Radioactive Decay

First Order ODE Theory

Scaling and Dimensionless Variables

Math 537 - Ordinary Differential Equations

Lecture Notes – Linear Differential Equations
and Scaling

Joseph M. Mahaffy,
[email protected]

Department of Mathematics and Statistics

Dynamical Systems Group
Computational Sciences Research Center
San Diego State University
San Diego, CA 92182-7720
http://jmahaffy.sdsu.edu

Fall 2019
Lecture Notes – Linear Differential Equations a
Joseph M. Mahaffy, [email protected] — (1/58)
 Radioactive Decay
First Order ODE Theory
Scaling and Dimensionless Variables

Outline

1 Radioactive Decay
Carbon Radiodating
Art Forgery
Radioactive Cascade
Quasi-steady state
Linear Operator
2 First Order ODE Theory
Existence and Uniqueness
Example

3 Scaling and Dimensionless Variables

Buckingham Pi Theorem
Launch Example
Atomic Bomb
Scaling Parameters in ODE
Perturbation
Mass-Spring

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (2/58)
 Radioactive Decay
Carbon Radiodating
First Order ODE Theory
Art Forgery
Scaling and Dimensionless Variables

Carbon Radiodating 1
Carbon Radiodating: One important application of radioactive
decay is the dating of biological specimens.

A living organism is continually changing its carbon with the

environment.
Plants directly absorb CO2 from the atmosphere.
Animals get their carbon either directly or indirectly from
plants.
Gamma radiation that bombards the Earth keeps the ratio of
14
C to 12 C fairly constant in the atmospheric CO2 .
Atomic tests in 1950s and 1960s increased gamma radiation.
Large releases of CO2 (global warming) are making the
dating less reliable.
Until recently, 14 C stays at a constant concentration until the
organism dies.
Lecture Notes – Linear Differential Equations a
Joseph M. Mahaffy, [email protected] — (3/58)
 Radioactive Decay
Carbon Radiodating
First Order ODE Theory
Art Forgery
Scaling and Dimensionless Variables

Carbon Radiodating 2
14
Modeling Carbon Radiodating: Radioactive carbon, C, decays
with a half-life of 5730 yr.
Living tissue shows a radioactivity of about 15.3 disintegrations
per minute (dpm) per gram of carbon.
The loss of 14 C from a sample at any time t is proportional to
the amount of 14 C remaining.
14
Let R(t) be the dpm per gram of C from an ancient object.
14
The differential equation for a gram of C
dR(t)
= −kR(t) with R(0) = 15.3.
dt
This differential equation has the solution:
ln(2)
R(t) = 15.3 e−kt , where k= 5730 = 0.000121.

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (4/58)
 Radioactive Decay
Carbon Radiodating
First Order ODE Theory
Art Forgery
Scaling and Dimensionless Variables

Example: Carbon Radiodating

Example Carbon Radiodating: Suppose that an object is

found to have a radioactive count of 5.2 dpm per g of carbon
Find the age of this object.
Solution: From above

5.2 = 15.3 e−kt or ekt = 15.3

5.2 = 2.94.
ln(2)
kt = ln(2.94) with k = 5730 = 0.000121.
ln(2.94)
Thus, t = k = 8915 yr, so the object is about 9000 yrs old.

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (5/58)
 Radioactive Decay
Carbon Radiodating
First Order ODE Theory
Art Forgery
Scaling and Dimensionless Variables

Art Forgery
Van Meegeren Art Forgery: At the end of WWII, H. A. Van Meegeren was
arrested for collaborating with the Germans in the sale of the painting “Woman
Taken in Adultery” by Jan Vermeer to Goering.1

Van Meegeren was considered a 3rd rate painter.

From prison to avoid charges of treason announced he forged this painting
and a number of other famous paintings.
Several of the paintings were extremely well-done, so several art experts
didn’t believe him.
He was just about to show how he created the masterpieces, including his
technique to age the paintings, when charges were changed to forgery.
Some of the paintings were easily shown to be forgeries, so he was convicted
and sent to prison.
However, his “Disciples of Emmaus” was so good, it fooled experts and was
certified genuine.
Scientists used the radioactive 210 Pb from lead oxide (white lead) to prove it
was a fake.

