature of a critical point. In our case, we see that f ( x, y) ≥ 0
for all ( x, y), since f ( x, y) is the sum of fourth and second powers of numbers and hence must
be nonnegative. But we also see that f (2, 1) = 0. Thus f ( x, y) ≥ 0 = f (2, 1) for all ( x, y), and
hence (2, 1) is, in fact, a global minimum for f .
98 CHAPTER 3. FUNCTIONS OF SEVERAL VARIABLES

2
+ y2 )
Example 3.22. Find all local maxima and minima of f ( x, y) = ( x2 + y2 ) e−( x .
Solution: First find the critical points; that is, solve ∇ f = 0. Since

∂f 2
+ y2 )
= 2 x(1 − ( x2 + y2 )) e−( x
∂x
∂f 2
+ y2 )
= 2 y(1 − ( x2 + y2 )) e−( x
∂y

then the critical points are (0, 0) and all points ( x, y) on the unit circle x2 + y2 = 1.
To use Theorem 3.7, we need the second-order partial derivatives:

∂2 f 2
+ y2 )
= 2[1 − ( x2 + y2 ) − 2 x2 − 2 x2 (1 − ( x2 + y2 ))] e−( x
∂ x2
∂2 f 2
+ y2 )
= 2[1 − ( x2 + y2 ) − 2 y2 − 2 y2 (1 − ( x2 + y2 ))] e−( x
∂ y2

∂2 f 2
+ y2 )
= −4 x y[2 − ( x2 + y2 )] e−( x
∂ y ∂x

∂2 f
At (0, 0), we have D = 4 > 0 and ∂ x2 (0, 0) = 2 > 0, so (0, 0) is a local minimum. However, for
points ( x, y) on the unit circle x2 + y2 = 1, we have

D = (−4 x2 e−1 )(−4 y2 e−1 ) − (−4 x ye−1 )2 = 0

and so the test fails. If we look at the graph of f ( x, y), as shown in Figure 3.5.2, it looks like
we might have a local maximum for ( x, y) on the unit circle x2 + y2 = 1. If we switch to using
polar coordinates ( r, θ ) instead of ( x, y) in R2 , where r 2 = x2 + y2 , then we see that we can write
2 2
f ( x, y) as a function g( r ) of the variable r alone: g( r ) = r 2 e−r . Then g ′ ( r ) = 2 r (1 − r 2 ) e−r ,
so it has a critical point at r = 1, and we can check that g ′′ (1) = −4 e−1 < 0, so the Second
Derivative Test from single-variable calculus says that r = 1 is a local maximum. But r = 1
corresponds to the unit circle x2 + y2 = 1. Thus, the points ( x, y) on the unit circle x2 + y2 = 1
are local maximum points for f .

Exercises
A
For Exercises 1–10, find all local maxima and minima of the function f ( x, y).
 3.5 Maxima and Minima 99

0.4
0.35
0.3
0.25
z 0.2
0.15 -3
0.1 -2
0.05
0 -1
-3
-2 0
-1 x
1
0
1 2
y 2
3 3

2 + y2 )
Figure 3.5.2 f (x, y) = (x2 + y2 )e−( x .

1. f ( x, y) = x3 − 3 x + y2 ; 2. f ( x, y) = x3 − 12 x + y2 + 8 y;

3. f ( x, y) = x3 − 3 x + y3 − 3 y; 4. f ( x, y) = x3 + 3 x2 + y3 − 3 y2 ;

5. f ( x, y) = 2 x3 + 6 x y + 3 y2 ; 6. f ( x, y) = 2 x3 − 6 x y + y2 ;
p
7. f ( x, y) = x2 + y2 ; 8. f ( x, y) = x + 2 y;

9. f ( x, y) = 4 x2 − 4 x y + 2 y2 + 10 x − 6 y; 10. f ( x, y) = −4 x2 + 4 x y − 2 y2 + 16 x − 12 y.

B
11. For a rectangular solid of volume 1000 cubic meters, find the dimensions that will min-
imize the surface area. (Hint: Use the volume condition to write the surface area as a
function of just two variables.)

12. Prove that if ( x0 , y0 ) is a local maximum or local minimum point for a smooth function
f ( x, y), then the tangent plane to the surface z = f ( x, y) at the point ( x0 , y0 , f ( x0 , y0 )) is
parallel to the x y-plane. (Hint: Use Theorem 3.6.)
C
13. Find three positive numbers x, y, z whose sum is 10 such that x2 y2 z is a maximum.
100 CHAPTER 3. FUNCTIONS OF SEVERAL VARIABLES

3.6 Numerical Methods

The types of problems that we solved in the previous section were examples of unconstrained
optimization problems. That is, we tried to find local (and perhaps even global) maximum
and minimum points of real-valued functions f ( x, y), where the points ( x, y) could be any
points in the domain of f . The method we used required us to find the critical points of
f , which meant having to solve the equation ∇ f = 0, which in general is a system of two
equations in two unknowns ( x and y). While this was relatively simple for the examples we
did, in general this will not be the case. It might be impossible to solve these equations by
elementary means.7
In a situation such as this, the only choice may be to find a solution using some numer-
ical method which gives a sequence of numbers which converge to the actual solution. For
example, Newton’s method for solving equations f ( x) = 0, which you probably learned in
single-variable calculus. In this section we will describe another method of Newton for find-
ing critical points of real-valued functions of two variables.
Let f ( x, y) be a smooth real-valued function, and define

∂2 f ∂2 f ³ ∂2 f ´2
D ( x, y) = 2
( x, y) 2
( x, y) − ( x, y) .
∂x ∂y ∂ y ∂x
Newton’s algorithm: Pick an initial point ( x0 , y0 ). For n = 0, 1, 2, 3, . . . , define:
¯ 2 ¯
¯∂ f ∂2 f ¯
¯ ∂ y2 ( xn , yn ) ∂ x ∂ y ( xn , yn )¯
¯ ¯
¯ ∂f ∂f ¯
¯ ∂ y ( xn , yn ) (x , y ) ¯
∂x n n
x n +1 = x n − ,
D ( xn , yn )
¯ 2 ¯ (3.15)
¯∂ f ∂2 f ¯
¯ ∂ x2 ( xn , yn ) ∂ x ∂ y ( xn , yn )¯
¯ ¯
¯ ∂f ∂f ¯
¯ ∂ x ( xn , yn ) (
∂y n n
x , y ) ¯
yn+1 = yn − .
D ( xn , yn )
Then the sequence of points ( xn , yn )∞
n=1 typically converges to a critical point. If there are
several critical points, then you will have to try different initial points to find them.
The choice of the formulas in (3.15) is motivated by the following fact, which can be
∂2 f ∂2 f
checked by direct calculations. Assume that the partial derivatives ∂ x2 ( x, y), ∂ x∂ y ( x, y) and
∂2 f
∂ y2
( x, y)
are constants; in other words, the function f ( x, y) can be expressed as a quadratic
polynomial in x and y, say

f ( x, y) = a + bx + c y + lx2 + mx y + n y2

for some constants a, b, c, l, m, n. Then for any choice ( x0 , y0 ) the formulas (3.15) returns a
critical point ( x1 , y1 ), which is unique in this case.
7 This is also a problem for the equivalent method (the Second Derivative Test) in single-variable calculus,

though one that is not usually emphasized.

 3.6 Numerical Methods 101

Example 3.23. Find all local maxima and minima of f ( x, y) = x3 − x y − x + x y3 − y4 .

Solution: First calculate the necessary partial derivatives:

∂f ∂f
= 3 x 2 − y − 1 + y3 , = − x + 3 x y2 − 4 y3
∂x ∂y

∂2 f ∂2 f ∂2 f
= 6x , = 6 x y − 12 y2 , = −1 + 3 y 2
∂ x2 ∂ y2 ∂ y ∂x

Notice that solving ∇ f = 0 would involve solving two third-degree polynomial equations in x
and y, which in this case can not be done easily.
We need to pick an initial point ( x0 , y0 ) for our algorithm. Looking at the graph of z = f ( x, y)
over a large region may help (see Figure 3.6.1 below), though it may be hard to tell where
the critical points are.

50000
0
-50000
-100000
-150000
-200000 -20
-250000 -15
-300000 -10
-350000 -5
-20
-15 0
-10 5 x
-5
0 10
5
y 10 15
15
20 20

Figure 3.6.1 f (x, y) = x3 − x y − x + x y3 − y4 for −20 ≤ x ≤ 20 and −20 ≤ y ≤ 20.

Notice in the formulas (3.15) that we divide by D , so we should pick an initial point where
D is not zero. And we can see that D (0, 0) = (0)(0) − (−1)2 = −1 6= 0, so take (0, 0) as our initial
point. Since it may take a large number of iterations of Newton’s algorithm to be sure that
we are close enough to the actual critical point, and since the computations are quite tedious,
we will let a computer do the computing. For this, we will write a simple program, using
the Java programming language, which will take a given initial point as a parameter and
102 CHAPTER 3. FUNCTIONS OF SEVERAL VARIABLES

then perform 100 iterations of Newton’s algorithm. In each iteration the new point will be
printed, so that we can see if there is convergence. The full code is shown in Listing 3.1.
//Program to find the critical points of f(x,y)=x^3-xy-x+xy^3-y^4
public class newton {
public static void main(String[] args) {
//Get the initial point (x,y) as command-line parameters
double x = Double.parseDouble(args[0]); //Initial x value
double y = Double.parseDouble(args[1]); //Initial y value
System.out.println("Initial point: (" + x + "," + y + ")");
//Go through 100 iterations of Newton’s algorithm
for (int n=1; n<=100; n++) {
double D = fxx(x,y)*fyy(x,y) - Math.pow(fxy(x,y),2);
double xn = x; double yn = y; //The current x and y values
if (D == 0) { //We can not divide by 0
System.out.println("Error: D = 0 at iteration n = " + n);
System.exit(0); //End the program
} else { //Calculate the new values for x and y
x = xn - (fyy(xn,yn)*fx(xn,yn) - fxy(xn,yn)*fy(xn,yn))/D;
y = yn - (fxx(xn,yn)*fy(xn,yn) - fxy(xn,yn)*fx(xn,yn))/D;
System.out.println("n = " + n + ": (" + x + "," + y + ")");
}
}
}
//Below are the parts specific to the function f
//The first partial derivative of f wrt x: 3x^2-y-1+y^3
public static double fx(double x, double y) {
return 3*Math.pow(x,2) - y - 1 + Math.pow(y,3);
}
//The first partial derivative of f wrt y: -x+3xy^2-4y^3
public static double fy(double x, double y) {
return -x + 3*x*Math.pow(y,2) - 4*Math.pow(y,3);
}
//The second partial derivative of f wrt x: 6x
public static double fxx(double x, double y) {
return 6*x;
}
//The second partial derivative of f wrt y: 6xy-12y^2
public static double fyy(double x, double y) {
return 6*x*y - 12*Math.pow(y,2);
}
//The mixed second partial derivative of f wrt x and y: -1+3y^2
public static double fxy(double x, double y) {
return -1 + 3*Math.pow(y,2);
}
}

Listing 3.1 Program listing for newton.java

 3.6 Numerical Methods 103

To use this program, you should first save the code in Listing 3.1 in a plain text file called
newton.java. You will need the Java Development Kit8 to compile the code. In the directory
where newton.java is saved, run this command at a command prompt to compile the code:
javac newton.java
Then run the program with the initial point (0, 0) with this command:
java newton 0 0
Below is the output of the program using (0, 0) as the initial point, truncated to show the
first 10 lines and the last 5 lines:

java newton 0 0
Initial point: (0.0,0.0)
n = 1: (0.0,-1.0)
n = 2: (1.0,-0.5)
n = 3: (0.6065857885615251,-0.44194107452339687)
n = 4: (0.484506572966545,-0.405341511995805)
n = 5: (0.47123972682634485,-0.3966334583092305)
n = 6: (0.47113558510349535,-0.39636450001936047)
n = 7: (0.4711356343449705,-0.3963643379632247)
n = 8: (0.4711356343449874,-0.39636433796318005)
n = 9: (0.4711356343449874,-0.39636433796318005)
n = 10: (0.4711356343449874,-0.39636433796318005)
...
n = 96: (0.4711356343449874,-0.39636433796318005)
n = 97: (0.4711356343449874,-0.39636433796318005)
n = 98: (0.4711356343449874,-0.39636433796318005)
n = 99: (0.4711356343449874,-0.39636433796318005)
n = 100: (0.4711356343449874,-0.39636433796318005)

As you can see, we appear to have converged fairly quickly (after only 8 iterations) to
what appears to be an actual critical point (up to Java’s level of precision), namely the point
(0.4711356343449874, −0.39636433796318005). It is easy to confirm that ∇ f = 0 at this
∂f ∂f
point, either by evaluating ∂ x and ∂ y at the point ourselves or by modifying our program to
also print the values of the partial derivatives at the point. It turns out that both partial
derivatives are indeed close enough to zero to be considered zero:
∂f
(0.4711356343449874, −0.39636433796318005) = 4.85722573273506 × 10−17
∂x
∂f
(0.4711356343449874, −0.39636433796318005) = −8.326672684688674 × 10−17
∂y
We also have D (0.4711356343449874, −0.39636433796318005) = −8.776075636032301 < 0,
so by Theorem 3.7 we know that (0.4711356343449874, −0.39636433796318005) is a saddle
point.
8 Available for free at http://www.oracle.com/technetwork/java/javase/downloads/
104 CHAPTER 3. FUNCTIONS OF SEVERAL VARIABLES

Since ∇ f consists of cubic polynomials, it seems likely that there may be three critical
points. The computer program makes experimenting with other initial points easy, and
trying different values does indeed lead to different sequences which converge:

java newton -1 -1
Initial point: (-1.0,-1.0)
n = 1: (-0.5,-0.5)
n = 2: (-0.49295774647887325,-0.08450704225352113)
n = 3: (-0.1855674752461383,-1.2047647348546167)
n = 4: (-0.4540060574531383,-0.8643989895639324)
n = 5: (-0.3672160534444,-0.5426077421319053)
n = 6: (-0.4794622222856417,-0.24529117721011612)
n = 7: (0.11570743992954591,-2.4319791238981274)
n = 8: (-0.05837851765533317,-1.6536079835854451)
n = 9: (-0.129841298650007,-1.121516233310142)
n = 10: (-1.004453014967208,-0.9206128022529645)
n = 11: (-0.5161209914612475,-0.4176293491131443)
n = 12: (-0.5788664043863884,0.2918236503332734)
n = 13: (-0.6985177124230715,0.49848120123515316)
n = 14: (-0.6733618916578702,0.4345777963475479)
n = 15: (-0.6704392913413444,0.4252025996474051)
n = 16: (-0.6703832679150286,0.4250147307973365)
n = 17: (-0.6703832459238701,0.42501465652421205)
n = 18: (-0.6703832459238667,0.4250146565242004)
n = 19: (-0.6703832459238667,0.42501465652420045)
n = 20: (-0.6703832459238667,0.42501465652420045)
...
n = 98: (-0.6703832459238667,0.42501465652420045)
n = 99: (-0.6703832459238667,0.42501465652420045)
n = 100: (-0.6703832459238667,0.42501465652420045)
∂f ∂f
Again, it is easy to confirm that both ∂ x and ∂ y vanish at the point
(−0.6703832459238667, 0.42501465652420045), which means it is a critical point. And
D (−0.6703832459238667, 0.42501465652420045) = 15.3853578526055 > 0
∂2 f
(−0.6703832459238667, 0.42501465652420045) = −4.0222994755432 < 0
∂ x2
so we know that (−0.6703832459238667, 0.42501465652420045) is a local maximum. An
idea of what the graph of f looks like near that point is shown in Figure 3.6.2, which does
suggest a local maximum around that point.
Finally, running the computer program with the initial point (−5, −5) yields the critical
point (−7.540962756992551, −5.595509445899435), with D < 0 at that point, which makes
it a saddle point.
 3.6 Numerical Methods 105

(−0.67, 0.42, 0.57)

0.6
0.4
0.2
0
z -0.2
-0.4 -1
-0.6
-0.8 -0.8
-1 -0.6
0
0.2 -0.4 x
0.4
0.6 -0.2
y 0.8
1 0

Figure 3.6.2 f (x, y) = x3 − x y − x + x y3 − y4 for −1 ≤ x ≤ 0 and 0 ≤ y ≤ 1.

We can summarize our findings for the function f ( x, y) = x3 − x y − x + x y3 − y4 :

(0.4711356343449874, −0.39636433796318005) : saddle point

(−0.6703832459238667, 0.42501465652420045) : local maximum
(−7.540962756992551, −5.595509445899435) : saddle point

The derivation of Newton’s algorithm, and the proof that it converges (given a “reason-
able” choice for the initial point) requires techniques beyond the scope of this text. See
R ALSTON and R ABINOWITZ for more detail and for discussion of other numerical methods.
Our description of Newton’s algorithm is the special two-variable case of a more general
algorithm that can be applied to functions of n ≥ 2 variables.
In the case of functions which have a global maximum or minimum, Newton’s algorithm
can be used to find those points. In general, global maxima and minima tend to be more
interesting than local versions, at least in practical applications. A maximization problem
can always be turned into a minimization problem (why?), so a large number of methods
have been developed to find the global minimum of functions of any number of variables.
This field of study is called nonlinear programming.
Many of these methods are based on the steepest descent technique, which is based on
an idea that we discussed in Section 2.4. Recall that the negative gradient −∇ f gives the
106 CHAPTER 3. FUNCTIONS OF SEVERAL VARIABLES

direction of the fastest rate of decrease of a function f . The crux of the steepest descent idea,
then, is that starting from some initial point, you move a certain amount in the direction
of −∇ f at that point. Wherever that takes you becomes your new point, and you then just
keep repeating that procedure until eventually (hopefully) you reach the point where f has
its smallest value. There is a “pure” steepest descent method, and a multitude of variations
on it that improve the rate of convergence, ease of calculation, etc. For more discussion of
this, and of nonlinear programming in general, see B AZARAA, S HERALI and S HETTY.

Exercises
C
1. Recall Example 3.21 from the previous section, where we showed that the point (2, 1) was
a global minimum for the function f ( x, y) = ( x − 2)4 + ( x − 2 y)2 . Notice that our computer
program can be modified fairly easily to use this function (just change the return values
in the fx, fy, fxx, fyy and fxy function definitions to use the appropriate partial derivative).
Either modify that program or write one of your own in a programming language of your
choice to show that Newton’s algorithm does lead to the point (2, 1). First use the initial
point (0, 3), then use the initial point (3, 2), and compare the results. Make sure that your
program attempts to do 100 iterations of the algorithm. Did anything strange happen
when your program ran? If so, how do you explain it? (Hint: Something strange should
happen.)

2. There is a version of Newton’s algorithm for solving a system of two equations

f 1 ( x, y) = 0 and f 2 ( x, y) = 0 ,

where f 1 ( x, y) and f 2 ( x, y) are smooth real-valued functions:

Pick an initial point ( x0 , y0 ). For n = 0, 1, 2, 3, . . . , define:
¯ ¯ ¯ ¯
¯ f 1 ( xn , yn ) f 2 ( xn , yn ) ¯¯ ¯ f 1 ( xn , yn ) f 2 ( xn , yn ) ¯¯
¯ ¯
¯ ∂ f1 ¯
¯ ( xn , yn ) ∂ f 2 ( xn , yn )¯
¯ ∂ f1 ¯
¯ ( xn , yn ) ∂ f 2 ( xn , yn )¯
∂y ∂y ∂x ∂x
x n +1 = x n − , yn+1 = yn + , where
D ( xn , yn ) D ( xn , yn )
∂ f1 ∂ f2 ∂ f1 ∂ f2
D ( xn , yn ) = ( xn , yn ) ( xn , yn ) − ( xn , yn ) ( xn , yn ) .
∂x ∂y ∂y ∂x
Then the sequence of points ( xn , yn )∞
n=1 converges to a solution. Write a computer program
that uses this algorithm to find approximate solutions to the system of equations

f 1 ( x, y) = sin( x y) − x − y = 0 and f 2 ( x, y) = e2 x − 2 x + 3 y = 0 .

Show that you get two different solutions when using (0, 0) and (1, 1) for the initial point
( x0 , y0 ).
 3.7 Lagrange Multipliers 107

3.7 Lagrange Multipliers

In Sections 2.5 and 2.6 we were concerned with finding maxima and minima of functions
without any constraints on the variables (other than being in the domain of the function).
What would we do if there were constraints on the variables? The following example illus-
trates a simple case of this type of problem.

Example 3.24. For a rectangle whose perimeter is 20 m, find the dimensions that will max-
imize the area.
Solution: The area A of a rectangle with width x and height y is A = x y. The perimeter P of
the rectangle is then given by the formula P = 2 x + 2 y. Since we are given that the perimeter
P = 20, this problem can be stated as:

Maximize : f ( x, y) = x y
given : 2 x + 2 y = 20

The reader is probably familiar with a simple method, using single-variable calculus, for
solving this problem. Since we must have 2 x + 2 y = 20, then we can solve for, say, y in
terms of x using that equation. This gives y = 10 − x, which we then substitute into f to get
f ( x, y) = x y = x(10 − x) = 10 x − x2 . This is now a function of x alone, so we now just have to
maximize the function f ( x) = 10 x − x2 on the interval [0, 10]. Since f ′ ( x) = 10 − 2 x = 0 ⇒ x = 5
and f ′′ (5) = −2 < 0, then the Second Derivative Test tells us that x = 5 is a local maximum
for f , and hence x = 5 must be the global maximum on the interval [0, 10] (since f = 0 at
the endpoints of the interval). So since y = 10 − x = 5, then the maximum area occurs for a
rectangle whose width and height both are 5 m.

Notice in the above example that the ease of the solution depended on being able to solve
for one variable in terms of the other in the equation 2 x + 2 y = 20. But what if that were not
possible (which is often the case)? In this section we will use a general method, called the
Lagrange multiplier method9 , for solving constrained optimization problems:

Maximize (or minimize) : f ( x, y) (or f ( x, y, z))

given : g( x, y) = c (or g( x, y, z) = c) for some constant c

The equation g( x, y) = c is called the constraint equation, and we say that x and y are con-
strained by g( x, y) = c. Points ( x, y) which are maxima or minima of f ( x, y) with the con-
dition that they satisfy the constraint equation g( x, y) = c are called constrained maximum
or constrained minimum points, respectively. Similar definitions hold for functions of three
variables.
The Lagrange multiplier method for solving such problems can now be stated:
9 Named after the French mathematician Joseph Louis Lagrange (1736–1813).
108 CHAPTER 3. FUNCTIONS OF SEVERAL VARIABLES

Theorem 3.8. Let f ( x, y) and g( x, y) be smooth functions, and suppose that c is a scalar
constant such that ∇ g( x, y) 6= 0 for all ( x, y) that satisfy the equation g( x, y) = c. Then to
solve the constrained optimization problem

Maximize (or minimize) : f ( x, y)

given : g( x, y) = c ,

find the points ( x, y) that solve the equation ∇ f ( x, y) = λ∇ g( x, y) for some constant λ. The
number λ is called the Lagrange multiplier and the point ( x, y) is called critical point of
f ( x, y) constrained by g( x, y) = c.
If there is a constrained maximum or minimum, then it must be such a critical point of
f ( x, y) constrained by g( x, y) = c.

Recall that ∇ g( x, y) is perpendicular to tangent line of the curve g( x, y) = c at the point

( x, y). Therefore, the condition ∇ f ( x, y) = λ∇ g( x, y) simply means that ∇ f ( x, y) is perpen-
dicular to tangent line of the curve g( x, y) = c at the point ( x, y). It should be intuitively
clear that if ∇ f ( x, y) is not perpendicular to the tangent line then slight movement along
the curve g( x, y) = c can increase and decrease the value f ( x, y); in particular, ( x, y) is nei-
ther minimum nor maximum point. A rigorous proof, however, requires use of the Implicit
Function Theorem, which is beyond the scope of this text.10
Note that the theorem only gives a necessary condition for a point to be a constrained
maximum or minimum. That is, if that minimum or maximum is achieved at some point
then this point must be critical, but the theorem says nothing about existence of minimum
and maximum points.
Let us discuss two important cases when existence of minima and maxima is guaranteed;
both of them follow from so called Extreme value theorem which is also beyond the scope of
this text.
Recall that a set is called bounded if it completely lies in a ball of sufficiently large radius.
The following condition guarantees existence of maximum and minimum; a proof is given in
T AYLOR and M ANN.

If the constraint equation g( x, y) = c as above describes a bounded set in R2 , then the

constrained maximum and minimum of f ( x, y) will occur at some points.

If the condition holds then by theorem above the maximum and minimum points have to be
one of the critical. It remains to find all the critical points ( x, y) and compare their values
f ( x, y); the maximum of these values is the global maximum of f ( x, y) with the constraint
g( x, y) = c; analogously, the minimal value if the global minimum.
Let us formulate a more general condition which guarantees existance of minimum, but
not maximum.

10 See T AYLOR and M ANN, § 6.8 for more detail.

 3.7 Lagrange Multipliers 109

If the set described by system g( x, y) = c, f ( x, y) ≤ d is not empty and bounded for some
d then the constrained minimum of f ( x, y) will occur at some points.

Again, to find the minimum we can find all the values f ( x, y) at the critical points ( x, y); the
minimal value is the global minimum of f ( x, y) with the constraint g( x, y) = c.
For example if f ( x, y) = x2 + y2 then the set described by f ( x, y) ≤ d is bounded for any d .
Therefore the above condition guarantees existence of minimum for any constrain g( x, y) = c.
Here is an analogous condition for maximum.
If the set described by system g( x, y) = c, f ( x, y) ≥ d is not empty and bounded for some
d then the constrained maximum of f ( x, y) will occur at some points.

x+
y=
6
xy = 1
x+
y=
2
x+
y=
−
2

Sometimes the answer depend on the hidden constants. For example, consider the func-
tion f ( x, y) = x + y and the constraint x y = 1. The equation ∇ f = λ∇ g from Theorem 3.8 takes
form
(1, 1) = λ( y, x).
Since yx = 1, we have two critical points: (1, 1) with the multiplier 1 and (−1, −1) with the
multiplier −1. Non of these points is maximum, nor minimum; in fact, f can positive and
negative values with arbitrary large absolute value. On the other hand, it might happen that
by the nature of the problem, x and y have to be positive (it is called implicit constraints)
then only one point (1, 1) satisfies is the constrained minimum point and the minimum is 2.
In a general case of that type the maximum or minimum of f ( x, y) will occur either at a
point ( x, y) satisfying ∇ f ( x, y) = λ∇ g( x, y) or at a “boundary” point of the set described by the
hidden constraints.
Similar thing happens in the Example 3.24 the constraint equation 2 x + 2 y = 20 describes
a line in R2 , which by itself is not bounded. However, there are “hidden” constraints, due
110 CHAPTER 3. FUNCTIONS OF SEVERAL VARIABLES

to the nature of the problem, namely 0 ≤ x, y ≤ 10, which cause that line to be restricted
to a line segment in R2 , which is bounded; the the endpoints of that line segment form the
“boundary”.

Example 3.25. For a rectangle whose perimeter is 20 m, use the Lagrange multiplier method
to find the dimensions that will maximize the area.
Solution: As we saw in Example 3.24, with x and y representing the width and height,
respectively, of the rectangle, this problem can be stated as:

Maximize : f ( x, y) = x y
given : g( x, y) = 2 x + 2 y = 20

Then solving the system of scalar and vector equations

g( x, y) = 20,
∇ f ( x, y) = λ∇ g( x, y)

for some λ means solving the system of three scalar equations

g( x, y) = 20,
∂f ∂g
=λ ,
∂x ∂x
∂f ∂g
=λ ,
∂y ∂y

namely:

2 x + 2 y = 20,
y = 2 λ,
x = 2 λ.

The general idea is to solve for λ in the last two equations, then set those expressions equal
(since they both equal λ) to solve for x and y. Doing this we get
y x
= λ = ⇒ x= y ,
2 2
so now substitute either of the expressions for x or y into the constraint equation to solve for
x and y:
20 = g( x, y) = 2 x + 2 y = 2 x + 2 x = 4 x ⇒ x = 5 ⇒ y = 5
There must be a maximum area, since the minimum area is 0 and f (5, 5) = 25 > 0, so the
point (5, 5) that we found (called a constrained critical point) must be the constrained maxi-
mum.
∴ The maximum area occurs for a rectangle whose width and height both are 5 m.
 3.7 Lagrange Multipliers 111

Example 3.26. Find the points on the circle x2 + y2 = 80 which are closest to and farthest
from the point (1, 2).
Solution: The distance d from any point ( x, y) to the point (1, 2) is
q
d = ( x − 1)2 + ( y − 2)2 ,

and minimizing the distance is equivalent to minimizing the square of the distance. Thus
the problem can be stated as:

Maximize (and minimize) : f ( x, y) = ( x − 1)2 + ( y − 2)2

given : g( x, y) = x2 + y2 = 80

Solving ∇ f ( x, y) = λ∇ g( x, y) means solving the following equations:

2( x − 1) = 2λ x ,
2( y − 2) = 2λ y

Note that x 6= 0 since otherwise we would get −2 = 0 in the first equation. Similarly, y 6= 0.
So we can solve both equations for λ as follows:

x−1 y−2
= λ = ⇒ x y − y = x y − 2x ⇒ y = 2x
x y

y
Substituting this into g( x, y) = x2 + y2 = 80 yields 5 x2 = 80, x2 + y2 = 80
so x = ±4. So the two constrained critical points are (4, 8) and (4, 8)
(−4, −8). Since f (4, 8) = 45 and f (−4, −8) = 125, and since there
must be points on the circle closest to and farthest from (1, 2), (1, 2) x
then it must be the case that (4, 8) is the point on the circle clos- 0
est to (1, 2) and (−4, −8) is the farthest from (1, 2) (see Figure
2.7.1).
Notice that since the constraint equation x2 + y2 = 80 describes (−4, −8)
a circle, which is a bounded set in R2 , then we were guaranteed
that the constrained critical points we found were indeed the Figure 3.7.1
constrained maximum and minimum.

The Lagrange multiplier method can be extended to functions of three variables.

Example 3.27.

Maximize (and minimize) : f ( x, y, z) = x + z

given : g( x, y, z) = x2 + y2 + z2 = 1
112 CHAPTER 3. FUNCTIONS OF SEVERAL VARIABLES

Solution: Solve the system of equations g( x, y, z) = 1, ∇ f ( x, y, z) = λ∇ g( x, y, z):

x 2 + y2 + z 2 = 1
1 = 2λ x,
0 = 2λ y,
1 = 2λ z,

The second equation implies λ 6= 0 (otherwise we would have 1 = 0), so we can divide by λ in
the third equation to get y = 0 and we can divide by λ in the first and last equations to get
x = 21λ = z. Substituting these expressions into the constraint equation x2 + y2 + z2 = 1 yields
the constrained critical points

( p1 , 0, p1 ) −1
and ( p −1
, 0, p ).
2 2 2 2

Since
f ( p1 , 0, p1 ) > f ( p
−1 −1
, 0, p ),
2 2 2 2
2 2 2
and since³the constraint
´ equation x + y + z = 1 describes a ³sphere (which
´ is bounded) in
3 1 1 − 1 − 1
R , then p , 0, p is the constrained maximum point and p , 0, p is the constrained
2 2 2 2
minimum point.

Note that solving the equation ∇ f ( x, y) = λ∇ g( x, y) means having to solve a system of two
(possibly nonlinear) equations in three unknowns, which as we have seen before, may not be
possible to do. And the 3-variable case can get even more complicated. All of this somewhat
restricts the usefulness of Lagrange’s method to relatively simple functions. Luckily there
are many numerical methods for solving constrained optimization problems, though we will
not discuss them here.11

Exercises
A
1. Find the constrained maxima and minima of f ( x, y) = 2 x + y given that x2 + y2 = 4.

2. Find the constrained maxima and minima of f ( x, y) = x y given that x2 + 3 y2 = 6.

3. Find the constrained minima of f ( x, y) = x2 + 3 y2 given that x y = 1 and show that there is
no constrained maxima.

4. Find the points on the circle x2 + y2 = 100 which are closest to and farthest from the point
(2, 3).
11 See B AZARAA, S HERALI and S HETTY.
 3.7 Lagrange Multipliers 113

B
5. Find the constrained maxima and minima of f ( x, y, z) = x + y2 + 2 z given that

4 x2 + 9 y2 − 36 z2 = 36.

