0% found this document useful (0 votes)

3K views473 pages

Christopher W. Kulp, Vasilis Pagonis - Classical Mechanics - A Computational Approach With Examples Using Mathematica and Python

Uploaded by

jaydeep4567

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

3K views473 pages

Christopher W. Kulp, Vasilis Pagonis - Classical Mechanics - A Computational Approach With Examples Using Mathematica and Python

Uploaded by

jaydeep4567

First edition published 2021

by CRC Press
6000 Broken Sound Parkway NW, Suite 300, Boca Raton, FL 33487-2742

and by CRC Press

2 Park Square, Milton Park, Abingdon, Oxon, OX14 4RN

c 2021 Taylor & Francis Group, LLC

CRC Press is an imprint of Taylor & Francis Group, LLC

Reasonable efforts have been made to publish reliable data and information, but the author and publisher
cannot assume responsibility for the validity of all materials or the consequences of their use. The authors
and publishers have attempted to trace the copyright holders of all material reproduced in this publication
and apologize to copyright holders if permission to publish in this form has not been obtained. If any
copyright material has not been acknowledged please write and let us know so we may rectify in any future
reprint.

Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, trans-
mitted, or utilized in any form by any electronic, mechanical, or other means, now known or hereafter
invented, including photocopying, microfilming, and recording, or in any information storage or retrieval
system, without written permission from the publishers.

For permission to photocopy or use material electronically from this work, access www.copyright.com or
contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923, 978-750-
8400. For works that are not available on CCC please contact [email protected]

Trademark notice: Product or corporate names may be trademarks or registered trademarks and are used
only for identification and explanation without intent to infringe.

Library of Congress Cataloging-in-Publication Data

Names: Kulp, Christopher W., author. | Pagonis, Vasilis, author.
Title: Classical mechanics : a computational approach, with examples using
mathematica and python / Christopher W. Kulp, Vasilis Pagonis.
Description: Boca Raton : CRC Press, 2020. | Includes bibliographical
references and index.
Identifiers: LCCN 2020021514 | ISBN 9781138495289 (paperback) | ISBN
9781138495173 (hardback) | ISBN 9781351024389 (ebook)
Subjects: LCSH: Mechanics. | Mechanics--Data processing. | Python (Computer
program language) | Mathematica (Computer file)
Classification: LCC QC127 .K85 2020 | DDC 531.0285/53--dc23
LC record available at https://lccn.loc.gov/2020021514

ISBN: 978-1-138-49517-3 (hbk)

ISBN: 978-1-138-49528-9 (pbk)
ISBN: 978-1-351-02438-9 (ebk)

Typeset in LMRoman
by Nova Techset Private Limited, Bengaluru & Chennai, India

Visit the [eResources]: www.routledge.com/9781138495289

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page v — #5

i i

Chris dedicates this book to his wife Gail, mother Linda, and his late father,
Chester. Without their support, this book would not have been possible.

Vasilis dedicates this book to his wife, Mary Jo Boylan, and to his students
at McDaniel College.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page vii — #7

i i

Contents

Preface xiii

Chapter 1 The Foundations of Motion and Computation 1

1.1 THE WORLD OF PHYSICS 1
1.2 THE BASICS OF CLASSICAL MECHANICS 3
1.2.1 The Basic Descriptors of Motion 3
1.2.1.1 Position and Displacement 3
1.2.1.2 Velocity 4
1.2.1.3 Acceleration 5
1.2.2 Mass and Force 5
1.2.2.1 Mass 6
1.2.2.2 Force 6
1.3 NEWTON’S LAWS OF MOTION 7
1.3.1 Newton’s First Law 7
1.3.2 Newton’s second law 8
1.3.3 Newton’s third law 10
1.4 REFERENCE FRAMES 11
1.5 COMPUTATION IN PHYSICS 13
1.5.1 The Use of Computation in Physics 14
1.5.2 Different Computational Tools 20
1.5.3 Some Warnings 22
1.6 CLASSICAL MECHANICS IN THE MODERN WORLD 22
1.7 CHAPTER SUMMARY 23
1.8 END-OF-CHAPTER PROBLEMS 24

Chapter 2 Single-Particle Motion in One Dimension 29

2.1 EQUATIONS OF MOTION 29
2.2 ORDINARY DIFFERENTIAL EQUATIONS 30
2.3 CONSTANT FORCES 32
2.4 TIME-DEPENDENT FORCES 36
2.5 AIR RESISTANCE AND VELOCITY-DEPENDENT FORCES 39
2.6 POSITION-DEPENDENT FORCES 45
2.7 NUMERICAL SOLUTIONS OF DIFFERENTIAL EQUATIONS 48

vii
i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page viii — #8

i i

viii Contents

2.8 CHAPTER SUMMARY 54

2.9 END-OF-CHAPTER PROBLEMS 55

Chapter 3 Motion in Two and Three Dimensions 61

3.1 POSITION, VELOCITY, AND ACCELERATION IN CARTESIAN
COORDINATE SYSTEMS 61
3.2 VECTOR PRODUCTS 69
3.2.1 The Dot Product 69
3.2.2 The Cross Product 71
3.3 POSITION, VELOCITY, AND ACCELERATION IN NON-CARTESIAN
COORDINATE SYSTEMS 75
3.3.1 Polar Coordinates 75
3.3.2 Position, Velocity, and Acceleration in Cylindrical Coordinates 82
3.3.3 Position, Velocity, and Acceleration in Spherical Coordinates 84
3.4 THE GRADIENT, DIVERGENCE, AND CURL 86
3.4.1 The Gradient 86
3.4.2 The Divergence 92
3.4.3 The Curl 94
3.4.4 Second Derivatives with the Del Operator 96
3.5 CHAPTER SUMMARY 97
3.6 END-OF-CHAPTER PROBLEMS 99

Chapter 4 Momentum, Angular Momentum, and Multiparticle Systems 107

4.1 CONSERVATION OF MOMENTUM AND NEWTON’S THIRD LAW 107
4.2 ROCKETS 111
4.3 CENTER OF MASS 113
4.4 NUMERICAL INTEGRATION AND THE CENTER OF MASS 118
4.4.1 Trapezoidal Rule 118
4.4.2 Simpson’s Rule 119
4.5 MOMENTUM OF A SYSTEM OF MULTIPLE PARTICLES 123
4.6 ANGULAR MOMENTUM OF A SINGLE PARTICLE 125
4.7 ANGULAR MOMENTUM OF MULTIPLE PARTICLES 126
4.8 CHAPTER SUMMARY 129
4.9 END-OF-CHAPTER PROBLEMS 131

Chapter 5 Energy 135

5.1 WORK AND ENERGY IN ONE-DIMENSIONAL SYSTEMS 135
5.2 POTENTIAL ENERGY AND EQUILIBRIUM POINTS IN
ONE-DIMENSIONAL SYSTEMS 139
5.3 WORK AND LINE INTEGRALS 146
5.4 THE WORK-KINETIC ENERGY THEOREM, REVISITED 149
5.5 CONSERVATIVE FORCES AND POTENTIAL ENERGY 150

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page ix — #9

i i

Contents ix

5.6 ENERGY AND MULTIPARTICLE SYSTEMS 154

5.7 CHAPTER SUMMARY 156
5.8 END-OF-CHAPTER PROBLEMS 157

Chapter 6 Harmonic Oscillations 163

6.1 DIFFERENTIAL EQUATIONS 163
6.2 THE SIMPLE HARMONIC OSCILLATOR 164
6.2.1 The Equation of Motion of the Simple Harmonic Oscillator 165
6.2.2 Potential and Kinetic Energy in Simple Harmonic Motion 167
6.2.3 The Simple Plane Pendulum as an Example of a Harmonic
Oscillator 168
6.3 NUMERICAL SOLUTIONS USING THE EULER METHOD FOR
HARMONIC OSCILLATIONS 170
6.4 DAMPED HARMONIC OSCILLATOR 172
6.4.1 Overdamped Oscillations 173
6.4.2 Underdamped Oscillation 174
6.4.3 Critically Damped Oscillations 176
6.5 ENERGY IN DAMPED HARMONIC MOTION 177
6.6 FORCED HARMONIC OSCILLATOR 179
6.7 ENERGY RESONANCE AND THE QUALITY FACTOR FOR DRIVEN
OSCILLATIONS 185
6.8 ELECTRICAL CIRCUITS 188
6.9 PRINCIPLE OF SUPERPOSITION AND FOURIER SERIES 191
6.9.1 The Principle of Superposition 191
6.9.2 Fourier Series 192
6.9.3 Example of Superposition Principle and Fourier Series 195
6.10 PHASE SPACE 198
6.11 CHAPTER SUMMARY 200
6.12 END-OF-CHAPTER PROBLEMS 202

Chapter 7 The Calculus of Variations 209

7.1 THE MOTIVATION FOR LEARNING THE CALCULUS OF VARIATIONS 209
7.2 THE SHORTEST DISTANCE BETWEEN TWO POINTS—SETTING UP
THE CALCULUS OF VARIATIONS 210
7.3 THE FIRST FORM OF THE EULER EQUATION 212
7.4 THE SECOND FORM OF THE EULER EQUATION 215
7.5 SOME EXAMPLES OF PROBLEMS SOLVED USING THE CALCULUS
OF VARIATIONS 216
7.5.1 The Brachistochrone Problem 216
7.5.2 Geodesics 219
7.5.3 Minimum Surface of Revolution 220
7.6 MULTIPLE DEPENDENT VARIABLES 223

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page x — #10

i i

x Contents

7.7 CHAPTER SUMMARY 225

i i

CHAPTER 1

The Foundations of Motion

and Computation

So, you have decided to—or are required to—learn the subject of classical mechanics. But,
what is classical mechanics? Is it fixing old cars? No, but a knowledge of classical mechanics
will help you understand how your car works! To understand the term, classical mechanics
let us first understand the term classical physics. Classical physics are the fields of physics
that don’t involve either quantum theory or the theory of relativity. Mechanics is the branch
of physics that deals with the actions of forces on an object that involve motion. So, classical
mechanics involves the study of forces on objects that are well-described without using
quantum theory and whose motion is nonrelativistic (i.e., cases where relativity is not
needed to correctly model the motion). In other words, classical mechanics is the physics of
your day-to-day world! An understanding of classical mechanics will help you understand
how to build roller coasters, merry-go-rounds, and airplanes. Our knowledge of classical
mechanics also allows us to make predictions on the motion of objects like comets and
punted footballs. The world you interact with on a daily basis is generally the world of
classical mechanics.
At first it might seem that classical mechanics is a dusty old subject that you need
to learn in order to get to the “interesting stuff” like quantum mechanics. While the field
of classical mechanics is one of the older subjects in physics, it is certainly not dusty! In
fact, there is a lot of intriguing current research done in classical systems. For example,
in Chapter 13 we will explore the field of nonlinear systems, which occur in in many of
the natural and social sciences, not only in physics, and they display interesting types of
behaviors including chaos.

1.1 THE WORLD OF PHYSICS

It is helpful to break up the field of physics into different branches. It should be noted that
such divisions are largely arbitrary but useful human constructs that help us understand the
world around us. As we sometimes tell students: Mother Nature doesn’t care if a particular
system is a thermodynamics problem or a classical mechanics problem; humans make those
distinctions in order to better understand how to describe and model the system. In fact,
most real-world problems involve multiple fields of physics. So with that in mind, we can
loosely break up physics into the following branches:
• Classical Mechanics deals with how forces cause motion in classical systems.

1.2 THE BASICS OF CLASSICAL MECHANICS

When thinking about any discipline of science, first ask what are the basic foundations of
the field? Sometimes these foundations are assumptions. For classical mechanics we can
start with space and time:
• Space serves as a background in which physical processes occur. Unlike in the theory of
general relativity, space has no effect on the behavior of physical systems. Furthermore,
length measurements are the same for every observer, unlike in the theory of special
relativity.
• Time progresses at the same rate for every observer (unlike in the theories of special
and general relativity).

Classical mechanics addresses the problem of predicting an object’s motion given the
forces acting on the object. For now, we will consider the object to be a point particle. The
advantage of working with point particles is that they have no size and no internal dynam-
ics. In addition, point particles do not rotate nor do they deform, further simplifying their
dynamics. It turns out that treating objects as point particles can be a very good approx-
imation for describing translational motion. In later chapters, we will study the physics
of extended bodies and rotational dynamics, where we will no longer restrict ourselves to
working with point particles.

1.2.1 The Basic Descriptors of Motion

In order to describe the translational motion of a particle, we need three quantities that we
will call the basic descriptors of motion: position, velocity, and acceleration. For rotational
motion, we will need three additional quantities: angular position, angular velocity, and
angular acceleration, and we will return to those later in the book.

1.2.1.1 Position and Displacement

The position of a particle is the location of the particle with respect to an origin and is
measured in meters. The meter is defined as the distance light travels in 1/299,792,458
seconds. Note that the second was defined in 1967 at the 13th meeting of the International
Committee on Weights and Measures. At that meeting the following definition was adopted:
“The second is the duration of 9,192,631,770 periods of the radiation corresponding to the
transition between the two hyperfine levels of the ground state of the cesium-133 atom.”
The position of the particle is typically described by using a vector whose components
consist of the particle’s distance from the origin along three perpendicular axes. For example,
in Cartesian coordinates the location of a particle can be described using:

r = xî + y ĵ + z k̂ (1.2.1)
where î, ĵ, and k̂ are the Cartesian unit vectors along the x, y, and z axes, respectively. The
variables x, y, and z in (1.2.1) give the distance between the particle and the origin along
each axis and are called the components of the vector. Also note that in (1.2.1): r = r(t),
x = x(t), y = y(t), and z = z(t). In this book, vectors are denoted by bold font. In Chapter
3, we will discuss vector quantities in more detail.
The vector in (1.2.1) tells us that in order to get to the location of the particle, r, one
needs to move a distance x along the x-axis (as denoted by î), turn and move a distance y
along the y-axis (as denoted by ĵ), and then turn and move a distance z along the z-axis

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 4 — #20

i i

4 Classical Mechanics: A Computational Approach

(as denoted by k̂). The notation used in this book is that a hat (the ˆ symbol) represents
a unit vector, a vector of length one. Hence, we can think of r as the sum of three vectors:
xî, y ĵ, and z k̂, each representing a displacement from the origin along one of the axes.
The displacement of the particle, ∆r = r − r0 , is the change of the particle’s position
from the position, r0 , to the position, r. The displacement is found by subtracting the two
vectors r0 and r, component by component,

∆r = r − r0 = (x − x0 )î + (y − y0 )ĵ + (z − z0 )k̂ (1.2.2)

You may recall that a vector quantity contains information about both magnitude
(amount) and direction. In this case, the vector ∆r tells us how far (magnitude) and in
which direction the particle traveled.
You may also see the unit vectors in the form x̂, ŷ, and ẑ, sometimes used to clarify
that one is working in the Cartesian coordinate system. In addition, sometimes it will be
easier to use more generic notation for vectors where r1 , r2 , and r3 are used instead of the
components x, y, and z and the unit vectors ê1 , ê2 , and ê3 are used instead of î, ĵ, and k̂.
This generic notation allows for a more compact method of writing vectors:
3
X
r = r1 ê1 + r2 ê2 + r3 ê3 = ri êi (1.2.3)
i=1

There are other coordinate systems which we will explore later in this book. Each coor-
dinate system will have its own components and unit vectors, however, the basic idea of
position and displacement will be the same.

1.2.1.2 Velocity
The velocity v of a particle is the particle’s displacement (change of position) per unit
time and is measured in meters per second (m/s). The instantaneous velocity is found by
computing the time derivative of the position vector:

dr ∆r
v= = lim (1.2.4)
dt ∆t→0 ∆t
where ∆r = r(t + ∆t) − r(t). All of the vectors we will come across here are differentiable,
and the limits will exist. Furthermore, the derivative in (1.2.4) will behave like derivatives
you have encountered before. Therefore, we can use the standard derivative rules:

d dr1 dr2
(r1 + r2 ) = + (1.2.5)
dt dt dt
d dc dr
(cr) =r + c (1.2.6)
dt dt dt
where c is a scalar function of time. The rules in (1.2.5) and (1.2.6) allow us to distribute
the derivative to each vector component. In addition, if the unit vectors are constant, like
they are in Cartesian coordinates (but not in others!), then we can compute the velocity
by:

i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 8 — #24

i i

8 Classical Mechanics: A Computational Approach

We know that acceleration is the change of velocity and hence, the first law states that
a nonzero net force is needed to cause an acceleration.
Finally, the first law provides a definition for the term equilibrium. A particle is in
equilibrium if the acceleration of the particle is equal to zero. Hence, one condition for
equilibrium is that the net force acting on a particle must be zero. We will later see that
when we study rotational dynamics, we will also need the net torque to be equal to zero as
an additional condition for equilibrium.
Before moving on to Newton’s second law, it is useful to compare our statement of the
first law to the more colloquial “An object in motion remains in motion, and an object
at rest stays at rest unless acted upon by an unbalanced force.” While there is nothing
technically wrong with this statement, notice that it is longer than our statement of the law
on the previous page. The colloquial version breaks down motion and rest as two different
behaviors. However, our use of the word velocity takes both states (motion and rest) into
account because being at rest simply means that the particle’s velocity is equal to zero. As
mentioned previously, the term motion is not one of our descriptors and therefore it is not
clear what is remaining constant. Additional terms can be added to explain motion, but at
the cost of conciseness. Finally, the term unbalanced force is also not as clear as the term
net force. The term net force means vector sum of forces, a clearly defined term. While the
colloquial statement of the First Law may use words familiar to most people, a physicist
would prefer a statement similar to the one in the box above, due to its use of precise
language.

1.3.2 Newton’s second law

Newton’s second law of motion is used to find a particle’s equations of motion, which are
equations that give the particle’s position, velocity, or acceleration at any point in time.
We will be using Newton’s second law as the central tool for mathematically describing the
motion of a particle throughout this book.
Newton’s Second Law

A particle’s time rate of change of linear momentum, p, is equal to the net force, F,
applied to the particle:

F = ṗ (1.3.1)

where the linear momentum of a particle with a mass m and velocity v is defined to be:
Linear Momentum of a Single Particle

p = mv (1.3.2)

Equation (1.3.1) might not be the way you are used to seeing Newton’s second law. If
we consider a system with constant mass m such as a single particle experiencing a nonzero
net force, then we have:
dp d
F= = (mv) (1.3.3)
dt dt
F = ma (1.3.4)
and we recover the more familiar form of the second law. However, when studying the
motion of an object whose mass m is changing with time, such as in the case of a rocket,

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 9 — #25

i i

The Foundations of Motion and Computation 9

the form of the second law presented in (1.3.1) will be the one needed in order to derive the
object’s equations of motion.
Although we will use the form of Newton’s second law F = ma most often for calcula-
tional purposes in this book, we find it helpful to think of the second law in the form:

F
a= (1.3.5)
m

We find this form to be more useful when understanding the concepts behind Newton’s
second law. Equation (1.3.5) tells us:
• Acceleration is in the same direction as the net force. There is no minus sign in front
of F to denote an acceleration in the opposite direction, nor is there any mathematical
transformation done to the vector F to rotate it.
• The magnitude of the acceleration is directly proportional to the magnitude of the net
force. In other words, net forces with a large magnitude produce larger magnitude
accelerations than net forces with small magnitudes. Mathematically, this can be seen
because F appears in the numerator of the fraction in the right-hand side of (1.3.5)
and by thinking of the equation as: |a| = |F|/m.
• Mass “resists” acceleration. The mass m appears in the denominator of the fraction
in the right-hand side of (1.3.5). Large denominators result in smaller overall fractions
when compared to a small denominator with the same numerator (i.e., 1/4 < 1/2). In
other words, with the same given force, objects with a larger mass experience a smaller
acceleration, and smaller mass objects experience a larger acceleration.
All three of the above bullet points are contained in the one simple equation (1.3.5)! This is
one of the reasons why physicists prefer math as the language for describing the universe.
A lot can be said in one simple equation. As a physicist, you should learn how to “read
equations” like we did above.
Notice the consistency between the first and second laws. If F = 0, i.e., a zero net force
is acting on the particle, then a = 0, and the particle’s velocity is not changing, just as
stated in the first law. Likewise, we could have restated Newton’s first law as: “A particle’s
momentum remains constant if no external net force acts on the particle.”
As we mentioned, the second law will be used to produce differential equations which
describe the motion of a particle. As a simple example, we will consider a particle moving
in one dimension (along a line) under the influence of a constant force, where a = ẍî and
F = F î. Recall that the dots above the x denote a second derivative with respect to time.
We can use (1.3.5) to write:

ẍ = F/m = a (1.3.6)
Equation (1.3.6) says that the solution x(t) is a function such that its second derivative is
equal to the constant, a. Recall from simple algebra that an equation like x + 3 = 10 says x is
the number that when added to 3 gives 10. In that case, the solution, x, is a number (x = 7).
The second-order differential equation (1.3.6) is similar to our algebra problem (but much
harder!), except instead of finding a number, you are asked to find a function. Differential
equations are the mathematical language used by physicists to describe the motion of a
particle. Be prepared to solve them both in closed form (analytically) and numerically. Just
to whet your appetite, if a is constant, the solution to (1.3.6) is:

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 10 — #26

i i

10 Classical Mechanics: A Computational Approach

1
x(t) = x0 + v0 t + at2 (1.3.7)
2
where x0 = x(0) and v0 = v(0), are the initial position and velocity, respectively. You can
check this for yourself by taking two derivatives of (1.3.7) and showing that it satisfies
(1.3.6) in the case of a constant a. In Chapter 2, we will go through the steps to derive
(1.3.7) and other solutions to differential equations. If you aren’t certain how we obtained
(1.3.7) from (1.3.6), don’t worry, you should understand it by the end of Chapter 2.
Finally, we can return to the weak equivalence principle. Suppose a particle is near the
surface of the Earth and experiences only the force of gravity (weight W ):

F =W
minertial a =mgravitational g (1.3.8)

where we have dropped the vector notation, assumed “downward” is the positive direction,
and used g for the acceleration due to gravity (9.8 m/s2 ). The mass minertial in this equation
is the m from Newton’s second law, i.e., it is the mass that “resists” acceleration, and we
call this the inertial mass. The mass mgravitational is the mass that is affected by gravity
and is called the gravitational mass. The weak principle of equivalence says that minertial =
mgravitational , and therefore the masses cancel, and a = g for a freely falling body near the
surface of the Earth. If the weak equivalence principle were not true, then the mass of an
object would affect its acceleration due to gravity. So far, physicists have not been able to
detect any mass dependence (outside experimental error) on an object’s acceleration due to
gravity, even in the most sensitive of experiments.

1.3.3 Newton’s third law

Newton’s third law is a statement that discusses the nature of interactions between two
particles. In order to state Newton’s third law, we need to consider two objects interacting
by exerting a force on each other. Let us define F12 to be the force on object 1 exerted by
object 2, and F21 is the force on object 2 exerted by object 1. Newton’s third law is then
stated:
Newton’s third law

If object 2 exerts a force F12 on object 1, then object 1 exerts a force F21 on object 2
such that:

F21 = −F12 (1.3.9)

Notice that the minus sign and lack of a scalar multiple in (1.3.9) says that the force
exerted by object 2 on object 1 is equal in magnitude (lack of scalar multiple which, if
present, would change the magnitude), but opposite in direction (denoted by the the minus
sign) to the force object 1 applies on 2. Many of the interaction forces we will study in this
book will be central forces. Central forces are forces that act along the line that joins the
centers of the two interacting objects, such as gravity and electrostatic forces, and obey
Newton’s third law. Velocity-dependent forces are not central, and the third law may not
apply. An example is the force between two moving electric charges; the Lorentz force is

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 11 — #27

i i

The Foundations of Motion and Computation 11

velocity-dependent, and the magnetic force vectors between the two charged particles do
not lie along the same line, hence the resulting net forces do not obey Newton’s third law.
Let us take (1.3.9) a little further:

F21 = − F12 (1.3.10)

m1 a1 = m2 (−a2 ) (1.3.11)

Not surprisingly, we see that the accelerations of each object are in opposite directions. If
we continue manipulating (1.3.11), we find:
m1 a2
=− (1.3.12)
m2 a1
where we have taken the magnitude of the acceleration vectors. Notice that ratio of the
accelerations is inverse to the ratio of the masses. In other words, if m1 > m2 then a1 < a2 ,
in order for the right-hand side of (1.3.12) to be a fraction greater than one. As an example,
consider the classic scenario of a mosquito hitting the windshield of a moving automobile.
Let the mosquito be object 2 and the car be object 1. Clearly, the automobile has more mass
than the mosquito, m1 > m2 , and according to (1.3.12), the acceleration of the mosquito is
greater than that of the automobile, a2 > a1 . It is a bad day for the mosquito. Another way
to interpret (1.3.12) is that an object cannot accelerate without another object accelerating
in the opposite direction.
Next, we need to address one other important fundamental concept for classical mechan-
ics: reference frames. Once we have a working understanding of reference frames, we will
have finished laying out the foundations of classical mechanics, and we will be ready to do
some physics!

1.4 REFERENCE FRAMES

All of the basic descriptors of motion are measured with respect to a reference frame. A
reference frame is a choice of origin, spatial and temporal, as well as a set of axes with respect
to which all measurements are made. For example, if you are holding your cell phone in a car
moving at a constant speed of 60 MPH while your friend is driving, then you observe your
phone to be at rest. However, a bystander on the side of the road will observe your phone to
be moving at 60 MPH. Who is correct? They both are, because the measurement of velocity
depends on the reference frame. In this case, your frame is moving with the car while the
bystander’s is at rest on the side of the road (ignoring Earth’s motion). By choosing the
right reference frame, you may be able to simplify a particular physics problem, in this case
a moving phone versus one at rest. For another example, in introductory physics you no
doubt worked on problems involving inclined planes. It is well known that choosing one axis
to be parallel to the incline greatly simplifies the free-body diagram for the problem. Other
changes of reference frame may simply involve changing the time for which t = 0.
Another important point to mention about the above example is that the two reference
frames are moving relative to one another with a constant velocity. This will make the
accelerations measured in each frame the same. Consider the following example, shown in
Figure 1.1, with two reference frames S and S 0 , where S 0 is moving with respect to S
with a velocity u = d,˙ and d(t) is the distance between the two reference frames at a time,
t. Note that in this case, primed variables denote the reference frame they belong to not
differentiation. In each frame, an observer is measuring the location of the black dot. The

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 12 — #28

i i

12 Classical Mechanics: A Computational Approach

y y'
xS'

x x'
S S'
Figure 1.1: The frame S 0 is moving at a constant velocity with respect to the fixed frame,
S. The particle (black dot) is measured to have a position xS in the reference frame S and
xS 0 in the reference frame S 0 . The distance between the references frames is d.

observer in S measures the black dot to be located at xS . Likewise, the observer in S 0

measures the black dot to be at xS 0 .
Now suppose the observers in each frame want to communicate to each other the motion
of the black dot. To keep things simple, we will consider motion only along the x-direction.
The coordinate transformation between the two reference frames:

xS = xS 0 + d (1.4.1)
allows the two frames to consistently describe the motion in each frame. First, we will
consider the case where S 0 is moving at a constant speed relative to S, in other words, u = d˙
is constant. Differentiating (1.4.1) with respect to time will give us the transformations of
the velocity between coordinate systems:

ẋS =ẋS 0 + u (1.4.2)

vS =vS 0 + u (1.4.3)

Hence, if the observers in each frame wanted to determine if they are consistently measuring
the velocity of the particle, then they can insert their measured velocities into (1.4.3) in
order to check the other’s result. Next, we compute the acceleration transformation:

ẍS =ẍS 0 + u̇ (1.4.4)

aS =aS 0 (1.4.5)

where we used the fact that u (the relative velocity between frames) is constant. Note that
(1.4.5) says that the measured accelerations in the two frames are the same, thus (1.4.5)

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 13 — #29

i i

The Foundations of Motion and Computation 13

also tells us that the forces measured on the particle in each reference frame are the same
and therefore Newton’s laws hold in their typical form in each frame. In other words, if we
measured the acceleration of the particle in each reference frame, we would find that the
measured acceleration can be accounted for by considering all of the forces acting on the
particle. The reference frame, S 0 is called an inertial reference frame because it is moving
with a constant velocity. Newton’s second law holds in inertial reference frames.
Let us now contrast that with the case of S 0 accelerating with respect to S. In this case,
(1.4.3) doesn’t change, and (1.4.5) becomes:

aS = aS 0 + u̇ (1.4.6)
and u̇ 6= 0. Now the two measured accelerations are not the same, and therefore the measured
forces in each frame are different. In particular, the acceleration of a particle in a noninertial
reference frame cannot be accounted for by summing all of the forces acting on the particle.
For example, on certain carnival rides that involve rotation, you will experience an apparent
outward force which is not caused by any forces acting on you from the ride itself. This is
an example of a noninertial reference frame where Newton’s laws no longer hold in their
standard form because there is an acceleration measured in S that is not apparent in S 0 . As
we will see in a later chapter, we can modify Newton’s laws in order to address the case of
noninertial frames. Such modifications involve treating the acceleration u̇ as coming from
an inertial force, a force that is not created by physical interactions, but rather is due to an
accelerating frame. While the surface of the Earth typically approximates an inertial frame,
the Earth does rotate and revolve around the Sun; hence, the velocity of a reference frame
“glued” to the Earth’s surface is not constant. There are cases where the noninertial nature
of the “glued” frame needs to be accounted for. Such cases include long-distance motion
such as missile trajectories and the motion of wind and water currents.

1.5 COMPUTATION IN PHYSICS

An important part of this book is the inclusion of computation in solving problems. Compu-
tation is defined by Merriam-Webster [MW()] as “the act or action of computing.” The term
“compute” is further defined by Merriam-Webster [MW()] as “to determine by especially
mathematical means” and “to determine or calculate by means of a computer.” It is the
latter definition that is addressed in this book. So when we say “computation” in this book,
we mean using a computer to solve physics and mathematics problems, either analytically
or numerically. Physicists use computers in a variety of ways to solve problems. In addition
to providing instruction in classical mechanics, this book will also provide you instruction
on how to use computers to solve physics and mathematics problems.
Extensive experience in coding is not necessary in order to begin reading this book.
In fact, we will mainly rely on commands that are already a part of software packages.
That said, you will pick up the coding that you need along the way. In this section, we
will provide a motivation for how computing is used in physics and why it is important.
Also in this section, we will discuss different types of programming languages, and some
direction on how to go about learning to code. Coding has become a fundamental skill for
a physicist. We believe that all students should have coding experience before graduating
with an undergraduate degree in physics. Hopefully, by the time you have finished this book,
you’ll have a thorough understanding of both classical mechanics and how to use computers
to solve physics problems. Think of it as a two-for-one deal!

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 14 — #30

i i

14 Classical Mechanics: A Computational Approach

1.5.1 The Use of Computation in Physics

To understand how computation is used in physics, we will demonstrate the solution to some
problems analytically (also known as “by hand”) and using computation. At this time, it
is not important to understand all of the physics in the problems, nor is it necessary to
understand the computational methods used to solve the problems; we will cover those
topics in detail later in this text. The important thing to take away from this section is an
appreciation of how computing is used and why computing is important.
Let us start with a simple physics problem that can be “solved by hand.”

Example 1.1: Velocity as a function of time

Consider a particle that is moving along a line with a constant acceleration, a. If at
time t = 0 the particle’s velocity is v0 , find the formula for the particle’s velocity as a
function of time.
Solution:
We know from (1.2.11) that acceleration is the first derivative of velocity with respect
to time:
dv
=a (1.5.1)
dt
Next, we separate variables and integrate both sides of the equation. Think of separa-
tion of variables as multiplying by dt in order to get all of the v 0 s on one side and the t0 s
on the other. After separation of variables (1.5.1) becomes:
Z v(t) Z t
0
dv = adt0 (1.5.2)
v0 0

where we matched the lower and upper limits on each side. Notice that the lower-limit on
the right-hand side is t = 0, and the lower limit on the left-hand side is the value of v at
t = 0, similar for the upper limits. In addition, we included a prime on our variables of
integration to distinguish them from the limits of integration. Finally we integrate, noting
that a is constant:

v(t) = v0 + at (1.5.3)
The result is the velocity as a function of time, as requested by the problem.

Example 1.1 is an example of a problem that is “done by hand,” meaning that we were
able to perform the necessary mathematical manipulations to solve the differential equation
(1.5.1) without the aid of a computer. Equation (1.5.3) is called a “closed form” solution to
the differential equation (1.5.1) because it gives a specific solution consisting of functions
and mathematical operations. What is considered as “closed-form” is somewhat arbitrary
because, for example, a solution in the form of an infinite sum may not be considered in
“closed-form.”
As you will see in this book, solving physics problems very often involves finding the
solution to differential equations. Some of those equations have no closed-form solution,
while others are very difficult to find. In those cases, computation can be extremely helpful.
For the purposes of this book, there are two forms of computation:
• Symbolic Computation: involves using a computer to help find closed-form solutions to
differential equations, integrals, eigenvalues, and much more. You enter the equation

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 15 — #31

i i

The Foundations of Motion and Computation 15

or integral you want solved, and the computer program returns a closed-form solution.
You may have seen tables of derivatives and integrals; symbolic computation is a much
more sophisticated version of those.
• Numerical Computation: (also sometimes referred to as “numerics”) provides a list of
numerical values representing a solution. For example, a numerical solution to (1.5.1)
can be thought of as a table of values with two columns, t and v(t), where each row
contains a specific time, t, and the value of v at time, t. Numerical solutions are often
best represented as a graph of ordered pairs (and sometimes triplets, depending on
the dimension of the problem being solved).
As an example of each type of computation, we will use the programming languages, Math-
ematica and Python, to solve Example 1.1 both symbolically and numerically. Again,
the point of the next two examples is to provide an illustration of each type of com-
putation. We follow each example with an explanation of what is going on in the
code. Please note that all of the code that appears in this book can be downloaded at
www.routledge.com/9781138495289

Example 1.2: An example of symbolic computation

Using symbolic computation, solve the differential equation from Example 1.
Solution:
The code for using Mathematica to solve (1.5.1) is shown below.
solution = DSolve[{v 0 [t] == a, v[0] == v0}, v, t];

v[t]/.solution[[1]]//TraditionalForm

OUTPUT: at + v0
The code for using the interpreted-language Python to solve 1.5.1 is shown below.

from sympy import ∗

init printing ()

v = Function ( ’ v ’ )
t = Symbol ( ’ t ’ , r e a l = True , p o s i t i v e = True )
a = Symbol ( ’ a ’ , r e a l = True )

general soln = dsolve ( Derivative (v( t ) , t ) − a , v( t ))

print ( g e n e r a l s o l n . rhs )

OUTPUT
C1 + a ∗ t

The first line in the Mathematica code for Example 1.2 uses Mathematica’s DSolve com-
mand to produce a solution to the differential equation. Notice that the differential equation
and its initial condition appear as the first argument of the command (in between the curly
brackets). Notice also the “==” as opposed to “=” in the equation. Often in programming
languages, the single equal sign is used for variable assignment, whereas the double-equal
sign is used for equivalence. The first line of code stores the solution generated by DSolve in

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 16 — #32

i i

16 Classical Mechanics: A Computational Approach

the variable solution (notice the single equal sign). The second argument v tells Mathemat-
ica that v is the function for which we are solving, and the third argument identifies t as the
independent variable. The second line in the code outputs the solution. The command Tra-
ditionalForm displays the output in a more readable format; without using TraditionalForm,
Mathematica would display the output differently. Notice that Mathematica’s solution is
the same as the one we derived in Example 1.1.
The Python code for Example 1.2 is much more involved than the Mathematica code.
The first line, “from sympy import *” causes Python to import the SymPy library. Python,
by itself cannot perform symbolic manipulations. We needed to import the SymPy library
in order to expand Python’s capabilities. Libraries are written by experienced programmers,
and they include functions which can be used by any code that imports the library. In this
case, the SymPy library includes functions like dsolve which solve differential equations
in closed form. Note that this is not the same as the Mathematica command, DSolve.
Commands from one language typically cannot be used in another. In order to make a
distinction between the two languages, we will preserve the capitalization used in each
language. For example, if we write DSolve, then we are discussing Mathematica’s command
for solving differential equation because in Mathematica the D and S are capitalized in
the command. The letters D and S are not capitalized in Python and, therefore, when we
discuss solving differential equations in Python, we will write the command dsolve, as it
appears in the language of Python.
A word of warning: libraries, while useful, may not be free of bugs. It is best practice to
test any library functions you are using in your code to make sure they behave as expected.
Python libraries like SciPy, NumPy, and SymPy are well-tested and well-documented. Doc-
umentation for these and other Python libraries can be found online.
Mathematica automatically recognizes a, vo, and t as symbols to be manipulated. In
addition, the Mathematica syntax, v[t], allows Mathematica to automatically recognize v
as a function. However, Python by itself would think of a, vo, t as variables for assignment
and would not recognize v(t) as a function. Therefore, all of the symbol identifications
need to be specified in Python. The Symbol and Function commands from the SymPy
library do exactly that. The general solution of the differential equation is given by the
SymPy command dsolve, whose arguments are the differential equation and the function
for which the differential equation is being solved. Notice that in Python, the differential
equation is written such that the right-hand side is equal to zero, and only the left-hand
side of the differential equation is entered. At the time of this writing, Python’s dsolve
command cannot solve the differential equation with the initial conditions at the same time
(except for power series solutions), unlike Mathematica’s DSolve. With more code, one could
perform the substitutions, however, they are not included here in order to prevent (further)
confusion. The output of the Python program is included below the word, OUTPUT. The
word, OUTPUT, and the line below it are not part of the Python code; we included it only
to show the output of the program. In this book, we will often show the output of a Python
program in this manner. Note the the C1 in the output represents a constant of integration.
In Example 1.2, we see that Mathematica requires a much shorter code to solve this
problem than Python because Mathematica is designed, in part, for symbolic manipulations.
Computer algebra systems (CAS) like Mathematica and Maple are very convenient and
easy to use when performing symbolic computations. The disadvantage is that they are
both proprietary and more expensive than other options, such as the open-source free CAS
SageMath (https://www.sagemath.org/).
Next, we will look at an example of a numeric computation and compare the two lan-
guages in that context.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 17 — #33

i i

The Foundations of Motion and Computation 17

Example 1.3: An example of numeric computation

Using a = 9.8 m/s2 and an initial velocity of v(0) = v0 = 1 m/s, find and graph a
numerical solution to the differential equation from Example 1.1.
Solution:
The code for using Mathematica to solve (1.5.1) is shown below.
a = 9.8;

solution = NDSolve[{v 0 [t] == a, v[0]==1.0}, v, {t, 0, 3}];

Plot[v[t]/.solution, {t, 0, 3}, Frame → True, Axes → False,

FrameLabel → {“time (sec)”, “velocity (m/s)”}, BaseStyle → {FontSize → 18},
ImageSize → Large]

25
velocity (m/s)

0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
time (sec)

The Python code used to solve (1.5.1) is shown below. For brevity, we did not include
the resulting graph which is the same as the one above.

import numpy a s np
from s c i p y . i n t e g r a t e import o d e i n t
import m a t p l o t l i b . p y p l o t a s p l t

vo = 1

def velderiv (v , t ) :
a = 9.8
dvdt = a
r e t u r n dvdt

t i m e s = np . l i n s p a c e ( 0 , 3 , 3 0 )
v e l o c i t y = o d e i n t ( v e l d e r i v , vo , t i m e s )

plt . p l o t ( times , v e l o c i t y )
plt . x l a b e l ( ’ time ( s e c ) ’ )
plt . y l a b e l ( ’ v e l o c i t y (m/ s ) ’ )
plt . show ( )

i i

1.5.2 Different Computational Tools

A carpenter has a toolbox filled with a variety of tools, each tool designed for a specific
task. The same is true for computation: there are different types of programming languages,
each ideal for a variety of tasks. A computer programming language (here after, we will
just call them “languages”) is a special language programmers use to give a computer
instructions on how to perform a specific task. In this book, those tasks will focus on
performing the mathematics needed to solve physics problems. There are many different
computer languages used by physicists to solve problems; each language has its own set of
advantages and disadvantages. Just like a carpenter can use a screwdriver as a hammer,
computer languages can be used in applications they might not have been originally intended
for, to varying degrees of success. It is important for a physicist to be comfortable with a
few different programming languages so that he or she will be prepared for a variety of
problems.
In this book, we will divide languages up into three types: computer algebra systems
(CAS), interpreted languages, and compiled languages. This division is not intended to be
complete and cover all languages, but it will establish a way of thinking for how we will
approach languages in this book. Keep in mind that many languages will actually fit in
multiple categories, for example any language can be compiled or interpreted, but we will
assign languages by type in the way they are most often used.
• Computer Algebra System (CAS): CAS is software that can manipulate mathe-
matical expressions, similar to the ways you have learned in your mathematics courses.
CAS can solve a variety of mathematical problems in closed form, as we saw in Exam-
ple 1.2. Besides symbolic manipulations, CAS often has numerical components and its
own programming language. For example, the CAS Mathematica is also an interpreted
language (see below). Examples of CAS are: Mathematica, Maple, and SageMath.
• Interpreted Language: An interpreted language is one that can be executed directly
from the source code using software called an interpreter. Many people feel that inter-
preted languages are easier to learn, have a shorter development time (i.e., it takes
less time to write code), and are easier to debug (find and fix errors in code). Inter-
preted languages are generally easier to run on multiple operating systems, i.e., the
same Python (an interpreted language) script can be run on a Windows and a Linux
computer, provided both computers have Python installed on them. Examples of
interpreted languages include: Python, Mathematica, R, MATLAB, and JavaScript.
• Compiled Language: A compiled language is one whose source code needs to be
compiled by software called a compiler, which transforms the source code into machine
code and creates an executable. The executable is then run in order to perform the
instructions contained in the program. If changes are made to the program, it needs to
be recompiled, unlike an interpreted language. Likewise, the compiler creates instruc-
tions for the specific type of computer being used. Hence the executable created by
compiling a program on a Windows computer will not work on a Mac. Compiled lan-
guages typically have a shorter run-time compared to interpreted languages, although
for simple computational problems, the difference can be minimal. Examples of com-
piled languages are: FORTRAN, C, and C++.
The above categories are not intended to be strict divisions. We already mentioned that
theoretically, any language can be either interpreted or compiled. In addition, although
we will tend to think of Mathematica as a CAS in this book, Mathematica is also an
interpreted language which can be used to solve any numerical or symbolic computational

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 21 — #37

i i

The Foundations of Motion and Computation 21

problem. Python, an interpreted language, has a library called SymPy, which, as we saw
above, allows Python to become a CAS.
The type of language that you use depends on the nature of the problem you are trying
to solve. If you need help simplifying the mathematical form of an equation, a CAS is prob-
ably your best choice. While SymPy may be capable of mathematical manipulations, you
may find Mathematica easier to use for that problem, since Mathematica was specifically
designed for, among other things, symbolic manipulations. Interpreted languages are help-
ful for problems that are not computationally intensive, such as numerically solving certain
differential equations. The speed of compiled languages makes them ideal for computation-
ally intensive programs, such as climate models or modeling the motion of many stars in a
galaxy.
In this book, we are not going to be tied to one specific language. The focus of this book
is on classical mechanics and computation, not on a particular programming language.
The problems we will solve in this book are not computationally intensive, i.e., they won’t
demand a lot of computer resources or time. Hence, we will not be using a compiled language
in this book. For the sake of consistency, we chose to demonstrate the computations using
two languages, Mathematica and Python. The exercises and examples done in this book
can be done in any language you prefer, and, when appropriate, the code will be written in
such a way that it will be easy to reproduce in your language of choice. However, we felt
it would be best to avoid using too many different languages in the examples done in the
text.
Our choice of Mathematica and Python as example languages is not random. Both lan-
guages are used extensively in physics, and both languages are mature. The code that we will
write will focus on core commands that have been around for a long time and are not likely to
become outdated anytime soon. Python is free to use and available online for any platform.
You’ll likely want to download a distribution like Anaconda (https://www.anaconda.com)
which comes with an integrated developer environment (IDE). Mathematica is not free,
however, there are open source alternatives, such as SageMath which is available online
(https://www.sagemath.org/). SageMath was developed as an open-source alternative to
Mathematica. While there will be syntax differences between Mathematica and SageMath,
the symbolic calculations done in this text in Mathematica can, for the most part, be done
in SageMath.
You do not need to have experience with any programming languages to understand the
material presented in this book. Such previous experience will undoubtedly help in under-
standing the code contained in the book. However, we will introduce the necessary program-
ming concepts as they are needed, and we assume no programming experience. Mathematica
has extensive and very useful help files that we recommend going through. If you’d like to get
a head start on Python, we recommend Code Academy (https://www.codecademy.com/)
and the most recent edition of [Kinder and Nelson(2015)]. A quick way of getting your
feet wet and learning some basic programming skills is through using your favorite search
engine. Searches such as: “How to solve a differential equation in Mathematica” or “How
to integrate an equation in Python” often lead to many useful pages, which will at least
give you the syntax for how to solve a specific problem. While this is no substitution for an
introductory computer science course or a course in numerical analysis, it will certainly get
you started.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 22 — #38

i i

22 Classical Mechanics: A Computational Approach

1.5.3 Some Warnings

We will end this section with some warnings. First, the availability of computational tools to
solve problems does not mean that you do not need to know how to do math by hand! There
are several reasons for this. The first reason is that computer algorithms can and do give
wrong answers. Computational algorithms have significantly improved over the decades, and
wrong answers from them are not as common as in the early days of computing. However,
you need to be able to identify a wrong answer when you see one. A proficiency with
mathematics is still needed for identifying wrong answers. Furthermore, it is not unusual
for one to perform additional algebraic manipulation to the output of a CAS in order for
the result to be in a more convenient or insightful form. We will see some examples in this
book where the output of a CAS will be technically correct, but not as useful for describing
the physics of the situation as compared to a result that we would have obtained “by hand.”
Second, a computer will provide a solution with no context. The techniques and critical
thinking that you learn by doing mathematics will train you in how to understand and
interpret computer-generated solutions, so that you can put the solution in the proper
context. For example, does your solution have appropriate limits? Does the solution conform
to known laws of physics? A person skilled in mathematics and physics is still required to
interpret computer-generated solutions (at least for now...). In particular, when one becomes
skilled at mathematics, symbolic algorithms should be considered as a tool to help one
focus on the physics without getting encumbered by lots of mathematics. In a sense, you
should think of symbolic algorithms as being similar to a calculator. Only after learning
arithmetic “by hand,” should one begin to use a calculator. In the same sense, mastery of
algebra, trigonometry, and calculus, should be a prerequisite for extensive use of symbolic
algorithms.
Third, who programs computers to solve math problems? The people who know how to
solve the math problems. Hence, we need people who can do math, in order to ensure that
we will have future generations of algorithms to assist us.
The final warning comes from the nature of the algorithms presented in this text. The
authors do not claim that the codes contained within are the most efficient means of solving
the problems in the text. We tended to produce codes that focused on pedagogy over
efficiency. We encourage you to rewrite the programs contained within, to see if you can
improve them and make them more efficient. By playing around with the codes, you’ll get
a better understanding of the algorithms and programming in general. Have fun!

1.6 CLASSICAL MECHANICS IN THE MODERN WORLD

Students sometimes come to a classical mechanics text wondering about the relevance of
the material to modern-day physics. They have questions about why they need to learn
classical mechanics when “all of the interesting stuff” is in the quantum realm or deals with
relativity. While some of the most exciting popular science books deal with the realms of
quantum mechanics and cosmology, there is plenty of active research going on in classical
systems.
One of the most exciting fields of research in classical systems is in nonlinear and com-
plex systems. Nonlinear systems is a field of science that attempts to understand systems
whose governing equations are nonlinear, and often not solvable in closed-form. These types
of systems include large amplitude oscillators, weather, predator-prey models, and certain
chemical reactions. Notice that two of these three examples are not from physics. That
is because nonlinear systems is a multidisciplinary field and one in which physicists have
made many contributions. Complex systems, simply put, are large systems of many inter-

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 23 — #39

i i

The Foundations of Motion and Computation 23

acting agents whose collective behavior cannot be explained by the dynamics of the indi-
vidual agents alone. Examples of complex systems include power grids, global climate, and
economies. Again, these examples aren’t strictly “physics problems” but rather problems to
which physicists have contributed greatly. Progress in the fields of nonlinear and complex
systems has required people trained in physics, notably classical mechanics and statistical
mechanics.
Another important field involving classical mechanics is fluid mechanics and turbulence.
There is a famous story of Werner Heisenberg, that on his deathbed he said, “When I
meet God, I am going to ask him two questions: Why relativity? And why turbulence? I
really believe he will have an answer for the first.” Variations of this story are told about
other physicists as well. The point is that turbulence continues to be a difficult problem in
classical mechanics. Improvements in our understanding of turbulence will greatly benefit
many technologies that involve fluid flow, such as pipelines and flight.
The paragraphs above contain just a few examples of the relevance of classical mechanics
to the modern world. We have left out many other examples such as mechanical engineering,
oceanography, and space flight. The point is that ignoring classical mechanics or treating
it simply as a step to get to something else, is a mistake. A fundamental understanding
of classical systems is critical to becoming a complete physicist. In many ways, classical
mechanics serves as the foundation of the field of physics itself. A physicist needs a solid
foundation of classical mechanics in order to understand other sub fields of physics.
One of the many benefits of learning classical mechanics is that it gives you the ability
to identify what factors influence a system (physical or otherwise) and how to express those
factors mathematically. Much of this ability centers on the idea of learning how to specify
a system, identify the forces acting on the system, and then writing down the relevant
equations of motion that govern the system. It is a powerful and useful way of thinking,
that will be a critical component of your training as a physicist and is a valuable skill
regardless of how you end up using your physics training.

1.7 CHAPTER SUMMARY

Motion is described using the basic descriptors of motion, displacement, velocity, and
acceleration. In Cartesian coordinates the position, velocity, and acceleration of a particle
are:
r = xî + y ĵ + z k̂ (1.7.1)
dr dx dy dz
v= = î + ĵ + k̂ = vx î + vy ĵ + vz k̂ = ẋî + ẏ ĵ + ż k̂ (1.7.2)
dt dt dt dt
dv dvx dvy dvz
a= = î + ĵ + k̂ = ax î + ay ĵ + az k̂ = ẍî + ÿ ĵ + z̈ k̂ (1.7.3)
dt dt dt dt
where î, ĵ, and k̂ are the unit vectors.
When describing the motion of an object, two additional quantities become important,
the mass of the object and the net force acting on the object. The inertial mass of an
object minertial is a measure of the object’s inertia, where as the gravitational mass of
an object mgravitational is the mass that determines the gravitational force acting on it.
However,the weak principle of equivalence says that minertial = mgravitational .
A change in an object’s motion is caused by a nonzero net external force acting on the
object. Newton’s laws of motion describe how forces change the motion of an object.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 24 — #40

i i

24 Classical Mechanics: A Computational Approach

Newton’s First Law: A particle’s velocity remains constant if the net force acting on
the object is zero.
Newton’s second law: A particle’s time rate of change of momentum dp/dt, is equal to
the net force F applied to the particle :

F = dp/dt = ṗ (1.7.4)

Newton’s third law: If object 2 exerts a force F12 on object 1, then object 1 exerts a force
F21 on object 2 such that:
F21 = −F12 (1.7.5)
All measurements are made with respect to a reference frame. A reference frame is a
choice of origin (spatial x and temporal t), and a set of axes with respect to which all
measurements are made.
Two reference frames S, S 0 are called inertial reference frames when Newton’s laws
hold, and the forces measured in S 0 are the same forces as those measured in S.
Two reference frames S, S 0 are called noninertial reference frames when Newton’s laws
no longer hold in their standard form, because there is an acceleration measured in S that
is not apparent in S 0 .
Computation is an important tool for physicists. In this book, we’ll focus on two
types of computation, symbolic and numerical. Symbolic computation involves using a
computer to help find closed-form solutions to differential equations, integrals, eigenvalues,
etc. Numerical computation or “numerics,” provides a list of numerical values representing
a solution to differential equations, integrals, eigenvalues, etc.
Computer programming languages can be loosely categorized as a computer alge-
bra systems, interpreted languages, and compiled languages. A computer algebra system
(CAS) is a software that can manipulate mathematical expressions, and can solve a vari-
ety of mathematical problems in closed-form. An interpreted language is one that can be
executed directly from the source code, using software called an interpreter. A compiled
language is one whose source code needs to be compiled by software called a compiler,
which transforms the source code into machine code.

1.8 END-OF-CHAPTER PROBLEMS

with the initial condition f (x = 0) = 3. Plot the solution.

26. Numerically solve the differential equation:
df
= 5 sin(x) − 4e−x
dx

for the initial condition f (0) = 0. Plot the solution for x = 0 to 3 with a step size of
0.1.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 28 — #44

i i

28 Classical Mechanics: A Computational Approach

27. Numerically solve the coupled differential equations:

dx
= 0.6x − 1.2xy,
dt
dy
= xy − y,
dt
using the initial conditions x(0) = 40 and y(0) = 7. Plot x as a function of time, y as
a function of time, and y as a function of x.
28. Use symbolic computation to perform the integral:
Z
x2 cos(x)dx.

29. Numerically evaluate the definite integral:

Z 2π
x2 e2x dx.
0

30. Numerically evaluate the definite double integral:

Z 1Z 2
xydxdy.
0 0

If you are using Python, you will want to consider the nquad command from the SciPy
integrate library.
31. Numerically evaluate the integral:
Z 2 Z 1−y
xydxdy.
0 0

As with Problem 30, you will want to consider using the nquad command if you are
using Python.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 29 — #45

i i

CHAPTER 2

Single-Particle Motion in
One Dimension

In this chapter, we will examine one-dimensional motion, i.e., motion along a line. It is
sometimes the case that a particle’s motion need only to be described along one direc-
tion. Furthermore, a careful study of one-dimensional motion will be a useful foundation
for understanding more general motion in higher dimensions. In this chapter, we will give
several examples of solving Newton’s second law, F = ma in one dimension. We will con-
sider several types of forces: both constant and those which depend on time F (t), velocity
F (v), and position F (x). In addition, we will discuss and demonstrate two different uses
of computers to solve physics problems: how to use computer algebra systems (CAS) to
obtain the analytical solutions of Newton’s second law, and how to obtain numerical solu-
tions of ordinary differential equations (ODE) using software packages and by using the
Euler method.

2.1 EQUATIONS OF MOTION

To begin our study of one-dimesional motion, we first need to make some assumptions about
the object whose motion we are examining. One fundamental assumption in this chapter is
that the object being studied is a point particle. In order to mathematically describe the
motion of a particle under the influence of a force, we need to find the particle’s equations of
motion. The equations of motion of a particle are the equations which describe its position,
velocity, and acceleration as a functions of time. Equations of motion can be in the form of
algebraic equations, or in the form of differential equations.
As we will see, the equations of motion of a particle can be found by solving Newton’s
second law as a differential equation. In this chapter, we will focus on one-dimensional
motion, where the force vector and the particle’s displacement are along the same line (but
not necessarily in the same direction—the direction could be horizontal or vertical). Because
all vectors in a given problem lay along the same line, we drop the vector notation in all the
equations. A negative sign between two quantities will denote vectors that lay in opposite
directions along the same line.
Newton’s second law in one dimension is:

and is performed by treating the derivative as a fraction and multiplying both sides of
(2.2.1) by dt,

dx =7dt (2.2.2)
Z Z
dx = 7dt (2.2.3)

x =7t + c (2.2.4)

To solve (2.2.1), we carried out the integral after separating out variables. Note that
both integrals would produce constants of integration, but since both are constants, we can
combine them into one arbitrary constant. You can double-check the solution by computing
the derivative of 7t + c, to check that it satisfies (2.2.1).
What happens in the case of second-order ODEs? Second-order ODEs are common
in physics. There are many techniques to solve them, but we will demonstrate only one.
Consider the ODE,

d2 x
=7 (2.2.5)
dt2
Separation of variables does not make sense here, because we typically do not integrate
terms like d2 x. However, we can define a new variable v, such that, v = dx/dt. Then (2.2.5)
becomes,
dv
=7 (2.2.6)
dt
which we know gives the answer v(t) = 7t + c1 , where c1 is the constant of integration.
However, we want x(t), so we use v = dx/dt:

dx
=7t + c1 (2.2.7)
Z dt Z
dx = (7t + c1 ) dt (2.2.8)

x(t) =3.5t2 + c1 t + c2 (2.2.9)

where c2 is the constant of integration obtained by performing the above integral. Hence, we
pick up an additional constant of integration in our solution, giving two arbitrary constants
for the solution of the second-order ODE (2.2.5). Loosely speaking, we see that the number
of arbitrary constants in the solution of an nth -order ODE is equal to n, because we need to
do n integrations in order to solve the equation, and each integration produces an arbitrary
constant.
Next, we return to (2.2.1). Suppose (2.2.1) was an equation we wanted to use in order
to find the position of a particle as a function of time. The infinite number of solutions is
not helpful. Which solution describes the actual path taken by the particle? In order to
specify the particular solution for an ODE, we need to include initial conditions, the value
of our function at a particular time (normally at t = 0). Suppose we know that at t = 0, the
particle is at a position, x(0) = 3. Then we can solve for the arbitrary constant by inserting
the initial condition into our general solution,

x(0) = (7) (0) + c = 3 (2.2.10)

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 32 — #48

i i

32 Classical Mechanics: A Computational Approach

which gives c = 3. Our particular solution is then, x(t) = 7t + 3. A different initial condition
will give a different particular solution. Now suppose we wanted to find a particular solution
to (2.2.5); in that case one initial condition will not be enough because it will leave one
arbitrary constant. Hence, we will need to specify both x(t) and dx/dt at a particular time
(usually t = 0). Suppose that x(0) = 3 and v(0) = 1, where v = dx/dt. Then we have:

x(0) = 3.5 (0)2 + c1 (0) + c2 =3 (2.2.11)

v(0) = 7 (0) + c1 =1 (2.2.12)

where (2.2.12) is the derivative of the general solution evaluated at t = 0. The result here
is that c1 = 1 and c2 = 3, and the particular solution is x(t) = 3.5t2 + t + 3. In summary, in
order to solve for the particular solution of an nth -order ODE, we need n initial conditions.
In addition, we can also specify x using knowledge of the value of x at two different times,
as opposed to knowing initial values of x and its first derivative. In classical mechanics, it is
most common to know the initial conditions of the position and velocity. However, in other
fields, such as electromagnetism and thermodynamics, it is often more common to know the
value of a function, say the temperature, at two different locations. In this case, we have
what is known as a boundary value problem, and the conditions that provide the constants
of integration are known as boundary conditions.
There are many other properties of ODEs such as linear superposition, that we will
explore in this book as we need them. For now, we know enough about ODEs to get started.
Let’s get back to the physics.

2.3 CONSTANT FORCES

Consider the case where a constant net force, F = F0 , acts on a particle constrained to move
along a line. In this case, Newton’s second law (2.1.1) gives:
dv F0
= =a (2.3.1)
dt m
where a is a constant because F0 is constant. You may recall (2.3.1) from Example 1.1.
Equation (2.3.1) is an example of a first-order differential equation. We can solve this
through the process of separation of variables and then integrating the resulting equation.
The first step in separating variables is to multiply both sides of the equation by dt:

dv = adt (2.3.2)
which can be integrated to yield,
Z
v(t) = a dt = at + c1 (2.3.3)

where c1 is the constant of integration. The constant, c1 , can be found using initial conditions
of v(t0 ) = v0 , where v0 is the initial velocity at the initial time t0 . When t0 = 0, (2.3.3) gives
v(0) = c1 or c1 = v0 . Therefore, the solution to the differential equation, (2.3.1) is,

v(t) = v0 + at. (2.3.4)

Before moving forward, we should mention that there is an alternate method for finding
(2.3.3), which can be found by explicitly inserting the initial and final conditions when
integrating (2.3.2):

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 33 — #49

i i

Single-Particle Motion in One Dimension 33

Z v(t) Z t
0
dv = adt0 (2.3.5)
v0 t0
v(t) − v0 =a(t − t0 ) (2.3.6)

where we have introduced primes to the variables of integration in order to distinguish them
from the limits of integration. Notice that the lower limit in the left-hand side of (2.3.5)
corresponds to the value of v(t) when t = t0 , the lower limit of the right-hand side of (2.3.5),
and with similar considerations for the upper limits. It is very important that the limits
match on both sides of the equation.
Next, we can get an equation for x(t) by writing v = dx/dt :
dx
= at + v0 , (2.3.7)
dt
and separating variables we obtain:
Z x(t) Z t
dx0 = (at0 + v0 )dt0 (2.3.8)
x0 t0
Notice how all of the time-dependent and constant terms are on the same side of the
equation. If we had subtracted v0 from both sides of the equation before integrating, we
would have dx/dt − v0 , which does not make sense because v0 is meaningless without a
R

variable of integration. In this context, separation of variables always involves multiplication

and division, not addition and subtraction. Choosing t0 = 0 and performing the integral
results in:
1
x(t) = at2 + v0 t + x0 (2.3.9)
2
Together, equations (2.3.4) and (2.3.9) are the only equations you need to know, in order
to solve for the position and velocity of a particle moving in one-dimension and experiencing
a constant net force.
Finally, if we want v(x) we can solve (2.1.4):

dv F0
v= (2.3.10)
dx m
vdv =adx (2.3.11)
Z v Z x
0 0
v dv =a dx0 (2.3.12)
v0 x0
v − v02 =2a(x − x0 )
2
(2.3.13)

where a = F0 /m was used in (2.3.11). We could have also obtained this result by eliminating
t between equations (2.3.3) and (2.3.9). The box below summarizes all of the constant force
equations, sometimes called the kinematic equations. These are equations that you should
memorize.

kinematic equations (a=constant)

v(t) = vo + at (2.3.14)

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 34 — #50

i i

34 Classical Mechanics: A Computational Approach

1
x(t) = xo + vo t + at2 (2.3.15)
2

v(x)2 = vo2 + 2a (x − xo ) (2.3.16)

In the case of a freely falling particle near the surface of the Earth, we use a = −g =
−9.8 m/s2 (assuming “down” is in the negative direction), where g is the acceleration due
to gravity. The kinematic equations become:

v(t) =v0 − gt (2.3.17)

1
y(t) =y0 + v0 t − gt2 (2.3.18)
2
v(y)2 =v02 − 2g(y − y0 ) (2.3.19)

and by dividing with the mass m we find the constant acceleration:

a = g sin θ − µg cos θ (2.3.24)

This acceleration can then be used in (2.3.15) and (2.3.14) to get x(t) and v(t), respec-
tively.