1. M. Braun, Differential Equations and Their Applications, Springer-Verlag, 1983.

Lecture Notes – Linear Differential Equations a
Joseph M. Mahaffy, [email protected] — (6/58)
 Radioactive Decay
Carbon Radiodating
First Order ODE Theory
Art Forgery
Scaling and Dimensionless Variables

Radioactive Cascade
Radioactive Lead 210 Pb: One source of the stable element lead, 206 Pb, is
through a series of decaying radioactive elements starting with uranium, 238 U
(half-life 4.5 billion years), and ending with lead, 206 Pb.
We focus on the cascade of elements from radium, 226 Ra to lead, 206 Pb.

h1 h2 h3 h4
226 Ra −→ 222 Rn −→ 218 Po −→ 214 Pb −→ 214 Bi

h5 h6 h7 h8 h9
−→ 214 Po −→ 210 Pb −→ 210 Bi −→ 210 Po −→ 206 Pb

The half-lives are given by hi with

h1 = 1600 yr h2 = 3.82 da h3 = 3.05 min h4 = 26.8 min h5 = 19.7 min
h6 = 0.16 msec h7 = 22 yr h8 = 5.0 da h9 = 138 da -

This radioactive cascade system can be formulated into a system of

ordinary differential equations. This course will study large linear
systems of ODEs.
The decay rates span a wide range, and this course will examine how to
manage some multi-scale problems.
Often try to connect theory to practical problems.
Lecture Notes – Linear Differential Equations a
Joseph M. Mahaffy, [email protected] — (7/58)
 Radioactive Decay
Carbon Radiodating
First Order ODE Theory
Art Forgery
Scaling and Dimensionless Variables

Radioactive ODE System

ODE System: Define x0 as the amount of 226 Ra, x1 as the amount of 222 Rn, x2
as the amount of 218 Po, etc.
ln(2)
Define the decay rates ki = hi
, i = 1, ..., 9.
The linear nonhomogeneous system of ODEs satisfy:

ẋi = ki−1 xi−1 − ki xi , i = 1, ..., 9,

where k0 x0 is the rate of decay of 226 Ra × amount of 226 Ra.

With x = [x1 , x2 , ..., x9 ]T , this can be written as the ODE system:

ẋ = Ax + B,

where
−k1
 
0 ... ... 0
−k2 k 0 x0 0 ... 0
 
 k1 0 ... 0 
0 0 ... 0
 
 .. ..   
A=

0 k2 −k3 . .

 B=

.. .. .. ..



 .. .. .. ..


 . . . . 
 . . . . 0  0 0 ... 0
0 ... 0 k8 −k9
Lecture Notes – Linear Differential Equations a
Joseph M. Mahaffy, [email protected] — (8/58)
 Radioactive Decay
Carbon Radiodating
First Order ODE Theory
Art Forgery
Scaling and Dimensionless Variables

Radioactive – Quasi-steady state

Half-lives: The earlier table shows the half-lives in this radioactive
cascade varying from h1 = 1600 yr to h6 = 0.16 msec, which is an
extremely wide range of time scales.
When time scales vary by orders of magnitude, one often invokes a
quasi-steady state assumption that the specific equation is so fast
that it is effectively in a temporary steady state or equilibrium, so
ẋi = 0.
This reduces the dimension of the ODE by creating an algebraic
equation:

ẋi = 0 = ki−1 xi−1 − ki xi or ki−1 xi−1 = ki xi ,

which can be substituted into the system of ODEs.
This course will examine slow and fast time scales, scaling of
variables, and possibly study multi-time scale problems.
Lecture Notes – Linear Differential Equations a
Joseph M. Mahaffy, [email protected] — (9/58)
 Radioactive Decay
Carbon Radiodating
First Order ODE Theory
Art Forgery
Scaling and Dimensionless Variables

Art Forgery

Van Meegeren Art Forgery: White lead has been used in paintings for over
2000 years.

The smelting process to obtain lead removes much of the radium, which
decays to 210 Pb.
The method to detect forgeries looks at the ratio of 210 Pb to 226 Ra.

An authentic Vermeer would have an age of over 300 yrs, while the Van
Meegeren forgeries would be fairly recent.
With these time scales we can assume that the 226 Ra has a roughly constant
amount.
The other intermediate elements are on a fast time scale allowing
quasi-steady state approximations.
If y(t) is the amount of 210 Pb, then these assumptions result in the scalar
ODE:
ln(2)
ẏ = r − ky, where k = 22 .

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (10/58)
 Radioactive Decay
Carbon Radiodating
First Order ODE Theory
Art Forgery
Scaling and Dimensionless Variables

Linear ODE
Radioactive Decay: The ODE for a radioactive substance, y,
which has a constant source, satisfies:

ẏ + ky = r.

This ODE is

Linear, as the dependent variable, y, and its derivative only

appears linearly.
First Order, as the equation only has the first derivative.
Non-homogeneous, as the constant, r, appears on the rhs.

We use operator notation:

 
d
L[y] = r, where L[y] = + k y.
dt
Lecture Notes – Linear Differential Equations a
Joseph M. Mahaffy, [email protected] — (11/58)
 Radioactive Decay
Carbon Radiodating
First Order ODE Theory
Art Forgery
Scaling and Dimensionless Variables

Linear Operator

Linear Operator: We show that L is a linear operator.

d


L[αy1 + βy2 ] = dt + k [αy1 + βy2 ]

d
= dt (αy1 + βy2 ) + k(αy1 + βy2 )

= α dy dy2
dt + β dt + αky1 + βky2
1

   
= α dy dt
1
+ ky 1 + β dy2
dt + ky2

= αL[y1 ] + βL[y2 ]

Homogeneous problem is L[y] = 0.