6. Find the volume of the largest rectangular parallelepiped with edges parallel to the coor-
dinate axis that can be inscribed in the ellipsoid

x 2 y2 z 2
+ + =1 .
a2 b 2 c 2

C
7. Let ( x0 , y0 ) be a minimum point of smooth function f ( x, y) with the constraint g( x, y) ≤ c.
Assume g( x0 , y0 ) = c and ∇ g( x0 , y0 ) 6= 0. Show that ∇ f ( x0 , y0 ) = λ∇ g( x0 , y0 ) for some λ ≤
0. (Hint: Note that ( x0 , y0 ) is also a minimum point of smooth function f ( x, y) with the
constraint g( x, y) = c and the points ( x0 , y0 ) − t∇ g( x0 , y0 ) satisfy the constraint inequality
for small positive t.)
4 Multiple Integrals

4.1 Double Integrals

In single-variable calculus, differentiation and integration are thought of as inverse opera-
tions. For instance, to integrate a function f ( x) it is necessary to find the antiderivative of f ,
that is, another function F ( x) whose derivative is f ( x). Is there a similar way of defining in-
tegration of real-valued functions of two or more variables? The answer is yes, as we will see
shortly. Recall also that the definite integral of a nonnegative function f ( x) ≥ 0 represented
the area “under” the curve y = f ( x). As we will now see, the double integral of a nonnegative
real-valued function f ( x, y) ≥ 0 represents the volume “under” the surface z = f ( x, y).
Let f ( x, y) be a continuous function such that f ( x, y) ≥ 0 for all ( x, y) on the rectangle
R = {( x, y) : a ≤ x ≤ b, c ≤ y ≤ d } in R2 . We will often write this as R = [a, b] × [ c, d ]. For any
number x∗ in the interval [a, b], slice the surface z = f ( x, y) with the plane x = x∗ parallel to
the yz-plane. Then the trace of the surface in that plane is the curve f ( x∗, y), where x∗ is
fixed and only y varies. The area A under that curve (that is, the area of the region between
the curve and the x y-plane) as y varies over the interval [ c, d ] then depends only on the
value of x∗. So using the variable x instead of x∗, let A ( x) be that area (see Figure 4.1.1).
z
z = f (x, y)

c d y
0 A(x)
a
x

b
x R

Figure 4.1.1 The area A(x) varies with x

rd
Then A ( x) = c f ( x, y) d y since we are treating x as fixed, and only y varies. This makes
sense since for a fixed x the function f ( x, y) is a continuous function of y over the interval
[ c, d ], so we know that the area under the curve is the definite integral. The area A ( x) is a
function of x, so by the “slice” or cross-section method from single-variable calculus we know

114
 4.1 Double Integrals 115

that the volume V of the solid under the surface z = f ( x, y) but above the x y-plane over the
rectangle R is the integral over [a, b] of that cross-sectional area A ( x):

 
wb wb wd
V = A ( x) dx =  f ( x, y) d y dx (4.1)
a a c

We will always refer to this volume as “the volume under the surface”. The above expression
uses what are called iterated integrals. First the function f ( x, y) is integrated as a func-
tion of y, treating the variable x as a constant (this is called integrating with respect to y).
That is what occurs in the “inner” integral between the square brackets in equation (4.1).
This is the first iterated integral. Once that integration is performed, the result is then an
expression involving only x, which can then be integrated with respect to x. That is what
occurs in the “outer” integral above (the second iterated integral). The final result is then
a number (the volume). This process of going through two iterations of integrals is called
double integration, and the last expression in equation (4.1) is called a double integral.
Notice that integrating f ( x, y) with respect to y is the inverse operation of taking the
partial derivative of f ( x, y) with respect to y.
Also, we could have taken the area of cross-sections under the surface which were parallel
to the xz-plane, which would then depend only on the variable y, so that the volume would
be
 
wd wb
V =  f ( x, y) dx d y . (4.2)
c a

It turns out that in general1 the order of the iterated integrals does not matter. Also, we will
usually discard the brackets and simply write

wd wb
V = f ( x, y) dx d y , (4.3)
c a

where it is understood that the fact that dx is written before d y means that the function
f ( x, y) is first integrated with respect to x using the “inner” limits of integration a and b,
and then the resulting function is integrated with respect to y using the “outer” limits of
integration c and d . This order of integration can be changed if it is more convenient.

Example 4.1. Find the volume V under the plane z = 8 x + 6 y over the rectangle R = [0, 1] ×
[0, 2].

1 due to Fubini’s Theorem. See Ch. 18 in T AYLOR and M ANN.

116 CHAPTER 4. MULTIPLE INTEGRALS

Solution: We see that f ( x, y) = 8 x + 6 y ≥ 0 for 0 ≤ x ≤ 1 and 0 ≤ y ≤ 2, so:

w2 w1
V = (8 x + 6 y) dx d y
0 0
w2 µ ¯ x =1 ¶
2 ¯
= 4x + 6x y ¯ dy
x =0
0
w2
= (4 + 6 y) d y
0
¯2
= 4 y + 3 y2 ¯
¯
0
= 20

Suppose we had switched the order of integration. We can verify that we still get the same
answer:
w1 w2
V = (8 x + 6 y) d y dx
0 0
w1 µ ¯ y=2 ¶
2¯
= 8x y + 3 y ¯ dx
y=0
0
w1
= (16 x + 12) dx
0
¯1
= 8 x2 + 12 x ¯
¯
0
= 20

Example 4.2. Find the volume V under the surface z = e x+ y over the rectangle R = [2, 3] ×
[1, 2].
Solution: We know that f ( x, y) = e x+ y > 0 for all ( x, y), so
w2 w3
V = e x+ y dx d y
1 2
w2 µ ¯ x =3 ¶
x+ y ¯
= e ¯ dy
x =2
1
w2
= ( e y+3 − e y+2 ) d y
1
¯2
= e y+3 − e y+2 ¯
¯
1
 4.1 Double Integrals 117

= e5 − e4 − ( e4 − e3 ) = e5 − 2 e4 + e3

rb
Recall that for a general function f ( x), the integral a f ( x) dx represents the difference of
the area below the curve y = f ( x) but above the x-axis when f ( x) ≥ 0, and the area above the
curve but below the x-axis when f ( x) ≤ 0. Similarly, the double integral of any continuous
function f ( x, y) represents the difference of the volume below the surface z = f ( x, y) but above
the x y-plane when f ( x, y) ≥ 0, and the volume above the surface but below the x y-plane when
f ( x, y) ≤ 0. Thus, our method of double integration by means of iterated integrals can be
used to evaluate the double integral of any continuous function over a rectangle, regardless
of whether f ( x, y) ≥ 0 or not.

w2πwπ
Example 4.3. Evaluate sin( x + y) dx d y.
0 0
Solution: Note that f ( x, y) = sin( x + y) is both positive and negative over the rectangle [0, π] ×
[0, 2π]. We can still evaluate the double integral:

w2πwπ w2π³ ¯ x =π ´
¯
sin( x + y) dx d y = − cos( x + y) ¯ dy
x =0
0 0 0
w2π
= (− cos( y + π) + cos y) d y
0
¯2π
¯
= − sin( y + π) + sin y ¯ = − sin 3π + sin 2π − (− sin π + sin 0)
0
= 0

Exercises
A
For Exercises 1–4, find the volume under the surface z = f ( x, y) over the rectangle R .

1. f ( x, y) = 4 x y, R = [0, 1] × [0, 1] 2. f ( x, y) = e x+ y , R = [0, 1] × [−1, 1]

3. f ( x, y) = x3 + y2 , R = [0, 1] × [0, 1] 4. f ( x, y) = x4 + x y + y3 , R = [1, 2] × [0, 2]

For Exercises 5–12, evaluate the given double integral.

w1 w2 w1 w2
2
5. (1 − y) x dx d y 6. x( x + y) dx d y
0 1 0 0
118 CHAPTER 4. MULTIPLE INTEGRALS

w2 w1 w2 w1
7. ( x + 2) dx d y 8. x( x y + sin x) dx d y
0 0 −1 −1
w/2w1
π wπ πw/2
2
9. x y cos( x y) dx d y 10. sin x cos( y − π) dx d y
0 0 0 0
w2 w4 w1 w2
11. x y dx d y 12. 1 dx d y
0 1 −1 −1

13. Let M be a constant. Show that

wd wb
M dx d y = M ( d − c)( b − a).
c a
 4.2 Double Integrals Over a General Region 119

4.2 Double Integrals Over a General Region

In the previous section we got an idea of what a double integral over a rectangle represents.
We can now define the double integral of a real-valued function f ( x, y) over more general
regions in R2 .
Suppose that we have a region R in the x y-plane that is bounded on the left by the vertical
line x = a, bounded on the right by the vertical line x = b (where a < b), bounded below by
a curve y = g 1 ( x), and bounded above by a curve y = g 2 ( x), as in Figure 4.2.1(a). We will
assume that g 1 ( x) and g 2 ( x) do not intersect on the open interval (a, b) (they could intersect
at the endpoints x = a and x = b, though).
y y

y = g 2 (x)
d

R
x = h 1 (y)
x = h 2 (y)
y = g 1 (x) c
R
x x
0 a b 0

r br g ( x) r d r h2 ( y)
(a) Vertical slice: a g 2( x) f ( x, y) d y dx (b) Horizontal slice: c h 1 ( y)
f ( x, y) dx d y
1

Figure 4.2.1 Double integral over a nonrectangular region R

Then using the slice method from the previous x section, the double integral of a real-valued
function f ( x, y) over the region R , denoted by f ( x, y) d A , is given by
R
 
x wb gw
2 ( x)

f ( x, y) d A =  f ( x, y) d y dx (4.4)
R a g 1 ( x)

This means that we take vertical slices in the region R between the curves y = g 1 ( x) and
y = g 2 ( x). The symbol d A is sometimes called an area element or infinitesimal, with the A
signifying area. Note that f ( x, y) is first integrated with respect to y, with functions of x as
the limits of integration. This makes sense since the result of the first iterated integral will
have to be a function of x alone, which then allows us to take the second iterated integral
with respect to x.
Similarly, if we have a region R in the x y-plane that is bounded on the left by a curve
x = h 1 ( y), bounded on the right by a curve x = h 2 ( y), bounded below by the horizontal line
120 CHAPTER 4. MULTIPLE INTEGRALS

y = c, and bounded above by the horizontal line y = d (where c < d ), as in Figure 4.2.1(b)
(assuming that h 1 ( y) and h 2 ( y) do not intersect on the open interval ( c, d )), then taking
horizontal slices gives
 
x wd hw
2 ( y)

f ( x, y) d A =  f ( x, y) dx d y (4.5)
R c h 1 ( y)

Notice that these definitions include the case s when the region R is a rectangle. Also, if
f ( x, y) ≥ 0 for all ( x, y) in the region R , then f ( x, y) d A is the volume under the surface
R
z = f ( x, y) over the region R .

Example 4.4. Assume the region R is defined by the inequalities x2 ≤ y and y2 ≤ x. Rewrite
double integral x
f ( x, y) d A,
R

as an iterated integral.

Solution: Let us try to write the iterated integral as in (4.5).

First we need to find the projection of R to y-axis. In other words, we need to find all y
such that ( x, y) belongs to R for some y. The later holds if y4 ≥ y or equivalently if y ∈ [0, 1];
in other words c = 0 and d = 1.
Now for fixed y, such that 0 ≤ y ≤ 1 we need to find all values x such that
p
( x, y) belongs to R . The later holds if y2 ≤ x ≥ y. In other words, h 1 ( y) = y2
p
and h 2 ( y) = y. That is,
p 
x w1 w y
f ( x, y) d A =  f ( x, y) dx d y.
R 0 y2

Note that the region R does not change if you switch x and y. Therefore, the same integral
could be written as p 
x w1 w x
f ( x, y) d A =  f ( x, y) d y dx.
R 0 x2

Example 4.5. Find the volume V under the plane z = 8 x + 6 y over the plane region R defined
by the inequalities 0 ≤ x ≤ 1 and 0 ≤ y ≤ 2 x2 }.
 4.2 Double Integrals Over a General Region 121

y
Solution: The region R is shown in Figure 3.2.2. Using vertical slices we
y = 2x2
get:

x R
V = (8 x + 6 y) d A
x
R
  0 1
w1 2wx2
=  (8 x + 6 y) d y dx Figure 4.2.2
0 0
w1 µ ¯ y=2 x2 ¶
2¯
= 8x y + 3 y ¯ dx
y=0
0
w1
= (16 x3 + 12 x4 ) dx
0
¯1
= 4 x4 + 12
5 x 5¯
¯ = 4 + 12
5 =
32
5 = 6.4
0

y
We get the same answer using horizontal slices (see Figure 3.2.3): 2

x p
x= y/2
V = (8 x + 6 y) d A R
R
  x
w2 w1 0 1
 
=  (8 x + 6 y) dx d y
0
p Figure 4.2.3
y/2

w2 µ ¯ x =1 ¶
4 x2 + 6 x y ¯ p
¯
= dy
x= y/2
0
w2 p w2 p
= (4 + 6 y − (2 y + p6 y y )) d y = (4 + 4 y − 3 2 y3/2 ) d y
2
0 0
p ¯2 p p
= 4 y + 2 y2 − 6 2 5/2 ¯
5 y ¯ = 8+8− 6 2 32
5 = 16 − 48
5 =
32
5 = 6.4
0

Example 4.6. Find the volume V of the solid bounded by the three coordinate planes and
the plane 2 x + y + 4 z = 4.
Solution: The
s solid is shown in Figure 4.2.4(a) with a typical vertical slice. The volume V
is given by f ( x, y) d A , where f ( x, y) = z = 14 (4 − 2 x − y) and the region R , shown in Figure
R
122 CHAPTER 4. MULTIPLE INTEGRALS

y
4

z
y = −2x + 4
(0, 0, 1) 2x + y + 4z = 4

y R
0 (0, 4, 0)
x
x (2, 0, 0) 0 2
(a) (b)

Figure 4.2.4

4.2.4(b), is R = {( x, y) : 0 ≤ x ≤ 2, 0 ≤ y ≤ −2 x + 4}. Using vertical slices in R gives

x
1
V = 4 (4 − 2 x − y) d A
R
w2 −2wx+4
" #
1
= 4 (4 − 2 x − y) d y dx
0 0
w2 µ ¯ y=−2 x+4 ¶
− 81 (4 − 2 x − y)2 ¯
¯
= dx
y=0
0
w2
1 2
= 8 (4 − 2 x) dx
0
¯2
1
(4 − 2 x)3 ¯ = 64 4
¯
= − 48 48 = 3
0

For a general region Rs, which may not be one of the types of regions we have considered so
far, the double integral f ( x, y) d A is defined as follows. Assume that f ( x, y) is a nonnega-
R
tive real-valued function and that R is a bounded region in R2 , so it can be enclosed in some
rectangle [a, b] × [ c, d ]. Then divide that rectangle into a grid of subrectangles. Only consider
the subrectangles that are enclosed completely within the region R , as shown by the shaded
subrectangles in Figure 4.2.5(a). In any such subrectangle [ xi , xi+1 ] × [ y j , y j+1 ], pick a point
( xi∗ , y j∗ ). Then the volume under the surface z = f ( x, y) over that subrectangle is approxi-
mately f ( xi∗ , y j∗ ) ∆ xi ∆ y j , where ∆ xi = xi+1 − xi , ∆ y j = y j+1 − y j , and f ( xi∗ , y j∗ ) is the height and
∆ xi ∆ y j is the base area of a parallelepiped, as shown in Figure 4.2.5(b). Then the total vol-
ume under the surface is approximately the sum of the volumes of all such parallelepipeds,
namely XX
f ( x i∗ , y j∗ ) ∆ x i ∆ y j , (4.6)
j i
 4.2 Double Integrals Over a General Region 123

where the summation occurs over the indices of the subrectangles inside R . If we take
smaller and smaller subrectangles, so that the length of the largest diagonal of the subrect-
angles goes to 0, then the subrectangles begin to fill more and more of the region R , and so
the above sum approaches
s the actual volume under the surface z = f ( x, y) over the region R .
We then define f ( x, y) d A as the limit of that double summation (the limit is taken over all
R
subdivisions of the rectangle [a, b] × [ c, d ] as the largest diagonal of the subrectangles goes
to 0).

y
d

z
z = f (x, y)
∆ yj
∆ xi f (x i∗ , y j∗ )
y j +1

(x i∗ , y j∗ )
y j y j +1 y
yj
0

c xi
x x i +1
0 a (x i∗ , y j∗ ) R
x i x i +1 b x
(a) Subrectangles inside the region R (b) Parallelepiped over a subrectangle,
with volume f ( x i∗ , y j∗ ) ∆ x i ∆ y j

Figure 4.2.5 Double integral over a general region R

A similar definition can be made for a function f ( x, y) that is not necessarily always non-
negative: just replace each mention of volume by the negative volume in the description
above when f ( x, y) < 0. In the case of a region of the type shown in Figure 4.2.1, using
s the def-
inition of the Riemann integral from single-variable calculus, our definition of f ( x, y) d A
R
reduces to a sequence of two iterated integrals.
Finally, the region R does not have to be bounded. We can evaluate improper double inte-
grals (that is, over an unbounded region, or over a region which contains points where the
function f ( x, y) is not defined) as a sequence of iterated improper single-variable integrals.

Example 4.7. Evaluate

w∞ 1/wx2
2 y d y dx.
1 0
124 CHAPTER 4. MULTIPLE INTEGRALS

Solution:
w∞ 1/wx2 w∞µ ¯ y=1/ x2 ¶
2¯
2 y d y dx = y ¯ dx
y=0
1 0 1
w∞ ¯∞
x−4 dx = − 31 x−3 ¯ = 0 − (− 13 ) = 1
¯
= 3
1
1

Exercises
A
For Exercises 1–8, evaluate the given double integral.
w1 w1 wπ wy
2 2. sin x dx d y
1. 24 x y d y dx
p 0 0
0 x

wx
w2 ln w2 w2 y 2
3. 4 x d y dx 4. e y dx d y
1 0 0 0
w/2wy
π w∞w∞ 2
+ y2 )
5. cos x sin y dx d y 6. x ye−( x dx d y
0 0 0 0

w2 wy w1 wx2
7. 1 dx d y 8. 2 d y dx
0 0 0 0

For Exercises 9–10 evaluate x

f ( x, y) d A,
R
where
9. f ( x, y) = x y and R is the intersection of the unit disc x2 + y2 ≤ 1 and the positive quadrant.

10. f ( x, y) = x2 + y and R is the triangle with vertices (0, 0), (2, 0) and (0, 1).

11. Find the volume V of the solid bounded by the three coordinate planes and the plane
x + y + z = 1.

12. Find the volume V of the solid bounded by the three coordinate planes and the plane
3 x + 2 y + 5 z = 6.
B
s
13. Explain why the double integral 1 d A gives the area of the region R . For simplicity,
R
you can assume that R is a region of the type shown in Figure 4.2.1(a).
C
 4.2 Double Integrals Over a General Region 125

14. Prove that the volume of a tetrahedron with mutually perpendic- c

ular adjacent sides of lengths a, b, and c, as in Figure 3.2.6, is abc
6 .
b
(Hint: Mimic Example 4.6, and recall from Section 1.5 how three a
noncollinear points determine a plane.)

15. Show how Exercise 12 can be used to solve Exercise 10. Figure 4.2.6

B
For Exercises 16–17 rewrite double integral
x
f ( x, y) d A,
R

as an iterated integral. (Hint: try to visualize the region.)

16. If the region R is defined by the inequalities 2 y2 ≤ x ≤ 1 + y2 .

17. If the region R is defined by the inequalities x ≤ 2 y ≤ 4 x ≤ 1.

126 CHAPTER 4. MULTIPLE INTEGRALS

4.3 Triple Integrals

Our definition of a double integral of a real-valued function f ( x, y) over a region R in R2 can
be extended to define a triple integral of a real-valued function f ( x, y, z) over a solid S in R3 .
We simply proceed as before: the solid S can be enclosed in some rectangular parallelepiped,
which is then divided into subparallelepipeds. In each subparallelepiped inside S , with sides
t ∆ z, pick a point ( x∗ , y∗ , z∗ ). Then define the triple integral of f ( x, y, z)
of lengths ∆ x, ∆ y and
over S , denoted by f ( x, y, z) dV , by
S
y XXX
f ( x, y, z) dV = lim f ( x∗ , y∗ , z∗ ) ∆ x ∆ y ∆ z , (4.7)
S

where the limit is over all divisions of the rectangular parallelepiped enclosing S into sub-
parallelepipeds whose largest diagonal is going to 0, and the triple summation is over all the
subparallelepipeds inside S . It can be shown that this limit does not depend on the choice
of the rectangular parallelepiped enclosing S . The symbol dV is often called the volume
element.
Physically, what does the triple integral represent? We saw that a double integral could
be thought of as the volume under a two-dimensional surface. It turns out that the triple
integral simply generalizes this idea: it can be thought of as representing the hypervolume
under a three-dimensional hypersurface w = f ( x, y, z) whose graph lies in R4 . In general,
the word “volume” is often used as a general term to signify the same concept for any n-
dimensional object (including length in R1 , area in R2 and volume in R3 ). It may be hard to
get a grasp on the concept of the “volume” of a four-dimensional object, but at least we now
know how to calculate that volume!
In the case where S is a rectangular parallelepiped [ x1 , x2 ] × [ y1 , y2 ] × [ z1 , z2 ], that is, S =
{( x, y, z) : x1 ≤ x ≤ x2 , y1 ≤ y ≤ y2 , z1 ≤ z ≤ z2 }, the triple integral is a sequence of three iterated
integrals, namely
y wz2 wy2 wx2
f ( x, y, z) dV = f ( x, y, z) dx d y dz , (4.8)
S z1 y1 x1

where the order of integration does not matter. This is the simplest case.
A more complicated case is where S is a solid which is bounded below by a surface z =
g 1 ( x, y), bounded above by a surface z = g 2 ( x, y), y is bounded between two curves h 1 ( x) and
h 2 ( x), and x varies between a and b. Then

y wb hw2 ( x) g 2w( x,y)

f ( x, y, z) dV = f ( x, y, z) dz d y dx . (4.9)
S a h 1 ( x) g 1 ( x,y)

Notice in this case that the first iterated integral will result in a function of x and y (since its
limits of integration are functions of x and y), which then leaves you with a double integral of
a type that we learned how to evaluate in Section 3.2. There are, of course, many variations
 4.3 Triple Integrals 127

on this case (for example, changing the roles of the variables x, y, z), so as you can probably
tell, triple integrals can be quite tricky. At this point, just learning how to evaluate a triple
integral, regardless of what it represents, is the most important thing. We will see some
other ways in which triple integrals are used later in the text.

w3 w2 w1
Example 4.8. Evaluate ( x y + z) dx d y dz.
0 0 0
Solution:
w3 w2 w1 w3 w2 µ ¯ x =1 ¶
1 2 ¯
( x y + z) dx d y dz = 2 x y + xz ¯ d y dz
x =0
0 0 0 0 0
w3 w2 ¡ ¢
1
= 2 y+ z d y dz
0 0
w3 µ ¯ y=2 ¶
1 2 ¯
= 4y + yz ¯ dz
y=0
0
w3
= (1 + 2 z) dz
0
¯3
= z + z2 ¯ = 12
¯
0

Example 4.9. Evaluate

w1 1w− x 2−wx− y
( x + y + z) dz d y dx.
0 0 0
Solution:
w1 1w− x 2−wx− y w1 1w− xµ ¯ z=2− x− y ¶
1 2¯
( x + y + z) dz d y dx = ( x + y) z + 2 z ¯ d y dx
z =0
0 0 0 0 0
w1 1w− x ¡ ¢
= ( x + y)(2 − x − y) + 21 (2 − x − y)2 d y dx
0 0
w1 1w− x¡ ¢
= 2 − 21 x2 − x y − 12 y2 d y dx
0 0
w1 µ ¯ y=1− x ¶
1 2 1 2 1 3¯
= 2y − 2 x y − x y − 2 x y − 6 y ¯ dx
y=0
0
w1 ¡ 11 ¢
= 6 − 2 x + 16 x3 dx
0
128 CHAPTER 4. MULTIPLE INTEGRALS

¯1
11 2 1 4¯ 7
= 6 x − x + 24 x ¯ = 8
0

Note that the volume V of a solid in R3 is given by

y
V = 1 dV . (4.10)
S

Since the function being integrated is the constant 1, then the above triple integral reduces
to a double integral of the types that we considered in the previous section if the solid is
bounded above by some surface z = f ( x, y) and bounded below by the x y-plane z = 0. There
are many other possibilities. For example, the solid could be bounded below and above by
surfaces z = g 1 ( x, y) and z = g 2 ( x, y), respectively, with y bounded between two curves h 1 ( x)
and h 2 ( x), and x varies between a and b. Then

y wb hw2 ( x) g 2w( x,y) wb hw2 ( x)

V= 1 dV = 1 dz d y dx = ( g 2 ( x, y) − g 1 ( x, y)) d y dx
S a h 1 ( x) g 1 ( x,y) a h 1 ( x)

just like in equation (4.9). See Exercise 10 for an example.

Exercises
A
For Exercises 1–8, evaluate the given triple integral.
w3 w2 w1 w1 wx wy
1. x yz dx d y dz 2. x yz dz d y dx
0 0 0 0 0 0
wπ wx wx y w1 wz wy 2
3. x2 sin z dz d y dx 4. ze y dx d y dz
0 0 0 0 0 0
we wy 1/
wy w2 wy wz2
2

2
5. x z dx dz d y 6. yz dx dz d y
1 0 0 1 0 0
w2 w4 w3 w1 1w− x 1−wx− y
7. 1 dx d y dz 8. 1 dz d y dx
1 2 0 0 0 0

9. Let M be a constant. Show that

wz2 wy2 wx2
M dx d y dz = M ( z2 − z1 )( y2 − y1 )( x2 − x1 ).
z1 y1 x1

B
 4.3 Triple Integrals 129

10. Find the volume V of the solid S bounded by the three coordinate planes, bounded above
by the plane x + y + z = 2, and bounded below by the plane z = x + y.

11. Let S be the solid defined by the inequalities x2 − 1 ≤ y ≤ 1 − z2 . Rewrite the triple
integral y
f ( x, y, z) dV ,
S

as an iterated integral.

C
12. Show that
wb wz wy wb
( b − x )2
f ( x) dx d y dz = 2 f ( x) dx.
a a a a

(Hint: Think of how changing the order of integration in the triple integral changes the
limits of integration.)
130 CHAPTER 4. MULTIPLE INTEGRALS

4.4 Numerical Approximation of Multiple Integrals

As you have seen, calculating multiple integrals is tricky even for simple functions and
regions. For complicated functions, it may not be possible to evaluate one of the iterated in-
tegrals in a simple closed form. Luckily there are numerical methods for approximating the
value of a multiple integral. The method we will discuss is called the Monte Carlo method.
The idea behind it is based on the concept of the average value of a function, which you
learned in single-variable calculus. Recall that for a continuous function f ( x), the average
value f¯ of f over an interval [a, b] is defined as

1 w
b
f¯ = f ( x) dx . (4.11)
b−a a

The quantity b − a is the length of the interval [a, b], which can be thought of as the
“volume” of the interval. Applying the same reasoning to functions of two or three variables,
we define the average value of f ( x, y) over a region R to be

1 x
f¯ = f ( x, y) d A , (4.12)
A (R )
R

where A (R ) is the area of the region R , and we define the average value of f ( x, y, z) over a
solid S to be
1 y
f¯ = f ( x, y, z) dV , (4.13)
V (S )
S

where V (S ) is the volume of the solid S . Thus, for example, we have

x
f ( x, y) d A = A (R ) f¯ . (4.14)
R

The average value of f ( x, y) over R can be thought of as representing the sum of all the
values of f divided by the number of points in R . However, we can not take the sum literary
since there are an infinite number of points in any region (in fact, uncounably many — one
can not enumerate them by natural numbers). But what if we took a very large number N
of random points in the region R (which can be generated by a computer) and then took the
average of the values of f for those points, and used that average as the value of f¯? This is
exactly what the Monte Carlo method does. So in formula (4.14) the approximation we get
is s
x f 2 − ( f¯)2
f ( x, y) d A ≈ A (R ) f¯ ± A (R ) , (4.15)
N
R

where
1 X N 1 X N
f¯ = f ( xi , yi ) and f2 = ( f ( xi , yi ))2 , (4.16)
N i =1 N i =1
 4.4 Numerical Approximation of Multiple Integrals 131

with the sums taken over the N random points ( x1 , y1 ), . . ., ( xN , yN ). The ± “error term” in
formula (4.15) does not really provide hard bounds on the approximation. It represents a
single standard deviation from the expected value of the integral. That is, it provides a likely
bound on the error. Due to its use of random points, the Monte Carlo method is an example
of a probabilistic method (as opposed to deterministic methods such as Newton’s method,
which use a specific formula for generating points).
For example, we can use formula (4.15) to approximate the volume V under the plane
z = 8 x + 6 y over the rectangle R = [0, 1] × [0, 2]. In Example 4.1 in Section 3.1, we showed
that the actual volume is 20. Below is a code listing (montecarlo.java) for a Java program
that calculates the volume, using a number of points N that is passed on the command line
as a parameter.

//Program to approximate the double integral of f(x,y)=8x+6y

//over the rectangle [0,1]x[0,2].
public class montecarlo {
public static void main(String[] args) {
//Get the number N of random points as a command-line parameter
int N = Integer.parseInt(args[0]);
double x = 0; //x-coordinate of a random point
double y = 0; //y-coordinate of a random point
double f = 0.0; //Value of f at a random point
double mf = 0.0; //Mean of the values of f
double mf2 = 0.0; //Mean of the values of f^2
for (int i=0;i<N;i++) { //Get the random coordinates
x = Math.random(); //x is between 0 and 1
y = 2 * Math.random(); //y is between 0 and 2
f = 8*x + 6*y; //Value of the function
mf = mf + f; //Add to the sum of the f values
mf2 = mf2 + f*f; //Add to the sum of the f^2 values
}
mf = mf/N; //Compute the mean of the f values
mf2 = mf2/N; //Compute the mean of the f^2 values
System.out.println("N = " + N + ": integral = " + vol()*mf + " +/- "
+ vol()*Math.sqrt((mf2 - Math.pow(mf,2))/N)); //Print the result
}
//The volume of the rectangle [0,1]x[0,2]
public static double vol() {
return 1*2;
}
}

Listing 4.1 Program listing for montecarlo.java

The results of running this program with various numbers of random points (for instance,
java montecarlo 100) are shown below:
132 CHAPTER 4. MULTIPLE INTEGRALS

N = 10: 19.36543087722646 +/- 2.7346060413546147

N = 100: 21.334419561385353 +/- 0.7547037194998519
N = 1000: 19.807662237526227 +/- 0.26701709691370235
N = 10000: 20.080975812043256 +/- 0.08378816229769506
N = 100000: 20.009403854556716 +/- 0.026346782289498317
N = 1000000: 20.000866994982314 +/- 0.008321168748642816

As you can see, the approximation is fairly good. As N → ∞, it can be shown p that the
Monte Carlo approximation converges to the actual volume (on the order of O ( N ), in com-
putational complexity terminology).
In the above example the region R was a rectangle. To use the Monte Carlo method for
a nonrectangular (bounded) region R , only a slight modification is needed. Pick a rectangle
R̃ that encloses R , and generate random points in that rectangle as before. Then use those
points in the calculation of f¯ only if they are inside R . There is no need to calculate the area
of R for formula (4.15) in this case, since the exclusion of points not inside R allows you to
use the area of the rectangle R̃ instead, similar to before.
For instance, in Example 4.5 we showed that the volume under the surface z = 8 x + 6 y
over the nonrectangular region R = {( x, y) : 0 ≤ x ≤ 1, 0 ≤ y ≤ 2 x2 } is 6.4. Since the rectangle
R̃ = [0, 1] × [0, 2] contains R , we can use the same program as before, with the only change
being a check to see if y < 2 x2 for a random point ( x, y) in [0, 1] × [0, 2]. Listing 4.2 below
contains the code (montecarlo2.java):

//Program to approximate the double integral of f(x,y)=8x+6y over the

//region bounded by x=0, x=1, y=0, and y=2x^2
public class montecarlo2 {
public static void main(String[] args) {
//Get the number N of random points as a command-line parameter
int N = Integer.parseInt(args[0]);
double x = 0; //x-coordinate of a random point
double y = 0; //y-coordinate of a random point
double f = 0.0; //Value of f at a random point
double mf = 0.0; //Mean of the values of f
double mf2 = 0.0; //Mean of the values of f^2
for (int i=0;i<N;i++) { //Get the random coordinates
x = Math.random(); //x is between 0 and 1
y = 2 * Math.random(); //y is between 0 and 2
if (y < 2*Math.pow(x,2)) { //The point is in the region
f = 8*x + 6*y; //Value of the function
mf = mf + f; //Add to the sum of the f values
mf2 = mf2 + f*f; //Add to the sum of the f^2 values
}
}
mf = mf/N; //Compute the mean of the f values
mf2 = mf2/N; //Compute the mean of the f^2 values
System.out.println("N = " + N + ": integral = " + vol()*mf +
 4.4 Numerical Approximation of Multiple Integrals 133

" +/- " + vol()*Math.sqrt((mf2 - Math.pow(mf,2))/N));

}
//The volume of the rectangle [0,1]x[0,2]
public static double vol() {
return 1*2;
}
}

Listing 4.2 Program listing for montecarlo2.java

The results of running the program with various numbers of random points (for instance,
java montecarlo2 1000) are shown below:

N = 10: integral = 6.95747529014894 +/- 2.9185131565120592

N = 100: integral = 6.3149056229650355 +/- 0.9549009662159909
N = 1000: integral = 6.477032813858756 +/- 0.31916837260973624
N = 10000: integral = 6.349975080015089 +/- 0.10040086346895105
N = 100000: integral = 6.440184132811864 +/- 0.03200476870881392
N = 1000000: integral = 6.417050897922222 +/- 0.01009454409789472

To use the Monte Carlo method to evaluate triple integrals, you will need to generate
random triples ( x, y, z) in a parallelepiped, instead of random pairs ( x, y) in a rectangle, and
use the volume of the parallelepiped instead of the area of a rectangle in formula (4.15) (see
Exercise 2). For a more detailed discussion of numerical integration methods, see P RESS et
al.