2.4 TIME-DEPENDENT FORCES

Now consider a case where a particle constrained to move along a line is experiencing
a net force F (t), which is dependent on time. Time-dependent forces arise in a variety of
applications, one common occurrence is when a sinusoidal external force acts on the system.
In the cases where the net force F = F (t), Newton’s second law 2.1.1 gives:

dv F (t)
= (2.4.1)
dt m
and by separating variables and integrating from t0 = t0 to t we find :

Zt
1
v(t) − v(t = 0) = F (t0 ) dt0 (2.4.2)
m
t0 =t0

By using the initial conditions v(t = 0) = v0 :

1 t
Z
v(t) = v0 + F (t0 ) dt0 (2.4.3)
m t0 =0

Once we know v(t), we can find x(t) by integrating v = dx/dt and using the initial
conditions x(t0 ) = x0 to obtain:

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 37 — #53

i i

Single-Particle Motion in One Dimension 37

Z t
x(t) − x0 = v(t0 ) dt0 (2.4.4)
t0 =0

Next, we will demonstrate an example of how to find the equations of motion for a
particle experiencing a time-dependent force. We will demonstrate the solution both by
hand and using the CAS Mathematica. Refer to Chapter 1 for comments on how to use
Mathematica to solve differential equations in closed form.

Example 2.3: A decreasing force, F(t)

The force acting on a mass m is decreasing with time according to F (t) = F0 /(t2 + 1)
where t is the elapsed time. Find the position x(t) and the velocity v(t) with the initial
conditions x(0) = x0 = 0 and v(0) = v0 = 0.
Solution:
This is the case of a time-dependent force F = F (t), so we can use (2.4.3) with v0 = 0:

1 t 1 t
Z Z
0 0 F0 F0
v(t) = v0 + F (t ) dt = dt = tan−1 (t) (2.4.5)
m t0 =0 m t0 =0 t2 + 1 m
Does our v(t) make sense? To check, we ask what happens when t → ∞? In the limit
of t → ∞, F → 0, and we would expect that v becomes constant. In our particular case,
as t → ∞, tan−1 (t) → π/2, and therefore, our solution v(t) follows our expectation of the
velocity approaching a constant. Next, we find x(t) using (2.4.4) with x0 = 0:

Z t
x(t) − x0 = v(t0 ) dt0 (2.4.6)
t0 =0
F0 t
Z
= tan−1 t0 dt0 (2.4.7)

m t0 =0
1

F0
= t tan−1 (t) − ln t2 + 1 (2.4.8)

m 2

In this case, as t → ∞, we know that v → constant, and therefore, we would expect

f = −f (v)v̂ (2.5.1)
where the unit vector v̂ = v/v is along the direction of the object’s velocity (or motion) and
the minus sign is included to explicitly show that the direction of the air resistance force
is opposite of the direction of the velocity. The magnitude of the air resistance f (v) is a
function of velocity which generally takes the form,

f (v) = bv + cv 2 (2.5.2)
and therefore has both a linear and a quadratic component in v. For spheres in air the
values of the constants b and c are [Taylor(2005)]:

b =1.6 × 10−4 D N·s/m

(2.5.3)
c =0.25D 2
N·s /m2
2

where D is the diameter of the sphere in meters. Fortunately, one usually doesn’t need to
include both the linear and quadratic terms for air resistance. In order to determine which
term, linear or quadratic, is the most important in a given situation, one can compute the
ratio:

cv 2 0.25D2 v 2
γ= = = 1.6 × 103 Dv (2.5.4)
bv 1.6 × 10−4 Dv
Note that v should be in meters per second, and D in meters. If γ >> 1, then the quadratic
term dominates, and the linear term can be neglected. If γ << 1, then the linear term
dominates, and the quadratic term can be neglected. However, if γ ≈ 1, then both the
linear and quadratic terms need to be included. So for a 0.22 m soccer ball, the quadratic
term dominates for speeds approximately greater than 0.003 m/s (or 3 mm/s), while for
lower speeds the linear term dominates. Both linear and quadratic terms would need to be
included in the air resistance formula if the soccer ball is traveling near 3 mm/s.
To obtain equations for position and velocity as functions of time, we will consider a
generic form for F (v). Newton’s second law (2.1.1) gives:

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 40 — #56

i i

40 Classical Mechanics: A Computational Approach

dv F (v)
= (2.5.5)
dt m
Rearranging the above equation gives:
dv
dt = m (2.5.6)
F (v)
and by integrating from t0 = t0 to t , and using the initial conditions v(t0 ) = v0 we find :
v
dv 0
Z
t − t0 = m (2.5.7)
v 0 =v0 F (v 0 )

By solving (2.5.7), we can find v(t), the velocity as a function of time. Once we know
v(t), we can again use (2.4.4) to find x(t) with the initial condition x(t0 ) = x0 .
We can follow a similar procedure for finding v(x). Newton’s Second Law gives:
dv
mv = F (v) (2.5.8)
dx
After rearranging, we obtain:
v
v 0 dv 0
Z
x − x0 = m (2.5.9)
v 0 =v0 F (v 0 )

Next, we will show some examples involving velocity-dependent forces.

Example 2.4: Air resistance varying linearly with speed

The force of air resistance on a mass m depends linearly on the velocity v according
to F (v) = −bv. Find x(t) and v(t) with the initial conditions v(0) = v0 and x(0) = x0 .
Solution:
We use (2.5.7) to find x(t) with t0 = 0:
Z v
dv 0 m v dv 0
Z
m v
t=m = − = − ln (2.5.10)
0
v =v0 F (v 0) b 0
v =v0 v 0 b v0

Solving for v(t) we obtain:

bt
v(t) = v0 e− m (2.5.11)
We also find x(t) from (2.4.4):
Z t Z t
0 0 bt0
x(t) − x0 = v(t ) dt = v0 e− m dt0 (2.5.12)
t0 =t0 t0 =t0

which gives:
mv0 bt

x(t) = x0 + 1 − e− m (2.5.13)
b
Here is the solution using Mathematica:

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 41 — #57

i i

Single-Particle Motion in One Dimension 41

ODEsolution = DSolve[{m ∗ x00 [t] == −b ∗ x0 [t], x[0] == xo, x0 [0] == vo}, x[t], t];

position = Collect[x[t]/.ODEsolution[[1]], {vo, m}]

bt

mvo 1−e− m
OUTPUT: b + xo

velocity = D[position, t]
bt
OUTPUT: e− m vo

vo = 1; xo = 0; m = 1; b = 1;

GraphicsGrid[{{Plot[position, {t, 0, 10}, Frame → True, Axes → False, FrameLabel →

{“time, s”, “x, m”}, BaseStyle → {FontSize → 24}, PlotRange → All], Plot[velocity,
{t, 0, 10}, Frame → True, Axes → False, FrameLabel → {“time, s”, “v, m/s”}, BaseStyle →
{FontSize → 24}, PlotRange → All]}}]

1.0 1.0
0.8 0.8
v, m/s

0.6 0.6
x, m

0.4 0.4
0.2 0.2
0.0 0.0
0 2 4 6 8 10 0 2 4 6 8 10
time, s time, s

The Collect command gathers together terms that involve the same powers of the
objects in the curly brackets (second argument of Collect). It is a useful command for
simplifying algebraic terms and for identifying coefficients.

Example 2.5: Air resistance in the presence of gravity

Consider the case of a falling object near the surface of the Earth. Suppose in this case,
the air resistance is linear. Then the net force on the falling body depends on the velocity
v according to
F (v) = −bv + mg (2.5.14)
where we have defined +y to be vertically downward so that the body’s motion is in
the positive direction. Find the displacement y(t) and the velocity v(t), with the initial
conditions v(0) = v0 and y(0) = y0 .
Solution:
We use (2.5.7) to find v(t) with t0 = 0:
Z v mg
dv 0 b −v

m
t=m = − ln mg (2.5.15)
v 0 =v0 −bv + mg
0 b b − v0

Note that we have factored out the constant b from the denominator of the integral
before performing the integration.
Solving for v(t) we obtain:

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 42 — #58

i i

42 Classical Mechanics: A Computational Approach

mg mg − bt
v(t) = + v0 − e m (2.5.16)
b b
We also find y(t) from (2.4.4):
Z t Z t
0 0 mg mg − bt0
y(t) − y0 = v(t ) dt = + v0 − e m dt0 (2.5.17)
t0 =t0 t0 =t0 b b

which gives after integrating:

2
mg m g mv0 − bt
y(t) = y0 + t+ − e m (2.5.18)
b b2 b
bt
When t → ∞ we can substitute e− m → 0 in (2.5.16) to obtain the terminal velocity:
mg
vterminal = (2.5.19)
b
Hence, our solution matches the expected physics. As the object falls, its velocity
increases due to the force of gravity. However, at some point, the object is moving fast
enough that the drag force equals that of the object’s weight. At that point, the net force
on the object is zero, and the velocity remains at a constant value called the terminal
velocity, vterminal for the duration of its fall.

Figure 2.3 shows the plots of x(t) and v(t) found in Example 2.5. Notice that both plots
show the velocity of the object coming to a constant value as the object falls for a long
period of time.

1200 50
1000 40
800
v, m/s

30
x, m

600
20
400
200 10
0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
time, s t, s

0.8
velocity, m/s

0.6

Out[ ]=
0.4 Linear
Quadratic
0.2

0.0
0 5 10 15 20
time

Figure 2.4: Comparison of the velocity of falling body in the presence of linear and quadratic
air resistance. The two curves have been scaled, so that they produce the same result as
t → ∞.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 45 — #61

i i

Single-Particle Motion in One Dimension 45

2.6 POSITION-DEPENDENT FORCES

In the cases where the force is a function of the position F = F (x), we can use Newton’s
second law in the form
dv
F (x) = mv (2.6.1)
dx
By separating the variables v and x, and using the initial conditions x = x0 and v = v0
when t = 0 , we obtain:

1 1
Z x Z v
F (x0 )dx0 = mv 0 dv 0 = mv 2 − mv02 (2.6.2)
x0 v0 2 2
Solving for the velocity v(x) as a function of the position x:
sZ
2 x
1
v (x) = F (x0 )dx0 + mv02 (2.6.3)
m x0 2

By substituting v = dx/dt, separating the variables, and integrating with the initial
condition x = x0 when t = t0 , we obtain the relationship between position x and time t:

 
x
dx0
Z
m
t − t0 = (2.6.4)
2
 qR 
x
x0 F (x )dx + 2 mv0
x0 0 1 2

After performing the integral in equation (2.6.4), one would then try to invert this
formula to find x(t), something that is not always easy or possible to do. Often, one relies
on computer methods in such cases.
The next example will involve one of the most important problems in all of classical
mechanics, simple harmonic motion (SHM). In fact, we will devote a whole chapter to it! As
we will see in a later chapter, SHM is a common model used for small amplitude oscillations.
Although SHM is often thought of as a “mass on a spring,” it is also a useful model for
other small amplitude oscillations including pendulums, small amplitude water waves, and
loudspeakers—to name a few. Besides presenting the SHM, the next example also includes
a demonstration of using Python to find and plot the solution to a differential equation
in closed form. Python’s SymPy package was used in Chapter 1, to solve a first-order
differential equation. Here we will demonstrate the solution of a second-order differential
equation using SymPy.

Example 2.7: Simple harmonic motion

A mass m is under the influence of a position-dependent force, F = F (x) = −kx,
which is proportional to the distance x the mass is from an equilibrium position. The
mathematical form of F is, of course, the well-known Hooke’s law for springs. Find x(t)
and v(t) with the initial conditions v(0) = 0 and x(0) = x0 6= 0 at t = 0.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 46 — #62

i i

46 Classical Mechanics: A Computational Approach

Solution:
In this case (2.6.3) becomes:
s Z
2 x
1 2 1
r
v= kx dx + mv0 =
2 k x20 − x2 (2.6.5)

− 0 0
m x0 2 m 2

which gives the velocity v(x) as a function of the position x. Next we will find x(t) by
using (2.6.4):
Z x x r
dx0
r
m −1 x m −1 x π
t= = sin = sin − (2.6.6)
2
q
k x0 k x0
m (x0 − x )
x0 k 2 02
x0

or after some rearranging:

r ! r !
k π k
x(t) = x0 sin t+ = x0 cos t (2.6.7)
m 2 m
The result is that the mass oscillates about the equilibrium
q position x0 . Hence we find
that the natural frequency ω of the oscillator is ω = m k
.
Algorithm 1 is the solution using the SymPy library in Python. Also note that while
not commented-out, the lines starting with the word OUTPUT: are not code and are
included only to show the output from the line above. Do not include such lines in your
own code.
The code used to solve Example 2.7 is similar to what we used in Chapter 1. After
importing everything in the SymPy library (the * means “everything”), we specify that the
variables m, k, t are real and positive. Python needs to be told which variables are to be
treated as symbols and which variables are to be treated as functions. Next, the differential
equation is defined using the variable diffeq. The Eq command creates an equation with
the first argument as the left-hand side of the equation and the second argument as the
right-hand side. Hence, that line of code is the same as writing, mẍ + kx = 0. Note that the
command diff, represents the derivative of x with respect to time (two t’s in the argument
means second derivative). The next line solves the differential equation using dsolve. At
the time of this writing, the dsolve command in Python cannot include initial conditions,
unless one wants to obtain power series solutions to the differential equation. Therefore,
the next seven lines of code insert the initial conditions. To solve for the initial conditions,
we need to first isolate the right-hand side of the solution. Next, we identify C1 and C2
as symbols, even though they appear in the SymPy-produced solution. The lines starting
with ic1 and ic2 apply the initial conditions by setting x(0) = x0 and ẋ(0) = 0, respectively.
Then the resulting two equations are solved for the two unknowns, and the solution is stored
in solution ics. Finally, we obtain the particular solution by substituting the values of C1
and C2, stored in solution ics, into the general solution.
Like in Chapter 1, the code needed to solve Example 2.7 in Python is more complicated
than that used by Mathematica to solve previous examples in this chapter. Python needs
to be told which terms are symbols for manipulation and which terms are functions. In
addition, SymPy’s dsolve currently has no way of including initial conditions into the solu-
tion of the ODE, so more than half of the lines in Example 2.7’s algorithm are used to find
the arbitrary constants and plug them into the solution. As we’ve seen with Mathematica,
finding the constants using initial conditions is easy by using its DSolve command.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 47 — #63

i i

Single-Particle Motion in One Dimension 47

from sympy import ∗

x = Function ( ’ x ’ )
t = Symbol ( ” t ” , r e a l=True , p o s i t i v e=True )
k = Symbol ( ” k ” , r e a l=True , p o s i t i v e=True )
m = Symbol ( ”m” , r e a l=True , p o s i t i v e=True )

d i f f e q = Eq (m∗x ( t ) . d i f f ( t , t ) + k∗x ( t ) , 0 )

solution = dsolve ( diffeq , x( t ))

print ( s o l u t i o n . rhs )

OUTPUT: C1∗ s i n ( s q r t ( k ) ∗ t / s q r t (m) ) + C2∗ c o s ( s q r t ( k ) ∗ t / s q r t (m) )

equation = s o l u t i o n . rhs
C1 , C2 = symbols ( ’ C1 , C2 ’ )
xo = symbols ( ’ xo ’ )
i c 1 = Eq ( e q u a t i o n . s ub s ( t , 0 ) , xo )
i c 2 = Eq ( e q u a t i o n . d i f f ( t ) . s u b s ( t , 0 ) , 0 )
s o l u t i o n i c s = s o l v e ( [ i c 1 , i c 2 ] , ( C1 , C2 ) )
print ( solution ics )

OUTPUT: {C1 : 0 , C2 : xo }

p a r t i c u l a r = s i m p l i f y ( e q u a t i o n . su bs ( s o l u t i o n i c s ) )
print ( particular )

OUTPUT: xo ∗ c o s ( s q r t ( k ) ∗ t / s q r t (m) )

Algorithm 1: The Python code for Example 2.7.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 48 — #64

i i

48 Classical Mechanics: A Computational Approach

We didn’t need to specify that t, m, and k are real and positive. If we had only defined
those variables as symbols, Python would have still solved the differential equation. However,
Python would produce a cosh function instead of the cos function. The argument of the cosh
function produced has a negative sign under a radical. Using the relationship, cosh(ix) =
cos(x) we would get the same answer we obtained by hand and by specifying the nature of
the variables. It is not unusual to have to do additional algebra on results provided by CAS
algorithms, in order to get a result that is useful. This is an additional reason not to forget
your math skills and not to rely solely on the computer!
Not all differential equations can be easily solved in closed form and some cannot be
solved in closed form at all. In cases where closed-form solutions are not possible to obtain,
we can solve the differential equation using a numerical method. In the next section, we will
discuss a simple method of numerically solving ordinary differential equations.

2.7 NUMERICAL SOLUTIONS OF DIFFERENTIAL EQUATIONS

Sometimes differential equations are either very difficult to solve in closed form or have
no closed-form solution. In these cases, we rely on numerical algorithms. The truth is that
as a physicist, you will likely run into more differential equations that require numerical
solutions than those that can be solved by hand! Therefore, it is important to understand
how to solve a differential equation using numerical methods.
Consider an ordinary differential equation (ODE) of the following form:
dx
= f (x) (2.7.1)
dt
where f (x) is a well-behaved function of x. Note that here f (x) is an arbitrary function of
position, not the force acting on the system. What we are about to demonstrate works on
differential equations encountered in any field not just physics. To solve (2.7.1) numerically,
we need to approximate the derivative on the left-hand side. The approximation of the
derivative is done by removing the limit in Newton’s definition of the derivative:

dx x (t + dt) − x (t)
≈ . (2.7.2)
dt dt
If the value of dt is small enough, the right-hand side of the above equation will give a
good approximation of the derivative. We can then think of solving the differential equation
by finding the current value of x = x(t + dt) by using past values, x(t). This process is called
Euler’s method for solving differential equations. By inserting (2.7.2) into (2.7.1), we obtain
the formula for Euler’s method:

x(t + dt) = x(t) + dt f (x(t)). (2.7.3)

We can interpret (2.7.3) as using the tangent line at the current value of x, x(t), in order
to approximate the next value of x, x(t + dt). Over a short distance dt, we can expect this
local linear approximation to be a good method of finding x(t + dt) from x(t).
To implement Euler’s algorithm, follow these steps in your program of choice:
1. Define dt to be a value for the time interval small enough to produce the correct
solution.
2. Define an empty array x for the function, which is the solution to the ODE. Some
programming languages require you to define the length, or number of elements in
the array, when you define the array. If that is the case, define the length of x to be

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 49 — #65

i i

Single-Particle Motion in One Dimension 49

equal to tmax /dt (rounded up to the nearest integer), where tmax is the largest value
of t for which you want to know x(t).
3. Loop over an index i in order to enter values in the array using:

x[i] = x[i − 1] + dt ∗ f (x[i]) (2.7.4)

The ith element of the array x, denoted by x[i], contains the value of the function x at
t = idt. How do you know if the value of dt that you chose gives a good solution? It is
sometimes helpful to run your code with a first guess of dt, and then repeat for a smaller
value of dt. If the solution doesn’t change using the new value of dt, then that is a good sign
that you have a good value of dt and an accurate representation of the solution. Of course
it is not a proof, but the method can work well as you try to get a sense of the solution
to a differential equation. It is worth mentioning here that the errors in numerical methods
such as the Euler method are well known, and although we will not go into them here, they
are covered in typical numerical analysis texts such as [Press et al. (2007)].
You might wonder, why not just use a very small value, say, dt = 0.000001? One reason
is that if you did that, then to find the solution to t = 1, the loop would take 106 iterations!
However, a more important reason is that if dt is very small, then global truncation errors
become an issue for your solution. Local truncation error is the error that occurs at each
iteration of a numerical method. Global truncation error is the cumulative error caused by
many iterations. Hence, small values of dt both increase computing time and can introduce
more error. Furthermore, the Euler method can be unstable, meaning that the numerical
solution can grow very large when the exact solution does not. If you cannot obtain a
solution with reasonable accuracy using Euler’s method, then you will need to investigate
higher-order numerical solvers such as the four-order Runge-Kutta method, which we will
discuss in a later chapter. Even more sophisticated methods involve an adaptive value of
dt.
You might be wondering how to solve second-order ordinary differential equations
(ODE), which are more common in mechanics due to Newton’s second law. To solve a
second-order ODE, we break it up into two first-order ODEs and use the Euler method for
each one. Consider the ODE resulting from Newton’s second law in the previous example
of linear air resistance:

ma = − bv + mg (2.7.5)
d2 x dx
m =−b + mg (2.7.6)
dt2 dt
where we have written a = d2 x/dt2 and v = dx/dt. To create two first-order ODEs from
(2.7.6), we treat the velocity as a separate variable:
dx
=v (2.7.7)
dt
then (2.7.6) becomes:
dv b
=g− v (2.7.8)
dt m
The resulting equations for the Euler method are therefore:
x(t + dt) =x(t) + dt v(t)
(2.7.9)

k
v(t + dt) =v(t) + dt g − v(t)
m

i i

m, k , g = 1 . 0 , 0 . 2 , 9 . 8
dt = 0 . 0 1
tmax = 3 0 . 0
num steps = i n t ( tmax/ dt )

#d e f i n e f u n c t i o n s
d e f dxdt ( v e l ) :
return vel

d e f dvdt ( v e l ) :
r e t u r n g − k/m ∗ v e l

#d e f i n e i n i t i a l c o n d i t i o n s and a r r a y s
x = np . z e r o s ( num steps +1)
v = np . z e r o s ( num steps +1)

#perform the Euler step algorithm

for i in r a n g e ( 1 , num steps +1):
x[ i ] = x [ i −1] + dt ∗ dxdt ( v [ i −1])
v[ i ] = v [ i −1] + dt ∗ dvdt ( v [ i −1])

time = np . a r a n g e ( 0 , tmax+dt , dt )

f i g , axes = p l t . subplots (1 , 2 , f i g s i z e =(10 ,4))

a x e s [ 0 ] . p l o t ( time , x )
a x e s [ 0 ] . s e t x l a b e l ( ” time , s ” )
a x e s [ 0 ] . s e t y l a b e l ( ” x , m” )
a x e s [ 1 ] . p l o t ( time , v )
a x e s [ 1 ] . s e t x l a b e l ( ” time , s ” )
a x e s [ 1 ] . s e t y l a b e l ( ” v , m/ s ” )
f i g . s a v e f i g (” l i n e a r a i r r e s i s t a n c e . eps ”)

Algorithm 2: The solution to Example 2.5 using an Euler’s method algorithm.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 52 — #68

i i

52 Classical Mechanics: A Computational Approach

The output of Algorithm 2 is shown in Figure 2.5. Notice that the results obtained from
the Python program look similar to the results obtained from Mathematica. That is a good
sign! When we plot both results on the same graph, we find that the graphs are actually
identical. When developing your own numerical analysis programs, such as the Euler method
algorithm, it is often wise to check the results of your new algorithm with another algorithm
that you suspect works fine. In this case, the Mathematica DSolve command produced the
same analytical form as we obtained by hand. Therefore, we verified Mathematica’s result
with our own calculation and then verified our Euler method result (found using Python)
with that of Mathematica’s result. Ultimately, we can be confident that our Euler method
program works well, at least in the case of this one particular ODE. Furthermore, our Euler
method’s success here makes us more confident of the results we would obtain by using our
algorithm on new problems.

1200 50

1000 40
800 30
v, m/s
x, m

600
20
400
200 10
0 0
0 10 20 30 0 10 20 30
time, s time, s

Figure 2.5: The result of Algorithm 2

The ability to solve differential equations numerically allows you to address interesting
problems that you could not do by hand. The next example involves three different forces
acting on a system, it demonstrates the power of Euler’s method.

Example 2.8: Nonlinear forces

Consider a particle with mass m = 1 kg, which is experiencing a Hooke’s law restoring
force, F (x) = −kx, a quadratic drag, F (v) = −cv 2 , and a sinusoidal external driving force,
F (t) = A cos(ωt), where k = 3 N/m, c = 0.5 Ns2 /m, ω = 1.0 rad/s, and A = 1.0 N. The
resulting equation of motion can be found using Newton’s second law:

mẍ + kx + cẋ2 = A cos(ωt) (2.7.10)

Next, we insert the given values for each parameter to obtain:

ẍ + 3x + 0.5ẋ2 = cos(t) (2.7.11)

If the particle’s initial position and velocity is x(0) = 0.0 m and v(0) = 0.0 m/s, respec-
tively, find the particle’s position and velocity as a function of time.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 53 — #69

i i

Single-Particle Motion in One Dimension 53

Solution:
We will be using Euler’s method to solve the differential equation. In order to use
Euler’s method, we need to turn the second-order ODE into two coupled first-order ODEs.
This is done by using v = ẋ:

ẋ =v
(2.7.12)
v̇ = − 3x − 0.5v 2 + cos(t)
In the algorithm below, we removed the comments and graphing commands, but those
are similar to the previous problem involving linear air resistance. Furthermore, we added
the resulting graphs of x(t) and v(t) after the algorithm.

import numpy a s np

dt = 0 . 0 1
tmax = 4 0 . 0
num steps = i n t ( tmax/ dt )

d e f dxdt ( v e l ) :
return vel

d e f dvdt ( pos , v e l , t ) :
r e t u r n np . c o s ( t )−3∗ pos − 0 . 5 ∗ v e l ∗∗2

x = np . z e r o s ( num steps )
v = np . z e r o s ( num steps )

f o r i i n r a n g e ( 1 , num steps ) :
x [ i ] = x [ i −1] + dt ∗ dxdt ( v [ i −1])
v [ i ] = v [ i −1] + dt ∗ dvdt ( x [ i −1] , v [ i −1] , i ∗ dt )

time = np . l i n s p a c e ( 0 , tmax , num steps )

1.0
2
0.5

1
velocity, m/s
position, m

0.0

−0.5 0

−1.0
−1
−1.5

−2.0 −2

0 10 20 30 40 0 10 20 30 40
time, s time, s
Notice that we couldn’t pass the loop variable i into the function computing the cosine.
By itself, i is just an integer and does not represent time. The time at step i is idt.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 54 — #70

i i

54 Classical Mechanics: A Computational Approach

The position versus time graph is irregular. Notice that the particle position is not
continuously increasing or decreasing, nor is it periodic (it does not exactly repeat itself).
The differential equation governing this particle’s behavior is nonlinear, and nonlinear
systems can display interesting behavior. We’ll discuss nonlinear systems in more detail
in Chapter 13.

Example 2.8 can also be used to demonstrate the challenge of choosing a good value
for dt. By choosing dt = 0.01 in Example 2.8, we obtained a good solution for 0 ≤ t ≤ 40.
However, nonlinear equations can be tricky. If we choose dt = 0.02, the Euler method will
become unstable and the solution goes to infinity (try it!).

2.8 CHAPTER SUMMARY

Newton’s second law give the equation of motion for a particle. This equation of motion
is a differential equation and can be written in three forms:

d2 x dv dv
F =m =m = mv
dt2 dt dx

F = m dv
dt is a first-order ODE, because the highest derivative is a first derivative.

2
F = m ddt2x is a second-order ODE, because the highest derivative is a second derivative.

The number of arbitrary constants in the general solution of an ODE is equal to the
order of the ODE.

The method of separation of variables is performed by treating the derivative as a

fraction, and multiplying both sides of the equation by dt (or dx etc) and collecting terms
involving the dependent variable (e.g., x(t)) on one side and terms involving the indepen-
dent variable on the other.
When the acceleration a = constant, Newton’s second law gives the kinematic equations:
1
v(t) = v0 + at x(t) = x0 + v0 t + at2 v(x)2 = v02 + 2a (x − x0 )
2
When the force is a function of time only F (t), Newton’s second law can be integrated
using:
1 t
Z
v(t) = v0 + F (t0 ) dt0
m t0 =0
When the force is a function of speed only F (v), Newton’s second law can be integrated
using: Z v
dv 0
t − t0 = m
v 0 =0 F (v )
0

The magnitude of the air resistance f (v) has, in general, both a linear and a quadratic
component in v.
f (v) = bv + cv 2
We find the terminal velocity for a falling object by setting the air resistance force to
be equal to the weight of the object, f (v) = bv + cv 2 = mg.

(b) How long will it take her to fall if the parachute opens? Assume that the parachute
opens immediately and that the drag force on Sarah with the parachute open is
described by,
1
f = − cd ρAv 2 (2.9.3)
2
where the drag coefficient, cd = 1.5 for the parachute, ρ = 1.22 kg/m3 is the
density of air, and A = 46 m2 is the area of the parachute.

25. Consider a falling object whose mass is 1 kg and experiences an air resistance force
of F (v) = −c |v| v, where c = 0.5 Ns2 /m2 . Assuming the falling object starts at rest,
use a computer to calculate the time it takes the object to fall a distance of 1 km.
Does the velocity of the falling object ever become constant? In other words, is there
a terminal velocity for this force?
Section 2.6: Position-Dependent Forces
26. A particle has a velocity v(x) = ax−n . Find the force acting on the particle. If the
particle starts at rest at the origin, find its position and velocity as functions of time.
27. Find the velocity v(x) and position x(t) of a particle starting at rest at the origin
experiencing the force F = a + bx with a, b > 0.
28. A particle experiences a force causing it to move in the +x-direction with velocity,
v(x) = cxe−bx . Find the work done by the force if the particle starts at the origin and
travels to infinity (note: in this case, infinity is a useful mathematical tool meaning
“very far”). The work is defined by:
Z x
W= F · dx
0

29. Compute the value of the acceleration due to gravity as a function of an object’s
height y above the Earth’s surface. Plot your result from y = 0 to 25 km. Hint: Use
Newton’s universal law of gravitation:
Gm1 m2
Fg = − (2.9.4)
r2
where G = 1.67 × 10−11 Nm2 /kg2 is the universal gravitation constant, m1 and m2 are
the masses of the interacting bodies, and r is the distance between the centers of m1
and m2 . Assume that the falling object is a point particle and ignore air resistance.
30. As found in Problem 29, the acceleration of a falling object is not constant with height.
It turns out that the acceleration varies as:
9.8 m/s2
g= 2 (2.9.5)
1 + Rye

where Re = 6.37 × 106 m is the radius of the Earth, y is the altitude of the object
above the Earth’s surface, and g is in m/s2 . Compute the time it takes for a m =1
kg object with a diameter D =1 m to fall from an altitude of 25 km and land on the
Earth’s surface, assuming quadratic air resistance. What is the object’s final velocity?
You will need a computer to solve this problem.

i i

Motion in Two and Three

Dimensions

In the previous chapter, the kinematics and dynamics aspects of particle motion were exam-
ined in one dimension by using the scalar position parameter x. In this chapter, we will first
describe the motion of particles in two and three dimensions using Cartesian coordinates
(x, y, z). Describing motion in two and three dimensions requires the use of vectors. After
reviewing the basic properties of vectors and introducing the dot and cross products, we
will introduce vector derivatives. Vector derivatives are necessary in order to compute the
velocity and acceleration vectors. We then describe polar, cylindrical, and spherical coordi-
nates, and demonstrate how to describe a particle’s position, velocity, and acceleration in
those coordinate systems. Finally, we conclude the chapter with special vector derivatives
commonly used in all fields of physics, such as the gradient, the divergence, and the curl.

3.1 POSITION, VELOCITY, AND ACCELERATION IN CARTESIAN COORDI-

NATE SYSTEMS
In order to describe a particle’s motion in two and three dimensions, we will need to use
vectors. In this section the main properties of vectors are reviewed, and they are applied to
the description of position, velocity, and acceleration of a particle in three dimensions. In
Chapter 1, we introduced the concepts of position, velocity, and acceleration using Cartesian
coordinates. In this section, we will expand on that discussion. The expanded discussion of
the basic descriptors of motion in Cartesian coordinates will prepare us to describe those
quantities in other coordinate systems.
A point P1 in three-dimensional space can be represented by a vector A from the origin
(0, 0, 0) to point P1 as shown in Figure 3.1. Such a vector A has three components along
the x-, y-, and z-axes, and can be written in the form A = [Ax , Ay , Az ] or in Cartesian
coordinate form as:
A = Ax î + Ay ĵ + Az k̂ (3.1.1)
where î, ĵ, k̂ are the unit vectors along the x, y, and z axis, respectively.
The magnitude A of the vector A is found from the Pythagorean theorem:
q
A = |A| = A2x + A2y + A2z (3.1.2)

The magnitude of a vector is often thought of as its length and would correspond to a
quantity measured from the vector. For example, suppose A represents the position of

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 62 — #78

i i

62 Classical Mechanics: A Computational Approach

a particle located at the point A = 3 mî + 4 mĵ. The magnitude of A is |A| = 5 m and
represents the distance between the particle and the origin.
It is often useful to know the direction in which a vector points. This is easily found
for simple vectors like A = 3î, which points along the x-axis. However, other vectors like
A = aî + bĵ do not lie along either the x- or y-axes. In that case, we need a unit vector which
points in the same direction as the vector A but has a length of one. We can obtain a unit
vector Â in the direction of any vector A, by dividing A by its magnitude:
A A
Â = =q (3.1.3)
|A| A2x + A2y + A2z

Example 3.1: Finding unit vectors.

Find a unit vector along the direction of the vector a=[1, 3, −1].
Solution:
We write the vector as a = ax î + ax ĵ + ax k̂ = 1î + 3ĵ − 1k̂. In order to find the unit
vector, divide the vector with its magnitude:

a 1î + 3ĵ − 1k̂ 1 3 1

â = q =√ = √ î + √ ĵ − √ k̂
a2x + a2y + a2z 1 +3 +1
2 2 2 11 11 11

If we know the coordinates of two points P1 and P2 in space, then we can find the vector
connecting these two points by subtracting their respective x-, y-, z-components. For exam-
ple, in Figure 3.1, let r1 be the vector from the origin to P1 , and r2 the vector from the
origin to P2 . The vector r12 from P1 to P2 , is given by (see Figure 3.1):

Vector r12 Connecting Two Points in Space

r12 = r2 − r1 (3.1.4)

Now that we have covered some of the basic concepts behind vectors, we can apply
them to the basic descriptors of motion. As discussed in Chapter 1, the position vector
function r(t) = (x(t), y(t), z(t)) for a particle in three dimensions as described by Cartesian
coordinates is written as:
Position Vector

r(t) = x(t)î + y(t)ĵ + z(t)k̂ (3.1.5)

This is a vector-valued function, and x(t), y(t), z(t) are the scalar functions representing
the position along the x-, y-, and z-axes at time t, respectively. The vector r(t) has its tail
at the origin, and its head at the position of the particle at time t. The magnitude of the
position vector is equal to the distance between the particle and the origin.

Velocity Vector
dr dx dy dz
v(t) = = î + ĵ + k̂ (3.1.11)
dt dt dt dt

Because the velocity vector is the derivative of the position vector, the velocity vector
is going to be always tangent to the path followed by the particle.

Figure 3.2: The vector function r is illustrated here at times t and t + ∆t. In the limit
∆t → 0, the vector ∆r/∆t becomes the derivative vector function dr/dt, which is tangent
to the curve.

The magnitude v(t) of the velocity vector v(t) is called the speed:
s
2 2 2
dx dy dz
v(t) = |v(t)| = + + (3.1.12)
dt dt dt

Equation (3.1.12) can also be written in terms of the arc length s along the curve C describ-
ing the motion, as measured from some initial position along the curve. In order to use the
arc length, we need to define an infinitesimal displacement along the curve,

ds = dxî + dy ĵ + dz k̂ (3.1.13)

The differential distance ds is the magnitude of the infinitesimal displacement and is given
by q
ds = (dx)2 + (dy)2 + (dz)2 (3.1.14)
Equation (3.1.12) can now be written in terms of the arc length s, as:

ds
v(t) = |v(t)| = (3.1.15)
dt
A unit tangent vector T along the curve C describing the motion is obtained by dividing
the velocity vector v(t) by its magnitude v:

v(t)
T= (3.1.16)
v(t)

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 65 — #81

i i

Motion in Two and Three Dimensions 65

Note that the unit tangent vector T = v̂ points along the direction of the velocity and is
always tangent to the curve C. An example of how to calculate the unit tangent vector is
given in Example 3.2.
Similarly, the derivative of the velocity vector dv/dt is the acceleration vector function,
which in Cartesian coordinates is:
Acceleration Vector
dv d2 x dy 2 d2 z
a(t) = = 2 î + 2 ĵ + 2 k̂ (3.1.17)
dt dt dt dt

Example 3.2: Finding unit tangent vectors.

The position of a particle along a curve C in 3D space is given by the parametric
equations x = sin(2.0t), y = t2 , z = 5.0, where x, y, and z are measured in meters. (a)
Find the magnitude of the velocity and acceleration at time t = 1.00 s. (b) Find a general
expression for the unit vector T that is tangent to the curve C at time t.
Solution:
(a) The magnitude of the velocity vector is the speed v:
s
2 2 2 q
dx dy dz
v(t) = + + = [2.0 cos(2.0t)]2 + (2t)2
dt dt dt

The magnitude of the acceleration vector is:

s
2 2 2 2 2 2 q
d x d y d z
a(t) = + + = [−4.0 sin(2.0t)]2 + (2)2
dt2 dt2 dt2

At time t = 1.00 s these magnitudes are:

q
v(1) = [2.0 cos(2.0)]2 + (2)2 = 2.2 m/s
q
a(1) = [4.0 sin(2.0)]2 + (2)2 = 4.2 m/s2

(b) The unit vector T, which is tangent to the curve C at any time t, is found from:

1

v(t) dx(t) dy(t) dz(t)
T= = î + ĵ + k̂
v(t) v(t) dt dt dt

By evaluating the derivatives and substituting the speed v(t) from part (a):
1 1
T= [2 cos(2t)] î + 2tĵ = q [2 cos(2t)] î + 2tĵ
v(t)
[2 cos(2t)]2 + (2t)2

The next example shows how to calculate the position vector when one is given an
acceleration vector. As you might expect, the problem involves integrating the acceleration.
Example 3.3: Finding position and velocity from acceleration vector.
The acceleration of a particle moving in two dimensional space is given by the vector:

a(t) = sin (2.0t) î + cos (2.0t) ĵ

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 66 — #82

i i

66 Classical Mechanics: A Computational Approach

where the vector a is measured in m/s2 . Find the velocity vector v(t) and the position
vector r(t). Assume that the particle is stationary and is at the origin at time t = 0.
Solution:
Since a(t) = dv(t)/dt, we can find the velocity and position by integration. Just like dif-
ferentiating a vector, integrating a vector can be done one component at a time:
Z
v(t) = a(t)dt
Z
= sin (2t) î + cos (2t) ĵ dt
Z Z
= sin (2t) îdt + cos (2t) ĵdt

cos (2t) sin (2t)

=− î + ĵ + c
2 2
where c is a constant vector of integration, which can be found from the initial conditions
at time t = 0. Since the particle is stationary at t = 0, we must have v(0) = 0, i.e.,

cos (0) sin (0)

v(0) = − î + ĵ + c = 0
2 2

which gives c = 21 m/s î.

The position is found by integrating the velocity vector:

cos (2t) sin (2t) 1

Z Z
r(t) = v(t)dt = − î + ĵ + î dt
2 2 2
1 1 1
= − sin (2t) î − cos (2t) ĵ + tî + d
4 4 2
where d is a second constant vector of integration, which can be found from the initial
position condition r(0) = 0. Substituting r(0) = 0 we find
1 1 1
r(0) = − sin (0) î − cos (0) ĵ + 0î + d = 0
4 4 2

which gives d = − 41 m ĵ. Again, we have used the fact that the unit vectors î and ĵ in

Cartesian coordinates are constant, and therefore can be placed outside of each integral.

The three vector functions r(t), v(t), and a(t) depend on the single parameter t, rep-
resenting time. If we are given the explicit functions x(t), y(t), z(t) then the path of the
moving particle can be plotted in a so-called parametric plot in three dimensions using a
computer, see Examples 3.4 and 3.5 that follow. A parametric plot uses the functions x(t),
y(t), and z(t) to graph a curve that represents the path taken by a particle. Each point in
the curve corresponds to the x-, y-, and z-coordinates for various times t.
In the case of two-dimensional motion, if we are given explicit functions x(t) and y(t), one
may be able to eliminate the variable t from the simultaneous equations x = x(t), y = y(t).
If one of these equations can be solved for t, the expression obtained can be substituted
into the other equation to obtain an expression for y(x). In some cases, there is no single
equation in closed form that is equivalent to the parametric equations. A similar process
can be used in three dimensions in order to get z(x, y). Examples 3.4 and 3.5 show how to
create parametric plots in two and three dimensions in Mathematica and Python.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 67 — #83

i i

Motion in Two and Three Dimensions 67

Example 3.4: Parametric equations for motion of particle in 2D.

Use the appropriate parametric plot commands in Python and in Mathematica to
plot the motion of a particle in two dimensions when the (x, y) position is given by the
parametric equations:
x = 2 cos(t) y = sin(t).
where the parameter t is in the range from 0 to 2π.
Solution:
The time parameter t can be eliminated from the x(t) and y(t) equations using cos(t) =
x/2 and sin(t) = y to obtain:
x 2 y 2
cos2 t + sin2 t = + =1
2 1
This equation is of the form
x 2 y 2
=1 +
a b
which is the equation of an ellipse on the xy-plane, with center at (0, 0) and with semi-
major axis a = 2 and semi-minor axis b = 1. The following Python code creates a parametric
plot in two dimensions of this equation.

from sympy import ∗

from sympy . p l o t t i n g import ∗
t = symbols ( ’ t ’ )
f i g = symbols ( ’ f i g ’ )
f i g , p l o t p a r a m e t r i c (2∗ cos ( t ) , s i n ( t ) , ( t , 0 , 2 ∗ pi ) )

The equivalent code in Mathematica is shown below.

ParametricPlot[{2Cos[t], Sin[t]}, {t, 0, 2Pi}]
Both the Python code the Mathematica code produce the same result. The result of
the Mathematica code is shown below.

y
1.2

0.6
Out[ ]=

x
-2.4 -1.2 1.2 2.4
-0.6

-1.2

Example 3.5: Parametric equations for motion of particle in 3D.

Plot the motion of particle in three dimensions when the (x, y, z) positions are given
by the parametric equations:
x = d + a cos(t) y = e + b sin(t) z = ct

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 68 — #84

i i

68 Classical Mechanics: A Computational Approach

where the parameter t is in the range 0 to T .

Solution:
The time parameter t can be eliminated from the x(t), y(t) equations by solving the
equations for cos(t) and sin(t) and substituting in the identity cos2 (t) + sin2 (t) = 1:
2 2
x−d y−e
+ = cos2 (t) + sin2 (t) = 1
a b

This is the equation of an ellipse on the xy-plane with center at (d, e). Since z = ct, one
then expects the motion to be in the form of an elliptical helix moving along the z-axis
with a uniform speed c.
Next we present Python and Mathematica codes for plotting the motion, with the
numerical values a = 2, b = 4, c = 1, d = 1, e = 2, T = 12. The Python code is below.

import m a t p l o t l i b a s mpl
from m p l t o o l k i t s . mplot3d import Axes3D
import numpy a s np
import m a t p l o t l i b . p y p l o t a s p l t

mpl . rcParams [ ’ l e g e n d . f o n t s i z e ’ ] = 10

fig = plt . figure ()

ax = f i g . gca ( p r o j e c t i o n= ’ 3d ’ )
t h e t a = np . l i n s p a c e (−2∗np . pi , 2 ∗ np . pi , 1 0 0 )
z = np . l i n s p a c e ( 0 , 1 2 , 1 0 0 )
r = 1
x = (1+2∗ np . c o s ( t h e t a ) )
y = (2+4 ∗ np . s i n ( t h e t a ) )
ax . p l o t ( x , y , z , l a b e l= ’ p a r a m e t r i c c u r v e ’ )
ax . l e g e n d ( )

p l t . show ( )

The equivalent Mathematica code is below.

ParametricPlot3D[{1 + 2 ∗ Cos[t], 2 + 4 ∗ Sin[t], t}, {t, 0, 12}]
The programs produce similar results, the Mathematica output is shown below.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 69 — #85

i i

Motion in Two and Three Dimensions 69

3.2 VECTOR PRODUCTS

There are two methods of multiplying vectors together, the dot product and the cross
product. Both methods are very useful in physics.

3.2.1 The Dot Product

The dot product is an algebraic operation that describes a mathematical operation A · B
between two vectors A and B. The alternative name scalar product emphasizes the scalar
(rather than vector) nature of the result. The dot product of two vectors A = [Ax , Ay , Az ]
and B = [Bx , By , Bz ] is defined as:

A · B = Ax Bx + Ay By + Az Bz (3.2.1)

An alternative definition of the dot product is:

A · B = AB cos θ (3.2.2)

where A and B are the magnitude of the vectors A and B, respectively, and θ is the angle
between the two vectors as shown in Figure 3.3. Notice that the dot product is multiplying
the magnitude of B by the component of A, which is parallel to B. A physical example
involving the dot product is the definition of work. The work done by a constant force can be
computed as W = F · r, where a force F is applied to a particle along a displacement r. The
component of the force doing work on the particle is parallel to the particle’s displacement.
The dot product is used to compute the work because we need the component of F parallel
to r.
Equation (3.2.2) can be used to obtain the angle between two vectors:
A·B Ax Bx + Ay By + Az Bz
cos θ = = q q . (3.2.3)
AB A2x + A2y + A2z Bx2 + By2 + Bz2

The dot product can also be interpreted geometrically as the scalar projection of vector
A on the direction of vector B, as shown in Figure 3.3. This is useful if you want to know
the component of the vector A along the direction of the vector B. First, we create the
unit vector B̂ =B/ |B| pointing in the direction of B. The scalar projection is equal to
AB = A · B̂ = |A| cos θ. The vector projection of A onto B is, projB A = AB B̂.

Figure 3.3: The dot product represents the product of the scalar projection of vector A on
the unit vector B̂ in the direction of vector B: A · B̂ = |A| cos θ.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 70 — #86

i i

70 Classical Mechanics: A Computational Approach

There are several algebraic relationships for the dot products of two vectors A and B,
and most of them are intuitive. The following properties hold if A, B, and C are real vectors
and c is a scalar.
The dot product is commutative:

A·B = B·A (3.2.4)

The dot product is distributive over vector addition:

A · (B + C) = A · B + A · C (3.2.5)

Likewise, with a scalar c we can show:

A · (cB) = c (A · B) (3.2.6)

Two nonzero vectors A and B are orthogonal if and only if

A·B = 0 ←→ A⊥B (3.2.7)

The geometric interpretation of the orthogonality of vectors is that they are perpendicular,
i.e., θ = π/2. Finally, the product rule applies to the dot product:
d dA dB
(A · B) = ·B+A· (3.2.8)
dt dt dt

Example 3.6: Calculating the dot product.

Find the dot product of the vectors A = 3î + 4ĵ − k̂ and B = î − ĵ − k̂ analytically, and
also by using appropriate codes in Python and Mathematica.
Solution:
The scalar dot product is found by multiplying and adding the coefficients of the unit
vectors î, ĵ, k̂,
A · B = (3î + 4ĵ − k̂) · (î − ĵ − k̂)
= (3)(1) + (4)(−1) + (−1)(−1)
=0
The following are the Python and Mathematica codes for the dot product.
In the Python code, the command ReferenceFrame defines the reference frame N , and
the unit vectors in this reference frame are denoted by N.x, N.y, N.z so that for example,
the vector 3î + 4ĵ − k̂ is written in the code as 3 ∗ N.x + 4 ∗ N.y − 1 ∗ N.z. The dot(a,b)
command is used to evaluate the dot product of the two vectors.

from sympy . p h y s i c s . v e c t o r import ∗

N = ReferenceFrame ( ’N ’ )
print ( dot ( 3 ∗N. x+4∗N. y−1∗N. z , 1∗N. x−1∗N. y−1∗N. z ) )

OUTPUT: 0

The Mathematica code that follows uses the command Dot[A, B], and the vectors
A, B are entered in the form of lists of their Cartesian components.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 71 — #87

i i

Motion in Two and Three Dimensions 71

Dot[{3, 4, −1}, {1, −1, −1}]

OUTPUT: 0

Example 3.7: Finding unit vectors using dot products.

The position of a particle along a curve C in three-dimensional space is given by the
parametric equations x = t2 , y = t, z = 5, where quantities are in SI units.
(a) Find a unit vector T tangent to the curve C at any time t.
(b) Show that the vector dT /dt is perpendicular to the unit vector T at any time t.
(c) Find a unit vector N perpendicular to the unit vector T at time t = 1 s.
Solution:
(a) The velocity vector is:

dx(t) dy(t) dz(t)

v(t) = î + ĵ + k̂ = (2t) î + (1) ĵ = 2tî + ĵ
dt dt dt
and the unit vector T tangent to the curve C at any time t is found as in Example 3.2:

v(t) 2tî + ĵ 2tî + ĵ

T= =q =√
v(t) 2
(2t) + (1) 2 4t2 + 1

√
At t = 1 s we find the unit tangent vector T = 2î + ĵ / 5.
(b) Since T is a unit vector with a magnitude of 1, the dot product |T|2 = T · T = 1.
By differentiating |T|2 with respect to time t, we obtain:

dT dT dT
T + T = 2T =0
dt dt dt
Therefore the vector dT/dt is perpendicular to the tangent unit vector T at any time
t, and is found from: " #
dT d 2tî + ĵ (2) î + (−4t) ĵ
= √ =
dt dt 4t + 1
2
(4t2 + 1)
3/2

vu !2 !2 √
dT u 2 −4 20
= + =
t
dt (5)3/2 (5)3/2 (5)3/2

The desired unit vector N in this perpendicular direction is obtained by dividing dT/dt
by its magnitude:
2î−4ĵ
dT
(5)3/2 2î − 4ĵ
N= dt
= √ = √
| dT
dt |
20 20
(5)3/2

3.2.2 The Cross Product

The cross product or vector product is an operation on two vectors in three-dimensional
space. The cross product of two vectors A and B is denoted by A × B, and is defined as

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 72 — #88

i i

72 Classical Mechanics: A Computational Approach

a vector C that is perpendicular to both vectors A and B as shown in Figure 3.4(a). The
direction of the cross product vector A × B is given by the right-hand rule shown in Figure
3.4(b). The magnitude of this vector is equal to the area of the parallelogram formed by
vectors A and B, as shown in Figure 3.4(c).

Figure 3.4: (a) The cross product in respect to a right-handed coordinate system; (b) Finding
the direction of the cross product by the right-hand rule; (c) The area of a parallelogram
formed by two vectors A and B, is equal to the magnitude of the cross product of these
two vectors.

The magnitude of the cross product is defined by the formula

a × b = ab sin θ n̂ (3.2.9)
where θ is the angle between a and b (0 ≤ θ ≤ π), a = |a| and b = |b| are the magnitudes of
vectors a and b, and n̂ is a unit vector perpendicular to the plane containing a and b in the
direction given by the right-hand rule as in Figure 3.4(a). If the vectors a and b are parallel
(i.e., the angle θ between them is either 0 radians or π radians), then the cross product of
a and b is the zero vector 0.
The magnitude of the cross product is the product of the magnitude of the vector a,
and the component of b perpendicular to a. You may recall from your introductory physics
course that you can compute the torque applied to a door by using the formula N = r × F,
where N is the torque applied by the force F, and r is the vector which points from the
axis of rotation to the point where the force is applied. For a door, the vector r lies along
the door, pointing from the hinges to the edge of the door, where the force to open it is
applied. We know that the component of the force responsible for the rotation of the door
is perpendicular to the door, and that the component of the force parallel to the door is not
causing the rotation. Therefore, the cross product is used to compute the torque, because
we need the component of the force perpendicular to r.
The cross product of two vectors a = [a1 , a2 , a3 ] and b = [b1 , b2 , b3 ] can also be repre-
sented by the determinant of a formal matrix:

î ĵ k̂
a × b = a1 a2 a3 (3.2.10)

b1 b2 b3

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 73 — #89

i i

Motion in Two and Three Dimensions 73

This determinant can be expanded by using cofactor expansion along the first row to yield:

a a3 a1 a3
a × b = 2 ĵ + a1 a2 k̂ (3.2.11)

î −
b2 b3 b1 b3 b1 b2
which gives the components of the resulting cross product vector directly:

a × b = (a2 b3 − a3 b2 )î + (a3 b1 − a1 b3 )ĵ + (a1 b2 − a2 b1 )k̂ (3.2.12)

Like the dot product, the cross product has some familiar algebraic identities. The cross
product is distributive with addition:

a × (b + c) = a × b + a × c (3.2.13)

Likewise, with a scalar c, we can show:

(ca) × b = a × (cb) = c (a × b) (3.2.14)

Unlike the dot product, the cross-product is anticommutative, i.e.,

b×a = −(a × b) (3.2.15)

This property is easily concluded by considering the right-hand rule; when the two vectors
are interchanged, the sign of the cross-product vector is reversed. The product rule for
derivatives also applies to the cross product:
d da db
(a × b) = ×b+a× (3.2.16)
dx dx dx

Example 3.8: Example of cross product with Python and Mathematica.

Find the cross product of the vectors A = 3î + 4ĵ − k̂ and B = î − ĵ − k analytically,
and also by using appropriate codes in Python and Mathematica.
Solution:
The cross product is found by using the determinant:

î ĵ k̂
4 −1 3 −1 3 4

A × B = 3 4 −1 = ĵ + k̂ = −5î + 2ĵ − 7k̂

î −
−1 −1 1 −1 1 −1

1 −1 −1

The following are the Python and Mathematica codes for the cross product.

from sympy . p h y s i c s . v e c t o r import ∗

N = ReferenceFrame ( ’N ’ )
print ( c r o s s ( 3 ∗N. x+4∗N. y−1∗N. z , 1∗N. x−1∗N. y−1∗N. z ) )

OUTPUT:
− 5∗N. x + 2∗N. y − 7∗N. z

Cross[{3, 4, −1}, {1, −1, −1}]

OUTPUT: {−5, 2, −7}

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 74 — #90

i i

74 Classical Mechanics: A Computational Approach

Example 3.9: Finding unit vectors using cross products.

Find a unit vector perpendicular to both vectors a = [1, 3, −1] and b = [4, −2, 1].
Solution:
We first calculate the cross product of the two vectors since we know that it will be
perpendicular to both vectors. The cross product is:

î ĵ k̂
3 −1 1 −1 1 3

c = a × b = 1 3 −1 = ĵ + k̂ = 1î − 5ĵ − 14k̂.

î −
−2 1 4 1 4 −2

4 −2 1

To find the unit vector, divide this vector by its magnitude. The magnitude of a × b
is: q √
|a × b| = 12 + 52 + (−14)2 = 222
The desired unit vector is

c 1î − 5ĵ − 14k̂ 1 5 14

ĉ = q =√ =√ î − √ ĵ − √ k̂
c2x + c2y + c2z 1 +3 +1
2 2 2 222 222 222

In the Mathematica code below we use the Cross command to find the cross product,
and we use the Norm[c] command to find the magnitude of the given vector.
a = {1, 3, −1};
b = {4, −2, 1};

c = Cross[a, b]
OUTPUT: {1, −5, −14}

Norm[c] √
OUTPUT: 222

unitc = c/Norm[c]
n q o
OUTPUT: √ 1 , − √ 5 , −7
222 222
2
111

A useful mathematical construction involving the cross product is the triple product,
which is used in two different forms.
The scalar triple product of three vectors is defined as
a · (b × c) (3.2.17)
Notice that this expression means that we first calculate the cross product of vectors b
and c, followed by the dot product between the vectors a and b × c. The magnitude of the
triple scalar product represents the volume V of the parallelepiped with edges a, b, and
c. Since these three vectors a, b, c can be interchanged, the following three scalar triple
products are equal to each other and to the volume V :
a · (b × c) = b · (c × a) = c · (a × b) = V (3.2.18)
The vector triple product is the cross product of a vector with the result of another cross
product and is related to the dot product by the following identity:
a × (b × c) = b(a · c) − c(a · b) (3.2.19)

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 75 — #91

i i

Motion in Two and Three Dimensions 75

The mnemonic “BAC minus CAB” is often used to remember the order of the vectors
in the right-hand side of this equation. This formula is used in physics to simplify vector
calculations. Another useful identity relates the cross product to the scalar triple product:

(a × b) × (a × c) = (a · (b × c))a (3.2.20)

Example 3.10: Evaluation of triple products using Mathematica.

Use Mathematica to prove the triple product identities in (3.2.18) and (3.2.19) for any
vectors a, b, and c.
Solution:
By combining the Dot and Cross commands, the identities are verified in the code
below. The two sides of the identities are set equal by using the Boolean equality sign ==
and Mathematica returns True for the logical statements.
a = {ax, ay, az};
b = {bx, by, bz};
c = {cx, cy, cx};

Dot[a, Cross[b, c]] == Dot[b, Cross[c, a]]

OUTPUT: ay(−bxcx + bzcx) + az(−bycx + bxcy) + ax(bycx − bzcy) == by(axcx −

azcx) + bz(aycx − axcy) + bx(−aycx + azcy)

Dot[a, Cross[b, c]] == Dot[b, Cross[c, a]]//Simplify

OUTPUT: True

Cross[a, Cross[b, c]] == bDot[a, c] − cDot[a, b]//Simplify

OUTPUT: True

3.3 POSITION, VELOCITY, AND ACCELERATION IN NON-CARTESIAN

COORDINATE SYSTEMS
In this section, we will discuss how to describe a particle’s motion in coordinate systems
other than Cartesian coordinates. The new coordinate systems often exploit symmetries
in the problem, making the non-Cartesian coordinate systems a more natural means of
describing the motion of the particle. For example, the motion of a particle constrained to
move along a circle can be more easily described in polar coordinates than it is in Cartesian
coordinates.

3.3.1 Polar Coordinates

Figure 3.5 shows the position (x,y) of a particle located at the point P on the xy-plane.
Using polar coordinates, the position can also be described by the angle θ with the x-axis
and the radius r representing the distance from the origin.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 76 — #92

i i

76 Classical Mechanics: A Computational Approach

r̂

P
rsin r

rcos X
Figure 3.5: The radius r and angle θ locate a particle in polar coordinates, and they also
represent a complex number: z = x + iy = r(cos θ + i sin θ) = r eiθ . The two unit vectors r̂
and θ̂θ (in blue in the e-book) in polar coordinates change direction as the particle moves in
the xy plane along a curved path (in red in the e-book).

The relationship between Cartesian (x,y) coordinates and the polar coordinates r, θ is:

From Polar to Cartesian Coordinates

x = r cos θ (3.3.1)

y = r sin θ (3.3.2)

From Cartesian to Polar Coordinates

r = x2 + y 2 (3.3.3)
p

θ = tan−1 (y/x) (3.3.4)

We can also write the position vector as a complex number in the following form by
using Euler’s equations:

z = x + iy = r(cos θ + i sin θ) = reiθ (3.3.5)

and the complex conjugate of this complex number is:

z̄ = r(cos θ − i sin θ) = re−iθ (3.3.6)

By adding and subtracting the last two equations, we obtain Euler’s formulas for sine
and cosine functions. These equations provide a powerful connection between analysis and
trigonometry and should be memorized.

Euler’s Identities
eiθ + e−iθ
cos θ = (3.3.7)
2

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 77 — #93

i i

Motion in Two and Three Dimensions 77

eiθ − e−iθ
sin θ = (3.3.8)
2i
eiθ = cos θ + i sin θ (3.3.9)
e−iθ = cos θ − i sin θ (3.3.10)

While polar coordinates can be convenient for describing the position of a particle,
especially a particle constrained to move along an arc or a circle (red in the e-book in
Figure 3.5), the unit vectors require special attention. In Cartesian coordinates, the unit
vectors î and ĵ always point along the x- and y-axes respectively, regardless of the particle’s
location. However, the directions of the unit polar vectors r̂ and θ̂ change with the location
of the particle. This can be clearly seen in Figure 3.5.
The unit vector r̂ always points in the radially outward direction. The position vector
of the particle moving in the xy-plane can be written in terms of the unit vector r̂ as:

r = rr̂ (3.3.11)

In order to find the velocity and acceleration in polar coordinates, we first need to write
the unit vectors r̂ and θ̂θ in terms of î and ĵ. Using the geometry shown in Figure 3.5, we
can write:

r̂ = cos θ î + sin θ ĵ (3.3.12)

θ̂θ = − sin θ î + θ̂ cos θ ĵ (3.3.13)

where as expected θ̂ = |r̂| = 1, and the two unit vectors are perpendicular to each other,

i.e., θ̂·r̂ = 0.
The unit vectors r̂ and θ̂ vary in direction with time, and we can calculate their time
derivatives by using the chain rule:

d d h i dθ dθ dθ dθ
r̂ = cos θî + sin θĵ = − sin θ î + cos θ ĵ = − sin θî + cos θ ĵ (3.3.14)
dt dt dt dt dt dt

By combining the previous two equations we obtain:

dr̂ dθ
= θ̂θ (3.3.15)
dt dt

By working in a similar manner we obtain the time derivative of the unit vector θ̂:

dθ̂θ dθ
= − r̂ (3.3.16)
dt dt
We can now calculate the velocity vector in polar coordinates by taking the first time
derivative of the position vector r:

dr d dr dr̂
v= = (rr̂) = r̂ + r (3.3.17)
dt dt dt dt
By substituting (3.3.15) into this equation, we obtain the velocity vector in polar coor-
dinates:
dr dr dθ
v= = r̂ + r θ̂θ (3.3.18)
dt dt dt

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 78 — #94

i i

78 Classical Mechanics: A Computational Approach

The two components of the velocity in this equation have distinct physical meanings.
dt r̂ represents the radial component of the velocity, which points along
The first term vr = dr
the radius r. The second term v θ = r dθ dt θ̂ = rω θ̂ is proportional to the angular velocity
θ θ
ω = dθ/dt, and represents the velocity of the particle along the θ̂θ direction. This second
component represents the velocity of uniform circular motion.
From (3.3.18), we obtain the infinitesimal displacement ds between two points in space
in polar coordinates by canceling the differential time dt from both sides of this equation:

ds = drr̂ + rdθθ̂θ (3.3.19)

Based on this expression, the infinitesimal area element dA in polar coordinates can then
be written as:
dA = dxdy = rdθdr (3.3.20)
and we can calculate double integrals of any function f (x, y) in polar coordinates by using
the transformation:
Z Z Z Z
f (x, y) dxdy = f (r cos θ, r sin θ) r dr dθ (3.3.21)

In summary, the unit vectors in polar coordinates and their derivatives are given by:

Unit Vectors in Polar Coordinates

θ̂θ = − sin θî + cos θĵ (3.3.22)

r̂ = cos θî + sin θĵ (3.3.23)

d dθ d dθ
r̂ = θ̂θ θ̂θ = − r̂ (3.3.24)
dt dt dt dt

The following example shows how to create polar plots in Python and in Mathematica.