Non-homogeneous problem is L[y] = r(t).

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (12/58)
 Radioactive Decay
Carbon Radiodating
First Order ODE Theory
Art Forgery
Scaling and Dimensionless Variables

Initial Value Problem

Solution of Initial Value Problem: Consider the problem:
ẏ + ky = r, y(t0 ) = y0 .

This problem is readily solved with an integrating factor.

By multiplying the equation above by µ(t) = ekt , the left hand side
becomes an exact differential:
 
ekt dy d
ekt y(t) = rekt .


dt + ky = dt

Integrating produces
Z t Z t
d
eks y(s) ds = reks ds,


ds
t0 t0
or
r kt
ekt y(t) − ekt0 y0 = e − ekt0 .


k
Lecture Notes – Linear Differential Equations a
Joseph M. Mahaffy, [email protected] — (13/58)
 Radioactive Decay
Carbon Radiodating
First Order ODE Theory
Art Forgery
Scaling and Dimensionless Variables

Initial Value Problem

Solution of Initial Value Problem: It follows that the solution of

ẏ + ky = r, y(t0 ) = y0 ,

is r 
y(t) = y0 e−k(t−t0 ) + 1 − e−k(t−t0 ) .
k

Equilibrium: Note the equilibrium occurs when ẏ = 0, so from the ODE we see
kye = r or
r
ye = .
k
From the solution above, we see:

r
lim y(t) = .
t→∞ k

It follows that ye = kr is an asymptotically stable equilibrium with all solutions

approaching ye for large time.
We also note that the homogeneous ODE has the eigenvalue, λ = −k < 0.
Lecture Notes – Linear Differential Equations a
Joseph M. Mahaffy, [email protected] — (14/58)
 Radioactive Decay
Carbon Radiodating
First Order ODE Theory
Art Forgery
Scaling and Dimensionless Variables

Art Forgery

Van Meegeren Art Forgery: Experts believed the “Disciples of Emmaus” was
so good that it must be a real Jan Vermeer.
Samples of the white paint with white lead
were analyzed for 210 Pb and 226 Ra.
As a surrogate for 210 Pb, 210 Po was
measured at 8.5 distingrations/min
(quasi-steady state).
The 226 Ra decay was 0.8 distingrations/min.
These data translate into ky(t) = 8.5 and
r = 0.8 when measured.
Find ky0 when the pigment was formulated.
If this was a real Vermeer, then t − t0 ≈ 300 yr.

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (15/58)
 Radioactive Decay
Carbon Radiodating
First Order ODE Theory
Art Forgery
Scaling and Dimensionless Variables

Art Forgery

Van Meegeren Art Forgery: From the solution of the radioactive decay
problem, the data from the white lead measurements, and the half-life of 210 Pb,
we have:  
ky(t) = 8.5 ≈ ky0 e−300k + r 1 − e−300k ,

ln(2)
where k = 22
≈ 0.0315 and r = 0.8.
It follows that

ky0 ≈ (8.5 − 0.8)e300k + 0.8 ≈ 98, 049 disintegrations/min.

The richest known sources of 226 Ra have only a few thousand disintegrations/min.
It follows that the source of the white pigment could not have been refined 300 yr
ago, so the painting is significantly more recent.
This painting was a Van Meegeren forgery created in the 1930s.

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (16/58)
 Radioactive Decay
Existence and Uniqueness
First Order ODE Theory
Example
Scaling and Dimensionless Variables

Linear Differential Equation 1

Consider the Linear Differential Equation:

y 0 + p(t)y = g(t), with y(t0 ) = y0 . (1)

Assume p and g are continuous on an open interval I : α < t < β with

t0 ∈ (α, β), so p and g are integrable on I.

Definition (Integrating Factor)

Consider an undetermined function µ(t) with

d dy dµ(t)
[µ(t)y] = µ(t) + y.
dt dt dt
The function µ(t) is an integrating factor for (1) if it satisfies the
differential equation
dµ(t)
= p(t)µ(t).
dt
Lecture Notes – Linear Differential Equations a
Joseph M. Mahaffy, [email protected] — (17/58)
 Radioactive Decay
Existence and Uniqueness
First Order ODE Theory
Example
Scaling and Dimensionless Variables

Linear Differential Equation 2

The differential equation for the integrating factor is

dµ(t) 1 dµ(t)
= p(t)µ(t) or = p(t).
dt µ(t) dt
d(ln(µ(t))) 1 dµ(t)
Note that dt = µ(t) dt .

From the Fundamental Theorem of Calculus:

Z t
ln(µ(t)) − ln(µ(t0 )) = p(s)ds.
t0

It follows that the general integrating factor satisfies

Rt
p(s)ds
µ(t) = e t0
.
(Note: ln(µ(t0 )) = 0, as µ(t0 ) = 1.)