Exercises
C
1. Write a program that uses the Monte Carlo method to approximate the double integral
x
e x y d A,
R

where R = [0, 1] × [0, 1]. Show the program output for N = 10, 100, 1000, 10000, 100000
and 1000000 random points.

2. Write a program that uses the Monte Carlo method to approximate the triple integral
y
e x yz dV ,
S

where S = [0, 1] × [0, 1] × [0, 1]. Show the program output for N = 10, 100, 1000, 10000,
100000 and 1000000 random points.
134 CHAPTER 4. MULTIPLE INTEGRALS

3. Repeat Exercise 1 with the region R = {( x, y) : −1 ≤ x ≤ 1, 0 ≤ y ≤ x2 }.

4. Repeat Exercise 2 with the solid S = {( x, y, z) : 0 ≤ x ≤ 1, 0 ≤ y ≤ 1, 0 ≤ z ≤ 1 − x − y}.

5. Use the Monte Carlo method to approximate the volume of a sphere of radius 1.
x2 y2 2
6. Use the Monte Carlo method to approximate the volume of the ellipsoid 9 + 4 + z1 = 1.
 4.5 Change of Variables in Multiple Integrals 135

4.5 Change of Variables in Multiple Integrals

Given the difficulty of evaluating multiple integrals, the reader may be wondering if it is
possible to simplify those integrals using a suitable substitution for the variables. The an-
swer is yes, though it is a bit more complicated than the substitution method which you
learned in single-variable calculus.
Recall that if you are given, for example, the definite integral

w2 p
x3 x2 − 1 dx ,
1

then you would make the substitution

u = x2 − 1 ⇒ x2 = u + 1
du = 2 x dx

which changes the limits of integration

x=1 ⇒ u=0
x=2 ⇒ u=3

so that we get

w2 p w2 p
x3 x2 − 1 dx = 1 2
2 x · 2x x2 − 1 dx
1 1
w3 p
1
= 2 ( u + 1) u du
0
w3 ³ ´
= 1
2 u3/2 + u1/2 du , which can be integrated to give
0
p
14 3
= 5 .

Let us take a different look at what happened when we did that substitution, which will give
some motivation for how substitution works in multiple integrals. First, we let u = x2 − 1.
On the interval of integration [1, 2], the function x 7→ x2 − 1 is strictly increasing (and maps
[1, 2] onto [0, 3]) and hence has an inverse function (defined on the interval [0, 3]). That is,
on [0, 3] we can define x as a function of u, namely
p
x = g( u) = u + 1 .
p
Then substituting that expression for x into the function f ( x) = x3 x2 − 1 gives
p
f ( x) = f ( g( u)) = ( u + 1)3/2 u ,
136 CHAPTER 4. MULTIPLE INTEGRALS

and we see that

dx
= g ′ ( u) ⇒ dx = g ′ ( u) du
du
dx = 21 ( u + 1)−1/2 du ,

so since

g(0) = 1 ⇒ 0 = g−1 (1)

g(3) = 2 ⇒ 3 = g−1 (2)

then performing the substitution as we did earlier gives

w2 w2 p
f ( x) dx = x3 x2 − 1 dx
1 1
w3 p
1
= 2 ( u + 1) u du , which can be written as
0
w3 p
= ( u + 1)3/2 u · 21 ( u + 1)−1/2 du , which means
0

w2 g−w1 (2)
f ( x) dx = f ( g( u)) g ′ ( u) du .
1 g−1 (1)

In general, if x = g( u) is a one-to-one, differentiable function from an interval [ c, d ] (which

you can think of as being on the “ u-axis”) onto an interval [a, b] (on the x-axis), which means
that g ′ ( u) 6= 0 on the interval ( c, d ), so that a = g( c) and b = g( d ), then c = g−1 (a) and d =
g−1 ( b), and
wb g−w1 ( b)
f ( x) dx = f ( g( u)) g ′ ( u) du . (4.17)
a g −1 ( a )

This is called the change of variable formula for integrals of single-variable functions, and it
is what you were implicitly using when doing integration by substitution.
This formula turns out to be a special case of a more general formula which can be used
to evaluate multiple integrals. We will state the formulas for double and triple integrals
involving real-valued functions of two and three variables, respectively. We will assume
that all the functions involved are continuously differentiable and that the regions and solids
involved all have “reasonable” boundaries. The proof of the following theorem is beyond the
scope of the text.2
2 See T AYLOR and M ANN, § 15.32 and § 15.62 for all the details.
 4.5 Change of Variables in Multiple Integrals 137

Theorem 4.1. Change of Variables Formula for Multiple Integrals

Let x = x( u, v) and y = y( u, v) define a one-to-one mapping of a region R ′ in the uv-plane onto
a region R in the x y-plane such that the determinant
¯ ¯
¯ ∂x ∂x ¯
¯ ¯
¯ ¯
J ( u, v) = ¯¯ ∂ u ∂v ¯¯ (4.18)
¯ ∂y ∂y¯
¯ ¯
∂ u ∂v

is never 0 in R ′ . Then
x x
f ( x, y) d A ( x, y) = f ( x( u, v), y( u, v)) | J ( u, v) | d A ( u, v) . (4.19)
R R′

We use the notation d A ( x, y) and d A ( u, v) to denote the area element in the ( x, y) and ( u, v)
coordinates, respectively.
Similarly, if x = x( u, v, w), y = y( u, v, w) and z = z( u, v, w) define a one-to-one mapping of
a solid S ′ in uvw-space onto a solid S in x yz-space such that the determinant
¯ ¯
¯ ∂x ∂x ∂x ¯
¯ ¯
¯ ∂ u ∂v ∂w ¯
¯ ¯
¯ ∂y ∂y ∂y ¯
J ( u, v, w) = ¯
¯ ¯
¯ (4.20)
¯ ∂ u ∂v ∂w ¯
¯ ∂z ∂z ∂z ¯
¯ ¯
¯ ¯
∂ u ∂v ∂w

is never 0 in S ′ , then
y y
f ( x, y, z) dV ( x, y, z) = f ( x( u, v, w), y( u, v, w), z( u, v, w)) | J ( u, v, w) | dV ( u, v, w) . (4.21)
S S′

The determinant J ( u, v) in formula (4.18) is called the Jacobian of x and y with respect
to u and v, and is sometimes written as
∂( x, y)
J ( u, v) = . (4.22)
∂( u, v)
Similarly, the Jacobian J ( u, v, w) of three variables is sometimes written as
∂( x, y, z)
J ( u, v, w) = . (4.23)
∂( u, v, w)
Notice that formula (4.19) is saying that d A ( x, y) = | J ( u, v) | d A ( u, v), which you can think of
as a two-variable version of the relation dx = g ′ ( u) du in the single-variable case.
The following example shows how the change of variables formula is used.
x x− y
Example 4.10. Evaluate e x+ y d A , where R = {( x, y) : x ≥ 0, y ≥ 0, x + y ≤ 1}.
R
138 CHAPTER 4. MULTIPLE INTEGRALS

Solution: First, note that evaluating this double integral without using substitution is prob-
ably impossible, at least in a closed form. By looking at the numerator and denominator of
the exponent of e, we will try the substitution u = x − y and v = x + y. To use the change of
variables formula (4.19), we need to write both x and y in terms of u and v. So solving for
x and y gives x = 21 ( u + v) and y = 12 (v − u). In Figure 4.5.1 below, we see how the mapping
x = x( u, v) = 21 ( u + v), y = y( u, v) = 21 (v − u) maps the region R ′ onto R in a one-to-one manner.
y v
1
x= 2 (u + v) 1
1
1
y= 2 (v − u) R′
x+ y =1
u = −v u=v
R x u
0 1 −1 0 1

Figure 4.5.1 The regions R and R ′

Now we see that

¯ ¯
¯ ∂x ∂ x ¯¯ ¯ ¯
¯
¯ ¯ ¯ 1 1 ¯ 1
¯ ¯
¯1¯ 1
J ( u, v) = ¯¯ ∂ u ∂v ¯ = ¯¯ 2 2 ¯
¯ = ⇒ | J ( u, v) | = ¯¯ ¯¯ = ,
¯ ∂y ∂ y ¯¯ ¯ −1 1 ¯ 2 2 2
¯ ¯ 2 2
∂u ∂v

so using horizontal slices in R ′ , we have

x x− y
x
e x+ y d A = f ( x( u, v), y( u, v)) | J ( u, v) | d A
R R′
w1 wv u
1
= ev 2 du dv
0 −v
w1 ³ u ¯u=v ´
v v ¯
= 2 e ¯ u=−v dv
0
w1
v −1
= 2 ( e − e ) dv
0
µ ¶
v2 ¯1 1 1 e2 − 1
( e − e −1 ) ¯ =
¯
= e− =
4 0 4 e 4e

The change of variables formula can be used to evaluate double integrals in polar coordi-
nates. Letting
x = x( r, θ ) = r cos θ and y = y( r, θ ) = r sin θ ,
 4.5 Change of Variables in Multiple Integrals 139

we have
¯ ¯
¯ ∂x ∂ x ¯¯ ¯ ¯
¯
∂θ ¯¯ = ¯¯ cos θ − r sin θ ¯¯
¯ ¯
J ( u, v) = ¯¯ ∂ r ¯ = r cos2 θ + r sin2 θ = r ⇒ | J ( u, v) | = | r | = r ,
¯∂y ∂ y ¯¯ ¯ sin θ r cos θ ¯
¯ ¯
∂r ∂θ

so we have the following formula:

Double Integral in Polar Coordinates
x x
f ( x, y) dx d y = f ( r cos θ , r sin θ ) r dr d θ , (4.24)
R R′

where the mapping x = r cos θ , y = r sin θ maps the region R ′ in the r θ -plane onto the
region R in the x y-plane in a one-to-one manner.

Example 4.11. Find the volume V inside the paraboloid z = x2 + y2 for 0 ≤ z ≤ 1.

Solution: Using vertical slices, we see that z

x 2 + y2 = 1
x x 1
V = (1 − z) d A = (1 − ( x2 + y2 )) d A ,
R R

where R = {( x, y) : x2 + y2 ≤ 1} is the unit disk in R2 (see

Figure 3.5.2). In polar coordinates ( r, θ ) we know that
x2 + y2 = r 2 and that the unit disk R is the set R ′ = {( r, θ ) :
0 ≤ r ≤ 1, 0 ≤ θ ≤ 2π}. Thus, y
0
w2πw1 x
V = (1 − r 2 ) r dr d θ Figure 4.5.2 z = x 2 + y2
0 0
wπw1
2
= ( r − r 3 ) dr d θ
0 0
w2πµ ¯ ¶
4 ¯ r =1
r2
= 2 − r4 ¯ dθ
r =0
0
w2π
1
= 4 dθ
0
π
=
2

p
Example 4.12. Find the volume V inside the cone z = x2 + y2 for 0 ≤ z ≤ 1.
140 CHAPTER 4. MULTIPLE INTEGRALS

Solution: Using vertical slices, we see that z x 2 + y2 = 1

x xµ q ¶ 1
V = (1 − z) d A = 2 2
1− x + y dA ,
R R

where R = {( x, y) : x2 + y2 ≤ 1} is the unit disk in R2

(see Figure 3.5.3). In polar coordinates ( r, θ ) we know y
p 0
that x2 + y2 = r and that the unit disk R is the set
x
R ′ = {( r, θ ) : 0 ≤ r ≤ 1, 0 ≤ θ ≤ 2π}. Thus, p
Figure 4.5.3 z= x 2 + y2
w2πw1
V = (1 − r ) r dr d θ
0 0
wπw1
2
= ( r − r 2 ) dr d θ
0 0
w2πµ ¯ ¶
3 ¯ r =1
r2
= 2 − r3 ¯ dθ
r =0
0
w2π
1
= 6 dθ
0
π
=
3

In a similar fashion, it can be shown (see Exercises 5–6) that triple integrals in cylindrical
and spherical coordinates take the following forms:
Triple Integral in Cylindrical Coordinates
y y
f ( x, y, z) dx d y dz = f ( r cos θ , r sin θ , z) r dr d θ dz , (4.25)
S S′

where the mapping x = r cos θ , y = r sin θ , z = z maps the solid S ′ in r θ z-space onto the
solid S in x yz-space in a one-to-one manner.

Triple Integral in Spherical Coordinates

y y
f ( x, y, z) dx d y dz = f (ρ sin φ cos θ , ρ sin φ sin θ , ρ cos φ) ρ 2 sin φ d ρ d φ d θ ,
S S′
(4.26)
where the mapping x = ρ sin φ cos θ , y = ρ sin φ sin θ , z = ρ cos φ maps the solid S ′ in ρφθ -
space onto the solid S in x yz-space in a one-to-one manner.

Example 4.13. For a > 0, find the volume V inside the sphere S = x2 + y2 + z2 = a2 .
 4.5 Change of Variables in Multiple Integrals 141

Solution: We see that S is the set ρ = a in spherical coordinates, so

y w2πwπ wa
V = 1 dV = 1 ρ 2 sin φ d ρ d φ d θ
S 0 0 0
w w ρ 3 ¯ρ = a
2π π µ ¶ w2πwπ a3
¯
= ¯ sin φ d φ d θ = sin φ d φ d θ
3 ρ =0 3
0 0 0 0
w2πµ a3 ¯ φ= π ¶ w2π 2a3 4π a 3
¯
= − cos φ ¯ dθ = dθ = .
3 φ= 0 3 3
0 0

Exercises
A
1. Find the volume V inside the paraboloid z = x2 + y2 for 0 ≤ z ≤ 4.
p
2. Find the volume V inside the cone z = x2 + y2 for 0 ≤ z ≤ 3.

B
3. Find the volume V of the solid inside both x2 + y2 + z2 = 4 and x2 + y2 = 1.
p
4. Find the volume V inside both the sphere x2 + y2 + z2 = 1 and the cone z = x 2 + y2 .

5. Prove formula (4.25). 6. Prove formula (4.26).

s ¡ x+ y ¢ ¡ x− y ¢
7. Evaluate sin 2 cos 2 d A , where R is the triangle with vertices (0, 0), (2, 0) and
R
(1, 1). (Hint: Use the change of variables u = ( x + y)/2, v = ( x − y)/2.)

8. Find the volume of the solid bounded by z = x2 + y2 and z2 = 4( x2 + y2 ).

x2 y2
9. Find the volume inside the elliptic cylinder a2
+ b2 ≤ 1 for 0 ≤ z ≤ 2.

C
2 y2 2
10. Show that the volume inside the ellipsoid ax2 + b2 + zc2 = 1 is 4π3abc . (Hint: Use the change
of variables x = au, y = bv, z = cw, then consider Example 4.13.)

11. Show that

w +∞
+∞ w w +∞
+∞ w
f ( x, y) dx d y = f ( x + y, x + 2 y) dx d y
−∞ −∞ −∞ −∞

For any smooth function f ( x, y) which vanishes outside of a bounded region in the plane.
142 CHAPTER 4. MULTIPLE INTEGRALS

4.6 Application: Center of Mass

y
Recall from single-variable calculus that for a region
R = {( x, y) : a ≤ x ≤ b, 0 ≤ y ≤ f ( x)} in R2 that represents y = f (x)
a thin, flat plate (see Figure 3.6.1), where f ( x) is a con-
tinuous function on [a, b], the center of mass of R has R
coordinates ( x̄, ȳ) given by ( x̄, ȳ) x
0 a b
My Mx
x̄ = and ȳ = ,
M M Figure 4.6.1 Center of mass of R

where

wb ( f ( x))2 wb wb
Mx = dx , My = x f ( x) dx , M= f ( x) dx , (4.27)
a
2 a a

assuming that R has uniform density, i.e the mass of R is uniformly distributed over the
region. In this case the area M of the region is considered the mass of R (the density is
constant, and taken as 1 for simplicity).
In the general case where the density of a region (or lamina) R is a continuous function
δ = δ( x, y) of the coordinates ( x, y) of points inside R (where R can be any region in R2 ) the
coordinates ( x̄, ȳ) of the center of mass of R are given by

My Mx
x̄ = and ȳ = , (4.28)
M M
where x x x
My = xδ( x, y) d A , Mx = yδ( x, y) d A , M= δ( x, y) d A , (4.29)
R R R

The quantities M x and M y are called the moments (or first moments) of the region R about
the x-axis and y-axis, respectively. The quantity M is the mass of the region R . To see this,
think of taking a small rectangle inside R with dimensions ∆ x and ∆ y close to 0. The mass
of that rectangle is approximately δ( x∗ , y∗ )∆ x ∆ y, for some point ( x∗ , y∗ ) in that rectangle.
Then the mass of R is the limit of the sums of the masses of all such rectangless inside R as
the diagonals of the rectangles approach 0, which is the double integral δ( x, y) d A .
R
Note that the formulas in (4.27) represent a special case when δ( x, y) = 1 throughout R in
the formulas in (4.29).

Example 4.14. Find the center of mass of the region R = {( x, y) : 0 ≤ x ≤ 1, 0 ≤ y ≤ 2 x2 }, if the

density function at ( x, y) is δ( x, y) = x + y.
 4.6 Application: Center of Mass 143

y
Solution: The region R is shown in Figure 3.6.2. We have
y = 2x2
x
M = δ( x, y) d A
R R

w1 2wx2 x
= ( x + y) d y dx 0 1
0 0
w1
à ¯ y=2 x2 ! Figure 4.6.2
y2 ¯¯
= xy+ dx
2 ¯ y=0
0
w1
= (2 x3 + 2 x4 ) dx
0
¯1
x4 2 x5 ¯¯ 9
= + ¯ =
2 5 0 10

and

x x
Mx = yδ( x, y) d A My = xδ( x, y) d A
R R
w1 2wx2 w1 2wx2
= y( x + y) d y dx = x( x + y) d y dx
0 0 0 0
w1 w1
à ¯ y=2 x2 ! à ¯ y=2 x2 !
x y2 y3 ¯¯ 2 x y2 ¯¯
= + dx = x y+ dx
2 3 ¯ y=0 2 ¯ y=0
0 0
w1 8x 6 w1
= (2 x5 +
) dx = (2 x4 + 2 x5 ) dx
3
0 0
¯1 ¯1
x6 8 x7 ¯¯ 5 2 x5 x6 ¯¯ 11
= + ¯ = = + ¯ = ,
3 21 0 7 5 3 0 15

so the center of mass ( x̄, ȳ) is given by

My 11/15 22 Mx 5/7 50
x̄ = = = , ȳ = = = .
M 9/10 27 M 9/10 63
Note how this center of mass is a little further towards the upper corner of the region R
¡ ¢
than when the density is uniform (use the formulas in (4.27) to show that ( x̄, ȳ) = 43 , 53 in
that case). This makes sense since the density function δ( x, y) = x + y increases as ( x, y)
approaches that upper corner, where there is quite a bit of area.
144 CHAPTER 4. MULTIPLE INTEGRALS

In the special case where the density function δ( x, y) is a constant function on the region
R , the center of mass ( x̄, ȳ) is called the centroid of R .
The formulas for the center of mass of a region in R2 can be generalized to a solid S in R3 .
Let S be a solid with a continuous mass density function δ( x, y, z) at any point ( x, y, z) in S .
Then the center of mass of S has coordinates ( x̄, ȳ, z̄), where

M yz M xz Mx y
x̄ = , ȳ = , z̄ = , (4.30)
M M M

where
y y y
M yz = xδ( x, y, z) dV , M xz = yδ( x, y, z) dV , Mx y = zδ( x, y, z) dV , (4.31)
S S S
y
M = δ( x, y, z) dV . (4.32)
S

In this case, M yz , M xz and M x y are called the moments (or first moments) of S around the
yz-plane, xz-plane and x y-plane, respectively. Also, M is the mass of S .

Example 4.15. Find the center of mass of the solid S = {( x, y, z) : z ≥ 0, x2 + y2 + z2 ≤ a2 }, if

the density function at ( x, y, z) is δ( x, y, z) = 1.
z
Solution: The solid S is just the upper hemisphere inside the sphere a
of radius a centered at the origin (see Figure 3.6.3). So since the
density function is a constant and S is symmetric about the z-axis,
( x̄, ȳ, z̄)
then it is clear that x̄ = 0 and ȳ = 0, so we need only find z̄. We have y
y y 0 a
M = δ( x, y, z) dV = 1 dV = V olume(S ).
S S x
Figure 4.6.3
But since the volume of S is half the volume of the sphere of radius
3 3
a, which we know by Example 4.13 is 4π3a , then M = 2π3a . And
y
Mx y = zδ( x, y, z) dV
S
y
= z dV , which in spherical coordinates is
S
w2π πw/2wa
= (ρ cos φ) ρ 2 sin φ d ρ d φ d θ
0 0 0

w2π πw/2 wa
à !
3
= sin φ cos φ ρ dρ dφ dθ
0 0 0
 4.6 Application: Center of Mass 145

w2π πw/2
a4
= 4 sin φ cos φ d φ d θ
0 0

w2π πw/2
a4
Mx y = 8 sin 2φ d φ d θ (since sin 2φ = 2 sin φ cos φ)
0 0

w2πµ ¯φ=π/2 ¶
a4 ¯
= − 16 cos 2φ ¯ dθ
φ= 0
0

w2π
a4
= 8 dθ
0
π a4
= ,
4
so
π a4
Mx y 4 3a
z̄ = = = .
M 2πa3 8
3
¡ ¢
Thus, the center of mass of S is ( x̄, ȳ, z̄) = 0, 0, 38a .

Exercises
A
For Exercises 1–5, find the center of mass of the region R with the given density function
δ( x, y).

1. R = {( x, y) : 0 ≤ x ≤ 2, 0 ≤ y ≤ 4 }, δ( x, y) = 2 y

2. R = {( x, y) : 0 ≤ x ≤ 1, 0 ≤ y ≤ x2 }, δ( x, y) = x + y

3. R = {( x, y) : y ≥ 0, x2 + y2 ≤ a2 }, δ( x, y) = 1
p
4. R = {( x, y) : y ≥ 0, x ≥ 0, 1 ≤ x2 + y2 ≤ 4 }, δ( x, y) = x 2 + y2

5. R = {( x, y) : y ≥ 0, x2 + y2 ≤ 1 }, δ( x, y) = y

B
For Exercises 6–10, find the center of mass of the solid S with the given density function
δ( x, y, z).

6. S = {( x, y, z) : 0 ≤ x ≤ 1, 0 ≤ y ≤ 1, 0 ≤ z ≤ 1 }, δ( x, y, z) = x yz

7. S = {( x, y, z) : z ≥ 0, x2 + y2 + z2 ≤ a2 }, δ( x, y, z) = x2 + y2 + z2
146 CHAPTER 4. MULTIPLE INTEGRALS

8. S = {( x, y, z) : x ≥ 0, y ≥ 0, z ≥ 0, x2 + y2 + z2 ≤ a2 }, δ( x, y, z) = 1

9. S = {( x, y, z) : 0 ≤ x ≤ 1, 0 ≤ y ≤ 1, 0 ≤ z ≤ 1 }, δ( x, y, z) = x2 + y2 + z2

10. S = {( x, y, z) : 0 ≤ x ≤ 1, 0 ≤ y ≤ 1, 0 ≤ z ≤ 1 − x − y}, δ( x, y, z) = 1

C
11. Let F be a figure in the upper half-plane; denote as ( x0 , y0 ) its center of mass and as A
its the area. Show that
2π A y0
is the volume of the body of revolution obtained by rotating F around x-axis.
 4.7 Application: Probability and Expected Value 147

4.7 Application: Probability and Expected Value

In this section we will briefly discuss some applications of multiple integrals in the field of
probability theory. In particular we will see ways in which multiple integrals can be used to
calculate probabilities and expected values.

Probability
Suppose that you have a standard six-sided (fair) die, and you let a variable X represent
the value rolled. Then the probability of rolling a 3, written as P ( X = 3), is 61 , since there
are six sides on the die and each one is equally likely to be rolled, and hence in particular
the 3 has a one out of six chance of being rolled. Likewise the probability of rolling at most a
3, written as P ( X ≤ 3), is 36 = 21 , since of the six numbers on the die, there are three equally
likely numbers (1, 2, and 3) that are less than or equal to 3. Note that P ( X ≤ 3) = P ( X =
1) + P ( X = 2) + P ( X = 3). We call X a discrete random variable on the sample space (or
probability space) Ω consisting of all possible outcomes. In our case, Ω = {1, 2, 3, 4, 5, 6}. An
event A is a subset of the sample space. For example, in the case of the die, the event X ≤ 3
is the set {1, 2, 3}.
Now let X be a variable representing a random real number in the interval (0, 1). Note
that for any real number x in (0, 1), it makes no sense to consider P ( X = x) since it must
be 0 (why?). Instead, we consider the probability P ( X ≤ x), which is given by P ( X ≤ x) = x.
The reasoning is this: the interval (0, 1) has length 1, and for x in (0, 1) the interval (0, x)
has length x. So since X represents a random number in (0, 1), and hence is uniformly
distributed over (0, 1), then

length of (0, x) x
P ( X ≤ x) = = = x.
length of (0, 1) 1

We call X a continuous random variable on the sample space Ω = (0, 1). An event A is a
subset of the sample space. For example, in our case the event X ≤ x is the set (0, x).
In the case of a discrete random variable, we saw how the probability of an event was
the sum of the probabilities of the individual outcomes comprising that event (for instance,
P ( X ≤ 3) = P ( X = 1) + P ( X = 2) + P ( X = 3) in the die example). For a continuous random
variable, the probability of an event will instead be the integral of a function, which we will
now describe.
Let X be a continuous real-valued random variable on a sample space Ω in R. For simplic-
ity, let Ω = (a, b). Define the distribution function F of X as

F ( x) = P ( X ≤ x) , for −∞ < x < ∞ (4.33)


 1, for x ≥ b


= P ( X ≤ x), for a < x < b (4.34)


 0, for x ≤ a .
148 CHAPTER 4. MULTIPLE INTEGRALS

Suppose that there is a nonnegative, continuous real-valued function f on R such that

wx
F ( x) = f ( y) d y , for −∞ < x < ∞ , (4.35)
−∞

and
w∞
f ( x) dx = 1 . (4.36)
−∞

Then we call f the probability density function for X . We thus have

wx
P ( X ≤ x) = f ( y) d y , for a < x < b . (4.37)
a

Also, by the Fundamental Theorem of Calculus, we have

F ′ ( x) = f ( x) , for −∞ < x < ∞. (4.38)

Example 4.16. Let X represent a randomly selected real number in the interval (0, 1). We
say that X has the uniform distribution on (0, 1), with distribution function

1, for x ≥ 1


F ( x) = P ( X ≤ x) = x, for 0 < x < 1 (4.39)


0, for x ≤ 0 ,

and probability density function

(
1, for 0 < x < 1
f ( x) = F ′ ( x) = (4.40)
0, elsewhere.

In general, if X represents a randomly selected real number in an interval (a, b), then X has
the uniform distribution function

 1, for x ≥ b


x
F ( x) = P ( X ≤ x) = b−a , for a < x < b (4.41)


 0, for x ≤ a ,

and probability density function

(
1
′ b−a , for a < x < b
f ( x) = F ( x) = (4.42)
0, elsewhere.
 4.7 Application: Probability and Expected Value 149

Example 4.17. A famous distribution function is given by the standard normal distribution,
whose probability density function f is

1 2
f ( x) = p e− x /2 , for −∞ < x < ∞. (4.43)
2π

This is often called a “bell curve”, and is used widely in statistics. Since we are claiming that
f is a probability density function, we should have

w∞ 1 2
p e− x /2 dx = 1 (4.44)
−∞ 2π

by formula (4.36), which is equivalent to

w∞ 2 p
/2
e− x dx = 2π . (4.45)
−∞

We can use a double integral in polar coordinates to verify this integral. First,

w∞ w∞ w∞ w∞
à !
−( x2 + y2 )/2 − y2 /2 − x2 /2
e dx d y = e e dx d y
−∞ −∞ −∞ −∞

w∞ w∞
à !à !
− x2 /2 − y2 /2
= e dx e dy
−∞ −∞

w∞
à !2
− x2 /2
= e dx
−∞

since the same function is being integrated twice in the middle equation, just with different
variables. But using polar coordinates, we see that

w∞ w∞ w2πw∞
−( x2 + y2 )/2 2
/2
e dx d y = e−r r dr d θ
−∞ −∞ 0 0

w2πµ ¯r=∞ ¶
¯
− r 2 /2 ¯
= −e ¯ dθ
0 r =0

w2π w2π
0
= (0 − (− e )) d θ = 1 d θ = 2π ,
0 0
150 CHAPTER 4. MULTIPLE INTEGRALS

and so

w∞
à !2
− x2 /2
e dx = 2π , and hence
−∞

w∞ 2 p
/2
e− x dx = 2π .
−∞

In addition to individual random variables, we can consider jointly distributed random

variables. For this, we will let X , Y and Z be three real-valued continuous random variables
defined on the same sample space Ω in R (the discussion for two random variables is similar).
Then the joint distribution function F of X , Y and Z is given by

F ( x, y, z) = P ( X ≤ x, Y ≤ y, Z ≤ z) , for −∞ < x, y, z < ∞. (4.46)

If there is a nonnegative, continuous real-valued function f on R3 such that

wz wy wx
F ( x, y, z) = f ( u, v, w) du dv dw , for −∞ < x, y, z < ∞ (4.47)
−∞ −∞ −∞

and
w∞ w∞ w∞
f ( x, y, z) dx d y dz = 1 , (4.48)
−∞ −∞ −∞

then we call f the joint probability density function for X , Y and Z . In general, for a 1 < b 1 ,
a 2 < b 2 , a 3 < b 3 , we have

wb3 wb2 wb1

P (a 1 < X ≤ b 1 , a 2 < Y ≤ b 2 , a 3 < Z ≤ b 3 ) = f ( x, y, z) dx d y dz , (4.49)
a3 a2 a1

with the ≤ and < symbols interchangeable in any combination. A triple integral, then, can
be thought of as representing a probability (for a function f which is a probability density
function).

Example 4.18. Let a, b, and c be real numbers selected randomly from the interval (0, 1).
What is the probability that the equation ax2 + bx + c = 0 has at least one real solution x?
 4.7 Application: Probability and Expected Value 151

Solution: We know by the quadratic formula that there is at least c

one real solution if b2 − 4ac ≥ 0. So we need to calculate P ( b2 − 4ac ≥
1
0). We will use three jointly distributed random variables to do this.
1
First, since 0 < a, b, c < 1, we have c= 4a

p p
b2 − 4ac ≥ 0 ⇔ 0 < 4ac ≤ b2 < 1 ⇔ 0 < 2 a c ≤ b < 1 , R1 R2
a
0 1 1
where the last relation holds for all 0 < a, c < 1 such that 4

1 Figure 4.7.1 Region

0 < 4ac < 1 ⇔ 0 < c < . R = R1 ∪ R2
4a

Considering a, b and c as real variables, the region R in the ac-plane where the above
relation holds is given by R = {(a, c) : 0 < a < 1, 0 < c < 1, 0 < c < 41a }, which we can see is a
union of two regions R 1 and R 2 , as in Figure 3.7.1 above.
Now let X , Y and Z be continuous random variables, each representing a randomly se-
lected real number from the interval (0, 1) (think of X , Y and Z representing a, b and c,
respectively). Then, similar to how we showed that f ( x) = 1 is the probability density func-
tion of the uniform distribution on (0, 1), it can be shown that f ( x, y, z) = 1 for x, y, z in (0, 1)
(0 elsewhere) is the joint probability density function of X , Y and Z . Now,
p p
P ( b2 − 4ac ≥ 0) = P ((a, c) ∈ R, 2 a c ≤ b < 1) ,
p p
so this probability is the triple integral of f (a, b, c) = 1 as b varies from 2 a c to 1 and as
(a, c) varies over the region R . Since R can be divided into two regions R 1 and R 2 , then the
required triple integral can be split into a sum of two triple integrals, using vertical slices in
R:
1/4w1
w w1 w1 1/4
wa w1
P ( b2 − 4ac ≥ 0) = 1 db dc da + 1 db dc da
p p p p
0 0 2 a c 1/4 0 2 a c
| {z } | {z }
R1 R2
1/4w1
w p p w1 1/4
wa p p
= (1 − 2 a c) dc da + (1 − 2 a c) dc da
0 0 1/4 0
w
1/4µ
p ¯ c =1 ¶ w1 µ p ¯ c=1/4a ¶
c − 34 a c3/2 ¯ c − 34 a c3/2 ¯
¯ ¯
= da + da
c =0 c =0
0 1/4
w
1/4 w1
4p
¡ ¢ 1
= 1− 3 a da + 12a da
0 1/4
¯1/4 ¯1
8 3/2 ¯¯ 1 ¯
= a− a ¯ + ln a ¯¯
9 0 12 1/4
152 CHAPTER 4. MULTIPLE INTEGRALS

µ ¶ µ ¶
1 1 1 1 5 1
= − + 0− ln = + ln 4
4 9 12 4 36 12
5 + 3 ln 4
P ( b2 − 4ac ≥ 0) = ≈ 0.2544
36

In other words, the equation ax2 + bx + c = 0 has about a 25% chance of being solved!