Example 3.11: Polar plots using Python and Mathematica.

Create a polar plot of the equation: θ = 2πr .
Solution:
The following are the Python and Mathematica codes for creating the polar plot. As
the angle θ increases from 0 to 4π , the radius increases from 0 to 2.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 79 — #95

i i

Motion in Two and Three Dimensions 79

import numpy a s np
import m a t p l o t l i b . p y p l o t a s p l t

r = np . a r a n g e ( 0 , 2 , 0 . 0 1 )
t h e t a = 2 ∗ np . p i ∗ r

ax = p l t . s u b p l o t ( 1 1 1 , p r o j e c t i o n= ’ p o l a r ’ )
ax . p l o t ( t h e t a , r )
ax . s e t r m a x ( 2 )
ax . s e t r t i c k s ( [ 0 . 5 , 1 , 1 . 5 , 2 ] )
ax . g r i d ( True )
ax . s e t t i t l e ( ” Example o f a p o l a r p l o t i n Python ” , va= ’ bottom ’ )
p l t . s a v e f i g ( ’ p o l a r p l o t e x a m p l e . png ’ , d p i =600)
p l t . show ( )

PolarPlot[θ/2π, {θ, 0, 4π}]

The Mathematica and Python codes have similar outputs. The output for the Python
code is shown below.

Polar coordinates are very useful in describing the motion of celestial bodies around
each other, as we will explore in detail in a later chapter. The following example shows
how polar coordinates can simplify the mathematical description of the elliptical motion of
planets around the Sun.
Example 3.12: Ellipses in polar coordinates.
The equation of an ellipse with semi-major axes a, b in Cartesian coordinates is:

(x/a)2 + (y/b)2 = 1

or in parametric form
x = a cos θ y = b sin θ.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 80 — #96

i i

80 Classical Mechanics: A Computational Approach

(a) Use the symbolic capabilities of Mathematica to show that the general equation for
this ellipse in polar coordinates is:
ab
r= √ (3.3.25)
a − c2 cos2 θ
2

√
where c = a2 − b2 (with a > b). √
(b) Show that by shifting the x-axis by the quantity c = a2 − b2 , the equation of the
ellipse in polar coordinates simplifies to:

a 1 − e2

b2 /a
r= = (3.3.26)
1 + (c/a) cos θ 1 + e cos θ

where e = c/a = 1 − b2 /a2 (with a > b) is the eccentricity of the ellipse.

(c) Plot the above expressions in Mathematica using the PolarPlot command.
Solution:
(a) The following Mathematica code performs the necessary algebraic steps for deriving
(3.3.25), and then plots the ellipse with parameters
√ a = 2, b = 1. After defining the para-
metric forms x = a cos θ and y = b sin θ = b 1 − cos2 θ, the code uses the Solve command
to find the distance r as a function of cos θ. The command “replace with”/. is used to
define a parameter rellipse for the desired equation r = r(θ). Since there are two solutions
obtained from the Solve command, we chose the positive solution by selecting the second
(positive) square root.
x = r ∗ Cos[θ]; y = r ∗ Sqrt[1 − Cos[θ]∧ 2];
sol = Solve[(x/a)∧ 2 + (y/b)∧ 2 == 1, r]
   
   
OUTPUT: r → −q 1
, r → q 1
 1 + Cos[θ]2 − Cos[θ]2   1 + Cos[θ]2 − Cos[θ]2 
b2 a2 b2 b2 a2 b2

rellipse = r/.sol[[2]]

OUTPUT: q 1
1 + Cos[θ]2 − Cos[θ]2
b2 a2 b2

PolarPlot[rellipse/.{a → 2, b → 1}, {θ, 0, 2π}]

0.8

0.4
In[ ]:=
x
-1.8 -1.2 -0.6 0.6 1.2 1.8
-0.4

-0.8

(b) The following code carries out the algebra for deriving (3.3.26), and plots the ellipse.
As the polar plot shows, the resulting ellipse from (3.3.26) is shifted along the x-axis

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 81 — #97

i i

Motion in Two and Three Dimensions 81

√ √ √
by c = a2 − b2 = 22 − 12 = 3 . After some simple manipulation of the Mathematica
output, we obtain (3.3.26), which is the commonly used equation for an ellipse in polar
coordinates.
c = Sqrt[a∧ 2 − b∧ 2];
x = r ∗ Cos[θ]; y = r ∗ Sqrt[1 − Cos[θ]∧ 2];
sol2 = Solve[((x + c)/a)∧ 2 + (y/b)∧ 2 == 1, r]//FullSimplify;
rellipse2 = r/.sol2[[1]]

b2
OUTPUT: √
a+ (a−b)(a+b)Cos[θ]

PolarPlot[rellipse2/.{a → 2, b → 1}, {θ, 0, 2π}]

1.0

0.5

x
-4 -3 -2 -1 1
-0.5

-1.0

Using the derivatives of the unit vectors dt

d
θ̂ and dt
d
r̂, we can now also calculate the
acceleration vector as the second derivative of the position in polar coordinates (see the
end-of-chapter problems):
" 2 #
d2 r
2
dv dθ d θ dr dθ
a= = −r r̂ + r 2 + 2 θ̂θ (3.3.27)
dt dt2 dt dt dt dt

The
h 2two terms2in
i this equation again have distinct physical meanings. The first term
ar = dt2 − r dt
d r dθ
r̂ represents the radial component of the acceleration, which points
h 2 i
along the radius r. The first part of this radial term is ddt2r r̂ and represents the rate of
2
change of the radial velocity dr/dt. The second part of this radial term is −r dθ
dt = −rω 2 ,
d2 r
and points in the opposite direction from the dt2
term, and represents the centripetal
d2 θ
acceleration ac = v 2 /r. The second term aθ = (r dt2 + 2 dr
dt dt )θ̂ also contains two terms; the
dθ θ
2
first term r ddt2θ = r dω
dt = rα is the angular acceleration along the θ̂ direction. The last term
θ
represents the Coriolis force and will be discussed in Chapter 10.
In summary, the position, velocity, and acceleration vectors in polar coordinates are given
by:

Position, Velocity, and Acceleration in Polar Coordinates

r = rr̂ (3.3.28)

dr dr dθ
v= = r̂ + r θ̂θ (3.3.29)
dt dt dt

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 82 — #98

i i

82 Classical Mechanics: A Computational Approach

" 2 #
d2 r
2
dv dθ d θ dr dθ
a= = − r r̂ + r + 2 θ̂θ (3.3.30)
dt dt2 dt dt2 dt dt

3.3.2 Position, Velocity, and Acceleration in Cylindrical Coordinates

A point (x, y, z) in space can be expressed in cylindrical coordinates (r, θ, z) as shown in
Figure 3.6. Cylindrical coordinates are most useful when the problem to be solved has a
cylindrical symmetry. It is useful to remember that cylindrical coordinates (r, θ, z) are a
direct extension of polar coordinates (r, θ) from two to three dimensions, by adding the
z-coordinate.

Figure 3.6: Cylindrical coordinates (r, θ, z) and the corresponding unit vectors r̂, θ̂, ẑ.

The transformation from Cartesian to polar coordinates is made by the following rela-
tions:
From Cylindrical to Cartesian Coordinates

x = r cos θ (3.3.31)

y = r sin θ (3.3.32)
z=z (3.3.33)

From Cartesian to Cylindrical Coordinates

p
r = x2 + y 2 + z 2 (3.3.34)

tan θ = y/x (3.3.35)

z=z (3.3.36)

We can define three unit vectors r̂, θ̂θ , ẑ in cylindrical coordinates, which are shown in
Figure 3.6. Since cylindrical coordinates are essentially an extension of polar coordinates in
three dimensions, we can write the unit vectors immediately by analogy with the equations
in the previous section.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 83 — #99

i i

Motion in Two and Three Dimensions 83

Unit Vectors in Cylindrical Coordinates

r̂ = cos θî + sin θĵ (3.3.37)

θ̂θ = − sin θî + cos θĵ (3.3.38)

ẑ = k̂ (3.3.39)

Similar to the unit vectors in polar coordinates, the unit vectors in cylindrical coordinates
change direction with the location of the particle, with the exception of ẑ, which is identical
to k̂ and does not change direction. Note also that the cylindrical unit vectors are related
to the Cartesian unit vectors by a rotation matrix of the form:

cos θ sin θ 0
    
r̂ î
θ̂θ  = − sin θ cos θ 0  ĵ  (3.3.40)
ẑ 0 0 1 k̂

We can compute the position and velocity vectors in cylindrical coordinates in a manner
similar to that used for polar coordinates. The difference is that for cylindrical coordinates
there is an additional direction, ẑ. However, ẑ does not change direction with position and
is therefore constant. Hence, the position, velocity, and acceleration vectors in cylindrical
coordinates are given by:

Position, Velocity, and Acceleration Vectors in Cylindrical Coordinates

r = rr̂ (3.3.41)

dr dθ dz
v= r̂ + r θ̂θ + ẑ (3.3.42)
dt dt dt
" 2 #
dv 2
d r dθ 2
d θ dr dθ d2 z
a= = − r r̂ + (r + 2 ) θ
θ̂ + ẑ (3.3.43)
dt dt2 dt dt2 dt dt dt2

This section concludes with some additional useful expressions. By analogy to (3.3.19),
which was derived for polar coordinates, the infinitesimal displacement ds between two
points in space in cylindrical coordinates can be written as:

ds = drr̂ + rdθθ̂θ + dzẑ (3.3.44)

When computing infinitesimal areas, it is important to keep in mind the specific area on
the cylinder that is relevant to your problem. It is common to work with the infinitesimal
area element that is on the surface of the body of the cylinder. The infinitesimal area element
dA on the surface of a cylinder of radius r, can be written as:

dA = rdθdz (3.3.45)

Keep in mind, however, that if one is interested in the top or bottom of the cylinder, then
the polar infinitesimal area element, dA = rdrdθ, applies. An infinitesimal volume element
in cylindrical coordinates is given by:

dV = dxdydz = rdrdθdz (3.3.46)

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 84 — #100

i i

84 Classical Mechanics: A Computational Approach

and we can calculate triple integrals of any function f (x, y, z) by using the transformation
into cylindrical coordinates:
ZZZ ZZZ
f (x, y, z) dxdydz = f (r cos θ, r sin θ, z) r dr dθ dz (3.3.47)

These triple integrals will become useful in calculating, for example, the center of mass and
moment of inertia of an object in cylindrical coordinates.

3.3.3 Position, Velocity, and Acceleration in Spherical Coordinates

In spherical coordinates, we specify the coordinates of every point in space by two angles
(θ , φ) and by the radial distance r, as shown in Figure 3.7. It is important to remember
that the polar angle θ can range between 0 and π, while the azimuthal angle φ can vary
from 0 to 2π. Spherical coordinates are useful for problems involving spherical symmetry,
a common occurrence in classical mechanics.

Figure 3.7: Spherical coordinates (r, θ, φ). The angle φ is defined on the xy-plane and can
vary between 0 to 2π, while the azimuthal angle θ can vary between 0 to π only.

The Cartesian coordinates x, y, z are related to the spherical coordinates (r , θ , φ ) by:

From Spherical to Cartesian Coordinates

x = r sin θ cos φ (3.3.48)

y = r sin θ sin φ (3.3.49)

z = r cos θ (3.3.50)

From Cartesian to Spherical Coordinates

p
r = x2 + y 2 + z 2 (3.3.51)

tan φ = y/x (3.3.52)

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 85 — #101

i i

Motion in Two and Three Dimensions 85

z z
cos θ = =p (3.3.53)
r x + y2 + z2
2

Similar to the unit vectors in cylindrical coordinates, the orientation of the spherical
coordinate unit vectors r̂, θ̂θ , φ̂
φ depends on the position of the particle. We define these three
unit vectors as shown in Figure 3.7.
These unit vectors r̂, θ̂θ , φ̂φ for spherical coordinates are related to the Cartesian unit
vectors by the following equations:
Unit Vectors in Spherical Coordinates

r̂ = sin θ cos φî + sin θ sin φĵ + cos θk̂ (3.3.54)

θ̂θ = cos θ cos φî + cos θ sin φĵ − sin θk̂ (3.3.55)
φ = − sin φî + cos φĵ
φ̂ (3.3.56)

Like the cylindrical coordinate unit vectors, the spherical unit vectors are related to the
Cartesian unit vectors by an orthogonal rotation matrix:

sin θ cos φ sin θ sin φ cos θ

    
r̂ î
 θ̂θ  = cos θ cos φ cos θ sin φ − sin θ  ĵ  (3.3.57)
φ
φ̂ − sin φ cos φ 0 k̂
In order to calculate velocity and acceleration in spherical coordinates, we need the time
derivatives of the unit vectors. They are given by (see end-of-chapter problems):
r̂˙ = θ̇θ̂θ + φ̇ sin θφ̂
φ (3.3.58)
˙
θ̂θ = −θ̇r̂ + φ̇ cos θφ̂ φ (3.3.59)
˙
φ = −φ̇ sin θr̂ − φ̇ cos θθ̂
φ̂ (3.3.60)
By using these equations, it can be shown that the velocity vector in spherical coordinates
is given by:
v = ṙr̂ + rθ̇θ̂θ + rφ̇ sin θφ̂
φ (3.3.61)
Multiplying both sides of this equation by dt, we obtain the infinitesimal displacement ds
between two points in spherical coordinates:
ds = drr̂ + rdθθ̂θ + rdφ sin θφ̂
φ (3.3.62)
By multiplying the coefficients of the unit vectors in this equation, the infinitesimal volume
element dV in spherical coordinates is found to be:
dV = dxdydz = r2 sin θdrdθdφ (3.3.63)
We can also calculate triple integrals in spherical coordinates by using the transformation
equation:
ZZZ ZZZ
f (x, y, z) dx dy dz = f (r sin θ cos φ, r sin θ sin φ, r cos θ)r2 sin θ dr dθ dφ (3.3.64)

Finally, the acceleration vector in spherical coordinates is given by (see end-of-chapter

problems):
˙
a = r̈ − rφ̇2 sin2 θ − rθ̇2 r̂ + rθ̈ + 2ṙθ̇ − rφ̇2 sin θ cos θ θ̂θ

+ rφ̈ sin θ + 2ṙφ̇ sin θ + 2rθ̇φ̇ cos θ φ̂ (3.3.65)

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 86 — #102

i i

86 Classical Mechanics: A Computational Approach

Position, Velocity, and Acceleration Vectors in Spherical Coordinates

r = rr̂ (3.3.66)

v = ṙr̂ + rθ̇θ̂θ + rφ̇ sin θφ̂

φ (3.3.67)

a = r̈ − rφ̇2 sin2 θ − rθ̇2 r̂ + rθ̈ + 2ṙθ̇ − rφ̇2 sin θ cos θ θ̂θ

+ rφ̈ sin θ + 2ṙφ̇ sin θ + 2rθ̇φ̇ cos θ φ̂ (3.3.68)

3.4 THE GRADIENT, DIVERGENCE, AND CURL

There are several important spatial derivatives (i.e., derivatives that are taken with respect
to a spatial coordinate) that occur in physics. These are called the gradient, divergence, and
curl. In this section we will present these derivatives, along with physical interpretations.

3.4.1 The Gradient

Consider a scalar function f (x). Suppose we want to know how much the function f (x)
changes when we change its argument x by a small amount, say from x to x + dx. A Taylor
series expansion of f (x) gives:

df
f (x + dx) = f (x) + dx + · · · (3.4.1)
dx
We can ignore terms of dx2 and higher, because dx is very small. Next, we define the
differential df = f (x + dx) − f (x) and interpret it as the change of the value of f when we
change x by a small amount dx. Then from (3.4.1) we find:

df
df = dx (3.4.2)
dx
Equation (3.4.2) can be interpreted as follows. The amount that f changes when x is changed
(by a small amount), is equal to the product of the rate of change of f with respect to x,
and the amount by which x is changed (dx). This is similar to the familiar equation d = vt,
which says the distance traveled by a particle is the product of how fast it is traveling (i.e.,
the rate v at which the distance changes with time), and of the length of time t traveled.
Next, we can extend (3.4.2) to scalar functions in three dimensions. Let us consider a
scalar function f (x, y, z) in Cartesian coordinates. We would like to know how f (x, y, z)
changes when we change each of its variables by a small amount. A multivariable Taylor
series expansion gives:
∂f ∂f ∂f
f (x + dx, y + dy, z + dz) = f (x, y, z) + dx + dy + dz + · · · (3.4.3)
∂x ∂y ∂z

In this expression, ∂f ∂f ∂f
∂x , ∂y , ∂z are, of course, the partial derivatives of f with respect to
(x, y, z) respectively. Similar to the one-dimensional case, we define the differential as df =
f (x + dx, y + dy, z + dz) − f (x, y, z) and interpret it as the change of f as we change all of
its variables. Using (3.4.3), we can write in dot product form:

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 87 — #103

88 Classical Mechanics: A Computational Approach

Now we ask the question, in what direction does the scalar field f change most rapidly?
We can compute the dot product as:
df
= n̂ · ∇f = |n̂| |∇f | cos θ (3.4.12)
dn
where θ is the angle between n̂ and ∇f . The direction in which the directional derivative is
greatest would occur when n̂ · ∇f is maximal, i.e., when the angle θ equals zero. Therefore,
∇f points in the direction of the most rapid increase of f . Likewise, the direction of greatest
decrease in the scalar field is −∇f . As an example, consider a surface whose height above
sea level at a point (x, y) is H(x, y). The gradient of H at a point (x, y) is a vector pointing
in the direction of the steepest slope, or steepest grade at that point. The steepness of the
slope at that point is given by the magnitude of the gradient vector.

Example 3.13: The gradient of a scalar field

Consider the scalar field:
f (x, y) = e−(x +y )
2 2

(a) Compute and plot the gradient vector of f . (b) Plot the level sets of this function
by using a contour plot to show curves of constant f .
Solution:
The gradient of f can be found by computing:

∂ ∂
e−(x +y )
2 2
∇f = î + ĵ
∂x ∂y
−(x2 +y 2 )
î − 2ye−(x +y ) ĵ
2 2
= − 2xe
Next, we can produce the vector plot and the contour plot in Mathematica. The code
needed to compute ∇f , the contour plot, and the vector field plot is shown below.

f = Exp − x2 + y 2 ;

gradf = Grad[f, {x, y}]

n 2 2 2 2
o
OUTPUT: −2e−x −y x, −2e−x −y y

Show[ContourPlot[f, {x, −1, 1}, {y, −1, 1}, ColorFunction → GrayLevel],

VectorPlot[gradf, {x, −1, 1}, {y, −1, 1}, VectorStyle → Red,
VectorPoints → 10], BaseStyle → {FontSize → 14}, ImageSize → Large]
Notice that Mathematica outputs the gradient as a list. Additionally, the command
Grad is used to compute the gradient, and Mathematica needs to know what variables are
needed for differentiation. Note also that by default, Mathematica computes the gradient
in Cartesian coordinates. As we will see below, the gradient is different in cylindrical
and spherical coordinates. Mathematica can compute symbolically the gradient in those
coordinates as well, when the user specifies which coordinate system to use. The contour
and vector field plot are shown as follows.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 89 — #105

i i

Motion in Two and Three Dimensions 89

1.0

0.5

0.0

-0.5

-1.0
-1.0 -0.5 0.0 0.5 1.0

The contour curves are the level sets of f , with lighter colors corresponding to greater
values of f . The vectors represent the gradient vector field. The size of the arrows cor-
responds to the magnitude of the gradient at that point. Notice that the vectors are
perpendicular to the level sets.

The gradient plays a very important role in classical mechanics. For example, as we will
show in Chapter 5, for a particle with potential energy V (x, y, z), the force vector F (x, y, z)
acting on the particle is given by the negative gradient of the potential energy:
∂V ∂V ∂V
F (x, y, z) = −∇V (x, y, z) = − î − ĵ − k̂ (3.4.13)
∂x ∂y ∂z
Although we will discuss the relationship between force, energy, and equilibrium states of a
mechanical system in more detail in Chapter 5, it is worth noting that the above equation
says that forces point in the direction of most rapidly decreasing potential energy (because
of the minus sign).

Example 3.14: Finding the gradient analytically and using CAS

Find the gradient of the scalar function U (x, y, z) = xyz analytically, and also by using
appropriate codes in Python and Mathematica.
Solution:
The gradient is found by finding the partial derivatives:

∂U ∂U ∂U ∂(xyz) ∂(xyz) ∂(xyz)

∇U (x, y, z) = î + ĵ + k̂ = î + ĵ + k̂ = yz î + xz ĵ + xy k̂
∂x ∂y ∂z ∂x ∂y ∂z
The following are the Python and Mathematica codes for the gradient.
In the Python code, the command ReferenceFrame defines the reference frame R, and
the unit vectors in this reference frame are denoted by R.x, R.y, and R.z. The Cartesian
coordinates x, y, and z are denoted by R[0], R[1], and R[2] correspondingly, and the gra-
dient is calculated with the command gradient(f, R) where f = xyz is the given scalar
function.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 90 — #106

i i

90 Classical Mechanics: A Computational Approach

from sympy . p h y s i c s . v e c t o r import ∗

R = ReferenceFrame ( ’R ’ )

s1 = R[ 0 ] ∗R[ 1 ] ∗R[ 2 ]
print ( g r a d i e n t ( s1 , R) )

OUTPUT:
R y∗ R z ∗R. x + R x∗ R z ∗R. y + R x∗R y∗R. z

In the Mathematica code, we simply use the Grad[f, {x, y, z}] command. Notice that
the results are in a list, where the first element of the list is the x-component, the second
element is the y-component, and the third element is the z-component.

Grad[x ∗ y ∗ z, {x, y, z}]

OUTPUT: {y z,x z,x y}

Example 3.15: The directional derivative

Find the directional derivative of a scalar field f = x2 y + xz at the point (1, 2, −1) and
in the direction of the vector a = 2î − 2ĵ + k̂.
Solution:
We first obtain the unit vector â in the given direction by dividing a by its magnitude:

a 2î − 2ĵ + k̂ 2 2 1
â = q =p = î − ĵ + k̂
a2x + a2y + a2z 22 + 22 + (1)2 3 3 3

Next find the gradient of f :

∂f ∂f ∂f
∇f = î + ĵ + k̂ = (2xy + z)î + x2 ĵ + xk̂
∂x ∂y ∂z
Finally we find the directional derivative from its definition as the dot product of the
unit vector â and the gradient ∇, and then substitute the given point (1, 2, −1):
2 2 1 5
â · ∇f = (2xy + z) + x2 (− ) + x =
3 3 3 3

In order to get the form of the gradient in other coordinate systems, we return to the
fact that the total differential df of the scalar function is given by:
∂f ∂f ∂f ∂f ∂f ∂f
df = dx + dy + dz = dr + dθ + dz = ∇f · dr (3.4.14)
∂x ∂y ∂z ∂r ∂θ ∂z

where in Cartesian coordinates dr = dxî + dy ĵ + dz k̂. We can use this equation to obtain
expressions for the gradient in cylindrical and spherical coordinates, as follows.
The gradient of any scalar function f in cylindrical coordinates (r, θ, z) will be of the
general form:
∇f (x, y, z) = (∇f )r r̂ + (∇f )θ θ̂ + (∇f )z k̂ (3.4.15)

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 91 — #107

i i

Motion in Two and Three Dimensions 91

where (∇f )r , (∇f )θ , (∇f )z are the components of the gradient along the unit vectors r̂, θ̂, ẑ
in cylindrical coordinates. From (3.3.19) we recall that dr in cylindrical coordinates is given
by:
dr = drr̂ + rdθθ̂ + dzẑ (3.4.16)
By evaluating the dot product of (3.4.15) and (3.4.16), and substituting into (3.4.14), we
obtain:
∂f ∂f ∂f
dr + dθ + dz = (∇f )r dr + (∇f )θ rdθ + (∇f )z dz (3.4.17)
∂r ∂θ ∂z
By equating the corresponding coefficients of (dr, dθ, dz) on the two sides of (3.4.17), we
find:
∂f 1 ∂f ∂f
(∇f )r = , (∇f )θ = , (∇f )z =
∂r r ∂θ ∂z
Therefore by using (3.4.15), the expression of the gradient in cylindrical coordinates is:

∂f 1 ∂f ∂f
∇f (r, θ, z) = r̂ + θ̂ + ẑ (3.4.18)
∂r r ∂θ ∂z
The gradient in spherical coordinates can also be obtained with the same method, and is
of the form:
∂f 1 ∂f 1 ∂f
∇f (r, θ, φ) = r̂ + θ̂ + φ̂ (3.4.19)
∂r r ∂θ r sin θ ∂φ
Note that our work above has also given us the del operator in various coordinate systems:

The Gradient in Different Coordinate Systems

∂ ∂ ∂
∇ =î + ĵ + k̂ Cartesian Coordinates (3.4.20)
∂x ∂y ∂z
∂ 1 ∂ ∂
∇ =r̂ + θ̂ + ẑ Cylindrical Coordinates (3.4.21)
∂r r ∂θ ∂z
∂ 1 ∂ 1 ∂
∇ =r̂ + θ̂ + φ̂ Spherical Coordiantes (3.4.22)
∂r r ∂θ r sin θ ∂φ

The “derivative” properties of the gradient are shown in the following identity, which is
similar to the derivative product rule:

∇(f g) = f ∇g + g∇f (3.4.23)

“master˙book˙file” — 2020/9/5 — 9:32 — page 93 — #109

i i

Motion in Two and Three Dimensions 93

In physical terms, the divergence of a three-dimensional vector field is the extent to

which the vector field flow behaves like a source or a sink at a given point. It is a local
measure of its “outgoingness”—the extent to which there is more “material” exiting an
infinitesimal region of space than entering it. If the divergence is nonzero at some point,
then there must be a source or sink at that position. Note that when we use the terms
source and sink, we are imagining a vector like the velocity vector field of a fluid in motion.
If the divergence of a vector field A(x, y, z) is zero everywhere in space, this vector field
is called solenoidal.

Solenoidal Vector Fields A(x,y,z)

∇ · A = 0 ←→ A(x, y, z) is a solenoidal vector field (3.4.25)

If u, v are any vector functions and f is any scalar function, the divergence also obeys the
following product rule:
∇ · (f v) = f (∇ · v) + v · ∇f (3.4.26)
However, note that the divergence of a cross-vector product is less intuitive:

∇ · (u × v) = v · (∇ × u) − u · (∇ × v) (3.4.27)

Example 3.17: The divergence of a vector field

Find the divergence of the vector function v = xyz î + xyz ĵ + xyz k̂ analytically and also
by using appropriate codes in Python and Mathematica.
Solution:
The divergence is a scalar, found by adding the partial derivatives:

∂(xyz) ∂(xyz) ∂(xyz)

∇·v = + + = yz + xz + xy
∂x ∂y ∂z
The following are the Python and Mathematica codes for the gradient.
To check if a vector field is solenoidal, we use the package physics.vector from the
SymPy library. The function is solenoidal is used to check if the field is solenoidal. In this
example, the function returns the value False in Python, so that this field is not solenoidal.

from sympy . p h y s i c s . v e c t o r import ∗

R = ReferenceFrame ( ’R ’ )

v1 = R [ 0 ] ∗ R [ 1 ] ∗ R [ 2 ] ∗ (R. x+R. y+R. z )

print ( d i v e r g e n c e ( v1 , R) )
print ( i s s o l e n o i d a l ( v1 ) )

OUTPUT:
R x∗R y + R x∗ R z + R y∗ R z
False

Div[{x ∗ y ∗ z, x ∗ y ∗ z, x ∗ y ∗ z}, {x, y, z}]

OUTPUT: x y+x z+y z

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 94 — #110

i i

94 Classical Mechanics: A Computational Approach

Note that the divergence takes on a different form in cylindrical and spherical coordinates.
To compute the divergence in these coordinate systems, choose the appropriate form of the
del operator from (3.4.7), and compute the dot product of the appropriate form of the Del
Operator and the vector of interest. The results are (see also the problems at the end of
this chapter):

The Divergence in Cartesian, Cylindrical, and Spherical Coordinates

∂vx ∂vy ∂vz

∇·v = + + (3.4.28)
∂x ∂y ∂z
1 ∂ (r vr ) 1 ∂ (vθ ) ∂vz
∇·v = + + (3.4.29)
r ∂r r ∂θ ∂z
1 2v
1 1 ∂ vφ

∂ r r ∂
∇·v = 2 + (vθ sin θ) + (3.4.30)
r ∂r r sin θ ∂θ r sin θ ∂φ

3.4.3 The Curl

The divergence uses the del operator to measure the outflow rate of a vector field. We can
also use the del operator to measure the rotational properties of a vector field, called the
curl. The curl of a vector field v(x, y, z) = vx î + vy ĵ + vz k̂ is a vector function that can be
represented as:

∂vz ∂vy ∂vx ∂vz ∂vy ∂vx
∇×v = − î + − ĵ + − k̂ (3.4.31)
∂y ∂z ∂z ∂x ∂x ∂y
From a physical point of view, the curl at a point (x, y, z) is proportional to the on-axis
torque to which a tiny pinwheel would be subjected, if it were centered at that point.
The cross-product operation can be written as a pseudo-determinant in Cartesian coor-
dinates:
î ĵ k̂

∇ × v = ∂x (3.4.32)
∂ ∂ ∂
∂y ∂z
vx vy vz
If u and v are any vector functions and f is any scalar function, the curl also follows a
product rule:
∇ × (f v) = [(∇f ) × v] + [f ∇ × v] (3.4.33)
However, the curl of a cross product is more complex:
∇ × (u × v) = (u ∇ · v) − (v ∇ · u) + (v · ∇)u − (u · ∇)v (3.4.34)
The curl takes a different form in cylindrical and spherical coordinates. To compute the curl
in spherical and cylindrical coordinates, choose the appropriate form of the del operator from
(3.4.20), (3.4.21), and (3.4.22) and evaluate the cross product between del and the vector
of interest. The results are:
The Curl in Cartesian, Cylindrical, and Spherical Coordinates

∂vz ∂vy ∂vx ∂vz ∂vy ∂vx
∇×v = − î + − ĵ + − k̂ (3.4.35)
∂y ∂z ∂z ∂x ∂x ∂y

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 95 — #111

i i

Motion in Two and Three Dimensions 95

1 ∂vz ∂vθ 1 ∂ (rvθ ) ∂vr

∂vr ∂vz
∇×v = − r̂ + − θ̂ + − ẑ (3.4.36)
r ∂θ ∂z ∂z ∂r r ∂r ∂θ

1 1 1 ∂vr

∂ ∂vθ ∂
∇×v = vφ sin θ − r̂ +

− rvφ θ̂
r sin θ ∂θ ∂φ r sin θ ∂φ ∂r
1 ∂ (rvθ ) ∂vr

+ − φ̂ (3.4.37)
r ∂r ∂θ

The curl has many uses in physics. For example, as we will study in more detail in
Chapter 5, the necessary and sufficient condition for a force F (x, y, z) to be conservative is
that its curl equals zero:

Conservative force field F(x,y,z)

∇ × F = 0 ←→ F(x, y, z) is a conservative force field (3.4.38)

Example 3.18: Calculating the curl using Python and Mathematica.

Find the curl of the vector function v = (y 2 z)î + (xz) ĵ + (xy) k̂ analytically, and also
by using appropriate codes in Python and Mathematica.
Solution:
The curl is found by finding the pseudo-determinant:

î ĵ k̂
∂(y 2 z) ∂(xy) ∂(xz) ∂(y 2 z)

∂(xy) ∂(xz)
∇×v = ∂x ∂y ∂z =
∂ ∂ ∂ − î+ − ĵ+ − k̂

2 ∂y ∂z ∂z ∂x ∂x ∂y
y z xz xy

∇ × v = (x − x) î + y 2 − y ĵ + (z − 2yz) k̂ = (0) î + y 2 − y ĵ + (z − 2yz) k̂

The following are Python and Mathematica codes for the gradient.
We first compute the curl using the function curl from the Python SymPy library
package physics.vector. Notice that the curl is nonzero.
To check if a vector field is conservative we use the function is conservative which is
also from the SymPy library package physics.vector. In this example, the function returns
the value False, so that this field is not conservative.

from sympy . p h y s i c s . v e c t o r import ∗

R = ReferenceFrame ( ’R ’ )

v2 = R [ 1 ] ∗ R [ 1 ] ∗ R [ 2 ] ∗ R. x + R [ 0 ] ∗ R [ 2 ] ∗ R. y + R [ 0 ] ∗ R [ 1 ] ∗ R. z
print ( c u r l ( v2 , R) )
print ( i s c o n s e r v a t i v e ( v2 ) )

OUTPUT:
( R y ∗∗2 − R y ) ∗R. y + (−2∗R y∗ R z + R z ) ∗R. z
False

In Mathematica, we use the command Curl to find the curl of the vector function.
Notice that the curl is not equal to zero and, therefore, the vector field is not conservative.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 96 — #112

i i

96 Classical Mechanics: A Computational Approach

Curl[{y ∧ 2 ∗z, x ∗ z, x ∗ y}, {x, y, z}]

OUTPUT: 0, −y + y 2 , z − 2yz

3.4.4 Second Derivatives with the Del Operator

The Laplace operator or Laplacian ∇2 is a scalar operator that can be applied to either
vector or scalar fields. The Laplacian is defined as:

∂2 ∂2 ∂2
∇2 = ∇ · ∇ = + + (3.4.39)
∂x2 ∂y 2 ∂z 2
The Laplacian is ubiquitous throughout modern mathematical physics, appearing in
Laplace’s equation, Poisson’s equation, the heat equation, the wave equation, and the
Schrödinger equation—to name a few.
When the del operator acts on the gradient, divergence, and curl of a vector, it produces
five possible second derivatives; the use of the scalar Laplacian and vector Laplacian gives
two more. Here are these “second derivatives” created by applying the del operator.

div (grad f ) = ∇ · (∇f ) (3.4.40)

curl (grad f ) = ∇ × (∇f ) (3.4.41)
∆f = ∇2 f (3.4.42)
grad (div v) = ∇(∇ · v) (3.4.43)
div (curl v) = ∇ · (∇ × v) (3.4.44)
curl (curl v) = ∇ × (∇ × v) (3.4.45)
∆v =∇2 v (3.4.46)

In the above equations, we included the older notation of grad, div, curl, and ∆ with the
more modern del ∇ notation. You may encounter the older notation in some other books.
These second derivatives are of interest, principally because they are not always unique or
independent of each other. As long as the functions v and f are well-behaved, two of these
identities are always zero:

Two Important Vector Identities

curl (grad f ) = ∇ × (∇f ) = 0 (3.4.47)

div (curl v) = ∇ · (∇ × v) = 0 (3.4.48)

The first identity states that the curl of the gradient of any scalar field is always equal to
zero, and the second one states that the divergence of the curl of any vector field is always
equal to zero.
Two of the second derivatives are always equal:

div (grad f ) = ∇ · (∇f ) = ∇2 f = ∆f (3.4.49)

The remaining three vector derivatives are related by the equation:

∇ × (∇ × v) = ∇(∇ · v) − ∇2 v (3.4.50)

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 97 — #113

i i

Motion in Two and Three Dimensions 97

Example 3.19: Second derivatives

Consider the function V (x, y, z) = x + yz + z 2 in three dimensions. Calculate the gra-
dient, the divergence of the gradient, the curl of the gradient, and the Laplacian of this
function.
Solution:
The gradient is
∂V ∂V ∂V
∇V = î + ĵ + k̂ = (1)î + (z)ĵ + (y + 2z)k̂
∂x ∂y ∂z
The divergence of the gradient is the dot product of the gradient and the del operator:

∂ ∂V ∂ ∂V ∂ ∂V ∂2V ∂2V ∂2V

∇ · (∇V ) = ( )+ ( )+ ( )= 2
+ 2
+
∂x ∂x ∂y ∂y ∂z ∂z ∂x ∂y ∂z 2

Calculating the second derivatives: ∇ · (∇V ) = 2

This also represents the Laplacian of V , since

∂2 ∂2 ∂2
2
+ 2 + 2 = ∇ · ∇ = ∇2
∂x ∂y ∂z

To find the curl of the gradient, calculate the cross product, (∇ × ∇V ). This quantity
is always zero, as seen in (3.4.47) for the curl of a gradient.

3.5 CHAPTER SUMMARY

Motion in two and three dimensions is described using vectors. In this book, we use the
following conventions for a vector, its magnitude, and a unit vector:
q A A
A = Ax î + Ay ĵ + Az k̂ A = |A| = A2x + A2y + A2z Â = =q
|A| Ax + A2y + A2z
2

The dot product between two vectors A and B can be summarized by the following
relationships:
d dA dB
A · B = Ax Bx + Ay By + Az Bz = AB cos θ (A · B) = B+A
dt dt dt
The cross product between two vectors A and B can be summarized by the following
relationships:
a × b = |a||b| sin θ n̂

î ĵ k̂
a2 a3 a1 a3 a a2
a × b = a1 a3 = ĵ + 1

a2 î − k̂
b1 b2 b3 b1 b3 b1 b2
b2 b3
d da db
a × b = −b × a (a × b) = × b + a×
dt dt dt
There are several relationships that can simplify expressions with multiple dot or cross
products. They are:

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 98 — #114

i i

98 Classical Mechanics: A Computational Approach

The vector triple product: a × (b × c) = b(a · c) − c(a · b)

The scalar triple product: a · (b × c) = b · (c × a) = c · (a × b) = V
In addition, we often need to perform derivatives of vectors with respect to spatial coor-
dinates. Those derivatives are the gradient, divergence, and curl, written here in Cartesian
coordinates:
∂f ∂f ∂f
∇f (x, y, z) = î + ĵ + k̂
∂x ∂y ∂z
∂vx ∂vy ∂vz
∇·v = + +
∂x ∂y ∂z

î ĵ k̂

∂ = ∂vz −
∂vy ∂vx ∂vz ∂vy ∂vx
∇ × v = ∂x + +
∂ ∂
∂y ∂z
î − ĵ − k̂
∂y ∂z ∂z ∂x ∂x ∂y
vx vy vz

The gradient gives the direction of most rapid change in a scalar field. The divergence
measures the magnitude of a vector field’s source or sink at a given point. The curl at a
point is proportional to the on-axis torque to which a tiny pinwheel is subjected.
The directional derivative df /dn in the direction n̂ = nx î + ny ĵ + nz k̂ is:
df ∂f ∂f ∂f
= n̂ · ∇f = nx + ny + nz
dn ∂x ∂y ∂z
A force F will be conservative if and only if its curl is zero: ∇×F = 0
Second derivatives of vector and scalar fields are also important in mechanics. One
such example, the Laplacian ∇2 is a scalar operator that can be applied to either vector
or scalar fields:

∂2 ∂2 ∂2
∇2 = ∆ = 2
+ 2+ 2
∂x ∂y ∂z

∇ × (∇f ) = 0

∇·∇×v = 0

Motion in two and three dimesions can sometimes be best described by non-Cartesian
coordinates. Those coordinates often involve using Euler’s identities:

eiθ + e−iθ eiθ − e−iθ

cos θ = sin θ =
2 2i

eiθ = cos θ + i sin θ e−iθ = cos θ − i sin θ

Polar coordinates are related to Cartesian coordinates using:

x = r cos θ y = r sin θ

r = x2 + y 2 θ = tan−1 (y/x)
p

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 99 — #115

i i

Motion in Two and Three Dimensions 99

The unit vectors in polar coordinates are:

r̂ = cos θ î + sin θ ĵ θ̂θ = − sin θ î + θ̂ cos θ ĵ r = rr̂

(A × B) · (C × D) = (A · C) (B · D) − (A · D) (B · C)

A × (B × C) + B × (C × A) + C × (A × B) = 0
(A × B) · [(B × C) × (C × A)] = (A · [B × C])2

16. An object with mass m = 2 kg has a position vector given by r =

3t2 i–5 sin t j + 5 cos t k m. Use the symbolic capabilities of Mathematica to eval-

uate the following:

(a) Find the magnitude of the angular momentum vector at t = π sec
(b) Find the angle of the position and velocity vectors at time t = π sec
(c) Find the torque τ = r × F on the body at t = π sec, where F is the force acting
on the body.
17. The equation of motion of a particle P with a mass m is given by

d2 r
m = f (r)r̂
dt2
where r is the position vector of P measured from an origin O, r̂ is a unit vector in
the direction r, and f (r) is a function of the distance of P from O.
(a) Show that r × dr
dt =c where c is a constant vector.
(b) Interpret physically the cases f (r) < 0 and f (r) > 0.
(c) Interpret the result in (a) geometrically.
(d) Describe how the results obtained relate to the motion of the planets in our solar
system.

Section 3.3: Position, Velocity, and Acceleration in Non-Cartesian Coordinates

18. A mass m hangs from a string of length L, so that it moves on a horizontal circular
motion. The string makes an angle Θ with the vertical line. Calculate the period T of
the rotational motion.
19. Show that the acceleration vector in polar coordinates is given by:

dθ 2
h 2 i 2
a = dv
dt = d r
dt 2 − r dt r̂ + (r ddt2θ + 2 dr
dt dt )θ̂
dθ

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 102 — #118

i i

102 Classical Mechanics: A Computational Approach

One possible method is by taking the time derivative of the velocity vector

dr dr dθ
v= = r̂ + r θ̂
dt dt dt

and then using the derivatives of the unit vectors d

dt θ̂ and d
dt r̂ derived in this chapter.
20. A point mass m is placed at the top of the frictionless surface of a sphere of radius
R , and slides down the surface of the sphere. At what point does the mass leave the
spherical surface?
21. Consider a mass m moving in the xy-plane under the influence of a force given by the
expression Fθ = mṙθ̇ and Fr = 0 , where (r, θ) are polar coordinates. Show that

r = C θ̇−1
√
ṙ = A ln r + B
where A, B, C are constants.

22. A wheel of radius R is rolling without slipping on a flat frictionless surface.

(a) Show that the path of a point on the circumference of the wheel is a cycloid
curve:
x = Rθ + R sin θ y = R + R cos θ
where θ is the angle of rotation of the wheel around its center of mass.

(b) Find the velocity and acceleration of this point on the surface of the rolling wheel.
(c) The curvature k at a point of a curve is defined as

|a × v|
k=
|v|3
where a and v are the acceleration and velocity vectors of the wheel. Find the
curvature k at a point located at the top of the cycloid.
23. Find the radius of curvature at the top of the parabolic path of a projectile by assuming
no air resistance. The curvature k at a point of a curve is defined as

|a × v|
k=
|v|3
where a, v are the acceleration and velocity vectors of the object.

24. Find expressions for the velocity and the acceleration vectors in spherical coordinates.

Section 3.4: The Gradient, Divergence, and Curl

25. Represent the vector A = z î − 2xĵ + y k̂ in cylindrical coordinates. Thus determine Ar ,
Aθ , and Az , such that A = Ar r̂ + Aθ θ̂ + Az k̂.
26. Cylindrical and spherical coordinates are examples of orthogonal curvilinear systems.
In such coordinate systems we can obtain expressions for the gradient, divergence and
curl by using the following general procedure. Let r be some point in space and qi

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 103 — #119

i i

Motion in Two and Three Dimensions 103

represent the coordinate system, we can define three vectors ei which are orthogonal
to each other by evaluating the partial derivatives:
∂r
ei = ei · ej = 0 (i = 1, 2, 3 i 6= j)
∂qi
These ei vectors can be normalized by dividing by their length, and the normalized
basis vectors are then:
ei ei
êi = =
hi |ei |
The lengths hi of the basis vectors êi are very useful functions known as scale factors
of the coordinates. In the case of Cartesian coordinates, qi = x, y, z and

∂r ∂r ∂r
êi = , , = î, ĵ, k̂
∂x ∂y ∂z
and it is easy to verify that hi = 1.The infinitesimal length in this orthogonal coordi-
nate system is found from:
√ q
ds = dr · dr = h21 dq12 + h22 dq22 + h23 dq32
The corresponding infinitesimal volume elements are calculated from:
dV = h1 h2 h3 dq1 dq2 dq3
(a) Find expressions for the unit vectors êi in cylindrical coordinates, and determine
the corresponding scale factors.
(b) Find expressions for the infinitesimal arc length ds2 and infinitesimal volume dV .

27. Repeat the previous problem for spherical coordinates, i.e., find the unit vectors êi ,
the square of the element of arc length ds2 , the corresponding scale factors and the
infinitesimal volume dV for spherical coordinates.
28. The gradient of a scalar field, f , the divergence and curl of a vector field
F = (F1 , F2 , F3 ), and the Laplacian of a scalar field f can be found using the
following general expressions:

ê1 ∂f ê2 ∂f ê3 ∂f

∇f = + +
h1 ∂q1 h2 ∂q2 h3 ∂q3
1

∂ ∂ ∂
∇·F = (F1 h2 h3 ) + (F2 h3 h1 ) + (F3 h1 h2 )
h1 h2 h3 ∂q1 ∂q2 ∂q3

ê1 ∂ ∂ ê2 ∂ ∂
∇×F = (h3 F3 ) − (h2 F2 ) + (h1 F1 ) − (h3 F3 )
h2 h3 ∂q2 ∂q3 h3 h1 ∂q3 ∂q1

ê3 ∂ ∂
+ (h2 F2 ) − (h1 F1 )
h1 h2 ∂q1 ∂q2
h1 ê1 h2 ê2 h3 ê3

1 ∂
∂ ∂
=
h1 h2 h3 ∂q1 ∂q ∂q

h F h F2 h F3
1 1 2 2 3 3
1

∂ h h
2 3 ∂f ∂ h3 h1 ∂f ∂ h1 h2 ∂f
∇ f=
2
+ +
h1 h2 h3 ∂q1 h1 ∂q1 ∂q2 h2 ∂q2 ∂q3 h3 ∂q3

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 104 — #120

i i

104 Classical Mechanics: A Computational Approach

where the hi are the scale factors of the basis vectors, êi , which were defined in
Problems 26 and 27.
(a) Use the given expressions to obtain the gradient, divergence, and curl in cylindrical
coordinates.
(b) Obtain the divergence for spherical coordinates.
29. Consider the function V (x, y, z) = x4 + y 5 + z 6 in three dimensions. Calculate the
following quantities and indicate which of these is a vector and which is a scalar.

(a) ∇ ∇2 V

(b) (∇V × ∇V ) · (∇V )

(c) ∇ × ∇ × V î where î is the unit vector in the x-direction.
(d) ∇2 (∇V )
(e) ∇ (∇V · ∇V )
30. Given the function f = y 2 + x3 find the quantities:
(a) Curl of the gradient of f
(b) Laplacian of f
(c) Divergence of the gradient of f
31. Consider the function V (x, y) = x2 + y 2 in two dimensions.
(a) Graph the equipotential lines V (x, y) = 1, V (x, y) = 4, and V (x, y) = 9 on the
xy−plane.
(b) Graph the gradient of V (x, y) on the xy−plane.
(c) What does this gradient represent physically, if V (x, y) is the electric potential
on the xy-plane?
32. Graph the vector fields

V (x, y) = y î + xĵ V (x, y) = −xî − y ĵ V (x, y) = y î + xĵ + z k̂

33. Prove that ∇2 1

= 0.

r

34. Prove: ∇ · (φA) = (∇φ) · A + φ (∇ · A), where φ is a scalar function and A is a vector
function.

35. Prove ∇ · r/r3 = 0, where r is the position vector.

36. If v = ω × r, prove ω = 1
2 curl v where ω is a constant vector.

37. Find ∇φ if (a) φ = ln |r|, (b) φ = 1r .

F1 = F12 + Fext
1 (4.1.2)

107

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 108 — #124

i i

108 Classical Mechanics: A Computational Approach

where F12 is the force on Particle 1 exerted by Particle 2 (the “internal force”), and Fext 1
is the net external force acting on Particle 1. Similarly, Particle 2 experiences a net force:
F2 = F21 + Fext
2 (4.1.3)
Let p1 and p2 represent the linear momenta of the two particles, then Newton’s second law
states that
ṗ1 = F1 = F12 + Fext
1 (4.1.4)
and
ṗ2 = F2 = F21 + Fext
2 (4.1.5)
Now, we define the total momentum vector as P = p1 + p2 , to be the sum of the momenta
of each particle. Then,
Ṗ = ṗ1 + ṗ2 = Fext
1 + F2 =F
ext ext
(4.1.6)
Hence, we see that if there are no total net external forces Fext = Fext
1 + F2
ext then Ṗ = 0,

or to put it another way:

The Law of Conservation of Linear Momentum

If Fext = 0, then P = constant.

The total momentum of a system is conserved if no external forces act on that system. This
is an important law in physics, one that we will use to solve problems in cases where the
external forces add to zero.
Example 4.1: Inelastic collision between two bodies
Consider two objects with masses m1 and m2 and moving with velocities v1 and v2 ,
respectively. The two masses collide and stick together, moving away from the collision as
one object with a velocity v. Ignoring external forces occurring during the collision find
the velocity of the objects immediately after the collision.
Solution:
This type of collision is called a perfectly inelastic collision because the two bodies
are stuck together after the collision. Because we can ignore the external forces during
the collision, we can use conservation of momentum to solve this problem. The total
momentum before the collision is:

Pinitial = m1 v1 + m2 v2

and the total momentum after the collision is:

Pfinal = (m1 + m2 ) v

Note that because the two objects stick together after the collision, both masses have the
same velocity, hence the right-hand side of Pfinal above. Conservation of momentum tells
us that Pinitial = Pfinal and therefore:
m1 v1 + m2 v2
v=
m1 + m2
Notice that the final velocity is the weighted average of the initial velocities, where the
masses of each object serve as the weights for the average. While this problem may seem
to be simple, it is an important application of the conservation of momentum; it is used

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 109 — #125

i i

Momentum, Angular Momentum, and Multiparticle Systems 109

to solve problems in nuclear physics, astrophysics, and accident reconstruction–to name a

few.

Example 4.2: Elastic collision between two bodies

A particle of mass m1 = 0.10 kg moving with a velocity of v0 = 10.0 m/s î collides
with another particle that is at rest and has a mass m2 = 0.20 kg. After the collision,
each particle moves as shown in the figure below, at an angle θ = π/6 rad. The collision is
elastic, meaning that the kinetic energy of each particle is conserved. Find the speed v1 of
mass m1 , and the speed v2 and angle φ for mass m2 after the collision.
m1 v1

v0 θ
m1 φ
m2
m2
v2
Solution:
You may remember from your introductory physics course that in an elastic collision
both the total momentum mv and total kinetic energy 1/2mv 2 are conserved. We will
discuss kinetic energy more in depth in the next chapter. For now, your knowledge from
your introductory physics course will suffice to solve this problem.
Only mass m1 is moving before the collision, and the problem states that the velocity
is in the x-direction. Therefore,
pinitial = m1 v0 î
The momentum after the collision involves both particles. By analyzing the velocity vectors
v1 and v2 into their x and y components. We can write:

pfinal = (m1 v1 cos θ + m2 v2 cos φ) î + (m1 v1 sin θ − m2 v2 sin φ) ĵ

Using conservation of momentum, we equate pinitial = pfinal . In addition, because the

collision is elastic, we also equate the initial and final kinetic energies. Together, the two
conservation laws lead to three equations:

m1 v0 = m1 v1 cos θ + m2 v2 cos φ
0 = m1 v1 sin θ − m2 v2 sin φ
1 1 1
m1 v02 = m1 v12 + m2 v22
2 2 2
The top two equations represent the total momentum in the x and y directions respec-
tively, and the final equation represents the conservation of kinetic energy. In addition,
we know that v0 = 10.0 m/s, m1 = 0.10 kg, and m2 = 0.20 kg. As you might imagine, this
problem can involve a lot of algebra. However, a CAS will be able to assist us with that
algebra. The following is an example of code written in Mathematica, which can be used
to solve this problem.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 110 — #126

i i

110 Classical Mechanics: A Computational Approach

θ = π/6;
v0 = 10;
m1 = 0.1;
m2 = 0.2;

Solve[{m1v0 == m1v1Cos[θ] + m2v2Cos[φ], m1v1Sin[θ] == m2v2Sin[φ], 0.5

m1*v0∧ 2 == 0.5*m1*v1*∧ 2 + 0.5*m2*v2∧ 2}, {v1, v2, φ}]

OUTPUT: Solve::ifun: Inverse functions are being used by Solve, so some solutions
may not be found; use Reduce for complete solution information.

{{v1 → −3.56822, v2 → −6.6056, φ → 3.00613}, {v1 → −3.56822, v2 → 6.6056, φ →

−0.135459}, {v1 → 9.34172, v2 → −2.52311, φ → −1.95894}, {v1 → 9.34172, v2 →
2.52311, φ → 1.18266}}
We begin the program by defining our known variables, the final direction θ of mass
m1 , the initial velocity of m1 (v0 ), and the values of the masses m1 and m2 . The Solve
command gives a warning that some solutions may not be found by Mathematica. This
is not uncommon when using a CAS to solve transcendental equations. However, four
solutions are found. The first three solutions include negative values for v1 and v2 . These
are non-physical results because v1 and v2 are speeds, and speed is the positive magnitude
of the velocity vector. The fourth solution found by Mathematica contains all positive
values, which is in agreement with the problem’s diagram. Finally, we need to remember
that Mathematica, like any calculator, is not respecting significant figures. Therefore, our
final answer is: v1 = 9.3 m/s, v2 = 2.5 m/s, and φ = 1.2 rad. It should be noted, that
although the warning suggests using the command Reduce in this case, the set of solutions
produced by Reduce includes the ones found by Solve.
As we have said before, the availability of CAS programs is not a substitute for knowing
how to do algebra. The reader should verify the solution by hand.

Next, we study the case where three or more particles are involved. The questions arises:
Do we still expect momentum to be conserved when there are many interacting particles?
Let’s begin by examining the net force acting on a collection of N interacting particles. The
net force on the ith particle is: X
Fi = Fext
i + Fij (4.1.7)
j6=i

where as previously stated Fext

i is the net external force acting on the ith particle, and Fij
is the force of interaction between particle j and particle i. The summation term in the
equation above denotes the net internal force acting on the particle. The net force acting on
the system would then be obtained by summing the equation above over the N particles:
N
X N
X N X
X
F= Fi = i +
Fext Fij (4.1.8)
i=1 i=1 i=1 j6=i

Equation (4.1.8) can be rewritten as:

N
X N X
X
F= i +
Fext (Fij + Fji ) (4.1.9)
i=1 i=1 j>i

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 111 — #127

i i

Momentum, Angular Momentum, and Multiparticle Systems 111

If this rewriting is not clear to you, set N = 3 and write out the terms in (4.1.8); you
will find that you can rewrite the sum as a collection of two terms, as shown previously.
Newton’s third law tells us that Fji = −Fij , and therefore the second term in the previous
equation is zero leading to:
XN
Ṗ = F = Fext
i (4.1.10)
i=1

The Law of Conservation of Linear Momentum for a System of N Particles

N
X
If Fext
i =0 then P = constant (4.1.11)
i=1

In other words, the momentum of the system of particles is conserved if there are no external
forces acting on the system, like in the case of two-particle systems.

4.2 ROCKETS
A practical example of conservation of momentum is rocket propulsion. The rocket has a
challenge of moving forward without pushing against something. For example, an automobile
moves forward by turning its wheels so that it pushes against the ground. However, a rocket
in space has nothing to push against. Instead, the rocket solves this problem by ejecting
mass, similar to the recoil of a gun. Before a gun is fired, the bullet and the gun are at rest.
However, once the bullet is fired, it is propelled forward. In order for the total momentum to
be conserved, the gun has to move in the direction opposite that of the bullet. The velocities
of the gun and the bullet are, of course, very different because the gun has more mass than
the bullet does.
We begin by considering a rocket flying horizontally and assume that the rocket is not
experiencing any external forces. The rocket has momentum p(t) = mv at time t, where m
and v are the mass and velocity of the rocket. The momentum of the rocket is measured
by an observer in a reference frame that is at rest relative to the rocket. A short time later
denoted by t + dt, the rocket has expelled a small amount of mass dm0 , which is moving
at a speed u relative to the rocket, as illustrated in Figure 4.1. The speed u is sometimes
called the exhaust speed. The ejection of mass leads to an increase in the rocket’s speed by
a small amount dv, as measured in the rest frame. Therefore, the momentum of the system
immediately after the mass was ejected is:

p(t + dt) = (m − dm0 )(v + dv) + dm0 (v − u) (4.2.1)

where the second term above is the momentum of the ejected mass relative to the rest
frame.

dm' v+dv
m-dm'
u
Figure 4.1: A rocket after ejecting a small amount of mass dm0 at a speed relative to the
rocket, u.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 112 — #128

i i

112 Classical Mechanics: A Computational Approach

Applying conservation of momentum:

p(t) = p(t + dt) (4.2.2)

mv = (m − dm0 )(v + dv) + dm0 (v − u) (4.2.3)
m dv = u dm0 (4.2.4)
m dv = − u dm (4.2.5)

where the last line was obtained by noting that the change in mass of the rocket is
dm = −dm0 . Furthermore, we ignored the term dm0 dv because both dm0 and dv are small,
therefore their product is negligible. Dividing both sides by dt we obtain:

m v̇ = −u ṁ (4.2.6)

The left-hand side of (4.2.6) is Newton’s second law (F = ma), and therefore we see that
the net force acting on the rocket is equal to −u ṁ, which is sometimes called the thrust.
Note that ṁ < 0, so the thrust is positive (and points to the right in Figure 4.1). We can
also write (4.2.6) as:
dm
dv = −u (4.2.7)
m
This equation can be integrated by assuming a constant u. If at t = 0 the rocket’s speed is
v0 and its mass is m0 , we obtain:
m
0
v − v0 = u ln (4.2.8)
m
which gives the velocity of the rocket as a function of its mass m. Equation (4.2.8) tells us
that in order to make a rocket go as fast as possible, engineers need to create rockets with
large exhaust speed u, and with a large initial to final mass ratio m0 /m. The final speed of
the rocket will be obtained once all of the fuel is burned. Rockets with multiple stages that
are ejected can help maximize m0 /m, so that the rocket can achieve even greater speeds.
Next, we study the case of a rocket moving vertically, experiencing the force of gravity.
Our goal is to find an equation which describes the velocity of the rocket. We will use a
coordinate system where the positive y-direction is upward, in the opposite direction of the
force of gravity. In this case, gravity is an external force acting on the rocket, and therefore:

dp
= − mg (4.2.9)
dt
dp = − mgdt (4.2.10)
p(t + dt) − p(t) = − mgdt (4.2.11)
mdv + udm = − mgdt (4.2.12)
dv dm
m +u = − mg (4.2.13)
dt dt
In order to obtain an equation for v, we introduce the constant burn rate α:
dm
α=− (4.2.14)
dt
Using our definition of burn rate, the last line of (4.2.13) becomes:

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 113 — #129

i i

Momentum, Angular Momentum, and Multiparticle Systems 113

dv α
=−g+ u (4.2.15)
dt m
αu
dv = −g + dt (4.2.16)
g m
u
dv = − dm (4.2.17)
α m
Integrating the above equation using v(0) = v0 and m(0) = m0 produces:
g m
0
v − v0 = (m − m0 ) + u ln (4.2.18)
α m
This equation gives the velocity of the rocket as a function of mass m when the burn rate α
is constant. Sometimes (4.2.18) is rewritten by integrating the definition of the burn rate:
m − m0 = −αt. In that case, (4.2.18) becomes:
m
0
v − v0 = −gt + u ln (4.2.19)
m
The advantage of using (4.2.19) is that one can see two terms which affect the rocket’s
motion. The first term, −gt, is the standard kinematics term for a particle in free fall, and
represents the effect of gravity on the velocity. The second term, u ln(m0 /m), is the same
term we obtained for the horizontal motion of the rocket and represents the effect of thrust
on the rocket’s velocity.