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (18/58)
 Radioactive Decay
Existence and Uniqueness
First Order ODE Theory
Example
Scaling and Dimensionless Variables

Linear Differential Equation 3

Multiplying the Linear Differential Equation by µ(t) gives:

µ(t) y 0 + p(t)y = d


dt
[µ(t)y] = µ(t)g(t),

which upon integration gives:

Z t
µ(t)y(t) − y0 = µ(s)g(s)ds.
t0

It follows that the unique solution to the linear ODE (1) is

Z t 
1
y(t) = µ(t)
µ(s)g(s)ds + y0 ,
t0

where Rt
t0 p(s)ds
µ(t) = e .

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (19/58)
 Radioactive Decay
Existence and Uniqueness
First Order ODE Theory
Example
Scaling and Dimensionless Variables

Linear Differential Equation 4

The linear ODE has the following existence and uniqueness result.
Theorem
If the functions p and g are continuous on an open interval
I : α < t < β containing a point t = t0 , then there exists a unique
function y = φ(t) that satisfies the differential equation

y 0 + p(t)y = g(t)

for each t in I with the initial condition

y(t0 ) = y0 ,

where y0 is an arbitrary prescribed initial value.

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (20/58)
 Radioactive Decay
Existence and Uniqueness
First Order ODE Theory
Example
Scaling and Dimensionless Variables

Linear Differential Equation 5

The summary result for the linear ODE is the following.

Theorem (Solution of 1st Order Linear DE)

Assume the 1st order linear ODE given by (1) and the conditions on p
and g of the previous theorem. Then there exists a unique solution for
t ∈ (α, β) given by
Z t 
1
y(t) = µ(t) µ(s)g(s)ds + y0 ,
t0

where Rt
p(s)ds
µ(t) = e t0
.

We note that the existence and uniqueness result for the 1st order
nonlinear ODE is more complicated and may be examined later in
the course.
Lecture Notes – Linear Differential Equations a
Joseph M. Mahaffy, [email protected] — (21/58)
 Radioactive Decay
Existence and Uniqueness
First Order ODE Theory
Example
Scaling and Dimensionless Variables

Falling Object with Drag 1

Example: Consider an object falling under the force of gravity with
resistance linearly increasing in time and proportional to the velocity
of the object.
By Newton’s Law we can write this as a balance of forces:
dv
m = −ktv − mg,
dt
k
which when divided by the mass, m, and defining α = m becomes the
linear ODE:
dv
+ αtv = −g, with v(0) = v0 .
dt

The integrating factor is given by:

Rt 2
αs ds
µ(t) = e 0 = eαt /2
.
Lecture Notes – Linear Differential Equations a
Joseph M. Mahaffy, [email protected] — (22/58)
 Radioactive Decay
Existence and Uniqueness
First Order ODE Theory
Example
Scaling and Dimensionless Variables

Falling Object with Drag 2

2
Example: With the integrating factor, µ(t) = eαt /2 , the solution
becomes:
Z t
−αt2 /2 −αt2 /2 2
v(t) = v0 e +e (−g)eαs /2 ds,
0
Z t
2 2 2
−αt /2 −αt /2
v(t) = v0 e −ge eαs /2 ds.
0

What is the limiting velocity for this model?

2
The first term, v0 e−αt /2
, clearly tends to zero as t → ∞.
However, what about
t
Rt 2
eαs /2 ds
Z
2 2
lim e−αt /2
eαs /2
ds = lim 0
?
t→∞ 0 t→∞ eαt2 /2

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (23/58)
 Radioactive Decay
Existence and Uniqueness
First Order ODE Theory
Example
Scaling and Dimensionless Variables

Falling Object with Drag 3

2
Example: Since eαs /2 > 1, the integral is unbounded, which implies
both numerator and denominator tend to +∞.
So we apply L’Hôspital’s Rule, which gives:
R t αs2 /2 2
0
e ds eαt /2 1
lim αt2 /2 = lim αt2 /2 = lim = 0.
t→∞ e t→∞ αt e t→∞ αt

It follows that:
lim v(t) = 0,
t→∞

so the falling object would eventually have sufficient drag to halt the
object.

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (24/58)
 Radioactive Decay
Existence and Uniqueness
First Order ODE Theory
Example
Scaling and Dimensionless Variables

Parameters and Scaling

Parameters: Crucial to any modeling problem are the parameters.

These often describe important properties of the system.

Parameters are often fit to data.
Variations in parameters may have a critical role in the
dynamical behavior.
Questions arise on what is called identifiability of the
parameters.
More parameters can complicate the understanding of key
underlying behaviors.