Expected Value
The expected value E X of a random variable X can be thought of as the “average” value of
X as it varies over its sample space. If X is a discrete random variable, then
X
EX = x P ( X = x) , (4.50)
x

with the sum being taken over all elements x of the sample space. For example, if X repre-
sents the number rolled on a six-sided die, then

6
X 6
X 1
EX = x P ( X = x) = x = 3.5 (4.51)
x =1 x =1 6

is the expected value of X , which is the average of the integers 1–6.

If X is a real-valued continuous random variable with probability density function f , then

w∞
EX = x f ( x) dx . (4.52)
−∞

For example, if X has the uniform distribution on the interval (0, 1), then its probability
density function is
(
1, for 0 < x < 1
f ( x) = (4.53)
0, elsewhere,

and so
w∞ w1 1
EX = x f ( x) dx = x dx = . (4.54)
−∞
2
0

For a pair of jointly distributed, real-valued continuous random variables X and Y with
joint probability density function f ( x, y), the expected values of X and Y are given by

w∞ w∞ w∞ w∞
EX = x f ( x, y) dx d y and EY = y f ( x, y) dx d y , (4.55)
−∞ −∞ −∞ −∞

respectively.
 4.7 Application: Probability and Expected Value 153

Example 4.19. If you were to pick n > 2 random real numbers from the interval (0, 1), what
are the expected values for the smallest and largest of those numbers?
Solution: Let U1 , . . . ,Un be n continuous random variables, each representing a randomly
selected real number from (0, 1) with the uniform distribution on (0, 1). Define random vari-
ables X and Y by
X = min(U1 , . . . ,Un ) and Y = max(U1 , . . . ,Un ) .
Then it can be shown3 that the joint probability density function of X and Y is
(
n( n − 1)( y − x)n−2 , for 0 ≤ x ≤ y ≤ 1
f ( x, y) = (4.56)
0, elsewhere.

Thus, the expected value of X is

w1 w1
EX = n( n − 1) x( y − x)n−2 d y dx
0 x
w1 µ ¯ y=1 ¶
n −1 ¯
= nx( y − x) ¯ dx
y= x
0
w1
= nx(1 − x)n−1 dx , so integration by parts yields
0
1 ¯1
= − x(1 − x)n − (1 − x)n+1 ¯
¯
n+1 0
1
EX = ,
n+1
and similarly (see Exercise 3) it can be shown that

w1 wy n
EY = n( n − 1) y( y − x)n−2 dx d y = .
n+1
0 0

So, for example, if you were to repeatedly take samples of n = 3 random real numbers from
(0, 1), and each time store the minimum and maximum values in the sample, then the aver-
age of the minimums would approach 14 and the average of the maximums would approach
3
4 as the number of samples grows. It would be relatively simple (see Exercise 4) to write a
computer program to test this.

Exercises
B
3 See Ch. 6 in H OEL, P ORT and S TONE.
154 CHAPTER 4. MULTIPLE INTEGRALS

1. Evaluate the integral

w∞ 2
e− x dx
−∞

using anything you have learned so far.

2. For σ > 0 and µ > 0, evaluate

w∞ 1 2 2
p e−( x−µ) /2σ dx.
−∞ σ 2π

n
3. Show that EY = n +1 in Example 4.19

C
4. Write a computer program (in the language of your choice) that verifies the results in
Example 4.19 for the case n = 3 by taking large numbers of samples.

5. Repeat Exercise 4 for the case when n = 4.

6. For continuous random variables X , Y with joint probability density function f ( x, y),
define the second moments E ( X 2 ) and E (Y 2 ) by
w∞ w∞ w∞ w∞
2 2 2
E( X ) = x f ( x, y) dx d y and E (Y ) = y2 f ( x, y) dx d y ,
−∞ −∞ −∞ −∞

and the variances Var( X ) and Var(Y ) by

Var( X ) = E ( X 2 ) − (E X )2 and Var(Y ) = E (Y 2 ) − (EY )2 .

Find Var( X ) and Var(Y ) for X and Y as in Example 4.19.

7. Continuing Exercise 6, the correlation ρ between X and Y is defined as

E ( X Y ) − (E X )(EY )
ρ = p ,
Var( X ) Var(Y )
where
w∞ w∞
E( X Y ) = x y f ( x, y) dx d y.
−∞ −∞

Find ρ for X and Y as in Example 4.19.

(Note: The quantity E ( X Y ) − (E X )(EY ) is called the covariance of X and Y .)

8. In Example 4.18 would the answer change if the interval (0, 100) is used instead of (0, 1)?
Explain.
5 Line and Surface Integrals

5.1 Line Integrals

In single-variable calculus you learned how to integrate a real-valued function f ( x) over an
interval [a, b] in R1 . This integral (usually called a Riemann integral) can be thought of as
an integral over a path in R1 , since an interval (or collection of intervals) is really the only
kind of “path” in R1 . You may also recall that if f ( x) represented the force applied along the
x-axis to an object at position x in [a, b], then the work W done in moving that object from
position x = a to x = b was defined as the integral:

wb
W = f ( x) dx.
a

In this section, we will see how to define the integral of a function (either real-valued or
vector-valued) of two variables over a general curve (also called path) in R2 . This definition
will be motivated by the physical notion of work. We will begin with real-valued functions of
two variables.
In physics, the intuitive idea of work is that

Work = Force × Distance .

Assume you move a an object of unit weight along a curve C in R2 and want to find the work
of the force which works against the friction. Suppose f ( x, y) is the coefficient of friction at
the point ( x, y). In this case the force has magnitude f ( x, y) and it is applied in the direction
of motion along C (see Figure 5.1.1 below).

y t = ti q
C
∆s i ≈ ∆ xi 2 + ∆ yi 2
t=a ∆ yi
t = t i +1
∆ xi
t=b
x
0

Figure 5.1.1 Curve C : x = x(t), y = y(t) for t in [a, b]

155
156 CHAPTER 5. LINE AND SURFACE INTEGRALS

We will assume for now that the function f ( x, y) is continuous and real-valued, so we only
consider the magnitude of the force. Partition the interval [a, b] as follows:

a = t 0 < t 1 < t 2 < · · · < t n−1 < t n = b , for some integer n ≥ 2

q a typical subinterval [ t i , t i+1 ] the distance ∆ s i traveled

As we can see from Figure 5.1.1, over
along the curve is approximately ∆ xi 2 + ∆ yi 2 , by the Pythagorean Theorem. Thus, if the
subinterval is small enough then the work done in moving the object along that piece of the
curve is approximately
q
Force × Distance ≈ f ( xi∗ , yi∗ ) ∆ xi 2 + ∆ yi 2 , (5.1)

where ( xi∗ , yi∗ ) = ( x( t i ∗), y( t i ∗)) for some t i ∗ in [ t i , t i+1 ], and so

−1
nX q
W ≈ f ( xi∗ , yi∗ ) ∆ xi 2 + ∆ yi 2 (5.2)
i =0

is approximately the total amount of work done over the entire curve. But since
s
µ ¶ µ ¶
∆ xi 2 ∆ yi 2
q
2 2
∆ xi + ∆ yi = + ∆t i ,
∆t i ∆t i

where ∆ t i = t i+1 − t i , then

s
−1
nX µ ¶2 µ ¶2
∆ xi ∆ yi
W ≈ f ( xi∗ , yi∗ ) + ∆t i . (5.3)
i =0 ∆t i ∆t i

Taking the limit of that sum as the length of the largest subinterval goes to 0, the sum over
xi ∆ yi
all subintervals becomes the integral from t = a to t = b, ∆ ′ ′
∆ t i and ∆ t i become x ( t) and y ( t),
respectively, and f ( xi∗ , yi∗ ) becomes f ( x( t), y( t)), so that

wb q
W = f ( x( t), y( t)) x ′ ( t)2 + y ′ ( t)2 dt . (5.4)
a

The integral on the right side of the above equation gives us our idea of how to define,
for any real-valued function f ( x, y), the integral of f ( x, y) along the curve C , called a line
integral:

Definition 5.1. For a real-valued function f ( x, y) and a curve C in R2 , parametrized by

x = x( t), y = y( t), a ≤ t ≤ b, the line integral of f ( x, y) along C with respect to arc length
s is
w wb q
f ( x, y) ds = f ( x( t), y( t)) x ′ ( t)2 + y ′ ( t)2 dt . (5.5)
C a
 5.1 Line Integrals 157

The symbol ds is the differential of the arc length function

wt q
s = s( t) = x ′ ( u)2 + y ′ ( u)2 du , (5.6)
a

which you may recognize from Section 1.9 as the length of the curve C over the interval [a, t],
for all t in [a, b]. That is,
q
′
ds = s ( t) dt = x ′ ( t)2 + y ′ ( t)2 dt , (5.7)

by the Fundamental Theorem of Calculus.

r
For a general real-valued function f ( x, y), what does the line integral C f ( x, y) ds rep-
resent? The preceding discussion of ds gives us a clue. You can think of differentials as
infinitesimal lengths. So if you think of f ( x, y) as the height of a picket fence along C , then
f ( x, y) ds can be thought of as approximately the area of a section of r that fence over some
infinitesimally small section of the curve, and thus the line integral C f ( x, y) ds is the total
area of that picket fence (see Figure 5.1.2).

f ( x, y)

C ds

x
0

Figure 5.1.2 Area of shaded rectangle = height × width ≈ f (x, y) ds

Example 5.1. Use a line integral to show that the lateral surface area A of a right circular
cylinder of radius r and height h is 2π rh.
158 CHAPTER 5. LINE AND SURFACE INTEGRALS

Solution: We will use the right circular cylinder with base circle C z
given by x2 + y2 = r 2 and with height h in the positive z direction r
(see Figure 4.1.3). Parametrize C as follows:

x = x( t) = r cos t , y = y( t) = r sin t , 0 ≤ t ≤ 2π h = f (x, y)

Let f ( x, y) = h for all ( x, y). Then y

0
w wb q x C : x 2 + y2 = r 2
A = f ( x, y) ds = f ( x( t), y( t)) x ′ ( t)2 + y ′ ( t)2 dt
C a Figure 5.1.3
w2π p
= h (− r sin t)2 + ( r cos t)2 dt
0
w2π p
= h r sin2 t + cos2 t dt
0
w2π
= rh 1 dt = 2π rh
0

Note in Example 5.1 that if we had traversed the circle C twice (that is, let t vary from 0
to 4π) then we would have gotten an area of 4π rh — twice the desired area, even though the
curve itself is still the same (namely, a circle of radius r ). Also, notice that we traversed the
circle in the counter-clockwise direction. If we had gone in the clockwise direction, using the
parametrization

x = x( t) = r cos(2π − t) , y = y( t) = r sin(2π − t) , 0 ≤ t ≤ 2π , (5.8)

then it is easy to verify (see Exercise 12) that the value of the line integral is unchanged.
In general, it can ber shown (see Exercise 15) that reversing the direction in which a curve
C is traversed leaves C f ( x, y) ds unchanged, for any f ( x, y). If a curve C has a parametriza-
tion x = x( t), y = y( t), a ≤ t ≤ b, then denote by −C the same curve as C but traversed in the
opposite direction. Then −C is parametrized by

x = x( a + b − t) , y = y( a + b − t ) , a≤t≤b , (5.9)

and we have w w
f ( x, y) ds = f ( x, y) ds . (5.10)
C −C
Notice that our definition of the line integral was with respect to the arc length parameter
s. We can also define
w wb
f ( x, y) dx = f ( x( t), y( t)) x ′ ( t) dt (5.11)
C a
 5.1 Line Integrals 159

as the line integral of f ( x, y) along C with respect to x, and

w wb
f ( x, y) d y = f ( x( t), y( t)) y ′ ( t) dt (5.12)
C a

as the line integral of f ( x, y) along C with respect to y.

In the derivation of the formula for a line integral, we used the idea of work as force
multiplied by distance. However, we know that force is actually a vector. So it would be
helpful to develop a vector form for a line integral. For this, suppose that we have a function
f( x, y) defined on R2 by
f( x, y) = P ( x, y) i + Q ( x, y) j

for some continuous real-valued functions P ( x, y) and Q ( x, y) on R2 . Such a function f is

called a vector field on R2 . It is defined at points in R2 , and its values are vectors in R2 . For
a curve C with a smooth parametrization x = x( t), y = y( t), a ≤ t ≤ b, let

r( t) = x( t) i + y( t) j

be the position vector for a point ( x( t), y( t)) on C . Then r′ ( t) = x′ ( t) i + y ′ ( t) j and so

w w wb wb
P ( x, y) dx + Q ( x, y) d y = P ( x( t), y( t)) x ′ ( t) dt + Q ( x( t), y( t)) y ′ ( t) dt
C C a a
wb
= (P ( x( t), y( t)) x ′ ( t) + Q ( x( t), y( t)) y ′ ( t)) dt
a
wb
= f( x( t), y( t)) · r′ ( t) dt
a

by definition of f( x, y). Notice that the function f( x( t), y( t)) · r ′ ( t) is a real-valued function on
[a, b], so the last integral on the right looks somewhat similar to our earlier definition of a
line integral. This leads us to the following definition:

Definition 5.2. For a vector field f( x, y) = P ( x, y) i + Q ( x, y) j and a curve C with a smooth

parametrization x = x( t), y = y( t), a ≤ t ≤ b, the line integral of f along C is
w w w
f · dr = P ( x, y) dx + Q ( x, y) d y (5.13)
C C C
wb
= f( x( t), y( t)) · r ′ ( t) dt , (5.14)
a

where r( t) = x( t) i + y( t) j is the position vector for points on C .

160 CHAPTER 5. LINE AND SURFACE INTEGRALS

We use the notation d r = r ′ ( t) dt = dx i + d y j to denote the differential of the vector-valued

function r. The line integral in Definition 5.2 is often called a line integral of a vector field
to distinguish it from the line integral in Definition 5.1 which is called a line integral of a
scalar field. For convenience we will often write
w w w
P ( x, y) dx + Q ( x, y) d y = P ( x, y) dx + Q ( x, y) d y ,
C C C

where it is understood that the line integral along C is being applied to both P and Q . The
quantity P ( x, y) dx + Q ( x, y) d y is known as a differential form. For a real-valued function
F ( x, y), the differential of F is dF = ∂∂Fx dx + ∂∂Fy d y. A differential form P ( x, y) dx + Q ( x, y) d y
is called exact if it equals dF for some function F ( x, y).
Recall that if the points on a curve C have position vector r( t) = x( t) i + y( t) j, then r ′ ( t) is a
tangent vector to C at the point ( x( t), y( t)) in the direction of increasing t (which we call the
direction of C ). Since C is a smooth curve, then r ′ ( t) 6= 0 on [a, b] and hence

r ′ ( t)
T( t) = ° ′ °
° r ( t) °

is the unit tangent vector to C at ( x( t), y( t)). Putting Definitions 5.1 and 5.2 together we get
the following theorem:

Theorem 5.1. For a vector field f( x, y) = P ( x, y) i + Q ( x, y) j and a curve C with a smooth

parametrization x = x( t), y = y( t), a ≤ t ≤ b and position vector r( t) = x( t) i + y( t) j,
w w
f · dr = f · T ds , (5.15)
C C

r ′ ( t)
where T( t) = k r ′ ( t) k is the unit tangent vector to C at ( x( t), y( t)).

If the vector field f( x, y) represents the force moving an object along a curve C , then the work
W done by this force is w w
W = f · T ds = f · dr . (5.16)
C C

r 2
Example 5.2. Evaluate C (x + y2 ) dx + 2 x y d y, where:

(a) C : x = t , y = 2t , 0≤t≤1

(b) C : x = t , y = 2 t2 , 0≤t≤1
 5.1 Line Integrals 161

Solution: Figure 4.1.4 shows both curves. y

(1, 2)
′ ′ 2
(a) Since x ( t) = 1 and y ( t) = 2, then
w w1 ¡ ¢
( x2 + y2 ) dx + 2 x y d y = ( x( t)2 + y( t)2 ) x ′ ( t) + 2 x( t) y( t) y ′ ( t) dt
C 0 x
w1 ¡ ¢ 0 1
= ( t2 + 4 t2 )(1) + 2 t(2 t)(2) dt
Figure 5.1.4
0
w1
= 13 t2 dt
0
¯1
13 t3 ¯¯ 13
= =
3 ¯0 3
(b) Since x ′ ( t) = 1 and y ′ ( t) = 4 t, then
w w1 ¡ ¢
( x2 + y2 ) dx + 2 x y d y = ( x( t)2 + y( t)2 ) x ′ ( t) + 2 x( t) y( t) y ′ ( t) dt
C 0
w1 ¡ ¢
= ( t2 + 4 t4 )(1) + 2 t(2 t2 )(4 t) dt
0
w1
= ( t2 + 20 t4 ) dt
0
¯1
t3 ¯
5¯ 1 13
= + 4t ¯ = + 4 =
3 0 3 3
So in both cases, if the vector field f( x, y) = ( x2 + y2 ) i + 2 x y j represents the force moving an
object from (0, 0) to (1, 2) along the given curve C , then the work done is 13 3 . This may lead
you to think that work (and more generally, the line integral of a vector field) is independent
of the path taken. However, as we will see in the next section, this is not always the case.
Although we defined line integrals over a single smooth curve, if C is a piecewise smooth
curve, that is
C = C1 ∪ C2 ∪ . . . ∪ C n
is the union of smooth curves C 1 , . . . , C n , then we can define
w w w w
f · dr = f · d r1 + f · d r2 + . . . + f · d rn
C C1 C2 Cn

where each ri is the position vector of the curve C i .

r 2
Example 5.3. Evaluate C (x + y2 ) dx + 2 x y d y, where C is the polygonal path from (0, 0) to
(0, 2) to (1, 2).
162 CHAPTER 5. LINE AND SURFACE INTEGRALS

y
Solution: Write C = C 1 ∪ C 2 , where C 1 is the curve given by x = 0, y = t,
2 (1, 2)
0 ≤ t ≤ 2 and C 2 is the curve given by x = t, y = 2, 0 ≤ t ≤ 1 (see Figure C2
4.1.5). Then
w w C1
( x2 + y2 ) dx + 2 x y d y = ( x2 + y2 ) dx + 2 x y d y
x
C C1
w 0 1
+ ( x2 + y2 ) dx + 2 x y d y
Figure 5.1.5
C2

w2 ¡ ¢ w1 ¡ ¢
2 2
= (0 + t )(0) + 2(0) t(1) dt + ( t2 + 4)(1) + 2 t(2)(0) dt
0 0

w2 w1
= 0 dt + ( t2 + 4) dt
0 0
¯1
t3 ¯ 1 13
= + 4 t ¯¯ = + 4 =
3 0 3 3

Line integral
rb notation varies quite a bit. For example, in physics it is common to see the
notation a f · d l, where it is understood that the limits of integration a and b are for the
underlying
r parameter t of the curve, and the letter l signifies length. Also, the formulation
C f · T ds from Theorem 5.1 is often preferred in physics since it emphasizes the idea of
integrating the tangential component f · T of f in the direction of T (that is, in the direction
of C ), which is a useful physical interpretation of line integrals.
Exercises
A
For Exercises 1–4, calculate w
f ( x, y) ds
C
for the given function f ( x, y) and curve C .

1. f ( x, y) = x y; C : x = cos t, y = sin t, 0 ≤ t ≤ π/2

x
2. f ( x, y) = ; C : x = t, y = 0, 0 ≤ t ≤ 1
x2 + 1
3. f ( x, y) = 2 x + y; C : polygonal path from (0, 0) to (3, 0) to (3, 2)

4. f ( x, y) = x + y2 ; C : path from (2, 0) counterclockwise along the circle x2 + y2 = 4 to the

point (−2, 0) and then back to (2, 0) along the x-axis

5. Use a line integral to find the lateral surface area of the part of the cylinder
x2 + y2 = 4 below the plane x + 2 y + z = 6 and above the x y-plane.
 5.1 Line Integrals 163

For Exercises 6–11, calculate w

f · dr
C
for the given vector field f( x, y) and curve C .
6. f( x, y) = i − j; C : x = 3 t, y = 2 t, 0 ≤ t ≤ 1

7. f( x, y) = y i − x j; C : x = cos t, y = sin t, 0 ≤ t ≤ 2π

8. f( x, y) = x i + y j; C : x = cos t, y = sin t, 0 ≤ t ≤ 2π

9. f( x, y) = ( x2 − y) i + ( x − y2 ) j; C : x = cos t, y = sin t, 0 ≤ t ≤ 2π

10. f( x, y) = x y2 i + x y3 j; C : the polygonal path from (0, 0) to (1, 0) to (0, 1) to (0, 0)

11. f( x, y) = ( x2 + y2 ) i; C : x = 2 + cos t, y = sin t, 0 ≤ t ≤ 2π

B
12. Verify that the value of the line integral in Example 5.1 is unchanged when using the
parametrization of the circle C given in formulas (5.8).

13. Show that if f ⊥ r ′ ( t) at each point r( t) along a smooth curve C , then

w
f · d r = 0.
C

14. Show that if f points in the same direction as r ′ ( t) at each point r( t) along a smooth
curve C , then w w
f · d r = k f k ds.
C C

C
15. Prove that w w
f ( x, y) ds = f ( x, y) ds.
C −C
(Hint: Use formulas (5.9).)

16. Let C be a smooth curve with arc length L, and suppose that f( x, y) = P ( x, y) i + Q ( x, y) j
is a vector field such that k f( x, y) k ≤ M for all ( x, y) on C . Show that
¯w
¯ ¯
¯
¯ ¯
¯ f · d r ¯ ≤ ML.
¯ ¯
C
¯r ¯ r
¯ b ¯ b
(Hint: Recall that ¯ a g( x) dx ¯ ≤ a | g( x) | dx for Riemann integrals.)
rb
17. Prove that the Riemann integral a f ( x) dx is a special case of a line integral.
164 CHAPTER 5. LINE AND SURFACE INTEGRALS

5.2 Properties of Line Integrals

We know from the previous section that for line integrals of real-valued functions (scalar
fields), reversing the direction in which the integral is taken along a curve does not change
the value of the line integral:
w w
f ( x, y) ds = f ( x, y) ds (5.17)
C −C

For line integrals of vector fields, however, the value does change. To see this, let f( x, y) =
P ( x, y) i + Q ( x, y) j be a vector field, with P and Q continuously differentiable functions. Let
C be a smooth curve parametrized by x = x( t), y = y( t), a ≤ t ≤ b, with position vector r( t) =
x( t) i + y( t) j (we will usually abbreviate this by saying that C : r( t) = x( t) i + y( t) j is a smooth
curve). We know that the curve −C traversed in the opposite direction is parametrized by
x = x(a + b − t), y = y(a + b − t), a ≤ t ≤ b. Then

w wb d
P ( x, y) dx = P ( x(a + b − t), y(a + b − t)) ( x(a + b − t)) dt
a
dt
−C
wb
= P ( x(a + b − t), y(a + b − t)) (− x ′ (a + b − t)) dt (by the Chain Rule)
a
wa
= P ( x( u), y( u)) (− x ′ ( u)) (− du) (by letting u = a + b − t)
b
wa
= P ( x( u), y( u)) x ′ ( u) du
b
wb wa wb
= − P ( x( u), y( u)) x ′ ( u) du , since =− , so
a b a
w w
P ( x, y) dx = − P ( x, y) dx
−C C

since we are just using a different letter ( u) for the line integral along C . A similar argument
shows that w w
Q ( x, y) d y = − Q ( x, y) d y ,
−C C

and hence
w w w
f · dr = P ( x, y) dx + Q ( x, y) d y
−C −C −C
w w
= − P ( x, y) dx + − Q ( x, y) d y
C C
 5.2 Properties of Line Integrals 165

w w
à !
= − P ( x, y) dx + Q ( x, y) d y
w wC C

f · dr = − f · dr . (5.18)
−C C

The above formula can be interpreted in terms of the work done by a force f( x, y) (treated
as a vector) moving an object along a curve C : the total work performed moving the object
along C from its initial point to its terminal point, and then back to the initial point moving
backwards along the same path, is zero. This is because when force is considered as a vector,
direction is accounted for.
The preceding discussion shows the importance of always taking the direction of the curve
into account when using line integrals of vector fields. For this reason, the curves in line
integrals are sometimes referred to as directed curves or oriented curves.
Recall that our definition of a line integral required that we have a parametrization x =
x( t), y = y( t), a ≤ t ≤ b for the curve C . But as we know, any curve has infinitely many
parametrizations. So could we get a different value for a line integral using some other
parametrization of C , say, x = x̃( u), y = ỹ( u), c ≤ u ≤ d ? If so, this would mean that our
definition is not well-defined. Luckily, it turns out that the value of a line integral of a
vector field is unchanged as long as the direction of the curve C is preserved by whatever
parametrization is chosen:

Theorem 5.2. Let f( x, y) = P ( x, y) i + Q ( x, y) j be a vector field, and let C be a smooth curve

parametrized by x = x( t), y = y( t), a ≤ t ≤ b. Suppose that t = α( u) for c ≤ u ≤ d , such that
a = α( c), b = α( dr), and α ′ ( u) > 0 on the open interval ( c, d ) (that is, α( u) is strictly increasing
on [ c, d ]). Then C f· d r has the same value for the parametrizations x = x( t), y = y( t), a ≤ t ≤ b
and x = x̃( u) = x(α( u)), y = ỹ( u) = y(α( u)), c ≤ u ≤ d .

Proof: Since α( u) is strictly increasing and maps [ c, d ] onto [a, b], then we know that t =
α( u) has an inverse function u = α−1 ( t) defined on [a, b] such that c = α−1 (a), d = α−1 ( b),
1
and du ′
dt = α ′ ( u) . Also, dt = α ( u) du, and by the Chain Rule

d x̃ d dx dt x̃ ′ ( u)
x̃ ′ ( u) = = ( x(α( u))) = = x ′ ( t) α ′ ( u ) ⇒ x ′ ( t) =
du du dt du α ′ ( u)

so making the susbstitution t = α( u) gives

wb α−w1 ( b)
x̃ ′ ( u) ′
′
P ( x( t), y( t)) x ( t) dt = P ( x(α( u)), y(α( u))) (α ( u) du)
a
α ′ ( u)
α−1 (a)
wd
= P ( x̃( u), ỹ( u)) x̃ ′ ( u) du ,
c
166 CHAPTER 5. LINE AND SURFACE INTEGRALS

r
which shows that C P r ( x, y) dx has the same value for both parametrizations. A similar
argument
r shows that C Q ( x, y) d y has the same value for both parametrizations, and hence
C f · d r has the same value.
QED

Notice that the condition α ′ ( u) > 0 in Theorem 5.2 means that the two parametrizations
move along C in the same direction. That was not the case with the “reverse” parametriza-
tion for −C : for u = a + b − t we have t = α( u) = a + b − u ⇒ α ′ ( u) = −1 < 0.
r
Example 5.4. Evaluate the line integral C ( x2 + y2 ) dx + 2 x y d y from Example 5.2, Section
4.1, along the curve C : x = t, y = 2 t2 , 0 ≤ t ≤ 1, where t = sin u for 0 ≤ u ≤ π/2.
dt
Solution: First, we notice that 0 = sin 0, 1 = sin(π/2), and du = cos u > 0 on (0, π/2). So by
Theorem 5.2 we know that if C is parametrized by

x = sin u , y = 2 sin2 u , 0 ≤ u ≤ π/2

r
then C ( x2 + y2 ) dx + 2 x y d y should have the same value as we found in Example 5.2, namely
13
3 . And we can indeed verify this:

w w/2
π
¡ ¢
2 2
( x + y ) dx + 2 x y d y = (sin2 u + (2 sin2 u)2 ) cos u + 2(sin u)(2 sin2 u)4 sin u cos u du
C 0
w/2
π
¡ ¢
= sin2 u + 20 sin4 u cos u du
0
¯π/2
sin3 u 5 ¯
¯
= + 4 sin u ¯
3 0
1 13
= +4 =
3 3
In other words, the line integral is unchanged whether t or u is the parameter for C .

By a closed curve, we mean a curve C whose initial point and terminal point are the
same; that is, for C : x = x( t), y = y( t), a ≤ t ≤ b, we have ( x(a), y(a)) = ( x( b), y( b)).
A simple closed curve is a closed curve which does not intersect itself. Note that any
closed curve can be regarded as a union of simple closed curves (think of the loops in a figure
eight). We use the special notation
z z
f ( x, y) ds and f · dr
C C

to denote line integrals of scalar and vector

w fields, respectively, along closed curves. In some
w
older texts you may see the notation or  to indicate a line integral traversing a closed
curve in a counterclockwise or clockwise direction, respectively.
 5.2 Properties of Line Integrals 167

Ï Ï

t=a t=b t=a

t=b
Î Î
C C
(a) Closed (b) Not closed

Figure 5.2.1 Closed vs nonclosed curves

So far, the examples we have seen of line integrals (for instance, Example 5.2) have had
the same value for different curves joining the initial point to the terminal point. That is,
the line integral has been independent of the path joining the two points. As we mentioned
before, this is not always the case. The following theorem gives a necessary and sufficient
condition for this path independence:
r
Theorem 5.3. In a region R , theuline integral C f · d r is independent of the path between
any two points in R if and only if C f · d r = 0 for every closed curve C which is contained in
R.

u
Proof: Suppose that C f · d r = 0 for every closed curve C which is contained in R . Let P1
and P2 be two distinct points in R . Let C 1 be a curve in R going from P1 to P2 , and let C 2
be another curve in R going from P1 to P2 , as in Figure 4.2.2.
Then C =uC 1 ∪ −C 2 is a closed curve in R (from P1 to C1
P1 ), and so C f · d r = 0. Thus, Ï
z
0 = f · dr P1 P2
C
w w
Ï
= f · dr + f · dr
C2
C1 −C 2
w w Figure 5.2.2
= f · dr − f · d r , and so
C1 C2
r r
· ·
C 1 f · d r = C 2 f · d r. This proves path independence.

r
Conversely, suppose that the line integral C f · d r is independent of the path between any
two points in R . Let C be a closed curve contained in R . Let P1 and P2 be two distinct points
168 CHAPTER 5. LINE AND SURFACE INTEGRALS

on C . Let C 1 be a part of the curve C that goes from P1 to P2 , and let C 2 be the remaining
part of C that goes from P1 to P2 , again as in Figure 4.2.2. Then by path independence we
have
w w
f · dr = f · dr
C1 C2
w w
f · dr − f · dr = 0
C1 C2
w w
f · dr + f · d r = 0 , so
C1 −C 2
z
f · dr = 0
C

since C = C 1 ∪ −C 2 . QED

Clearly, the above theorem does not give a practical way to determine path independence,
since it is impossible to check the line integrals around all possible closed curves in a region.
What it mostly does is give an idea of the way in which line integrals behave, and how seem-
ingly unrelated line integrals can be related (in this case, a specific line integral between
two points and all line integrals around closed curves).
Recall that if z = f ( x, y) is a continuously differentiable function of x and y, and both
x = x( t) and y = y( t) are differentiable functions of t, then

dz ∂ z dx ∂z d y
= + . (5.19)
dt ∂ x dt ∂ y dt

This is multivariable version of the Chain Rule, see Theorem 3.3 and Corollary 3.4. We
will now use this version of Chain Rule to prove the following sufficient condition for path
independence of line integrals:

Theorem 5.4. Let f( x, y) = P ( x, y) i + Q ( x, y) j be a vector field in some region R , with P and

Q continuously differentiable functions on R . Let C be a smooth curve in R parametrized
by x = x( t), y = y( t), a ≤ t ≤ b. Suppose that there is a real-valued function F ( x, y) such that
∇F = f on R . Then w
f · d r = F (B ) − F ( A ) , (5.20)
C

where A = ( x(a), y(a)) and B = ( x( b), y( b)) are the endpoints of C . Thus, the line integral is
independent of the path between its endpoints, since it depends only on the values of F at
those endpoints.
 5.2 Properties of Line Integrals 169

r
Proof: By definition of ·
C f · d r, we have

w wb ¡ ¢
f · dr = P ( x( t), y( t)) x ′ ( t) + Q ( x( t), y( t)) y ′ ( t) dt
C a
wb µ ∂F dx ∂F d y
¶
∂F ∂F
= + dt (since ∇F = f ⇒ = P and = Q)
a
∂ x dt ∂ y dt ∂x ∂y
wb
= F ( x( t), y( t))′ dt (by the Chain Rule in Theorem 3.3)
a
¯b
¯
= F ( x( t), y( t)) ¯ = F (B) − F ( A )
a

by the Fundamental Theorem of Calculus.