4.3 CENTER OF MASS

Problems involving momentum often involve multiple objects, such as in the case of col-
lisions. In situations where the system of interest involves multiple particles or the object
under study is not a point particle, the concept of center of mass becomes useful. Let us
begin by considering a system of N point particles. The location R of the center of mass
for a system of particles is defined as:
Center of Mass for a System of Discrete Particles
N
1 X
R= mi ri (4.3.1)
M
i=1

N N N
1 X 1 X 1 X
X= mi xi Y = m i yi Z= mi zi (4.3.2)
M M M
i=1 i=1 i=1

Here ri P
is the position of the ith particle, which has mass mi . The total mass of the system
is M = mi . Equation (4.3.1) is actually three equations, one for each coordinate of R.
For example, in Cartesian coordinates the position of the ith particle is ri = xi î + yi ĵ + zi k̂
and we can write R = X î + Y ĵ + Z k̂.
The center of mass is the weighted average location of the mass in the system, with
the weights being the masses mi . It is a weighted average, similar to how grades are often
calculated in a course where maybe 50% of your final grade may be based on mid-semester
exams, 20% on homework assignments, and 30% on a final examination. The location of
the center of mass will be closer to the heavier particles. The next example illustrates this
point.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 114 — #130

i i

114 Classical Mechanics: A Computational Approach

Example 4.3: Center of mass of the Earth-Sun system

Compute the location of the center of mass of the Earth-Sun system.
Solution:
We choose a coordinate system such that the center of the Sun is at the origin. We will
also approximate the Earth’s orbit as circular and assume that the plane of the Earth’s
orbit is constant. Therefore, we can work in polar coordinates and only need to find the
distance of the center of mass of the system from the Sun’s center. The distance between
the Earth and the Sun is approximately 1.5 × 1011 m, and the masses of the Earth and
Sun are 5.9 × 1024 kg and 2.0 × 1030 kg, respectively. Therefore, in our coordinate system,
rSun = 0 and rEarth = 1.5 × 1011 m. Inserting these numbers into (4.3.1) gives:

(mEarth ) (rEarth ) 5.9 × 1024 kg 1.5 × 1011 m

= = 4.6 × 105 m.
mSun + mEarth 5.9 × 1024 kg + 2.0 × 1030 kg
The radius of the Sun is approximately 7.0 × 108 m. Therefore, the center of mass of the
Earth-Sun system is well inside the Sun, and practically coincides with the center of the
Sun.

If we want to extend (4.3.1) to continuous distributions of mass, i.e., objects that are not
well-described as point particles, then we need to make some changes. One way to approach
a continuous mass distribution is by thinking of it as a collection of infinitesimal masses
dm. Then the sum over discrete particles in (4.3.1) becomes an integral over infinitesimal
mass elements dm. The center of mass equations then become:

Center of Mass for a Continuous Mass Distribution

1
Z Z
R= rdm M = dm (4.3.3)
M
1 1 1
Z Z Z
X= xdm Y = ydm Z = zdm (4.3.4)
M M M

In the equations above, the total mass is found from M = dm, and like (4.3.1) equation
R

(4.3.3) actually consists of three equations, one for each coordinate. Typically, one does not
integrate (4.3.3) using the mass. Most often the integration is done over the length, area,
or volume of the object, when the object does not have a uniform density.
The following box summarizes how to calculate the total mass M by using the density
of the object in one-, two-, and three-dimensional situations. In the case of one-dimensional
objects, we use a linear mass density λ (mass per unit length in units of kg/m). In the
case of two-dimensional objects, we use a surface density σ (mass per unit area in units of
kg/m2 ), and for three-dimensional objects we use the familiar volume density ρ (mass per
unit volume in units of kg/m3 ).

Evaluating the Mass dm for 1D, 2D and 3D Objects

Z
dm = λdx M = λdx (4.3.5)
Z
dm = σdA M= σdA (4.3.6)

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 115 — #131

i i

Momentum, Angular Momentum, and Multiparticle Systems 115

Z
dm = ρdV M= ρdV (4.3.7)

The next example will demonstrate how to apply (4.3.3) to a continuous mass distribu-
tion.
Example 4.4: Center of mass of an isosceles triangle
A lamina of uniform mass per unit area σ, is shaped into an isosceles
√ triangle shown
below. The triangle has two sides of length a and the base has length 2a. Find the center
of mass of this triangular lamina.
y

a a
dm
dy
dx
x
2a
Solution:
First note that due to the symmetry of the triangle, the center of mass lies along
the y-axis. Computing the center of mass of a mass distribution involves identifying two
components. The first component to identify is the mass element dm, shown in the diagram
of the triangle. Cartesian coordinates are a natural choice of coordinates for this problem.
Therefore, dm = σdxdy will be used for the mass element.
The second component needed to find the center of mass are the limits of the integral in
(4.3.3). The limits of the integral range over the size of the object. We included a horizontal
gray strip near the top of the triangle to illustrate that the width of the triangle varies
with the height, with the strip becoming narrower
√ as it moves up the triangle. The right
side of√the strip is bound by the line y = a/ 2 − x√and the left side is bound by the line
y = a/ 2 + x, because the triangle √ has a height a/ 2. Therefore, we can √ consider the left
end of the strip to be at x = y − a/ 2, and the right end to be at x = a/ 2 − y.
The
R y-coordinate of the center of mass can then be found by computing Y =
1/M ydm:

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 116 — #132

i i

116 Classical Mechanics: A Computational Approach

√ √
1 y=a/ 2 Z a/ 2−y
Z
Y = √ yσdxdy
M y=0 x=y−a/ 2
Z a/√2 a/√2−y
σ
= yx dy

M 0 √
y−a/ 2
√
Z a/ 2
σ a
= 2y
√ − y dy
M 0 2
√
a/ 2
σ ay 2 2y 3

= √ −
M 2 3

0
a3 σ
= √
6 2M
To finish the problem, we need to find the total mass M .

√ √ √
y=a/ 2 Z a/ 2−y a/ 2
σa2
Z Z Z
a
M= dm = σdxdy = σ 2 √ − y dy =
y=0
√
x=y−a/ 2 0 2 2

By using the value for M found above we find that:

a
Y = √
3 2
We can use a CAS like Mathematica to evaluate the integrals, even if the limits are not
constant. Below is an example of how to use Mathematica to solve the problem. Note that
we use the command /. which means replace. The line Y /. m → M asks Mathematica
to take the values stored as M (a √ σ/2) and insert that where the letter m appears in the
2

expression stored in Y (a σ/6m 2).

3
h n o n oi
Y = Integrate σ∗y
m , y, 0, √a
2
, x, y − √a , √a − y
2 2

We will sometimes need to use numerical integration techniques when finding the center
of mass of objects that are irregularly shaped or have complicated density functions. In
this section we will present some basic numerical integration techniques and apply them to
problems involving center of mass.

4.4.1 Trapezoidal Rule

The most basic numerical integration algorithm is called the trapezoidal rule. As the name
suggests, the trapezoidal rule approximates the area under the curve by using a single
trapezoid as shown in Figure 4.2.

f(x)

f(b)
f(a)

x
a b
Figure 4.2: The area under the curve f (x) between x = a and x = b can be approximated
using the trapezoidal rule, which estimates the area using the dashed trapezoid shown in
the figure.

The trapezoidal area under the curve between the points x = a and x = b is found by
taking the sum of the triangle and rectangle shown in Figure 4.2:

i i

122 Classical Mechanics: A Computational Approach

Mathematica’s built-in numerical integration routine NIntegrate, and Python’s numerical

integration routine nquad from the SciPy library.

Example 4.6: The center of mass of a non-uniform isosceles triangle

Consider the triangle in Example 4.3, but now with a non-uniform density σ = 2y,
where the constant has units that produce a density measured in kg/m2 . Compute the
center of mass for the triangle with a = 1 m.
Solution:
In this case, the solution is very similar to that of Example 4.4, but now the triangle
has a non-uniform density. The limits of integration are the same, with a = 1. However, dm
is changed due to the non-uniform density. In this case, dm = σdxdy = 2ydxdy. Because
the density changes only in the y-direction, we still expect that the x-coordinate of the
center of mass is zero. The integrals we need to solve are:
Z y=a/√2 Z a/√2−y
1
Y = √ 2y dxdy
2
M y=0 x=y−a/ 2
Z y=a/√2 Z a/√2−y
M= √ 2ydxdy
y=0 x=y−a/ 2

First, we use Mathematica’s NIntegrate command to solve these integrals. NIntegrate

uses an adaptive algorithm, which recursively subdivides the interval of integration as
needed, instead of choosing the number of divisions ahead of time (as done in the extended
Simpson’s rule). This can be advantageous when the function to be integrated changes
rapidly over some intervals while remaining nearly constant over others. Notice that in the
Mathematica program, we included the result for X to show that it is zero.
a = 1.0;
mass = NIntegrate 2 ∗ y, y, 0, a 2 , x, y − a 2 ,a 2 −y ;
√ √ √

Y = Integrate 2 ∗ y 2 , y, 0, a 2 , x, y − a 2 ,a 2 −y mass
√ √ √

OUTPUT: 0.353553

X = Integrate 2 ∗ y ∗ x, y, 0, a 2 , x, y − a 2 ,a 2 −y mass
√ √ √

OUTPUT: 0.
Next we perform the integration using the numerical integration routine, nquad, from
SciPy’s integrate library. The command nquad uses the command quad over multiple
variables of integration. The command quad also comes from SciPy’s integrate library,
and uses a FORTRAN 77 library called QUADPACK [Piessens et al. (1983)] to evaluate
an integral. The code to perform the integration is shown in Algorithm 2.
Notice that we can include the bounds as a function; this allows us to use the non-
constant limits of integration necessary to describe the sides of the triangle. In addition
to finding Y , we also calculated X to show that X = 0, as expected.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 123 — #139

i i

Momentum, Angular Momentum, and Multiparticle Systems 123

from s c i p y import i n t e g r a t e
import numpy a s np

a = 1.0

d e f sigma ( x , y ) :
return 2.0∗y

def fx (x , y ) :
r e t u r n x∗ sigma ( x , y )

def fy (x , y ) :
r e t u r n y∗ sigma ( x , y )

d e f y bounds ( ) :
r e t u r n [ 0 , 1 / np . s q r t ( 2 . 0 ) ]

d e f x bounds ( y ) :
r e t u r n [ y−1/np . s q r t ( 2 . 0 ) , 1 / np . s q r t (2.0) − y ]

Y = i n t e g r a t e . nquad ( fy , [ x bounds , y bounds ] ) [ 0 ]

X = i n t e g r a t e . nquad ( fx , [ x bounds , y bounds ] ) [ 0 ]
mass = i n t e g r a t e . nquad ( sigma , [ x bounds , y bounds ] ) [ 0 ]

p r i n t ( ’ The x−c o o r d i n a t e i s : ’+ s t r (X/ mass ) )

p r i n t ( ’ The y−c o o r d i n a t e i s : ’+ s t r (Y/ mass ) )

OUTPUT
The x−c o o r d i n a t e i s : 0 . 0
The y−c o o r d i n a t e i s : 0 . 3 5 3 5 5 3 3 9 0 5 9 3 2 7 3 8 4

Algorithm 4: Example of center of mass calculation in Python.

4.5 MOMENTUM OF A SYSTEM OF MULTIPLE PARTICLES

Next, consider a system of N point particles with a total mass M . We begin by calculating
the net force acting on the whole system. We know from Section 4.1 that the internal forces
between particles cancel out and that the net force acting on the system is the sum of
external forces. The net force F acting on the particles is:
N N
X d2 X d2
F= mi r̈i = 2
mi ri = 2 M R = M R̈ (4.5.1)
dt dt
i=1 i=1

where we used the definition of the center of mass from (4.3.1). Therefore, the system of
particles moves like a single particle of mass M , acted upon by the external forces. We can
also compute the net momentum of the system of particles:
N N N
X X d X d
P= pi = mi ṙi = mi ri = M R = M Ṙ (4.5.2)
dt dt
i=1 i=1 i=1

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 124 — #140

i i

124 Classical Mechanics: A Computational Approach

The total momentum of the system is equal to the momentum of the system’s center of
mass. Again, we can think of the system as a single particle of mass M located at the
system’s center of mass. The time derivative of the total momentum gives:

Ṗ = M R̈ = F (4.5.3)

In other words, the total linear momentum of the system is conserved if there are no net
external forces acting on the system.
In summary, we can think of the net external forces as acting on the center of mass of
the system. The results of this section also hold for continuous mass distributions, but the
summations would need to be replaced by integrals.
When it comes to translational motion, the above result allows us to think of the object
or system of particles as a point particle located at the object’s center of mass. The mass of
the point particle would equal the total mass of the object or the total mass of the system
of particles. Since the external forces act on the center of mass, we can simply follow the
motion of the object’s center of mass. For example, an American football is not a point
particle. However, if the football is punted, its center of mass would follow a parabolic
trajectory because the center of mass is being acted upon by the force of gravity, and obeys
the equations established for projectile motion in Chapter 3.

Example 4.7: A falling chain

Consider a chain of length a and uniform density ρ, hanging from one end above a
table at time t = 0, as shown by the dashed line in the figure. At time t = 0, the end of
the chain is released from a height a above the table, and the chain is allowed to fall onto
the table. Find the force the table exerts on the chain as a function of the distance x from
the originally fixed end.

Solution:
The chain can be thought of as a system of particles. The force on the center of mass
of the chain is:
Fnet = M g − FN
where M g is the weight of the chain and FN is the force exerted by the table on the chain.
We have chosen a coordinate system where +x is in the downward direction, so that speed
of the falling chain is ẋ > 0. The mass of the falling part of the chain is ρ(a − x), because
the length of the chain that has yet to fall is (a − x). The momentum of the falling part

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 125 — #141

i i

Momentum, Angular Momentum, and Multiparticle Systems 125

of the chain is then:

P = mv = ρ(a − x)ẋ
We know that Ṗ = Fnet , therefore the time derivative of the momentum is:

Ṗ = ρ(a − x)ẍ − ρẋ2

√
For free fall we know that ẋ = 2gx and ẍ = g. By substituting ẋ and ẍ in the last equation
and collecting terms, we obtain:
Ṗ = ρg(a − 3x)
Finally substituting Ṗ = Fnet = M g − FN and using the total mass of the chain M = ρa,
we find:

ρg(a − 3x) = ρag − FN

FN = 3ρgx
As the distance x increases, the force the table exerts on the chain varies from FN = 0
at time t = 0, up to FN = 3ρga when the chain has fallen completely on the table.

4.6 ANGULAR MOMENTUM OF A SINGLE PARTICLE

Consider a particle of mass m moving with momentum p = mv as shown in Figure 4.3. The
angular momentum of the particle is defined by:

Angular Momentum

` = r×p (4.6.1)

Because the position r of the particle is measured relative to the origin, the angular
momentum ` is also relative to the origin. The linear momentum p does not depend on the
origin, while by contrast the angular momentum vector ` depends on the choice of origin.
The angular momenta of multiple particles can only be compared (and added) if all angular
momenta are measured with respect to the same origin. Further, note that the particle does
not need to be revolving around the point O, in order for the particle to have an angular
momentum relative to O.
The cross product relationship between r and p tells us that the direction of angular
momentum is found using the right-hand rule. In this case, we can imagine taking the
fingers of our right hand and pointing them in the direction of r. Next, we can sweep (or
curl) our fingers in the direction N
of p resulting in the thumb, which points in the direction
of ` , pointing into the page. The symbol in Figure 4.3 represents the angular momentum
vector pointing into the page. If ` pointed out of the page, we would use a circle with a dot
inside to represent the direction of the vector.
If we compute the time derivative of the angular momentum, we obtain:

`˙ = ṙ × p + r × ṗ = m (ṙ × ṙ) + r × ṗ = r × ṗ (4.6.2)

By using Newton’s second law ṗ = F we find:

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 126 — #142

i i

126 Classical Mechanics: A Computational Approach

r ℓ

O x
Figure 4.3: A particle of mass m at a location r relative to the origin O is moving with
momentum p, and has an angular momentum ` = r × p relative to the origin. In this case,
the angular momentum of the particle is into the page.

Newton’s Second Law of Rotation

`˙ = r × F = N (4.6.3)

Therefore, the time rate of change of the angular momentum is the torque, N = r × F.
Like the angular momentum, the torque is measured relative to the origin, the same origin
as the angular momentum. Equation (4.6.3) is sometimes referred to as Newton’s second
law for rotation, even though rotational motion is not needed in order for the particle to
experience a torque, or to have angular momentum.

4.7 ANGULAR MOMENTUM OF MULTIPLE PARTICLES

Next, we consider the total angular momentum (relative to the origin) for a system of
discrete particles. Again, the results are the same for a continuous mass distribution. Figure
4.4 illustrates a system of N particles. The mass of the ith particle in Figure 4.4 is mi , located
at a position ri relative to the origin, and at a position r0i relative to the center of mass
(empty circle in Figure 4.4). Notice that R + r0i = ri , where R is the location of the center
of mass relative to the origin.
We will calculate the total angular momentum (relative to the origin) for the system of
particles in Figure 4.4. Each particle has an angular momentum `i = ri × pi . Therefore the
total angular momentum L for the system is:
N
X N
X
L= ri × pi = ri × mi ṙi (4.7.1)
i=1 i=1

By substituting ri = R + r0i we obtain:

N
X
L= r0i + R × mi ṙ0i + Ṙ (4.7.2)

i=1

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 127 — #143

i i

Momentum, Angular Momentum, and Multiparticle Systems 127

ri'=ri-R

mi
R

x
Figure 4.4: A collection of N particles. Each particle mi is at a location ri relative to the
origin and a position r0i relative to the center of mass. The empty circle is the center of mass
of the system.

By expanding the cross product into four terms:

N
X
L= r0i × ṙ0i + r0i × Ṙ + R × ṙ0i + R × Ṙ (4.7.3)

mi
i=1

Of the four terms above, two are of immediate interest:

N N
!
X X
mi r0i × Ṙ = mi r0i (4.7.4)

× Ṙ
i=1 i=1
N N
!
X d X
mi R × ṙ0i = R× mi r0i (4.7.5)

dt
i=1 i=1

P In each case, the term mi r0i appears. However, mi r0i = mi (ri − R) = mi ri −

P P P P
mi R = M R − M R = 0 from the definition of the center of mass. Therefore, only two
terms survive in (4.7.3), and the total angular momentum is:
N
X
L= r0i × ṙ0i + R × Ṙ

mi
i=1
(4.7.6)
XN N
X
= r0i × mi ṙ0i +R×

mi Ṙ
i=1 i=1

The final result is the following important equation:

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 128 — #144

i i

128 Classical Mechanics: A Computational Approach

Angular Momentum of a System of Particles

N
X
L= r0i × p0i + R × P (4.7.7)

i=1

which says that the total angular momentum of the system is the sum of two terms:Pthe first
term is the sum of the angular momenta of the particles about the center of mass ( r0i × p0i
); the second term is the angular momentum of the center of mass about the origin (R × P).
If we compute the time derivative of L we find,
N
X
L̇ = ri × ṗi (4.7.8)
i=1
which is the net torque N acting on the system (recall that ṗ = F). Next, we will examine
the net torque acting on the system in detail:
N
X N
X
N= ri × ṗi = ri × Fi (4.7.9)
i=1 i=1
where ṗi = Fi is the net force acting on mass mi . This net force acting on mi is the sum of
external and internal forces, therefore:
 
N
X X
N= i +
ri × Fext Fij  (4.7.10)
i=1 j6=i
N
X N X
X
= ri × Fext
i + (ri × Fij ) (4.7.11)
i=1 i=1 j6=i

We will now use the same identity as in working with (4.1.8):

N N X
X X
N= ri × Fext
i + [(ri × Fij ) + (rj × Fji )] (4.7.12)
i=1 i=1 j>i

Using Newton’s third law Fji = −Fij yields:

N
X N X
X
N= ri × Fext + (ri − rj ) × Fij (4.7.13)

i
i=1 i=1 j>i
Note that if the internal forces Fij are central forces, then the vector Fij lies along the
line joining mi and mj . Therefore, for central forces Fij lies along the vector ri − rj , and
(ri − rj ) × Fij = 0. In this case, the second term in (4.7.13) is zero, and the net torque is:
N
X N
X
N= ri × Fext
i = Next
i = Next (4.7.14)
i=1 i=1
In other words, the net torque on the system is equal to the sum of the external torques.
Similar to the net force, the sum of the internal torques is zero (if the internal forces are
central forces).
In summary, similar to the the single-particle case, Newton’s second law for rotation can
be written as,

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 129 — #145

i i

Momentum, Angular Momentum, and Multiparticle Systems 129

Newton’s second law for Rotation

L̇ = Next (4.7.15)

Hence, the net angular momentum of the system is conserved only if the net external
torques are zero. The conservation of angular momentum can then be written as:

The Law of Conservation of Angular Momentum

If Next = 0 then L = constant

Example 4.8: The total angular momentum of a system of particles

Consider two particles each with a mass of 1.0 kg. At time t, Particle 1 is located at
r1 = 3.0 mî + 2.0 mĵ and is moving at a velocity of v1 = 2.0 m/s î. Particle 2 is located at
r2 = −3.0 m î + 2.0 m ĵ and is moving at a velocity of v2 = −2.0 m/s ĵ. Compute the total
angular momentum of this system of particles, with respect to the origin.
Solution:
Using (4.6.1) for Particle 1:

î ĵ k̂
`1 = m1 (r1 × v1 ) = 1.0 3.0 2.0 0 = −4.0 kg m2 /s k̂

2.0 0 0

where we dropped the units in the determinant to simplify the equation. Next, we follow
the same steps for Particle 2:

î ĵ k̂
`2 = m2 (r2 × v2 ) = 1.0 3.0 2.0 0 = 6.0 kg m2 /s k̂

0.0 2.0 0

Finally we add the angular momenta:

L = `1 + `2 = 2.0 kg m2 /s k̂

Note that we could add angular momenta, in this case, because each of the individual
angular momenta were calculated with respect to the same origin.

4.8 CHAPTER SUMMARY

The law of conservation of linear momentum states:
If Fext = 0 then P = constant.
In an elastic collision both the total momentum and total kinetic energy are conserved.
In an inelastic collision, only momentum is conserved.
The momentum of a system of N particles is conserved if there are no external forces
acting on the system:
N
X
Ṗ = F = Fext
i
i=1

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 130 — #146

i i

130 Classical Mechanics: A Computational Approach

If Fext = 0, then P = constant.

A system of N particles has a center of mass located at R:

N
1 X
R= mi ri
M
i=1

The center of mass for a continuous mass distribution is located at:

1
Z
R= rdm
M
Use these expressions to evaluate the mass dm and M for one-, two-, and three-dimensional
objects: Z
dm = λdx M= λdx
Z
dm = σdA M= σdA
Z
dm = ρdV M= ρdV

The trapezoid rule and Simpson’s rule are used for numerically evaluating definite integrals.
For translational motion, we can think of a system of particles as a point particle
located at the object’s center of mass, and of the net external forces acting on the center
of mass of the system. If M is the total mass and P is the total momentum:

Ṗ = M R̈ = F

The angular momentum ` of a single particle is the cross product of position vector r and
linear momentum vector p:
` = r×p
and the torque N is the cross product of position vector r and force vector F:

`˙ = r × F = N

The angular momentum L of a system of particles is the sum of two terms, the angular
momenta of the particles about the center of mass, and that of the center of mass about
the origin (R × P):
XN
L= r0i × p0i + R × P

i=1

where R is the location of the center of mass relative to the origin, and P is the total
momentum. If we differentiate L with respect to time, we find Newton’s second law for
Rotation:
L̇ = Next
From Newton’s second law for rotation, we can write the law of conservation of angular
momentum for a system of particles:
If Next = 0, then L = constant

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 131 — #147

i i

Momentum, Angular Momentum, and Multiparticle Systems 131

21. Consider a chain in a pile on the floor. The chain has a linear mass density of λ = M/L,
where M is the total mass of the chain and L is the total length of the chain. You
grab one end of the chain and pull on it with a constant force F . If there is no friction
between the links of the chain or between the chain and the floor, compute the position
of the end of the chain as a function of time.
22. A rope of length 2L and uniform mass density λ = M/2L (where M is the mass of the
rope) hangs in equilibrium over a peg in a wall. The rope is given a small perturbation
such that it begins to freely fall, but never leaves the peg. Find the velocity of the
rope as it leaves the peg.
23. A chain of length L and uniform mass density λ = M/L hangs in a U shape from the
ceiling, with its two ends close together. At time t = 0, one end of the chain is released.
Find the tension in the chain at the point of suspension, after the released end has
fallen a distance x. You may assume that the end of the chain which is released is
experiencingpfree fall. Plot the ratio of the tension to weight as a function of t/tf ,
where tf = 2L/g, the time it takes for the released end to fall a distance, L.
24. A rope of mass M and length L is stretched out along a table such that a length x0
is hanging over the edge of the table. Neglecting friction, if the rope is released at
rest, show that the time it takes
pfor the rope to completely slide off of the table is
t = T cosh−1 (L/x0 ), where T = L/g.
25. Consider the rope in Problem 24. Find the time it takes for the rope to slide off the
table if µ is the coefficient of friction between the rope and the table.
Sections 4.6 and 4.7: Angular Momentum of a Single Particle and of Multiple
Particles
26. Evidence from many sources suggest that Earth’s day was shorter in its past compared
to its current rotational period. Evidence from ancient corals suggest that 350 million
years ago, the Earth took 385 days to revolve around the Sun [Hadhazy(2001)]. The
period of a year hasn’t changed, the Earth was rotating more quickly. What is the
magnitude of the effective torque slowing the Earth’s rotation over the last 350 million
years?
27. Suppose you tie a ball of mass m to the end of a string and whirl the ball around your
head, so that the ball travels in a circular path whose plane is parallel to the ground.
The string starts with a length L, and the ball completes one revolution in a time T .
Now suppose that as you whirl the ball, you shorten the string at a slow and constant
rate to a length of L/4. What is the ball’s new period of revolution?

28. Consider a rigid rotating object. By breaking the object up into many small pieces
of mass mi , show that the components of the object’s angular momentum parallel to
the axis of rotation can be written as L = Iω, such that:
X
I= mi ri2
where ri is the position of the mass element mi relative to the origin.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 134 — #150

i i

134 Classical Mechanics: A Computational Approach

29. Let L be the angular momentum of a system of particles relative to the origin O. Let
L0 be the angular momentum of the same system with respect to another origin O0 .
Find an equation for L0 in terms of L. You will need to consider the vector rO0 , the
position of O0 relative to O.
30. Prove that the total gravitational torque about the center of mass of a system of
particles is zero.
31. Consider a system of two particles experiencing the force of interaction:

f12 = k [(r2 − r1 )]
where k is a positive constant and f12 is the force Particle 1 exerts on Particle 2. What
is the net internal torque acting on the system? Discuss your result.
32. Consider a system of two particles experiencing the force of interaction:

i i

136 Classical Mechanics: A Computational Approach

where primes are used to distinguish between variables of integration and the limits of
integration. The unit of work, the Joule is 1 J = 1 N·m. Note that work is negative if the
object’s displacement is in the opposite direction of the force. For simplicity, we will focus
only on positive work right now.
Recall from Chapter 2 that if F = F (x), then we can write:
dv
F (x) = mv (5.1.2)
dx
which can be written as:
1

d
F (x) = mv 2 (5.1.3)
dx 2
Notice that the right-hand side of the above equation is the derivative of the quantity
T = 21 mv 2 , called the kinetic energy:

dT
F (x) = (5.1.4)
dx
Separating variables and integrating:
Z T Z x
dT 0 = F (x0 )dx0 (5.1.5)
T0 x0

and using the definition of kinetic energy we find:

Work-Kinetic Energy Theorem

Zx
1 1
mv 2 − mv02 = F (x0 )dx0 (5.1.6)
2 2
x0

Equation (5.1.6) is called the work-kinetic energy theorem, which states that work needs
to be done by the force F (x) in order to change the kinetic energy of a system. For example,
the force of gravity does work on a falling object, and its kinetic energy increases.
Additional insight into the nature of work can be obtained from (5.1.6). If the change
of the particle’s kinetic energy ∆T is positive, then positive work is being done by F and
the system is gaining kinetic energy. If however ∆T < 0, then F (x) is doing negative work
and the particle is losing kinetic energy.
Next, we will add another restriction to F (x). If the work done by the F (x) is indepen-
dent of the path taken by the particle from x0 to x, then we say that force is conservative.
In a later section, we will study forces in two and three dimensions and develop a deeper
mathematical formalism for conservative forces. However, in one dimension all forces of the
form F = F (x) are conservative. Let’s consider the force of gravity exerted by an object
near the Earth’s surface. In this case, the force is the weight W = −mg ĵ, where the negative
sign is introduced to denote that the weight points downward, in the negative y-direction.
The work done by gravity is,
Z y
W= −mgdy 0 = −mg(y − y0 ) (5.1.7)
y0

where we used y for vertical displacements. Notice that the work done by gravity depends
only on the end points of the particle’s motion.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 137 — #153

i i

Energy 137

Consider a falling rock (in a vacuum, of course!). We noted earlier that the rock will
increase its kinetic energy as it falls. Where does that kinetic energy come from? According
to (5.1.6), the force of gravity does work on the rock and (5.1.7) tells us how much work
is done by gravity. Another way of thinking of the process is to say that during the fall,
the rock’s energy is changing from one form into another. While the kinetic energy is the
energy associated with the motion of the rock, there is an additional energy associated
with the configuration of the rock-Earth system. In this case, the important parameter
is the rock’s height above the ground. In fact, (5.1.7) supports the idea of configuration-
dependent energy. The term y − y0 in (5.1.7) says that the work done by gravity depends
on how the distance between the Earth’s surface and the rock changes during the motion.
That change of distance is a change in the configuration of the rock-Earth system. The
energy associated with the configuration of the rock-Earth system, or any system, is called
the potential energy. The potential energy function V (x) in one-dimension is defined by:

dV (x)
F (x) = − (5.1.8)
dx
Notice that (5.1.8) says that forces act so as to decrease the potential energy along x.
Hence, in our rock-Earth example, the force of gravity is decreasing the rock’s potential
energy and increasing the rock’s kinetic energy. Furthermore, this example illustrates that
forces transfer one form of energy into another, in this case, potential energy transferring
to kinetic.
To find the potential energy for a given force, F(x), we need to integrate the force:
Z x
V (x) − V (x0 ) = − F (x0 )dx0 (5.1.9)
x0

For example, the gravitational potential energy is:

Z y
V (y) − V (y0 ) = − −mgdy 0 = mg (y − y0 ) (5.1.10)
y0

If we choose y0 = 0 (and V (y0 ) = 0) then we recover the familiar equation, V = mgy.

Equation (5.1.9) can be rewritten as W = −∆V and tells us that potential energy also
measures the ability of the system to do work. By inserting (5.1.8) into (5.1.6) we get:
Zx
1 1 dV (x0 )

mv 2 − mv02 = T − T0 = − dx0 = V (x0 ) − V (x) (5.1.11)
2 2 dx0
x0

or, after simplification:

Conservation of Mechanical Energy

T + V (x) = T0 + V (x0 ). (5.1.12)

This equation is known as the conservation of mechanical energy, stating that the sum
of the kinetic and potential energies, called the total mechanical energy, is always constant
if only conservative forces act on the system. Conservative forces conserve energy!
Conservative forces are very useful in physics as they allow us to work with a scalar
quantity, the mechanical energy E = T + V . Mechanical energy is a useful way of solving
physics problems because instead of needing to keep track of the magnitude and direction

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 138 — #154

i i

138 Classical Mechanics: A Computational Approach

of vectors, as one needs to do with forces, all one needs to do is keep track of how the value
of the different types of energies are changing as the particle moves. Looking at (5.1.12), we
see that solving problems using conservation of energy requires only that we compare one
state of the system to another. For example, suppose we drop a rock starting at rest from
a height of 5 m above the ground and we want to know the kinetic energy of the rock after
it fell 2 m. All we need is to know is the initial kinetic energy (which is zero) and the initial
and final potential energies in order to obtain the final kinetic energy (and thus the speed).
We don’t need information about the state of the system during the actual fall!
Equation (5.1.12) also tells us that if the potential energy of the system decreases, its
kinetic energy must increase and vice versa, in order to maintain the equality of both sides
of the equation. The ability to interpret the meanings behind equations such as (5.1.8) and
(5.1.12) is an important skill for physicists.

Example 5.1: Simple harmonic motion and conservation of energy

Consider a mass m on a spring with spring constant k performing simple harmonic
motion with amplitude A. The force is given by Hooke’s law F = −kx where x is the
displacement from equilibrium.
(a) Find the total mechanical energy in terms of A.
(b) Find an expression for the turning points in the motion, the points at which the
velocity v of the particle is zero and the mass turns around, as a function of the total
mechanical energy E and spring constant k.
Solution:
(a) We will solve this problem using energy conservation. In order to use energy conser-
vation, we need to first find the potential energy associated with the spring. The potential
energy associated with the force of the spring is called the elastic potential energy and can
be found using:
1
Z x
V (x) − V (x0 ) = − (−kx0 )dx0 = k x2 − x20 (5.1.13)

x0 2
Therefore, we can write the elastic potential energy as V (x) = 1/2kx2 . When the mass
reaches the amplitude of oscillation, x0 = A, the mass has traveled as far from equilibrium
as it can and comes to a stop. Therefore, v0 = 0 and (5.1.12) becomes:

T0 + V (x0 ) =T + V (x)
1 1 1
0 + kA2 = mv 2 + kx2
2 2 2
Therefore, using:
1 1 1
E = mv 2 + kx2 = kA2 (5.1.14)
2 2 2
we find that the total mechanical energy in terms of A is E = 21 kA2 .
(b) At the turning points xTURN the velocity vTURN = 0 so (5.1.14) gives:
1 1
E = 0 + kxTURN = kA2
2 2
Therefore, the expression for the turning points in the motion of a simple harmonic
oscillator is:
xTURN = ±A = 2E/k (5.1.15)
p

In the next section, we will discuss turning points in more detail.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 139 — #155

i i

Energy 139

What happens if F (x) is not conservative? Friction is an example of a nonconservative

force, however we can still use (5.1.1) to find the work done by friction. Recall from your
introductory physics course that the force of friction F = µFN , where µ is the coefficient
of kinetic friction and FN is the normal force, which in this case is constant in magnitude.
The work done by friction is then:
Z x
Wnc = −µFN dx0 = −µFN (x − x0 ) (5.1.16)
x0

where we used Wnc to denote that the work is done by a nonconservative force. Note that
we explicitly included the minus sign in (5.1.16) because the force of friction opposes the
direction of motion, and we consider the displacement to be in the positive direction.
Consider the case of a particle, again constrained to move along a line, experiencing
multiple forces, both conservative and nonconservative. The net work Wnet is the total
work done by all of the forces. The net work can be written as Wnet = Wc + Wnc , where Wc
is the work done by the conservative forces and Wnc is the work done by nonconservative
forces. In regards to the work above, Wnet = ∆T and Wc = −∆V . Therefore,

∆T =Wnc − ∆V (5.1.17)
Wnc =∆T + ∆V (5.1.18)
Wnc =∆E (5.1.19)

or, in other words, the work done by nonconservative forces is equal to the change of the
system’s mechanical energy. So in cases where conservative forces act on the system, kinetic
energy gets exchanged for potential energy and vice versa. However, when nonconservative
forces act on the system, then mechanical energy can be lost or gained. An object falling
through air loses potential energy to both kinetic energy and work done by friction heating
the object and the surrounding air. However, if we account for the heat, we will find that
the total energy is conserved.
In the next section, we will study how to use potential energy to describe the motion of
an object.

5.2 POTENTIAL ENERGY AND EQUILIBRIUM POINTS IN

ONE-DIMENSIONAL SYSTEMS
An equilibrium point is a point in the system’s motion where both the net force and net
torque are equal to zero. Since in this chapter we are not dealing with rotational motion, we
will focus only on the condition that F net = 0. Knowing the location of equilibrium points in
a system can give us insight into the system’s behavior. The location of equilibrium points
can be found using the potential energy. Equation (5.1.8) tells us that a particle is in an
equilibrium state if:
dV
=0 (5.2.1)
dx
x0

Hence, the local maxima or minima of the potential energy gives the location of the system’s
equilibrium points, x0 . As we will see, the plot of V (x) can be used to describe the motion
of a particle.
Physical systems can be in three types of equilibria: stable, unstable, and neutral. An
example of a stable equilibrium is shown at the position near x = 0.7 mm in Figure 5.1.
This corresponds to values for which the potential energy V (x) is a local minimum. In

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 140 — #156

i i

140 Classical Mechanics: A Computational Approach

other words, d2 V /dx2 > 0 at the equilibrium position. In the case of stable equilibrium,
the forces acting on the particle will tend to restore the equilibrium. The restoring nature
of the force can be seen in the graph of V (x), by looking at the slope of V (x) on either
side of the equilibrium point. Note that the sign of the force is opposite of the sign of the
potential energy’s slope. Hence a displacement to the right of the equilibrium, there V (x)
has a positive slope, is met with a negative (left pointing) force. Similarly, to the left of
the equilibrium, V (x) has a negative slope, and therefore a particle displaced to the left of
equilibrium experiences a positive (right pointing) force.
An unstable equilibrium is shown at x = 0 in Figure 5.1. This corresponds to values for
which the potential energy V (x) is a local maximum, or in other words, d2 V /dx2 < 0 at
the equilibrium position. This means that if the particle is displaced an arbitrarily small
distance from the equilibrium state, the force causes it to move even farther away. By
studying the slope of V (x) on either side of the equilibrium, we can see that the particle
experiences a positive (rightward) force when displaced to the right of equilibrium, and a
negative (leftward) force when displaced to the left of the equilibrium position, hence the
force acts to move the particle farther from equilibrium.

1.00

0.95
Potential V(x)

0.90

0.85

0.80

0.75
-1.0 -0.5 0.0 0.5 1.0
position x (mm)

Figure 5.1: A graph of a potential energy function which shows stable equilibrium points
near x = 0.7 and x = −0.7 and an unstable equilibrium at the origin (x = 0).

A final type of equilibrium not shown in Figure 5.1 is called neutral equilibrium. A neutral
equilibrium corresponds to situations for which the potential energy V (x) is a constant, and
hence its derivative is zero. In the case of neutral equilibrium, the system will tend to remain
in equilibrium if displaced by a small amount.

Example 5.2: Classifying the equilibrium position of a system

Consider the object shown in the following figure. The object consists of a hemisphere
with a cylinder on top. The center of the hemisphere is at y = R, and when the object
is upright, the center of mass of the object is located a height y = h above the point of
contact with the ground. The object is tipped an angle θ with respect to the vertical. Find
the potential energy of the object when it is tilted. Show that if h > R, then the upright
equilibrium position is unstable.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 141 — #157

i i

Energy 141

CM
h-R
h
H
R θ R

Solution:
The figure on the left shows the object in its equilibrium state (upright). The figure
on the right shows the object tilted an angle θ from the vertical. The vertical line passing
through the center (circle) and center of mass (square) denotes the equilibrium position.
From the second diagram we can see that the height of the center of mass is H = R + (h −
R) cos θ. Therefore, the potential energy is:

V = mg [R + (h − R) cos θ] (5.2.2)

Next, to get the stability of the equilibrium state, we take two derivatives of V :

dV
= − mg(h − R) sin θ (5.2.3)
dθ
d2 V
= − mg(h − R) cos θ (5.2.4)
dθ2
Notice that the first derivative dV /dθ correctly identifies θ = 0 as an equilibrium con-
dition, since sin θ = 0 at θ = 0. Furthermore, if θ = 0 is inserted into the second derivative,
we find that the second derivative is negative when h > R. Therefore, the equilibrium at
θ = 0 is unstable when h > R. Additionally, the upright position is stable if h < R, since
in that case the second derivative is positive.

Energy can be used to get a qualitative understanding of a particle’s motion. Let’s

consider a particle whose potential energy V (x) is that shown in Figure 5.2. We can imagine
the particle’s motion being similar to a ball rolling along a “frictionless track” in the shape of
V (x). For example, suppose we started the particle at rest at x = x1 . This would be similar
to holding the ball on the “V (x) track” at x1 . If we released the ball, we would expect it
to roll along the track speeding up as it approaches the local minimum, then slowing down
and coming to a stop at x = x2 , where it would reach its maximum allowed height, V (x2 ).
The ball would then turn around and head back toward x1 and repeat the motion. The
points x1 and x2 are called turning points.
While the “frictionless track” described in the above paragraph is useful for visualizing
the particle’s motion, we will now examine the motion between the turning points x1 and x2
in more detail. Again, consider the case where the particle starts at rest at the point, x = x1 ,
in Figure 5.2. At x = x1 , the particle has energy E1 and all of the energy is in the form
of potential energy. When released, a force to the right (V (x) has a negative slope at x1 )
pushes the ball to the stable equilibrium (local minimum). At the stable equilibrium near
x = 0.5, the particle would have its highest velocity because it has reached its lowest potential

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 142 — #158

i i

142 Classical Mechanics: A Computational Approach

Potential V(x)
50

40 x3 E2
Out[ ]=

30
x1 x2
20
E1
10
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Position x
Figure 5.2: A potential energy function demonstrating turning points.

energy. Past the equilibrium point, the force begins to oppose the particle’s velocity (note
that V (x) has a positive slope in this region). Because the potential energy is increasing, the
particle’s kinetic energy would decrease in order to conserve the total mechanical energy. At
x = x2 , the particle has lost all of its kinetic energy and comes to a stop. However, it is still
experiencing a leftward force, and so the particle begins to move back towards x1 , where it
will once again come to a stop and move back towards x2 . We see that the resulting motion
is an oscillation between the turning points, x1 and x2 . Both methods of description, the
“frictionless track” and the more detailed analysis in this paragraph, tell us that a particle
with an initial position of x0 = x1 and initial velocity v0 = 0, will oscillate about the stable
equilibrium reaching an amplitude of x1 to the left of equilibrium and x2 to the right of
equilibrium.
We can, of course, use mathematics to describe this motion. Using energy conservation,
E = T + V = mv 2 /2 + V (x), we can solve for the velocity as a function of position:

2
r
v=± (E − V (x)) (5.2.5)
m
For a particle with energy E = E1 , we can see from Figure 5.2 that V (x1 ) = E1 and V (x2 ) =
E1 and therefore, using (5.2.5), the velocity is zero at the turning points, x = x1 and x = x2 .
Furthermore, for x > x2 and x < x1 , the velocity is a complex number because V (x) > E in
those regions. Complex velocities are not physical, and this means that the particle does not
have enough energy to access those regions. For a particle with E = E1 , the regions x < x1
and x > x2 are called classically forbidden regions. According to the physics of classical
mechanics, the particle cannot be in classically forbidden regions because it doesn’t have
enough energy to access those regions. However, such regions are accessible in quantum
mechanical systems.
Using v = dx/dt, we can find the position by integrating (5.2.5):
Z x
dx0
t − t0 = q (5.2.6)
x0 2
m (E − V (x 0 ))

Equation (5.2.6) can be a difficult way to solve for x(t) in closed form, for all but the most
simple of potential energy functions.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 143 — #159

i i

Energy 143

We can repeat our qualitative analysis for the motion of a particle starting at rest at the
unstable equilibrium x3 in Figure 5.2. In that case, if the particle is perturbed to the right,
it increases its velocity as the potential decreases. The graph ends at x = 3, so we cannot
tell what will happen to the particle for x > 3. If the particle is deflected to the left, then it
will increase in speed until it passes through the stable equilibrium, after which it will slow
down until it reaches x ≈ 0.1, before turning around. Mathematically, the particle will then
reach the unstable equilibrium coming to a stop.
Notice that both Figures 5.1 and 5.2 have local minima. The motion around local min-
ima corresponds to an oscillatory motion, if the energy of the particle is low enough. Let
us understand why that is the case using mathematics. Consider a particle near a stable
equilibrium x = x0 , hence, the equilibrium position x0 needs to be a local minimum of
V (x). Next, we ask the question, what is the mathematical form of V (x) near the stable
equilibrium point? In order to obtain the mathematical form of V (x) near a stable equi-
librium, x0 , we will need to perform a Taylor expansion in terms of x around an energy
minimum (x = x0 ). A Taylor expansion gives a polynomial approximation of a function near
a particular point. The general form of the Taylor expansion is:
1
V (x) = V (x0 ) + (x − x0 )V 0 (x0 ) + (x − x0 )2 V (2) (x0 ) + O(x − x0 )3 (5.2.7)
2
where O(x − x0 )3 indicates terms of the order of (x − x0 )3 and higher, and the notation
V 0 (x0 ) and V (2) (x0 ) indicate the first and second derivatives of V evaluated at x = x0 ,
respectively. If we consider the case where the displacement from equilibrium is small, then
x − x0 is small and terms O(x − x0 )3 and higher are negligible. Because V (x0 ) is a minimum,
V 0 (x0 ) = 0, so the linear term drops out:
1
V (x) = V (x0 ) + (x − x0 )2 V (2) (x0 ) + O(x − x0 )3 (5.2.8)
2
The constant term V (x0 ) is arbitrary and thus may also be dropped. This is equivalent to
saying that we can set the zero point of the energy to be any value, hence we set V (x0 ) = 0.
Or to put it another way, constant energy terms do not affect the particle’s motion, because
the force acting on the particle is the derivative of the potential energy, and therefore
constant terms do not contribute to the force. With all of this taken into account, we obtain
the form of V (x) for the simple harmonic oscillator :
1 1
V (x) ≈ (x − x0 )2 V (2) (x0 ) = k(x − x0 )2 (5.2.9)
2 2
where we set k = V (2) (x0 ). Hence, we find that Hooke’s Law appears any time we are
studying small amplitude oscillations about a stable equilibrium. Thus, given an arbitrary
potential energy function V (x) with a local minimum (and a non-vanishing positive second
derivative), we can use the solution to the simple harmonic oscillator found in Example
2.7 to provide a solution for small perturbations around the equilibrium point. As the
perturbations get larger, however, we need to include higher order terms in the Taylor
expansion of V (x), and Hooke’s Law no longer applies. These so-called nonlinear oscillators
can be very difficult to work with. We will study nonlinear oscillators in Chapter 13.
It is worth mentioning that (5.2.9) gives us a method of finding the frequency of small
oscillations about the stable equilibrium. Recall thatpthe frequency of oscillations for a
particle trapped in a Hooke’s Law potential is ω = k/m. Therefore, the frequency of
small oscillations about a stable equilibrium x0 can be found from:
r r
k V (2) (x0 )
ω= = (5.2.10)
m m

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 144 — #160

i i

144 Classical Mechanics: A Computational Approach

Example 5.3: The Lennard-Jones potential

The potential energy of the interaction between a pair of neutral atoms or molecules
is often described by the Lennard-Jones potential function, also referred to as the 6-12
potential:
σ 12 σ 6
V (x) = 4 −
x x

where σ, , x, and V are in appropriate SI units. Setting σ = 1 and = 1/4, determine the
type of equilibrium possible in this potential, and obtain a parabolic approximation of
this potential near the equilibrium point.

Solution:
We find the equilibrium point by setting the derivative of V (x) equal to zero and
solving for x:
dV (x)
dx = 12x−13 − 6x−7 = 0

x = 21/6 = 1.122

We can easily verify that this is a stable equilibrium point by examining the second
derivative. We can use Mathematica to expand the function V (x) around the equilibrium
point, and plot both the series approximation and the potential. The command NSolve is
used to find the equilibrium point x0 = 1.122, and the command Series is used to obtain
the first two terms of the series expansion:
V (x) = −0.25 + 7.143(x − x0 )2

V = 1
x12
− x16 ;

solution = NSolve[D[V, x] == 0&&x > 0, x];

xo = x/.solution[[1]];
Print[“Equilibrium point = ”, xo]
OUTPUT: Equilibrium point = 1.12246

expansion = Series[V, {x, xo, 2}]//Normal//Chop;

Print[“Series expansion, V(x) = ”, expansion]
OUTPUT: Series expansion, V(x) = − 0.25 + 7.1433(−1.12246 + x)2

Plot[{V, expansion}, {x, 0.95, 2}, Frame->True,

FrameLabel → {“position, x”, “potential, V(x)”}, BaseStyle → {FontSize → 18},
PlotRange → {All, {−0.3, 0.5}}, PlotStyle → {{Normal, Black}, {Black, Dashed}},
PlotLegends → {“Lenard-Jones”, “Approximation”}, ImageSize → Large]

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 145 — #161

i i

Energy 145

0.4
Lenard-Jones
Approximation

potential, V(x)
0.2

Out[ ]=

0.0

-0.2

1.0 1.2 1.4 1.6 1.8 2.0

position, x

In the plot above, the dashed line is the series approximation near the stable equilib-
rium, and the solid line is the Lenard-Jones potential. Notice that the approximation fails
to provide a reliable estimate of the potential energy the farther the particle moves from
the equilibrium position.
The code above introduces some new Mathematica commands. The double ampersand,
&& is Mathematica’s AND operator. It tells Mathematica to solve for when V 0 (x) = 0
AND x > 0. The commands Normal and Chop, are used to turn the series expansion into
a result that can be used for further calculations.

Example 5.4: Motion near a stable equilibrium point

Consider the potential energy illustrated in Figure 5.2, V (x) = a/x + bx2 − cx3 , where
a = 4 Jm, b = 25 J/m2 , and c = 8 J/m3 . Numerically solve for the position as a function of
time for a mass of m = 1 kg whose initial conditions are x(0) = 0.21 m (very close to x1 in
Figure 5.2) and ẋ(0) = 0. Describe the resulting motion by using Euler’s method.
Solution:
We begin by differentiating V (x) with respect to x in order to obtain the force F (x):

dV a
F (x) = − = 2 − 2bx + 3cx2
dx x
The next step is to use Newton’s second law in order to create a second-order ODE,
which can be solved for x(t):
a
mẍ = 2 − 2bx + 3cx2
x
This can be reduced to two first-order equations using the variable, y ≡ ẋ:

ẋ =y
1 a
ẏ = − 2bx + 3cx2
m x2
Next, we insert the known values of a, b, and c, and m in order to use Euler’s method
to solve for x(t) and y(t) for the given initial condition: x(0) = 0.21 and y(0) = 0. Below
is a Python implementation, where the algorithm needed to generate the graph of x(t) is
removed for brevity.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 146 — #162

i i

146 Classical Mechanics: A Computational Approach

m, a , b , c = 1 , 4 , 2 5 , 8
x0 = 0 . 2 1 #i n i t i a l p o s i t i o n
y0 = 0 #i n i t i a l v e l o c i t y

def force (x ) :
r e t u r n a/x ∗∗2 − 2∗b∗x + 3∗ c ∗x ∗∗2

x , y = [ x0 ] , [ y0 ]
dt = 0 . 0 0 0 1
t = np . a r a n g e ( 0 , 2 , dt )

f o r i i n r a n g e ( 1 , l e n ( t ) ) : #Euler ’ s method
x . append ( x [ i −1] + dt ∗y [ i −1])
y . append ( y [ i −1] + dt ∗ f o r c e ( x [ i − 1 ] ) )

1.0

0.9

0.8

0.7
position (m)

0.6

0.5

0.4

0.3

0.2
0.0 0.5 1.0 1.5 2.0
time (s)

The graph of x(t) is below the Python program. Notice that the motion is periodic,
oscillating between points x = 0.21 and x = 0.95, as suggested in Figure 5.2.

5.3 WORK AND LINE INTEGRALS

Next, we extend the concept of work to higher dimensional systems, i.e., cases where the
force F = F(r). In this case, the work done by the force F moving an object between points
r1 and r2 is generalized from (5.1.1) to be:
Work
Z r2
W (r1 → r2 ) = F · dr (5.3.1)
r1

Notice the dot product between the force F and the infinitesimal displacement dr in
(5.3.1). The dot product tells us that work calculations involve only the component of the
force parallel to the displacement, i.e., the component of the force perpendicular to the
displacement does no work. Note that (5.3.1) is also applicable to one-dimensional systems,
and we no longer need the restriction that F is in the same direction as dr.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 147 — #163

i i

Energy 147

The right hand side of (5.3.1) is a line integral. In other words, the integral is evaluated
along a specific path. The value of W can depend on the path taken by the particle. It is
common to rewrite (5.3.1) including the path C followed by the particle:
Z
W= F · dr (5.3.2)
C

In order to calculate integrals of this type, one can break them up into component integrals,
by writing out the dot product of the components of the force vector F = Fx î + Fy ĵ + Fz k̂
and the differential dr =dxî + dy ĵ + dz k̂, as follows:
Z Z Z
W= Fx dx + Fy dy + Fy dy (5.3.3)
C C C
and then evaluate the individual integrals along the path C. The next several examples
demonstrate how to calculate path integrals.

Example 5.5: Calculating the work done by a force along a straight line path
Find the work done by the force F = xy î − y 2 ĵ along the straight line path connecting
the origin to the point (2, 1) as indicated in the figure below.

where we used the trigonometric identity cos(2θ) = cos2 θ − sin2 θ. We can use the infinites-
imal displacement in polar coordinates, dr = r̂dr + θ̂rdθ, but this simplifies to dr = rdθθ̂,
because the particle is not moving radially away from the origin (i.e., r = 1 and dr = 0).
After inserting F and dr into (5.3.1) and setting r = 1, we obtain:
Z π/2
W= (−2 cos θ sin2 θ)dθ = −2/3
0

5.4 THE WORK-KINETIC ENERGY THEOREM, REVISITED

Recall that in one dimension, the kinetic energy T = 1/2mv 2 ; in two and three-dimensions
the kinetic energy takes on the form:
1 1
T = mv 2 = mv · v (5.4.1)
2 2
If we compute the time derivative of T, we get:
dT 1
= m (v̇ · v + v · v̇) = mv̇ · v = F · v (5.4.2)
dt 2

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 150 — #166

i i

150 Classical Mechanics: A Computational Approach

dT = F · (vdt) (5.4.3)
dT = F · dr (5.4.4)

After integrating this equation from r0 to r, we obtain:

The Work-Kinetic Energy Theorem

Z r
T − T0 = F · dr (5.4.5)
r0

where T0 is the value of the kinetic energy at position r0 . Equation (5.4.5) is the work-kinetic
energy theorem and is similar to (5.1.6) derived for one-dimensional systems.

5.5 CONSERVATIVE FORCES AND POTENTIAL ENERGY

Recall that for one-dimensional systems, we defined a conservative force F (x), such that
the work done by F (x) to move a particle between two points x0 and x, is independent of
the path taken from x0 to x. This can be easily generalized to two- and three-dimensional
systems. A conservative force is one that depends only on a particle’s position F(r) and
the work done by F(r) to move a particle from positions r0 to r is independent of the path
taken between the points. We can then define the potential energy to be:
Z r
V (r) = −W (r0 → r2 ) = − F(r0 ) · dr 0 (5.5.1)
r0

where we have set a reference point r0 such that V (r0 ) = 0. The above definition works for
our purposes because we have a conservative force. If the force was not conservative, then
the integral in (5.5.1) would also depend on the path taken by the particle and would not
be simply a function of r.
Now suppose that a conservative force, F(r), moves a particle from a point r0 to the
point r2 along two different paths. One path goes directly from the reference point r0 to
r2 , while the other path goes from r0 to r1 and then from r1 to r2 . Because the work is
conservative, the work done along the two paths is the same:

W (r0 → r2 ) = W (r0 → r1 ) + W (r1 → r2 ) (5.5.2)

Solving for W (r1 → r2 ):

W (r1 → r2 ) =W (r0 → r2 ) − W (r0 → r1 ) (5.5.3)

W (r1 → r2 ) = − V (r2 ) + V (r1 ) (5.5.4)

Or, the work done by the conservative force F(r) is related to the potential energy
through the formula:

W (r1 → r2 ) = − [V (r2 ) − V (r1 )] = −∆V (5.5.5)

Combining equation (5.5.5) with (5.5.1), we see that the work done by a conservative force
along a closed-loop path must be equal to zero, since the change of potential energy along
that path is zero.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 151 — #167

i i

Energy 151

Combining the Work-Kinetic Energy Theorem (5.4.5) with (5.5.1), produces:

Z r2
∆T = F(r0 ) · dr 0 (5.5.6)
r1
∆T =W (r1 → r2 ) (5.5.7)
∆T = − ∆V (5.5.8)

and therefore:
∆(T + V ) = 0 (5.5.9)
which tells us that the total mechanical energy E = T + V remains constant as the particle’s
position changes from r1 to r2 .
In the case where nonconservative forces work on the system, (5.1.19) still applies for
two- and three-dimensional systems. However, in this case, (5.4.1) is used to compute T ,
(5.5.1) is used to compute V , and the work done by the nonconservative force Fnc (r) is
found using: Z r2
Wnc = Fnc (r0 ) · dr0 (5.5.10)
r1

We have seen that the potential energy V (x) can be found from the conservative force F,
by using (5.5.1). Now suppose we know V and want to find F. We know from our discussion
of one-dimensional systems that F = −dV /dx; however, how does that generalize to higher
dimensions? To derive the relationship between F and V in three dimensions, we begin by
considering an infinitesimal displacement r → r + dr over which a conservative force F(r) is
acting on a particle. The work done by that force is:

dW = F(r) · dr = Fx dx + Fy dy + Fz dz (5.5.11)
where we are working in Cartesian coordinates for simplicity, but any other coordinate
system can be used. Continuing with the same displacement r → r + dr, the work done by
a conservative force can also be computed as:

dW = − dV (5.5.12)
= − [V (r + dr) − V (r)] (5.5.13)

∂V ∂V ∂V
= − V (r) + dx + dy + dz + · · · − V (r) (5.5.14)
∂x ∂y ∂z

∂V ∂V ∂V
=− dx + dy + dz (5.5.15)
∂x ∂y ∂z

where r = xî + y ĵ + z k̂ , and in (5.5.15) we kept only linear terms in the multivariable Taylor
series. Comparing the coefficients of dx, dy, and dz in Equations (5.5.11) and (5.5.15), we
find:
∂V ∂V ∂V
Fx = − Fy = − Fz = − (5.5.16)
∂x ∂y ∂z

In other words, we can equate the first equation in (5.5.11) with the last equation in (5.5.15)
to show that the force is related to the potential energy through the gradient.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 152 — #168

i i

152 Classical Mechanics: A Computational Approach

Figure 5.3: Surfaces of constant gravitational potential energy and the direction of gravita-
tional force, Fg .

Potential energy and force in three dimensions

→
− ∂V ∂V ∂V
F = −∇V (x, y, z) = − î + ĵ + k̂ (5.5.17)
∂x ∂y ∂z
Z
V = − F · dr (5.5.18)

Notice the minus sign in (5.5.17). The gradient points in the direction of the most rapid
increase in a scalar function. In the case of (5.5.17), we see that forces point in the direction
of the most rapidly decreasing potential energy. Consider the gravitational potential energy
as illustrated in Figure 5.3. Points at the same altitude all have the same potential energy.
We can think of the dashed lines in Figure 5.3 as surfaces of constant potential energy.
As the altitude increases, the potential energy increases. Hence, the direction of the most
rapid increase in V is upward i.e., perpendicular to the ground (shown as the rectangle).
Because Fg = −∇V, the force vector points along the direction of the most rapid decrease
in V , which is downward.
A conservative force F can be written as the gradient of the potential energy. From
Chapter 3, we recall that the curl of a gradient of any scalar function V is always zero:

∇ × F = −∇ × ∇V = 0 (5.5.19)

Therefore, one method to identify whether or not a force is conservative is to compute its
curl. A force F will be conservative if and only if its curl is zero:

Conservative Force

∇×F = 0 (5.5.20)

Example 5.8: Proving a force is nonconservative

Show that the force F = xy î − y 2 ĵ from Examples 5.5, 5.6, and 5.7 is not conservative.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 153 — #169

i i

Energy 153

Solution:
Calculate the curl of the force:

∂Fz ∂Fy ∂Fx ∂Fz ∂Fy ∂Fx
∇×F = − î + − ĵ + − k̂
∂y ∂z ∂z ∂x ∂x ∂y

with Fx = xy, Fy = −y 2 , Fz = 0 this becomes:

∇ × F = (0 − 0) î + (0 − 0) ĵ + (0 − x) k̂ = − xk̂ 6= 0

Hence the force is nonconservative, as we expected based on the results of Examples

5.5, 5.6, and 5.7.