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (25/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Buckingham Pi Theorem

Theorem (Buckingham Pi Theorem)

Let q1 , q2 , q3 , ..., qn be n dimensional variables that are physically
relevant in a given problem and that are inter-related by an
(unknown) dimensionally homogeneous set of equations. These can be
expressed via a functional relationship of the form:

F (q1 , q2 , ...qn ) = 0 or equivalently q1 = f (q2 , ...qn ).

If k is the number of fundamental dimensions required to describe the

n variables, then there will be k primary variables and the remaining
variables can be expressed as (n − k) dimensionless and independent
quantities or Pi groups, Π1 , Π2 , ..., Πn−k . The functional relationship
can tbe reduced to the much more compact form:

Φ(Π1 , Π2 , , Πn−k ) = 0 or equivalently Π1 = Φ(Π2 , , Πn−k ).

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (26/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Rayleigh’s Method of Dimensional Analysis

Rayleigh’s method of dimensional analysis

Gather all the independent variables that are likely to influence the
dependent variable.
If R is a variable that depends upon independent variables
R1 , R2 , R3 , ..., Rn , then the functional equation can be written as
R = F (R1 , R2 , R3 , ..., Rn ).
Write the above equation in the form R = CR1a R2b R3c ...Rn m , where C is a

dimensionless constant and a, b, c, ..., m are arbitrary exponents.

Express each of the quantities in the equation in some base units in which
the solution is required.
By using dimensional homogeneity, obtain a set of simultaneous equations
involving the exponents a, b, c, ..., m.
Solve these equations to obtain the value of exponents a, b, c, ..., m.
Substitute the values of exponents in the main equation, and form the
non-dimensional parameters by grouping the variables with like exponents.

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (27/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Dimensional Analysis
Dimensional Analysis – Primary Units
There are a number of primary units:
Length Mass Time Amount Temperature Electricity Luminosity
L M T N Q I C

Example 1: Newton’s Law of Force is given by

F = ma

This could be written

F
− 1 = 0,
ma
which gives the dimensionless quantity
F
Π= , so f (Π) = Π − 1.
ma
Lecture Notes – Linear Differential Equations a
Joseph M. Mahaffy, [email protected] — (28/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Dimensional Analysis – Example 1

Example - Launching: Consider launching an object with critical
quantities: m = mass, v = launch velocity, h = maximum height, and
g = acceleration gravity
Choose:

[m] = M [v] = LT −1 [h] = L [g] = LT −2

Create the dimensionless quantity:

Π = ma v b hc g d

Analyze the exponents for quantities M , L, and T , so to be

dimensionless

a=0 b+c+d=0 − b − 2d = 0.

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (29/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Dimensional Analysis – Example 2

Example (cont): There are 4 coefficients a, b, c, and d for the 3
dimensional variables M , L, and T , leaving one free parameter.
With the one degree of freedom, we take d = c and c = 1, then the
coefficients become

a=0 b = −2 c=1 d = 1.

The dimensionless variable is

 
hg hg
Π= 2 f (Π) = f = 0.
v v2

It follows that
hg kv 2
=k or h= .
v2 g

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (30/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Dimensional Analysis – Example 3

Example (cont): Since
kv 2
h= ,
g
it follows that the height of a launch depends only on the quantity
v 2 /g.

The height of a launch is independent of the mass.

The height of a launch varies as the square of the velocity.
The height of a launch is inversely proportional to the
acceleration of gravity.

It follows that doubling the launch velocity increases the height of the
launch by a factor of 4.
On the moon with gravity, g6 , the height of the launch increases by a
factor of 6.
Lecture Notes – Linear Differential Equations a
Joseph M. Mahaffy, [email protected] — (31/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Dimensional Analysis – Atomic Bomb 1

Example – Atomic Bomb: Sir Geoffrey Taylor F.R.S., The

formation of a blast wave by a very intense explosion: II. The atomic
explosion of 1945, Proc. R. Soc. Lond., A, (1950)
This article used a movie of the Trinity test with dimensional analysis
to estimate the power of the explosion.
Pictures of the White Sands, NM test in 1945 showed the radius of
explosion:

Time, t (sec) Blast Radius, R (m) Time, t (sec) Blast Radius, R (m)
0.00038 25.4 0.0008 34.2
0.00052 28.8 0.00094 36.3
0.00066 31.9 0.00108 38.9

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (32/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Dimensional Analysis – Atomic Bomb 2

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (33/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Dimensional Analysis – Atomic Bomb 3

Atomic Bomb (cont): Assume that the radius, R, of Atomic blast

depends only on time, t, ambient density, ρ, and Energy, E, of the
explosion – we ignore other effects.
From the Buckingham Pi Theorem, the dimensionless variable
satisfies:
Π = Ra E b tc ρd ,
where
M L2 M
[R] = L [E] = [t] = T [ρ] = ,
T2 L3
so b d
M L2
 
M
Π = La Tc .
T2 L3

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (34/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Dimensional Analysis – Atomic Bomb 4

Atomic Bomb (cont): From before, the dimensionless variable

satisfies: b  d
M L2

M
Π = La T c
.
T2 L3

From the coefficients above we have

a + 2b − 3d = 0 (L)
b+d = 0 (M )
−2b + c = 0 (T )

There is one degree of freedom, so let b = 1, then

a = −5 b=1 c=2 d = −1.