QED

Theorem 5.4 can be thought of as the line integral version of the Fundamental Theorem
of Calculus. A real-valued function F ( x, y) such that ∇F ( x, y) = f( x, y) is called a potential
for f. A conservative vector field is one which has a potential.
r
Example 5.5. Recall from Examples 5.2 and 5.3 in Section 4.1 that the line integral C ( x2 +
y2 ) dx + 2 x y d y was found to have the value 13
3 for three different curves C going from the
point (0, 0) to the point (1, 2). Use Theorem 5.4 to show that this line integral is indeed path
independent.
Solution: We need to find a real-valued function F ( x, y) such that
∂F ∂F
= x 2 + y2 and = 2x y .
∂x ∂y

Suppose that ∂∂Fx = x2 + y2 , Then we must have F ( x, y) = 13 x3 + x y2 + g( y) for some function

g( y). So ∂∂Fy = 2 x y + g ′ ( y) satisfies the condition ∂∂Fy = 2 x y if g ′ ( y) = 0; that is, g( y) = K , where
K is a constant. Since any choice for K will do (why?), we pick K = 0. Thus, a potential
F ( x, y) for f( x, y) = ( x2 + y2 ) i + 2 x y j exists, namely
1 3
F ( x, y) = x + x y2 .
3
r
Hence the line integral C ( x2 + y2 ) dx + 2 x y d y is path independent.
Note that we can also verify that the value of the line integral of f along any curve C going
from (0, 0) to (1, 2) will always be 13
3 , since by Theorem 5.4
w 1 3 1 13
f · d r = F (1, 2) − F (0, 0) = (1) + (1)(2)2 − (0 + 0) = + 4 = .
3 3 3
C
170 CHAPTER 5. LINE AND SURFACE INTEGRALS

A consequence of Theorem 5.4 in the special case where C is a closed curve, so that the
endpoints A and B are the same point, is the following important corollary:
z
Corollary 5.5. If a vector field f has a potential in a region R , then f · d r = 0 for any closed
C
curve C in R . Equivalently, z
∇F · d r = 0
C

for any real-valued function F ( x, y).

z
Example 5.6. Evaluate x dx + y d y for C : x = 2 cos t, y = 3 sin t, 0 ≤ t ≤ 2π.
C
Solution: The vector field f( x, y) = x i + y j has a potential F ( x, y):

∂F 1 2
= x ⇒ F ( x, y) = x + g( y) , so
∂x 2
∂F 1
= y ⇒ g ′ ( y) = y ⇒ g ( y) = y2 + K
∂y 2

1 2 1 2
for any constant K , so F ( x, y) = x + y is a potential for f( x, y). Thus,
2 2
z z
x dx + y d y = f · dr = 0
C C

x2 y2
by Corollary 5.5, since the curve C is closed (it is the ellipse 4 + 9 = 1).

Exercises
A
z
1. Evaluate ( x2 + y2 ) dx + 2 x y d y for C : x = cos t, y = sin t, 0 ≤ t ≤ 2π.
C
w
2. Evaluate ( x2 + y2 ) dx + 2 x y d y for C : x = cos t, y = sin t, 0 ≤ t ≤ π.
C

3. Is there a potential F ( x, y) for f( x, y) = y i − x j? If so, find one.

4. Is there a potential F ( x, y) for f( x, y) = x i − y j? If so, find one.

5. Is there a potential F ( x, y) for f( x, y) = x y2 i + x3 y j? If so, find one.

 5.2 Properties of Line Integrals 171

B
6. Let f( x, y) and g( x, y) be vector fields, let a and b be constants, and let C be a curve in R2 .
Show that w w w
(a f ± b g) · d r = a f · d r ± b g · d r .
C C C
r
7. Let C be a curve whose arc length is L. Show that C 1 ds = L.

8. Let f ( x, y) and g( x, y) be continuously differentiable real-valued functions in a region R .

Show that z z
f ∇ g · dr = − g ∇ f · dr
C C

for any closed curve C in R . (Hint: Use Exercise 21 in Section 2.4.)

−y
9. Let f( x, y) = x2 + y2
i + x2 +x y2 j for all ( x, y) 6= (0, 0), and C : x = cos t, y = sin t, 0 ≤ t ≤ 2π.
(a) Show that
f = ∇F,
for F ( x, y) = tan−1 ( y/ x).
(b) Show that z
f · d r = 2π.
C

Does this contradict Corollary 5.5? Explain.

C
10. Let g( x) and h( y) be differentiable functions, and let f( x, y) = h( y) i + g( x) j. For which
g( x) and h( y), the vector field f is potential? Find the potential F ( x, y) for all these cases.
172 CHAPTER 5. LINE AND SURFACE INTEGRALS

5.3 Green’s Theorem

We will now see a way of evaluating the line integral of a smooth vector field around a simple
closed curve. A vector field f( x, y) = P ( x, y) i + Q ( x, y) j is smooth if its component functions
P ( x, y) and Q ( x, y) are smooth. Green’s Theorem relates the line integral around a closed
curve with a double integral over the region inside the curve:

Theorem 5.6. (Green’s Theorem) Let R be a region in R2 whose boundary is a simple

closed curve C which is piecewise smooth. Let f( x, y) = P ( x, y) i + Q ( x, y) j be a smooth vector
field defined on both R and C . Then
z x µ ∂Q ∂P ¶
f · dr = − dA , (5.21)
∂x ∂ y
C R

where C is traversed so that R is always on the left side of C .

Proof: We will prove the theorem in the case for a simple region R , that is, where the
boundary curve C can be written as C = C 1 ∪ C 2 in two distinct ways:

C 1 = the curve y = y1 ( x) from the point X 1 to the point X 2 (5.22)

C 2 = the curve y = y2 ( x) from the point X 2 to the point X 1 , (5.23)

where X 1 and X 2 are the points on C farthest to the left and right, respectively; and

C 1 = the curve x = x1 ( y) from the point Y2 to the point Y1 (5.24)

C 2 = the curve x = x2 ( y) from the point Y1 to the point Y2 , (5.25)

where Y1 and Y2 are the lowest and highest points, respectively, on C . See Figure 4.3.1.
y

y = y2 (x)
d
Y2
Î
X 2 x = x2 (y)
x = x1 (y) X 1 R

ÏC
Y1
c
y = y1 (x)
x
a b
Figure 5.3.1

Integrate P ( x, y) around C using the representation C = C 1 ∪ C 2 given by (4.23) and (4.24).

Since y = y1 ( x) along C 1 (as x goes from a to b) and y = y2 ( x) along C 2 (as x goes from b to
 5.3 Green’s Theorem 173

a), as we see from Figure 4.3.1, then we have

z w w
P ( x, y) dx = P ( x, y) dx + P ( x, y) dx
C C1 C2
wb wa
= P ( x, y1 ( x)) dx + P ( x, y2 ( x)) dx
a b
wb wb
= P ( x, y1 ( x)) dx − P ( x, y2 ( x)) dx
a a

wb
= − (P ( x, y2 ( x)) − P ( x, y1 ( x))) dx
a

wb µ ¯ y= y2 ( x) ¶
¯
= − P ( x, y) ¯ dx
y= y1 ( x)
a

wb yw2 ( x) ∂P ( x, y)
= − d y dx (by the Fundamental Theorem of Calculus)
a y1 ( x)
∂y

x ∂P
= − dA .
∂y
R

Likewise, integrate Q ( x, y) around C using the representation C = C 1 ∪ C 2 given by (4.25)

and (4.26). Since x = x1 ( y) along C 1 (as y goes from d to c) and x = x2 ( y) along C 2 (as y goes
from c to d ), as we see from Figure 4.3.1, then we have
z w w
Q ( x, y) d y = Q ( x, y) d y + Q ( x, y) d y
C C1 C2
wc wd
= Q ( x1 ( y), y) d y + Q ( x2 ( y), y) d y
d c
wd wd
= − Q ( x1 ( y), y) d y + Q ( x2 ( y), y) d y
c c

wd
= (Q ( x2 ( y), y) − Q ( x1 ( y), y)) d y
c

wd µ ¯ x= x2 ( y) ¶
¯
= Q ( x, y) ¯ dy
x= x1 ( y)
c
174 CHAPTER 5. LINE AND SURFACE INTEGRALS

wd xw2 ( y) ∂Q ( x, y)
= dx d y (by the Fundamental Theorem of Calculus)
c x1 ( y)
∂x

x ∂Q
= d A , and so
∂x
R

z z z
f · dr = P ( x, y) dx + Q ( x, y) d y
C C C
x ∂P x ∂Q
= − dA + dA
∂y ∂x
R R
x µ ∂Q ∂P
¶
= − dA .
∂x ∂y
R

QED

Example 5.7. Evaluate z

( x2 + y2 ) dx + 2 x y d y,
C

where C is the boundary (traversed counterclockwise) of the region R = { ( x, y) : 0 ≤ x ≤

1, 2 x2 ≤ y ≤ 2 x }.
Solution: R is the shaded region in Figure 4.3.2. By Green’s Theorem, for y
(1, 2)
P ( x, y) = x2 + y2 and Q ( x, y) = 2 x y, we have 2

z x µ ∂Q ∂P ¶
2 2 C
( x + y ) dx + 2 x y d y = − dA
∂x ∂ y
C R
x x x
= (2 y − 2 y) d A = 0dA = 0 . 0 1
R R
Figure 5.3.2

We actually already knew that the answer was zero. Recall from Example 5.5 in Section
4.2 thatuthe vector field f( x, y) = ( x2 + y2 ) i + 2 x y j has a potential function F ( x, y) = 13 x3 + x y2 ,
and so C f · d r = 0 by Corollary 5.5.

Though we proved Green’s Theorem only for a simple region R , the theorem can also
be proved for more general regions; in particular to regions which admit subdivision into
 5.3 Green’s Theorem 175

simple regions.1 It includes regions bounded by few closed curves. For such regions, the
“outer” boundary and the “inner” boundaries are traversed so that R is always on the left
side.

C1 Î C1
R1 R1

Î
Ï C2 C3 C2

Ï
Î

Î
R2 R2
Ï
Ï

(a) Region R with one hole (b) Region R with two holes

Figure 5.3.3 Multiply connected regions

The idea for why Green’s Theorem holds for such regions is shown in Figure 5.3.3 above.
The idea is to cut region R so that it is divided into simple subregions. For example, in
Figure 5.3.3(a) the region R is the union of the regions R 1 and R 2 , which are divided by the
slits indicated by the dashed lines. Those slits are part of the boundary of both R 1 and R 2 ,
and we traverse then in the manner indicated by the arrows. Notice that along each slit the
boundary of R 1 is traversed in the opposite direction as that of R 2 , which means that the line
integrals of f along those slits cancel each other out. Assuming that Green’s Theorem holds
for R 1 and R 2 , we get
z x µ ∂Q ∂P ¶ z x µ ∂Q ∂P ¶
f · dr = − d A and f · dr = − dA .
∂x ∂ y ∂x ∂ y
bdy R1 bdy R2
of R 1 of R 2

But since the line integrals along the slits cancel out, we have
z z z
f · dr = f · dr + f · dr ,
C 1 ∪C 2 bdy bdy
of R 1 of R 2

and so
z x µ ∂Q ∂P
¶ x µ ∂Q ∂P
¶ x µ ∂Q ∂P
¶
f · dr = − dA + − dA = − dA ,
∂x ∂y ∂x ∂y ∂x ∂y
C 1 ∪C 2 R1 R2 R

which shows that Green’s Theorem holds in the region R . A similar argument shows that
the theorem holds in the region with two holes shown in Figure 5.3.3(b).
1 See T AYLOR and M ANN, § 15.31 for a discussion of some of the difficulties involved when the boundary curve

is “complicated”.
176 CHAPTER 5. LINE AND SURFACE INTEGRALS

Example 5.8. Let f( x, y) = P ( x, y) i + Q ( x, y) j, where

−y x
P ( x, y) = and Q ( x, y) = ,
x 2 + y2 x 2 + y2

and let R = { ( x, y) : 0 < x2 + y2 ≤ 1 }. For the boundary curve Cu: x2 + y2 = 1, traversed counter-
clockwise, it was shown in Exercise 9(b) in Section 4.2 that C f · d r = 2π. But

∂Q y2 − x 2 ∂P x µ ∂Q ∂P ¶ x
= 2 = ⇒ − d A = 0dA = 0 .
∂x ( x + y2 )2 ∂y ∂x ∂ y
R R

This would seem to contradict Green’s Theorem. However, note that R is not the entire
region enclosed by C , since the point (0, 0) is not contained in R . That is, R has a “hole” at
the origin, so Green’s Theorem does not apply.
y
If we modify the region R to be the annulus R =
{ ( x, y) : 1/4 ≤ x2 + y2 ≤ 1 } (see Figure 4.3.3), and take 1
the “boundary” C of R to be C = C 1 ∪ C 2 , where C 1 is C1
the unit circle x2 + y2 = 1 traversed counterclockwise

Î
R
and C 2 is the circle x2 + y2 = 1/4 traversed clockwise, 1/2
then it can be shown (see Exercise 8) that C2

Ï
x
z 0 1/2 1
f · dr = 0 .
C

s ³ ∂Q ´
We would still have ∂x
− ∂∂Py d A = 0, so for this R
R
we would have
Figure 5.3.4 The annulus R
z x µ ∂Q ∂P
¶
f · dr = − dA ,
∂x ∂y
C R

which shows that Green’s Theorem holds for the annular region R .

We know from Corollary 5.5 that when a smooth vector field f( x, y) = P ( x, y) i + Q ( x, y) j on

a regionuR (whose boundary is a piecewise smooth, simple closed curve C ) has a potential in
R , then C f · d r = 0. And if the potential F ( x, y) is smooth in R , then ∂∂Fx = P and ∂∂Fy = Q , and
so we know that
∂2 F ∂2 F ∂P ∂Q
= ⇒ = in R .
∂ y ∂x ∂x ∂ y ∂y ∂x
∂P ∂Q
Conversely, if ∂y
= ∂x
in R then

z x µ ∂Q ∂P
¶ x
f · dr = − dA = 0dA = 0 .
∂x ∂y
C R R
 5.3 Green’s Theorem 177

For a simply connected region R (that is, a region with no holes), the following can be
shown:

The following statements are equivalent for a simply connected region R in R2 :

(a) f( x, y) = P ( x, y) i + Q ( x, y) j has a smooth potential F ( x, y) in R

w
(b) f · d r is independent of the path for any curve C in R
C
z
(c) f · d r = 0 for every simple closed curve C in R
C

∂P ∂Q
(d) = in R (in this case, the differential form P dx + Q d y is exact)
∂y ∂x

A
For Exercises 1–4, use Green’s Theorem to evaluate the given line integral around the curve
C , traversed counterclockwise.
z
1. ( x2 − y2 ) dx + 2 x y d y; C is the boundary of R = { ( x, y) : 0 ≤ x ≤ 1, 2 x2 ≤ y ≤ 2 x }
C
z
2. x2 y dx + 2 x y d y; C is the boundary of R = { ( x, y) : 0 ≤ x ≤ 1, x2 ≤ y ≤ x }
C
z
3. 2 y dx − 3 x d y; C is the circle x2 + y2 = 1
C
z 2 2
4. ( e x + y2 ) dx + ( e y + x2 ) d y; C is the boundary of the triangle with vertices (0, 0), (4, 0)
C
and (0, 4)

5. Is there a potential F ( x, y) for f( x, y) = ( y2 + 3 x2 ) i + 2 x y j? If so, find one.

6. Is there a potential F ( x, y) for f( x, y) = ( x3 cos( x y) + 2 x sin( x y)) i + x2 y cos( x y) j? If so, find

one.

7. Is there a potential F ( x, y) for f( x, y) = (8 x y + 3) i + 4( x2 + y) j? If so, find one.

8. Show that z
a dx + b d y = 0
C

for any constants a, b and any closed simple curve C .

B
178 CHAPTER 5. LINE AND SURFACE INTEGRALS

u
9. For the vector field f as in Example 5.8, show directly that C f · d r = 0, where C is the
boundary of the annulus R = { ( x, y) : 1/4 ≤ x2 + y2 ≤ 1 } traversed so that R is always on
the left.

10. Evaluate z
e x sin y dx + ( y3 + e x cos y) d y,
C

where C is the boundary of the rectangle with vertices (1, −1), (1, 1), (−1, 1) and (−1, −1),
traversed counterclockwise.

C
11. For a region R bounded by a simple closed curve C , show that the area A of R is
z z 1z
A = − y dx = xdy = x d y − y dx ,
2
C C C

where C is traversed
s so that R is always on the left. (Hint: Use Green’s Theorem and the
fact that A = R 1 d A .)

In the following exercises, use Exercise 11 to find the area bounded by curve. (You should
figure out how the curve traversed around the region it bounds.)

12. The curve (sin t, sin(2 t)) for 0 ≤ t ≤ π.

13. The deltoid curve (2 cos t + cos 2 t, 2 sin t − sin 2 t) for 0 ≤ t ≤ 2π. (The deltoid curve is shown
on the diagram; you can assume without proof that it has no self-intesections.)
 5.4 Surface Integrals and the Divergence Theorem 179

5.4 Surface Integrals and the Divergence Theorem

In Section 4.1 we learned how to integrate along a curve. We will now learn how to perform
integration over a surface in R3 , such as a sphere or a paraboloid. Recall from Section 1.8
how we identified points ( x, y, z) on a curve C in R3 , parametrized by x = x( t), y = y( t), z = z( t),
a ≤ t ≤ b, with the terminal points of the position vector

r( t) = x( t)i + y( t)j + z( t)k for t in [a, b].

The idea behind a parametrization of a curve is that it “transforms” a subset of R1 (nor-

mally an interval [a, b]) into a curve in R2 or R3 (see Figure 5.4.1).

z
(x(t), y(t), z(t))
(x(a), y(a), z(a))
x = x(t) C
y = y(t) r(t) (x(b), y(b), z(b))
z = z(t)
R1 y
a t b 0
x

Figure 5.4.1 Parametrization of a curve C in R3

Similar to how we used a parametrization of a curve to define the line integral along the
curve, we will use a parametrization of a surface to define a surface integral. We will use
two variables, u and v, to parametrize a surface Σ in R3 : x = x( u, v), y = y( u, v), z = z( u, v),
for ( u, v) in some region R in R2 (see Figure 5.4.2).

v R2
z
Σ

R x = x(u, v)
y = y(u, v)
(u, v) z = z(u, v) r(u, v)
y
0
u x

Figure 5.4.2 Parametrization of a surface Σ in R3

In this case, the position vector of a point on the surface Σ is given by the vector-valued
180 CHAPTER 5. LINE AND SURFACE INTEGRALS

function
r( u, v) = x( u, v)i + y( u, v)j + z( u, v)k for ( u, v) in R .
∂r ∂r
Since r( u, v) is a function of two variables, define the partial derivatives ∂u
and ∂v
for ( u, v)
in R by

∂r ∂x ∂y ∂z
( u, v) = ( u, v)i +
( u, v)j + ( u, v)k , and
∂u ∂u ∂u ∂u
∂r ∂x ∂y ∂z
( u, v) = ( u, v)i + ( u, v)j + ( u, v)k .
∂v ∂v ∂v ∂v

The parametrization of Σ can be thought of as “transforming” a region in R2 (in the uv-

plane) into a 2-dimensional surface in R3 . This parametrization of the surface is sometimes
called a patch, based on the idea of “patching” the region R onto Σ in the grid-like manner
shown in Figure 5.4.2.
In fact, those gridlines in R lead us to how we will define a surface integral over Σ. Along
the vertical gridlines in R , the variable u is constant. So those lines get mapped to curves on
Σ, and the variable u is constant along the position vector r( u, v). Thus, the tangent vector
to those curves at a point ( u, v) is ∂∂rv . Similarly, the horizontal gridlines in R get mapped to
curves on Σ whose tangent vectors are ∂∂ur .
Now take a point ( u, v) in R as, say, the lower left corner of one of the rectangular grid
sections in R , as shown in Figure 5.4.2. Suppose that this rectangle has a small width and
height of ∆ u and ∆v, respectively. The corner points of that rectangle are ( u, v), ( u + ∆ u, v),
( u + ∆ u, v + ∆v) and ( u, v + ∆v). So the area of that rectangle is A = ∆ u ∆v. Then that rectangle
gets mapped by the parametrization onto some section of the surface Σ which, for ∆ u and
∆v small enough, will have a surface area that is very close to the area of the parallelogram
which has adjacent sides r( u + ∆ u, v) − r( u, v) (corresponding to the line segment from ( u, v)
to ( u + ∆ u, v) in R ) and r( u, v + ∆v) − r( u, v) (corresponding to the line segment from ( u, v)
to ( u, v + ∆v) in R ). Combining our usual notion of a partial derivative (see Definition 3.3
in Section 2.2) with that of the derivative of a vector-valued function (see Definition 2.3 in
Section 1.8) applied to a function of two variables, we have

∂r r( u + ∆ u, v) − r( u, v)
≈ , and
∂u ∆u
∂r r( u, v + ∆v) − r( u, v)
≈ ,
∂v ∆v
and so the surface area element d σ is approximately
° ° ° °
° (r( u + ∆ u, v) − r( u, v)) × (r( u, v + ∆v) − r( u, v)) ° ≈ ° (∆ u ∂r ) × (∆v ∂r ) ° = ° ∂r × ∂r ° ∆ u ∆v
° ° ° ° ° °
° ∂u ∂v ° ° ∂ u ∂v °

by Theorem 1.13 in Section

° 1.4.
° Thus, the total surface area S of Σ is approximately the sum
of all the quantities ° ∂∂ur × ∂∂rv ° ∆ u ∆v, summed over the rectangles in R . Taking the limit of
 5.4 Surface Integrals and the Divergence Theorem 181

that sum as the diagonal of the largest rectangle goes to 0 gives

x°
° ∂r
°
∂r °
S = ° ∂ u × ∂v ° du dv . (5.26)
° °
R

We will write the double integral on the right using the special notation
x x°
° ∂r
°
∂r °
dσ = ° ∂ u × ∂v ° du dv . (5.27)
° °
Σ R

This is a special case of a surface integral over the surface Σ, where the surface area element
d σ can be thought of as 1 d σ. Replacing 1 by a general real-valued function f ( x, y, z) defined
in R3 , we have the following:

Definition 5.3. Let Σ be a surface in R3 parametrized by x = x( u, v), y = y( u, v),

z = z( u, v), for ( u, v) in some region R in R2 . Let r( u, v) = x( u, v)i + y( u, v)j + z( u, v)k be the
position vector for any point on Σ, and let f ( x, y, z) be a real-valued function defined on some
subset of R3 that contains Σ. The surface integral of f ( x, y, z) over Σ is
x x °
° ∂r ∂r °
°
f ( x, y, z) d σ = f ( x( u, v), y( u, v), z( u, v)) °
° × ° du dv . (5.28)
∂ u ∂v °
Σ R

In particular, the surface area S of Σ is

x
S = 1 dσ . (5.29)
Σ

Example 5.9. A torus T is a surface obtained by revolving a circle of radius a in the yz-plane
around the z-axis, where the circle’s center is at a distance b from the z-axis (0 < a < b), as
in Figure 5.4.3. Find the surface area of T .
Solution: For any point on the circle, the line segment from the center of the circle to that
point makes an angle u with the y-axis in the positive y direction (see Figure 5.4.3(a)). And
as the circle revolves around the z-axis, the line segment from the origin to the center of that
circle sweeps out an angle v with the positive x-axis (see Figure 5.4.3(b)). Thus, the torus
can be parametrized as:

x = ( b + a cos u) cos v , y = ( b + a cos u) sin v , z = a sin u , 0 ≤ u ≤ 2π , 0 ≤ v ≤ 2π

So for the position vector

r( u, v) = x( u, v)i + y( u, v)j + z( u, v)k

= ( b + a cos u) cos v i + ( b + a cos u) sin v j + a sin u k
182 CHAPTER 5. LINE AND SURFACE INTEGRALS

z
z
(y − b)2 + z2 = a2
y
a (x, y, z)
u y v
0 a

b x
(a) Circle in the yz-plane (b) Torus T

Figure 5.4.3

we see that
∂r
= −a sin u cos v i − a sin u sin v j + a cos u k
∂u
∂r
= −( b + a cos u) sin v i + ( b + a cos u) cos v j + 0k ,
∂v
and so computing the cross product gives
∂r ∂r
× = −a( b + a cos u) cos v cos u i − a( b + a cos u) sin v cos u j − a( b + a cos u) sin u k ,
∂u ∂v
which has magnitude ° °
° ∂r ∂r °
° ∂ u × ∂v ° = a( b + a cos u) .
° °

Thus, the surface area of T is

x
S = 1 dσ
Σ

w2πw2π°
° ∂r
°
∂r °
= ° ∂ u × ∂v ° du dv
° °
0 0

w2πw2π
= a( b + a cos u) du dv
0 0

w2πµ ¯u=2π ¶
2 ¯
= abu + a sin u ¯ dv
u =0
0

w2π
= 2πab dv
0
 5.4 Surface Integrals and the Divergence Theorem 183

= 4π2 ab

z
Assume that a surface Σ is given by a collection of charts.
Note that for each chart r( u, v) in the collection, the vectors ∂∂ur
and ∂∂rv are tangent to the surface. Therefore, their crossproduct
∂r
∂u
( u, v) × ∂∂rv ( u, v) is normal to Σ at the point with position vector y
r( u, v). 0
Assume further that at each point P of the surface Σ one can
choose a unit normal vector n in such a way such that for every
x
chart r( u, v) in the collection n at the point with position vector
Figure 5.4.4
r( u, v), the crossproduct ∂∂ur ( u, v) × ∂∂rv ( u, v) and is codirectional
with n. In this case Σ is called oriented and the vector field n is
called outward unit normal vector of Σ.

Definition 5.4. Let Σ be an oriented surface in R3 and let f( x, y, z) be a vector field defined
on some subset of R3 that contains Σ. The surface integral of f over Σ is
x x
f · dσ = f · n dσ , (5.30)
Σ Σ

where, at any point on Σ, n is the outward unit normal vector to Σ.

In particular, if Σ is given by a single chart r( u, v) = x( u, v)i + y( u, v)j + z( u, v)k defined
on a plane region R then
x x ¡ ∂r ¢
f · dσ = f( x( u, v), y( u, v), z( u, v)) · ∂u
( u, v) × ∂∂rv ( u, v) du dv.
Σ R

Note in the above definition that the dot product inside the integral on the right is a real-
valued function, and hence we can use Definition 5.3 to evaluate the integral.

s
Example 5.10. Evaluate the surface integral f · d σ, where f( x, y, z) = yzi + xzj + x yk and Σ
Σ
is the part of the plane x + y + z = 1 with x ≥ 0, y ≥ 0, and z ≥ 0, with the outward unit normal
n pointing in the positive z direction (see Figure 4.4.5).
184 CHAPTER 5. LINE AND SURFACE INTEGRALS

Solution: Since the vector v = (1, 1, 1) is normal to the plane x + y + z = 1 z

1
(why?), then ³ dividing´ v by its length yields the outward unit normal n
vector n = p1 , p1 , p1 . We now need to parametrize Σ. As we can see
3 3 3 Σ
from Figure 4.4.5, projecting Σ onto the x y-plane yields a triangular 0
y
region R = { ( x, y) : 0 ≤ x ≤ 1, 0 ≤ y ≤ 1 − x }. Thus, using ( u, v) instead of 1
( x, y), we see that 1 x+ y+ z = 1
x
x = u, y = v, z = 1 − ( u + v), for 0 ≤ u ≤ 1, 0 ≤ v ≤ 1 − u Figure 5.4.5

is a parametrization of Σ over R (since z = 1 − ( x + y) on Σ). So on Σ,

µ ¶
1 1 1 1
f · n = ( yz, xz, x y) · p , p , p = p ( yz + xz + x y)
3 3 3 3
1 1
= p (( x + y) z + x y) = p (( u + v)(1 − ( u + v)) + uv)
3 3
1
= p (( u + v) − ( u + v)2 + uv)
3
for ( u, v) in R , and for r( u, v) = x( u, v)i + y( u, v)j + z( u, v)k = ui + vj + (1 − ( u + v))k we have
° °
∂r ∂r ° ∂r ∂r ° p
× = (1, 0, −1) × (0, 1, −1) = (1, 1, 1) ⇒ ° ° × °= 3 .
∂ u ∂v ∂ u ∂v °
Thus, integrating over R using vertical slices (indicated by the dashed line in Figure 4.4.5)
gives
x x
f · dσ = f · n dσ
Σ Σ
x °
° ∂r ∂r °
°
= (f( x( u, v), y( u, v), z( u, v)) · n) °
° × ° dv du
∂ u ∂v °
R

w1 1w−u 1 p
= p (( u + v) − ( u + v)2 + uv) 3 dv du
0 0
3

w1 (u + v)2
à ¯v=1−u !
( u + v)3 uv2 ¯¯
= − + du
2 3 2 ¯ v =0
0

w1 µ 1 u 3 u2 5 u3
¶
= + − + du
6 2 2 6
0

¯1
u u2 u3 5 u4 ¯¯ 1
= + − + ¯ = .
6 4 2 24 0 8
 5.4 Surface Integrals and the Divergence Theorem 185

Computing surface integrals can often be tedious, especially when the formula for the
outward unit normal vector at each point of Σ changes. The following theorem provides an
easier way in the case when Σ is a closed surface, that is, when Σ encloses a bounded
solid in R3 . For example, spheres, cubes, and ellipsoids are closed surfaces, but planes and
paraboloids are not.

Theorem 5.7. (Divergence Theorem) Let Σ be a closed surface in R3 which bounds a

solid S , and let f( x, y, z) = f 1 ( x, y, z)i + f 2 ( x, y, z)j + f 3 ( x, y, z)k be a vector field defined on some
subset of R3 that contains Σ. Then
x y
f · dσ = div f dV , (5.31)
Σ S

where
∂ f1 ∂ f2 ∂ f3
div f = + + (5.32)
∂x ∂y ∂z
is called the divergence of f.

The proof of the Divergence Theorem is very similar to the proof of Green’s Theorem. It is
first proved for the simple case when the solid S is bounded above by one surface, bounded
below by another surface, and bounded laterally by one or more surfaces. The proof can then
be extended to more general solids.2

s
Example 5.11. Evaluate f · d σ, where f( x, y, z) = xi + yj + zk and Σ is the unit sphere
Σ
x2 + y2 + z2 = 1.
Solution: We see that div f = 1 + 1 + 1 = 3, so
x y y
f · dσ = div f dV = 3 dV
Σ S S
y 4π(1)3
= 3 1 dV = 3 vol(S ) = 3 · = 4π .
3
S

s
In physical applications, the surface integral f · d σ is often referred to as the flux of f
Σ
through the surface Σ. For example, if f represents the velocity field of a fluid, then the flux
is the net quantity of fluid to flow through the surface Σ per unit time. A positive flux means
there is a net flow out of the surface (that is, in the direction of the outward unit normal
vector n), while a negative flux indicates a net flow inward (in the direction of −n).
2 See T AYLOR and M ANN, § 15.6 for the details.
186 CHAPTER 5. LINE AND SURFACE INTEGRALS

The term divergence comes from interpreting div f as a measure of how much a vector
field “diverges” from a point. This is best seen by using another definition of div f which is
equivalent3 to the definition given by formula (5.32). Namely, for a point ( x, y, z) in R3 ,

1x
div f( x, y, z) = lim f · dσ , (5.33)
V →0 V
Σ

where V is the volume enclosed by a closed surface Σ around the point ( x, y, z). In the
limit, V → 0 means that we take smaller and smaller closed surfaces around ( x, y, z), which
means that the volumes they enclose are going to zero. It can be shown that this limit is
independent of the shapes of those surfaces. Notice that the limit being taken is of the
ratio of the flux through a surface to the volume enclosed by that surface, which gives a
rough measure of the flow “leaving” a point, as we mentioned. Vector fields which have zero
divergence are often called solenoidal fields.
The following theorem is a simple consequence of formula (5.33).

Theorem 5.8. If the flux of a vector field f is zero through every closed surface containing a
given point, then div f = 0 at that point.

Proof: By formula (5.33), at the given point ( x, y, z) we have

1x
div f( x, y, z) = lim f · d σ for closed surfaces Σ containing ( x, y, z), so
V →0 V
Σ
1
= lim (0) by our assumption that the flux through each Σ is zero, so
V →0 V

= lim 0
V →0
= 0. QED

Lastly, we note that sometimes the notation

{ {
f ( x, y, z) d σ and f · dσ
Σ Σ

is used to denote surface integrals of scalar and vector fields, respectively, over closed sur-
faces.

Exercises
A
3 See S CHEY, p. 36–39, for an intuitive discussion of this.
 5.4 Surface Integrals and the Divergence Theorem 187

For Exercises 1–4, use the Divergence Theorem to evaluate the surface integral
x
f · dσ
Σ

of the given vector field f( x, y, z) over the surface Σ.

1. f( x, y, z) = xi + 2 yj + 3 zk, Σ : x2 + y2 + z2 = 9

2. f( x, y, z) = xi + yj + zk, Σ : boundary of the solid cube S = { ( x, y, z) : 0 ≤ x, y, z ≤ 1 }

3. f( x, y, z) = x3 i + y3 j + z3 k, Σ : x2 + y2 + z2 = 1

4. f( x, y, z) = 2i + 3j + 5k, Σ : x2 + y2 + z2 = 1

B
5. Show that the flux of any constant vector field through any closed surface is zero.

6. Evaluate the surface integral from Exercise 2 without using the Divergence Theorem;
that is, using only Definition 5.3, as in Example 5.10. Note that there will be a different
outward unit normal vector to each of the six faces of the cube.
s
7. Evaluate the surface integral f · d σ, where f( x, y, z) = x2 i + x yj + zk and Σ is the part of
Σ
the plane 6 x + 3 y + 2 z = 6 with x ≥ 0, y ≥ 0, and z ≥ 0, with the outward unit normal n
pointing in the positive z direction.