Example 5.9: Proving a force is conservative and finding the associated poten-
tial energy
Consider the force F = −k y î + xĵ − z k̂ .
(a) Show that it is conservative.
(b) Find the potential energy associated with this force.
Solution:

î ĵ k̂ h i
∇ × F = ∂x (−k) = + 0 + (1 1) =0
∂ ∂ ∂
∂y ∂z −k î(0) ĵ k̂ −
y x −z

So F is conservative and there exists a potential energy V associated with the force F.
Z Z Z Z
V = − F · dr = − Fx dx − Fy dy − Fy dy
Z Z Z
V =k ydx + k xdy − zdz

We must have
∂V ∂V ∂V
F = −∇V = − ,− ,−
∂x ∂y ∂z
So
∂V
Fx = −
= −ky
∂x
Integrating with respect to x while keeping y, z as constants:

V (x, y, z) = −kyx + g(y, z)

where g(y, z) is a function to be determined. We must also have:

∂V ∂g
Fy = − = −kx = kx −
∂y ∂y
where we evaluated the ∂V∂y . Integrating this equation with respect to y while keeping x, z
as constants, we obtain:

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 154 — #170

i i

154 Classical Mechanics: A Computational Approach

g(x, y) = 2kxy + h(z)

where h(z) is a function to be determined. Now we have:

V (x, y, z) = −kyx + 2kxy + h(z) = kxy + h(z)

We must also have:

∂V ∂h
Fz = − = kz =
∂z ∂z
Integrating with respect to z , we get:
k
h(z) = − z 2 + c
2
where c is an arbitrary constant which we set to c = 0. Therefore:
k
V (x, y, z) = kxy − z 2
2
We must check the answer:

∂V ∂V ∂V
F = −∇V = − ,− ,− = −ky î − kxĵ + kz k̂ X
∂x ∂y ∂z

5.6 ENERGY AND MULTIPARTICLE SYSTEMS

Consider a collection of N discrete particles, each located at ri for i = 1, 2, . . . , N , with a
center of mass located at R, shown as the empty circle in Figure 5.4. The mass of each
particle is mi and the velocity of each particle is vi . Note that what follows can be repeated
for a continuous mass distribution, but discrete sums will become integrals. The ith particle
is located at a position ri relative to the origin, and at a position r0i relative to the center
of mass. Notice that R + r0i = ri , where R is the location of the center of mass relative to
the origin.

ri'=ri-R

mi
R

x
Figure 5.4: A collection of discrete particles. The center of mass is located at R and is
denoted by the empty circle.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 155 — #171

i i

Energy 155

First, we will calculate the kinetic energy of the system, by using a method similar to
our calculation of the total momentum of a system of N particles in Chapter 4. The total
kinetic energy is:
N
1
X
T= 2
mi vi (5.6.1)
2
i=1

By using ri = R + r0i
and ṙi = Ṙ + ṙ0i
(recall that a dot above a variable means differentiation
of that variable with respect to time), we can write vi2 as:
2
vi2 = ṙi · ṙi = ṙ0i + Ṙ · ṙ0i + Ṙ = vi0 + 2 ṙ0i · Ṙ + V2 (5.6.2)

where vi0 = ṙ0i is the velocity of the ith particle relative to the center of mass, and V = Ṙ is
the velocity of the center of mass. Inserting (5.6.2) into (5.6.1) gives:
N N
! N
1 X 1

X
0 2 d X 0
T= + Ṙ · mi ri + 2
(5.6.3)

mi vi mi V
2 dt 2
i=1 i=1 i=1

The middle term in the above equation contains mi r0i , which in Chapter 4 was shown to
P
be zero. Hence, the total kinetic energy becomes:

Kinetic Energy of a System of N Particles

N
1 1

0 2
X
T = MV +
2
(5.6.4)

mi vi
2 2
i=1

The first term in (5.6.4) is the kinetic energy of the center of mass, and the second term
in (5.6.4) corresponds to the kinetic energy of each particle that is moving relative to the
center of mass.
For potential energy, we need to consider both external and internal forces and their
corresponding potential energies. The potential energy of the particle with mass mi is:
X
Vi = Viext + int
Vi,j (5.6.5)
j6=i

where Viext is the potential energy due to the external force acting on mi , and Vi,j int is

the potential energy due to the interaction force between mi and mj . Examples of such
interaction forces could be gravitational,
electrostatic, or any other central forces. In general
for central forces we have Vi,j
int = V int r − r , meaning that the potential energy depends

i,j i j
on the relative distance between the two particles. Therefore the total potential energy of
the system is:
XN N X
X
V = Viext + int
(5.6.6)

Vi,j ri − rj
i=1 i=1 j6=i

In the case of rigid bodies the particles are at fixed distances, and therefore the internal
potential energy can be ignored, since it will be constant for each particle. Hence, for a rigid
body we will only have to worry about external forces. We will study rigid bodies in more
detail in Chapter 11.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 156 — #172

i i

156 Classical Mechanics: A Computational Approach

5.7 CHAPTER SUMMARY

The work W done by the force F (x) on a particle moving along a line from position x0 to
position x is: Z x
W= F x0 dx0

x0

The Work-Kinetic Energy Theorem states that work needs to be done by the force
F (x) in order to change the kinetic energy of a system.
Zx
1 1
mv 2 − mvo2 = F (x0 )dx0
2 2
xo

The potential energy function V (x) associated with the force F (x) in one dimension is
defined by:
Z x
V (x) − V (x0 ) = − F (x0 )dx0
x0

The force F (x) associated with a potential energy V (x) can be found through differenti-
ation:

dV (x)
F (x) = −
dx
A particle is in an equilibrium state at x = x0 if:

dV
=0
dx
x0

Physical systems can exist in three types of equilibria: stable, unstable, and neutral
depending on the sign of d2 V /dx2 at the equilibrium point x = x0 .
The velocity is zero at the turning points of motion, so all of the particle’s energy is in
the form of potential energy.
Hooke’s law appears any time we are studying small amplitude oscillations about a
stable equilibrium, by doing a Taylor expansion of the potential V (x):
1 1
V (x) ≈ (x − x0 )2 V (2) (x0 ) = k(x − x0 )2
2 2
In two and three dimensions, the work done by the force F moving an object between
points r1 and r2 is: Z r2
W (r1 → r2 ) = F · dr
r1

The value of W can depend on the path C taken by the particle:

Z
W= F · dr
C

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 157 — #173

i i

Energy 157

In order to calculate integrals of this type, one can break them up into component integrals:
Z Z Z
W= Fx dx + Fy dy + Fz dz
C C C

Potential energy and force in three dimension are:

Z
∂V ∂V ∂V
F = −∇V (x, y, z) = − i+ j+ k V = − F · dr
∂x ∂y ∂z
Rr
For a conservative force: ∇ × F = 0, and the work r12 F · dr is independent of the path
taken between the two points r1 and r2 .
Conservation of mechanical energy states that the sum of the kinetic and potential
energy, called the total mechanical energy E, is always constant only if conservative forces
act on the system. In short: Conservative forces conserve energy.

E = T + V (x) = To + V (xo )

The kinetic energy of a system of N particles is the sum of the kinetic energy of the
center of mass as well as the kinetic energy of each particle that is moving relative to the
center of mass:
N
1 1

0 2
X
T = MV + 2

mi vi
2 2
i=1

25. Consider the force F = −2z 3 î + (3 + 2yez )ĵ + (y 2 ez − 6xz 2 )k̂.

(a) Show that it is conservative.
(b) Find the potential energy associated with this force.
26. Prove that the electrostatic force F = kq1 q2 /r2 r̂ on a point charge q1 at position r due
to a fixed charge q2 at the origin is conservative. The constant k = 1/4π0 is positive
and depends on the permittivity of free space, 0 . Find the potential energy when the
charges have the same sign. What happens to the potential energy when the charges
have a different sign? Explain your answer.
27. For each potential energy function, compute the associated force:
(a) V = k x2 + y 2 + z 2 .

(b) V = −kr cos θ, where k is a positive constant and r and θ are polar coordinates.
(c) V = −kr cos φ + cr2 sin θ, where k and c are positive constants and r, θ, and φ
are the typical spherical coordinates.

i i

In this section, we make some introductory remarks about the mathematical representation
of oscillators in general. We will then apply the ideas presented in this section to specific
oscillators in later sections. Oscillating systems are fundamentally important in physics
and engineering systems, but interest in their study extends to almost every branch of
science. From a mathematical point of view, oscillating systems are often described by
nonhomogeneous linear ordinary differential equations of the general form:

dn x dn−1 x dx
An + An−1 + .. + A1 + A0 x = f (t) (6.1.1)
dtn dtn−1 dt
where x = x(t) measures the oscillator’s displacement from a stable equilibrium, A0 , . . . , An
are constants, and f (t) is a function of time t. The order of a differential equation is the
highest derivative in the equation, and the equation is called homogeneous when the function
f (t) = 0. Equation (6.1.1) is called a linear differential equation because it consists only
of terms that are linear in x(t) and their derivatives. In general, equations that describe
oscillations are nonlinear in x and its derivatives. However, as we will see later in this
chapter, in the limit of small displacements from equilibrium, the equations of motion
usually take a linear form similar to that of (6.1.1).
In this chapter, we will first study homogeneous linear second-order differential equations
of the form:

163

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 164 — #180

i i

164 Classical Mechanics: A Computational Approach

Homogeneous Second-Order Linear Differential Equation

d2 x dx
A2 + A1 + A0 x = 0 (6.1.2)
dt2 dt

If x1 (t), x2 (t) are two solutions of this equation, then the general solution xc (t) of (6.1.2)
is a linear combination of these two solutions, in the form:

Solution of Homogeneous Equation

xc (t) = C1 x1 (t) + C2 x2 (t) (6.1.3)

where C1 , C2 are two arbitrary constants. These constants C1 , C2 can be determined by

the initial conditions of the system at time t = 0, which are usually given in the form of
the initial position x(0) and the initial speed v(0) = ẋ(0). We will find that a homogeneous
second-order linear differential equation can be used to describe the motion of a particle
experiencing a restoring force (responsible for the oscillation) and a drag force, called a
damping force.
When external forces are applied to the oscillator, we will need a nonhomogeneous linear
second-order differential equation to describe the motion. A nonhomogeneous linear second-
order differential equation takes the form:

Nonhomogeneous Second-Order Linear Differential Equation

d2 x dx
A2 2
+ A1 + A0 x = f (t) (6.1.4)
dt dt

In the general theory of differential equations, the solution of this equation is given by
the sum of the solution of the homogeneous equation xc (t) from (6.1.3), sometimes called
the complementary solution, and a particular solution xp (t) of the full nonhomogeneous
equation. The particular solution xp (t) is any solution of the equation:

d2 xp dxp
A2 2
+ A1 + A0 xp = f (t) (6.1.5)
dt dt
The general solution x(t) of a nonhomogeneous second-order differential equation then takes
the form:
General Solution of Nonhomogeneous Equation

x(t) = xc (t) + xp (t) = C1 x1 (t) + C2 x2 (t) + xp (t) (6.1.6)

6.2 THE SIMPLE HARMONIC OSCILLATOR

In this section, we study one of the most important problems in classical mechanics, the
simple harmonic oscillator (SHO). Simple harmonic motion occurs when an object displaced
a small distance from a stable equilibrium experiences a restoring force towards the equi-
librium state. An important example of simple harmonic motion is a mass attached to a
massless spring with no friction.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 165 — #181

x(t) = C1 ei ω0 t + C2 e−i ω0 t (6.2.8)

Note that in general C1 and C2 are arbitrary complex constants.

By using Euler’s identities ei ω0 t = cos (ω0 t) + i sin (ω0 t) and e−i ωo t = cos (ω0 t) −
i sin (ω0 t), we can change the exponentials in this equation into trigonometric functions.
We find that the general solution must be of the form:

x(t) = C1 [cos ( ω0 t) + i sin (ω0 t)] + C2 [cos (ω0 t) − i sin (ω0 t)] (6.2.9)

This equation can be written in two equivalent mathematical forms:

Two Forms of the Solution x(t) for the SHO

x(t) = A cos ( ω0 t) + B sin ( ω0 t) (6.2.10)

x(t) = C cos ( ω0 t + φ) (6.2.11)

p
with C = A2 + B 2 tan φ = B/A (6.2.12)

We can rewrite (6.2.9) as (6.2.10) by requiring the coefficients A and B to be real

numbers. In this case, we require that C2∗ = C1 , where C2∗ is the complex conjugate of C2 .
Complex conjugation changes the sign of the imaginary part of a complex number. Hence,
if c = a + ib, where a and b are real numbers, then c∗ = a − ib.
We can also rewrite (6.2.9) as (6.2.11) using the trigonometric identity cos(a + b) =
cos a cos b − sin a sin b.
From the trigonometric forms of the solution of the SHO, we see that the quantity ω0
represents the angular frequency of the oscillating mass. The angular frequency ω0 is called
the natural frequency of the oscillator and is the frequency of small amplitude oscillations.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 167 — #183

i i

Harmonic Oscillations 167

The Natural Frequency of a Simple Harmonic Oscillator

r
k
ω0 = (6.2.13)
m

Returning to Equations (6.2.10) and (6.2.11), A, B, and C are now real constants which
can be determined by the initial conditions, which specify the state of the system at a given
time (usually at t = 0). For example, given the initial conditions v(0) = v0 and x(0) = x0 at
t0 = 0, we then have from (6.2.10) with t = 0:

x(0) = A cos 0 + B sin 0 = A = x0 (6.2.14)

and so A = x0 .
The speed of the mass v(t) is found by taking the derivative of (6.2.10) with respect to
time:
dx
v(t) = = −Aω0 sin ( ω0 t) + B ω0 cos ( ω0 t) (6.2.15)
dt
By using the given initial condition v(0) = v0 at t0 = 0 :

v(0) = ω0 [−A sin 0 + B cos 0 ] = ω0 B = v0 (6.2.16)

and so B = v0 /ω0 . Therefore the complete solution for these initial conditions is:
v0
x(t) = x0 cos ( ω0 t) + sin ( ω0 t) (6.2.17)
ω0
The above solution is the same result we obtained in Chapter 2, by substituting a
solution of the form x = A cos(ωt − φ) in the original equation mẍ = −kx.

6.2.2 Potential and Kinetic Energy in Simple Harmonic Motion

Let us write the position and velocity for a mass undergoing simple harmonic motion in
the form x = A cos(ω0 t) and v = −Aω0 sin(ω0 t). The corresponding potential and kinetic
energies are varying with time t, according to: V (t) = 21 k x(t)2 and T (t) = 21 m v(t)2 . The
total energy of the system is then:
1 1
E = T + V = mω02 A2 sin2 (ω0 t) + kA2 cos2 (ω0 t) (6.2.18)
2 2
However, because ω02 = k/m, we get:
1
E = kA2 (6.2.19)
2
which is the maximum potential energy of the system. Hooke’s law represents a conservative
force, therefore during the oscillation, energy is transferred between kinetic and potential,
but their sum always remains the same. At the maximum displacement from the equilibrium
position, we have x = A, and the potential energy is a maximum and kinetic energy is zero.
When the oscillator is at the equilibrium x = 0, the kinetic energy is maximum and potential
energy is zero. In-between the maximum displacement and equilibrium, the potential and
kinetic energies change. For example, as the mass moves towards equilibrium, the potential
energy decreases and the kinetic energy increases, however as mentioned earlier, their sum
remains constant.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 168 — #184

i i

168 Classical Mechanics: A Computational Approach

We can calculate the average kinetic energy < T > in one period of oscillation by taking
a time average of T over the period τ = 2π/ω0 as follows:

1 τ1 1 τ1 1
Z Z
<T >= mv 2 dt = mA2 ω02 cos2 (ω0 t)dt = mA2 ω02 (6.2.20)
τ 0 2 τ 0 2 4
Rτ
where we used the integral value τ1 0 cos2 (ω0 t)dt = 21 .
Similarly, we can find the average potential energy < V > in one period of oscillation by
taking a time average of V (t) :

1 τ1 2 1 τ1 2 2 1 1
Z Z
<V >= kx dt = kA sin (ω0 t)dt = kA2 = mA2 ω02 (6.2.21)
τ 0 2 τ 0 2 4 4

In conclusion:
1 1 1
<V >=<T >= mA2 ω02 = kA2 = E (6.2.22)
4 4 2
where we used k = mω 2 and E = 21 kA2 . This equation shows that the time averages of both
the kinetic and potential energies within one period of oscillation are equal to one-half the
total energy E.

Time Averages of Energy

1
< V >time = < T >time = E (6.2.23)
2

This equation is an example of the well-known virial theorem in mechanics. Since the
potential and kinetic energies are also functions of the position x, we can also evaluate their
space averages over a period. The result for these space averages is:

Space Averages of Energy

1 1
< V >space = E < T >space = E (6.2.24)
6 3

6.2.3 The Simple Plane Pendulum as an Example of a Harmonic Oscillator

The plane pendulum is one of the simplest physical systems which can exhibit harmonic
oscillations for small displacements from an equilibrium position. In this section, we calculate
the period τ of the simple pendulum for oscillations with a small amplitude, and find
expressions for the kinetic and potential energies as functions of time.
Figure 6.2 shows a pendulum consisting of a mass m which can swing freely in a vertical
plane. A massless rod of length L connects the mass to the pendulum’s pivot point. When
the mass is displaced sideways from its equilibrium position, it is subject to a restoring force
due to gravity. This force will accelerate the mass back toward the equilibrium position.
The time for one complete cycle is the period τ .

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 169 — #185

i i

Harmonic Oscillations 169

Figure 6.2: The simple plane pendulum.

We can find the equation of motion for the simple plane pendulum by applying Newton’s
second law to the tangential axis only, F = −mg sin θ = ma and a = −g sin θ. The linear
2
acceleration a = ddt2s , where s is the arc length, is related to the angle θ by the formula
d2 s
s = Lθ and is given by dt2
= Lθ̈. Therefore we obtain:

Lθ̈ = −g sin θ (6.2.25)

By substituting ω02 = g/L, we obtain the equation of motion for the simple plane pendulum:

θ̈ + ω02 sin θ = 0 (6.2.26)

For small displacements from equilibrium, we can approximate sin θ ≈ θ, and the equation
becomes:
Equation of Motion for Small Pendulum Oscillations

f i g , a x e s = p l t . s u b p l o t s ( 1 , 2 , f i g s i z e =(10 ,4) )
a x e s [ 0 ] . p l o t ( time , x )
a x e s [ 0 ] . s e t x l a b e l ( ” time , s ” )
a x e s [ 0 ] . s e t y l a b e l ( ”x , m” )
a x e s [ 1 ] . p l o t ( time , T, ” r−−” , time , V, ”b” )
a x e s [ 1 ] . s e t x l a b e l ( ” time , s ” )
a x e s [ 1 ] . s e t y l a b e l ( ”V−e n e r g y=Blue , T−e n e r g y=Red , J ” )
f i g . s a v e f i g ( ” ex6 −1. e p s ” )

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 172 — #188

i i

172 Classical Mechanics: A Computational Approach

2.0 2.00
1.5 1.75

V-energy=Blue,T-energy=Red, J
1.0 1.50
0.5 1.25
x, m

0.0 1.00
−0.5 0.75
−1.0 0.50
−1.5 0.25
−2.0 0.00
0 2 4 6 8 10 0 2 4 6 8 10
time, s time, s

Figure 6.3: The Python code output, using the Euler method to integrate the equation for
simple harmonic motion.

6.4 DAMPED HARMONIC OSCILLATOR

Next, we add a linear resistive force to the simple harmonic oscillator. The resistive force
will be modeled similar to the linear drag forces discussed in Chapter 2. The resistive force
is sometimes called damping and is present, for example, when the oscillator is moving
through a fluid such as air. The equation for the damped harmonic oscillator contains an
additional friction term which is assumed to be linearly dependent on the speed in the form
F = −bv, where b depends on the size and shape of the oscillator, as well as other physical
parameters such as the type of fluid the oscillator moves through, and the coefficient of
friction between the oscillator and a supporting surface. The complete equation for the
damped harmonic oscillator is obtained now from F = ma = −kx − bv. By dividing with the
mass, m, we obtain:
d2 x b dx k
+ + x=0 (6.4.1)
dt2 m dt m
We now define a constant γ called the damping parameter, in order to express the amount
of damping present in the system:
γ = b/2m (6.4.2)
and the differential equation becomes:

Equation of Motion for the Damped Harmonic Oscillator

ẍ + 2γ ẋ + ω02 x = 0 (6.4.3)

where we used again ω0 = k/m for the natural frequency of the undamped oscillator.
p

Once more we look for solutions that have the form Ceλt , where C is a constant. By
substituting this into (6.4.3) and dividing all terms by Ceλt , we find:

λ2 + 2γλ + ω02 = 0 (6.4.4)

This is solved using the quadratic formula:

q
λ± = −γ ± γ 2 − ω02 (6.4.5)

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 173 — #189

i i

Harmonic Oscillations 173

This implies a pair of solutions eλt , corresponding to the two roots of the quadratic. The
general solution is a linear combination of these two solutions:
p p
−γ+ γ 2 −ω02 t −γ− γ 2 −ω02 t
p p
−γt γ 2 −ω02 t − γ 2 −ω02 t
x(t) = A0 e + A1 e =e A0 e + A1 e

q (6.4.6)
or by introducing the parameter ω ≡ γ 2 − ω02 :

General Solution x(t) for the Damped Harmonic Oscillator

x(t) = e−γt A0 eωt + A1 e−ωt (6.4.7)

q q
ω= γ 2 − ω02 = (b/2m)2 − (k/m) (6.4.8)

The exact mathematical form of this equation x(t) q depends on the numerical values of
γ and ω0 , which determine whether the parameter ω = γ 2 − ω02 is real, imaginary, or zero.
Therefore there are three physically different behaviors of the damped harmonic oscillator as
follows: overdamped oscillations (γ > ω0 ), underdamped oscillations (γ < ω0 ), and critically
damped oscillations (γ = ω0 ).

6.4.1 Overdamped Oscillations

In the case of overdamped oscillations,
q we have γ > ω0 , which means that the square root
appearing in the expression ω = γ − ω02 is real, and (6.4.7) will be written as the sum of
2

two real exponential functions. The resulting motion is that the system returns to equilib-
rium by an exponential decay without oscillating. Alternatively, one can write the solution
as a linear combination of hyperbolic functions (see Problem 24, in the end of chapter
problems for this section):

Solution x(t) for an Overdamped Oscillator

x(t) = A0 eωt + A1 e−ωt e−γt (6.4.9)

x(t) = [B0 sinh (ωt) + B1 cosh (ωt)] e−γt (6.4.10)

q
ω = γ 2 − ω02 γ > ω0 (6.4.11)

The constants A0 , A1 , B0 , B1 can be found from the initial conditions of the oscillator.
Example 6.2 illustrates the motion of an overdamped oscillator using Mathematica to solve
for the equations of motion in a mass-spring system exhibiting overdamped behavior.

Example 6.2: Mathematica code for overdamped motion

Use Mathematica to obtain the solution x(t) of (6.4.3), with parameters m = 1 kg,
k = 1 N/m, x(0) = 1 m, v(0) = 0 m/s and b = 3.0 Ns/m.
Solution:
For the given values of the parameters, ω 2 = γ 2 −ω02 = (b/2m)2 −(k/m) = (3.0/2)2 −1 =
1.25 > 0. Since this quantity is positive, we expect overdamped motion. The analytical

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 174 — #190

i i

174 Classical Mechanics: A Computational Approach

q
solution given below from Mathematica is a linear combination of a cosh γ 2 − ω02 t and
q
a sinh γ − ω0 t term, in agreement with the general form of the analytical solution
2 2

given by (6.4.10) for the overdamped motion.

sol1 = Assuming[b∧ 2 − 4 ∗ k ∗ m > 0, DSolve[{m ∗ x00 [t] + k ∗ x[t] + b ∗ x0 [t] == 0, x[0] == xo,
x0 [0] == 0}, x[t], t]]//FullSimplify;
x1 = x[t]/.sol1[[1]]
 h√ i
√ bSinh b2 −4kmt
bt b2 −4kmt 2m
OUTPUT: e− 2m xo Cosh 2m + √ 
b2 −4km

m = 1; k = 1; b = 3; xo = 1;
Plot[x1, {t, 0, 15}, FrameLabel → {“Time t,s”, “Position x(t)”}, Frame → True,
LabelStyle → (FontSize → 20), ImageSize → Medium]
1.0
Position x(t)

0.8
0.6
0.4
0.2
0.0
0 2 4 6 8 10 12 14
Time t,s

6.4.2 Underdamped Oscillation

q
In the case of underdamped oscillations, we have γ < ω0 , and ω = γ 2 − ω02 is imaginary.
The system oscillates at a reduced frequency compared to the undamped case, with the
amplitude gradually decreasing to zero. Mathematically, it is preferable
q to get rid of the
imaginary quantities by using the real frequency parameter ωd = ω02 − γ 2 , expressing the
general solution as a linear combination of trigonometric functions (see Problem 25 in the
end of chapter problems for this section):

Two Forms of Solutions x(t) for an Underdamped Oscillator

x(t) = [A0 sin (ωd t) + A1 cos (ωd t)] e−γt (6.4.12)

x(t) = Ae−γt cos (ωd t − φ) (6.4.13)

q
ωd = ω02 − γ 2 γ < ω0 (6.4.14)

The solution of the underdamped case oscillates repeatedly through the equilibrium
point as the amplitude of oscillation decays to zero. This is unlike the overdamped case,
where the oscillator may not pass through equilibrium at all, as its amplitude decays to
zero. This is an important difference in the behaviors of the two oscillators. Also notice from

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 175 — #191

i i

Harmonic Oscillations 175

q
(6.4.12) and (6.4.13) that the parameter ωd = ω02 − γ 2 = k/m − γ 2 is the frequency of
p

the underdamped oscillator, which is smaller than the natural frequency ωo = k/m.
p

The expression Ae−γt in (6.4.12) and (6.4.13) represents the decrease of the amplitude
as a function of time t for underdamped motion. Notice that the amplitude of oscillation
decays exponentially with time. The quantity τA = 1/γ has dimensions of time, in order for
the argument of the exponential e−γt to be dimensionless, and characterizes how fast the
amplitude of oscillation reaches zero. As a general rule of thumb for exponential decaying
functions, the amplitude of oscillation is considered to have reached a value of effectively
zero after five characteristic times, i.e., after t ∼
= 5/γ.
For small damping, i.e., when when γ << ω0 , we can use the binomial approximation
∼ 1 − ax to obtain:
(1 − x)a =

γ2
q q
2∼
ωd = ω0 − γ = ω0 1 − (γ/ω0 ) = ω0 1 − 2
2 2 γ << ω0 (6.4.15)
2ω0

Example 6.3 gives a Mathematica program for a mass-spring system following under-
damped behavior.

Example 6.3: Mathematica code for underdamped motion

Use Mathematica to obtain the solution x(t) of (6.4.3), with parameters m = 1 kg,
k = 1 N/m, x(0) = 1 m, v(0) = 0 m/s and b = 0.1 Ns/m.
Solution:
For the given values of the parameters, ω 2 = γ 2 − ω02 = (b/2m)2 − (k/m) = −0.9975 < 0.
Since this quantity is negative, we expect underdamped motion. The following is Math-
ematica code that uses the DSolve command to obtain the analytical solution of (6.4.3)
with the given initial conditions x(0) = 1 m , v(0) = 0 m/s . The parameter x1 in the code
represents the analytical solution, and the Assuming command tells Mathematicaq that
2
(k/m) − (b/2m) > 0. The analytical solution is a linear combination of a cos 2
ω0 − γ 2 t
q
and a sin ω02 − γ 2 t term, multiplied by the exponential decay term e−γt . This is in
agreement with the general form of the underdamped solution given by (6.4.12).
sol1 = Assuming[b∧ 2 − 4 ∗ k ∗ m < 0, DSolve[{m ∗ x”[t] + k ∗ x[t] + b ∗ x0 [t] == 0, x[0] == xo,
x0 [0] == 0}, x[t], t]]//FullSimplify;
x1 = x[t]/.sol1[[1]]
 h√ i
√ bSin −b2 +4kmt
bt −b2 +4kmt 2m
OUTPUT: e− 2m xo Cos 2m + √ 
−b2 +4km

m = 1; k = 1; b = 0.1; xo = 1;
Plot[x1, {t, 0, 70}, FrameLabel → {“Time t,s”, “Position x(t)”}, Frame → True,
LabelStyle → (FontSize → 20), ImageSize → Medium]

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 176 — #192

i i

176 Classical Mechanics: A Computational Approach

1.0

Position x(t)
0.5

0.0

-0.5

0 10 20 30 40 50 60 70
Time t,s

6.4.3 Critically Damped Oscillations

q
In the case of critically damped oscillations, we have γ = ω0 , and ω = γ 2 − ω02 = 0. A
critically damped system returns to equilibrium as quickly as possible without oscillating.
This is often desired for the damping of systems, such as doors and car suspensions. In this
case, there is only one root of the quadratic equation λ = −ω0 . In addition to the solution
x(t) = eλt , a second solution is given by the function x(t) = teλt (see Problem 27 from the
end of chapter problems for this section). The complete solution is a linear combination of
these two solutions:
Solution x(t) for a Critically Damped Oscillator

x(t) = (A + Bt) e−ω0 t (6.4.16)

where A and B are determined by the initial conditions of the system. A comparison of x(t)
for the underdamped, overdamped, and critically damped oscillations is shown in Figure
6.4, where we can easily see that the critically damped oscillator (dash dot curve) returns
to equilibrium in a shorter time than other cases. Note that the same parameter values are
used for m, k, x(0), and v(0) to produce all three curves, while the value of b is variable.

1.0
Underdamped
Overdamped
0.5 Critical
Position x(t)

0.0
Out[ ]=

-0.5

0 2 4 6 8 10 12 14
Time t,s

Figure 6.4: Dependence of the harmonic oscillator behavior on the amount of damping, show-
ing underdamped (oscillatory curve), overdamped cases (solid line), and critically damped
(dashed curve). The code which produces these graphs is given in Examples 6.2, 6.3, and
6.4.

Example 6.4 gives a Mathematica program for a mass-spring system following critically
damped behavior.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 177 — #193

178 Classical Mechanics: A Computational Approach

these approximations the term γ sin (ωd t) is much smaller that the term ωd cos (ωd t) and
can be dropped in (6.5.3), to obtain:
dx
= A0 e−γt [ωd cos (ωd t) ] (6.5.4)
dt
Substituting into (6.5.1) we obtain:
1 2 1 2
E = m A0 e−γt ωd cos (ωd t) + k A0 sin (ωd t) e−γt (6.5.5)
2 2
By collecting terms:
1
E = A20 e−2γt mωd2 cos2 (ωd t) + k sin2 (ωd t) (6.5.6)

2
Substituting ωd ' ω0 = k/m we obtain:
p

1 1
E = A20 e−2γt k cos2 (ωd t) + k sin2 (ωd t) = kA20 e−2γt (6.5.7)

2 2
The above equation shows that the energy decays exponentially with time. The char-
acteristic decay time is τE = 1/(2γ). We previously saw in (6.4.12) that the corresponding
characteristic decay time for the amplitude is τA = 1/γ.
In summary, the total energy E and the rate of decrease of the total energy dE/dt for
the cases of weak damping γ << ωd , are given by:

Energy and Energy Loss for Underdamped Oscillator, γ << ωd

1
E = kA20 e−2γt (6.5.8)
2
dE
= −γkA20 e−2γt = −2γE (6.5.9)
dt

The total energy lost within a small time interval ∆t can be estimated from:
dE
∆E ∼
= ∆t = −2γE∆t (6.5.10)
dt
where we used (6.5.9). It is customary to define the dimensionless Quality factor or Q-factor
Q as the ratio Q = 2π|E/∆E|, for a time interval ∆t equal to the period τ = 2π/ωd . By
substituting this value of ∆t = τ = 2π/ωd in this definition of Q:

E 1 ωd
Q = 2π = 2π = (6.5.11)

∆E 2γτ 2γ

Q-factor for Underdamped Motion

ωd ∼ ωo
Q= = (6.5.12)
2γ 2γ

In section 6.7, we will provide a more in-depth discussion of the quality factor, in con-
nection with the concept of energy resonance of a forced harmonic oscillator.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 179 — #195

i i

Harmonic Oscillations 179

6.6 FORCED HARMONIC OSCILLATOR

Next, we consider the case where an external force F0 cos(ωt) is acting on a damped oscilla-
tor. Notice that the driving force has an amplitude of F0 and a frequency of ω, sometimes
referred to as the drive frequency. This case is referred to as the forced or driven harmonic
oscillator. The net force is then:

mẍ = −kx − bẋ + F0 cos(ωt)

The general equation for the driven system is then:

General Differential Equation for a Forced Harmonic Oscillator

ẍ + 2γ ẋ + ω02 x = D cos(ωt) (6.6.1)

where γ = b/2m, ω02 = k/m, and D = F0 /m. You might be wondering why we limit ourselves
to studying driving forces in the form of cosine functions. Clearly not all driving forces in
nature are in the form of cosine or sine functions. As we will show later in this chapter,
any periodic force (under certain conditions, which are often met in physical problems) can
be approximated using a series of sines and cosines. Therefore, the case of cosine or sine
driving forces is applicable widely and for many physical systems.
Note that (6.6.1) is a nonhomogeneous ordinary differential equation. As we discussed in
the introduction of this chapter on the general theory of differential equations, the solution
of nonhomogeneous equations is given by the sum of two parts, the homogeneous solution
xc (t), plus the particular solution xp (t) of the full nonhomogeneous equation:

x(t) = xc (t) + xp (t) (6.6.2)

We already saw in the previous section that the general solution xc (t) of the homogeneous
equations is given by (6.4.7):
q
xc (t) = e−γt A0 eωt + A1 e−ωt ω = (b/2m)2 − (k/m) (6.6.3)

The solution xc (t) is commonly referred to as the transient solution, because the exponential
decay term e−γt causes xc (t) to decay to zero. After the transient solution has decayed to
zero, the only solution left is xp (t) (assuming that xp 6= 0), and is therefore known as
the steady-state solution. Many physical systems exhibit both a transient and steady-state
behavior. Physicists are often (but not always!) more interested in the steady-state behavior
of the system, because that is the system’s long-term behavior.
Before we study the solution x(t) for the full equation (6.6.1), it is instructive to study
the case in which damping is absent. When no friction is present, the driven oscillator is
described by b = 0, so we obtain:

ẍ + ω02 x = D cos(ωt) (6.6.4)

We try a solution of the form x = A cos (ωt − φ), and after we substitute and collect terms
we obtain:
−Aω 2 cos (ωt − φ) + ω02 A cos (ωt − φ) =D cos(ωt)
(6.6.5)
A ω02 − ω 2 cos (ωt − φ) =D cos(ωt)

This will have two solutions, when φ = 0 and cos (ωt − φ) = cos (ωt), and when φ = π where
cos (ωt − φ) = − cos (ωt). The amplitude A must be always positive, so we can write the two

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 180 — #196

i i

180 Classical Mechanics: A Computational Approach

possible solutions in the form:

D
A= φ = 0 and ω < ω0 (6.6.6)
ω02 − ω 2

D
A= φ = π and ω > ω0 (6.6.7)
ω 2 − ω02

and the solution of this driven oscillator with no damping is:

D
x = A cos (ωt) = cos (ωt) φ = 0 and ω < ω0 (6.6.8)
ω02 − ω 2

D
x = A cos (ωt) = cos (ωt) φ = π and ω > ω0 (6.6.9)
ω 2 − ω02
A plot of the amplitude A and of the phase difference φ as a function of the external
frequency ω is shown in Figure 6.5. We see that the amplitude of the oscillation is maximum
when the external frequency ω = ω0 , and that there is a sharp change in the amplitude and
in the phase at this frequency. This is our first demonstration of amplitude resonance, where
there is a large response from the oscillator when it is driven at a frequency that matches
the natural frequency of the oscillator. We will comment on resonance in more detail later
in this section.

20 5

4
Phase difference

15
Amplitude A

3
10
2

5
1

0 0
0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0
External frequency ω/ωo External frequency ω ωo

Figure 6.5: The amplitude A and phase difference φ as a function of the external frequency
ω, in the special case of driven simple harmonic motion with no damping.

A more realistic situation is the case where damping is not zero. For this case, the
standard method is to try substituting a solution x = A cos(ωt − φ) in (6.6.1). The choice
of a cosine solution is motivated by the fact that the right-hand side of (6.6.1) contains a
cosine term with a frequency ω. We need a function x(t) whose derivatives will ultimately
give a cosine with the same frequency. We know that the first derivative of the cosine is
sine. Therefore we introduce a phase term φ, which if chosen properly will allow all of the
terms on the left-hand side of (6.6.1) to add up to a cosine, and match the right-hand
side. To simplify the mathematics, we write our trial solution as a complex exponential,
x = A ei(ωt−φ) , and also replace the term D cos(ωt) in this equation with D eiωt . We will
carry out the algebra
and then at the end, we will take the real part of the solution, because
cos(ωt) = Re eiωt . By substituting and taking the derivatives, we obtain:

− Aω 2 ei(ωt−φ) + 2γωiAei(ωt−φ) + ω02 A ei(ωt−φ) = D eiωt (6.6.10)

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 181 — #197

i i

Harmonic Oscillations 181

By canceling out the eiωt from all terms and multiplying by eiφ = cos φ + i sin φ on both
sides, we obtain:
− Aω 2 + 2γωiA + ω02 A = D eiφ = D (cos φ + i sin φ) (6.6.11)
By equating the real parts on the two sides of this equation and also equating the imaginary
parts on the two sides to each other, we obtain:
A ω02 − ω 2 = D cos φ (6.6.12)

2γωA = D sin φ (6.6.13)

Solving this system of equations for the amplitude A and for the phase angle φ, we
obtain the particular solution xp (t) of the nonhomogeneous differential equation (6.6.1)
for the driven harmonic oscillator. Finding φ from the above equations is simple; all one
needs to do is divide the second equation by the first. Finding A involves squaring each
equation and adding them, and recalling that sin2 φ + cos2 φ = 1. The complete solution
x(t) = xc (t) + xp (t) is then given by:

General Solution x(t) for the Forced Harmonic Oscillator

x(t) = xp (t) + A cos (ωt − φ) (6.6.14)

dx dxp
v= = − Aω sin (ωt − φ) (6.6.15)
dt dt
F0 /m
A= q 2 (6.6.16)
ω02 − ω 2 + 4γ 2 ω 2
2γω
tan φ = (6.6.17)
ω02 − ω 2

The Python code in Example 6.5 numerically solves (6.6.1) using the Euler method
described in Chapter 2. The results are plotted in Figure 6.6, where the left graph is a
plot of x(t) and the right graph is a plot of v(t). Notice that in Figure 6.6, the transient
behavior is dominant for t < 60 s, and that the sinusoidal steady-state behavior dominates
for t > 60 s.
Example 6.5: Python code for driven harmonic oscillator
Write a Python code which numerically solves the differential equation for the driven
SHO, ẍ + 2γ ẋ + ω02 x = D cos(ωt), by using the Euler method described in Chapter 2. Use
the numerical values m = 1 kg, k = 1 N/m , b = 15 Ns/m , D = 5 m/s2 and ω = 3 s−1 , from
t = 0 to t = 80 s. Plot both the position x(t) and the velocity v(t). Discuss the behavior of
the solution at small times and at large times. Use initial conditions x(0) = 2 and v(0) = 0.
Solution:
The following Python code clearly shows the change of behavior of the solutions x(t)
and v(t) as time progresses. The initial part of the functions contains an exponentially
decaying transient component. At longer times, the solutions approach the steady-state
solutions, which oscillate with time, and with a constant amplitude. Note that the red (in
the e-book) curled arrows in the code denote a line continuation, i.e., the text following a
curly arrow is a continuation of the line above it. Those arrows are not to be included in
your code.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 182 — #198

i i

182 Classical Mechanics: A Computational Approach

import numpy a s np
import m a t p l o t l i b . p y p l o t a s p l t

#d e f i n e c o n s t a n t s , time s t e p s i z e , and i n t e g r a t i o n time

m, k , b = 1 . 0 , 1 . 0 , 15
dt = 0 . 0 1
tmax = 8 0 . 0
num steps = int ( tmax/ dt )

#d e f i n e i n i t i a l c o n d i t i o n s and a r r a y s
x = np . z e r o s ( num steps )
v = np . z e r o s ( num steps )
x [ 0 ] = 2.0
v [ 0 ] = 0.0

#perform t h e E u l e r s t e p a l g o r i t h m
f o r i in range ( 1 , num steps ) :
x [ i ] = x [ i −1] + dt ∗ v [ i −1]
v [ i ] = v [ i −1] + dt ∗(−k/m ∗x [ i ]− dt ∗b/m ∗v [ i −1]+5∗ np . c o s
,→ ( 3 ∗ dt ∗ i ) )
time = np . l i n s p a c e ( 0 , tmax , num steps )
f i g , a x e s = p l t . s u b p l o t s ( 1 , 2 , f i g s i z e =(10 ,4) )
a x e s [ 0 ] . p l o t ( time , x )
a x e s [ 0 ] . s e t x l a b e l ( ” time , s ” , f o n t s i z e =18)
a x e s [ 0 ] . s e t y l a b e l ( ”x , m” , f o n t s i z e =18)
a x e s [ 1 ] . p l o t ( time , v )
a x e s [ 1 ] . s e t x l a b e l ( ” time , s ” , f o n t s i z e =18)
a x e s [ 1 ] . s e t y l a b e l ( ”v , m/ s ” , f o n t s i z e =18)
f i g . s a v e f i g ( ” ex4 . 5 . e p s ” )

4
2 3

2
1
1
v, m/s
x, m

0
0
−1

−2
−1
−3

−4
−2
0 20 40 60 80 0 20 40 60 80
time, s time, s
Figure 6.6: The result of running the Python code in Example 6.5 for the externally driven
harmonic oscillator. The effect of the transient part of the solution is clearly seen for times
t < 60 s, while the steady-state solution is dominant for t > 60 s.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 183 — #199

i i

Harmonic Oscillations 183

In addition, we can plot the amplitude A and phase angle φ as a function of frequency
using (6.6.16) and (6.6.17). The plots of the amplitude A and of the phase angle φ as a
function of the external frequency ω are shown in Figure 6.7. When ω << ω0 , we find that
the phase difference φ goes to zero, while for ω >> ω0 we find that the phase goes to π (180
degrees) .

Figure 6.7: Plots of the amplitude A (ω) and phase angle φ (ω) in the case of externally-
driven damped harmonic motion.

One of the most striking results shown in Figure 6.7 is the amplitude A (ω) as a function
of frequency. Notice that similar to the undamped case, amplitude resonance can occur.
The damping prevents the amplitude from blowing up to infinity at resonance. Although
the small value of γ = 0.01 results in a large amplitude at resonance, it is not infinite; it
simply is beyond the scale of the graph here. Notice that the amplitude resonance occurs
when ω ≈ ω0 , because of the small values of γ used to make Figure 6.7. Next, we will
compute the exact frequency at which resonance occurs.
The resonance frequency is the value of the drive frequency ω, which causes amplitude
resonance, and is found by computing:

dA
=0 (6.6.18)
dω

ω=ωr
q
We find that A is a maximum when ω = ωr = ω02 − 2γ 2 , where ωr is the resonant frequency.
The maximum value of A defined as Amax is then:

Resonance Frequency and Resonance Amplitude for a Forced Harmonic

Oscillator
q q
ωr = ωd2 − γ 2 = ω02 − 2γ 2 (6.6.19)

F /m
Amax = q0 (6.6.20)
2γ ω02 − γ 2

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 184 — #200

i i

184 Classical Mechanics: A Computational Approach

Example 6.6: Mathematica code for amplitude resonance

Use Mathematica to obtain the solution x(t) of (6.6.1) and plot the amplitude A of
the driven oscillation as a function of the external driving frequency ω. Use the numerical
values m = 1 kg, ω0 = 1 rad/s, F0 = 1.0 N, γ = 0.1 rad/s.
Solution:
By using the DSolve command as shown in the code, Mathematica produces the ana-
lytical solution which clearly contains the sum of xc (t) and xp (t). The variable x in the
code represents the full solution, while the variable steady represents xp (t). By using the
replacement rule /.Exp[t− ] → 0, the two exponential terms in xh (t) are set to zero, and
we are left with the expression steady which represents xp (t). By subtracting the variable
steady from x, we obtain the variable trans which represents the transient solution xh (t).
The parameter u is a list (created by Mathematica’s Table command) of external frequen-
cies ω and of the corresponding maximum amplitude A(ω). These maximum values are
obtained by using the command NMaxValue in Mathematica. Finally the list u is plotted
using the ListPlot command, and we obtain the desired graph of the maximum amplitude
A(ω) as a function of the external driving frequency ω.
sol = DSolve[{x00 [t] + ωo∧ 2 ∗ x[t] + 2 ∗ γ ∗ x0 [t] == Fo/m ∗ Cos[ω ∗ t]}, x[t], t];
x = x[t]/.sol[[1]]//Simplify

OUTPUT:√ √
Fo((−ω 2 +ωo2 )Cos[tω]+2γωSin[tω])

−t γ+ γ 2 −ωo2 t −γ+ γ 2 −ωo2
e C[1] + e C[2] +
m 4γ 2 ω 2 +(ω 2 −ωo2 )
2

steady = x/.Exp[t ] → 0

Fo((−ω 2 +ωo2 )Cos[tω]+2γωSin[tω])

OUTPUT:
m 4γ 2 ω 2 +(ω 2 −ωo2 )
2

trans = x − steady
√ √
−t γ+ γ 2 −ωo2 t −γ+ γ 2 −ωo2
OUTPUT: e C[1] + e C[2]

m = 1; γ = 0.1; Fo = 1; ωo = 1;
u = Table[{ω, NMaxValue[steady, t]}, {ω, 0, 3, .02}];
ListPlot[u, FrameLabel → {“Driving frequency ω, Hz”, “Amplitude A(ω)”},
Frame → True, BaseStyle → (FontSize → 18), Joined → True, ImageSize → Large]
5

4
Amplitude A(ω)

0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Driving frequency ω, Hz

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 185 — #201

186 Classical Mechanics: A Computational Approach

order to find ∆ω, we need to find the two points ω1 and ω2 on this graph, which correspond
to half of the maximum of the peak.

Figure 6.8: Plot of the time averaged energy < E > from (6.7.4), as a function of the
driving frequency ω. The value of γ = 0.03 results in a lightly damped harmonic motion,
corresponding to a large Q-factor.

To calculate Q, we assume a system with small damping (i.e., a sharp resonance), and
2
we substitute ω ' ω0 in (6.7.4), except for the term ω02 − ω 2 , to obtain:

F02 ω02 1
<E>= (6.7.6)
2m ω 2 − ω 2 2 + 4γ 2 ω 2

0 0

By approximating ω0 ' ω, the term ω02 − ω 2 = (ω0 + ω) (ω0 − ω) ' 2ω0 (ω0 − ω), and the
previous equation becomes:

F02 1
<E>= (6.7.7)
8m (ω0 − ω)2 + γ 2

Next, taking the derivative of < E > with respect to the driving frequency ω and setting
it equal to zero, we find that the maximum of this function occurs at ω = ω0 . Therefore,
the average energy resonance occurs when ω = ω0 . By substituting this value in (6.7.4), the
maximum time average of the total energy is given by:

F02
< E >max = (6.7.8)
8mγ 2
Next, we need to find the frequency at which the average energy < E > drops to half its
maximum value. We substitute (6.7.8) in (6.7.7) to obtain:

1 1 F02 F02 1
< E >max = = (6.7.9)
2 2 8mγ 2 8m (ω0 − ω)2 + γ 2

which simplifies to:

1 1
= (6.7.10)
2γ 2 (ω0 − ω)2 + γ 2
The solution to this quadratic equation in ω gives us the two values ω1 and ω2 of the driving
frequency, at which the time average of the total energy drops to half its maximum value:

ω1 = γ + ω0 ω2 = ω0 − γ (6.7.11)

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 187 — #203

i i

Harmonic Oscillations 187

The full width at half maximum is given by:

∆ω = ω1 − ω2 = 2γ (6.7.12)

Therefore from the definition of the Q-factor in (6.7.5), we find the ratio of the resonance
frequency ω0 over the FWHM ∆ω:

The Quality Factor Q for a Driven Oscillator with Small Damping

ωd ω0
Q= = (6.7.13)
∆ω 2γ

This is the same simple equation we derived for underdamped motion, and involves only
the coefficients of the second-order differential equation describing most resonant systems,
electrical or mechanical. In electrical systems, the stored energy is the sum of energies stored
in lossless inductors and capacitors; the lost energy is the sum of the energies dissipated
in resistors per cycle. In mechanical systems, the stored energy is the maximum possible
stored total energy, i.e., the sum of the potential and kinetic energies at some point in time,
while the lost energy is that lost to damping forces such as friction.

Example 6.7. Mathematica code for energy resonance

Use Mathematica to obtain the solution x(t) of (6.6.1) and plot the maximum total
energy E of the driven oscillation as a function of the external driving frequency ω. Use
the numerical values m = 1 kg, ω0 = 1 Hz, F0 = 1.0 N,γ = 0.1 rad/s.
Solution:
The code for this example is similar to the code used in Example 6.6. By using the
DSolve command as shown in the code, Mathematica produces the variable steady rep-
resenting xp (t) as before. The code also evaluates analytically the corresponding speed
vp (t) = dxp /dt, and the total energy E(t) = 12 mvp2 + 12 kx2p . The parameter u is a list of
external frequencies ω and of the corresponding maximum energy E(ω), which are obtained
by using the command NMaxValue as previously. Finally the list u is plotted using the
ListPlot command, and we obtain the desired graph of the maximum average energy E(ω)
as a function of the external driving frequency ω.
It is notable that the energy resonance E(ω) shown in this example is much narrower
than the corresponding amplitude resonance A(ω) studied in the previous example.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 188 — #204

i i

188 Classical Mechanics: A Computational Approach

sol = DSolve[{x00 [t] + ωo∧ 2 ∗ x[t] + 2 ∗ γ ∗ x0 [t] == (Fo*/m ∗ Cos[ω ∗ t]}, x[t], t];
x = x[t]/.sol[[1]]//Simplify

OUTPUT:√ √
Fo((−ω 2 +ωo2 )Cos[tω]+2γωSin[tω])

−t γ+ γ 2 −ωo2 t −γ+ γ 2 −ωo2
e C[1] + e C[2] +
m 4γ 2 ω 2 +(ω 2 −ωo2 )
2

steady = x/.Exp[t ] → 0

Fo((−ω 2 +ωo2 )Cos[tω]+2γωSin[tω])

OUTPUT:
m 4γ 2 ω 2 +(ω 2 −ωo2 )
2

en = (1/2) ∗ m ∗ steady∧ 2

Fo2 ((−ω 2 +ωo2 )Cos[tω]+2γωSin[tω])

2
OUTPUT: 2 2
2m 4γ 2 ω 2 +( )

ω 2 −ωo2

m = 1; γ = 0.1; Fo = 1; ωo = 1;
u = Table[{ω, NMaxValue[en, t]}, {ω, 0, 3, 0.02}];
ListPlot[u, PlotRange → All, FrameLabel → {“Driving frequency ω, Hz”, “Energy E(ω)”},
Frame->True, LabelStyle → {FontSize → 20}, Joined → True, ImageSize → Medium]

12
10
Energy E(ω)

8
6
Out[ ]=
4
2
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Driving frequency ω, Hz

6.8 ELECTRICAL CIRCUITS

The equation of motion for the forced harmonic oscillator can also be used to describe a
driven RLC circuit. In this section, we develop the equations of motion for the RLC circuit,
and we explore what is sometimes referred to as an electrical-mechanical analogy. To begin,
consider an RLC circuit as shown in Figure 6.9.

Figure 6.9: RLC series circuit with external voltage V (t), resistance R, inductance L, and
capacitance C.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 189 — #205

i i

Harmonic Oscillations 189

The governing differential equation can be found by substituting into Kirchhoff’s voltage
law (KVL), the constitutive equation for each of the three elements. From the KVL,

V (t) + VR + VL + VC = 0 (6.8.1)
where VR = −IR is the voltage drop across the resistor, VL = is the voltage drop
−L dI
dt
Q
across the inductor, VC = − C is the voltage drop across the capacitor, and V (t) is the
time varying voltage from the source. Note that we are using the normal conventions that
I is the current in the circuit and Q is the charge on the capacitor. Furthermore, L is the
inductance of the inductor, R is the resistance of the resistor, C is the capacitance of the
capacitor, and the current is I = Q̇. For simplicity, we will assume that V (t) = V0 cos(ωt).
Substituting these expressions in the KVL equation:
dI Q
L + IR + = V0 cos(ωt) (6.8.2)
dt C
Dividing by L leads to the second-order differential equation:
General Differential Equation for RLC Series Circuit

d2 Q(t) R dQ(t) 1 V0
2
+ + Q(t) = cos(ωt) (6.8.3)
dt L dt LC L

This equation has the exact same mathematical form as equation (6.6.1) for a driven
damped harmonic oscillator:
d2 x b dx k F0
+ + x= cos(ωt) (6.8.4)
dt2 m dt m m
where (6.6.1) is written without the substitution γ = b/2m.
Because (6.8.3) and (6.6.1) have the same mathematical form, we can reuse the solution
of (6.6.1) in order to get the solution of (6.8.3). The mathematical similarities imply that the
physical behaviors of the two systems are similar. The forced damped harmonic oscillator
has two solutions: a transient solution, which decays as t → ∞, and a steady-state solution
consisting of a sinusoidal oscillation with a frequency equal to the drive frequency. Physically,
these solutions are describing the displacement of the mass, m. Because of the similarities
between (6.8.3) and (6.6.1), we know that the RLC circuit will also have two solutions, a
transient, which will also decay, and a steady-state sinusoidal solution, which has the same
frequency as the voltage source in the circuit. In the case of the RLC circuit, it is the value
of the charge on the capacitor that is oscillating.
Furthermore, we can then draw exact analogies between mechanical systems and elec-
trical systems. These analogies are shown in Table 6.1 and are found by comparing the
locations of the variables in (6.8.3) to those in (6.8.4). The simplest analogy is of that
between the displacement x of the mass and the charge Q on the capacitor. A more inter-
esting analogy exists between L and m. The inductance L in (6.8.3) appears in the same
location as m in (6.6.1). This implies that L is taking on a role similar to that of mass. This
analogy makes physical sense if you recall that the inductance L measures the inductor’s
ability to resist changes in current (Q̇), while the mass m is a measure of inertia, an object’s
ability to resist changes in velocity (ẋ).
It is also possible to use the analogies to identify qualitative characteristics of the RLC
circuit’s behavior. For example, we know that the coefficient of x in (6.6.1) is the angular
frequency ω02 of oscillation for an undamped oscillator. We can conclude from (6.8.3) that
the oscillation frequency of a LC circuit (where R = 0) will be given by:

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 190 — #206

i i

190 Classical Mechanics: A Computational Approach

Oscillation Frequency for LC Series Circuit

1
ω0 = √ (6.8.5)
LC

because 1/LC is the coefficient of Q in (6.8.3).

Mechanical Series RLC Circuit

Position x Charge Q
dQ
Velocity dx
dt Current dt

Mass m Inductance L
Spring Constant k Inverse Capacitance 1/C
Damping γ Resistance R
q q
Natural Frequency k
m
1
LC

mẍ + γ ẋ + kx = F LQ̈ + RQ̇ + Q/C = V

Table 6.1: Analogy between electrical RLC circuit in series and the mechanical driven har-
monic oscillator.

Finally, we can derive an expression about the energy of the RLC circuit. We begin by
multiplying (6.8.3) by Q̇ and L:
1
LQ̇Q̈ + RQ̇2 + QQ̇ = V0 cos(ωt)Q̇ (6.8.6)
C
Notice that:
1 d 1 d
Q̇Q̈ = Q̇2 and QQ̇ = Q2

2 dt 2 dt
Substitution of the previous equations into (6.8.6) gives:
d 1 11 2

LQ̇2 + Q = −RQ̇2 + V0 cos(ωt)Q̇ (6.8.7)
dt 2 2C
The 12 LQ̇2 and 12 C1 Q2 terms in (6.8.7) are the energy stored in the inductor and capacitor,
respectively. We can see that if R = 0 and if there is no voltage source in the circuit V0 = 0,
then the right-hand side of (6.8.7) is zero, so the total energy of the circuit is conserved.
The terms on the right-hand side describe how energy is either lost or gained in the system.
Recall from your introductory physics course that the power dissipated by a resistor is
P = −I 2 R, which is the first term on the right-hand side of (6.8.7). The second term on the
right-hand side of (6.8.7) is describing the rate at which energy is supplied by the variable
voltage source. Notice that V0 cos(ωt)Q̇ is in the form of V I, which is the formula for electric
power.
Analogies can be a powerful tool in physics, which can be used to take an understanding
of one system, and apply it to another mathematically similar system. As with any analogy
(physics-related or not), be careful not to take the analogy too far. Doing so can lead to
confusion at best and to a complete misunderstanding of the system at worst.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 191 — #207

i i

Harmonic Oscillations 191

6.9 PRINCIPLE OF SUPERPOSITION AND FOURIER SERIES

So far we have studied the response of an oscillator that is being driven by a continuous
periodic force. However, not all driving forces are continuous. Consider a mother pushing
her child on a swing. The force applied by the mother is not continuous, but it is periodic
with a period τ . How can we mathematically model such forces? We could describe a force
that is “off” for a time τ /2 and then constant for a time τ /2 by using a piecewise definition:
(
0 0 ≤ t < τ /2
F (t) = (6.9.1)
F0 τ /2 ≤ t ≤ τ

however, such a definition could be difficult to work with analytically. We should note
that modern CAS like Mathematica can handle such equations in a variety of applications,
however, there is much value in learning how to work with such cases analytically. So, the
question remains: how do we represent such periodic discontinuous functions in a way that
lends itself to analytical solutions? The answer, which we will demonstrate in this section,
is that we can represent periodic discontinuous functions as a sum of sines and cosines
called a Fourier series. This interesting result is known as Fourier’s theorem and applies
to periodic functions that are continuous, or have a finite number of discontinuities over
the period τ . Since most functions describing physical processes have only a finite number
of discontinuities over an interval t ∈ [0, τ ], we can consider Fourier’s theorem to be widely
applicable.
Fourier’s theorem is only one part of the solution of handling discontinuous periodic
drive functions. By representing drive functions as a series of sines and cosines, we need to
be able to solve equations of the form:

ẍ + 2γ ẋ + ω02 x = a0 + a1 cos(ωt) + a2 cos(2ωt) + · · · + b1 sin(ωt) + b2 sin(2ωt) + · · · (6.9.2)

In other words, we need to know how to solve linear ordinary differential equations that
have multiple terms on the right-hand side. Because the differential equation is linear in x
and its derivatives, we can use the powerful concept of the principle of superposition, which
will allow us to break up the differential equation into pieces and then sum the solutions of
the individual pieces.

6.9.1 The Principle of Superposition

Consider a differential equation of the following form:

d2 x
2
dx d d
a 2 + b + cx = a 2 + b + c x = F (t) (6.9.3)
dt dt dt dt
which is very similar to the equations describing oscillations that we have worked with so
far. We can rewrite such an equation as L̂(x) = F (t), where the differential operator L̂ is
defined as:
d2 d
L̂ ≡ a 2 + b + c (6.9.4)
dt dt
The operator L̂ is linear if it satisfies:

L̂ (c1 x1 + c2 x2 ) = c1 L̂ (x1 ) + c2 L̂ (x2 ) (6.9.5)

where c1 and c2 are arbitrary constants, and x1 and x2 are two solutions of the differential
equation. It is straightforward to show that L̂ defined in (6.9.4) is a linear operator.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 192 — #208

i i

192 Classical Mechanics: A Computational Approach

Next, we consider a differential equation of the form:

L̂(x) = c1 F1 (t) + c2 F2 (t) (6.9.6)

where L̂ is defined as in (6.9.4). Suppose we know two functions x1 and x2 , such that
L̂ (x1 ) = F1 (t) and L̂ (x2 ) = F2 (t). In other words, x1 is the solution to (6.9.6) without F2 (t)
on the right-hand side, and x2 is the solution without F1 (t) present. Then:

L̂ (c1 x1 + c2 x2 ) = c1 L̂ (x1 ) + c2 L̂ (x2 ) = c1 F1 (t) + c2 F2 (t) (6.9.7)

Therefore, the sum of the individual solutions x1 and x2 is the solution to the full differential
equation (6.9.6). We can, of course, extend this to an arbitrary number of terms on the
right-hand side:
N
X
L̂ (x) = ci Fi (t) (6.9.8)
i=1

which would then have the solution:

N
X
x(t) = xc + ci xi (6.9.9)
i=1

where xi satisfied L̂ (xi ) = Fi (t) and xc is the complementary solution of the ODE. The
ability to add solutions in the above manner is called the principle of superposition and
applies only to linear differential equations. In Chapter 13, we will study nonlinear systems,
and the principle of superposition will not apply to those systems.
The principle of superposition will be critical in dealing with Fourier series representa-
tions of discontinuous drive forces. When we represent the discontinuous drive force as a
Fourier series, the resulting differential equation will be of the form:

ẍ + 2γ ẋ + ω02 x = a0 + a1 cos(ωt) + a2 cos(2ωt) + · · · + b1 sin(ωt) + b2 sin(2ωt) + · · · (6.9.10)

We now know that in order to solve this type of equation, all we need to do is solve the
system of equations:

ẍi + 2γ x˙i + ω02 xi = ai cos(iωt) and ẍ + 2γ ẋ + ω02 x = bi sin(iωt) (6.9.11)

for i = 0, . . . , N . Note that for i = 0, cos(0) = 1, and we obtain the solution for the constant
term on the right-hand side of the differential equation. Similarly sin(0) = 0 gives the homo-
geneous solution. While there may be a large number of equations to solve, there are ways
that the process can be generalized. For example, notice that in the above case, by changing
the value of i, we are simply changing the drive frequency in the case of i > 0. Next, we
need to learn how to represent a discontinuous periodic function as a Fourier series.

6.9.2 Fourier Series

In this section, we will introduce an important result due to French mathematician and
physicist Joseph Fourier (1768–1830). Let us consider a function f (t) of the real variable t,
which is periodic with a period τ , such that f (t)=f (t + τ ). Fourier’s theorem states that,
under certain conditions which typically hold in mechanics, this periodic function can be
written as the sum of sine and cosine terms in the form of a Fourier series:

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 193 — #209

i i

Harmonic Oscillations 193

General Mathematical Form of Fourier Series

N
a0 X
f (t) = + an cos(nωt) + bn sin(nωt) (6.9.12)
2
n=1

where the lowest angular frequency is ω = 2π/τ , and the higher frequencies are integer
multiples of the fundamental angular frequency ω. The coefficients an and bn represent
the amplitudes of the various waves, and are known as Fourier coefficients. The Fourier
coefficients an and bn are computed by using the following general integral relationships:
(
0 m 6= n
Z τ /2
cos (nωt) cos (mωt) dt =
−τ /2 τ /2 m = n 6= 0
(
0 m 6= n
Z τ /2
sin (nωt) sin (mωt) dt = (6.9.13)
−τ /2 τ /2 m = n 6= 0
Z τ /2
cos (nωt) sin (mωt) dt =0 for all integers n and m
−τ /2

The calculation involves multiplying both sides of (6.9.12) with either cos(2πmt/τ ) or
sin(2πmt/τ ), and integrating over the whole period. The results of this calculation are
known as Fourier’s trick to many generations of physicists:

Fourier’s Trick for Finding the Fourier Amplitude Coefficients

2 τ /2
Z
an = f (t) cos(nωt)dt (6.9.14)
τ −τ /2

2 τ /2
Z
bn = f (t) sin(nωt)dt (6.9.15)
τ −τ /2

It is not always possible to compute coefficients of all values of n, so a finite Fourier series
approximates the original function f (t), and the approximation improves as N increases.
The Fourier series is generally presumed to converge everywhere except at discontinuities,
where it converges to the mean of the values of f (t) before and after the discontinuity.

Example 6.8: Fourier series calculation

Find the Fourier series for the following periodic sawtooth function:
x
f (x) = , for − π < x < π
π
f (x + 2πj) = f (x), for − ∞ < x < ∞ and j ∈ Z.

Solution:
Before finding the Fourier series, it is often helpful to plot the function f (x).

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 194 — #210

i i

194 Classical Mechanics: A Computational Approach

1.0

0.5

f(x)
0.0

-0.5

-1.0
-5 0 5
Position x

Although in this case the independent variable is x, the function f (x) is periodic in x, so
all of the formulas for Fourier series still apply. Next, it is important to identify the period.
As can be seen from the graph (and the definition of f ), the period is τ = 2π. Using the
formulas (6.9.14) and (6.9.15), we can compute the Fourier coefficients with ω = 2π/τ = 1.
First we find the an coefficients:
1 π x
Z
an = cos( 2πnx
2π ) dx
π −π π
2 Z π
1
= x cos(nx) dx
π −π
2 π
1 x sin(nx)
=
π n

−π

therefore we see that an = 0 for all values of n, because sin(nπ) = 0 for all integers n. Next,
we compute the bn coefficients.