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (35/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Dimensional Analysis – Atomic Bomb 5

Atomic Bomb (cont): From the dimensionless variable, we

write  2 1/5
−5 2 −1 Et
Π = R Et ρ or R=k .
ρ
The Taylor article goes to some length to show that k ≈ 1 and ρ ≈ 1.
Air has ρ = 1.2 kg/m3 at sea level, and White sands is at 1200 m,
which has a density of 1.03 kg/m3
It follows that
R = (Et2 )1/5 ,
which is a power law or allometric model and
1 2
ln(R) = ln(E) + ln(t).
5 5

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (36/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Dimensional Analysis – Atomic Bomb 6

Below are graphs of the data and the ln of the data:

45 3.8

40
3.7
35
3.6
30
Radius (m)

25 3.5

ln(R)
20 R = 604.15t0.4024 3.4
15
3.3
10 ln(R) = 0.4024 ln(t) + 6.4038
3.2
5

0 3.1
0 0.2 0.4 0.6 0.8 1 1.2 -8 -7.5 -7 -6.5
Time (sec) ×10-3 ln(t)

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (37/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Dimensional Analysis – Atomic Bomb 7

From before we have the power law or allometric model
1 2
R = (Et2 )1/5 or ln(R) = ln(E) + ln(t),
5 5
and the slope of the logarithmic graph from the data agrees with the
coefficient obtained by dimensional analysis.
From the data we obtain the intercept, so
1
ln(E) = 6.4038,
5
which is readily solved for E giving the energy of the atomic blast as

E = e32.02 = 8.05 × 1013 J.

Scientists running experiments at the blast site measured the power of

the trinity atomic blast as 9 × 1013 J.
Lecture Notes – Linear Differential Equations a
Joseph M. Mahaffy, [email protected] — (38/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Scaling Parameters in ODE 1

Consider the ODE given by:

dy
+ αp(ωt)y = βq(νt),
dt
which has the 4 parameters: α, β, ω, and ν.
We can reduce the parameters to 2 by rescaling the dependent
variable, y, and time, t:

y = Az and τ = Bt.

Positive: Easier to determine types of qualitative behavior and

fewer parameters to fit.
Negative: Scaled parameters may not match natural kinetic
parameters and may be hard to unravel fitting experiments.

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (39/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Scaling Parameters in ODE 2

The scaled ODE becomes:
dz ω

ν


AB + Aαp Bτ z = βq Bτ ,
dτ
which is equivalent to
dz α ω

β ν


+ Bp Bτ z= AB q Bτ .
dτ
This reduces to 2 parameters by taking:
α β
B =1 and AB = 1,
β
or B = α and A = α.
This becomes the scaled ODE:
dz
+ p (ω̂τ ) z = q (ν̂τ ) ,
dτ
where ω̂ = ω/α and ν̂ = ν/α.
Lecture Notes – Linear Differential Equations a
Joseph M. Mahaffy, [email protected] — (40/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Perturbation 1
The previous scaled ODE gave:
dz
+ p (ω̂τ ) z = q (ν̂τ ) ,
dτ
where ω̂ = ω/α and ν̂ = ν/α.
Suppose the time arguments of p and q vary rapidly, so
ω ν
ω̂ = α 1 and ν̂ = α  1,

or better written

ω̂ = 1ε ω̂0 and ν̂ = 1ε ν̂0 with ε  1 or 1

ε  1.

The ODE can be written:

dz ω̂0

ν̂0


+p ε τ z=q ε τ ,
dτ
where p and q are varying quickly.
Lecture Notes – Linear Differential Equations a
Joseph M. Mahaffy, [email protected] — (41/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Perturbation 2

With the ODE:

dz ω̂0

ν̂0


+p ε τ z=q ε τ ,
dτ
dz dz dT 1 dz
we scale again with T = τ /ε, so dτ = dT dτ = ε dT .
This results in the following:
1 dz
+ p (ω̂0 T ) z = q (ν̂0 T ) ,
ε dT
or
dz
+ εp (ω̂0 T ) z = εq (ν̂0 T ) ,
dT
which is solved from our linear ODE technique.