8. Use a surface integral to show that the surface area of a sphere of radius r is 4π r 2 . (Hint:
Use spherical coordinates to parametrize the sphere.)

9. Use a surface integral

p to show that the surface area of a right circular cone of radiush R
and height h is πR h2 + R 2 . (Hint: Use the parametrization x = r cos θ , y = r sin θ , z = R r ,
for 0 ≤ r ≤ R and 0 ≤ θ ≤ 2π.)
x2 y2 2
10. The ellipsoid a2
+ b2 + zc2 = 1 can be parametrized using ellipsoidal coordinates

x = a sin φ cos θ , y = b sin φ sin θ , z = c cos φ , for 0 ≤ θ ≤ 2π and 0 ≤ φ ≤ π.

Show that the surface area S of the ellipsoid is

wπ w2π q
S = sin φ a2 b2 cos2 φ + c2 (a2 sin2 θ + b2 cos2 θ ) sin2 φ d θ d φ .
0 0

(Note: The above double integral can not be evaluated by elementary means. For specific
values of a, b and c it can be evaluated using numerical methods. An alternative is to
express the surface area in terms of elliptic integrals.4 )
4 B OWMAN, F., Introduction to Elliptic Functions, with Applications, New York: Dover, 1961, § III.7.
188 CHAPTER 5. LINE AND SURFACE INTEGRALS

C
11. Use Definition 5.3 to prove that the surface area S over a region R in R2 of a surface
z = f ( x, y) is given by the formula
xr ³ ´ ³ ´2
∂f 2 ∂f
S = 1+ ∂x
+ ∂y dA .
R

(Hint: Think of the parametrization of the surface.)

 5.5 Stokes’ Theorem 189

5.5 Stokes’ Theorem

So far the only types of line integrals which we have discussed are those along curves in R2 .
But the definitions and properties which were covered in Sections 4.1 and 4.2 can easily be
extended to include functions of three variables, so that we can now discuss line integrals
along curves in R3 .

Definition 5.5. For a real-valued function f ( x, y, z) and a curve C in R3 , parametrized by

x = x( t), y = y( t), z = z( t), a ≤ t ≤ b, the line integral of f ( x, y, z) along C with respect to
arc length s is

w wb q
f ( x, y, z) ds = f ( x( t), y( t), z( t)) x ′ ( t)2 + y ′ ( t)2 + z ′ ( t)2 dt . (5.34)
C a

The line integral of f ( x, y, z) along C with respect to x is

w wb
f ( x, y, z) dx = f ( x( t), y( t), z( t)) x ′ ( t) dt . (5.35)
C a

The line integral of f ( x, y, z) along C with respect to y is

w wb
f ( x, y, z) d y = f ( x( t), y( t), z( t)) y ′ ( t) dt . (5.36)
C a

The line integral of f ( x, y, z) along C with respect to z is

w wb
f ( x, y, z) dz = f ( x( t), y( t), z( t)) z ′ ( t) dt . (5.37)
C a

r
Similar to the two-variable case, if f ( x, y, z) ≥ 0 then the line integral C f ( x, y, z) ds can be
thought of as the total area of the “picket fence” of height f ( x, y, z) at each point along the
curve C in R3 .

Vector fields in R3 are defined in a similar fashion to those in R2 , which allows us to define
the line integral of a vector field along a curve in R3 .
190 CHAPTER 5. LINE AND SURFACE INTEGRALS

Definition 5.6. For a vector field f( x, y, z) = P ( x, y, z) i + Q ( x, y, z) j + R ( x, y, z) k and a curve C

in R3 with a smooth parametrization x = x( t), y = y( t), z = z( t), a ≤ t ≤ b, the line integral
of f along C is
w w w w
f · dr = P ( x, y, z) dx + Q ( x, y, z) d y + R ( x, y, z) dz (5.38)
C C C C
wb
= f( x( t), y( t), z( t)) · r ′ ( t) dt , (5.39)
a

where r( t) = x( t) i + y( t) j + z( t) k is the position vector for points on C .

Similar to the two-variable case,rif f( x, y, z) represents the force applied to an object at a

point ( x, y, z) then the line integral C f · d r represents the work done by that force in moving
the object along the curve C in R3 .
Some of the most important results we will need for line integrals in R3 are stated below
without proof (the proofs are similar to their two-variable equivalents).

Theorem 5.9. For a vector field f( x, y, z) = P ( x, y, z) i + Q ( x, y, z) j + R ( x, y, z) k and a curve

C with a smooth parametrization x = x( t), y = y( t), z = z( t), a ≤ t ≤ b and position vector
r( t) = x( t) i + y( t) j + z( t) k, w w
f · d r = f · T ds , (5.40)
C C
′
r ( t)
where T( t) = k r ′ ( t) k is the unit tangent vector to C at ( x( t), y( t), z( t)).

Theorem 5.10. (Chain Rule) If w = f ( x, y, z) is a continuously differentiable function of

x, y, and z, and x = x( t), y = y( t) and z = z( t) are differentiable functions of t, then w is a
differentiable function of t, and

dw ∂w dx ∂w d y ∂w dz
= + + . (5.41)
dt ∂ x dt ∂ y dt ∂ z dt

Also, if x = x( t 1 , t 2 ), y = y( t 1 , t 2 ) and z = z( t 1 , t 2 ) are continuously differentiable function of

( t 1 , t 2 ), then5
∂w ∂w ∂ x ∂w ∂ y ∂w ∂ z
= + + (5.42)
∂ t1 ∂ x ∂ t1 ∂ y ∂ t1 ∂ z ∂ t1
and
∂w ∂w ∂ x ∂w ∂ y ∂w ∂ z
= + + . (5.43)
∂ t2 ∂ x ∂ t2 ∂ y ∂ t2 ∂ z ∂ t2

5 See T AYLOR and M ANN, § 6.5 for a proof.

 5.5 Stokes’ Theorem 191

Theorem 5.11. Let f( x, y, z) = P ( x, y, z) i + Q ( x, y, z) j + R ( x, y, z) k be a vector field in some

solid S , with P , Q and R continuously differentiable functions on S . Let C be a smooth
curve in S parametrized by x = x( t), y = y( t), z = z( t), a ≤ t ≤ b. Suppose that there is a
real-valued function F ( x, y, z) such that ∇F = f on S . Then
w
f · d r = F (B ) − F ( A ) , (5.44)
C

where A = ( x(a), y(a), z(a)) and B = ( x( b), y( b), z( b)) are the endpoints of C .

z
Corollary 5.12. If a vector field f has a potential in a solid S , then f · d r = 0 for any closed
z C
curve C in S (that is, ∇F · d r = 0 for any real-valued function F ( x, y, z)).
C

Example 5.12. Let f ( x, y, z) = z and let C be the curve in R3 parametrized by

x = t sin t , y = t cos t , z=t , 0 ≤ t ≤ 8π .

r
Evaluate C f ( x, y, z) ds. (Note: C is called a conical helix. See Figure 5.5.1).
Solution: Since x ′ ( t) = sin t + t cos t, y ′ ( t) = cos t − t sin t, and z ′ ( t) = 1, we have

x ′ ( t)2 + y ′ ( t)2 + z ′ ( t)2 = (sin2 t + 2 t sin t cos t + t2 cos2 t) + (cos2 t − 2 t sin t cos t + t2 sin2 t) + 1
= t2 (sin2 t + cos2 t) + sin2 t + cos2 t + 1
= t2 + 2 ,

so since f ( x( t), y( t), z( t)) = z( t) = t along the curve C , then

w w8π q
f ( x, y, z) ds = f ( x( t), y( t), z( t)) x ′ ( t)2 + y ′ ( t)2 + z ′ ( t)2 dt
C 0
w8π p
= t t2 + 2 dt
0
µ ¶ ¯8π
1 2 3/2 ¯
¯ 1³ 2 3/2
p ´
= ( t + 2) ¯ = (64 π + 2) − 2 2 .
3 0 3

Example 5.13. Let f( x, y, z) =rx i + y j + 2 z k be a vector field in R3 . Using the same curve C
from Example 5.12, evaluate C f · d r.
192 CHAPTER 5. LINE AND SURFACE INTEGRALS

30

25
t = 8π
20

15
z
10

5 -25
t=0 -20
-15
0-25 -10
-5
-20 -15 -10 0
-5 5 x
0 5 10 10
15 20 15
y 25 30 2520

Figure 5.5.1 Conical helix C

x2 y2
Solution: Note that F ( x, y, z) = 2 + 2 + z2 is a potential for f( x, y, z) (that is, ∇F = f). So by
Theorem 5.11 we know that
w
f · d r = F (B) − F ( A ) , where A = ( x(0), y(0), z(0)) and B = ( x(8π), y(8π), z(8π)), so
C
= F (8π sin 8π, 8π cos 8π, 8π) − F (0 sin 0, 0 cos 0, 0)
= F (0, 8π, 8π) − F (0, 0, 0)
(8π)2
= 0+ + (8π)2 − (0 + 0 + 0) = 96π2 .
2

We will now discuss a generalization of Green’s Theorem in R2 to orientable surfaces in

R , called Stokes’ Theorem. A surface Σ in R3 is orientable if there is a continuous vector
3

field N in R3 such that N is nonzero and normal to Σ (that is, perpendicular to the tangent
plane) at each point of Σ. We say that such an N is a normal vector field.
 5.5 Stokes’ Theorem 193

z
For example, the unit sphere x2 + y2 + z2 = 1 is orientable, since the N
continuous vector field N( x, y, z) = x i + y j + z k is nonzero and normal
to the sphere at each point. In fact, −N( x, y, z) is another normal −N
vector field (see Figure 4.5.2). We see in this case that N( x, y, z) is y
what we have called an outward normal vector, and −N( x, y, z) is an 0
inward normal vector. These “outward” and “inward” normal vec-
tor fields on the sphere correspond to an “outer” and “inner” side,
x
respectively, of the sphere. That is, we say that the sphere is a two-
Figure 5.5.2
sided surface. Roughly, “two-sided” means “orientable”. Other ex-
amples of two-sided, and hence orientable, surfaces are cylinders,
paraboloids, ellipsoids, and planes.
You may be wondering what kind of surface would not have two sides. An example is the
Möbius strip, which is constructed by taking a thin rectangle and connecting its ends at
the opposite corners, resulting in a “twisted” strip (see Figure 5.5.3).

AA
→

→
A B

−→

B A
(a) Connect A to A and B to B along the ends (b) Not orientable

Figure 5.5.3 Möbius strip

If you imagine walking along a line down the center of the Möbius strip, as in Figure
5.5.3(b), then you arrive back at the same place from which you started but upside down!
That is, your orientation changed even though your motion was continuous along that center
line. Informally, thinking of your vertical direction as a normal vector field along the strip,
there is a discontinuity at your starting point (and, in fact, at every point) since your vertical
direction takes two different values there. The Möbius strip has only one side, and hence is
nonorientable.6
For an orientable surface Σ which has a boundary curve C , pick a unit normal vector n
such that if you walked along C with your head pointing in the direction of n, then the
surface would be on your left. We say in this situation that n is a positive unit normal vector
and that C is traversed n-positively. We can now state Stokes’ Theorem:

6 For further discussion of orientability, see O’N EILL, § IV.7.

194 CHAPTER 5. LINE AND SURFACE INTEGRALS

Theorem 5.13. (Stokes’ Theorem) Let Σ be an orientable surface in R3 whose boundary

is a simple closed curve C , and let f( x, y, z) = P ( x, y, z)i + Q ( x, y, z)j + R ( x, y, z)k be a smooth
vector field defined on some subset of R3 that contains Σ. Then
z x
f · dr = (curl f) · n d σ , (5.45)
C Σ

where µ ¶ µ ¶ µ ¶
∂R ∂Q ∂P ∂R ∂Q ∂P
curl f = − i + − j + − k, (5.46)
∂y ∂z ∂z ∂x ∂x ∂y
n is a positive unit normal vector over Σ, and C is traversed n-positively.

Proof: As the general case is beyond the scope of this text, we will prove the theorem only
for the special case where Σ is the graph of z = z( x, y) for some smooth real-valued function
z( x, y), with ( x, y) varying over a region D in R2 .
Projecting Σ onto the x y-plane, we see that the closed z
Σ : z = z(x, y)
curve C (the boundary curve of Σ) projects onto a closed n
curve C D which is the boundary curve of D (see Fig-
ure 4.5.4). Assuming that C has a smooth parametriza-
tion, its projection C D in the x y-plane also has a smooth
C
parametrization, say y
C D : x = x ( t ) , y = y( t ) , a ≤ t ≤ b , 0
D (x, y)
3
and so C can be parametrized (in R ) as x
CD
C : x = x( t) , y = y( t) , z = z( x( t), y( t)) , a ≤ t ≤ b ,
Figure 5.5.4
since the curve C is part of the surface z = z( x, y). Now, by the Chain Rule (Theorem 3.3),
for z = z( x( t), y( t)) as a function of t, we know that
∂z ∂z
z ′ ( t) = x ′ ( t) + y ′ ( t) ,
∂x ∂y
and so
z w
f · dr = P ( x, y, z) dx + Q ( x, y, z) d y + R ( x, y, z) dz
C C
wb µ µ
∂z ∂z
¶¶
′ ′ ′ ′
= P x ( t) + Q y ( t) + R x ( t) + y ( t) dt
a
∂x ∂y
wb µµ ∂z
¶ µ ¶
∂z ′
¶
= P +R x ′ ( t) + Q + R y ( t) dt
a
∂x ∂y
w
= P̃ ( x, y) dx + Q̃ ( x, y) d y ,
CD
 5.5 Stokes’ Theorem 195

where
∂z
P̃ ( x, y) = P ( x, y, z( x, y)) + R ( x, y, z( x, y)) ( x, y) , and
∂x
∂z
Q̃ ( x, y) = Q ( x, y, z( x, y)) + R ( x, y, z( x, y)) ( x, y)
∂y
for ( x, y) in D . Thus, by Green’s Theorem applied to the region D , we have
z x µ ∂Q̃ ∂P̃ ¶
f · dr = − dA . (5.47)
∂x ∂ y
C D

Thus,
µ ¶
∂Q̃ ∂ ∂z
= Q ( x, y, z( x, y)) + R ( x, y, z( x, y)) ( x, y) , so by the Product Rule we get
∂x ∂x ∂y
µ ¶ µ ¶
∂ ∂ ∂z ∂ ∂z
= (Q ( x, y, z( x, y))) + R ( x, y, z( x, y)) ( x, y) + R ( x, y, z( x, y)) ( x, y) .
∂x ∂x ∂y ∂x ∂ y
Now, by formula (5.42) in Theorem 5.10, we have
∂ ∂Q ∂ x ∂Q ∂ y ∂Q ∂ z
(Q ( x, y, z( x, y))) = + +
∂x ∂x ∂x ∂ y ∂x ∂z ∂x
∂Q ∂Q ∂Q ∂ z
= ·1 + ·0 +
∂x ∂y ∂z ∂x
∂Q ∂Q ∂ z
= + .
∂x ∂z ∂x
Similarly,
∂ ∂R ∂R ∂ z
(R ( x, y, z( x, y))) = + .
∂x ∂x ∂z ∂x
Thus,
µ ¶
∂Q̃ ∂Q ∂Q ∂ z ∂R ∂R ∂ z ∂ z ∂2 z
= + + + + R ( x, y, z( x, y))
∂x ∂x ∂z ∂x ∂x ∂z ∂x ∂ y ∂x ∂ y
2
∂Q ∂Q ∂ z ∂R ∂ z ∂R ∂ z ∂ z ∂ z
= + + + + R .
∂x ∂z ∂x ∂x ∂ y ∂z ∂x ∂ y ∂x ∂ y
In a similar fashion, we can calculate

∂P̃ ∂P ∂P ∂ z ∂R ∂ z ∂R ∂ z ∂ z ∂2 z
= + + + + R .
∂y ∂y ∂z ∂ y ∂ y ∂x ∂z ∂ y ∂x ∂ y ∂x
So subtracting gives
µ ¶ µ ¶ µ ¶
∂Q̃ ∂P̃ ∂Q ∂R ∂ z ∂R ∂P ∂ z ∂Q ∂P
− = − + − + − (5.48)
∂x ∂y ∂z ∂ y ∂x ∂x ∂z ∂ y ∂x ∂y
196 CHAPTER 5. LINE AND SURFACE INTEGRALS

∂2 z ∂2 z
since ∂x ∂ y
= ∂ y ∂x
by the smoothness of z = z( x, y). Hence, by equation (5.47),
z x µ µ ∂R ¶
∂Q ∂ z
µ
∂P
¶
∂R ∂ z
µ
∂Q ∂P
¶¶
f · dr = − − − − + − dA (5.49)
∂y ∂z ∂x ∂z ∂x ∂ y ∂x ∂y
C D

after factoring out a −1 from the terms in the first two products in equation (5.48).
Now, recall from Section 2.3 (see p.76) that the vector N = − ∂∂xz i − ∂∂ zy j + k is normal to the
tangent plane to the surface z = z( x, y) at each point of Σ. Thus,

N − ∂∂xz i − ∂∂ zy j + k
n = ° ° = r
°N° ¡ ¢2 ³ ´2
1 + ∂∂xz + ∂∂ zy

is, in fact, a positive unit normal vector to Σ (see Figure 4.5.4). Hence, using the
parametrization r( x, y) = x i + y j + z( x, y) k, for ( x,ry) in D , of the surface Σ, we have
¡ ¢ ³ ´2
° × ° = 1 + ∂ z 2 + ∂ z . So we see that us-
∂r ∂z ∂r ∂z
° ∂r ∂r °
∂x
= i + ∂x
k and ∂y
= j + ∂y
k, and so ∂x ∂y ∂x ∂y
ing formula (5.46) for curl f, we have
x x °
° ∂r ∂r °
°
·
(curl f) n d σ = ·
(curl f) n °° × °dA
∂x ∂ y °
Σ D
x µµ ∂R ∂Q ¶ µ ∂P ∂R ¶ µ ∂Q ∂P ¶ ¶ µ ∂ z ∂z
¶
= − i+ − j+ − k · − i− j+k dA
∂y ∂z ∂z ∂x ∂x ∂ y ∂x ∂y
D
x µ µ ∂R ∂Q ¶ ∂ z µ ∂P ∂R ¶ ∂ z µ ∂Q ∂P ¶¶
= − − − − + − dA ,
∂y ∂z ∂x ∂z ∂x ∂ y ∂x ∂ y
D

which, upon comparing to equation (5.49), proves the Theorem.

QED

Note: The condition in Stokes’ Theorem that the surface Σ have a (continuously vary-
ing) positive unit normal vector n and a boundary curve C traversed n-positively can be
expressed more precisely as follows: if r( t) is the position vector for C and T( t) = r ′ ( t)/k r ′ ( t) k
is the unit tangent vector to C , then the vectors T, n, T × n form a right-handed system.
Also, it should be noted that Stokes’ Theorem holds even when the boundary curve C is
piecewise smooth.

Example 5.14. Verify Stokes’ Theorem for f( x, y, z) = z i + x j + y k when Σ is the paraboloid

z = x2 + y2 such that z ≤ 1 (see Figure 4.5.5).
 5.5 Stokes’ Theorem 197

z
Solution: The positive unit normal vector to the surface C
z = z( x, y) = x2 + y2 is 1 n

− ∂∂xz i − ∂∂ zy j + k −2 x i − 2 y j + k
n = r = p ,
¡ ∂ z ¢2 ³
∂z
´2 1 + 4 x 2 + 4 y2
1 + ∂x + ∂ y
Σ
y
and curl f = (1 − 0) i + (1 − 0) j + (1 − 0) k = i + j + k, so 0
x
q Figure 5.5.5 z = x 2 + y2
(curl f) · n = (−2 x − 2 y + 1)/ 1 + 4 x2 + 4 y2 .

Since Σ can be parametrized as r( x, y) = x i + y j + ( x2 + y2 ) k for

( x, y) in the region D = { ( x, y) : x2 + y2 ≤ 1 }, then

x x °
° ∂r ∂r °
°
(curl f) · n d σ = ·
(curl f) n °° × °dA
∂x ∂ y °
Σ D
x −2 x − 2 y + 1 q
= p 1 + 4 x 2 + 4 y2 d A
1 + 4 x 2 + 4 y2
D
x
= (−2 x − 2 y + 1) d A , so switching to polar coordinates gives
D
w2πw1
= (−2 r cos θ − 2 r sin θ + 1) r dr d θ
0 0
w2πw1
= (−2 r 2 cos θ − 2 r 2 sin θ + r ) dr d θ
0 0
w2πµ 3 3
¯
2 ¯ r =1
¶
= − 23r cos θ − 23r sin θ + r2 ¯ dθ
r =0
0
w2π ¡ ¢
= − 23 cos θ − 23 sin θ + 12 d θ
0
¯2π
= − 23 sin θ + 23 cos θ + 12 θ ¯ = π .
¯
0

The boundary curve C is the unit circle x2 + y2 = 1 laying in the plane z = 1 (see Figure
198 CHAPTER 5. LINE AND SURFACE INTEGRALS

4.5.5), which can be parametrized as x = cos t, y = sin t, z = 1 for 0 ≤ t ≤ 2π. So

z w2π
f · dr = ((1)(− sin t) + (cos t)(cos t) + (sin t)(0)) dt
C 0
w2πµ 1 + cos 2 t
¶ µ
1 + cos 2 t
¶
= − sin t + dt here we used cos2 t =
2 2
0
t sin 2 t ¯¯2π
= cos t + + ¯ = π.
2 4 0
So we see that z x
f · dr = (curl f ) · n d σ,
C Σ
as predicted by Stokes’ Theorem.
The line integral in the preceding example was far simpler to calculate than the surface
integral, but this will not always be the case.
2 y2
Example 5.15. LetuΣ be the elliptic paraboloid z = x4 + 9 for z ≤ 1, and let C be its boundary
curve. Calculate C f · d r for f( x, y, z) = (9 xz + 2 y)i + (2 x + y2 )j + (−2 y2 + 2 z)k, where C is
traversed counterclockwise.
Solution: The surface is similar to the one in Example 5.14, except now the boundary curve C
2 y2
is the ellipse x4 + 9 = 1 laying in the plane z = 1. In this case, using Stokes’ Theorem is easier
than computing the line integral directly. As in Example 5.14, at each point ( x, y, z( x, y)) on
2 y2
the surface z = z( x, y) = x4 + 9 the vector
− ∂∂xz i − ∂∂ zy j + k 2y
− 2x i − 9 j + k
n = r = q ,
¡ ∂ z ¢2 ³ ∂ z ´2 x2 4 y2
1 + ∂x + ∂ y 1 + 4 + 9

is a positive unit normal vector to Σ. And calculating the curl of f gives

curl f = (−4 y − 0)i + (9 x − 0)j + (2 − 2)k = −4 y i + 9 x j + 0 k ,
so
2y
(−4 y)(− 2x ) + (9 x)(− 9 ) + (0)(1) 2x y − 2x y + 0
(curl f) · n = q = q = 0,
2
x 2 4y x2 4 y2
1+ 4 + 9 1+ 4 + 9
and so by Stokes’ Theorem
z x x
f · dr = (curl f) · n d σ = 0 dσ = 0 .
C Σ Σ
u
In physical applications, for a simple closed curve C the line integral C f· d r is often called
the circulation of f around C . For example, if E represents the electrostatic field u due to a
point charge, then it turns out7 that curl E = 0, which means that the circulation C E · d r = 0
7 See Ch. 2 in R EITZ, M ILFORD and C HRISTY .
 5.5 Stokes’ Theorem 199

by Stokes’ Theorem. Vector fields which have zero curl are often called irrotational fields.
In fact, the term curl was created by the 19th century Scottish physicist James Clerk
Maxwell in his study of electromagnetism, where it is used extensively. In physics, the
curl is interpreted as a measure of circulation density. This is best seen by using another
definition of curl f which is equivalent8 to the definition given by formula (5.46). Namely, the
value of n · (curl f) at a point ( x, y, z), is

1z
lim f · d r, (5.50)
S →0 S
C

where S is the surface area of a surface Σ containing the point ( x, y, z) and with a simple
closed boundary curve C and positive unit normal vector n at ( x, y, z). In the limit, think of
the curve C shrinking to the point ( x, y, z), which causes Σ, the surface it bounds, to have
smaller and smaller surface area. That ratio of circulation to surface area in the limit is
what makes the curl a rough measure of circulation density (that is, circulation per unit
area).

An idea of how the curl of a vector field is y

related to rotation is shown in Figure 4.5.6.
Suppose we have a vector field f( x, y, z) which
is always parallel to the x y-plane at each f
point ( x, y, z) and that the vectors grow larger
the further the point ( x, y, z) is from the y-
axis. For example, f( x, y, z) = (1 + x2 ) j. Think
of the vector field as representing the flow
of water, and imagine dropping two wheels
with paddles into that water flow, as in Fig-
ure 4.5.6. Since the flow is stronger (that x
is, the magnitude of f is larger) as you move 0
away from the y-axis, then such a wheel would
rotate counterclockwise if it were dropped to Figure 5.5.6 Curl and rotation
the right of the y-axis, and it would rotate
clockwise if it were dropped to the left of the y-axis. In both cases the curl would be nonzero
(curl f( x, y, z) = 2 x k in our example) and would obey the right-hand rule; that is, curl f( x, y, z)
points in the direction of your thumb as you cup your right hand in the direction of the rota-
tion of the wheel. So the curl points outward (in the positive z-direction) if x > 0 and points
inward (in the negative z-direction) if x < 0. Notice that if all the vectors had the same di-
rection and the same magnitude, then the wheels would not rotate and hence there would
be no curl (which is why such fields are called irrotational, meaning no rotation).
Finally, by Stokes’ Theorem, we know that if C is a simple closed curve in some solid region

8 See S CHEY, p. 78–81, for the derivation.

200 CHAPTER 5. LINE AND SURFACE INTEGRALS

S in R3 and if f( x, y, z) is a smooth vector field such that curl f = 0 in S , then

z x x x
f · dr = (curl f) · n d σ = 0 · n dσ = 0 dσ = 0 ,
C Σ Σ Σ

where Σ is any orientable surface inside S whose boundary is C (such a surface is some-
times called a capping surface for C ). So similar to the two-variable case, we have a three-
dimensional version of a result from Section 4.3, for solid regions in R3 which are simply
connected (that is, regions having no holes):

The following statements are equivalent for a simply connected solid region S in R3 :

(a) f( x, y, z) = P ( x, y, z) i + Q ( x, y, z) j + R ( x, y, z) k has a smooth potential F ( x, y, z) in S ;

w
(b) f · d r is independent of the path for any curve C in S ;
C
z
(c) f · d r = 0 for every simple closed curve C in S ;
C

∂R ∂Q ∂P ∂R ∂Q ∂P
(d) = , = , and = in S (that is, curl f = 0 in S ).
∂y ∂z ∂z ∂x ∂x ∂y

Part (d) is also a way of saying that the differential form P dx + Q d y + R dz is exact.

Example 5.16. Determine if the vector field f( x, y, z) = x yz i + xz j + x y k has a potential in R3 .

Solution: Since R3 is simply connected, we just need to check whether curl f = 0 throughout
R3 , that is,
∂R ∂Q ∂P ∂R ∂Q ∂P
= , = , and =
∂y ∂z ∂z ∂x ∂x ∂y
throughout R3 , where P ( x, y, z) = x yz, Q ( x, y, z) = xz, and R ( x, y, z) = x y. But we see that
∂P ∂R ∂P ∂R
= xy , =y ⇒ 6= for some ( x, y, z) in R3 .
∂z ∂x ∂z ∂x

Thus, f( x, y, z) does not have a potential in R3 .

Exercises
A r
For Exercises 1–3, calculate C f ( x, y, z) ds for the given function f ( x, y, z) and curve C .

1. f ( x, y, z) = z; C : x = cos t, y = sin t, z = t, 0 ≤ t ≤ 2π
x
2. f ( x, y, z) = + y + 2 yz; C : x = t2 , y = t, z = 1, 1 ≤ t ≤ 2
y
 5.5 Stokes’ Theorem 201

p
2 2 3/2
3. f ( x, y, z) = z2 ; C : x = t sin t, y = t cos t, z = 3 t , 0≤t≤1
r
For Exercises 4–9, calculate ·
C f · dr for the given vector field f( x, y, z) and curve C .

4. f( x, y, z) = i − j + k; C : x = 3 t, y = 2 t, z = t, 0 ≤ t ≤ 1

5. f( x, y, z) = y i − x j + z k; C : x = cos t, y = sin t, z = t, 0 ≤ t ≤ 2π

6. f( x, y, z) = x i + y j + z k; C : x = cos t, y = sin t, z = 2, 0 ≤ t ≤ 2π

7. f( x, y, z) = ( y − 2 z) i + x y j + (2 xz + y) k; C : x = t, y = 2 t, z = t2 − 1, 0 ≤ t ≤ 1

8. f( x, y, z) = yz i + xz j + x y k; C : the polygonal path from (0, 0, 0) to (1, 0, 0) to (1, 2, 0)

9. f( x, y, z) = x y i + ( z − x) j + 2 yz k; C : the polygonal path from (0, 0, 0) to (1, 0, 0) to (1, 2, 0)

to (1, 2, −2)

For Exercises 10–13, state whether or not the vector field f( x, y, z) has a potential in R3 (you
do not need to find the potential itself).

10. f( x, y, z) = y i − x j + z k 11. f( x, y, z) = a i + b j + c k (a, b, c constant)

12. f( x, y, z) = ( x + y) i + x j + z2 k 13. f( x, y, z) = x y i − ( x − yz2 ) j + y2 z k

B
For Exercises 14–15, verify Stokes’ Theorem for the given vector field f( x, y, z) and surface Σ.

14. f( x, y, z) = 2 y i − x j + z k; Σ : x2 + y2 + z2 = 1, z ≥ 0

15. f( x, y, z) = x y i + xz j + yz k; Σ : z = x 2 + y2 , z ≤ 1

16. Construct a Möbius strip from a piece of paper, then draw a line down its center (like
the dotted line in Figure 5.5.3(b)). Cut the Möbius strip along that center line completely
around the strip. How many surfaces does this result in? How would you describe them?
Are they orientable?