1 π x
Z
bn = sin( 2πnx
2π ) dx
π −π π
2 Z π
1
= x sin(nx) dx
π −π
2 2
=− cos(nπ) + 2 2 sin(nπ)
πn π n
Again, we use sin(nπ) = 0 for integer values of n. However, note that cos(nπ) = (−1)n .
Therefore, we find that:
2 (−1)n+1
bn =
πn
The resulting Fourier series is therefore:
∞
2 X (−1)n+1
f (x) = sin(nx)
π n
n=1

When x = ±π, the Fourier series converges to 0, which is the mean of the left and right
limits of f (x) at x = ±π.

We can use Python to numerically compute the Fourier coefficients. Using a computer
to numerically find Fourier coefficients can be helpful in cases where the function is compli-

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 195 — #211

i i

Harmonic Oscillations 195

Fourier series and f(x)

1.0

0.5

0.0

−0.5

−1.0

−3 −2 −1 0 1 2 3
x
Figure 6.10: Results from Python code for calculating the Fourier coefficients an , bn of a
periodic function. The graph shows a comparison of the original function (orange in the
e-book) and the approximation using the Fourier series (blue in the e-book).

cated. With a computer, we can find as many terms in the Fourier series as we need (and
have the resources/time for!).
Algorithm 5 uses SciPy’s quad command to find the first 10 terms of the Fourier series
of the sawtooth function from Example 6.8. Note that quad returns two arguments, the first
being the value of the integral and the second argument being the error. Notice that we
defined f (x) as its own function. In addition, we also defined two functions a coefficient and
b coefficient, which are functions of n, and correspond to an and bn . The coefficients an and
bn are numbers found by integrating f (x) cos(2πnx/tau) and f (x) sin(2πnx/tau), respec-
tively. In order to perform the integrals over x, quad needs the integrand to be defined as a
function of x, which is why we defined the additional functions a integrand and b integrand.
We then store the Fourier coefficients as an array. Finally, we need to plot the result-
ing Fourier series. In order to make the plot, we need to evaluate the Fourier series at
several values of x. In Algorithm 5, we used 100 values of x between x = −tau/2 = −π
and x = tau/2 = π. Recall that in NumPy if we compute a function of an array, such as
cos ([1, 2, 3]) we obtain [cos(1), cos(2), cos(3)], hence the variable fourier series is actually an
array. The definition of fourier series is a bit complex, so make sure you understand what
is being done. Finally, we plot the Fourier series and the original function together, and the
result is shown in Figure 6.10, where f (x) is the orange line and the 10-term Fourier series
is in blue. If we increased the number of terms (by changing the value of num fc), we would
get a better approximation.

6.9.3 Example of Superposition Principle and Fourier Series

To conclude this section, we will present an example of a problem where a damped harmonic
oscillator is driven by a discontinuous periodic force. The following problem ties together
the concepts of the principle of superposition and Fourier series.
Consider the following discontinuous periodic force F (t):
t
F (t) = mA for − jπ < x < jπ (6.9.16)
π
where m is the mass of the oscillator, j is an integer, and A has units of acceleration per
unit time. This is essentially the same function from Example 6.8. It may seem strange to

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 196 — #212

i i

196 Classical Mechanics: A Computational Approach

import numpy a s np
from s c i p y . i n t e g r a t e import quad
import m a t p l o t l i b . p y p l o t a s p l t

tau = 2 . 0 ∗ np . p i #p e r i o d o f f u n c t i o n
num fc = 10 #number o f terms i n F o u r i e r s e r i e s

def f ( x ) : #f u n c t i o n f o r which t h e F o u r i e r s e r i e s i s c a l c u l a t e d
return x/np . p i

def a c o e f f i c i e n t ( n ) : #F o u r i e r c o e f f i c e n t , a n
def a i n t e g r a n d ( x ) : return f ( x ) ∗np . c o s ( 2 . 0 ∗ np . p i ∗n∗x/ tau )
i n t v a l u e , e r r o r = quad ( a i n t e g r a n d ,− tau / 2 . 0 , tau / 2 . 0 )
return 2 . 0 / tau ∗ i n t v a l u e

def b c o e f f i c i e n t ( n ) : #F o u r i e r c o e f f i c e n t , a n
def b i n t e g r a n d ( x ) : return f ( x ) ∗np . s i n ( 2 . 0 ∗ np . p i ∗n∗x/ tau )
i n t v a l u e , e r r o r = quad ( b i n t e g r a n d ,− tau / 2 . 0 , tau / 2 . 0 )
return 2 . 0 / tau ∗ i n t v a l u e

#compute a n ’ s and b n ’ s

a = np . a r r a y ( [ a c o e f f i c i e n t ( n ) f o r n in range ( 0 , num fc ) ] )
b = np . a r r a y ( [ b c o e f f i c i e n t ( n ) f o r n in range ( 0 , num fc ) ] )

#compute t h e F o u r i e r s e r i e s as a f u n c t i o n o f x f o r one p e r i o d
num xvals = 100 #number o f v a l u e s
x v a l s = np . l i n s p a c e (−tau / 2 . 0 , tau / 2 . 0 , num xvals )

f o u r i e r s e r i e s = a [ 0 ] / 2 . 0 + sum ( [ a [ n ] ∗ np . c o s ( 2 . 0 ∗ np . p i ∗n∗ x v a l s /
tau ) + b [ n ] ∗ np . s i n ( 2 . 0 ∗ np . p i ∗n∗ x v a l s / tau ) f o r n in range ( 1 ,
num fc ) ] )

plt . plot ( x vals , f o u r i e r s e r i e s )

plt . plot ( x vals , f ( x vals ) )
plt . x l a b e l ( ’ x ’ , f o n t s i z e =18)
plt . y l a b e l ( ’ F o u r i e r s e r i e s and f ( x ) ’ , f o n t s i z e =18)
plt . s a v e f i g ( ’ f i g u r e 1 f o u r i e r . e p s ’ , format= ’ e p s ’ )

p l t . show ( )

Algorithm 5: Python algorithm for evaluating Fourier coefficients for the sawtooth
function.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 197 — #213

i i

Harmonic Oscillations 197

have m as a term in the formula of the driving force, but it will simplify the mathematics
later.
The driving force in (6.9.16) acts on a damped harmonic oscillator with damping param-
eter γ and natural frequency ω0 , whose equation of motion is then:

F (t)
ẍ + 2γ ẋ + ω02 x = (6.9.17)
m
In order to solve for x(t) using the analytical methods described in this chapter, we will
need to use the Fourier series for F (t). From Example 6.8, we know that:
∞
2mA X (−1)n+1
F (t) = sin(nωt) (6.9.18)
π n
n=1

where we changed the independent variable from x to t, included the multiplicative constant
A, and explicitly included ω into the Fourier series. Note that ω = 1 because τ = 2π. So the
resulting differential equation to solve is:
∞
2A X (−1)n+1
ẍ + 2γ ẋ + ω02 x = sin(nωt) (6.9.19)
π n
n=1

However, the work done in this chapter had cosine terms on the right-hand side of the
differential equation. In order to use the solutions developed in this chapter, we will need
to use the identity: sin(a) = cos(a − π/2). Doing so results in (6.9.19) becoming:
∞
2A X (−1)n+1
ẍ + 2γ ẋ + ω02 x = cos(nωt − π/2) (6.9.20)
π n
n=1

To solve (6.9.20), we consider individual equations of the form:

2A (−1)n+1
ẍn + 2γ ẋn + ω02 xn = cos(nωt − π/2) (6.9.21)
π n
We use trial solutions of the form xn (t) = An cos(nωt − π/2 − φn ) and follow the procedure
outlined for the solution of the driven damped harmonic oscillator. The result is:

2A(−1)n+1
An = q 2 (6.9.22)
nπ ω02 − (nω)2 + 4γ 2 (nω)2

2γnω
tan φn = (6.9.23)
ω02 − (nω)2
Notice that the above solution is the same as (6.6.16) and (6.6.17), however the constant
coefficient is now 2A(−1)n+1 /nπ, and the drive frequency is nω instead of ω. Finally, we
can use the principle of superposition to find the solution to (6.9.20):
∞
2A X (−1)n+1
x(t) = q cos(nωt − π/2 − φn ) (6.9.24)
nπ 2
n=1 ω02 − (nω)2 + 4γ 2 (nω)2
where φn is defined in (6.9.23).

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 198 — #214

i i

198 Classical Mechanics: A Computational Approach

6.10 PHASE SPACE

Phase space diagrams are a useful concept in physics and other disciplines. A phase space
diagram usually plots the position x(t) of a particle on the x-axis and its corresponding
speed v(t) on the y-axis. Alternatively, one can also use the momentum p = mv(t) on the
y-axis, instead of the speed v(t). The result is a graph that gives a qualitative description of
the particle’s motion. Phase space diagrams are important in analyzing nonlinear systems,
and we will discuss the phase space diagrams of nonlinear systems in Chapter 13. For now,
we will explore the phase space diagrams associated with the harmonic oscillator.
Let us consider an undamped simple harmonic oscillator whose position and speed are
described by:
x(t) = A cos(ω0 t) (6.10.1)
v(t) = dx/dt = −Aω0 sin(ω0 t) (6.10.2)
We can eliminate the time t in these equations by using the trig identity sin2 (ω0 t) +
cos2 (ω0 t)
= 1:
x2 v2
2
+ 2 2 =1 (6.10.3)
A A ω0
This clearly represents an ellipse on the xv plane. We say that the phase space repre-
sentation of a simple harmonic oscillator is an ellipse, with semimajor axis equal to the
amplitude of oscillation a = A, and the semimajor y-axis equal to b = Aω0 . We know that
for the SHO, E = 21 kA2 and ω02 = k/m. Therefore, we can rewrite (6.10.3) as:

x2 v2
+ =1 (6.10.4)
2E/k 2E/m

This is an ellipse with semimajor axispequal to the amplitude of oscillation a = 2E/k

and the semimajor y-axis equal to b = 2E/m. Examples of harmonic oscillators with
different energies, E, are shown in Figure 6.11. In this case, we see that ellipses with larger
area correspond to the trajectories with larger energies.

1
Figure 6.11: Phase space of simple harmonic oscillator.

A = 1; ωd = 1; φ = 0; γ = 0.1;
x = A ∗ Exp[−γ ∗ t]Cos[ωd ∗ t − φ];
v = D[x, t];

ParametricPlot[{x, v}, {t, 0, 100}, PlotRange → All]

0.5

Out[ ]= x
-0.5 0.5 1.0

-0.5

The above algorithm can be used to generate the phase diagram for any system whose
position x(t) is known. Recall that the D in the above code is Mathematica’s command for
differentiation. While the phase space diagram does not provide new information about the
behavior of the harmonic oscillator, we will find in Chapter 13 that phase space diagrams
will become a critical tool for understanding the possible behaviors of a nonlinear system.
In Chapter 13, we will develop a method of finding a system’s phase space diagram, without
solving analytically the system’s equations of motion. It is often the case in nonlinear
systems that the equations of motion cannot be solved analytically. Therefore the phase
space diagram may be the only tool available in getting information about the behavior of
the system.

6.11 CHAPTER SUMMARY

A spring-mass system is an example of simple harmonic motion and described by:
r
d2 x k
F (x) = −kx + ω0 x = 0
2
ω0 =
dt2 m
and has the equivalent solutions:

x(t) = A cos ( ω0 t) + B sin ( ω0 t) x(t) = C cos ( ω0 t + φ)

and total kinetic energy

1 1
E = T + V = mv 2 + kx2
2 2
A simple plane pendulum is also an example of simple harmonic motion when the angle
of oscillation is small. The differential equation is:
g
θ̈ + sin θ = 0
l

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 201 — #217

i i

Harmonic Oscillations 201

In the limit of small oscillations θ0 1, the solution of this equation is:

s
g l
θ(t) = C cos ( ω0 t + φ) ω0 =
2
τ = 2π
l g

The general equations describing the damped harmonic oscillator are:

d2 x dx k
q
−γt −ωt
+ 2γ + x=0 x(t) = e ωt
+ A1 e ω = γ 2 − ω02

A0 e
dt2 dt m
where the constant γ = b/2m is the damping
q parameter. There are three possible types of
solutions, depending on whether ω = γ − ω02 is real, imaginary, or zero: (a) overdamped
2

oscillations (γ > ω0 ), (b) underdamped oscillations (γ < ω0 ), and (c) critically damped
oscillations (γ = ω0 or ω = 0). The corresponding solutions are:

q
x(t) = A0 eωt + A1 e−ωt e−γt ω= γ 2 − ω02 γ > ω0 Overdamped

q
x(t) = Ae−γt cos (ωd t − φ) ωd = ω02 − γ 2 γ < ω0 Underdamped
−ω0 t
x(t) = (A + Bt) e γ = ω0 Critically Damped

The energy and energy loss for a weakly damped oscillator are:
1 dE
E = kA20 e−2γt γ << ωd = −2γE (6.11.1)
2 dt
The dimensionless Q-factor expresses the degree of damping present in an oscillator, with
higher Q indicating a lower rate of energy loss:
ω0
Q=
2γ

The forced damped oscillator is described by the amplitude A (ω) and phase angle φ (ω),
both are functions of the external driving frequency ω:

ẍ + 2γ ẋ + ω02 x = F0 cos(ωt) x(t) = xc (t) + A cos (ωt − φ)

F0 /m 2γω
A= q tan φ =
2 ω02 − ω 2
ω02 − ω 2 + 4γ 2 ω 2

where xc (t) is one of the three possible solutions of the homogeneous equation. This
oscillator exhibits the phenomenon of resonance, with the following resonance frequency
and resonance amplitude:

F /m
q q
ωr = ωd2 − γ 2 = ω02 − 2γ 2 Amax = q0
2γ ω02 − γ 2

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 202 — #218

i i

202 Classical Mechanics: A Computational Approach

There is an analogy between the charge Q(t) in electrical RLC circuits and the position
x(t) in mechanical oscillators:

d2 Q(t) R dQ(t) 1 d2 x b dx k
+ + Q(t) = 0 + + x=0
dt2 L dt LC dt2 m dt m
The general mathematical form of a Fourier series for any periodic function is:
N
a0 X
f (t) = + an cos(nωt) + bn sin(nωt)
2
n=1

The Fourier coefficients can be found using Fourier’s trick:

2 τ /2
2 τ /2
Z Z
an = f (t) cos(nωt)dt bn = f (t) sin(nωt)dt
τ −τ /2 τ −τ /2

Phase space diagrams are important in analyzing nonlinear systems. They are plots of
the position x(t) of a particle on the x-axis and its corresponding speed v(t) (or
momentum) on the y-axis.

6.12 END-OF-CHAPTER PROBLEMS

Section 6.1–6.2: Differential Equations and the Simple Harmonic Oscillator
1. A mass m moves with simple harmonic motion along the x-axis. At time t = 0 it is
located x0 meters from the origin and moving towards the origin with a speed of v0 .
(a) Find the position, velocity, and acceleration as a function of time t.
(b) Find the amplitude and period of the motion.

2. Starting from the general expression x(t) = A cos ( ω0 t) + B sin ( ω0 t) for a harmonic
oscillation, show that an equivalent form
√ of this equation is given by the expressions:
x(t) = C cos ( ω0 t + φ) where C = A2 + B 2 tan φ = B/A.
3. A mass m is attached to a spring of spring constant k and moves with underdamped
harmonic motion along the x-axis, with the air resistance given by F = −bv.

(a) Show that the difference in times between two successive maxima is given by:
2π
∆t = p
ωo2 − γ 2

where γ = b/2m and ωo =

p
k/m.

(b) Show that the ratio of the amplitudes of two successive maxima of the position is
given by:
R = eγ∆t

where ∆t is the expression in part (a).

4. A mass m is attached to a vertical spring of spring constant k, which at equilibrium
stretches by an amount a. At time t = 0 a force F (t) = A sin(ωt) is applied to the

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 203 — #219

i i

Harmonic Oscillations 203

mass, where ω = k/m. Describe qualitatively the position and the speed of this
p

mass (x(t), v(t)) as a function of time t. You may neglect any damping present in the
system.
5. A spring has a spring constant of k = 8 N/m and a mass m = 0.5 kg is attached to
it. At time t = 0 a force F = 6 sin (4t), in Newtons, is applied to the mass. Find an
analytical expression for the position x (t) at time t.
6. The differential equation for an oscillating system is:
d2 z
− 16z − 160 cos (6t) = 0
dt2
with the initial conditions z(0) = 0 and dt (0)
dz
= 0. Plot the solution z(t) for several
periods and interpret the results.
7. A cylinder of mass m and radius R floats with its axis vertically in a liquid of density
ρ. The cylinder is given a small vertical displacement downwards and is released. Find
the period of oscillation.
8. A mass m moves in a region of xy-plane where the force is given by F = −kxî − ky ĵ.
(a) Show that depending on the initial conditions of the problem, the motion will be
an ellipse, a parabola, or a hyperbola.
(b) Find the potential energy of this oscillator.
9. Consider a mass m moving on a vertical wire inside a gravitational field. The shape of
the wire is a cycloid, described by the equations: x = aθ − a sin θ and y = −a + a cos θ.
(a) Find the speed of the mass at the bottom of the cycloid wire, if it is released at
the origin (0, 0) at time t = 0.
(b) Find the period of oscillation.
10. A mass m which is on a frictionless table is connected to two fixed points on opposing
walls, by two springs of equal natural length, of negligible mass and spring constants
k1 and k2 respectively. The mass is displaced horizontally and then released. Prove
that the period of oscillation is τ = 2π m/(k1 + k2 )
p

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 205 — #221

i i

Harmonic Oscillations 205

16. Find the frequencies of vibration for the system of particles of equal masses m con-
nected by springs with the same spring constant k, as indicated in Figure 6.16.

Figure 6.16: Problem 6.16: Two coupled masses hanging from verticle springs.

17. A pendulum has length L = 1 m, mass m = 0.1 kg and the gravitational acceleration
is g = 9.8 m/s2 .
(a) Find the kinetic energy T and potential energy V as a function of time t, when
the pendulum is released from rest at an angle of 3 degrees from equilibrium. Plot
T, V and the sum T + V on the same graph using the programming language of
your choice.
(b) Repeat part (a) when the pendulum is released from rest at an angle of 80 degrees
from equilibrium. How are the graphs of T, V , T + V different from part (a)?
(c) Find the time averages < T > and < V > over a complete period of oscillation
in part (a) and compare them with the total mechanical energy E = T + V .
18. A pendulum has length L = 1 m, mass m = 0.1 kg, and is released from rest at an
angle of 3 degrees from equilibrium. Use a value of g = 9.8 m/s2 for the gravitational
acceleration.
(a) Find the tension T of the string as a function of time t, and plot it using using
the language of your choice.
(b) At what value of time t is the tension a maximum, and at what value is it a
minimum? Do these results make sense?
19. In this chapter, we saw that for small amplitude oscillations, the period of a pendulum
is τ = 2π L/g.
p

(a) Show that in the more general case of oscillations with any amplitude θ0 , the
period is given by the expression:
Z θ0
dθ
τ = 8L/g
p
√
0 cos θ − cos θ0
(b) By expanding the integrand in a series of the angle θ, show that the integration
can be performed term-by-term to obtain the following expression:
θ2

τ = 2π L/g 1 + 0
p
16

The Calculus of Variations

In the next chapter, we will develop a new way of finding a system’s equations of motion. The
new method will involve a branch of mathematics called the calculus of variations, which
we introduce in this chapter. The calculus of variations focuses on finding the extrema of a
mathematical object called a functional. The foundations of the calculus of variations were
laid by Isaac Newton and Gottfried Wilhelm Leibniz (1646–1716) and was further developed
by the brothers Jakob and Johann Bernoulli (1655–1705 and 1667–1748, respectively). The
first major contributions to the calculus of variations were made by Leonhard Euler (1707–
1783), Joseph-Louis Lagrange (1736–1813), and Pierre-Simon Laplace (1749–1827). In this
chapter, we will introduce the calculus of variations and use it to solve problems related to
physics.

7.1 THE MOTIVATION FOR LEARNING THE CALCULUS OF VARIATIONS

Up to this point in the book, we have primarily found a particle’s equation of motion by solv-
ing a second-order differential equation found by using Newton’s second law. This method
can be straightforward when the forces acting on the system are mathematically simple
and the problem is best described in Cartesian coordinates. However, many problems in
physics require the use of coordinate systems other than Cartesian coordinates. In Chapter
3, we computed velocities and accelerations in polar, cylindrical, and spherical coordinates.
We found that those quantities took on forms that are much more complicated than when
they are expressed in Cartesian coordinates. Because Newton’s second law uses acceleration,
problems involving the description of motion in spherical or cylindrical coordinates can be
quite messy and difficult to solve.
An additional complication with Newton’s second law comes from the fact that it uses
force to describe motion. While there are many mathematically simple forces which can act
on a particle, such as weight, the normal force, and the linear and quadratic drag forces;
forces of constraint can either be very complicated or completely unknown. Examples of a
force of constraint would be a particle constrained to move along the surface of a sphere or
a bead sliding along a curved wire. Even a rolling object can involve constraint forces. If
the forces of constraint are unknown, it may not be possible to use Newton’s second law to
find the equations of motion.
Granted that one knows all of the forces acting on the particle, Newton’s second law can
be used to describe the motion of particles even in the cases described above. However, the
mathematics can be quite difficult and messy. We need another way to formulate equations
of motion which will allow to us to more easily deal with non-Cartesian coordinate systems
and with constraint forces, even in cases where the force of constraint is unknown. One such

209

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 210 — #226

i i

210 Classical Mechanics: A Computational Approach

r2
y2
r1
y1 y=y(x)

x1 x2
Figure 7.1: A function y(x) which describes a path between two points, r1 = (x1 , y1 ) and
r2 = (x2 , y2 ).

alternative formulation involves the use of a variational principle. A variational principle

involves finding a function which either minimizes or maximizes a particular quantity, which
is itself a function. The calculus of variations provides a means of finding the function which
extremizes the quantity of interest. For example, a chain that is suspended at both ends will
take on a shape that minimizes the chain’s gravitational potential energy. In this case, the
shape of the chain is the “function to be found” and finding that function (finding the shape
of the chain) involves minimizing the chain’s gravitational potential energy, which is itself
a function. The chain’s gravitational potential energy is a function of the chain’s shape.
Hence, the chain’s gravitational potential energy is a “function of a function,” because the
chain’s gravitational potential energy function depends on the chain’s shape function. In
this case, the calculus of variations can be used to find the function that corresponds to the
shape of the chain that minimizes the chain’s gravitational potential energy. This will be
analogous to finding the value that minimizes a function using the methods of elementary
calculus.
In this chapter, we will present the calculus of variations and use it to demonstrate the
solution to a few mathematical problems, all of which are relevant to physics. In the next
chapter, we will present a method which allows us to formulate a system’s equations of
motion using variational principles and the calculus of variations.

7.2 THE SHORTEST DISTANCE BETWEEN TWO POINTS—SETTING UP

THE CALCULUS OF VARIATIONS
We will begin our development of the calculus of variations by finding the shortest distance
between two points on a plane. The setup of this problem will provide a concrete example
of what we are trying to achieve with the calculus of variations. Let’s begin by considering
two points on the Cartesian plane as shown in Figure 7.1.
The quantity to be minimized is the length L of the curve joining the points r1 =
(x1 , yp
1 ) and r2 = (x2 , y2 ). The length can be found by integrating the differential distance
ds = dx2 + dy 2 along a small part of the curve:
Z r2
L= ds (7.2.1)
r1

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 211 — #227

i i

The Calculus of Variations 211

where r1 and r2 denote the location of the first and second point, respectively. As you may
recall from your introductory calculus course, it is common to factor out the dx2 term from
ds in order to obtain:
Z x2 q
L= 1 + y 0 (x)2 dx (7.2.2)
x1

where we used:
dy
dy = dx = y 0 (x)dx (7.2.3)
dx
In the introductory paragraphs of this chapter we said, “[the] calculus of variations provides
a means of finding the function which extremizes the quantity of interest.” In this case, the
“quantity of interest” is the length L of the curve y(x) joining the points r1 and r2 . Notice
that the length of the curve depends on the function y(x) (or more specifically, its derivative),
hence our “quantity of interest” is a function of the function y(x) and, therefore, we can
write the length of the curve as L = L(y). The problem specifically states that we want to
find the path which minimizes the distance between the points r1 and r2 . In other words,
we want to find the function y ∗ (x) such that L(y ∗ ) is a minimum value of L. As we will
see, the calculus of variations will provide us with a means of how to find y ∗ (x).
Before showing how to solve the above problem, it will be useful to establish some
notation. Notice that (7.2.2) takes the form:

General Functional for the Calculus of Variations

Z x2
J(y) = f y(x), y 0 (x); x dx (7.2.4)

x1

In the case of (7.2.2), the integrand f [y(x), y 0 (x); x] = 1 + y 0 (x)2 is a function of only
p

y 0 (x). The use of square brackets denotes the functional nature of f , i.e., that, loosely
speaking, its arguments are functions. The semicolon is used to separate out the independent
variable in the argument of f from the functions. In other words, f depends on functions
y(x) and y 0 (x) but can also contain terms of those functions’ independent variable, x. It
should be noted that there are no hard rules for these two notations, square brackets and
semicolons, parentheses and a comma could have been used instead. Many of the problems
from the calculus of variations that we will study will involve integrands that are functions of
y(x), y 0 (x), and x. However, in this case, because both y and y 0 depend on x, the integrand
is actually only a function of one variable, x. Later on in this chapter, we will explore cases
where the integrand is a function of multiple variables. Furthermore, J is sometimes referred
to as a functional. A functional is a mathematical object that takes a function as an input
and returns a scalar as an output. In the case of (7.2.2), the functional L can be rewritten
as L(y), and takes the equation of the curve y(x) as an input and returns as the length of
the curve (a scalar) as the output.
Next, we will develop a method for finding the function y(x) that is an extremum of the
functional J(y). Or, continuing with our example, we will learn how to find the curve y ∗ (x)
that minimizes the path length between two points on a plane.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 212 — #228

i i

212 Classical Mechanics: A Computational Approach

7.3 THE FIRST FORM OF THE EULER EQUATION

Let us return to introductory calculus for a moment. A common problem from introductory
calculus was that of finding a number x∗ , which is an extremum of f (x). This is done by
using the condition:

df
=0 (7.3.1)
dx

∗x=x
In other words, the condition for the value x∗ to be an extremum of f is that the first
derivative of f must be zero at x∗ . Equivalently, the rate of change in the function f is zero
at the extremum. It may not surprise you that in order to find the function y(x) that is
an extremum of the functional J, we will also perform a derivative and identify a condition
for that derivative to be equal to zero. The condition that leads to a zero derivative will
also give us a means of finding the extremum, y(x). Note that for clarity of notation, we
will drop the asterisk representing the extremum function. Furthermore, the point x∗ is
sometimes called a stationary point because infinitesimal displacements from x∗ does not
change the value of f (x) since the slope of f (x) is zero at x∗ . Likewise, in the case of the
functional J, the extremum function y(x) makes J stationary. In other words, infinitesimal
changes in the function y(x) do not change the value of J.
Continuing with the introductory calculus problem, you probably recall that one can
easily distinguish between a local maximum and a local minimum by calculating the second
derivative of f (x) and evaluating it at the stationary point. As it turns out, determining
if y(x) maximizes or minimizes J can be very difficult. Fortunately for applications in
mechanics, it is only necessary to know that y(x) makes a certain functional stationary and
whether y(x) maximizes or minimizes the functional will not matter.
Let us now consider the problem of finding the function y(x) that makes the functional
Z x2
J(Y ) = f Y (x), Y 0 (x); x dx (7.3.2)

x1

stationary. It will be helpful to think of the “shortest distance between two points on a
plane” example from the previous section as we work through the following derivation.
However, keep in mind that J is not necessarily a path length. Let us return to Figure 7.1
for a visualization; however, this time we will start with the true stationary function y(x)
and vary it by adding a small perturbation η(x) such that η(x1 ) = η(x2 ) = 0. A perturbation
is a small additive deviation from a quantity. Hence, the perturbation results in a function
Y (x) = y(x) + aη(x), shown in Figure 7.2, which no longer makes the functional J stationary
because of the perturbation. The parameter a provides a measure of the amount that y(x)
is perturbed.
The functional J becomes:

Z x2
J = f [Y, Y 0 ; x]dx (7.3.3)
x
Z x1 2
= f [y + aη, y 0 + aη 0 ; x]dx (7.3.4)
x1

i i

214 Classical Mechanics: A Computational Approach

Z x2
∂J ∂f d ∂f
= − η(x)dx (7.3.9)
∂a x1 ∂Y dx ∂Y 0
Next, we need to evaluate (7.3.9) at a = 0, resulting in
Z x2
∂J ∂f d ∂f
= − η(x)dx = 0 (7.3.10)
∂a x1 ∂y dx ∂y 0
where we have set the derivative equal
R to zero to ensure that J is stationary.
Equation (7.3.10) is of the form, h(x)η(x)dx = 0, and must be true for any continuous
function η(x). It can be shown that if h(x)η(x) = 0 for any η(x) then h(x) = 0 for all x.
R

The end result is the Euler equation:

The Euler Equation

∂f d ∂f
− =0 (7.3.11)
∂y dx ∂y 0

which provides a means for finding the function y(x) that makes J stationary. Equation
(7.3.11) is also sometimes referred to as the first form of the Euler equation and provides
the promised condition for finding the extremum y because, as we will see, (7.3.11) will
produce a differential equation for y(x).
Next, we will use the Euler equation to find the path of the shortest distance between
two points on the Cartesian plane.

Example 7.1: The shortest path between two points on the Cartesian plane
Find the path that has the shortest length between two points (x1 , y1 ) and (x2 , y2 ) on
the Cartesian plane.
Solution:
We start with (7.2.2),
Z x2 q
L= 1 + y 0 (x)2 dx
x1

where we see that f = 1 + y 0 (x)2 . Next, we compute the derivatives needed for the Euler
p

equation:

∂f ∂f y0
=0 and = .
∂y 0
q
∂y
1 + (y 0 )2
Because the first term of the Euler equation is zero, we have
 
d  y0 =0
q
dx 2
1 + (y )
0

or, in other words, the term within the parentheses must be constant in x. Setting that
constant equal to c produces

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 215 — #231

i i

The Calculus of Variations 215

y0
q =c
1 + (y 0 )2
which leads to:
c
y0 = √ ≡m
1 − c2
The term on the right-hand side of the above equation is also a constant which we
defined to be m. As promised, the Euler equation leads to a differential equation from
which y can be found. It is easy to check that the differential equation y 0 = m results in:

y(x) = mx + b
where b is the constant of integration. Hence, the path with the shortest distance between
two points in the Cartesian plane is a straight line. No surprise. Computer algebra systems
sometimes have algorithms for handling problems involving the calculus of variations. The
following Mathematica code creates the Euler equation for this example.
<< VariationalMethods`

EulerEquations[Sqrt[1 + y 0 [x]∧ 2], y[x], x]

y 00 [x]
OUTPUT: − == 0
(1+y0 [x]2 )
3/2

y 00 [x]
DSolve − == 0, y[x], x
(1+y0 [x]2 )
3/2

OUTPUT: {{y[x] → C[1] + xC[2]}}

Similar to Python, Mathematica also has libraries which are called packages in Mathe-
matica. The << command tells Mathematica to import the VariationalMethods package,
which contains various functions which can be used to solve problems involving the cal-
culus of variations. The EulerEquations command creates the Euler equation for f . The
result of the command appears below it. Finally, we use DSolve in order to solve the dif-
ferential equation that results from the Euler equation. The result is the equation of a
line.

7.4 THE SECOND FORM OF THE EULER EQUATION

There is another form of the Euler equation which is useful when the integrand of J does not
explicitly depend on x, or in other words, ∂f /∂x = 0. Note that even with this condition,
df /dx is not necessarily zero. We begin by computing the derivative:

df ∂f dy ∂f dy 0
= + (7.4.1)
dx ∂y dx ∂y 0 dx
df ∂f ∂f
= y0 + y 00 0 (7.4.2)
dx ∂y ∂y

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 216 — #232

i i

216 Classical Mechanics: A Computational Approach

∂f df ∂f
y 00 = − y0 (7.4.3)
∂y 0 dx ∂y
In addition, we can compute:

d 0 ∂f ∂f d ∂f
y = y 00 0 + y 0 (7.4.4)
dx ∂y 0 ∂y dx ∂y 0
Finally, we insert (7.4.3) into (7.4.4),

d ∂f df ∂f d ∂f
y0 0 = − y0 + y0 (7.4.5)
dx ∂y dx ∂y dx ∂y 0

d ∂f ∂f df
= y0 − + (7.4.6)
dx ∂y 0 ∂y dx
The term in the square brackets in (7.4.6) is Euler’s equation, and if y(x) makes J
stationary, then that term is zero according to the first form of the Euler equation. Therefore
we have:

d 0 ∂f
f −y =0 (7.4.7)
dx ∂y 0
Using the same kind of argument as we did in Example 7.1, we can set the term inside the
parentheses equal to a constant and therefore:

The Second Form of the Euler Equation

∂f
f − y0 = constant (7.4.8)
∂y 0

The second form of the Euler equation is often useful when the dependent variable, x
does not explicity appear in f , i.e., ∂f /∂x = 0. An example of such a case is when finding
the mathematical form of a geodesic on a sphere (see below).

7.5 SOME EXAMPLES OF PROBLEMS SOLVED USING THE CALCULUS OF

VARIATIONS
In this section, we will demonstrate the solution to three problems using the calculus of
variations. In the first problem, we will study the so-called brachistochrone problem, which
is a well-known problem in the calculus of variations. The second problem will involve
computing the shortest path between two points on the surface of a sphere. The third
problem will involve finding the minimal surface of revolution between two points.

7.5.1 The Brachistochrone Problem

The first problem we will solve is the classic problem known as the brachistochrone problem
which was solved by Johann Bernoulli in 1696. The brachistochrone problem involves finding
the path that minimizes the time of travel for a particle moving between two points (x1 , y1 )
and (x2 , y2 ) in a conservative force field as shown in Figure 7.3. We will consider a particle of
mass m starting at rest at the point (x1 , y1 ). The particle experiences a constant conservative
force, which for our problem will be the force of gravity F = mg ĵ, and we will define the
positive y-direction to be downwards.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 217 — #233

i i

The Calculus of Variations 217

(x1,y1)
x

F
(x2,y2)

Figure 7.3: The solution of the brachistochrone problem will provide the formula for the
dotted path which minimizes the time of travel for a particle moving between the two points
(x1 , y1 ) and (x2 , y2 ) in a constant conservative force field, F = mg ĵ.

The time of travel between the two points is:

Z (x2 ,y2 )
ds
t= (7.5.1)
(x1 ,y1 ) v

In our previous example,pwe used ds = 1 + y 0 (x)2 dx, however, we are going to find it more
p

convenient to use ds = 1 + x0 (y)dy, where x0 = dx/dy. Next, we need to find v in order

cot θ = β sin (φ − α) (7.5.16)

Strictly speaking, with (7.5.16) we are finished as we have θ(φ), but it is difficult to
interpret this result. In order to clarify our result, we will multiply (7.5.16) by R sin θ on
both sides and expand sin (φ − α) = cos α sin φ − sin α cos φ,

R cos θ = β cos α (R sin θ sin φ) − β sin α (R sin θ cos φ) (7.5.17)

Notice that the terms inside the parenthese in (7.5.17) are the coordinate transformations:

x =R sin θ cos φ


y =R sin θ sin φ (7.5.18)
z =R cos θ



and constant coefficients, A ≡ β cos α and B ≡ β sin α. Therefore, in Cartesian coordinates,

(7.5.17) can be expressed as:

z = Ay − Bx (7.5.19)
which is the equation of a plane passing through the center of the sphere. Because we are
restricted to the surface of a sphere of radius R, the geodesic is the path made by the
intersection of the plane with the sphere’s surface. Hence, the geodesic is the great circle
that lies at the intersection of the plane and the surface of the sphere. The great circle
between two points is the path often used for air and sea navigation, because it provides
the path with the shortest distance between those points.

7.5.3 Minimum Surface of Revolution

Consider two parallel circular wire rings that were dipped in soapy water. The soap film
between the rings will then form a surface. What is the shape of that surface? The answer
can be found by finding the minimum surface of revolution formed by a line joining two
points. Why does the soap film take the shape of the minimum surface of revolution? The
answer has to do with the free energy of the soap film. The free energy of the soap film is
E = σA where σ is the film tension and A is the area of the film. In equilibrium the film’s
energy is minimized. Hence, minimizing the area is the same as minimizing the energy of
the film.
Consider two points (y1 , z1 ) and (y2 , z2 ) as shown in Figure 7.5. We can generate a
surface of revolution by revolving a line that joins (y1 , z1 ) and (y2 , z2 ) about the z-axis. We
want to find the equation of the line joining (y1 , z1 ) and (y2 , z2 ), which minimizes the area
of the surface of revolution.
We begin by writing down the equation for the area of the surface of revolution. The
area is found by integrating the shaded area dA over the whole surface. Note that Figure
7.5 shows only a part of the infinitesimal area dA. Further note that dA actually extends
around the surface. We can think of dA as being the surface area of a cylinder, which has
area 2πrh where r is the radius of the cylinder and h is the height of the cylinder. In

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 221 — #237

i i

The Calculus of Variations 221

z
(y2,z2)

(y1,z1)

x
Figure 7.5: The surface of revolution formed by a line joining the points (y1 , z1 ) and (y2 , z2 ).

this case, r would be the distance y from the z-axis to the curve defining the surface of
revolution. The height of the cylinder ds is the differential distance along the curve z(y),
ds = dy 2 + dz 2 . Therefore, our functional for the area is:
p

Z y2 p
A= 2πy 1 + z 02 dy (7.5.20)
y1

We want to find z(y) that minimizes the area of the surface of revolution. Therefore
√ we
factored dy out of ds to produce (7.5.20), where z 0 = dz/dy. Next, we insert f = y 1 + z 02
into the first form of the Euler equation (7.3.11),

yz 0

d
√ =0 (7.5.21)
dy 1 + z 02
Note that ∂f /∂z = 0. Therefore we can write,

yz 0
√ =a (7.5.22)
1 + z 02
where a is a constant. Solving (7.5.22) for z 0 we obtain,
dz a
=p (7.5.23)
dy y − a2
2

After separating variables and integrating, we obtain,

y
z = a cosh−1 +b (7.5.24)
a
where b is a constant of integration. Note that both a and b can be determined once the
points (y1 , z1 ) and (y2 , z2 ) are specified. We can solve (7.5.24) for y to obtain,

z −b
y = a cosh (7.5.25)
a

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 222 — #238

i i

222 Classical Mechanics: A Computational Approach

The next example shows how to visualize the result.

Example 7.3: The soap film problem

Plot the surface of the soap film that connects two circular rings. Assume that each
ring is parallel to the xy−plane. Furthermore, each ring is centered on the z−axis and has
a radius of 3 units. One ring is located at z = 1 and the other is located at z = −1, where
the z coordinate is measured in the same units as the radius.
Solution:
The soap film will take on the shape of the minimum surface of revolution between the
two rings. We can use (7.5.25),

z −b
y(z) = a cosh
a
as the curve that describes the shape of the film’s surface in the yz−plane. The curve will
have end points (y1 , z1 ) = (3, −1) and (y2 , z2 ) = (3, 1).
The first step in the solution is to use the end points to solve for a and b in (7.5.25).
Note that if we insert the end points into (7.5.25), we find that y(−1) = y(1). Furthermore,
cosh is an even function. Together, the condition that y(−1) = y(1) and the even nature
of the cosh function require that b = 0.
Next, we find a by numerically solving y(1) = 3,

“master˙book˙file” — 2020/9/5 — 9:32 — page 224 — #240

i i

224 Classical Mechanics: A Computational Approach

Z n
x2 X
∂J ∂f d ∂f
= − ηi (x)dx (7.6.3)
∂α x1 i=1 ∂Yi dx ∂Yi0

Each perturbation ηi (x) is independent of one another. Therefore, in order for the derivative
in (7.6.3) to be zero at a = 0, we must have,

Multivariable Euler Equation

∂f d ∂f
− =0 (7.6.4)
∂yi dx ∂yi0

In other words, we will have an Euler equation for each dependent function yi . As before,
each Euler equation will produce an ordinary differential equation. However, these ODEs
are often coupled and can be difficult to solve in closed form.
For an example of f depending on multiple dependent variables, we can reconsider the
first problem we studied, the shortest path between two points on the Cartesian plane. When
we first solved that problem, we assumed y = y(x). However, if we wanted to consider all
possible paths, then we would need to consider curves which are parameterized using a
variable t,

x = x(t) and y = y(t)

The above generalization allows us to look at curved paths, such as spirals in the Cartesian
plane that can’t be written as y = y(x). For example, one of the simplest parameterizations
is that for the unit circle, where x(t) = cos t and y(t) = sin t.
The
p infinitesimal displacement along a parameterized path in the Cartesian plane is
ds = dx2 + dy 2 , which can be rewritten and inserted into (7.2.1),
Z t2 q
L= x0 (t)2 + y 0 (t)2 dt (7.6.5)
t1

where x0 = dx/dt and y 0 = dy/dt. Using f = x02 + y 02 , and the Euler equations:
p

d ∂f ∂f d ∂f ∂f
− =0 and − =0
dt ∂x0 ∂x dt ∂y 0 ∂y
we find,

x0 y0
=a and =c (7.6.6)
x02 + y 02 x02 + y 02
p p

where a and c are constants. Note that we used the fact that ∂f /∂x = ∂f /∂y = 0, in order
to set ∂f /∂x0 and ∂f /∂y 0 equal to the constants a and c, respectively. By dividing the two
equations in (7.6.6), we get,

y0 dy c
0
= = =m (7.6.7)
x dx a
Integrating (7.6.7), we find y = mx + b, where b is the constant of integration. In other
words, we again get the equation for a line.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 225 — #241

i i

The Calculus of Variations 225

7.7 CHAPTER SUMMARY

The calculus of variations provides us a means of finding a function y(x) that extremizes
the functional:
Z x2
J= f [y(x), y 0 (x); x]dx
x1

Common problems in the calculus of variations involve finding the path of the shortest
distance between two points, finding the path that minimizes the time of travel between
two points, and finding the minimum surface area for a surface of revolution, to name a
few.
The function y(x) can be found using Euler’s equation:
d ∂f ∂f
0
− =0
dx ∂y ∂y
Euler’s equation will produce a differential equation, which can be solved for y(x). An
alternative form of the Euler’s equation is,
∂f
f − y0 = constant
∂y 0
which can be used when ∂f /∂x = 0.
Problems involving finding the extrema of a functional which depends on several depen-
dent variables {yi (x)}, for i = 1, 2, . . . , n, of the form,
Z x2
J= f [yi (x), yi0 (x); x]dx
x1

can be solved using multiple Euler equations,

d ∂f ∂f
− =0
dx ∂yi0 ∂yi
The result is typically a system of n-coupled ordinary differential equations, which can
be solved, either numerically or in closed form, for each yi (x).

7.8 END-OF-CHAPTER PROBLEMS

Section 7.2: The Shortest Distance Between Two Points—Setting Up the Cal-
culus of Variations
1. Fermat’s principle states that the path taken between two points by a light ray is the
path that can be traversed in the least time. Consider a light ray traveling between
Points 1 and 2 in the Cartesian plane. The points are in a medium with a spatially
varying index of refraction, n(x, y). Show that the path taken by the light ray must
minimize the functional,
Z 2
n(x, y)
q
t= 1 + y 0 (x)2 dx (7.8.1)
1 c

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 226 — #242

i i

226 Classical Mechanics: A Computational Approach

(x1,y1) (x2,y2)

θ1 θ2

(xr,yr)
x

Figure 7.6: Problem 7.5: The path of a light ray in the xy-plane reflecting off of a mirror
in the xz-plane.

where y 0 = dy/dx and the index of refraction is defined as the ratio:

c
n=
v
where c is the speed of light in vacuum, and v is the speed of light inside the medium.
2. What is the length of the path that joins two points on the surface of a cylinder
with radius R? Use cylindrical coordinates (r, θ, z) and assume that the curve can be
written as θ(z).
Sections 7.2 and 7.3: The First and Second Form of the Euler Equation
3. Find the path that makes the functional,
Z 1
J(y) = 1 − y 0 dx,
p
x2
0

stationary with y 0 = dy/dx. The path must connect the points (0, 0) and (1, 1).
4. Find the shortest path between two points on the surface of a cylinder with radius R.
Consult Problem 2, for how to set up the necessary functional.

5. Consider the light ray (solid line) reflecting off of the surface of a plane mirror at the
point (xr , yr ) as shown in Figure 7.6. The surface of the mirror is in the xz-plane,
and the light ray is in the xy-plane. Using Fermat’s principle from Problem 1, prove
the law of reflection. In other words, show that the angle of incidence θ1 is equal to
the angle of reflection θ2 in the figure below. Although not necessary, you may find
it helpful to set x1 = 0 and yr = 0 as suggested in Figure 7.6. To solve the problem,
compute the time it takes to travel along the path in Figure 7.6, then show that the
time is a minimum when θ1 = θ2 . You may start by assuming that the point (xr , yr )
is in the same vertical plane as (x1 , y1 ) and (x2 , y2 ). If you were not to make that
assumption, then explain how you would need to change your solution in order to
prove the law of reflection.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 227 — #243

i i

The Calculus of Variations 227

(x1,y1) y

θ1

(xr,yr) n1

n2
x
θ2

(x2,y2)

Figure 7.7: Problem 7.6: The path of a light ray in the xy-plane leaving a medium with an
index of refraction n1 and entering a medium with an index of refraction n2 .

6. Consider the light ray shown in Figure 7.7. The light ray starts in a medium with a
spatially uniform index of refraction n1 and, at the point (xr , yr ) enters a new medium
with a spatially uniform index of refraction n2 . Using Fermat’s principle from Problem
1 to minimize the time of travel between the two points (x1 , y1 ) and (x2 , y2 ), prove
Snell’s law, n1 sin θ1 = n2 sin θ2 and that the point of refraction (xr , yr ) is coplanar
with the points (x1 , y1 ) and (x2 , y2 ). As before, the index of refraction is defined as
the ratio n = vc , where c is the speed of light in vacuum and v is the speed of light
inside the medium.

7. Find the extremum of the functional,

Z 1
J= y 02 − y 2 + x dx

0

subject to the two boundary conditions: y(0) = 1, y 0 (0) = −1.

8. Show that
x2 cos α − 2xy sin α − y 2 cos α = β
(where α and β are constants) are the extrema of the functional,
Z p p
x2 + y 2 1 + y 02 dx

where y 0 = dy/dx.
Hint: Rewrite the functional in polar coordinates and use Mathematica’s Variational-
Methods package to get Euler’s equations.
9. Find the geodesics on the surface of the cone whose equation in cylindrical coordinates
is z = cr.
10. Imagine a ball rolling down the first half of the first loop of a cycloid (from x = 0 to
x = πa), defined by the equations:
y = a (1 − cos θ) x = a (θ − sin θ)
Show that the time of travel from x = 0 to x = a is equal to π a/g. Show that this
p

time is independent of the starting point.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 228 — #244

“master˙book˙file” — 2020/9/5 — 9:32 — page 230 — #246

i i

230 Classical Mechanics: A Computational Approach

8.1 AN INTRODUCTION TO THE LAGRANGIAN

In this section, we will motivate the material in the rest of the chapter by looking at a basic
example, the simple harmonic oscillator (SHO). Recall that a model system for the SHO is
a mass m connected to one end of a horizontal spring with a spring constant k. The other
end of the spring is connected to a support. The mass, connected to the spring slides along
a frictionless surface. We will use the variable x to measure the location of the mass relative
to equilibrium. In this case, we know that the kinetic and potential energies can be written
as,
1 1
T = mẋ2 and V = kx2 (8.1.1)
2 2
respectively. We know that for the SHO,
∂V
F =− = −kx (8.1.2)
∂x
is the force acting on the mass. Therefore, using Newton’s second law, F = mẍ, we can
write the equation of motion for the SHO as,

mẍ + kx = 0 (8.1.3)
You may recall that normally we define ω02 = k/m for the SHO, but for now we will leave
the equation of motion as it appears above.
There is another method of finding (8.1.3) that doesn’t directly involve Newton’s second
law. Notice that,

d ∂T d
= (mẋ) = ṗ = mẍ (8.1.4)
dt ∂ ẋ dt
Therefore, we can obtain (8.1.3) by writing,

d ∂T ∂V
+ =0 (8.1.5)
dt ∂ ẋ ∂x
Equation (8.1.5) is similar to the Euler equation in appearance, except that the functions
appearing in each derivative are different. We can make (8.1.5) look like the Euler equation
by defining a new function, the Lagrangian,

L ≡ T −V (8.1.6)
Notice that V does not depend on ẋ , and T does not depend on x, therefore ∂T /∂x = 0
and ∂V /∂ ẋ = 0 and we can write:

d ∂T d ∂L
= (8.1.7)
dt ∂ ẋ dt ∂ ẋ
and
∂V ∂L
=− (8.1.8)
∂x ∂x
And therefore we obtain from (8.1.5) the so-called Euler-Lagrange equation,

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 231 — #247

i i

Lagrangian and Hamiltonian Dynamics 231

The Euler-Lagrange Equation for One Degree of Freedom

d ∂L ∂L
− =0 (8.1.9)
dt ∂ ẋ ∂x

For the SHO,

1 1
L = mẋ2 − kx2 (8.1.10)
2 2
Inserting (8.1.10) into (8.1.9), we obtain (8.1.3). Hence we see that the equation of motion
for the SHO can be obtained by using the Euler-Lagrange equation.
We see that L = L(x, ẋ; t) and (8.1.9) implies that the integral L(x, ẋ; t)dt is stationary
R

i i

Lagrangian and Hamiltonian Dynamics 233

1 1
T = m ẋ2 + ẏ 2 = m`2 q̇ 2 (8.2.3)
2 2

V = mgy = −mg` cos q (8.2.4)

The resulting Lagrangian, L = T − V , is:
1
L (q, q̇) = m`2 q̇ 2 + mg` cos q (8.2.5)
2
Notice that the Lagrangian is a function of both the generalized coordinate q and the
generalized velocity q̇. The generalized velocity is a term used to describe the first derivative
of the generalized coordinate with respect to time.
Although we haven’t demonstrated that it is true for cases other than the simple har-
monic oscillator, we can say that the pendulum’s motion is the path q(t), which makes the
integral L (q, q̇) dt stationary. Therefore, the Euler-Lagrange equation,
R

d ∂L ∂L
− =0 (8.2.6)
dt ∂ q̇ ∂q
will produce the equation of motion. Inserting (8.2.5) into (8.2.6) gives the simple plane
pendulum equation,
g
q̈ + sin q = 0 (8.2.7)
`

8.3 HAMILTON’S PRINCIPLE

Now that we have developed an understanding of generalized coordinates, it is time that we
justify using the Euler-Lagrange equations to derive a system’s equations of motion. This
justification was done by Irish mathematician, Sir William Rowan Hamilton (1805–1865), in
papers published in 1834 [Hamilton(1834)] and 1835 [Hamilton(1835)]. Hamilton’s Principle
states:
Hamilton’s Principle

The path taken by a particle from the point r1 to the point rR2 during the time interval
t
t1 to t2 is the one that makes the action integral S = t12 Ldt stationary.

Hamilton’s principle is a cornerstone not just of classical physics, but also of quantum
physics. From Chapter 7, we know that the condition for S to be stationary is

The Euler-Lagrange Equation

d ∂L ∂L
− =0 (8.3.1)
dt ∂ q̇j ∂qj

where qj is the j th generalized coordinate, j = 1, 2, . . . , s, and s is the number of degrees of

freedom in the system.
How can we be certain that Hamilton’s principle produces the actual path that the
particle takes? We will demonstrate this using a system with one degree of freedom, but is
easily generalized to s degrees of freedom. We will follow the argument used to derive the

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 234 — #250

i i

234 Classical Mechanics: A Computational Approach

Q(t)=q(t)+aη(t)
q2

q1
q(t)

t1 t2
Figure 8.2: The function q(t) is the path that makes Ldt stationary and Q(t) is a pertur-
R

bation from q(t).

Euler equation in Chapter 7, and therefore we want to find a function q(t) that makes the
action integral,
Z t2
S= L (q, q̇) dt (8.3.2)
t1

stationary. Following the argument from Chapter 7, we will perturb q(t) using a function
η(t), which is zero at the limits of integration, as shown in Figure 8.2. The function Q(t) is
the perturbation of q(t) and the perturbation has a size a, which is constant.
Next, we will differentiate S with respect to a and the extrema function q(t) should be
found when,

∂S
=0 (8.3.3)
∂a

a=0
From Chapter 7, we know that the above derivative is zero when,

d ∂L ∂L
= (8.3.4)
dt ∂ q̇ ∂q
If we substitute,
1
L = mq̇ 2 − V (q) (8.3.5)
2
into (8.3.4), then we have:
d ∂V
(mq̇) = − (8.3.6)
dt ∂q
The right-hand side of (8.3.6) is the q component of the force acting on the particle.
Likewise, the left hand side of (8.3.6) is the time derivative of the momentum. Therefore,
we see that the path q(t) that makes the action integral S stationary is the one that
also solves Newton’s second law. Therefore, Hamilton’s principle does provide a means of
finding a system’s equation of motion. Note that the above work is not a proof of Hamilton’s
principle, but rather a demonstration of how it is capable of producing a system’s equation
of motion.

i i

Lagrangian and Hamiltonian Dynamics 237

R
x1
x2
m1

We use the horizontal dashed lined passing through the center of the Atwood machine
as the origin for our coordinate system. The coordinates x1 and x2 measure the distance
from the center of the Atwood machine to masses m1 and m2 , respectively. Because the
string is inextensible, the two coordinates x1 and x2 are not independent. For example,
as m1 falls down, m2 rises and vice versa. Therefore, there is one equation of constraint
among the coordinates:

` = x1 + x2 + πR (8.4.5)
The last term in the constraint is the length of the string that wraps around the pulley.
Of course an additional constraint is that the masses are each constrained to move only up
or down along a line, therefore y1 = y2 = z1 = z2 = 0, which are actually four constraints
(one on each coordinate) for a total of five constraints on the system (m = 5). Therefore
with two particles in the system N = 2, we have s = 3(2) − 5 = 1 degree of freedom for the
problem. We will rewrite the coordinates x1 and x2 in terms of the generalized coordinate
x using:

x1 = x (8.4.6)
x2 = ` − x − πR (8.4.7)

Notice that the velocities have the relationship, ẋ2 = −ẋ1 , and the accelerations have
the relationship, ẍ2 = −ẍ1 , which is what we would expect in this situation.
Next, we’ll find the kinetic and potential energies of the system. Notice that the pulley
has mass but the string does not. Therefore, the total kinetic energy includes the kinetic
energy of m1 , m2 , and the rotational kinetic energy of the pulley. You may recall from
your introductory physics course that rotational kinetic energy takes the form Trot = 21 Iω 2 ,
where I = 21 M R2 is the rotational inertia for the pulley, and ω = ẋ/R is the angular velocity
of the pulley. Therefore, the total kinetic energy for the system is:

1 1 1
T = m1 ẋ21 + m2 ẋ22 + Iω 2 (8.4.8)
2 2 2
2
1 1 1 ẋ
= m1 ẋ2 + m2 ẋ2 + I (8.4.9)
2 2 2 R

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 238 — #254

i i

238 Classical Mechanics: A Computational Approach

The potential energy takes the form of the gravitational potential energy. Measuring
upward as the positive direction, we have:

V = −m1 gx1 − m2 gx2 (8.4.10)

= −m1 gx − m2 g (` − x − πR) (8.4.11)

The resulting Lagrangian is:

1 1 1 I

L= m1 + m2 + ẋ2 + m1 gx + m2 g (` − x − πR) (8.4.12)
2 2 2 R2
Because s = 1, there is only one Euler-Lagrange equation for this system:

d ∂L ∂L
= (8.4.13)
dt∂ ẋ ∂x

d I
m1 + m2 + 2 ẋ = (m1 − m2 )g (8.4.14)
dt R

I
m1 + m2 + 2 ẍ = (m1 − m2 )g (8.4.15)
R

The generalized acceleration is:

(m1 − m2 )g
ẍ = (8.4.16)
m1 + m2 + M

2

where we used I = 12 M R2 . From the generalized acceleration, we can get the acceleration
of each mass because according to (8.4.5), ẍ1 = ẍ and ẍ2 = −ẍ.

Example 8.2 demonstrates several ideas. First, drawing a picture of the system can help
identify constraints on the coordinates. Second, Example 8.2 reminds us to carefully identify
all kinetic and potential energies in the system. It would be easy to overlook the rotational
kinetic energy of the pulley if you thought the pulley is massless, a common assumption
in introductory physics. Finally, Example 8.2 shows us that although it might be useful to
describe the motion in generalized coordinates (in this case x), the solution of the problem
needs to address what is being asked by the problem. In this case, we needed to make sure
that we showed how to express the acceleration of each mass in terms of the generalized
coordinate.
Example 8.3: Pendulum supported by a rotating disk
A pendulum is suspended from a massless disk with a radius R that is rotating with
a constant angular velocity ω as shown in the figure below. The pendulum consists of a
mass m attached to one end of a rigid massless rod of length `. The other end of the rod
is attached to the edge of the rotating disk. Find the equation of motion of the mass m.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 239 — #255

i i

Lagrangian and Hamiltonian Dynamics 239

y
ω
R x

ℓ
θ

m
Solution:
The coordinate system is defined in the figure where the coordinate axes are the dashed
lines. The mass is constrained to move in the xy-plane, hence z = 0 is our first equation
of constraint. There is an additional constraint which may not be immediately obvious.
However, if we write the location of the mass in terms of polar coordinates, then it will
become clear that this problem has only one degree of freedom. The coordinates of the
mass can be written as:

x = R cos (ωt) + ` sin θ and y = R sin (ωt) − ` cos θ (8.4.17)

In other words, we chose the origin to be the center of the rotating disk and the location
of the support (which is attached to the disk) appears as the first term in the equations
for x and y. We see that x and y are a function of the generalized coordinate θ. The x and
y component of the velocities are:

ẋ = Rω sin (ωt) + `θ̇ cos θ and ẏ = Rω cos (ωt) + `θ̇ sin θ (8.4.18)
Next, we will find the kinetic and potential energies. The kinetic energy T = 1
2 mv
2 can
be written using the velocity of the mass in the Cartesian plane, v = ẋî + ẏ ĵ,
1 1 1
T = mv 2 = mv · v = m ẋ2 + ẏ 2

2 2 2
The pendulum is in the Earth’s gravitational field, therefore the potential energy is
V = mgy. Inserting (8.4.17) and (8.4.18) into our equations for the kinetic and potential
energies results in the Lagrangian (after some algebra, see Problem 9),

1
L = m R2 ω 2 + `2 θ̇2 + 2R`ω θ̇ sin (θ − ωt) − mg (R sin (ωt) − ` cos θ) (8.4.19)

2
Inserting (8.4.19) into

d ∂L ∂L
− =0
dt ∂ θ̇ ∂θ
produces the equation of motion:

Rω 2 g
θ̈ + cos (θ − ωt) + sin θ = 0 (8.4.20)
` `

“master˙book˙file” — 2020/9/5 — 9:32 — page 241 — #257

i i

Lagrangian and Hamiltonian Dynamics 241

By inserting the constraints and (8.4.22) we can find the Lagrangian:

L =T − V
1
= m ṙ2 + r2 θ̇2 + ż 2 − mgz

2
1
= m ṙ2 + r2 ω 2 + 4r2 a2 ṙ2 − mgar2

2
Note that L = L (r, ṙ). To get the equation of motion for the bead, we can insert the
Lagrangian into the Euler-Lagrange equation,

d ∂L d m
= 2ṙ + 8r a ṙ = 1 + 4a r mr̈ + 8ma rṙ 
2 2 2 2 2 2

dt ∂ ṙ dt 2

(8.4.23)
∂L
=mr rω + 4a ṙ − 2ga
2 2



∂r
Equating the two terms above gives:

1 + 4a2 r2 r̈ + 4a2 rṙ2 + r ω 2 − 2ga = 0 (8.4.24)

We are interested in the case when the bead is traveling in a circle of radius R. There-
fore, inserting r̈ = ṙ = 0 and r = R into (8.4.24), we obtain

ω2
a=
2g

Example 8.4 demonstrates several important points. First, it demonstrates the need to
use the velocity in cylindrical coordinates. The velocity of a particle in polar, cylindrical,
and spherical coordinates was discussed in Chapter 3. It is important when computing the
kinetic energy that the velocity vector be expressed in the same coordinate system that you
are using to solve the problem. Second, although this problem didn’t specifically ask for the
equation of motion, we found it easier to answer the problem by finding it. Once we had the
general equation of motion for the bead (8.4.24), we inserted the specific conditions for the
motion of interest. Finally, Example 8.4 shows how the Lagrange formulation doesn’t need
to know all of the forces acting on the system. We do not know the mathematical form of the
force that keeps the bead on the wire. The aforementioned ignorance is no problem, all we
needed for the Lagrange formulation was how the value of the z coordinate was constrained.