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (42/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Perturbation 3

The integrating factor is:

RT
µ(T ) = eε 0
p(ω̂0 s)ds
,

which gives the solution:

RT RT
Z T Rs
z(T ) = z(0)e−ε 0
p(ω̂0 s)ds
+ εe−ε 0
p(ω̂0 s)ds
q(ν̂0 s)eε 0
p(ω̂0 u)du
ds.
0

However, by Taylor’s Theorem:

!2
T T
ε2
RT
Z Z
−ε p(ω̂0 s)ds
e 0 =1−ε p(ω̂0 s)ds + p(ω̂0 s)ds − ...
0 2! 0

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (43/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Perturbation 4

RT
With this expansion of e−ε 0
p(ω̂0 s)ds
inserted into the solution:
RT RT
Z T Rs
z(T ) = z(0)e−ε 0
p(ω̂0 s)ds
+ εe−ε 0
p(ω̂0 s)ds
q(ν̂0 s)eε 0
p(ω̂0 u)du
ds.
0

we can collect the terms based on the order of ε and obtain:

Z T Z T !
p(ω̂0 s)ds + O ε2 .


z(T ) = z(0) + ε q(ν̂0 s)ds − z(0)
0 0

For ε small the O ε2 terms are insignificant, which means the
solution is approximated by the first two terms of the ε expansion, a
significant reduction in computation.

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (44/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Perturbation Example 1

Example: Consider the following linear ODE:

dy
+ α sin(ωt)y = β cos(ωt).
dt

From before we scale the problem with

β
z = Ay and τ = Bt, taking B=α and A = α.

The rescaled problem is

dz ω

ω


+ sin ατ z = cos ατ .
dτ

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (45/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Perturbation Example 2
ω 1
Example: Suppose that ω = 10 and α = 0.1, so α ≡ ε = 100  1.
The rescaled problem becomes
dz τ

τ


+ sin ε z = cos ε .
dτ

With T = τε , the linear ODE is

dz
+ ε sin (T ) z = ε cos (T ) ,
dT
which has the solution:
Z T
z(T ) = z(0)e ε(cos(T )−1)
+ εeε(cos(T )−1)
cos(s)e−ε(cos(s)−1) ds.
0

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (46/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Perturbation Example 3

From Taylor’s Theorem, we have

eε(cos(T )−1) = 1 + ε(cos(T ) − 1) + O ε2 .

It follows that the solution can be approximated by

z(0) 1 + ε(cos(T ) − 1) + O ε2



z(T ) =
Z T
+ε 1 + ε(cos(T ) − 1) + O ε2 cos(s) 1 − ε(cos(s) − 1) + O ε2





ds,
0

 Z T 
+ O ε2 ,


z(T ) = z(0) + ε z(0)(cos(T ) − 1) + cos(s)ds
0
 
z(0) + ε z(0)(cos(T ) − 1) + sin(T ) + O ε2 .


z(T ) =

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (47/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Perturbation Example 4
The figure below shows both the approximate solution and actual
solution for z(0) = 2.
2.05

1.95

1.9
Approximation
Solution
1.85
0 10 20 30 40 50
Lecture Notes – Linear Differential Equations a
Joseph M. Mahaffy, [email protected] — (48/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Perturbation Example 5
The figure was produced by the MatLab program below.
1 z0 = 2 ; ep = 0 . 0 1 ;
2 tt = linspace (0 ,50 ,500) ;
3 z z = z0 + ep ∗ ( z0 ∗ ( c o s ( t t ) −1) + s i n ( t t ) ) ;
4 [ t1 , z1 ] = ode23 ( @perturb , t t , z0 ) ;
5 p l o t ( t t , zz , ’ b− ’ , ’ LineWidth ’ , 1 . 5 ) ;
6 h o l d on
7 p l o t ( t1 , z1 , ’ r− ’ , ’ LineWidth ’ , 1 . 5 ) ;
8 grid ;
9 l e g e n d ( ’ Approximation ’ , ’ S o l u t i o n ’ , ’ L o c a t i o n ’ , ’ s o u t h e a s t ’ ) ;
10 xlim ( [ 0 , 5 0 ] ) ;
11 ylim ( [ 1 . 8 5 , 2 . 0 5 ] ) ;

1 f u n c t i o n zp = p e r t u r b (T, z )
2 %P e r t u r b a t i o n ODE
3 ep = 0 . 0 1 ;
4 zp = −ep ∗ s i n (T) ∗ z + ep ∗ c o s (T) ;
5 end

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (49/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Mass-Spring 1
Mass-Spring Example: Consider a mass-spring system with spring
constant, k, and damping proportional to the velocity of the mass, cẋ.
Newton’s Law gives:

mẍ = −kx − cẋ.

This gives the ODE:

c k
ẍ + m ẋ + mx = 0,

where c ≥ 0 and m, k > 0.

We scale the time with τ = βt, so
d dτ d d
dt = dt dτ = β dτ , and

d2 d d d d d d d2
dτ d2





dt2 = dt dt = dt β dτ = β dt dτ = β dτ 2 dt = β 2 dτ 2.

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (50/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Mass-Spring 2
Mass-Spring Example: Let x(t) = x̂(βt) = x̂(τ ), then this scaling
gives:
2
β 2 ddτx̂2 + cβ dx̂ k
m dτ + m x̂ = 0.
k c
Let β 2 = m and define γ = √mk , then ignoring the hats, the scaled
mass-spring system with damping is:
ẍ + γ ẋ + x = 0.