C
17.sLet Σ be a closed surface and f( x, y, z) a smooth vector field. Show that
(curl f) · n d σ = 0. (Hint: Split Σ in half.)
Σ

18. Show that Green’s Theorem is a special case of Stokes’ Theorem.

202 CHAPTER 5. LINE AND SURFACE INTEGRALS

5.6 Gradient, Divergence, Curl and Laplacian

In this final section we will establish some relationships between the gradient, divergence
and curl, and we will also introduce a new quantity called the Laplacian. We will then show
how to write these quantities in cylindrical and spherical coordinates.
For a real-valued function f ( x, y, z) on R3 , the gradient ∇ f ( x, y, z) is a vector-valued func-
tion on R3 , that is, its value at a point ( x, y, z) is the vector
µ ¶
∂f ∂f ∂f ∂f ∂f ∂f
∇ f ( x, y, z) = , , = i + j + k
∂x ∂ y ∂z ∂x ∂y ∂z

in R3 , where each of the partial derivatives is evaluated at the point ( x, y, z). So in this way,
you can think of the symbol ∇ as being “applied” to a real-valued function f to produce a
vector ∇ f .
It turns out that the divergence and curl can also be expressed in terms of the symbol ∇.
This is done by thinking of ∇ as a vector in R3 , namely

∂ ∂ ∂
∇ = i + j + k. (5.51)
∂x ∂y ∂z

Here, the symbols ∂∂x , ∂∂y and ∂∂z are to be thought of as “partial derivative operators” that
will get “applied” to a real-valued function, say f ( x, y, z), to produce the partial derivatives
∂f ∂f ∂f ∂f
,
∂x ∂ y
and ∂ z . For instance, ∂∂x “applied” to f ( x, y, z) produces ∂ x .
Is ∇ really a vector? Strictly speaking, no, since ∂∂x , ∂∂y and ∂∂z are not actual numbers. But
it helps to think of ∇ as a vector, especially with the divergence and curl, as we will soon see.
The process of “applying” ∂∂x , ∂∂y , ∂∂z to a real-valued function f ( x, y, z) is normally thought of
as multiplying the quantities:
µ ¶ µ ¶ µ ¶
∂ ∂f ∂ ∂f ∂ ∂f
(f ) = , (f ) = , (f ) =
∂x ∂x ∂y ∂y ∂z ∂z

For this reason, ∇ is often referred to as the “del operator”, since it “operates” on functions.
For example, it is often convenient to write the divergence div f as ∇ · f, since for a vector
field f( x, y, z) = f 1 ( x, y, z)i + f 2 ( x, y, z)j + f 3 ( x, y, z)k, the dot product of f with ∇ (thought of as a
vector) makes sense:
µ ¶
∂ ∂ ∂
∇· f = i + j + k · ( f 1 ( x, y, z)i + f 2 ( x, y, z)j + f 3 ( x, y, z)k)
∂x ∂y ∂z
µ ¶ µ ¶ µ ¶
∂ ∂ ∂
= ( f1) + ( f2) + ( f3)
∂x ∂y ∂z
∂ f1 ∂ f2 ∂ f3
= + +
∂x ∂y ∂z
= div f
 5.6 Gradient, Divergence, Curl and Laplacian 203

We can also write curl f in terms of ∇, namely as ∇ × f, since for a vector field f( x, y, z) =
P ( x, y, z)i + Q ( x, y, z)j + R ( x, y, z)k, we have:
¯ ¯
¯
¯ i j k ¯
¯
¯ ∂ ∂ ∂ ¯
¯ ¯
∇×f = ¯ ¯
¯ ∂x ∂y ∂z ¯
¯ ¯
¯P ( x, y, z) Q ( x, y, z) R ( x, y, z)¯
µ ¶ µ ¶ µ ¶
∂R ∂Q ∂R ∂P ∂Q ∂P
= − i − − j + − k
∂y ∂z ∂x ∂z ∂x ∂ y
µ ¶ µ ¶ µ ¶
∂R ∂Q ∂P ∂R ∂Q ∂P
= − i + − j + − k
∂y ∂z ∂z ∂x ∂x ∂ y
= curl f

∂f ∂f ∂f
For a real-valued function f ( x, y, z), the gradient ∇ f ( x, y, z) = ∂x
i + ∂y j + ∂z
k is a vector
field, so we can take its divergence:

div ∇ f = ∇ · ∇ f
µ ¶ µ ¶
∂ ∂ ∂ ∂f ∂f ∂f
= i + j + k · i + j + k
∂x ∂y ∂z ∂x ∂y ∂z
µ ¶ µ ¶ µ ¶
∂ ∂f ∂ ∂f ∂ ∂f
= + +
∂x ∂x ∂y ∂y ∂z ∂z
∂2 f ∂2 f ∂2 f
= + +
∂ x2 ∂ y2 ∂ z2

Note that this is a real-valued function, to which we will give a special name:

Definition 5.7. For a real-valued function f ( x, y, z), the Laplacian of f , denoted by ∆ f , is

given by
∂2 f ∂2 f ∂2 f
∆ f ( x, y, z) = ∇ · ∇ f = + + . (5.52)
∂ x2 ∂ y2 ∂ z2

Example 5.17. Let r( x, y, z) = x i+ y j+ z k be the position vector field on R3 . Then k r( x, y, z) k2 =

r · r = x2 + y2 + z2 is a real-valued function. Find

(a) the gradient of k r k2

(b) the divergence of r

(c) the curl of r

(d) the Laplacian of k r k2

204 CHAPTER 5. LINE AND SURFACE INTEGRALS

Solution: (a) ∇k r k2 = 2 x i + 2 y j + 2 z k = 2 r
∂
(b) ∇ · r = ∂x
( x) + ∂∂y ( y) + ∂∂z ( z) = 1 + 1 + 1 = 3
(c) ¯ ¯
¯ i j k ¯¯
¯
¯ ¯
¯∂ ∂ ∂¯
∇× r = ¯ ¯ = (0 − 0) i − (0 − 0) j + (0 − 0) k = 0
¯ ∂x ∂y ∂ z ¯¯
¯
¯ x y z ¯
∂2 2 2
(d) ∆k r k2 = ∂ x2
( x2 + y2 + z2 ) + ∂∂y2 ( x2 + y2 + z2 ) + ∂∂z2 ( x2 + y2 + z2 ) = 2 + 2 + 2 = 6
2
Note that we could have calculated ∆k r k another way, using the ∇ notation along with parts
(a) and (b):
∆k r k2 = ∇ · ∇k r k2 = ∇ · 2 r = 2 ∇ · r = 2(3) = 6

Notice that in Example 5.17 if we take the curl of the gradient of k r k2 we get

∇ × (∇k r k2 ) = ∇ × 2 r = 2 ∇ × r = 2 0 = 0 .

The following theorem shows that this will be the case in general:

Theorem 5.14. For any smooth real-valued function f ( x, y, z), ∇ × (∇ f ) = 0.

Proof: We see by the smoothness of f that

¯ ¯
¯ i j k ¯¯
¯
¯ ∂ ∂ ∂ ¯¯
¯
¯ ¯
∇ × (∇ f ) = ¯ ∂ x ∂ y ∂ z ¯
¯ ¯
¯∂f ∂f ∂f ¯
¯ ¯
¯ ∂x ∂ y ∂z ¯
µ 2 ¶ µ 2 ¶ µ 2 ¶
∂ f ∂2 f ∂ f ∂2 f ∂ f ∂2 f
= − i − − j + − k = 0,
∂ y ∂z ∂z ∂ y ∂x ∂z ∂z ∂x ∂x ∂ y ∂ y ∂x

since the mixed partial derivatives in each component are equal. QED

Corollary 5.15. If a vector field f( x, y, z) has a potential, then curl f = 0.

Another way of stating Theorem 5.14 is that gradients are irrotational. Also, notice that
in Example 5.17 if we take the divergence of the curl of r we trivially get

∇ · (∇ × r) = ∇ · 0 = 0 .

The following theorem shows that this will be the case in general:
 5.6 Gradient, Divergence, Curl and Laplacian 205

Theorem 5.16. For any smooth vector field f( x, y, z), ∇ · (∇ × f) = 0.

The proof is straightforward and left as an exercise for the reader.

Corollary 5.17. The flux of the curl of a smooth vector field f( x, y, z) through any closed
surface is zero.

Proof: Let Σ be a closed surface which bounds a solid S . The flux of ∇ × f through Σ is
x y
(∇ × f ) · d σ = ∇ · (∇ × f ) dV (by the Divergence Theorem)
Σ S
y
= 0 dV (by Theorem 5.16)
S
= 0. QED

There is another method for proving Theorem

s 5.14 which can be useful, and is often used
in physics. Namely, if the surface integral f ( x, y, z) d σ = 0 for all surfaces Σ in some solid
Σ
region (usually all of R3 ), then we must have f ( x, y, z) = 0 throughout that region. The proof
is not trivial, and physicists do not usually bother to prove it. But the result is true, and can
also be applied to double and triple integrals.
For instance, to prove Theorem 5.14, assume that f ( x, y, z) is a smooth real-valued function
on R3 . Let C be a simple closed curve in R3 and let Σ be any capping surface for C (that is, Σ
is orientable and its boundary is C ). Since ∇ f is a vector field, then
x z
(∇ × (∇ f )) · n d σ = ∇ f · dr by Stokes’ Theorem, so
Σ C
= 0 by Corollary 5.12.

Since the choice of Σ was arbitrary, then we must have (∇× × (∇ f )) · n = 0 throughout R3 , where
n is any unit vector. Using i, j and k in place of n, we see that we must have ∇ × (∇ f ) = 0 in
R3 , which completes the proof.

Example 5.18. A system of electric charges has a charge density ρ ( x, y, z) and produces an
electrostatic field E( x, y, z) at points ( x, y, z) in space. Gauss’ Law states that
x y
E · d σ = 4π ρ dV
Σ S

for any closed surface Σ which encloses the charges, with S being the solid region enclosed
by Σ. Show that ∇ · E = 4πρ . This is one of Maxwell’s Equations.9
9 In Gaussian (or CGS) units.
206 CHAPTER 5. LINE AND SURFACE INTEGRALS

Solution: By the Divergence Theorem and Gauss’ Law, we have

y x
∇ · E dV = E · dσ
S Σ
y
= 4π ρ dV .
S

Combining the integrals gives

y
(∇ · E − 4πρ ) dV = 0 .
S

Since Σ and hence S was arbitrary, we get ∇ · E = 4πρ .

Exercises
A
For Exercises 1–6, find the Laplacian of the function f ( x, y, z).

1. f ( x, y, z) = x + y + z 2. f ( x, y, z) = x5 3. f ( x, y, z) = ( x2 + y2 + z2 )3/2
2
− y2 − z2
4. f ( x, y, z) = e x+ y+ z 5. f ( x, y, z) = x3 + y3 + z3 6. f ( x, y, z) = e− x

B
For Exercises 7–18, prove the given formula ( r = k r k is the length of the position vector field
r( x, y, z) = x i + y j + z k).

7. ∇ (1/ r ) = −r/ r 3 8. ∆ (1/ r ) = 0 9. ∇ · (r/ r 3 ) = 0 10. ∇ (ln r ) = r/ r 2

11. div(f + g) = div f + div g 12. curl(f + g) = curl f + curl g

13. div( f g) = f div g + g · ∇ f 14. div(f × g) = g · curl f − f · curl g

15. div(∇ f × ∇ g) = 0 16. curl( f g) = f curl g + (∇ f ) × g

17. curl(curl f) = ∇(div f) − ∆ f 18. ∆ ( f g) = f ∆ g + g ∆ f + 2(∇ f · ∇ g)

C
19. Prove Theorem 5.16.

20. Use f = u ∇v in the Divergence Theorem to prove:

(a) Green’s first identity:
y x
( u ∆ v + (∇ u) · (∇v)) dV = ( u ∇ v) · d σ
S Σ

(b) Green’s second identity:

y x
( u ∆ v − v ∆ u) dV = ( u ∇v − v ∇ u) · d σ
S Σ
 5.7 Other coordinate systems 207

21. Suppose that ∆ u = 0 (that is, u is harmonic) over R3 . Show that

x
∇u · d σ = 0
Σ

for any closed surface Σ.

5.7 Other coordinate systems

Often (especially in physics) it is convenient to use other coordinate systems when dealing
with quantities such as the gradient, divergence, curl and Laplacian. We will present the
formulas for these in cylindrical and spherical coordinates.
Recall from Section 1.7 that a point ( x, y, z) can be represented in cylindrical coordinates
( r, θ , z), where x = r cos θ , y = r sin θ , z = z. At each point ( r, θ , z), let er , eθ , e z be unit vectors
in the direction of increasing r , θ , z, respectively (see Figure 5.7.1). Then er , eθ , e z form an
orthonormal set of vectors. Note, by the right-hand rule, that e z × er = eθ .

ez
eρ
z eθ z eθ
(x, y, z) (x, y, z)
er
eφ
z φ ρ
y z y
0 0
x θ r x θ

y (x, y, 0) y (x, y, 0)
x x

Figure 5.7.1 Figure 5.7.2

Orthonormal vectors er , eθ , e z Orthonormal vectors eρ , eθ , eφ
in cylindrical coordinates in spherical coordinates

Similarly, a point ( x, y, z) can be represented in spherical coordinates (ρ , θ , φ), where x =

ρ sin φ cos θ , y = ρ sin φ sin θ , z = ρ cos φ. At each point (ρ , θ , φ), let eρ , eθ , eφ be unit vectors
in the direction of increasing ρ , θ , φ, respectively (see Figure 5.7.2). Then the vectors eρ , eθ ,
eφ are orthonormal. By the right-hand rule, we see that eθ × eρ = eφ .
We can now summarize the expressions for the gradient, divergence, curl and Laplacian
in Cartesian, cylindrical and spherical coordinates in the following tables:
208 CHAPTER 5. LINE AND SURFACE INTEGRALS

Cartesian ( x, y, z): Scalar function F ; Vector field f = f 1 i + f 2 j + f 3 k

∂F ∂F ∂F
gradient : ∇F = i + j + k
∂x ∂y ∂z
∂ f1 ∂ f2 ∂ f3
divergence : ∇ · f = + +
∂x ∂y
∂z
µ µ ¶ ¶ µ ¶
∂ f3 ∂ f1 ∂ f3
∂ f2 ∂ f2 ∂ f1
curl : ∇ × f = − i + − j + − k
∂y ∂z ∂z ∂x ∂x ∂y
∂2 F ∂2 F ∂2 F
Laplacian : ∆ F = + +
∂ x2 ∂ y2 ∂ z2

Cylindrical ( r, θ , z): Scalar function F ; Vector field f = f r er + f θ eθ + f z e z

∂F 1 ∂F ∂F
gradient : ∇F = er + eθ + ez
∂r r ∂θ ∂z
1 ∂ 1 ∂ fθ ∂fz
divergence : ∇ · f = (r f r ) + +
µr ∂ r ¶r ∂θ µ ∂ z ¶ µ ¶
1 ∂ f z ∂ fθ ∂ fr ∂ fz 1 ∂ ∂ fr
curl : ∇ × f = − er + − eθ + (r f θ ) − ez
r ∂θ ∂z ∂z ∂r r ∂r ∂θ
µ ¶
1 ∂ ∂F 1 ∂2 F ∂2 F
Laplacian : ∆ F = r + 2 2 +
r ∂r ∂r r ∂θ ∂ z2

Spherical (ρ , θ , φ): Scalar function F ; Vector field f = f ρ eρ + f θ eθ + f φ eφ

∂F 1 ∂F 1 ∂F
gradient : ∇F = eρ + eθ + eφ
∂ρ ρ sin φ ∂θ ρ ∂φ
1 ∂ 2 1 ∂ fθ 1 ∂
divergence : ∇ · f = 2 (ρ f ρ ) + + (sin φ f θ )
ρ ∂ρ ρ sin φ ∂θ ρ sin φ ∂φ
µ ¶ µ ¶
1 ∂ ∂ fφ 1 ∂ ∂ fρ
curl : ∇ × f = (sin φ f θ ) − eρ + (ρ f φ ) − eθ
ρ sin φ ∂φ ∂θ ρ ∂ρ ∂φ
µ ¶
1 ∂ fρ 1 ∂
+ − (ρ f θ ) eφ
ρ sin φ ∂θ ρ ∂ρ
µ ¶ µ ¶
1 ∂ ∂F 1 ∂2 F 1 ∂ ∂F
Laplacian : ∆ F = 2 ρ2 + + sin φ
ρ ∂ρ ∂ρ ρ 2 sin2 φ ∂θ 2 ρ 2 sin φ ∂φ ∂φ

The derivation of the above formulas for cylindrical and spherical coordinates is straight-
forward but extremely tedious. The basic idea is to take the Cartesian equivalent of the
quantity in question and to substitute into that formula using the appropriate coordinate
transformation. As an example, we will derive the formula for the gradient in spherical
coordinates.
 5.7 Other coordinate systems 209

Goal: Show that the gradient of a real-valued function F (ρ , θ , φ) in spherical coordinates is:

∂F 1 ∂F 1 ∂F
∇F = eρ + eθ + eφ
∂ρ ρ sin φ ∂θ ρ ∂φ

Idea: In the Cartesian gradient formula ∇F ( x, y, z) = ∂∂Fx i + ∂∂Fy j + ∂∂Fz k, put the Cartesian ba-
sis vectors i, j, k in terms of the spherical coordinate basis vectors eρ , eθ , eφ and functions of
ρ , θ and φ. Then put the partial derivatives ∂∂Fx , ∂∂Fy , ∂∂Fz in terms of ∂∂ρF , ∂∂θF , ∂∂φ
F
and functions
of ρ , θ and φ.

Step 1: Get formulas for eρ , eθ , eφ in terms of i, j, k.

We can see from Figure 5.7.2 that the unit vector eρ in the ρ direction at a general point
(ρ , θ , φ) is eρ = k rr k , where r = x i + y j + z k is the position vector of the point in Cartesian
coordinates. Thus,
r xi+ yj+ zk
eρ = = p ,
krk x 2 + y2 + z 2
p
so using x = ρ sin φ cos θ , y = ρ sin φ sin θ , z = ρ cos φ, and ρ = x2 + y2 + z2 , we get:

eρ = sin φ cos θ i + sin φ sin θ j + cos φ k

Now, since the angle θ is measured in the x y-plane, then the unit vector eθ in the θ
direction must be parallel to the x y-plane. That is, eθ is of the form a i + b j + 0 k. To figure
out what a and b are, note that since eθ ⊥ eρ , then in particular eθ ⊥ eρ when eρ is in the
x y-plane. That occurs when the angle φ is π/2. Putting φ = π/2 into the formula for eρ gives
eρ = cos θ i + sin θ j + 0 k, and we see that a vector perpendicular to that is − sin θ i + cos θ j + 0 k.
Since this vector is also a unit vector and points in the (positive) θ direction, it must be eθ :

eθ = − sin θ i + cos θ j + 0 k

Lastly, since eφ = eθ × eρ , we get:

eφ = cos φ cos θ i + cos φ sin θ j − sin φ k

Step 2: Use the three formulas from Step 1 to solve for i, j, k in terms of eρ , eθ , eφ .
This comes down to solving a system of three equations in three unknowns. There are
many ways of doing this, but we will do it by combining the formulas for eρ and eφ to
eliminate k, which will give us an equation involving just i and j. This, with the formula for
eθ , will then leave us with a system of two equations in two unknowns (i and j), which we
will use to solve first for j then for i. Lastly, we will solve for k.
First, note that
sin φ eρ + cos φ eφ = cos θ i + sin θ j
so that
sin θ (sin φ eρ + cos φ eφ ) + cos θ eθ = (sin2 θ + cos2 θ )j = j ,
210 CHAPTER 5. LINE AND SURFACE INTEGRALS

and so:
j = sin φ sin θ eρ + cos θ eθ + cos φ sin θ eφ

Likewise, we see that

cos θ (sin φ eρ + cos φ eφ ) − sin θ eθ = (cos2 θ + sin2 θ )i = i ,

and so:
i = sin φ cos θ eρ − sin θ eθ + cos φ cos θ eφ

Lastly, we see that:

k = cos φ eρ − sin φ eφ

Step 3: Get formulas for ∂∂ρF , ∂∂θF , ∂F

∂φ
in terms of ∂F ∂F ∂F
, , .
∂x ∂ y ∂z
By the Chain Rule, we have

∂F ∂F ∂ x ∂F ∂ y ∂F ∂ z
= + + ,
∂ρ ∂ x ∂ρ ∂ y ∂ρ ∂ z ∂ρ
∂F ∂F ∂ x ∂F ∂ y ∂F ∂ z
= + + ,
∂θ ∂ x ∂θ ∂ y ∂θ ∂ z ∂θ
∂F ∂F ∂ x ∂F ∂ y ∂F ∂ z
= + + ,
∂φ ∂ x ∂φ ∂ y ∂φ ∂ z ∂φ

which yields:
∂F ∂F ∂F ∂F
= sin φ cos θ + sin φ sin θ + cos φ
∂ρ ∂x ∂y ∂z
∂F ∂F ∂F
= −ρ sin φ sin θ + ρ sin φ cos θ
∂θ ∂x ∂y
∂F ∂F ∂F ∂F
= ρ cos φ cos θ + ρ cos φ sin θ − ρ sin φ
∂φ ∂x ∂y ∂z

Step 4: Use the three formulas from Step 3 to solve for ∂∂Fx , ∂∂Fy , ∂∂Fz in terms of ∂∂ρF , ∂∂θF , ∂∂φ
F
.
Again, this involves solving a system of three equations in three unknowns. Using a
similar process of elimination as in Step 2, we get:
µ ¶
∂F 1 ∂F ∂F ∂F
= ρ sin2 φ cos θ − sin θ + sin φ cos φ cos θ
∂x ρ sin φ ∂ρ ∂θ ∂φ
µ ¶
∂F 1 2 ∂F ∂F ∂F
= ρ sin φ sin θ + cos θ + sin φ cos φ sin θ
∂y ρ sin φ ∂ρ ∂θ ∂φ
µ ¶
∂F 1 ∂F ∂F
= ρ cos φ − sin φ
∂z ρ ∂ρ ∂φ

∂F ∂F ∂F
Step 5: Substitute the formulas for i, j, k from Step 2 and the formulas for , ,
∂x ∂ y ∂z
from
∂F ∂F ∂F
Step 4 into the Cartesian gradient formula ∇F ( x, y, z) = ∂x
i+ ∂y
j+ ∂z
k.
 5.7 Other coordinate systems 211

Doing this last step is perhaps the most tedious, since it involves simplifying 3 × 3 + 3 × 3 +
2 × 2 = 22 terms! Namely,
µ ¶
1 ∂F ∂F ∂F
∇F = ρ sin2 φ cos θ − sin θ + sin φ cos φ cos θ (sin φ cos θ eρ − sin θ eθ
ρ sin φ ∂ρ ∂θ ∂φ
+ cos φ cos θ eφ )
µ ¶
1 2 ∂F ∂F ∂F
+ ρ sin φ sin θ + cos θ + sin φ cos φ sin θ (sin φ sin θ eρ + cos θ eθ
ρ sin φ ∂ρ ∂θ ∂φ
+ cos φ sin θ eφ )
µ ¶
1 ∂F ∂F
+ ρ cos φ − sin φ (cos φ eρ − sin φ eφ ) ,
ρ ∂ρ ∂φ
which we see has 8 terms involving eρ , 6 terms involving eθ , and 8 terms involving eφ . But
the algebra is straightforward and yields the desired result:
∂F 1 ∂F 1 ∂F
∇F = eρ + eθ + eφ X
∂ρ ρ sin φ ∂θ ρ ∂φ

Example 5.19. In Example 5.17 we showed that ∇k r k2 = 2 r and ∆k r k2 = 6, where r( x, y, z) =

x i + y j + z k in Cartesian coordinates. Verify that we get the same answers if we switch to
spherical coordinates.
Solution: Since k r k2 = x2 + y2 + z2 = ρ 2 in spherical coordinates, let F (ρ , θ , φ) = ρ 2 (so that
F (ρ , θ , φ) = k r k2 ). The gradient of F in spherical coordinates is
∂F 1 ∂F 1 ∂F
∇F = eρ + eθ + eφ
∂ρ ρ sin φ ∂θ ρ ∂φ
1 1
= 2ρ eρ + (0) eθ + (0) eφ
ρ sin φ ρ
r
= 2ρ eρ = 2ρ , as we showed earlier, so
krk
r
= 2ρ = 2 r , as expected. And the Laplacian is
ρ
µ ¶ µ ¶
1 ∂ 2 ∂F 1 ∂2 F 1 ∂ ∂F
∆F = ρ + + 2 sin φ
ρ 2 ∂ρ ∂ρ ρ 2 sin2 φ ∂θ 2 ρ sin φ ∂φ ∂φ
1 ∂ 2 1 1 ∂ ¡ ¢
= (ρ 2ρ ) + 2 (0) + 2 sin φ (0)
ρ 2 ∂ρ ρ sin φ ρ sin φ ∂φ
1 ∂
= (2ρ 3 ) + 0 + 0
ρ 2 ∂ρ
1
= (6ρ 2 ) = 6 , as expected.
ρ2

Exercises
A
212 CHAPTER 5. LINE AND SURFACE INTEGRALS

1. Let f ( x, y, z) = ( x2 + y2 + z2 )3/2 in Cartesian coordinates. Find the Laplacian of f in spher-

ical coordinates.
2 2 2
2. Let f ( x, y, z) = e− x − y − z in Cartesian coordinates. Find the Laplacian of the function in
spherical coordinates.
z
3. Let f ( x, y, z) = in Cartesian coordinates. Find ∇ f in cylindrical coordinates.
x 2 + y2
4. For f( r, θ , z) = r er + z sin θ eθ + rz e z in cylindrical coordinates, find div f and curl f.

5. For f(ρ , θ , φ) = eρ + ρ cos θ eθ + ρ eφ in spherical coordinates, find div f and curl f.

C
6. Derive the gradient formula in cylindrical coordinates:

∂F 1 ∂F ∂F
∇F = er + eθ + ez .
∂r r ∂θ ∂z
Bibliography
Abbott, E.A., Flatland, 7th edition. New York: Dover Publications, Inc., 1952
Classic tale about a creature living in a 2-dimensional world who encounters a higher-
dimensional creature, with lots of humor thrown in.

Anton, H. and C. Rorres, Elementary Linear Algebra: Applications Version, 8th edition. New
York: John Wiley & Sons, 2000
Standard treatment of elementary linear algebra.

Bazaraa, M.S., H.D. Sherali and C.M. Shetty, Nonlinear Programming: Theory and Algo-
rithms, 2nd edition. New York: John Wiley & Sons, 1993
Thorough treatment of nonlinear optimization.

Farin, G., Curves and Surfaces for Computer Aided Geometric Design: A Practical Guide,
2nd edition. San Diego, CA: Academic Press, 1990
An intermediate-level book on curve and surface design.

Hecht, E., Optics, 2nd edition. Reading, MA: Addison-Wesley Publishing Co., 1987
An intermediate-level book on optics, covering a wide range of topics.

Hoel, P.G., S.C. Port and C.J. Stone, Introduction to Probability Theory, Boston, MA:
Houghton Mifflin Co., 1971
An excellent introduction to elementary, calculus-based probability theory. Lots of good exer-
cises.

Jackson, J.D., Classical Electrodynamics, 2nd edition. New York: John Wiley & Sons, 1975
An advanced book on electromagnetism, famous for being intimidating. Most of the mathemat-
ics will be understandable after reading the present book.

Marion, J.B., Classical Dynamics of Particles and Systems, 2nd edition. New York: Academic
Press, 1970
Standard intermediate-level treatment of classical mechanics. Very thorough.

O’Neill, B., Elementary Differential Geometry, New York: Academic Press, 1966
Intermediate-level book on differential geometry, with a modern approach based on differential
forms.

213
214 Bibliography

Pogorelov, A.V., Analytical Geometry, Moscow: Mir Publishers, 1980

An intermediate/advanced book on analytic geometry.

Press, W.H., S.A. Teukolsky, W.T. Vetterling and B.P. Flannery, Numerical Recipes in FOR-
TRAN: The Art of Scientific Computing, 2nd edition. Cambridge, UK: Cambridge Uni-
versity Press, 1992
An excellent source of information on numerical methods for solving a wide variety of problems.
Though all the examples are in the FORTRAN programming language, the code is clear enough
to implement in the language of your choice.

Protter, M.H. and C.B. Morrey, Analytic Geometry, 2nd edition. Reading, MA: Addison-
Wesley Publishing Co., 1975
Thorough treatment of elementary analytic geometry, with a rigor not found in most recent
books.

Ralston, A. and P. Rabinowitz, A First Course in Numerical Analysis, 2nd edition. New York:
McGraw-Hill, 1978
Standard treatment of elementary numerical analysis.

Reitz, J.R., F.J. Milford and R.W. Christy, Foundations of Electromagnetic Theory, 3rd edi-
tion. Reading, MA: Addison-Wesley Publishing Co., 1979
Intermediate text on electromagnetism.

Schey, H.M., Div, Grad, Curl, and All That: An Informal Text on Vector Calculus, New York:
W.W. Norton & Co., 1973
Very intuitive approach to the subject, from a physicist’s viewpoint. Highly recommended.

Taylor, A.E. and W.R. Mann, Advanced Calculus, 2nd edition. New York: John Wiley & Sons,
1972
Excellent treatment of n-dimensional calculus. A good book to study after the present book.
Many intriguing exercises.

Uspensky, J.V., Theory of Equations, New York: McGraw-Hill, 1948

A classic on the subject, discussing many interesting topics.

Weinberger, H.F., A First Course in Partial Differential Equations, New York: John Wiley &
Sons, 1965
A good introduction to the vast subject of partial differential equations.

Welchons, A.M. and W.R. Krickenberger, Solid Geometry, Boston, MA: Ginn & Co., 1936
A very thorough treatment of 3-dimensional geometry from an elementary perspective, in-
cludes many topics which (sadly) do not seem to be taught anymore.
Appendix A
Answers and Hints to Selected Exercises
Chapter 1 3.(a) (2, 1, 3) + t(1, 0, 1); (b) x = 2 + t, y = 1, z =
3 + t; (c) x − 2 = z − 3, y = 1; 5. x = 1 + 2 t, y =
Section 1.1 (p. 8) −2 + 7 t, z = −3 + 8 t; 7. 7.65; 9. (1, 2, 3);
p p p
1.(a)p 5; (b) 5; (c) 17; (d) 1; 11. 4 x − 4 y + 3 z − 10 = 0; 13. x − 2 y − z + p2 =
(e) 2 17; 2. Yes; 3. No. 0; 15. 11 x − 24 y + 21 z − 26 = 0; 17. 9/ 35;
19. x = 5 t, y = 2 + 3 t, z = −7 t; 21. (10, −2, 1).

Section 1.2 (p. 14)

Section 1.6 (p. 49)
1.(a) (³−4, 4, −3); (b) ´ (2, 6, −1);
p p
(c) −1 p5
, , −2
; (d) 41
; (e) 41 1. radius: 1, center: (2, 3, 5); 3. radius:
2 ;
p p
30 30 30 2
(f) (14, −6, 8); (g) (−7, 3, −4); (h) (−1, −6, 1); 5, center: (−1, −1, −1); 5. No intersection;
p
(i) (−2, −4, 2); (j) No. 3. No, k v k + k w k is 7.circle x2 + y2 = 4 in the planes z = ± 5;
y y
larger. 9. lines ax = b , z = 0 and ax = − b , z = 0;
³ ´
13. 22−ac , 22−bc , 0 .
Section 1.3 (p. 19)
1. 10; 3. 73.4◦ ; 5. 90◦ ; 7. 0◦ ;p 9. pYes,
Section 1.7 (p. 53)
p
since v · w = 0; 11. | v · w < 21p5 = 1.(a)
p| = 0 p (4, π3 , −1); (b) ( 17, π3 , 1.816); 3.
p p
k v k k w k; 13. k v + w k = 26 < 21 + 5 = (a) (2 7, 11π , 0); (b) (2 7, 11π , π ); 5.(a) r 2 +
6 6 2
k v k + k w k; 15. Hint: use Definition 1.6; z2 = 25; (b) ρ = 5; 7.(a) r 2 + 9 z2 = 36;
24. Hint: See Theorem 1.10(c). (b) ρ 2 (1 + 8 cos2 φ) = 36; 10. (a, θ , a cot φ);
12. Hint: Use the distance formula for Carte-
Section 1.4 (p. 31) sian coordinates.
1. (−5, −23, −24); p 3. (8, 4, −5); 5. 0;
7. 16.72; 9. 4 5; 11. 9; 13. 0 and Chapter 2
(8, −10, 2); 15. 14; 31. A circle of radius
1
k v k centered at the origin in the normal Section 2.1 (p. 63)
plane to v.
1. f ′ ( t) = (1, 2 t, 3 t2 ), x = 1 + t, y = z =
1; 3. f ′ ( t) = (−2 sin 2 t, 2 cos 2 t, 1); x = 1,
Section 1.5 (p. 41)
y = 2 t, z = t; 5. v( t) = (1, 1 − cos t, sin t),
1. (a) (2, 3, −2) + t(5, 4, −3); (b) x = 2 + 5 t, a( t) = (0, sin t, cos t); 9.(a) Line parallel to c;
y−3
y = 3 + 4 t, z = −2 − 3 t; (c) x−5 2 = 4 = z−+32 ; (b) Half-line parallel to c; (c) Hint: Think of

215
216 APPENDIX A: ANSWERS AND HINTS TO SELECTED EXERCISES

the functions as position vectors; 15. Hint: Section 3.3 (p. 86)
Theorem 1.16.
1. 2 x + 3 y − z − 3 = 0; 3. −2 x + y − z − p
2 = 0;
1 4 11
5. x + 2 y = z; 7. 2 ( x − 1) + 9 ( y − 2) + 12 ( z −
Section 2.3 (p. 72) p
2 11
p
3π 5 3 ) = 0;
9. 3 x + 4 y − 5 z = 0.
1. 2 ; 3. 2(53/2 − 8); 5. Replace t by
³³ ´2/3 ´.
27 s+16
2 − 4 9; 6. Hint: Use Theo- Section 3.4 (p. 91)
rem 2.1(e), Example 2.3, and Theorem 1.16; y
1. (2 x, 2 y); 3. ( p 2 x 2 , p 2 2 ); 5. ( 1x , 1y );
7. Hint: Use Exercise 6. 9. Hint: Use x + y +4 x + y +4
f ′ ( t) = k f( t) kT, differentiate that to get f ′′ ( t), 7. ( yz cos( x yz), xz cos( xpyz), x y cos( x yz));
′ ′′
put those expressions into f ( t) × f ( t), then 9. (2 x, 2 y, 2 z); 11. 2 2; 13. p1 ;
p 3
write T ′ ( t) in terms of N( t).; 11. T( t) = 15. 3 cos(1); 17. increase: (45, 20), de-
p1 (− sin t, cos t, 1), N( t) = (− cos t, − sin t, 0), crease: (−45, −20)
2
B( t) = p1 (sin t, − cos t, 1), κ( t) = 1/2
2
Section 3.5 (p. 98)
Chapter 3 1. local min. (1, 0); saddle pt. (−1, 0); 3. lo-
cal min. (1, 1); local max. (−1, −1); saddle pts.
Section 3.1 (p. 78)
(1, −1), (−1, 1); 5. local min. (1, −1), saddle
1. domain: R2 , range: [−1, ∞); 3. domain: pt. (0, 0); 7. local min. (0, 0); 9. local min.
{( x, y) : x2 + y2 ≥ 4}, range: [0, ∞); 5. domain: (−1, 1/2); 11. width = height = depth=10;
R3 , range: [−1, 1]; 7. 1; 9. does not exist; 13. x = y = 4, z = 2.
11. 2; 13. 2; 15. 0; 17. does not exist.