Example 8.5: A moving inclined plane

A frictionless inclined plane of mass M , length `, and incline θ is sliding along a level
horizontal frictionless surface, while a particle of mass m slides on the inclined plane. Find
the acceleration of the inclined plane and the particle.
Solution:
We begin by drawing a picture of the system.

i i

i i

244 Classical Mechanics: A Computational Approach

up to a particular order. In this section, we will focus on the fourth-order Runge-Kutta

(RK4) method which uses a fourth-order expansion. Runge-Kutta methods tend to be used
instead of the Euler method because for a given step size, RK methods tend to be more
accurate and the Euler method tends not to be very stable. Stability, in this context, is the
ability to faithfully reproduce the ODEs solution over long periods of time.
Before discussing the RK4 method further, it is worth pointing out that there are other
methods for numerically solving ODEs besides the Euler and RK methods. Other methods
include the Bulirsch-Stoer method and Predictor-Corrector methods. These methods use
extrapolation and other techniques for finding the solution. Bulirsch-Stoer and Predictor-
Corrector methods can be faster and more accurate methods than RK methods when solving
ODEs. However, RK methods are very common in physics and we will focus our presentation
on RK4 in particular. The interested reader is directed to [Press et al. (2007)] and the
Massachusetts Institute of Technology OpenCourseWare website [MIT(2018)], which has
an introductory course on numerical methods.
Similar to Chapter 2, we will focus on first-order ODEs of the form,

ẋ = f (x, t), (8.5.1)

which has a solution x(t). Recall that second order ODEs can be rewritten as a system of
coupled first-order ODEs. Therefore, we will focus on the solution of first-order equations.
Like with the Euler method, the initial condition, x (t0 ) must be specified. We will then
numerically solve for the solution up to a time tmax . Also, similar to the Euler method,
we will break the time range into n equally-spaced time steps of size ∆t = (tmax − t0 )/n,
creating a list of times, ti = t0 + i∆t. We then compute the value of x at each time step ti
(i = 1, . . . , n). We store the solution in an array x(ti ), which has n + 1 elements. The first
element of the array x is the initial condition x(t0 ). The equations for RK4 are

k1 =dt f (ti , xi )



∆t

k1 
k2 =dt f ti + , xi +


2 2





∆t

k2 
k3 =dt f ti + , xi + (8.5.2)
2 2 


k4 =dt f (ti + ∆t, xi + k3 )






k1 k2 k3 k4 
xi+1 =xi + + + + + O ∆t 
5 

6 3 3 6


The RK4 method computes the solution’s derivative once at each end point of an interval
and twice at the mid-point of the interval as shown in Figure 8.3. In Figure 8.3, the interval
is ∆t = ti+1 −ti and the circles represent the points in time where the derivative is evaluated.
The solid lines through the circles show the derivative of the solution (dashed line) at those
points.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 245 — #261

i i

Lagrangian and Hamiltonian Dynamics 245

xi+1

Example 8.6: Numerical solution to Example 8.3

Consider the equation of motion found for the driven pendulum in Example 8.3,

Rω 2 g
θ̈ + cos (θ − ωt) + sin θ = 0
` `
Find θ(t) for t = 0 to t = 20 seconds using the values R = 0.2 m, ω = 3π/2 rad/s, ` = 1.0
m, g = 9.8 m/s2 , and initial conditions θ(0) = 1 rad and θ̇(0) = 0 rad/s.
Solution:
The first step in solving the problem is to create two first-order ODEs from the equation
of motion. Defining a new variable v = θ̇, we have as our new equations:

θ̇ =v


Rω 2 g (8.5.4)
v̇ = − cos (θ − ωt) − sin θ
` `
Next, we need to modify (8.5.2) for a system of two equations. For simplicity of nota-
tion, we will define the following functions:

Rω 2 g
f (t, θ, v) = v and cos (θ − ωt) − sin θ
h(t, θ, v) = −
` `
Then for a system of two equations, (8.5.2) becomes:

dt k1 j1
k1 =dt f ti + , θi + , vi + j1 =dt h (ti , θi , vi )
2 2 2

dt k1 j1 dt k1 j1
k2 =dt f ti + , θi + , vi + j2 =dt h ti + , θi + , vi +
2 2 2 2 2 2

dt k2 j2 dt k2 j2
k3 =dt f ti + , θi + , vi + j3 =dt h ti + , θi + , vi +
2 2 2 2 2 2
k4 =dt f (ti + dt, θi + k3 , vi + j3 ) j4 =dt h (ti + dt, θi + k3 , vi + j3 )
1 1
θi+1 =θi + (k1 + 2k2 + 2k3 + k4 ) vi+1 =vi + (j1 + 2j2 + 2j3 + j4 )
6 6
Algorithm 7 uses the RK4 method to solve (8.5.4) in Python.
The red (in the e-book) curled arrows in the code above denote a line continuation i.e.,
the text following a curly arrow is a continuation of the line above it. Notice that we used
f and g to denote our derivatives. Also, notice that the method is similar to what was done
for (8.5.3) except now we need to include the appropriate steps for each variable θ(ki ) and
v(ji ). A plot of θ(t) for the parameters listed in the problem appears below. Notice that
the rotating disk causes the pendulum to deviate from the sinusoidal-like behavior of the
undriven simple plane pendulum. We encourage you to try Problem 26 and explore how
θ(t) changes as you change the parameters in this example.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 248 — #264

i i

248 Classical Mechanics: A Computational Approach

Of course, this isn’t the end of the story when it comes to ODE solvers. Modern ODE
solvers often include an adaptive step-size control, where the value of dt changes. The value
of dt may increase when the function isn’t changing rapidly, and decreases when the function
does change rapidly. Equipped with adaptive step size, an ODE solver like RK4 can become
significantly more computationally efficient. Adaptive step-size controllers are beyond the
scope of this book, but the interested reader is encouraged to consult [Press et al. (2007)]
for more information.
It is often the case that you will use a numerical ODE solver that comes packaged
with your software of choice, or one that is included in a particular library. While you may
never need to write an ODE solver yourself, it is important to understand how they work,
so that you know how to interpret the results that they give. Errors from either misuse,
instability, or a variety of other mathematical reasons do occur. You need to be able to
identify when they happen and not assume the “computer is always right.” Mathematica’s
NDSolve is a powerful numerical ODE solver, which has built-in adaptive step size. The
NDSolve command has many features, including the ability to choose which algorithm
is being used. The interested reader should consult Mathematica’s online documentation
[MMA(2018)] for more information. Likewise, Python has the command odeint in the SciPy
library, which is based on the FORTRAN library ODEPACK. We comment on odeint in
more detail in Chapter 13. We have already seen both NDSolve and odeint used in this
book.
In the next section, we will return to the Lagrange formalism and study more closely
systems with constraints.

8.6 CONSTRAINT FORCES AND LAGRANGE’S EQUATION WITH UNDE-

TERMINED MULTIPLIERS
In Section 8.2, we studied the example of the simple plane pendulum and we found that
there was a constraint on the coordinates x and y. For a pendulum of length `, there is the
following relationship between x and y,

x2 + y 2 = `2 (8.6.1)
Notice that (8.6.1) takes the form,

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 249 — #265

i i

Lagrangian and Hamiltonian Dynamics 249

import numpy a s np

#d e f i n e c o n s t a n t s
a = 0.2 #r a d i u s o f d i s k
omega = 3/2∗ np . p i #a n g u l a r v e l o c i t y o f d i s k
g = 9.8 #a c c e l e r a t i o n due t o g r a v i t y
e l l = 1.0 #l e n g t h o f pendulum

def f ( vel ) : #e q u a t i o n f o r ” t h e t a dot ”

return vel

d e f h ( time , a n g l e ) : #e q u a t i o n f o r ”v dot ”
r e t u r n −(a ∗omega ∗∗2/ e l l ) ∗np . c o s ( a n g l e −omega∗ time ) −(g / e l l ) ∗
,→ np . s i n ( a n g l e )

#d e i n e i n i t i a l c o n d i t i o n s
theta0 = 1.0 #i n i t i a l c o n d i t i o n f o r t h e t a
v0 = 0 #i n i t a l c o n d i t i o n f o r v
tmax = 2 0 . 0 #maximum time
dt = 0 . 0 1 #s t e p s i z e
n = i n t ( tmax/ dt ) #number o f s t e p s
t = np . l i n s p a c e ( 0 , tmax , n ) #a r r a y o f t i m e s
v = np . z e r o s ( n ) #a r r a y t h a t s t o r e s t h e s o l u t i o n
,→ f o r v
t h e t a = np . z e r o s ( n ) #a r r a y t h a t s t o r e s t h e s o l u t i o n
,→ f o r t h e t a

theta [ 0 ] = theta0
v [ 0 ] = v0

f o r i in range (1 , n) : #l o o p f o r f i n d i n g t h e s o l u t i o n
k1 = dt ∗ f ( v [ i −1])
j 1 = dt ∗h ( t [ i −1] , t h e t a [ i −1])

k2 = dt ∗ f ( v [ i −1]+ j 1 / 2 . 0 )
j 2 = dt ∗h ( t [ i −1]+dt / 2 , t h e t a [ i −1]+k1 / 2 . 0 )

k3 = dt ∗ f ( v [ i −1]+ j 2 / 2 . 0 )
j 3 = dt ∗h ( t [ i −1]+dt / 2 , t h e t a [ i −1]+k2 / 2 . 0 )

k4 = dt ∗ f ( v [ i −1]+ j 3 )
j 4 = dt ∗h ( t [ i −1]+dt , t h e t a [ i −1]+k3 )

t h e t a [ i ] = t h e t a [ i −1] + k1 / 6 . 0 + k2 / 3 . 0 + k3 / 3 . 0 + k4 / 6 . 0
v [ i ] = v [ i −1] + j 1 / 6 . 0 + j 2 / 3 . 0 + j 3 / 3 . 0 + j 4 / 6 . 0

Algorithm 7: The Python code for Example 8.6.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 250 — #266

i i

250 Classical Mechanics: A Computational Approach

f (x, y) = constant (8.6.2)

where f (x, y) = x2 + y 2 . We will call equations of the form (8.6.2) constraint equations and
in this section, we will show how the Lagrange formulation allows us to directly handle
constraints without the substitutions done in earlier chapters. The method is sometimes
called the method of Lagrange multipliers or Lagrange undetermined multipliers. In addition,
we will see that by using the constraint equation, we can also find the force causing the
constraint. In the case of the pendulum, the constraint force is the tension in the pendulum’s
rod.
Before developing the method of Lagrange multipliers, it is important to discuss types of
constraints. In this chapter, we will focus on holonomic constraints. Holonomic constraints
are constraint equations which can be written as an algebraic relationship between the
coordinates. Equation (8.6.2) is an example of such a constraint. Sometimes, holonomic
constraints are written in the form,

f (xi ; t) = 0 (8.6.3)
where the abbreviation xi says that f is a function of several coordinates (x, y, and z in
Cartesian coordinates) and the semicolon shows f ’s indirect dependence on time (through
the coordinates xi ). For example, we can rewrite (8.6.1) in the form of (8.6.2) by writing
f (x, y) = x2 + y 2 − `2 = 0. Constraints that must be written as relationships involving veloci-
ties are called nonholonomic constraints, unless the constraint equation can be integrated to
produce a relationship between coordinates. Although we won’t discuss nonholonomic con-
straints further, the interested reader can consult the classic text by Thornton and Marion
[Thornton and Marion(2004)].
Now we will derive Lagrange’s equations with undetermined multipliers. We will derive
the equations for a system with two degrees of freedom and then discuss a generalization for
systems with higher degrees of freedom. The system will have one holonomic constraint of
the form, f (x, y) = constant. We know that finding the equations of motion involves finding
functions x(t) and y(t) that make the action integral,
Z t2
S= (8.6.4)

L X, Ẋ, Y, Ẏ dt
t1

stationary. We then write,

)
X(a, t) =x(t) + aηx (t)
(8.6.5)
Y (a, t) =y(t) + aηy (t)
where x(t) and y(t) are the functions that make (8.6.4) stationary, and ηx (t) and ηy (t) are
perturbations from x and y with ηx (t1 ) = ηy (t1 ) = 0 and ηx (t2 ) = ηy (t2 ) = 0. By inserting
(8.6.5) into (8.6.4) and using (7.6.3) we obtain:
Z t2 Z t2
∂S ∂L d ∂L ∂L d ∂L
= − ηx dt + − ηy dt (8.6.6)
∂a t1 ∂X dt ∂ Ẋ t1 ∂Y dt ∂ Ẏ
However, (8.6.6) is only true for motions that satisfy the constraint f (X, Y ) = constant.
Therefore, along the path of the particle,
df ∂f ∂f
= ηx + ηy = 0 (8.6.7)
da ∂X ∂Y

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 251 — #267

i i

Lagrangian and Hamiltonian Dynamics 251

Recall that ∂X/∂a = ηx and ∂Y /∂a = ηy . Solving (8.6.7) for ηy /ηx and inserting the result
into (8.6.6), we obtain:
Z t2
∂S ∂L d ∂L ∂L d ∂L ∂f /∂Y
= − − − ηx dt (8.6.8)
∂a t1 ∂X dt ∂ Ẋ ∂Y dt ∂ Ẏ ∂f /∂X
Setting ∂S/∂a = 0 at a = 0 gives (recall that X(a = 0, t) = x(t), and similar for Y (0, t)),
−1 −1
∂L d ∂L ∂f ∂L d ∂L ∂f
− = − (8.6.9)
∂x dt ∂ ẋ ∂x ∂y dt ∂ ẏ ∂y
The left-hand side of (8.6.9) depends on derivatives with respect to x and ẋ, while the
right-hand side depends on derivatives with respect to y and ẏ. Note that x and y are both
functions of t. In order for the two sides of (8.6.9) to be equal, they must both be equal
to a function of only t. We will define that function to be −λ(t), and we call λ Lagrange’s
undetermined multiplier. We do not know λ, hence the name “undetermined multiplier.”
Setting each side of (8.6.9) equal to −λ we obtain,

∂L d ∂L ∂f 
− = − λ(t) 
∂x dt ∂ ẋ ∂x

(8.6.10)
∂L d ∂L ∂f 
− = − λ(t)  
∂y dt ∂ ẏ ∂x
Equations (8.6.10) along with f (x, y) = constant will produce the equations of motion for
the system.
Of course, not all systems are limited to only two degrees of freedom and one constraint
equation. In general, systems will have n degrees of freedom and m equations of constraint.
We have already seen how to generalize the Lagrange formulation for s degrees of freedom,
in which case we have s Euler-Lagrange equations to solve. However, multiple constraint
equations means that there will be multiple derivatives dfj /da (j = 1, . . . , m, with m equal
to the number of constraints) that we will have to include in the derivation above. The
result is:
Euler-Lagrange Equation with Undetermined Multipliers
m

∂L d ∂L X ∂fj 
− =− λj (t)

∂qi dt ∂ q˙i

j=1
∂qi (8.6.11)

fj (qi ; t) =0



where i = 1, . . . , n, j = 1, . . . , m, and we have used generalized coordinates qi . Notice that in

(8.6.11) there are m undetermined multipliers λj , one for each constraint equation fj (qi ; t) =
constant.
Before moving on to an example, there are two important points we should make. First,
you might be wondering why one would use Lagrange multipliers. Using Lagrange multipliers
can save one from doing a lot of messy algebra when it comes to solving f for the relationship
between coordinates.
√ For example, in the pendulum problem, if we solved f for y, we would
have y = ± `2 − x2 . We would have to deal with both the positive and negative roots.
However, as we will see in the example below, using Lagrange’s multipliers will help simplify
the procedure. That said, the power of modern computer algebra systems may make this
argument less compelling than it once was.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 252 — #268

f (r; t) = r − R = 0 (8.6.15)
The force of constraint is the force that keeps the particle on the sphere and ṙ = 0.
When the particle leaves the sphere’s surface, ṙ 6= 0; therefore, there will be an acceleration
in the radial direction. We can find the relevant constraint force by finding:
∂f
Frc = λ
.
∂r
We know ∂f /∂r, so we need to find λ. We will find λ by setting up the Euler-Lagrange
equations for the system.
Using polar coordinates, we can write down the kinetic and potential energies for the
system:
1
T = m ṙ2 + r2 θ̇2

2
V =mgr cos θ
1
L = m ṙ2 + r2 θ̇2 − mgr cos θ

2
where we used the horizontal line in the above figure as y = 0 and therefore the potential
energy can be written as mgy with y = r cos θ. The Euler-Lagrange equations are:

∂L d ∂L ∂f
− =−λ (8.6.16)
∂r dt ∂ ṙ ∂r

∂L d ∂L ∂f
− =−λ . (8.6.17)
∂θ dt ∂ θ̇ ∂θ

By inserting the Lagrangian L and the constraint equation, we obtain:

mrθ̇2 − mg cos θ − mr̈ = − λ (8.6.18)

−mgr sin θ − mr θ̈ − 2mrṙθ̇ =0
2
(8.6.19)

Next, we apply the equation of constraint, r = R and ṙ = r̈ = 0, because we want to

know the angle θ just before the particle leaves the sphere’s surface.

mRθ̇2 − mg cos θ = − λ
(8.6.20)
−mgR sin θ − mR2 θ̈ =0
In the first equation above, we have found that λ = λ(θ, θ̇). The second equation will
give us a way of finding θ(t),
g
θ̈ = sin θ (8.6.21)
R
While we could attempt to integrate (8.6.21) and solve for θ, we actually don’t need
to do that. What we really need is θ̇, which can be found from (8.6.21) using,

dθ̇ dθ̇ dθ dθ̇

θ̈ = = = θ̇ (8.6.22)
dt dθ dt dθ

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 254 — #270

i i

254 Classical Mechanics: A Computational Approach

By inserting (8.6.22) into (8.6.21), we get:

Z Z
g
θ̇dθ̇ = sin θdθ
R
After doing the integral, we find:
g
θ̇2 = 2 (1 − cos θ)
R
and inserting the result into (8.6.18) to get:

λ = mg (3 cos θ − 2)

The constraint force is then:

∂f
Frc = λ = mg (3 cos θ − 2)
∂r
Note that when the particle is on top of the sphere, the constraint force is mg. Finally, the
particle leaves the surface of the sphere when the constraint force is equal to zero, hence
we want the angle θ0 for which Frc = 0,

2

θ0 = cos−1
3

Notice that this angle does not depend on the mass of the particle or the radius of the
sphere!

8.7 CONSERVATION THEOREMS AND THE LAGRANGIAN

Besides being used to derive a system’s equations of motion, the Lagrangian can help us
identify conserved quantities in the system. In particular, we will look at conservation of
momentum and energy.

8.7.1 Conservation of Momentum

We already know that in Newtonian mechanics (using Newton’s second law to describe
a system), the total momentum of an isolated system of particles is conserved. We would
expect to find that the total momentum is conserved for the isolated system in the Lagrange
formulation as well. One important feature of isolated systems is that they are translationally
invariant. In other words, if you move each particle in the system by the same amount and
in the same direction, then there is no change in the system, other than the location of each
particle. We will see that translational invariance of the Lagrangian will lead to momentum
conservation in the system.
The connection between translational invariance of the Lagrangian and momentum con-
servation is a result of Noether’s theorem, which states that whenever there is a continuous
symmetry in the Lagrangian, there is a corresponding conservation law. Noether’s theorem
was proved by Emmy Noether (1882–1935) in 1915 and published in 1918 [Noether(1918)].
In this subsection, we will demonstrate the translational invariance of the Lagrangian, and
we will connect the symmetry to the conservation of momentum.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 255 — #271

i i

Lagrangian and Hamiltonian Dynamics 255

Consider a system of N isolated particles of mass mα , where α = 1, . . . , N . The loca-

tion of each particle is denoted by rα . Next we shift all of these particles through the
same displacement dr, which is our “translation,” rα → rα + dr. We need to show that the
Lagrangian will remain constant under the translation. In other words,

dL = L (rα + dr, ṙα ; t) − L (rα , ṙα ; t) = 0 (8.7.1)

Notice that there is no shift in the velocity in (8.7.1). We will discuss why below. Because
the Lagrangian consists of both the kinetic and potential energies, we will need to show
both dT = 0 and dV = 0.
We will first demonstrate dT = 0. Notice that the translation does not change the velocity
of the particles because dr is constant. Because the kinetic energy depends on the velocity,
T = 21 mṙ · ṙ, and the displacements do not change the velocities, the kinetic energy does not
change with the translation and therefore dT = 0.
The translation does affect the position of the particle. The careful reader might then
think that dV 6= 0 because potential energies are often position dependent. Wouldn’t the
gravitational potential energy change with translation? Yes! However, recall that this is an
isolated system, meaning that there are no external forces acting on the system. Further-
more, inter-particle forces are usually dependent on the distances between particles. Because
all of the particles experience the same displacement, the distances between particles will
not change. Hence, dV = 0, and therefore dL = 0.
Now that we have demonstrated the translational invariance of the Lagrangian, we
can make the connection between the symmetry and the conservation of momentum. For
simplicity, we will set N = 1 but the following holds for any number of particles. In addition,
we will use generalized coordinates
P with generalized unit vectors êi , again for simplicity of
notation, r = qi êi and dr = êi dqi . Hence, for a single particle under the translation,
P

∂L ∂T X
pi = = = aij q̇j (8.7.18)
∂ q̇i ∂ q̇i
j

Proof of (8.7.18) is left as Problem 31. Finally, we insert (8.7.18) into (8.7.12):
  
X X 
H=

 aij q̇j  q̇i − L



i j (8.7.19)
=2T − (T − V )




=T + V



In summary, as long as L is invariant to translations in time (i.e., not explicitly dependent

on time), the Hamiltonian is a conserved quantity. In addition, if the transformation from
Cartesian coordinates to generalized coordinates is time-independent, then the Hamiltonian
is equal to the total mechanical energy of the system. This is another demonstration of
Noether’s theorem.
Both the connections between the symmetries (translational invariance and time trans-
lational invariance of L), and the conservation laws (momentum and energy), result from
cyclic coordinates. A wise choice for a set of generalized coordinates would be one such that
there are as many cyclic coordinates as possible. If such a set could be found, then there
will be many conserved quantities which will likely simplify the equations of motion for our
system. Not only are simple equations of motion easier to solve, they are also easier to use
to gain insight into the system of interest.

8.8 HAMILTONIAN DYNAMICS

The Hamiltonian is an important quantity in both classical and quantum physics. In quan-
tum mechanics, the Hamiltonian is an important element of the Schrödinger equation which
is used to find a particle’s wave function. In classical mechanics, the Hamiltonian can be
used to derive a system’s equations of motion. In this section, we will show how to derive
Hamilton’s equations which will provide the equations of motion for a system in terms of
pi and qi .
We begin our derivation with another look at generalized momentum. Recall that the
generalized momentum can be found using the Lagrangian,
∂L
pi = (8.8.1)
∂ q̇i
Because the Lagrangian is a function of both generalized coordinates and generalized veloc-
ities, the momentum can be written as pi = pi (qj , q̇j ; t). The function for the momentum
can in general be inverted such that,

q̇j = q̇j (qk , pk ; t) (8.8.2)

If we combine (8.8.2) with the definition of the Hamiltonian,

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 259 — #275

i i

Lagrangian and Hamiltonian Dynamics 259

X
H= pi q̇i − L (qk , q̇k ; t) (8.8.3)
i
we find that the Hamiltonian is a function of generalized coordinates and generalized
momenta, H = H (qk , pk ; t).
Now we are ready to derive Hamilton’s equations of motion. We will derive Hamilton’s
equations for a system with one degree of freedom and leave the derivation for the s-degree of
freedom case as Problem 34. We begin by calculating the total differential of H = H(q, p; t),
∂H ∂H ∂H
dH = dq + dp + dt (8.8.4)
∂q ∂p ∂t
Next, we calculate dH using (8.8.3),

dH = q̇dp + pdq̇ − dL (8.8.5)

∂L ∂L ∂L
= q̇dp + pdq̇ − dq + dq̇ + dt (8.8.6)
∂q ∂ q̇ ∂t

∂L
= q̇dp + pdq̇ − ṗdq + pdq̇ + dt (8.8.7)
∂t
where we used,
d d ∂L ∂L
ṗ = (p) = =
dt dt ∂ q̇ ∂q
from the Euler-Lagrange equation. Collecting terms, (8.8.7) becomes,
∂L
dH = q̇dp − ṗdq − dt (8.8.8)
∂t
Finally, we equate the coefficients of dq and dp in (8.8.4) and (8.8.8) to get Hamilton’s
equations for a system with one degree of freedom,
∂H ∂H
q̇ = ṗ = − (8.8.9)
∂p ∂q
As you will show in Problem 34, Hamilton’s equations for a system of s degrees of
freedom are:
Hamilton’s Equations of Motion
∂H ∂H
q˙k = ṗk = − k = 1, . . . , s (8.8.10)
∂pk ∂qk

Hamilton’s equations of motion are also sometimes referred to as the canonical equations
of motion, and the description of a particle’s motion by these equations is referred to as
Hamiltonian dynamics. Regardless of the number of degrees of freedom, we also have the
relationship:
∂H ∂L
=− (8.8.11)
∂t ∂t
found by equating the coefficients of dt in (8.8.4) and (8.8.8).
Finally, if we divide (8.8.4) by dt, we find that,
dH ∂H
= (8.8.12)
dt ∂t

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 260 — #276

i i

260 Classical Mechanics: A Computational Approach

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 262 — #278

i i

262 Classical Mechanics: A Computational Approach

θ
Out[ ]=

-1

-2

-3
0 5 10 15 20
time

Example 8.9: Mass on a cylinder

Find Hamilton’s equations of motion for a particle of mass m constrained to move on
the surface of a frictionless cylinder of radius R. The particle is attracted to the origin by
a force F = −krr̂ where k is a positive constant, r̂ is the unit vector pointing from the
origin to the particle’s position, and r is the distance between the particle and the origin.
Solution:
An illustration of the system is shown below.

ρ=R
m

r
z

We will begin by choosing cylindrical coordinates to describe the system. Furthermore,

there is one equation of constraint,

x2 + y 2 = R 2 (8.8.15)
which can also be written as ρ2 = R2 . Therefore, there are two degrees of freedom in this
problem.
Next, we compute the kinetic energy. In cylindrical coordinates, v 2 = ρ̇2 + ρ2 θ̇2 + ż 2 .
However, using our constraint ρ̇ = 0, the kinetic energy of the particle becomes,
1
T = m R2 θ̇2 + ż 2 (8.8.16)

2
The potential energy is best found in Cartesian coordinates first. We are given that

F = −kr = −krr̂ = −k xî + y ĵ + z k̂ (8.8.17)

Therefore the potential energy is

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 263 — #279

i i

Lagrangian and Hamiltonian Dynamics 263

1
V = kr2
2 (8.8.18)
1
= k x2 + y 2 + z 2

2
If we apply the constraint (8.8.15), then the potential energy becomes
1
V = k R2 + z 2

2
We next compute the Lagrangian L = T − V ,
1 1
L = m R2 θ̇2 + ż 2 − k R2 + z 2 (8.8.19)

2 2
From the Lagrangian we see that the canonical momentum conjugate to θ is conserved
because θ is a cyclic coordinate. In order to find the Hamiltonian, we need to compute pθ
and pz ,
∂L ∂L
pθ = = mR2 θ̇ and pz = = mż (8.8.20)
∂ θ̇ ∂ ż
Note that pθ again is an angular momentum, whereas pz is the translational momen-
tum. Next, we use (8.8.3) to find the Hamiltonian,

H =pθ θ̇ + pz ż − L (8.8.21)
1 p 2 p 2 1

pθ pz θ z
=pθ + 2
+ 2
+ 2
(8.8.22)

pz − m R − k R z
mR2 m 2 mR2 m 2
p2
p 2 1
= θ 2 + z + k R2 + z 2 (8.8.23)

2mR 2m 2
Finally, we use the Hamiltonian to compute Hamilton’s equations,
∂H pz 

ż = → ż = 
∂pz m  


∂H pθ 


θ̇ = → θ̇ = 2


∂pθ mR (8.8.24)
∂H
ṗz = − → ṗz = − kz 



∂z 


∂H


ṗθ = − → ṗθ =0


∂θ
Notice that ṗz is equal to the z-component of the external force acting on the system.

8.9 ADDITIONAL EXPLORATIONS INTO THE HAMILTONIAN

You might be wondering why we would ever want to use the Hamilton formulation, especially
given the fact that we typically need to find the Lagrangian in order to get the Hamiltonian.
Why not just use the Lagrange formulation? The reason is that Hamilton’s formulation is
powerful from a theoretical standpoint. For example, in the case of cyclic coordinates, the
equations of motion are extremely simple. Consider Example 8.9 where we studied the mass

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 264 — #280

i i

264 Classical Mechanics: A Computational Approach

constrained to move along the surface of a cylinder. Recall that θ was a cyclic coordinate
in Example 8.9. The equations of motion relevant to θ were,
pθ
θ̇ = and ṗθ = 0.
mR2
Because pθ is constant, it is easy to find θ(t),
Z t
pθ pθ
θ(t) = 2
dt0 = 2
(t − t0 ) .
t0 mR mR
When the canonical momentum is constant, the equation of motion can be reduced to
quadrature (as above). If a coordinate system can be found such that all coordinates are
cyclic, then the equations of motion are trivial. In fact, it is possible to find such coordinate
systems. Those transformations were developed by Carl Gustov Jacob Jacobi (1804–1851)
and the work is referred to as the Hamilton-Jacobi theory. Hamilton-Jacobi theory is beyond
the scope of this book, but the motivated reader should consult [Goldstein et al. (2001)].
Furthermore, simply knowing the Hamiltonian alone can give us insight into all possible
behaviors of the system. This can be done by using the Hamiltonian to plot the system’s
phase space. Unlike in Chapter 6 where we plotted x versus v, we will plot p versus q
to create the phase space. In particular, we will plot the canonical momentum versus its
conjugate generalized coordinate. In general, the dimension of the phase space is 2s, where
s is the system’s number of degrees of freedom. Notice that H = H (qk , pk ) and will define
surfaces in the phase space. To understand the value of such plots, we will revisit the simple
pendulum, yet again.

Example 8.10: The phase space of the simple pendulum

Consider the simple plane pendulum described in Example 8.8. Using the Hamiltonian
found in Example 8.8, plot the phase space of the simple plane pendulum and interpret
the result.
Solution:
The Hamiltonian for the simple plane pendulum was found in Example 8.8,

p2
H(q, p) = − mg` cos q (8.9.1)
2m`2
where we have defined p = pθ and q = θ for simplicity of notation. The Mathematica code
to create a contour plot with m = 1.0 kg, ` = 1.0 m, and g = 9.8 m/s2 is shown below.
l = 1.0;
m = 1.0;
g = 9.8;
2
p
H[q , p ]:= 2∗m∗l 2 − m ∗ g ∗ l ∗ Cos[q];

ContourPlot[H[q, p], {q, −π, π}, {p, −10, 10}, ContourShading → None,
Contours → {−m ∗ g ∗ l ∗ Cos[0.5], −m ∗ g ∗ l ∗ Cos[2], m ∗ g ∗ l, 2 ∗ m ∗ g ∗ l, 3 ∗ m ∗ g ∗ l},
ContourStyle → {Black, Black, {Red, Dashed}, Black, Black}]
The resulting plot is shown as follows.

8.10 CHAPTER SUMMARY

The Lagrange formulation uses Hamilton’s principle, which says that the path a particle
will follow makes the action integral
Z t2
S= L (qi q̇i ; t) dt
t1

stationary. The functions qj (t) that make the action integral stationary can be found using
the Euler-Lagrange equation,

d ∂L ∂L
− =0
dt ∂ q̇i ∂qi
The Euler-Lagrange equation produces s second-order ODEs, where s is the number
of degrees of freedom in the system, which can be solved for qi (t).
In the case, where there are m holonomic constraints on the system, we can express
the constraints as m constraint equations fj and the Euler-Lagrange equation becomes
m
∂L d ∂L X ∂fj
− =− λj (t)
∂qi dt ∂ q˙i ∂qi
j=1

In addition to using the Lagrangian, the Hamiltonian can also be used to derive a
system’s equations of motion. The Hamiltonian is defined as,
s
X
H= pi q̇i − L.
i=1

Hamilton’s equations of motion are then

∂H ∂H
q˙i = and ṗi = − .
∂pi ∂qi

8.11 END-OF-CHAPTER PROBLEMS

Section 8.1: An Introduction to the Lagrangian
1. Consider a free particle of mass m moving in three dimensions. The particle is free of
any external forces. Find the Lagrangian and the equations of motion for the particle.
Identify any conserved quantities.
2. Consider a particle of mass m moving in three dimensions near the surface of the
Earth. Find the Lagrangian and the equations of motion for the particle. Identify any
conserved quantities.
3. Consider a particle of mass m = 1.0 kg moving in one dimension under the influence of
the force F = −kx3 . Find the Lagrangian and the equations of motion. Use a computer
to solve for x(t) in the case where k = 0.1 N/m3 , x(0) = 1.0 m, and ẋ(0) = 0 m/s.
Section 8.2: Generalized Coordinates and Degrees of Freedom
4. What is the number of degrees of freedom for a particle undergoing projectile motion
without air resistance? Are there any constraints on the motion? Assume that the

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 267 — #283

i i

Lagrangian and Hamiltonian Dynamics 267

projectile travels a short enough distance that the Coriolis force (see Chapter 10) is
negligible.
5. Consider a disk of radius R rolling down an inclined plane. What quantities are needed
to describe the state of the system? Are there constraints among those quantities?
Hint: The disk is rolling, without sliding.
6. A very small ball of mass m and radius a is constrained to roll along the surface of a
sphere of R a. The ball is constrained to move along geodesics on the sphere. How
many degrees of freedom does the ball have?
Section 8.3: Hamilton’s Principle
7. Consider a single-particle system of s degrees of freedom. The particle experiences a
force which has an associated potential energy V (qi ). Show that the functions qi (t)
Rt
for i = 1, . . . , s that make the action integral S = t12 L (qi , q̇i ) dt stationary also obey
Newton’s laws. In other words, extend the work done in Section 8.3 to a single-particle
system with s-degrees of freedom.
8. Repeat Problem 7 for an N -particle system.
Section 8.4 and 8.6: Examples of Lagrangian Dynamics
Note that the problems in this section combine problems for sections 8.4 and 8.6.
9. Fill in the necessary steps to calculate (8.4.19) and (8.4.20) in Example 8.3.
10. In Example 8.5, find the acceleration of the inclined plane and the particle in the limit
of M → 0. Discuss your results.
11. A simple plane pendulum made of a mass m and massless rod of length ` is attached
to a support that accelerates upward with an acceleration a. Find the equations of
motion and the period of small oscillations.
12. Find the equation of motion for a simple plane pendulum made of a mass m and a
massless rod of length ` which is attached to a support that oscillates horizontally,
such that xs = a cos (ωt), where xs is the location of the support on the horizontal
axis. The support has no vertical motion.
13. A simple plane pendulum made of a mass m and massless rod of length ` is attached to
a support that oscillates vertically, such that ys = a cos (ωt) , where ys is the location
of the support on the vertical axis. The support has no horizontal motion. Find the
equation of motion and period of small oscillations.
14. A double pendulum consists of two pendulums attached in series as shown in Figure
8.5 that follows. Find the equation of motion for the double pendulum.

15. Consider a simple pendulum of mass m and initial length `. After the pendulum
begins to swing, its length is shortened at a rate `˙ = −α, where α is a constant.
Use the Lagrangian to compute the equations of motion of the system. Is the energy
conserved in this system?
16. A small particle with mass m slides down a smooth circular wedge with mass M and
radius of curvature R as shown in Figure 8.6. The wedge is free to move horizontally
along a frictionless surface. Find the equation of motion for each object.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 268 — #284

i i

268 Classical Mechanics: A Computational Approach

ℓ1
θ1

ℓ2
θ

Figure 8.5: Problem 8.14: A double pendulum.

Figure 8.6: Problem 8.16: The circular wedge and mass.

17. Find the equations of motion for a single particle in three dimensions experiencing a
central force field. Use spherical coordinates.
18. A particle of mass m is constrained to move along a frictionless massless circular hoop
of radius a. The hoop rotates about its vertical diameter with a constant angular
speed ω. Find the equilibrium positions of the particle.
19. A massless spring of equilibrium length ` connects two masses, m1 and m2 . The masses
are free to rotate and oscillate on a horizontal frictionless surface. Find the equations
of motion for the particles.
20. Consider a mass attached to a massless spring with a spring constant k and equilib-
rium length ` that is allowed to swing on a frictionless plane. This is the so-called
elastic pendulum. Find the equation of motion for an elastic pendulum of mass m and
equilibrium length `.
21. A smooth horizontal circular wire of radius R rotates about a point on its perimeter. A
particle of mass m is constrained to move along the wire as shown in Figure 8.7. Find
the equation of motion for the particle. About what point does the particle oscillate?

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 269 — #285

i i

Lagrangian and Hamiltonian Dynamics 269

ω
x

Figure 8.7: Problem 8.21: A bead restricted to move on a rotating circular.

22. A rope of mass M and length L is stretched out along a frictionless table such that
a length z0 is hanging over the edge of the table. Find the equation of motion of the
rope if it is released from rest. Show that the rope requires a time
s
L
t= cosh−1 (L/z0 )
g
to completely slide off the table. You may remember this problem from Chapter 2,
where we did it using Newton’s laws.
23. A particle of mass m moves in a plane under the force F = −αr2 where F points
towards the origin. Find the equations of motion for the particle. Are there any con-
served quantities? Is so, what are they?
24. A sphere of radius a is constrained to roll without slipping on the inside of a cylinder
with a radius R. Find the equations of motion of the sphere and calculate the period
of small oscillations.
25. A particle of mass m is constrained to move on a frictionless wire is bent into the
shape of a helix with r = R and z = αθ, where R and α are positive constants. The
particle is under the influence of gravity, which points in the −z direction. Find the
equation of motion for the particle. What is the particle’s acceleration when R → 0?
Section 8.5: Numerical Solutions to ODE’s Using the Fourth-Order Runge-
Kutta Method
26. Repeat Example 8.6 for different values of the parameters R, ω, and `. Discuss the
different types of motions that result from changing the parameters.
27. Write a program that uses RK4 to numerically solve the equation of motion for the
simple harmonic oscillator, ẍ + ω02 x = 0 where ω0 = 1.0 rad/s with initial conditions
x(0) = 1.0 m and ẋ(0) = 0.5 m/s.
28. Using the same equation and values for the SHO in Problem 27, calcu-
late the error in x(t) at t = 3.0 s for the RK4 and Euler method by using
dt = 0.001, 0.01, 0.1, 0.2, and 1.0. We will define the error as,
exact value − result

exact value

where “exact value” is calculated from the closed-form solution of the SHO and
“result” is the result obtained by either the RK4 or Euler method.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 270 — #286

i i

270 Classical Mechanics: A Computational Approach

29. Use an RK4 program to solve the damped driven harmonic oscillator,

ẍ + 2β ẋ + ω02 x = A cos (ωt) ,

for ω = 1.0 rad/s, β = 2.0 rad/s, ω0 = 0.5 rad/s, and A = 0.25 m/s2 with initial con-
ditions x(0) = 1.0 m and ẋ(0) = 0 m/s.

30. The Lorenz equations,

ẋ =σ(y − x)
ẏ =rx − y − xz
ż =xy − bz

are an important set of equations in the field of nonlinear dynamics. Using an RK4
program, solve for x(t), y(t), and z(t) for parameter values: σ = 10, r = 28, and b = 8/3
and initial conditions x(0) = 0.1, y(0) = z(0) = 0. Plot each solution individually and
then make a three-dimensional plot which contains x(t) vs. y(t) vs. z(t).
Section 8.7: Conservation Theorems and the Lagrangian
31. Prove (8.7.18):
∂L ∂T X
pi = = = aij q̇j
∂ q̇i ∂ q̇i
j

32. A particle of mass m moves within a force field that has an associated potential energy
V = −k/r where k is a positive constant. What are the conserved quantities for this
particle?
33. Consider a system with s degrees of freedom and Lagrangian L(qk , q̇k ), where k =
1, . . . , s. Now, let F = F(qk ) be any function of the generalized coordinates, qk . We
can construct an additional Lagrangian,

L0 = L + dF/dt

Show that L and L0 give the same equations of motion.

Section 8.8: Hamiltonian dynamics
34. Derive Hamilton’s equations of motion for a system with s degrees of freedom.
35. A spherical pendulum is a simple pendulum that is not restricted to move in a plane.
Find the equations of motion for a spherical pendulum that has a fixed support and
consists of a massless rod of length ` attached to a mass m at the end of the rod.
36. Find Hamilton’s equations for the double-well potential, V = − 21 kx2 + 14 x4 , where k
and are positive constants.
37. A particle of mass m moves within a force field that has an associated potential energy
V = −k/r, where k is a positive constant. Find Hamilton’s equations for the particle.
38. Find Hamilton’s equations for the elastic pendulum described in Problem 20.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 271 — #287

i i

Lagrangian and Hamiltonian Dynamics 271

39. Find Hamilton’s equations for a particle of mass m confined to the surface of a cone,
which in cylindrical coordinates is described by r = az, inside a uniform gravitational
field. Show that for a given energy, there are maximum and minimum values of z for
which the particle is confined.
40. Consider the force
k
F = − e−λt (8.11.1)
x
acting on a particle of mass m moving in one dimension. Assume that the constants
k and λ are bothR positive. We can create a “pseudo potential energy,” using the rela-
tionship, V = − F dx. Using the pseudo potential energy find Hamilton’s equations
of motion for the particle. Compute ∂H/∂t and ∂L/∂t and show that they satisfy
(8.8.11):
∂H ∂L
=−
∂t ∂t
41. Recall the cycloid from the brachistochrone problem,

Central Forces and Planetary

Motion

In this chapter, we will first look at the general concept of central forces and their importance
in describing physical phenomena. We will begin with a general discussion of central forces.
Next we will examine the problem of two objects interacting via a central force, this is the
so called two-body problem. Our analysis of the two-body problem will lead to Kepler’s
laws of planetary motion. We will then examine the specific case of a planet orbiting a
star. Finally, we will study the three-body problem where three objects interact via central
forces. Solving for the motion of one of the objects in the three-body problem will require
us to use the computational skills developed so far in this book.

9.1 CENTRAL FORCES

In this section, we will outline some general properties of central forces.
A central force is a force on a particle that is directed along a line that joins the particle
and the force center (such as the origin). When two particles interact via a central force,
then the force is always directed along the line that joins the centers of the particles, as
shown in Figure 9.1, where the central force of interaction F is directed along a line that
joins the two particles. As an example, F may be Newton’s law of universal gravitation
(9.1.2), in which case F21 is the force of gravitational attraction experienced by m1 due to
the presence of m2 . Parallel to the force vectors in Figure 9.1 is a dashed line of length r,
the distance between the centers of m1 and m2 , and the unit vector r̂, which lies along the
the line joining the two masses. Note that the vector r̂ in Figure 9.1 should lay on top of
the vector F21 , but we placed it beside F21 for clarity in the image.
Central forces depend only on the distance r between the two interacting bodies. There-
fore, spherical coordinates provide a simple way of describing central forces, especially
because they do not depend on θ and φ. Hence, we can represent a central force gener-
ally in the form:

General Formula for Central Forces

F = f (r)r̂ (9.1.1)

where f (r) is a scalar function of the magnitude r of the position vector r, and r̂ is a unit
vector along the position vector as shown in Figure 9.1.

273

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 274 — #290

i i

274 Classical Mechanics: A Computational Approach

m1 r

F21
r

F12
m2
Figure 9.1: Two masses m1 and m2 interacting via a central force F.

The best-known examples of central forces in physics are Newton’s law of universal
gravitation FG (r) between two masses m1 and m2 , and the electrostatic Coulomb force
FC (r) between two electric changes q1 and q2 . The mathematical description of these two
forces is very similar:
Gm1 m2
FG (r) = − r̂ (9.1.2)
r2
1 q1 q2
FC (r) = r̂ (9.1.3)
4π0 r2

where G = 6.67×10−11 m3 kg−1 s−2 is the universal gravitational constant and 0 = 8.85 ×
10−12 m−3 kg−1 s4 A2 is the corresponding constant for electrical forces, called the permit-
tivity of free space.

9.1.1 Central Forces and the Conservation of Energy

A very important property of central forces is that they are conservative. This can be
easily seen by writing the general form of the curl of a conservative force F in spherical
coordinates:

1 1 1 ∂Fr

∂ ∂Fθ ∂
∇×F = Fφ sin θ − r̂ +

− rFφ θ̂
r sin θ ∂θ ∂φ r sin θ ∂φ ∂r
1 ∂ (rFθ ) ∂Fr

+ − φ̂ (9.1.4)
r ∂r ∂θ

In the case of central forces we have Fθ = Fφ = 0, ∂Fr

∂φ =0 , ∂Fr
∂θ = 0 so that (9.1.4) gives:

∇×F = 0 (9.1.5)
and therefore, F is a conservative force.
As we saw in Chapter 5, conservative forces can be associated with potential energy
functions V (r) = − F · dr. Hence, central forces have a potential energy associated with
R

them. The corresponding scalar gravitational potential energy VG and electrostatic potential

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 275 — #291

i i

Central Forces and Planetary Motion 275

energy VC are:
Gm1 m2
VG (r) = − (9.1.6)
r
1 q1 q2
VC (r) = (9.1.7)
4π0 r
Another well-known example of a central force is the intermolecular long-range van der
Waals force, which is of importance in several branches of science. In one of its simplest
forms, the van der Waals force between a pair of neutral atoms or molecules can be approx-
imated by the Lennard-Jones potential (also termed the L-J potential or 6-12 potential).
The mathematical form of the 6-12 potential VLJ (r) is:
hc c2 i
1
VLJ (r) = 12 − 6 (9.1.8)
r r
where c1 and c2 are constants with the appropriate SI units.
In the field of particle and atomic physics, the Yukawa short-range nuclear force is of
great importance, and the corresponding Yukawa potential VY (r) is represented mathemat-
ically by:
e−c5 r
VY (r) = c4 (9.1.9)
r
where c4 , and c5 are constants with the appropriate SI units.

9.1.2 Central Forces and the Conservation of Angular Momentum

Next we will show that central forces conserve angular momentum. Consider a system of
N particles, where mi , ri , and pi are the mass, position, and momentum of particle i with
i = 1, . . . , N . We know from Chapter 4 that the total torque N acting on a system of particles
is
N
X N
X
N= ri × ṗi = ri × Fi (9.1.10)
i=1 i=1

where Fi is the force acting on the particle with mass mi . In Chapter 4, we further simplified
this equation to show,
N
X N X
X
N= ri × Fext
i + (ri − rj ) × Fij (9.1.11)
i=1 i=1 j>i

where Fext is the sum of the external forces acting on mi and Fij is the force of interaction
between mi and mj . If we assume an isolated system then Fext = 0 and the first term
in (9.1.11) is equal to zero. Using a coordinate system similar to that in Figure 9.1, we
can write ri − rj = rij r̂ij , where r̂ij is a unit vector that lies along a line joining mi and
mj . Furthermore, if the force of interaction between the particles is a central force, then
Fij = Fij (r)r̂ij and the second term in (9.1.11) is: (rij r̂ij ) × (Fij r̂ij ) = 0. Therefore the net
torque acting on the system is zero, and angular momentum is conserved.
In the next section, we will examine the so-called two-body problem in detail. This
discussion will lead us to Kepler’s famous laws of planetary motion.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 276 — #292

i i

276 Classical Mechanics: A Computational Approach

9.2 THE TWO-BODY PROBLEM

In this section, we will consider the two-body Keplerian problem, where two masses are
interacting via a central force. Such two-body systems are very common in astronomy.
We begin with the system shown in Figure 9.2, where the masses m1 and m2 are at
locations r1 and r2 with respect to a shared origin, O. The center of mass of the two
particles is located at R relative to the origin as shown in Figure 9.2 and is denoted by a
white dot. The two masses interact with a central force F = F (r)r̂, not shown in the figure,
but F lies along the vector r similar to Figure 9.1.
m1

r1 r=r1-r2

R
m2

r2
O

Figure 9.2: The position vectors r1 and r2 and relative distance r = r1 − r2 of the two masses
interacting with a central force F (r)r̂. The white dot at the location R shows the location
of the system’s center of mass.

If r1 and r2 are the position vectors of the two masses m1 and m2 , respectively, then the
vector distance between the two masses is

r = r1 − r2 (9.2.1)

Newton’s second law F = ma for each mass can be written as:

m1 r̈1 = −F (r)r̂ (9.2.2)

m2 r̈2 = F (r)r̂ (9.2.3)

where r̂ is the unit vector in the direction of r. Note that we are using Newton’s third law
to write the above equations for an attractive central force. These two equations give:

1 1

r̈1 −r̈2 = − + F (r)r̂ (9.2.4)
m1 m2

m1 m2
(r̈1 −r̈2 ) = −F (r)r̂ (9.2.5)
m1 + m2
which can be rewritten as
Newton’s Second Law for the Reduced Mass µ

µr̈ = −F (r)r̂ (9.2.6)

m1 m2
µ= (9.2.7)
m1 + m2

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 277 — #293

i i

Central Forces and Planetary Motion 277

This equation tells us that the two-body system behaves just like a single mass µ with a
position vector r, moving under the influence of the central force F = F (r)r̂. This single
mass µ is called the reduced mass of the two-body system and represents an effective inertial
mass appearing in the two-body problem. Using the reduced mass µ allows us to study the
two-body problem as if it were a one-body problem. Note that the minus sign appears in
(9.2.6) because we assumed an attractive force. A repulsive force would give similar results
but without the minus sign in (9.2.6).
What is the reduced mass? The definition in (9.2.7) can be difficult to interpret. Consider
the Earth-Sun system. In this case the mass of the Sun mS is much greater than the mass
of the Earth, mE . If we use m1 = mS and m2 = mE in (9.2.7), we find that µ ≈ mE . Hence,
when we study the motion of the Earth-Sun system, we are essentially studying the motion
of the Earth. As the masses become more comparable, we cannot associate the reduced
mass with either one of the masses, as we will demonstrate later in Example 9.1.
When working with a system of multiple particles, we know that the system’s center of
mass can be useful in describing the behavior of the system. As we discussed in Chapter 4,
the position of the center of mass of the two-body system is defined by:
m1 r1 + m2 r2
R= (9.2.8)
m1 + m2
Because the translational motion of the system as a whole is not of interest (we are
focused on the particles’ orbits with respect to each other), we can set R = 0, as shown
in Figure 9.3a. In other words, we can place the center of mass at the origin. This new
coordinate system is different from that shown in Figure 9.2. As we will see, this new
coordinate system with the center of mass at the origin, will be very useful in describing
the orbits.

Figure 9.3: (a) The binary system can be best described using a system of coordinates axes
with the origin located at the center of mass of the system. (b) The motion can also be
described by Newton’s law (9.2.6), as the motion of the reduced mass µ, with position vector
r = r1 − r2 , moving with the relative velocity of the two masses v = ṙ1 − ṙ2 .

Using R = 0, we have,
m1 r1 + m2 r2 = 0 (9.2.9)
By combining (9.2.9) with the definition r = r1 − r2 for the position vectors r1 and r2 , we
obtain:
m2
r1 = r (9.2.10)
m1 + m2

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 278 — #294

i i

278 Classical Mechanics: A Computational Approach

m1
r2 = − r (9.2.11)
m1 + m2
The equations (9.2.10) and (9.2.11) allow us to relate the particles’ position to the
separation vector r. In addition, we can write the Lagrangian for the system as:
1 1
L = m1 ṙ21 + m2 ṙ22 − V (r) (9.2.12)
2 2
where V (r) = − F (r)dr is the potential energy associated with the central force F. Sub-
R

stituting the values of (9.2.10) and (9.2.11) in the Lagrangian (9.2.12), we find after some
simple algebra that the Lagrangian of the two-body system is:

Lagrangian of the Two-Body System

1
L = µ | ṙ |2 − V (r) (9.2.13)
2

Similar to Newton’s second law, the Lagrangian shows that we can reduce the two-body
problem to the motion of a single object with a mass equal to the reduced mass µ, moving
with the relative velocity of the two masses v = ṙ1 − ṙ2 .
Similarly, the total angular momentum of the two-body problem is found from:

dr1 dr2
L = m1 r1 × + m2 r2 × (9.2.14)
dt dt

Substituting (9.2.10) and (9.2.11) into (9.2.14) we can obtain (see Problem 10 at the end
of this chapter):

Angular Momentum of the Two-Body System

` = µ (r × ṙ) (9.2.15)

where we have replaced L (for two bodies) with ` (for one body). This equation tells us that
the total angular momentum of the two-body problem is equal to the angular momentum of
the reduced mass µ, located at the point r = r1 −r2 and moving with velocity v = ṙ = ṙ1 − ṙ2 .
Again, the result is that we can reduce the two-body problem to studying the motion of a
single particle of mass µ.
Furthermore, taking the time derivative of (9.2.15) results in `˙ = 0, hence just as before,
the angular momentum is conserved. One of the physical consequences of the conservation
of the angular momentum vector ` is that the motion of the mass µ has to remain on
the same plane (which also contains the force center), in order to keep the direction of `
constant at all times. This is shown for the special case of a circular orbit in Figure 9.4.
Because the motion of the particle is restricted to a plane, polar coordinates are a useful
coordinate system when dealing with central force problems. The constant nature of ` helps
explain why all of the orbits of the planets in our solar system are essentially co-planar and
continue to stay that way.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 279 — #295

i i

Central Forces and Planetary Motion 279

=r×p

θ r
p
μ

Figure 9.4: Relationship of the angular momentum vector `, position vector r and linear
momentum vector p = µv.

Next, we will compute the magnitude of the angular momentum in polar coordinates.
Recalling that in polar coordinates, ṙ = ṙr̂ + rθ̇θ̂θ , (9.2.15) yields:
Magnitude of the Angular Momentum of the Two-Body Problem

` = µr2 θ̇ (9.2.16)

We can also compute the total energy of the two-body system using E = T + V ,

1
E = µ ṙ2 + r2 θ̇2 + V (r) (9.2.17)

2
2
1 2 1 2

`
= µṙ + µr + V (r) (9.2.18)
2 2 µr2
which can be simplified to
Total Energy of the Two-Body System
2
`2 1 dr
E= + µ + V (r) (9.2.19)
2µr 2 2 dt

Again, the physical interpretation of this equation, and the others before it, is that the two-
body system can be replaced by a single reduced mass µ at a distance r =| r1 − r2 |from the
origin (i.e., the force center and the system’s center of mass), and moving with the velocity
v = ṙ = ṙ1 − ṙ2 . This is shown schematically in Figure 9.3(b).
While the reduced mass is a useful tool for performing the calculations above, what we
really want is to be able to calculate the motions of m1 and m2 . In the next section, we will
derive the equations of motion for the two-body problem.

9.3 EQUATIONS OF MOTION FOR THE TWO-BODY PROBLEM

When finding equations of motion, we often want to find the position of the particle as a
function of time. In principle, this can be done by solving (9.2.19) for ṙ:
s
dr 2 `2
ṙ = =± (E − V (r)) − (9.3.1)
dt µ 2µr2

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 280 — #296

i i

280 Classical Mechanics: A Computational Approach

which, in principle, can be integrated to get t(r) and then inverted to get r(t), once we have
specified V (r). However, r(θ) would be a more useful equation of motion because it would
give us a mathematical formula for the observed motion. We can substitute,

dr dr dθ dr `2
= = (9.3.2)
dt dθ dt dθ µr2
into (9.3.1) to obtain
Z
`dr
θ(r) = ± r (9.3.3)
`2
r2 2µ E − V (r) − 2µr 2

In principle, after specifying V (r), we could perform the integral to get a mathematical
form for θ(r), and then invert it to get r(θ). However, in general, the integral in (9.3.3) is
difficult to do, and the inversion is also not easy.
Computers can provide a means of obtaining numerical solutions, in which case it is often
easier working directly with Newton’s second law. In the next two examples, we show how
numerical solutions for an object’s orbit can be obtained by numerically solving Newton’s
second law.
Example 9.1: The orbits in a two-body gravitational system
Integrate Newton’s second law using Newton’s law of universal gravitation to represent
for the interaction force between two masses m1 and m2 , in order to obtain the orbits of
a binary star system with masses m1 = 1, m2 = 2 (in arbitrary units). In order to simplify
the presentation of the numerical results, use arbitrary units with G = 1. At time t = 0 the
reduced mass is located at x = 1, y = 0 and has a velocity with an x-component of vx = 0
and a y-component of vy = 1.
Solution:
As discussed in this section, Newton’s second law for the two masses becomes a single
equation for the reduced mass:
Gm1 m2
µr̈ = − r̂
r2
where we have used Newton’s law of universal gravitation as the central force. Now we
need to rewrite the above formula in terms of Cartesian coordinates. The unit vector is
given in Cartesian coordinates (x, y) by:

r xî + y ĵ
r̂ = =p
r x2 + y 2

By combining the previous two equations and using r2 = x2 + y 2 ,

d2 r Gm1 m2 xî + y ĵ Gm1 m2

µ = − = − x î + y ĵ
dt2 x 2 + y2 x2 + y 2 3 /2
p
(x 2 + y 2 )

This is equivalent to the following two coupled differential equations:

d2 x Gm1 m2 d2 y Gm1 m2
µ =− x µ =− y (9.3.4)
dt2 (x + y 2 )
2 3 /2 dt2 (x + y 2 )
2 3 /2

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 281 — #297

i i

Central Forces and Planetary Motion 281

In the Mathematica code below, the command NDSolve is used to solve the system of
differential equations (9.3.4), and the Table command is used to construct a list of the
pairs of position coordinates r = (x, y) from t = 0 to t = 10, in time steps of dt = 0.01. The
coordinate pairs are stored in the variable xyList. We see that the reduced mass µ follows
an elliptical orbit. As we will see later, this is an example of Kepler’s First Law.
We also plotted the positions of the two masses. The positions can be calculated from
r by using (9.2.10) and (9.2.11); this is done easily by multiplying the list xyList with the
factors m1 /(m1 + m2 ) and −m2 /(m1 + m2 ), to get r1 and r2 , respectively. Finally a graph
of the orbits of the two masses is produced by using the ListPlot command. Notice that
each mass follows an elliptical path and the center of mass (at the origin) is one of the
foci of each ellipse.

m1 = 1; m2 = 2; x0 = 1; y0 = 0; vx0 = 0; vy0 = 1;
µ = m1 ∗ m2/(m1 + m2);

sol = NDSolve[{x00 [t] == −(m1 + m2) ∗ x[t]/((x[t]∧ 2 + y[t]∧ 2)∧ (3/2)),

y 00 [t] == −(m1 + m2) ∗ y[t]/((x[t]∧ 2 + y[t]∧ 2)∧ (3/2)), x[0] == x0, y[0] == y0, x0 [0] ==
vx0, y 0 [0] == vy0}, {x, y}, {t, 0, 10}];

x1 = x/.sol[[1]];
y1 = y/.sol[[1]];
xyList = Table[{x1[t], y1[t]}, {t, 0, 10, .01}];

a = (Max[xyList[[All, 1]]] − Min[xyList[[All, 1]]])/2;

b = (Max[xyList[[All, 2]]] − Min[xyList[[All, 2]]])/2;
e = Sqrt[1 − b∧ 2/a∧ 2];

centerEllipse = Max[xyList[[All, 1]]] − a;

centerofMass = Max[xyList[[All, 1]]] − a − a ∗ e;

Print[“location of center of ellipse=”, centerEllipse]

Print[“location of CM=”, centerofMass]

or
d2 u µ
+ u = − 2 2 f u−1 (9.3.14)

dθ 2 ` u
Finally this equation can be written in terms of r by substituting u = 1/r:

Two-Body Equation of Motion in Terms of r(θ) and f (r)

d2 1 1 µr2

2
+ = − 2 f (r) (9.3.15)
dθ r r `

This equation is useful when we know the orbit in the form r(θ), and we want to evaluate
the central force f (r) that creates this orbit.

Example 9.2: Finding the central force when we know the orbit
The orbit r(θ) of a mass m moving inside a central force field, is given by the expression:

r(θ) = kθ

(a) Find the central force f (r) creating this orbit.