This 2nd order ODE is transformed into a system of 1st order

ODEs by letting x1 (τ ) = x(τ ) and x2 (τ ) = ẋ(τ ), so
ẋ1 = x2 ,
ẋ2 = −x1 − γx2 ,
which in matrix form is:
    
ẋ1 0 1 x1
= or ẋ = Ax.
ẋ2 −1 −γ x2
Lecture Notes – Linear Differential Equations a
Joseph M. Mahaffy, [email protected] — (51/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Mass-Spring System 1

Mass-Spring System: The matrix given by:

 
0 1
A= ,
−1 −γ

satisfies the characteristic equation:

−λ 1
det |A − λI| = = λ2 + γλ + 1 = 0.
−1 −λ − γ

The eigenvalues satisfy:

h p i
1
λ= 2 −γ ± γ 2 − 4 .

There are 3 cases, which we examine.

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (52/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Mass-Spring System 2
√
Case (i): Consider γ 2 − 4 > 0 (or c > 2 mk), which leads to two
real negative eigenvalues,
λ1 < λ2 < 0.
The associated eigenvectors solve:
    
−λi 1 v1 0
(A − λi I)vi = = ,
−1 −λi − γ v2 0
which gives the eigenvectors:
   
1 1
v1 = and v2 = .
λ1 λ2

The solution of the mass-spring problem is:

     
x1 (t) 1 λ1 t 1
= c1 e + c2 eλ2 t .
x2 (t) λ1 λ2
Lecture Notes – Linear Differential Equations a
Joseph M. Mahaffy, [email protected] — (53/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Mass-Spring System 3
Case (i): Consider the specific case when γ = 3, so
√ √
λ1 = 0.5(−3 − 5) ≈ −2.618 and λ2 = 0.5(−3 + 5) ≈ −0.382.
This is the over-damped mass-spring problem. Below is the solution x(0) = 5
and ẋ(0) = 5.

0
0 5 10 15

Phase Portrait Time Series

Lecture Notes – Linear Differential Equations a
Joseph M. Mahaffy, [email protected] — (54/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Mass-Spring System 4
√
Case (ii): Consider γ 2 − 4 < 0 (or c < 2 mk), which leads to two complex
eigenvalues with negative real part (unless γ = 0),
√ √
−γ± γ 2 −4 4−γ 2
λ1,2 = 2
= − γ2 ± iω, ω= 2
, (0 < γ < 2).

The associated eigenvectors solve:

    
−λ1 1 v1 0
(A − λ1 I)v1 = = ,
−1 −λ1 − γ v2 0

which gives the eigenvector:

   
1 1
v1 = = .
λ1 − γ2 + iω

The solution of the mass-spring problem is:

      
x1 (t) cos(ωt) sin(ωt)
= e−γt/2 c1 + c2 .
x2 (t) − γ2 cos(ωt) − ω sin(ωt) ω cos(ωt) − γ2 sin(ωt)

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (55/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Mass-Spring System 5
Case (ii): Consider the specific case when γ = 1, so
√
λ1,2 = 0.5(−1 ± i 3) ≈ −0.5 ± 0.866i.
This is the under-damped mass-spring problem. Below is the solution x(0) = 5
and ẋ(0) = 5.

-2
0 5 10 15

Phase Portrait Time Series

Lecture Notes – Linear Differential Equations a
Joseph M. Mahaffy, [email protected] — (56/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Mass-Spring System 6
When γ = 0, this system is undamped and the resulting solution produces a
center as shown in the Introduction notes.
√
Case (iii): Consider γ 2 − 4 = 0 (or c = 2 mk), which leads to the repeated
eigenvalue, λ = −1.
The associated eigenvector solves:
      
1 1 v1 0 1
(A + I)v1 = = , so v1 = .
−1 −1 v2 0 −1

This eigenvalue has an algebraic multiplicity = 2 and a geometric multiplicity

= 1.
The second solution to the ODE comes from the higher null space. (Solve
(A + I)w = v.)
The solution of this mass-spring problem is:
       
x1 (t) 1 1 1
= c1 e−t + c2 t+ e−t .
x2 (t) −1 −1 0

Later we’ll investigate details of this case more.

Lecture Notes – Linear Differential Equations a
Joseph M. Mahaffy, [email protected] — (57/58)
 Radioactive Decay Buckingham Pi Theorem
First Order ODE Theory Scaling Parameters in ODE
Scaling and Dimensionless Variables Mass-Spring

Mass-Spring System 5
Case (iii): This is the critically-damped mass-spring problem. Below is the
solution x(0) = 5 and ẋ(0) = 5.

7
6
5
4
3
2
1
0
-1
0 5 10 15

Phase Portrait Time Series

Lecture Notes – Linear Differential Equations a

Joseph M. Mahaffy, [email protected] — (58/58)