Section 3.6 (p. 106)

Section 3.2 (p. 83)
∂f ∂f ∂f 2. ( x0 , y0 ) = (0, 0) : → (0.2858, −0.3998);
1. ∂x
= 2 x, ∂y
= 2 y; 3. ∂x
= x( x2 + y + 4)−1/2 ,
∂f ∂f ∂f
( x0 , y0 ) = (1, 1) : → (1.03256, −1.94037)
∂y
= 21 ( x2 + y + 4)−1/2 ; 5. ∂x
= ye x y + y, ∂y
=
∂f ∂f ∂f
xe x y + x; 7. = 4 x3 , = 0; 9.= x( x + 2

∂ f
∂x ∂y
∂
∂x
f
Section 3.7 (p. 112)
y2 )−1/2 , ∂ y = y( x2 + y ) ; 11. ∂ x = 23x ( x2 +
2 −1/2
³ ´ ³ ´
−4 p
∂f
y + 4)−2/3 , ∂ y = 13 ( x2 + y + 4)−2/3 ; 13. ∂ x =
∂f 1. min. p , −2 ; max. p4 , p2 ; 3. min.
³ ´ 5 5 ³ 5 5
´
2 2 ∂f 2 2
−2 xe−( x + y ) , ∂ y = −2 ye−( x + y ) ; 15. ∂ x =
∂f p20 , p30 ; max. − 20
p ,− p 30
4. There is
13 13 13 13
∂f ∂2 f ∂2 f no global maximum, nor global minimum.
y cos( x y), ∂ y = x cos( x y); 17. ∂ x2 = 2, ∂ y2 = 2,
5. 8abc
p
∂2 f ∂2 f ∂2 f 3 3
∂x ∂ y
= 0; 19. ∂ x2 = ( y+4)( x2 + y+4)−3/2 , ∂ y2 =
∂2 f
− 14 ( x2 + y + 4)−3/2 , ∂ x ∂ y = − 12 x( x2 + y + 4)−3/2 ;
∂2 f ∂2 f ∂2 f
Chapter 4
21. ∂ x2 = y2 e x y , ∂ y2 = x2 e x y , ∂ x ∂ y = (1 +
∂2 f ∂2 f ∂2 f Section 4.1 (p. 117)
x y) e x y + 1; 23. ∂ x2 = 12 x2 , ∂ y2 = 0, ∂ x ∂ y = 0;
∂2 f ∂2 f ∂2 f
25. ∂ x2 = − x−2 , ∂ y2 = − y−2 , ∂ x ∂ y = 0 1. 1; 3. 7
12 ; 5. 76 ; 7. 5; 9. 12 ; 11. 15.
 217

Section 4.2 (p. 124) Section 5.2 (p. 170)

π
6. 14 ; 7. 2; 9. 61 ; 1. 0; 3. No; 4. Yes. F ( x, y) = x2 − y ; 5. No;
2
1. 1; 3. 8 ln 2 − 3; 5. 4; 2 2
10. 65 . 9. (b) No. Hint: Think of how F is defined;
10. Yes. F ( x, y) = ax y + bx + c y + d
Section 4.3 (p. 128)
1. 92 ; 3. (2 cos(π2 ) + π4 − 2)/4; 5. 1
6; 7. 6; Section 5.3 (p. 177)
10. 13
1. 16/15; 3. −5π; 5. Yes. F ( x, y) = x y2 + x3 ;
7. Yes. F ( x, y) = 4 x2 y + 2 y2 + 3 x
Section 4.4 (p. 133)
1. The values should converge to ≈ 1.318. Section 5.4 (p. 186)
(Hint: In Java the exponential function e x
can be obtained with Math.exp(x). Other 1. 216π; 2. 3; 3. 12π/5; 7. 15/4
languages have similar functions, otherwise
use e = 2.7182818284590455 in your pro- Section 5.5 (p. 200)
gram.) 2. ≈ 1.146; 3. ≈ 0.705; 4. ≈ 0.168. p p p
1. 2 2 π2 2. (17 17 − 5 5)/3 3. 2/5; 4. 2;
Section 4.5 (p. 141) 5. 2π(π − 1); 7. 67/15; 9. 6; 11. Yes;
13. No; 19. Hint: Think of how a vector
1. 8π; 3. 43π (8 − 33/2 ); 7. 1 − sin2 2 ; 9. 2πab field f( x, y) = P ( x, y) i + Q ( x, y) j in R2 can be
extended in a natural way to be a vector field
Section 4.6 (p. 145) in R3 .

1. (1, 8/3) 3. (0, 34πa ) 5. (0, 3π/16);

7. (0, 0, 5a/12); 9. (7/12, 7/12, 7/12) Section 5.6 (p. 206)
p
1. 0; 3. 12 x2 + y2 + z2 ; 5. 6( x + y + z);
Section 4.7 (p. 153)
p n 1
1. π; 2. 1; 6. Both are ( n+1)2 ( n+2)
; 7. n; Section 5.7 (p. 211)
2
1. 12ρ ; 2. (4ρ 2 − 6) e−ρ ; 3. − 2r3z er + r12 e z ;
Chapter 5 sin θ
5. div f = ρ2 − sin + cot φ, curl f = cot φ cos θ eρ +
φ
2eθ − 2 cos θ eφ 6. Hint: Start by showing that
Section 5.1 (p. 162)
er = cos θ i + sin θ j, eθ = − sin θ i + cos θ j, e z =
1. 1/2; 3. 23; 5. 24π; 7. −2π; 9. 2π 11. 0 k.
GNU Free Documentation License
Version 1.2, November 2002
Copyright ©2000,2001,2002 Free Software Foundation, Inc.

51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA

Everyone is permitted to copy and distribute verbatim copies of this license document, but
changing it is not allowed.

Preamble

The purpose of this License is to make a manual, textbook, or other functional and useful
document "free" in the sense of freedom: to assure everyone the effective freedom to copy
and redistribute it, with or without modifying it, either commercially or noncommercially.
Secondarily, this License preserves for the author and publisher a way to get credit for their
work, while not being considered responsible for modifications made by others.
This License is a kind of "copyleft", which means that derivative works of the document
must themselves be free in the same sense. It complements the GNU General Public License,
which is a copyleft license designed for free software.
We have designed this License in order to use it for manuals for free software, because free
software needs free documentation: a free program should come with manuals providing the
same freedoms that the software does. But this License is not limited to software manuals; it
can be used for any textual work, regardless of subject matter or whether it is published as a
printed book. We recommend this License principally for works whose purpose is instruction
or reference.

1. APPLICABILITY AND DEFINITIONS

This License applies to any manual or other work, in any medium, that contains a notice
placed by the copyright holder saying it can be distributed under the terms of this License.
Such a notice grants a world-wide, royalty-free license, unlimited in duration, to use that
work under the conditions stated herein. The "Document", below, refers to any such manual
or work. Any member of the public is a licensee, and is addressed as "you". You accept
the license if you copy, modify or distribute the work in a way requiring permission under
copyright law.
A "Modified Version" of the Document means any work containing the Document or a
portion of it, either copied verbatim, or with modifications and/or translated into another
language.

218
 219

A "Secondary Section" is a named appendix or a front-matter section of the Document

that deals exclusively with the relationship of the publishers or authors of the Document
to the Document’s overall subject (or to related matters) and contains nothing that could
fall directly within that overall subject. (Thus, if the Document is in part a textbook of
mathematics, a Secondary Section may not explain any mathematics.) The relationship
could be a matter of historical connection with the subject or with related matters, or of
legal, commercial, philosophical, ethical or political position regarding them.
The "Invariant Sections" are certain Secondary Sections whose titles are designated,
as being those of Invariant Sections, in the notice that says that the Document is released
under this License. If a section does not fit the above definition of Secondary then it is not
allowed to be designated as Invariant. The Document may contain zero Invariant Sections.
If the Document does not identify any Invariant Sections then there are none.
The "Cover Texts" are certain short passages of text that are listed, as Front-Cover Texts
or Back-Cover Texts, in the notice that says that the Document is released under this Li-
cense. A Front-Cover Text may be at most 5 words, and a Back-Cover Text may be at most
25 words.
A "Transparent" copy of the Document means a machine-readable copy, represented in
a format whose specification is available to the general public, that is suitable for revising
the document straightforwardly with generic text editors or (for images composed of pixels)
generic paint programs or (for drawings) some widely available drawing editor, and that
is suitable for input to text formatters or for automatic translation to a variety of formats
suitable for input to text formatters. A copy made in an otherwise Transparent file format
whose markup, or absence of markup, has been arranged to thwart or discourage subsequent
modification by readers is not Transparent. An image format is not Transparent if used for
any substantial amount of text. A copy that is not "Transparent" is called "Opaque".
Examples of suitable formats for Transparent copies include plain ASCII without markup,
Texinfo input format, LaTeX input format, SGML or XML using a publicly available DTD,
and standard-conforming simple HTML, PostScript or PDF designed for human modifica-
tion. Examples of transparent image formats include PNG, XCF and JPG. Opaque formats
include proprietary formats that can be read and edited only by proprietary word proces-
sors, SGML or XML for which the DTD and/or processing tools are not generally available,
and the machine-generated HTML, PostScript or PDF produced by some word processors for
output purposes only.
The "Title Page" means, for a printed book, the title page itself, plus such following pages
as are needed to hold, legibly, the material this License requires to appear in the title page.
For works in formats which do not have any title page as such, "Title Page" means the text
near the most prominent appearance of the work’s title, preceding the beginning of the body
of the text.
A section "Entitled XYZ" means a named subunit of the Document whose title either
is precisely XYZ or contains XYZ in parentheses following text that translates XYZ in an-
other language. (Here XYZ stands for a specific section name mentioned below, such as
"Acknowledgments", "Dedications", "Endorsements", or "History".) To "Preserve the
220 GNU FREE DOCUMENTATION LICENSE

Title" of such a section when you modify the Document means that it remains a section
"Entitled XYZ" according to this definition.
The Document may include Warranty Disclaimers next to the notice which states that
this License applies to the Document. These Warranty Disclaimers are considered to be
included by reference in this License, but only as regards disclaiming warranties: any other
implication that these Warranty Disclaimers may have is void and has no effect on the
meaning of this License.

2. VERBATIM COPYING
You may copy and distribute the Document in any medium, either commercially or non-
commercially, provided that this License, the copyright notices, and the license notice say-
ing this License applies to the Document are reproduced in all copies, and that you add no
other conditions whatsoever to those of this License. You may not use technical measures to
obstruct or control the reading or further copying of the copies you make or distribute. How-
ever, you may accept compensation in exchange for copies. If you distribute a large enough
number of copies you must also follow the conditions in section 3.
You may also lend copies, under the same conditions stated above, and you may publicly
display copies.

3. COPYING IN QUANTITY
If you publish printed copies (or copies in media that commonly have printed covers) of
the Document, numbering more than 100, and the Document’s license notice requires Cover
Texts, you must enclose the copies in covers that carry, clearly and legibly, all these Cover
Texts: Front-Cover Texts on the front cover, and Back-Cover Texts on the back cover. Both
covers must also clearly and legibly identify you as the publisher of these copies. The front
cover must present the full title with all words of the title equally prominent and visible.
You may add other material on the covers in addition. Copying with changes limited to the
covers, as long as they preserve the title of the Document and satisfy these conditions, can
be treated as verbatim copying in other respects.
If the required texts for either cover are too voluminous to fit legibly, you should put the
first ones listed (as many as fit reasonably) on the actual cover, and continue the rest onto
adjacent pages.
If you publish or distribute Opaque copies of the Document numbering more than 100,
you must either include a machine-readable Transparent copy along with each Opaque copy,
or state in or with each Opaque copy a computer-network location from which the general
network-using public has access to download using public-standard network protocols a com-
plete Transparent copy of the Document, free of added material. If you use the latter option,
you must take reasonably prudent steps, when you begin distribution of Opaque copies in
quantity, to ensure that this Transparent copy will remain thus accessible at the stated lo-
cation until at least one year after the last time you distribute an Opaque copy (directly or
through your agents or retailers) of that edition to the public.
 221

It is requested, but not required, that you contact the authors of the Document well before
redistributing any large number of copies, to give them a chance to provide you with an
updated version of the Document.

4. MODIFICATIONS

You may copy and distribute a Modified Version of the Document under the conditions of
sections 2 and 3 above, provided that you release the Modified Version under precisely this
License, with the Modified Version filling the role of the Document, thus licensing distribu-
tion and modification of the Modified Version to whoever possesses a copy of it. In addition,
you must do these things in the Modified Version:

A. Use in the Title Page (and on the covers, if any) a title distinct from that of the Document,
and from those of previous versions (which should, if there were any, be listed in the
History section of the Document). You may use the same title as a previous version if the
original publisher of that version gives permission.

B. List on the Title Page, as authors, one or more persons or entities responsible for au-
thorship of the modifications in the Modified Version, together with at least five of the
principal authors of the Document (all of its principal authors, if it has fewer than five),
unless they release you from this requirement.

C. State on the Title page the name of the publisher of the Modified Version, as the pub-
lisher.

D. Preserve all the copyright notices of the Document.

E. Add an appropriate copyright notice for your modifications adjacent to the other copy-
right notices.

F. Include, immediately after the copyright notices, a license notice giving the public per-
mission to use the Modified Version under the terms of this License, in the form shown
in the Addendum below.

G. Preserve in that license notice the full lists of Invariant Sections and required Cover
Texts given in the Document’s license notice.

H. Include an unaltered copy of this License.

I. Preserve the section Entitled "History", Preserve its Title, and add to it an item stating
at least the title, year, new authors, and publisher of the Modified Version as given on the
Title Page. If there is no section Entitled "History" in the Document, create one stating
the title, year, authors, and publisher of the Document as given on its Title Page, then
add an item describing the Modified Version as stated in the previous sentence.
222 GNU FREE DOCUMENTATION LICENSE

J. Preserve the network location, if any, given in the Document for public access to a Trans-
parent copy of the Document, and likewise the network locations given in the Document
for previous versions it was based on. These may be placed in the "History" section. You
may omit a network location for a work that was published at least four years before the
Document itself, or if the original publisher of the version it refers to gives permission.

K. For any section Entitled "Acknowledgments" or "Dedications", Preserve the Title of the
section, and preserve in the section all the substance and tone of each of the contributor
acknowledgments and/or dedications given therein.

L. Preserve all the Invariant Sections of the Document, unaltered in their text and in their
titles. Section numbers or the equivalent are not considered part of the section titles.

M. Delete any section Entitled "Endorsements". Such a section may not be included in the
Modified Version.

N. Do not retitle any existing section to be Entitled "Endorsements" or to conflict in title

with any Invariant Section.

O. Preserve any Warranty Disclaimers.

If the Modified Version includes new front-matter sections or appendices that qualify as
Secondary Sections and contain no material copied from the Document, you may at your
option designate some or all of these sections as invariant. To do this, add their titles to
the list of Invariant Sections in the Modified Version’s license notice. These titles must be
distinct from any other section titles.
You may add a section Entitled "Endorsements", provided it contains nothing but endorse-
ments of your Modified Version by various parties–for example, statements of peer review
or that the text has been approved by an organization as the authoritative definition of a
standard.
You may add a passage of up to five words as a Front-Cover Text, and a passage of up to
25 words as a Back-Cover Text, to the end of the list of Cover Texts in the Modified Version.
Only one passage of Front-Cover Text and one of Back-Cover Text may be added by (or
through arrangements made by) any one entity. If the Document already includes a cover
text for the same cover, previously added by you or by arrangement made by the same entity
you are acting on behalf of, you may not add another; but you may replace the old one, on
explicit permission from the previous publisher that added the old one.
The author(s) and publisher(s) of the Document do not by this License give permission to
use their names for publicity for or to assert or imply endorsement of any Modified Version.

5. COMBINING DOCUMENTS
You may combine the Document with other documents released under this License, under
the terms defined in section 4 above for modified versions, provided that you include in the
 223

combination all of the Invariant Sections of all of the original documents, unmodified, and
list them all as Invariant Sections of your combined work in its license notice, and that you
preserve all their Warranty Disclaimers.
The combined work need only contain one copy of this License, and multiple identical In-
variant Sections may be replaced with a single copy. If there are multiple Invariant Sections
with the same name but different contents, make the title of each such section unique by
adding at the end of it, in parentheses, the name of the original author or publisher of that
section if known, or else a unique number. Make the same adjustment to the section titles
in the list of Invariant Sections in the license notice of the combined work.
In the combination, you must combine any sections Entitled "History" in the various orig-
inal documents, forming one section Entitled "History"; likewise combine any sections En-
titled "Acknowledgments", and any sections Entitled "Dedications". You must delete all
sections Entitled "Endorsements".

6. COLLECTIONS OF DOCUMENTS
You may make a collection consisting of the Document and other documents released un-
der this License, and replace the individual copies of this License in the various documents
with a single copy that is included in the collection, provided that you follow the rules of this
License for verbatim copying of each of the documents in all other respects.
You may extract a single document from such a collection, and distribute it individually
under this License, provided you insert a copy of this License into the extracted document,
and follow this License in all other respects regarding verbatim copying of that document.

7. AGGREGATION WITH INDEPENDENT WORKS

A compilation of the Document or its derivatives with other separate and independent
documents or works, in or on a volume of a storage or distribution medium, is called an "ag-
gregate" if the copyright resulting from the compilation is not used to limit the legal rights
of the compilation’s users beyond what the individual works permit. When the Document
is included in an aggregate, this License does not apply to the other works in the aggregate
which are not themselves derivative works of the Document.
If the Cover Text requirement of section 3 is applicable to these copies of the Document,
then if the Document is less than one half of the entire aggregate, the Document’s Cover
Texts may be placed on covers that bracket the Document within the aggregate, or the elec-
tronic equivalent of covers if the Document is in electronic form. Otherwise they must ap-
pear on printed covers that bracket the whole aggregate.

8. TRANSLATION
Translation is considered a kind of modification, so you may distribute translations of the
Document under the terms of section 4. Replacing Invariant Sections with translations re-
quires special permission from their copyright holders, but you may include translations of
224 GNU FREE DOCUMENTATION LICENSE

some or all Invariant Sections in addition to the original versions of these Invariant Sections.
You may include a translation of this License, and all the license notices in the Document,
and any Warranty Disclaimers, provided that you also include the original English version
of this License and the original versions of those notices and disclaimers. In case of a dis-
agreement between the translation and the original version of this License or a notice or
disclaimer, the original version will prevail.
If a section in the Document is Entitled "Acknowledgments", "Dedications", or "History",
the requirement (section 4) to Preserve its Title (section 1) will typically require changing
the actual title.

9. TERMINATION

You may not copy, modify, sublicense, or distribute the Document except as expressly pro-
vided for under this License. Any other attempt to copy, modify, sublicense or distribute the
Document is void, and will automatically terminate your rights under this License. How-
ever, parties who have received copies, or rights, from you under this License will not have
their licenses terminated so long as such parties remain in full compliance.

10. FUTURE REVISIONS OF THIS LICENSE

The Free Software Foundation may publish new, revised versions of the GNU Free Doc-
umentation License from time to time. Such new versions will be similar in spirit to the
present version, but may differ in detail to address new problems or concerns. See http://
www.gnu.org/copyleft/.
Each version of the License is given a distinguishing version number. If the Document
specifies that a particular numbered version of this License "or any later version" applies to
it, you have the option of following the terms and conditions either of that specified version or
of any later version that has been published (not as a draft) by the Free Software Foundation.
If the Document does not specify a version number of this License, you may choose any
version ever published (not as a draft) by the Free Software Foundation.

ADDENDUM: How to use this License for your documents

To use this License in a document you have written, include a copy of the License in the
document and put the following copyright and license notices just after the title page:

Copyright ©YEAR YOUR NAME. Permission is granted to copy, distribute and/or

modify this document under the terms of the GNU Free Documentation License, Ver-
sion 1.2 or any later version published by the Free Software Foundation; with no
Invariant Sections, no Front-Cover Texts, and no Back-Cover Texts. A copy of the
license is included in the section entitled "GNU Free Documentation License".
 225

If you have Invariant Sections, Front-Cover Texts and Back-Cover Texts, replace the
"with...Texts." line with this:

with the Invariant Sections being LIST THEIR TITLES, with the Front-Cover Texts
being LIST, and with the Back-Cover Texts being LIST.

If you have Invariant Sections without Cover Texts, or some other combination of the
three, merge those two alternatives to suit the situation.
If your document contains nontrivial examples of program code, we recommend releas-
ing these examples in parallel under your choice of free software license, such as the GNU
General Public License, to permit their use in free software.
History
This section contains the revision history of the book. For persons making modifications to
the book, please record the pertinent information here, following the format in the first item
below.

1. VERSION: 1.0
Date: 2008-01-04
Author(s): Michael Corral
Title: Vector Calculus
Modification(s): Initial version

2. VERSION: 1.1
Date: 2016-12-04
Author(s): Anton Petrunin
Title: Corral’s Vector Calculus
Modification(s): Minor corrections and more exercises.

226
Index

Symbols B
1 ∞
C , C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 Bézier curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 bounded set . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
M x , M y . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
M x y , M xz , M yz . . . . . . . . . . . . . . . . . . . . . . . . 144 C
R2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 capping surface . . . . . . . . . . . . . . . . . . . . . . . 200
R3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Cauchy–Schwarz Inequality . . . . . . . . . . . 17
x̄ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .142 center of mass . . . . . . . . . . . . . . . . . . . . . . . . . 142
ȳ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142 centroid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
z̄ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .144 Chain Rule . . . . . . . . . . . . . . . . . . . . . . . . . 66, 89
δ( x, y) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142 change of variable . . . . . . . . . . . . . . . . 135, 137
∂( x, y, z) circulation. . . . . . . . . . . . . . . . . . . . . . . . . . . . .198
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
∂( u, v, w) closed curve . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
∂f closed surface . . . . . . . . . . . . . . . . . . . . . . . . . 185
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
t
∂x collinear . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126 conical helix . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
r . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115, 119 conservative field . . . . . . . . . . . . . . . . . . . . . 169
C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156, 159 constrained critical point . . . . . . . . . . . . . 107
er , eθ , e z , eρ , eφ . . . . . . . . . . . . . . . . . . . . . . . 207
continuity. . . . . . . . . . . . . . . . . . . . . . . . . . .57, 78
i, j, k . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
continuously differentiable . . . . . . . . . 57, 89
v . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89, 202
∇
coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
Cartesian . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
u
Σ
curvilinear . . . . . . . . . . . . . . . . . . . . . . . . . 51
C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .166
∂ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80 cylindrical . . . . . . . . . . . . . . . . . . . . 51, 208
D v f . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87 ellipsoidal. . . . . . . . . . . . . . . . . . . . . . . . .187
d r . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160 left-handed . . . . . . . . . . . . . . . . . . . . . . . . . . 2
polar . . . . . . . . . . . . . . . . . . . . . . . . . . 52, 139
A rectangular. . . . . . . . . . . . . . . . . . . . . . . . . .1
acceleration . . . . . . . . . . . . . . . . . . . . . . . . . . 2, 61 right-handed . . . . . . . . . . . . . . . . . . . . . . . . 1
angle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 spherical . . . . . . . . . . . . . . . . . . . . . . 51, 208
annulus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176 coplanar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
arc length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
area element . . . . . . . . . . . . . . . . . . . . . . . . . . 119 covariance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
average value . . . . . . . . . . . . . . . . . . . . . . . . . 130 critical point . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

227
228 Index

cross product . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

curl. . . . . . . . . . . . . . . . . . . . . . . . . .194, 203, 208 function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
cylinder. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .45 continuous . . . . . . . . . . . . . . . . . . . . . . . . . 78
scalar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
D vector-valued . . . . . . . . . . . . . . . . . . . . . . 56
density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
derivative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 G
directional . . . . . . . . . . . . . . . . . . . . . . . . . 87 Gaussian blur . . . . . . . . . . . . . . . . . . . . . . . . . . 79
mixed partial . . . . . . . . . . . . . . . . . . . . . . 82 global maximum . . . . . . . . . . . . . . . . . . . . . . . 93
partial . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80 global minimum . . . . . . . . . . . . . . . . . . . . . . . . 93
vector-valued function . . . . . . . . . . . . . 57 gradient . . . . . . . . . . . . . . . . . . . . . . . . . . . 89, 208
determinant . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 Green’s identities . . . . . . . . . . . . . . . . . . . . . 206
differential . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160 Green’s Theorem . . . . . . . . . . . . . . . . . . . . . . 172
differential form . . . . . . . . . . . . . . . . . . . . . . 160
directed curve . . . . . . . . . . . . . . . . . . . . . . . . . 165 H
direction angles . . . . . . . . . . . . . . . . . . . . . . . . 20 harmonic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
direction cosines . . . . . . . . . . . . . . . . . . . . . . . . 20 helicoid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
directional derivative. . . . . . . . . . . . . . . . . . .87 helix . . . . . . . . . . . . . . . . . . . . . . . . . . . 56, 66, 191
distance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .6 hyperbolic paraboloid . . . . . . . . . . . . . . . . . . 47
between points . . . . . . . . . . . . . . . . . . . . . . 6 hyperboloid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
from point to line . . . . . . . . . . . . . . . . . . 35 one sheet . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
point to plane . . . . . . . . . . . . . . 39, 44, 45 two sheets. . . . . . . . . . . . . . . . . . . . . . . . . .46
distribution function . . . . . . . . . . . . . . . . . . 147 hypersurface . . . . . . . . . . . . . . . . . . . . . . . . . . 126
joint. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .150 hypervolume . . . . . . . . . . . . . . . . . . . . . . . . . . 126
normal . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
divergence . . . . . . . . . . . . . . . . . . 185, 202, 208 I
Divergence Theorem . . . . . . . . . . . . . . . . . . 185 improper integral . . . . . . . . . . . . . . . . . . . . . 123
dot product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 integral
double integral . . . . . . . . . . . . . . . . . . . 115, 119 double . . . . . . . . . . . . . . . . . . . . . . . 115, 119
polar coordinates . . . . . . . . . . . . . . . . . 139 improper . . . . . . . . . . . . . . . . . . . . . . . . . . 123
doubly ruled surface . . . . . . . . . . . . . . . . . . . . 48 iterated . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
E multiple . . . . . . . . . . . . . . . . . . . . . . . . . . 114
ellipsoid . . . . . . . . . . . . . . . . . . . . . . 46, 141, 187 surface . . . . . . . . . . . . . . . . . . . . . . . 179, 181
elliptic cone. . . . . . . . . . . . . . . . . . . . . . . . . . . . .48 triple. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .126
elliptic paraboloid . . . . . . . . . . . . . . . . . . . . . . 47 involute . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
Euclidean space . . . . . . . . . . . . . . . . . . . . . . . . . 1 irrotational . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
exact differential form . . . . . . 160, 177, 200 iterated integral . . . . . . . . . . . . . . . . . . . . . . 115
expected value . . . . . . . . . . . . . . . . . . . . . . . . 152
extreme point . . . . . . . . . . . . . . . . . . . . . . . . . . 93 J
Jacobi identity. . . . . . . . . . . . . . . . . . . . . . . . . .32
F Jacobian. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .137
flux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185 joint distribution . . . . . . . . . . . . . . . . . . . . . . 150
 Index 229

L of revolution . . . . . . . . . . . . . . . . . . . . . . . 47
Lagrange multiplier. . . . . . . . . . . . . . . . . . .107 parallelepiped . . . . . . . . . . . . . . . . . . . . . . . . . . 26
lamina . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142 volume . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
Laplacian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208 parameter . . . . . . . . . . . . . . . . . . . . . . . . . . 33, 66
level curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 parametrization . . . . . . . . . . . . . . . . . . . . . . . . 66
limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 partial derivative . . . . . . . . . . . . . . . . . . . . . . . 80
vector-valued function . . . . . . . . . . . . . 57 partial differential equation . . . . . . . . . . . . 83
line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 path independence . . . . . . . . . . 167, 177, 200
intersection of planes . . . . . . . . . . . . . . 40 pedal curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
parallel . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 piecewise smooth curve . . . . . . . . . . . . . . . 161
parametric representation . . . . . . . . . 33 plane
perpendicular . . . . . . . . . . . . . . . . . . . . . . 36 coordinate . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
skew . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 Euclidean . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
symmetric representation . . . . . . . . . 34 line of intersection . . . . . . . . . . . . . . . . . 40
through two points . . . . . . . . . . . . . . . . . 35 normal form . . . . . . . . . . . . . . . . . . . . . . . 37
vector representation . . . . . . . . . . . . . . 33 normal vector . . . . . . . . . . . . . . . . . . . . . . 37
line integral . . . . . . . . . . . . . . . . . . . . . . 156, 159 point-normal form . . . . . . . . . . . . . . . . . 37
local maximum . . . . . . . . . . . . . . . . . . . . . . . . . 93 tangent . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
local minimum . . . . . . . . . . . . . . . . . . . . . . . . . 93 through three points . . . . . . . . . . . . . . . 38
position vector . . . . . . . . . . . . . . . . . 59, 60, 159
M potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142 probability . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 probability density function . . . . . . . . . . . 148
mixed partial derivative . . . . . . . . . . . . . . . . 82 projection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
Möbius strip . . . . . . . . . . . . . . . . . . . . . . . . . . 193
moment . . . . . . . . . . . . . . . . . . . . . . . . . . 142, 144 Q
momentum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 quadric surface . . . . . . . . . . . . . . . . . . . . . . . . . 46
Monte Carlo method . . . . . . . . . . . . . . . . . . 130
multiple integral . . . . . . . . . . . . . . . . . . . . . . 114
R
random variable . . . . . . . . . . . . . . . . . . . . . . 147
multiply connected . . . . . . . . . . . . . . . . . . . . 175
regular reparametrization . . . . . . . . . . . . . 66
N reparametrization . . . . . . . . . . . . . . . . . . . . . . 66
n-positive direction . . . . . . . . . . . . . . . . . . . 193 Riemann integral . . . . . . . . . . . . . . . . . . . . . 155
normal to a curve. . . . . . . . . . . . . . . . . . . . . . .90 right-hand rule . . . . . . . . . . . . . . . . . . . . . . . . . 22
normal vector field . . . . . . . . . . . . . . . . . . . . 192 ruled surface . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

O S
orientable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192 saddle point . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
sample space . . . . . . . . . . . . . . . . . . . . . . . . . . 147
P scalar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
paraboloid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 combination . . . . . . . . . . . . . . . . . . . . . . . . 13
elliptic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 scalar function . . . . . . . . . . . . . . . . . . . . . . . . . 58
hyperbolic . . . . . . . . . . . . . . . . . . . . . . 47, 94 scalar triple product . . . . . . . . . . . . . . . . . . . . 26
230 Index

Second Derivative Test . . . . . . . . . . . . . . . . . 95 direction. . . . . . . . . . . . . . . . . . . . . . . . . . . . .3

second moment. . . . . . . . . . . . . . . . . . . . . . . .154 magnitude . . . . . . . . . . . . . . . . . . . . . 3, 6, 7
second-degree equation. . . . . . . . . . . . . . . . .46 normal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
simple closed curve . . . . . . . . . . . . . . . . . . . 166 normalized . . . . . . . . . . . . . . . . . . . . . . . . . 12
simply connected . . . . . . . . . . . . . . . . . 177, 200 parallel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
smooth function . . . . . . . . . . . . . . . . . . . . 57, 95 perpendicular . . . . . . . . . . . . . . . . . . 16, 17
solenoidal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186 positive unit normal . . . . . . . . . . . . . . 193
span . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 scalar multiplication . . . . . . . . . . . . . . . . 9
sphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 subtraction . . . . . . . . . . . . . . . . . . . . . . . . . 10
spherical spiral . . . . . . . . . . . . . . . . . . . . . . . . . 59 tangent . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
standard normal distribution . . . . . . . . . 149 translation . . . . . . . . . . . . . . . . . . . . . . . . 5, 9
steepest descent . . . . . . . . . . . . . . . . . . . . . . . 106 unit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
stereographic projection . . . . . . . . . . . . . . . . 50 zero . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Stokes’ Theorem . . . . . . . . . . . . . . . . . 192, 194 vector field . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 normal . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
doubly ruled . . . . . . . . . . . . . . . . . . . . . . . 48 smooth . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
orientable . . . . . . . . . . . . . . . . . . . . . . . . . 192 vector triple product. . . . . . . . . . . . . . . . . . . . 27
ruled . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2, 61
two-sided . . . . . . . . . . . . . . . . . . . . . . . . . 193 volume element . . . . . . . . . . . . . . . . . . . . . . . 126
surface integral . . . . . . . . . . . . . . . . . . 179, 181
W
T wave equation . . . . . . . . . . . . . . . . . . . . . . . . . . 83
tangent plane . . . . . . . . . . . . . . . . . . . . . . . . . . 84 work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155, 190
torus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
trace . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 Z
triangle inequality . . . . . . . . . . . . . . . . . . . . . 18 zenith angle . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
triple integral . . . . . . . . . . . . . . . . . . . . . . . . . 126
cylindrical coordinates . . . . . . . . . . . .140
spherical coordinates . . . . . . . . . . . . . 140

U
uniform density . . . . . . . . . . . . . . . . . . . . . . . 142
uniform distribution . . . . . . . . . . . . . . . . . . 148
uniformly distributed . . . . . . . . . . . . . . . . . 147
unit disk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

V
variance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
vector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
addition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
angle between. . . . . . . . . . . . . . . . . . . . . .15
basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
components . . . . . . . . . . . . . . . . . . . . . . . . 13