(b) Find the orbit θ(t) with the initial condition θ(0) = 0.
Solution:
(a) We use (9.3.15) to evaluate the force:

`2 1 1
2
d
f (r) = − 2 +
µr dθ2 r r

We now have
d2 1 d2 1 2

= 2 = 3
dθ2 r dθ kθ kθ
and the f (r) equation above gives:

`2 2k 2 1 `2 2k 2 1

f (r) = − 2 + =− + 3
µr r3 r µ r5 r

So the force in this system is a linear combination of two terms 1/r5 and 1/r3. .
(b) We can find the orbit θ(t) by using the conservation of angular momentum:

dθ
µr2 θ̇ = µr2 =`
dt
and substituting r = kθ:
dθ µk 2 d (θ)3
µk 2 θ2 = =`
dt 3 dt
This can be integrated to yield:
3t`
θ3 (t) = +C
µk 2

B = 0. Let’s choose θ = 0 as the angular position of µ when it makes its closest approach
rmin to the force center (located at the center of mass). At this point, ṙ = 0 (because this
is a turning point in the orbit) and due to (9.3.11), u0 = 0. Hence, we are setting an initial
condition for u to be u(θ = 0) = 0. Because w0 = u0 then w0 (θ = 0) = 0 and we can choose
B = 0. Hence we have,

µk
u = A cos θ + (9.4.4)
`2
If we set e = `2 A/µk and α = `2 /µk, we get
1 1
= (1 + e cos θ) (9.4.5)
r α
which is the equation of a conic section in polar coordinates with one focus at the origin,
e is called the eccentricity and α is called the semi-latus rectum. Later in this chapter, we
will discuss the equation for conic sections in polar coordinates in more detail.
When studying planetary motion, Johannes Kepler (1571–1630) developed three laws of
planetary motion. Kepler’s first law stated that planets orbit in an ellipse and that the Sun
is at one of the foci of the ellipse. An ellipse, of course, is one example of a conic section.
However, starting from Newton’s second law, we have found a much more general result.
We have found that the motion of the reduced mass is of the form of a conic section: circle,
ellipse, parabola, or hyperbola. What conditions dictate which conic section the reduced
mass follows? In the next section, we will begin to answer this question.
However, before we move on, there is one piece of (9.4.5) that is still unsatisfactory, the
eccentricity e is in terms of the constant of integration A. It would be better if we could
rewrite e in terms of physical parameters, similar to α. We begin by using our choice of
coordinates where ṙ = 0 at r = rmin and rmin occurs at θ = 0, hence (9.4.5) gives
α
rmin = (9.4.6)
1+e
and the total energy at rmin is

`2 k
E= − (9.4.7)
2µrmin
2 rmin

Solving the two equations above for e we get

s
2E`2
e= 1+ (9.4.8)
µk 2

9.5 ORBITS IN A CENTRAL FORCE FIELD

In this section, we look at the general problem of determining the orbits of a particle with
a reduced mass µ moving in a central force field F(r) = −k/r2 r̂. We already know that
the orbits are conic sections. However, we would like to know, for example, under what
conditions is a particle’s orbit an ellipse, as opposed to a parabola?
We start by considering the equation for the total energy:

1 l2
E = T + V = µṙ2 + + V (r) (9.5.1)
2 2µr2

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 286 — #302

i i

286 Classical Mechanics: A Computational Approach

where we have returned to the generic potential energy. Note that for an inverse square law,
V = −k/r. The first term in (9.5.1) can be interpreted as a kinetic energy term, because it
is of the form of one half times a mass and the square of the radial velocity. It is convenient
to think of the second two terms as potential energy terms.
We notice that the term

l2 1
VC ≡ = µr2 θ̇2 (9.5.2)
2µr2 2
can be interpreted as a form of energy called the centrifugal potential energy VC , and the
corresponding force will be the centrifugal force FC :

∂VC l2
FC = − = 3 (9.5.3)
∂r µr

Note that FC = FC r̂ points radially outward from the center of mass (the force center for
the two-body problem). We can then write the effective potential energy Veff (r) as the sum
of two terms, the centrifugal potential VC and the potential energy V (r) corresponding to
the central force.
The Effective Potential in a Central Force Field

l2
Veff (r) = + V (r) = VC + V (r) (9.5.4)
2µr2

Figure 9.5 shows a plot of the effective potential Veff (r) as a function of the distance r, and for
an attractive potential V = −k1 /r where k1 = 2.5 in SI-units. The dashed lines in this graph
indicates the repulsive positive centrifugal potential VC = 1/r2 , and the attractive potential
V = −k1 /r. The thick solid line in this figure shows the sum of these two potentials, which
represents the effective potential Veff (r). When the distance r → 0 the effective potential
goes to infinity, while as r → ∞ the effective potential goes to zero. The horizontal lines in
this figure indicate two possible energies of the reduced mass, E0 = −1.55 J and E1 = −1.3 J.
Note that the value of these energies are arbitrarily chosen for illustrative purposes below.
We can use the methods of Chapter 5 in order to qualitatively describe the motion of
the reduced mass µ as a function of its total energy E. Recall, that we can think of the
reduced mass as a particle that is rolling along the track made by the shape of the Veff
graph. When E = E0 , the particle is fixed at the minimum of Veff . This means that the
value of r does not change. The result is that the reduced mass has a circular orbit with
r = r0 . When E = E1 < 0 (or more generally for E0 < E < 0), the particle rolls back and
forth between r = r1 and r = r2 . We know from Chapter 5 that such a “rolling back and
forth” means oscillatory motion. The reduced mass gets no closer to the force center than
r = r1 , and no farther than r = r2 . Hence the resulting motion is an ellipse with rmin = r1
and rmax = r2 . Note that we are using rmin and rmax as the closest and farthest distance
between the reduced mass and the force center, respectively. In general, we can see from
Figure 9.5 that when E < 0, the motion of the reduced mass is bound to periodic orbits.
Furthermore, from Figure 9.5 we can see that when E ≥ 0, the motion is unbound. The
reduced mass comes in from infinity and gets as close to the force center as r3 , which is
not shown in Figure 9.5. However, one can find r3 by solving E = Veff (r3 ), where E is the
energy of the reduced mass. The solution can be found either analytically or numerically.
The unbound orbits are parabolic or hyperbolic, depending on the value of E.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 287 — #303

i i

Central Forces and Planetary Motion 287

eccentricity energy orbit shape

e=0 E = Veff,min circle
0<e<1 Veff,min < E < 0 ellipse
e=1 E=0 parabola
e>1 E>0 hyperbola

Table 9.1: The energies and eccentricites associated with each type of orbit for the reduced
mass µ.

Figure 9.5: A plot of the effective potential Veff (r) = VC + V (r) as a function of the distance
r. The dashed lines indicate the repulsive positive centrifugal potential VC and the attractive
potential V (r) = −2.5/r. When E = E0 the motion is a circle and r = r0 . When E = E1 < 0,
the motion is bound between the two circles with radii r = r1 and r = r2 . When E > 0, the
motion is unbound.

So far in this section, we have learned that the energy of the reduced mass determines
the nature of its orbit. Now that we have a qualitative understanding of the orbit of µ, we
can use our knowledge of conic sections in order to determine the conditions under which
each orbit occurs. We know that conic sections have specific eccentricities, for example,
e = 0 for a circle. Using (9.4.8) we can now determine the energies of each type of orbit.
They are summarized in Table 9.1.

9.6 KEPLER’S LAWS OF PLANETARY MOTION

To conclude our work on the two-body problem, we will examine Kepler’s laws of planetary
motion in detail.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 288 — #304

i i

288 Classical Mechanics: A Computational Approach

9.6.1 Kepler’s First Law

As mentioned previously, Kepler’s first law states that planets orbit in an ellipse with the
Sun at one focus. We have found that the general motion is that of a conic section, however,
we will focus on ellipses in this section because of their importance in planetary and satellite
motion.
Let us begin our discussion by recalling (9.4.5),
s
1 1 `2 2E`2
= (1 + e cos θ) α= e = 1+ (9.6.1)
r α µk µk 2
and our choice of coordinates is such that r = rmin when θ = 0, and r = rmax when θ = π, as
shown in Figure 9.6, where the force center is at the focus F . From (9.4.5), we can easily
show,
α α
rmin = and rmax = (9.6.2)
1+e 1−e
In this section, we will assume that the reduced mass is a planet orbiting the Sun. Hence
we will use the terms perihelion as the planet’s distance of closest approach rmin to the Sun,
and aphelion as the planet’s greatest distance rmax from the Sun. We can define the major
axis 2a of the ellipse as

rmin + rmax = 2a (9.6.3)

and the quantity a is called the semimajor axis.

Figure 9.6: The motion of the planets around the Sun can be described by an ellipse which
has the Sun at one focus, F . The ellipse can be described mathematically using (9.4.5) in
polar coordinates (r, θ).

At θ = 90◦ and at θ = 270◦ the distance is equal to r = α, which is called the semi-latus
rectum. This quantity has dimensions of length, and depends on both the semi major axis
a of the elliptical motion, and on the eccentricity e:

α = a 1 − e2 (9.6.4)

The semi-latus rectum is shown in Figure 9.7.

The minor axis 2b is perpendicular to the major axis, and both can be related to the
energy of the orbit:

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 289 — #305

i i

Central Forces and Planetary Motion 289

α k
a= = (9.6.5)
1−e 2 2 |E|
α `
b= √ =p (9.6.6)
1−e 2 2µ |E|
Furthermore, we can rewrite the aphelion and perihelion distances as:

rmin = a(1 − e) and rmax = a(1 + e)

The eccentricity e can also be found from rmin and rmax :
rmax − rmin
e= (9.6.7)
rmax + rmin
The area of the ellipse is A = πab. In the special case of a circle the eccentricity e = 0,
resulting in rmin = rmax = a = b and A = πr2 .

Figure 9.7: Kepler’s first law places the Sun at the focus of an elliptical orbit

The next two examples shows how to plot the orbits of planets. The first example uses
data from the planets Mercury and Earth, while the second uses a numerical solution to
Newton’s second law.
Example 9.4: The orbits of Earth and Mercury around the Sun
The Earth’s distance from the Sun ranges from 147.5 million km (at perihelion), to
about 152.6 million km (at aphelion), while for Mercury these distances are from 46,000,000
to 70,000,000 km. Plot the orbits of Mercury and Earth around the Sun.
Solution:
The semi-major axis a can be found directly from the distances rmin and rmax for the
perihelion and aphelion by using (9.6.3), and the eccentricity e can be found by using
(9.6.7). Once the geometrical properties a, e are known, we know everything about the
orbits and we can plot them using the polar plot commands with the general equation
(9.6.1) for the elliptical orbits.
The Mathematica code uses the PolarPlot command to plot the two orbits and also
prints out the values of e for the two planets. In this polar plot the Sun is of course
at the origin (0, 0). Notice the almost completely circular orbit of the Earth (because the

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 290 — #306

i i

290 Classical Mechanics: A Computational Approach

eccentricity e = 0.0167 is close to zero), and the clearly elliptical orbit for Mercury (because
of the larger value of e = 0.21).

rmin = 147.5 ∗ 10∧ 6 ∗ 10∧ 3;

rmax = 152.6 ∗ 10∧ 6 ∗ 10∧ 3;

a = (rmin + rmax)/2;
eEarth = (rmax − rmin)/(rmin + rmax)//N ;

gr1 = PolarPlot[a ∗ (1 − eEarth∧ 2)/(1 + eEarth ∗ Cos[θ]), {θ, 0, 2 ∗ Pi},

Frame → True, LabelStyle → Medium, AxesLabel → {“x, m”, “y, m”},
PlotRange → All];

rmin = 46 ∗ 10∧ 6 ∗ 10∧ 3;

rmax = 70 ∗ 10∧ 6 ∗ 10∧ 3;

a = (rmin + rmax)/2;
eMercury = (rmax − rmin)/(rmin + rmax)//N ;

gr2 = PolarPlot[a ∗ (1 − eMercury∧ 2)/(1 + eMercury ∗ Cos[θ]),

{θ, 0, 2 ∗ Pi}, Frame → True, LabelStyle → Medium];

Print[“The eccentricity of Earth’s orbit is: ” <> ToString[eEarth]]

Print[“The eccentricity of Mercury’s orbit is: ” <> ToString[eMercury]]

OUTPUT:
The eccentricity of Earth’s orbit is: 0.0169943
The eccentricity of Mercury’s orbit is: 0.206897

Show[gr1,gr2]

2 × 1011

1 × 1011
y, m

0
Out[ ]=

-1 × 1011

-2 × 1011
-2 × 1011-1 × 1011 0 1 × 1011 2 × 1011
x, m

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 291 — #307

i i

Central Forces and Planetary Motion 291

Example 9.5: The orbit of planets around the Sun

Starting with the gravitational force, integrate Newton’s universal law of Gravity to
obtain the orbit of mass m around M , with m M . In order to simplify the presentation of
the numerical results, use arbitrary units with GM = 1, and the following initial conditions
at time t = 0: position x = 1, y = 0 and with initial velocity along the x and y axis equal
to vx = 0 and vy = 0.5 .
Solution:
The gravitational force acting between m and M is,

d2 r dv GM m
m 2
=m = − 2 r̂
dt dt r
Following a procedure similar to Example 9.1, we obtain the following two coupled differ-
ential equations:
d2 x GM
=− x (9.6.8)
dt2 (x + y 2 )
2 3 /2

d2 y GM
=− y (9.6.9)
dt2 (x + y 2 )
2 3 /2

In the Mathematica code that follows, the command NDSolve is used to solve the system
of differential equations (9.6.8) and (9.6.9), and the solutions x(t) and y(t) are stored in
the parameters x1 and y1 correspondingly. The Table command is used to construct the
pairs of position coordinates x(t) and y(t) from t = 0 to t = 3 s, in time steps of dt = 0.01 s
with the pairs stored in the parameter xyList.
The code calculates the semi-major axes a by evaluating the maximum and minimum
distance of the planet from the origin, and also calculates the eccentricity e and the location
of the two foci of the ellipse along the x-axis. The notation [[ ]] in Mathematica means
Part. The statement xyList[[All, 1]] tells Mathematica to take the first element of every
item in the list xyList. Correspondingly, the statement xyList[[All, 2]] tells Mathematica
to take the second element of every item in the list xyList. Finally a graph of the orbit
is produced by using the ListPlot command, and the location of the foci of the ellipse is
plotted by using the Graphics and Disk commands. Note that the points along the orbit
are sparse near the perihelion (0, 0) , but they are spaced much closer near the aphelion
(1, 0). This is because the planet moves much faster at perihelion than at aphelion.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 292 — #308

i i

292 Classical Mechanics: A Computational Approach

M = 1; x0 = 1; y0 = 0; vx0 = 0; vy0 = .5;

sol = NDSolve[{x00 [t] == −M ∗ x[t]/((x[t]∧ 2 + y[t]∧ 2)∧ (3/2)),

y 00 [t] == −M ∗ y[t]/((x[t]∧ 2 + y[t]∧ 2)∧ (3/2)), x[0] == x0, y[0] == y0, x0 [0] == vx0,
y 0 [0] == vy0}, {x, y}, {t, 0, 10}];

x1 = x/.sol[[1]];
y1 = y/.sol[[1]];
xyList = Table[{x1[t], y1[t]}, {t, 0, 10, .01}];

a = (Max[xyList[[All, 1]]] − Min[xyList[[All, 1]]])/2;

b = (Max[xyList[[All, 2]]] − Min[xyList[[All, 2]]])/2;
e = Sqrt[1 − b∧ 2/a∧ 2];

focusEllipse1 = Max[xyList[[All, 1]]] − a − a ∗ e;

focusEllipse2 = Max[xyList[[All, 1]]] − a + a ∗ e;
Print[“eccentricity=”, e ∗ a]
Print[“location of focus of ellipse=”, focusEllipse1]
Print[“location of focus of ellipse=”, focusEllipse2]

OUTPUT:
eccentricity=0.428563
location of focus of ellipse=0.0000153831
location of focus of ellipse=0.857142

Show[ListPlot[xyList], Graphics[{Black, Disk[{focusEllipse1, 0}, .02], Gray,

Disk[{focusEllipse2, 0}, .02]}]]

y
0.4

0.2

Out[ ]=
x
0.2 0.4 0.6 0.8 1.0

-0.2

-0.4

9.6.2 Kepler’s Second Law

Recall that the magnitude of the angular momentum vector is:

` = µr2 θ̇ (9.6.10)

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 293 — #309

i i

Central Forces and Planetary Motion 293

rdθ
dθ
r
Figure 9.8: A small wedge of the planet’s orbit traced out in a time dt.

From this magnitude we can now derive Kepler’s second law. We begin by considering a
small wedge of the planet’s orbit traced out in a time dt, as shown in Figure 9.8.
The wedge shown in Figure 9.8 forms an infinitesimal area element dA which in polar
coordinates can be written as:
r2
dA = dθ (9.6.11)
2
By dividing with dt and using (9.6.10), we obtain Kepler’s second law:

Kepler’s Second Law

dA r2 dθ `
= = = constant (9.6.12)
dt 2 dt 2µ

Note that Kepler’s second law holds for any central force. Equation (9.6.12) is the
mathematical form of Kepler’s second law which says that a line joining a planet to the Sun
sweeps out equal areas in equal times. This is illustrated in Figure 9.9. The result of Kepler’s
second law is that the planet’s speed increases as it reaches perihelion and decreases as it
approaches aphelion.

Figure 9.9: Kepler’s second law: A line connecting a planet to the Sun (S) sweeps out equal
areas in equal times. If the time TAB required to travel distance AB is equal to the time
TCD required to travel distance CD, then the swept out area (AREA)ABS must be equal
to the corresponding area (AREA)CDS .

Example 9.6 shows how to evaluate the angular momentum of the Earth as it orbits
around the Sun, and also how to estimate the Earth’s orbital velocity.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 294 — #310

i i

294 Classical Mechanics: A Computational Approach

Example 9.6: The orbital angular momentum of the Earth around the Sun
The speed of Earth at perihelion is 30,300 m/s, and its distance from the Sun ranges
from 147.3 million km (at perihelion), to about 152.6 million km (at aphelion). Evaluate:
(a) The orbital angular momentum of the Earth around the Sun.
(b) The orbital velocity of the Earth at aphelion.
(c) The speed of the Earth at the latus rectum point α of its elliptical orbit.
Solution:
(a) When the Earth is at the aphelion (or at the perihelion) of its orbit, the velocity
vector v = dr/dt is perpendicular to the position vector r, so that the radial component
of the velocity is zero (dr/dt = 0), therefore the magnitude of the angular momentum is
found from:
dr dr
= m | r || | sin 90◦ = mrv

| ` |= m r ×

dt dt
where v represents the magnitude of the orbital velocity at these two points and m =
5.97 × 1024 kg is the mass of the Earth. By using the given value of the speed at perihelion
and the given distance from the Sun, we can find the magnitude of the angular momentum
in SI units:

` = | ` |= mrperihelion vperihelion = 5.97 × 1024 kg 3.03 × 104 m/s 147.3 × 109 m

= 2.66 × 1040 kgm2 /s

(b) Since the magnitude of the angular momentum is conserved, we have:

` = mrperihelion vperihelion = mraphelion vaphelion

so that the orbital velocity at the aphelion is found from:

vaphelion = `/ m raphelion

By using this equation we find the speed at the aphelion in SI units:

2.66 × 1040 kgm2 /s

vaphelion = = 29, 200 m/s
(5.97 × 1024 kg ) × (152.6 × 109 m)
(c) The latus rectum point is located on the elliptical orbit of the Earth, at a distance
α vertically above the location of the Sun in Figure 9.7. By working as in the previous
example, we use the values of rperihelion and raphelion to find the values of the semi-major
axis a and the eccentricity e. Once we know a and e, we calculate the distance α from
(9.6.4). Finally we use conservation of angular momentum to find the vlatusrectum :

vlatusrectum = rperihelion vperihelion /α

which gives the value of vlatusrectum = 29, 800 m/s. Note that the velocity vector at the
latus rectum is not perpendicular to the position vector.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 295 — #311

i i

Central Forces and Planetary Motion 295

9.6.3 Kepler’s Third Law

Next we derive Kepler’s third law, in which he mathematically related the period of a
planet’s orbit to its semimajor axis. Using our more general result, we will derive the period-
distance relationship for the reduced mass.
We begin by integrating (9.6.12), Kepler’s second law,

2µ
dt = dA (9.6.13)
`
2µ
P= πab (9.6.14)
`
where P is the period, the time needed for a line joining the reduced mass and the force
center to sweep out the entire area of the ellipse. Next, we use (9.6.5) and (9.6.6) to express
the period P and as a function of energy E.
r
µ
P = πk |E|−3/2 (9.6.15)
2
By substituting now |E| = k/(2a) from Eq.(9.6.5), we obtain Kepler’s third law:

Kepler’s Second Law

4π 2 µ 3
P2 = a (9.6.16)
k

Kepler’s original statement was that the planet’s period squared was proportional to the
cube of its semimajor axis. Kepler didn’t know the proportionality constant. However, using
Newton’s second law, we have found it! If we consider the central force to be Newton’s law
of universal gravitation, the constant k = Gm1 m2 and the definition of the reduced mass
µ = m1 m2 / (m1 + m2 ), then Kepler’s third law becomes

4π 2 a3
P2 = (9.6.17)
G (m1 + m2 )
However, for a planet of mass m1 orbiting the Sun of mass m2 = M m1 , we have

4π 2 a3
P2 = (9.6.18)
GM
Example 9.7 illustrates Kepler’s third law with the Moons of Jupiter.

Example 9.7: The Moons of Jupiter

The following table lists the orbital properties of the four largest Moons of Jupiter,
discovered by Galileo Galilei. Use this data to find the mass of Jupiter.

Satellite Semi-major axis (km) Orbital period (days)

Io 421,700 1.7691
Europa 671,034 3.5512
Ganymede 1,070,412 7.1546
Callisto 1,882,709 16.689

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 296 — #312

i i

296 Classical Mechanics: A Computational Approach

Solution:
The data can be analyzed as shown in the Mathematica example below. First we set up
two lists called aValues and periods, which contain the semi-major axes a and the periods
P of the 4 Moons. The aValues list is converted in SI units by converting the units of
kilometers into meters, and the periods P are converted into SI units by converting days
into seconds. The command Transpose is used to create a new list called list, containing
the pairs of (a3 , P 2 ) for the four Moons. This is plotted using the ListPlot command, and
the graph is stored in the variable gr1. By using the FindFit command we fit a best line to
the graph, and we find the best slope of this line to have a slope of 3.11 × 10−16 (SI-units).
From Kepler’s law of the periods we have

4π 2 3
P2 = a
GM
where M is the mass of Jupiter. From this equation, the slope of a plot of P 2 vs a3 will
be equal to 4π 2 /(GM ).
By setting the best slope 3.11×10−16 = 4π 2 /(GM ) and using the NSolve command, we
obtain the mass of Jupiter as M = 1.89 × 1027 kg. This value is very close to the accepted
value of the mass of Jupiter MJ = 1.90 × 1027 kg.

G = 6.67408 ∗ 10∧ − 11;

aValues = {421700, 671034, 1070412, 1882709} ∗ 10∧ 3;

periods = {1.7691, 3.5512, 7.1546, 16.689} ∗ 3600 ∗ 24;

list = Transpose[{aValues∧ 3, periods∧ 2}];

gr1 = ListPlot[list, Frame → True, FrameLabel → {“a∧ 3 (m∧ 3)”, “P∧ 2 (s∧ 2)”},
PlotRange → All, LabelStyle → Large];

Print[“bestline fit = a+b*x with ”]

bestline = FindFit[list, a + b ∗ x, {a, b}, x]

Print[“bestlope =”, slope = b/.bestline]

Print[“Mass of Jupiter = ”]
NSolve[slope == 4 ∗ Pi∧ 2/(G ∗ M ), M ]

OUTPUT:

{a → 1.94945 × 106 , b → 3.1155753615415436`*∧ -16}

bestlope =3.1155753615415436`*∧ -16

“master˙book˙file” — 2020/9/5 — 9:32 — page 298 — #314

i i

298 Classical Mechanics: A Computational Approach

y
m3

R m2
r2

t
m 1-m x
r1
m1

Figure 9.10: The coordinate system for the planar restricted circular three-body problem.
The center of mass for m1 and m2 is at the origin. Mass m1 follows a circular orbit with
radius m about the center of mass and m2 follows a circular orbit of radius 1 − m. The third
body, m3 is located at the position R. Not shown in this figure is a vector r which points
from m1 to m2 and passes through the origin.

Noting that ω = 1 (because the period of m1 and m2 is 2π), we can write,

r 2 = (1 − m) cos t î + (1 − m) sin t ĵ (9.7.3)

Although not shown in Figure 9.10, the separation vector r = r2 − r1 points from m1 to m2
and passes through the origin. Noting that r = 1, we can write,

r1 = −m cos t î − m sin t ĵ (9.7.4)

The location of m3 can be written as

R = X î + Y ĵ (9.7.5)
We will use the Lagrangian to derive the equations of motion for the particle m3 . The
kinetic energy of the mass m3 is,
1
T3 = m3 Ẋ 2 + Ẏ 2 (9.7.6)

2
We find the potential energy using the formula, V = −Gm1 m2 /r where m and m2 are
the masses of the interacting particles and r is the distance between m and m2 . There are
two gravitational interactions that involve m3 , V13 is the potential energy between m1 and
m3 and V23 is the potential energy between m2 and m3 . They can be written as:

G(1 − m)m3
V13 = − (9.7.7)
r13
Gmm3
V23 = − (9.7.8)
r23

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 299 — #315

i i

Central Forces and Planetary Motion 299

where

2
r13 = (X + m cos t)2 + (Y + m sin t)2 (9.7.9)
2 2
2
r23 = (X − (1 − m) cos t) + (Y − (1 − m) sin t) (9.7.10)

are the distances between m3 and m1 and between m3 and m2 , respectively.

Because we are interested only in the motion of m3 (we already know the motion of m1
and m2 by our assumptions), we will write only the terms of the Lagrangian for the mass m3 .
The terms for the kinetic energies of m1 and m2 , and the term describing the gravitational
potential energy between m1 and m2 will not contribute to the equations of motion for
m3 because they are independent of the variables X, Ẋ, Y , and Ẏ . The Lagrangian of our
system (including only the terms relevant to m3 ) is L = T3 − (V13 + V23 ), which, when using
G = 1, can be written as,

1 (1 − m)m3 mm3
L = m3 Ẋ 2 + Ẏ 2 + + (9.7.11)
2 r13 r23
Note that (9.7.11) is time-dependent. We can eliminate the time-dependence by trans-
forming our coordinate system into a frame that is rotating with the masses m1 and m2 .
In this rotating frame m1 and m2 are at rest, with m1 located at the point (m, 0) and m2
located at the point (1 − m, 0). Let us use the notation (x, y) to describe the position of m3
in the rotating frame. The transformation that connects the coordinates (X, Y ) and (x, y)
is the rotation matrix

cos t − sin t

X x
= (9.7.12)
Y sin t cos t y
where the angle of rotation is θ = ωt, with ω = 1 due to our assumption about the period
of revolution for masses m1 and m2 . We can write out the matrix multiplication in (9.7.12)
to find,

X =x cos t − y sin t (9.7.13)

Y =x sin t + y cos t (9.7.14)

Using (9.7.13) and (9.7.14) we find,

i n i t i a l C o n d i t i o n 1 = [ −0.75 , 0 , 0 , 1]
t=np . l i n s p a c e ( 0 , 2 0 , 2 0 0 1 )
s o l 1 = o d e i n t ( threeBody , i n i t i a l C o n d i t i o n 1 , t )

initialCondition2 = [1.1 , 0 , 0 , 0.3]

t=np . l i n s p a c e ( 0 , 5 0 , 5 0 0 1 )
s o l 2 = o d e i n t ( threeBody , i n i t i a l C o n d i t i o n 2 , t )

initialCondition3 = [0.5 , 0.86 , 0 , 0.02]

t=np . l i n s p a c e ( 0 , 5 0 , 5 0 0 1 )
s o l 3 = o d e i n t ( threeBody , i n i t i a l C o n d i t i o n 3 , t )

initialCondition4 = [1.1 , 0.0 , 0 , 0.01]

t=np . l i n s p a c e ( 0 , 1 0 , 1 0 0 1 )
s o l 4 = o d e i n t ( threeBody , i n i t i a l C o n d i t i o n 4 , t )

f , ( ( ax1 , ax2 ) , ( ax3 , ax4 ) ) = p l t . s u b p l o t s ( 2 , 2 )

p l t . s u b p l o t s a d j u s t ( top = 0 . 9 9 , bottom =0.01 , h s p a c e =0.75 ,

,→ wspace =1.0)

ax1 . p l o t ( s o l 1 [ : , 0 ] , s o l 1 [ : , 1 ] )

ax2 . p l o t ( s o l 2 [ : , 0 ] , s o l 2 [ : , 1 ] )

ax3 . p l o t ( s o l 3 [ : , 0 ] , s o l 3 [ : , 1 ] )

ax4 . p l o t ( s o l 4 [ : , 0 ] , s o l 4 [ : , 1 ] )

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 304 — #320

i i

304 Classical Mechanics: A Computational Approach

9.9 END-OF-CHAPTER PROBLEMS

Section 9.1: Central Forces
1. Find the potential energy corresponding to the central force fields defined by the
following forces (K, α, β are constants).
(a) F = −K/r3
(b) F = α/r2 + β/r3
(c) F = −Kr
√
(d) F = 1/ r
(e) F = sin (πr).

2. A particle moves in the force field F = −K/r2 .

(a) Find the potential energy for this particle
(b) How much work is done by the force field in (a) moving the particle from a point
on the circle r = a > 0 to another point on the circle r = b > 0? Does the work
depend on the path? Explain.
3. A particle of mass m moves in a central force field defined by F = −K/r3 .
(a) Write an equation for the conservation of energy.
(b) Prove that if E is the total energy supplied to the particle, then its speed is given
by
q
v= K/(mr2 ) + 2E/m

4. A particle moves in a central force field defined by F = −Kr2 . It starts from rest at
a point on the circle r = a.
(a) Prove that when it reaches the circle r = b its speed will be
q
v = 2K(a3 − b3 )/3m

(b) Prove that the speed will be independent of the path.

5. A particle of mass m moves in a central force field F = −K/rn where K and n are
constants and K > 0. It starts from rest at r = a and arrives at r = 0 with finite speed
Vo .
(a) Prove that we must have n < 1.
(b) Prove that s
2Ka1−n
Vo =
m (n − 1)

6. Let L, M, and T represent the dimensions of length, mass and time, respectively. Find
the dimensions of the universal gravitational constant G.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 305 — #321

i i

Central Forces and Planetary Motion 305

Motion in Noninertial
Reference Frames

As we saw in Chapter 1, Newton’s second law is valid only for inertial reference frames.
However, it is sometimes the case where we need to describe the motion in a noninertial
reference frame. For example, long-range missile trajectories need to be described using
a noninertial reference frame, a frame fixed to the surface of the rotating (and therefore
noninertial) Earth. In Chapter 1 we found that using Newton’s second law in a noninertial
reference frame, resulted in the appearance of inertial forces . In this chapter, we will expand
on the idea of inertial forces; specifically, those arising in rotating reference frames. We will
discuss how a vector in a rotating reference frame can be described in a nonrotating frame,
and from that description, show how applying Newton’s second law in a rotating frame
results in inertial forces such as the Coriolis force and the centrifugal force. We will then
study how the Coriolis and centrifugal forces affect the motion of a particle near the surface
of the Earth. Finally, we will explore the famous problem of the Foucault pendulum and
projectile motion in a noninertial frame.

10.1 MOTION IN A NONROTATING ACCELERATING REFERENCE FRAME

Although we examined the case of a nonrotating accelerating frame in Chapter 1, it is
worthwhile revisiting it, this time using three dimensions. Consider a particle of mass m
whose motion is being described using two reference frames as shown in Figure 10.1.
The reference frame S 0 shown in Figure 10.1 is an inertial frame which we’ll imagine to
be at rest, and the frame S is accelerating away from S 0 with an acceleration A = R̈. In
Chapter 1, we denoted the moving frame with primes. In this chapter, we will switch our
notation for convenience and use primed variables to describe quantities measured in the
inertial frame S 0 and unprimed variables for quantities measured in the noninertial frame
S. We will be focusing on the motion of a particle in the noninertial frame and we do not
want to be complicating the notation with additional primes.
The position of the particle measured in S 0 can be written as r0 = R + r and the velocity
of the particle, as measured in the frame S 0 is therefore,

ṙ0 = ṙ + V (10.1.1)
where V = Ṙ is the instantaneous velocity of the frame S relative to S0. For example, if
S 0 is a stationary reference frame on the ground and S is fixed in an accelerating car, then
the particle’s velocity relative to the ground (ṙ0 ) is the velocity of the particle relative to

311

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 312 — #328

i i

312 Classical Mechanics: A Computational Approach

y m

r
y' r'
S x
R

S' x'
Figure 10.1: A particle of mass m located at a position r0 relative to the origin of an inertial
frame S 0 and at a location r relative to the origin of a noninertial frame S. The vector R
measures the position of the noninertial frame’s origin relative to the origin of the inertial
frame.

the car (ṙ) plus the velocity of the car (V). It should be noted that (10.1.1) only holds for
non-relativistic cases.
Next, we differentiate (10.1.1) to get the acceleration of the particle as measured in the
noninertial frame,

r̈ = r̈0 − A (10.1.2)
where A = V̇. Notice that similar to what was found in Chapter 1, an additional acceleration
measured in S 0 that is not measured in S. Multiplying (10.1.2) by the mass of the particle
m,

mr̈ =mr̈0 − mA (10.1.3)

F =F0 − mA (10.1.4)

where F = mr̈ are the forces acting on the particle as measured in the noninertial frame,
and F0 = mr̈0 are the forces acting on the system as measured by the noninertial frame. The
forces acting on the particle may include gravity, air resistance, friction, etc. These forces are
measured in both frames. However, there is an additional force measured in the noninertial
frame, −mA. This additional force is sometimes called an inertial force, a fictitious force,
or a pseudo force. In order to use Newton’s second law to describe motion in a noninertial
frame, we need to include the additional inertial force, which is not due to the particle’s
interaction with another body, but rather arises from the acceleration of the noninertial
frame.
Although the inertial force is sometimes called the fictitious force, inertial forces are real
to those moving in a noninertial reference frame. You have experienced noninertial forces
in an accelerating car as you are being pushed back on the seat, or on a rotating carnival
ride such as the Tilt-a-Whirl, where you are pushed against the side of a carriage as the
ride spins. In each case, the force is due to your inertia as opposed to the interaction with
another body, hence the name inertial since it is introduced in order to account for all of the
accelerations you are experiencing. By including inertial forces, we are able to use Newton’s
second law to describe motion in noninertial reference frames.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 313 — #329

i i

Motion in Noninertial Reference Frames 313

Example 10.1: A pendulum with an accelerating support

Consider a simple plane pendulum of mass m and a massless rod of length ` in a
downward uniform gravitational field g and whose support has a vertical upward acceler-
ation of a as shown in the diagram below. Find the frequency of small oscillations of the
pendulum.

θ ℓ

Solution:
According to (10.1.3), the force in the frame S which is accelerating with the pendulum
is,

F = (T + mg) − ma (10.1.5)
where T is the tension in the pendulum, g is the acceleration due to gravity, and a is a
vector that has a magnitude of a and points in the direction opposite of g. The term in
the parentheses of (10.1.5) is F0 , the sum of the forces acting on the system according to
Newton’s second law. We can rewrite (10.1.5) as

F = T + mgeff (10.1.6)
where geff = g −a is the effective gravitational field which includes the gravitational attrac-
tion and the acceleration of the pendulum’s support. It then follows that in the frame S,
the equation of motion of the pendulum for small oscillations is,
geff
θ̈ + θ=0 (10.1.7)
`
where θ is the angle that the pendulum makes with the vertical (as usual). From Chapter
6 we know that the coefficient in front of θ in (10.1.7) is the square of ω0 , the frequency
of small oscillations. Therefore, we find
r
|g − a|
ω0 = (10.1.8)
`
We see that the upward acceleration effectively reduces the gravitational field. The
result is that the accelerated frame is the same as having an additional gravitational
field (in this case, whose acceleration is in the opposite direction of g). The fact that
accelerating frames are indistinguishable from gravitational fields is an important element
in the general theory of relativity.

10.2 ANGULAR VELOCITY AS A VECTOR

Before we move on to rotating frames, we need to first reexamine angular velocity, in
particular, the vector nature of angular velocity. Euler’s rotation theorem says that any

ρ v

θ r

Figure 10.3: The northern hemisphere of the Earth showing the location of a point particle
(black dot) and its velocity, v.

In addition to a formal definition of angular velocity, we will also need a relationship

between the velocity of a particle and its angular velocity. You may recall from your intro-
ductory physics course that the tangential velocity v of a particle moving along a circle of
radius r with angular velocity ω is v = ωr. In this particular case, the axis of rotation went
through the center of the circle and was perpendicular to the plane of the circle. This simple
relationship between v and ω results from the fact that all of the motion is restricted to a
plane. That is not generally true.
Consider a particle, represented by the black dot in Figure 10.3 fixed to a location on
the Earth’s surface in the Northern Hemisphere. The angular velocity is essentially constant
and points from the South Pole to the North Pole. While every point on the Earth has the
same angular velocity, the particle’s latitude will affect the particle’s tangential velocity.
In Figure 10.3, the particle travels in a circle of radius ρ = |ρρ| as the Earth rotates. The
latitude of the particle is determined by the angle θ (the latitude is π/2 − θ). As θ increases,
the particle needs to travel a greater distance in the same amount of time (the angular
velocity is the same at all latitudes). Therefore, v = |v| must increase with an increasing θ.
We know that the tangential velocity of the particle is v = ρω from introductory physics.
However, ρ = r sin θ. Therefore, v = rω sin θ or we can write the tangential velocity as,

v = ω ×r (10.2.1)

Notice that (10.2.1) is the general form for the equation v = rω you learned in introductory
physics (where θ = π/2). We need to use (10.2.1) in this case because the origin of the
inertial frame is the center of the Earth, and the center of the Earth is not the center of
the circle along which the particle travels. Furthermore, the center of the Earth will serve
as the origin of the inertial frame that we will use in future problems.
Equation (10.2.1) is not unique to the velocity vector. Note that we can rewrite (10.2.1)
as,
dr
= ω ×r (10.2.2)
dt
which is not a formula unique to the vector r. In fact, for any vector Q that is constant in
the rotating reference frame, its time derivative as measured in the non-rotating frame is,

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 316 — #332

i i

316 Classical Mechanics: A Computational Approach

ê3
ω P
ê'3

r ê2
r'

R ê1

ê'2

ê'1
Figure 10.4: The inertial frame S 0 (primed unit vectors) and the noninertial rotating frame
S (unprimed unit vectors, red in e-book). The vector R represents the location of the origin
of the noninertial frame as measured in the inertial system. The vector ω represents the
angular velocity of the noninertial frame.

dQ
= ω ×Q (10.2.3)
dt
In the next section, we will provide a derivation of a more general form of (10.2.3) which
will include vectors that are not constant in the rotating reference frame.

10.3 TIME DERIVATIVES OF VECTORS IN ROTATING COORDINATE

FRAMES
Recall that in Section 10.1, we related quantities such as velocity and acceleration between
an inertial and noninertial frames. This was easily done because neither frame was rotating.
However, once we allow for the noninertial frame to rotate with respect to the inertial
frame, things become a bit more difficult. We will still be able to relate quantities such as
the acceleration measured in the rotating frame to the acceleration measured in the inertial
frame, but to do so, we will need to include the angular velocity of the rotating frame.
We begin by setting up two specific reference frames. Like in Section 10.1, we will use
primed coordinates to denote the inertial frame S 0 the fixed inertial frame, and unprimed
coordinates for for the noninertial frame S. The origin of the noninertial frame S is located
at a location R from the origin of S 0 and the noninertial frame is rotating with an angular
velocity ω with respect to S 0 . The two coordinate systems are shown in Figure 10.4.
In Figure 10.4, the inertial frame S 0 is represented by the primed unit vectors ê0i , where
we are using generalized unit vectors in order to simplify an upcoming derivation. The
noninertial frame S is represented by the unprimed unit vectors êi and are colored red in
the e-book. The unit vectors êi are fixed in the rotating frame. In other words, êi rotate with
the frame and, therefore, an observer in the rotating frame will measure êi to be constant
in both magnitude and direction. The vector ω (green in the e-book) represents the angular
velocity of the noninertial frame. The point P can be located using either coordinate system.
As measured in the inertial frame S, the point P is located at, r0 = R + r.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 317 — #333

i i

Motion in Noninertial Reference Frames 317

Now let’s consider an arbitrary vector Q which is measured in the noninertial frame.
The vector Q could be the position of a particle at the point P , the velocity of the particle,
or any other vector quantity. We will compute the time derivative of Q as measured in the
inertial (fixed) frame and the noninertial (rotating) frame using the following notation,

dQ
= the time derivative of Q relative to the inertial (fixed) frame S’
dt f

dQ
= the time derivative of Q relative to the noninertial (rotating) frame S.
dt r

We begin by writing Q in terms of the unit vectors that are fixed in the noninertial
frame,
X3
Q= Qi êi (10.3.1)
i=1

The coefficients Qi are the same in each frame. Observers in the inertial frame, would
see êi vary with time, but that would be the only difference between what is observed in
the two frames. If we calculate the time derivative of Q relative to the noninertial frame,
we would find,
3
dQ X
= Q̇i êi (10.3.2)
dt r
i=1

Because the scalar coefficients Qi are the same in each frame, we do not need to distin-
guish whether Q̇i is relative to the inertial or noninertial frame. Note that the derivative
doesn’t affect the unit vectors because they are constant relative to the noninertial frame.
Next, we compute the time derivative of Q relative to the fixed frame,
3 3
dQ X X dêi
= Q̇i êi + Qi (10.3.3)
dt f dt f
i=1 i=1

The first term on the right-hand side of (10.3.3) is what we found when computing the
time derivative relative to the noninertial frame. Notice that the second term in the right-
hand side of (10.3.3) is the time derivative of the vector êi , which is constant in the rotating
frame. We can use (10.2.3) to write,

dêi
= ω × êi (10.3.4)
dt f
Therefore, (10.3.3) becomes,
3 3
dQ X X
= Q̇i êi + ω × (Qi êi ) (10.3.5)
dt f i=1 i=1

or using (10.3.2):

Time Derivative of a Vector Q Relative to an Inertial Frame

dQ dQ
= + ω × Q. (10.3.6)
dt f dt r

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 318 — #334

i i

318 Classical Mechanics: A Computational Approach

Notice that (10.3.6) is what we would expect from our studies of translationally moving
reference frames. To better understand this, let Q = r, the position of a particle as measured
in the noninertial (rotating) frame. The first term in the right-hand side of (10.3.6) is the
velocity measured in the rotating frame. The second term is the translational velocity of
the particle at r due to the rotation of the noninertial frame. Hence, the velocity measured
by an observer in the fixed frame is the velocity of the particle in the moving frame plus the
velocity of the frame itself. Note that one result of (10.3.6) is that the angular acceleration
ω is the same in each frame because ω × ω = 0.
ω̇
Equation (10.3.6) tells us that a vector’s rate of change can be related between inertial
and noninertial frames. Newton’s second law provides a particle’s equations of motion by
stating that the net force acting on a particle is proportional to the particle’s change in
velocity as measured in an inertial frame. Thus, (10.3.6) allows us to relate the acceleration
of a particle in an inertial frame, found from Newton’s second law, to situations where the
observer is in a noninertial frame. In the next section, we will use (10.3.6) to find Newton’s
second law in a rotating frame.

10.4 NEWTON’S SECOND LAW IN A ROTATING FRAME

In order to find an equation which describes the motion of a particle in a rotating frame, we
will need to modify Newton’s second law for use in noninertial frames. To find the form of
Newton’s second law in a rotating frame S, we will use, as a reference, Figure 10.4. Suppose
that there is a particle with mass m located at the point P in Figure 10.4. We will allow
the noninertial frame to be both rotating and have a linear acceleration relative to S 0 . The
angular velocity of S will be ω . It will be helpful (but not necessary) to think of the inertial
frame as fixed.
In the inertial frame S 0 we can write Newton’s second law as,
0
0 dv
F =m (10.4.1)
dt f

where v0 is the velocity of the particle relative to the inertial frame S 0 and F0 is the net
force acting on the particle. The net force F0 includes the interaction forces between the
mass m and other bodies such as gravity, air resistance, friction, etc. In order to find the
form of Newton’s second law for a noninertial frame, we will need to rewrite the acceleration
as measured in the inertial frame (dv0 /dt)f in terms of the acceleration as measured in the
noninertial frame.
We begin by finding v0 . Figure 10.4 shows that,

r0 = R + r (10.4.2)
The velocity v0 is found by taking the time derivative of r0 relative to to the inertial frame.
0
0 dr dR dr
v = = + (10.4.3)
dt f dt f dt f
We use (10.3.6) on the second term in the right-hand side of (10.4.3) in order to find
how an observer in the inertial frame S 0 measures the velocity of the particle moving in the
noninertial frame S,
dr dr
= +ω ×r (10.4.4)
dt f dt r

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 319 — #335

i i

Motion in Noninertial Reference Frames 319

In Figure 10.5, a particle of mass m (not shown) is located at r in a reference frame

that is rotating relative to an inertial frame. In the inertial frame, an observer would see
the particle moving in a circle of radius ρ = r sin θ. In order to find the direction of the
centrifugal force, −m [ω ω × (ω ω × r)], we first need to find the direction of the vector ω × r.
The vector ω × r is perpendicular to both ω and r and therefore lies in the plane of the circle
as shown most clearly in Figure 10.5(b). Next, the direction of ω × (ω ω × r) is perpendicular
to both ω and ω × r and therefore points radially in towards the center of the circle. Finally,
the centrifugal force is −m [ω ω × (ω
ω × r)], which points in the opposite direction of ω × (ωω × r)
and therefore points radially outward from the center of the circle.
We can compute the magnitude of the centrifugal force from (10.4.18). Let us focus on
describing the motion along the circle in cylindrical coordinates ρ, φ, and z. Note that in
this case, we want an azimuthal angle to describe the motion around the circle of radius ρ,
as shown in Figure 10.5a. Hence we will use φ to denote that angle consistent with spherical
coordinates. Using cylindrical polar coordinates, we find ω × r = ωr sin θφ̂ φ. Furthermore,
ω ×(ωω × r) = −ω 2 r sin θρ̂ρ (note that ω = ωẑ in cylindrical polar coordinates with ẑ× φ̂
φ = −ρ̂ρ).
Therefore, we find that

Fcent = mρω 2 ρ̂ (10.4.19)

Hence, a particle moving along a circle would experience an outward force with a magnitude
mρω 2 . Notice that this outward force is consistent with the centripetal force you learned
about in introductory physics, Fc = −mrω 2 ρ̂ρ. Recall that the centripetal force is the net
force that keeps a particle moving along a circular path with a constant speed. In intro-
ductory physics you may have been told that the particle following uniform circular motion
doesn’t fall towards the force center because of its inertia, its velocity is tangent to the
circular path. In introductory physics, your work focused on inertial frames. Now that we

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 322 — #338

i i

322 Classical Mechanics: A Computational Approach

have an understanding of noninertial frames, we have another means of describing uniform

circular motion. From the perspective of the particle, which is in a noninertial frame, there
is a centrifugal force acting on it such that the net force is zero in the radial direction , and
the particle doesn’t fall towards the force center.
Because the Earth is rotating, objects near the surface of the Earth experience the cen-
trifugal force. The outward-pointing centrifugal force reduces the effective acceleration due
to gravity by an amount depending on the particle’s latitude. The next example illustrates
this.
Example 10.2: Gravity and the centrifugal force
Consider a particle of mass m at rest on the surface of the Earth. Find the effective
force acting on the particle and compute the difference in the effective acceleration due to
gravity between the North Pole and a point on the equator.
Solution:
It will be helpful to include a diagram of the particle’s location before we perform any
calculations. The diagram below shows a particle’s position RE on the Earth’s surface.

ω
-ω×(ω×r)
g0
ρ g

θ R
E

The position vector RE has a magnitude of 6, 371, 000 m and points in a direction from
the center of the Earth to the location of the particle on the Earth’s surface. The angle θ
is the same angle from Figure 10.5a and is related to the particle’s latitude λ such that
λ = π/2 − θ. To a stationary observer above the Earth, the particle appears to move in a
circle of radius ρ = RE sin θ.
According to (10.4.16), the effective force acting on the particle is,

F = F0 + mg0 − m [ω
ω × (ω
ω × RE )] (10.4.20)
where
GME
g0 = − 2 R̂E
RE
is the acceleration due to the gravitational attraction, found from Newton’s universal law of
gravitation, the Earth’s mass is ME = 5.972 × 1024 kg and F0 is the sum of all of the forces
other than gravity acting on the particle. The only inertial force acting on the particle is
the centripetal force. The Coriolis force is zero because the particle is not moving relative
to the Earth’s surface and, to good approximation, ω is a constant and therefore ω̇ ω = 0.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 323 — #339

i i

Motion in Noninertial Reference Frames 323

We can collect the second and third terms in (10.4.20) and define a new acceleration
due to gravity called an effective acceleration due to gravity g where,

g = g0 − ω × (ω
ω × r) (10.4.21)

and it is g that determines the acceleration due to free fall and the period of a pendulum at
a point on Earth. Note that the centrifugal acceleration effectively reduces the acceleration
due to gravity. In addition, a plumb line will lie along the direction of g and will slightly
deviate from true vertical because of the outward-directed centrifugal force shown in the
diagram above. Furthermore, the surfaces of the Earth’s oceans are perpendicular to g not
g0 .
We can compute the effective acceleration due to gravity at the North Pole and equator
by first finding the magnitude of g0 ,

6.674 × 10−11 Nm2 /kg 2 5.972 × 1024 kg

GME
g0 = 2 = 2 = 9.820 m/s2
RE (6.371 × 10 m)
6

Next, we compute the magnitude of the centripetal acceleration at the North Pole and
equator. We will use acent = RE ω 2 sin θρ̂ρ which comes from dividing (10.4.19) by the mass
m. Note that we want the component acent sin θ of acent which is parallel to RE . Note
that for the North Pole, θ = 0 and for the equator θ = π/2. In each case ω = 2π rad/day =
7.272 × 10−5 rad/s,

acent = RE ω 2 sin2 0 =0 m/s2 North Pole

π
acent = RE ω 2 sin2 =0.0337 m/s2 Equator
2
The effective gravitational acceleration is therefore, g = g0 − acent , and we find that the
effective acceleration due to gravity at the North Pole is 0.0337 m/s2 , or about 0.3%,
greater than at the equator. Furthermore, the centrifugal force is responsible for the oblate-
ness of the Earth. The Earth’s polar radius is 21.4 km greater than its equatorial radius!
Note that the additional bulge at the Earth’s equator actually means that the gravitational
acceleration at the poles is actually 0.5% greater than that at the equator.
A result of Example 10.2 is that in general, we can rewrite the effective force (10.4.16)
acting on a particle near the surface of the Earth as,

F = F0 + mg − 2m (ω
ω × v) (10.4.22)
where F0 is the sum of all of the interaction forces experienced by the particle excluding
its gravitational attraction to the Earth. We then combined mg0 with the centrifugal force
into a new force mg. The final term in (10.4.22) is the Coriolis force, which will be the topic
for the next section.

10.4.2 The Coriolis Force

When a particle is moving with a velocity v relative to a rotating reference frame, an
additional inertial force called the Coriolis force will be present. The Coriolis force is,

Fcor = −2m (ω
ω × v) (10.4.23)
It is clear from (10.4.23) that the Coriolis force depends on the velocity of a particle.
Particles that are not moving will not experience a Coriolis force. We can estimate the size

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 324 — #340

i i

324 Classical Mechanics: A Computational Approach

of the Coriolis force using (10.4.23). In the previous subsection, we used ω ≈ 7.3 × 10−5
rad/sec. Therefore a golf ball at a latitude of 45 degrees North traveling at 70 m/s will
experience a Coriolis force of 0.007 N, a very small effect considering its time of flight.
However, an intercontinental ballistic missile traveling at 7 km/s will experience a Coriolis
force of 0.7 N which could have a more considerable effect on the missile. However, missiles
are also affected by the weather and their courses are corrected by computer guidance
systems.
The direction of the Coriolis acceleration (and therefore Coriolis force) is determined
by the ω × v term. As shown in Figure 10.6, the Coriolis acceleration is to the right of
the velocity when the particle is moving in Earth’s Northern Hemisphere. In the Southern
Hemisphere, the Coriolis acceleration will be to the left of the particle’s velocity.
Besides missile trajectories, the Coriolis force has an important impact on Earth’s
weather, deflecting air currents in the atmosphere. In the Northern Hemisphere, air is
deflected toward the right, whichleads to high pressure to the right of the airflow, and
low pressure to the left (opposite in Southern Hemisphere), resulting in the counterclock-
wise circulation of air in the Northern Hemisphere and the clockwise circulation in the
Southern. The result can be hurricanes and typhoons.

-ω×v
Figure 10.6: The potential direction of the Coriolis acceleration, acor = −2 (ω ω × v), can
exert on a particle moving with a velocity v relative to the angular velocity ω of a rotating
reference frame. The acceleration deflects the particle towards its right in the Northern
Hemisphere and towards its left in the Southern Hemisphere.

Example 10.3: The horizontal deflection of a plumb line

Consider a particle of mass m falling freely in Earth’s gravitational field. The particle
starts at a height H above the Earth’s surface and is at a latitude of λ. Find the particle’s
horizontal deflection from the plumb line due to the Coriolis force.
Solution:
A diagram showing the coordinate system appears as follows.

i i

326 Classical Mechanics: A Computational Approach

Using g = −gêz and (10.4.26) we find that (10.4.24) becomes three equations,

ẍ =0 (10.4.27)
ÿ =2ωgt cos λ (10.4.28)
z̈ = − g (10.4.29)

which show us that there is no deflection to the south (x-direction) and the vertical motion
is a free fall. The deflection is towards the east (y-direction) and is found by integrating
(10.4.28) twice, once with with initial condition ẏ(0) = 0 and then a second time with
initial condition y(0) = 0. The result is,
1
y(t) = ωgt3 cos λ (10.4.30)
3
We can find the time of flight t by integrating (10.4.29) twice to get z(t) = z(0) −
1 2
(note ż(0) = 0) with z(0) = H. Solving for t, we find t = 2H/g and therefore the
p
2 gt
displacement from the plumb line of the falling particle is:
s
1 8H 3
y= ω cos λ (10.4.31)
3 g
How large is this? An object dropped from a height of 100 m at a latitude of 45◦ North
is deflected by 1.55 cm, if we neglect the effects of air resistance.

10.5 FOUCAULT PENDULUM

One of the most famous problems involving noninertial reference frames is the Foucault
pendulum, which was first devised by French physicist Jean Foucault (1819–1868). The
pendulum consists of a heavy mass m suspended from a tall ceiling by a long and light
wire. As the pendulum oscillates, it precesses in the horizontal plane. In some cases, the
pendulum is set up such that it knocks over pegs as it precesses. You might have seen a Fou-
cault pendulum in-person at a science museum or university. A quick search for “Foucault
pendulum” online will result in many videos of Foucault pendulums in action. Foucault
originally conceived the pendulum as a demonstration of Earth’s rotation. In 1851 he used
a 28-kg mass suspended from a 67-m wire from the dome of the Pantheon in Paris, France.
In this section, we will derive the equations of motion for the Foucault pendulum and show
how its precession period is proportional to the pendulum’s latitude.
We begin with a schematic diagram of the pendulum. Figure 10.7 shows a Foucault
pendulum of mass m connected to a long light wire of length `. The coordinate frame
illustrated is a noninertial frame whose xy-plane is tangent to the Earth’s surface at the
origin and whose z-axis is oriented along the local vertical direction. The angle α is small.
There are two forces acting on the Foucault pendulum, gravity, and tension. The force of
gravity mg uses the effective gravitational acceleration g which includes the centrifugal
force as previously discussed. The tension in the wire T is illustrated in Figure 10.7, which
also includes the x- and y-components of T. Note that the z-component of T will be
discussed later in the problem. Finally, we note that the Foucault pendulum’s motion is
largely confined to the xy-plane. In other words, ż ẋ and ż ẏ.
Assuming that the Earth’s angular acceleration is constant, which over the period of
the pendulum is an excellent approximation, we can use (10.4.22) to find the acceleration
of the mass m by dividing (10.4.22) by m and using F0 = T,

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 327 — #343

i i

Motion in Noninertial Reference Frames 327

T
ey

Tx
Ty
ex
mg

Figure 10.7: A Foucault pendulum of mass m and length ` suspended from a large vertical
height at a point on the z-axis.

T
a= + g − 2 (ω × v) (10.5.1)
m
To get x(t) and y(t) (note that with our assumptions z(t) = 0), we will need to find the
components of the vector equation (10.5.1), integrate and solve the resulting second order
ODE.
To get the ODE, we write the components of each of the vectors on the right-hand side of
(10.5.1) noting that a = ẍêx + ÿêy + z̈êz . We can see in Figure (10.7), that g = −gêz where
g is the magnitude of the local effective acceleration due to gravity. Next, the components
of the tension can be found using the fact that α is a small angle and ` is large,
x y
Tx ≈ − T Ty ≈ − T Tz ≈T (10.5.2)
` `
The Coriolis acceleration ω × v can be found using the method outlined in Example
10.3,

êx êy êz
ω × v = −ω cos λ 0 ω sin λ (10.5.3)

ẋ ẏ 0
= (−ẏω sin λ) êx + (ẋω sin λ) êy + (−ẏω cos λ) êz (10.5.4)

Therefore, inserting the vector components into (10.5.1) we obtain,

T x

ẍ = − + 2ẏω sin λ
m` 


T y

ÿ = − − 2ẋω sin λ (10.5.5)
m` 

T


z̈ = − g + 2ẏω cos λ 

m
The pendulum stays in the xy-plane, therefore z̈ ≈ 0. Furthermore, ẏω is small compared
to the acceleration due to gravity and the tension. Therefore, we can write T = mg using
the equation for z̈ in (10.5.5). The result is that we can rewrite (10.5.5) as,

i i

Motion in Noninertial Reference Frames 329

ez (vertical)

ey
(East)

ex
(South)

Figure 10.8: The coordinate system for a projectile launched in the yz-plane towards the
East, at an angle α with a speed v0 .

It is easier to get some physical insight into these equations if we rewrite (10.5.15) as a
matrix equation,

cos (ωz t) sin (ωz t) x (t)

0
x(t)
= (10.5.16)
y(t) − sin (ωz t) cos (ωz t) y 0 (t)
The appearance of the rotation matrix in (10.5.16) tells us that the Foucault pendulum’s
plane of oscillation is rotating through an angle ωz t with a rotation frequency of ωz = ω sin λ.
Therefore, the precession rate of the pendulum depends on the latitude of the pendulum.
We also see how the Foucault pendulum gives a demonstration of the Earth’s rotation as
its rotation rate is dependent on the Earth’s angular velocity.

10.6 PROJECTILE MOTION IN A NONINERTIAL FRAME

As a final section to this chapter, we will explore the problem of projectile motion in a
noninertial reference frame under the influence of Earth’s gravitation. We will compute the
Coriolis deflection of a projectile launched at a latitude λ = 50◦ North at an angle α = 45◦
with a speed of v0 = 200 m/s towards the East.
To address the problem of projectile motion in a noninertial frame, we will need to set
up a coordinate system as shown in Figure 10.8. We have shown the noninertial frame, the
frame in which observers are watching the projectile’s flight. This frame is in a location
in Earth’s Northern Hemisphere. The êy and êx directions correspond to east and south,
respectively. The vector êz is oriented to point along the vertical direction. The projectile
is launched in the yz-plane, such that if the Earth were not rotating, the projectile would
land at a location R = Rêy , or in other words, would land a distance R from the origin
along the y-axis.
By including the effect of the Earth’s rotation, we expect that the projectile will land
somewhere to the south of the y-axis. To find where the projectile lands, we will follow a
process similar to that done for finding the rotation rate of the Foucault pendulum in the
previous section. We begin by writing the acceleration of the projectile using (10.4.22),

a = g − 2 (ω × v) (10.6.1)

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 330 — #346

i i

330 Classical Mechanics: A Computational Approach

Other than gravity, there are no external forces acting on the system. Therefore, F0 = 0.
As with the Foucault pendulum, the effective acceleration due to gravity is g = −gêz . The
Coriolis acceleration is,

êx êy êz
ω × v = −ω cos λ 0 ω sin λ (10.6.2)

ẋ ẏ ż
ω × v = (−ẏω sin λ) êx + (żω cos λ + ẋω sin λ) êy + (−ẏω cos λ) êz (10.6.3)

Notice that we made no assumptions about the components of the projectile’s velocity.
Inserting the accelerations into (10.6.1), we obtain the following equations,

ẍ =2ẏω sin λ



ÿ = − 2ω (ż cos λ + ẋ sin λ) (10.6.4)
z̈ = − g + 2ẏω cos λ



The result is a system of coupled ODEs. While it is possible to argue some approxi-
mations to simplify (10.6.4) (such as 2ẏω cos λ g), we will seek a numerical solution of
(10.6.4) using Mathematica. Note that statements between (* and *) are comments.

(* Define the parameters *)

ω = 7.272 ∗ 10∧ (−5);
α = 45 ∗ Pi/180;
λ = 50 ∗ Pi/180;
v0 = 200;
g = 9.8;

(* Use parameters to compute the initial velocities *)

z0 = v0*Sin[α];
y0 = v0*Cos[α];

(* Find the solution to the ODEs *)

soln = NDSolve[{x00 [t] == 2 ∗ y 0 [t] ∗ ω*Sin[λ], y 00 [t] == −2 ∗ z 0 [t] ∗ ω*Cos[λ] − 2 ∗ x0 [t] ∗
ω*Sin[λ], z 00 [t] == −g + 2 ∗ y 0 [t] ∗ ω*Cos[λ], x0 [0] == 0, y 0 [0] == y0, z 0 [0] == z0, x[0] ==
0, y[0] == 0, z[0] == 0}, {x, y, z}, {t, 0, 2000}];

(* Find the time of flight *)

T = FindRoot[(z[t]/.soln) == 0, {t, 10, 2000}][[1, 2]]
OUTPUT: 28.9005

(* Find the Coriolis deflection *)

x[T ]/.soln
OUTPUT: {6.57713}

Algorithm 9: A Mathematica algorithm for solving the system of equations (10.6.4)

describing a projectile launched at an initial velocity of 200 m/s at and angle 45 degrees
above the horizon. The projectile is located at a lattitiude of 50 degrees North.
Algorithm 9 contains a Mathematica algorithm which uses NDSolve to solve the system
of coupled equations (10.6.4) for λ = 50 degrees North latitude, α = 45 degrees, and v0 = 200
m/s. We don’t know a priori the time of flight of the projectile, so we choose a suitably

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 331 — #347

i i

Motion in Noninertial Reference Frames 331

long time and have Mathematica find the root of z(t) using the FindRoot command. Once
we have the time of flight T , we need to evaluate the numerical solution at t = T . We find
that the deflection is about 6.5 m. The range of this projectile is around 4080 m, so the
deflection is small but significant if one needs high precision.

10.7 CHAPTER SUMMARY

It is sometimes necessary to describe the motion of a particle in a noninertial reference
frame. Although Newton’s second law is valid only in inertial frames, we can modify
Newton’s second law for use in a noninertial frame. In order to modify Newton’s second
law for a noninertial frame, we need to relate the time derivative of any vector Q relative
to a rotating noninertial frame rf , to that of an inertial frame. The relationship is

dQ dQ
= +ω ×Q
dt f dt r
where ω is the angular velocity of the noninertial frame with respect to the inertial frame.
Once we have a means of transforming a time derivative from a noninertial (rotating)
frame to an inertial frame, and vice versa, we can rewrite Newton’s second law for rotating
frames as,

ma = F0 − mA − 2m (ω
ω × v) − m (ω̇
ω × r) − m [ω
ω × (ω
ω × r)]
where:
a = the acceleration relative to the noninertial frame
F0 = the sum of the “interaction” forces acting on the particle (gravity,
friction)
A = the linear acceleration of the noninertial frame relative to an inertial one
−2 (ωω × v) = the Coriolis acceleration
ω × r = an acceleration due to the angular acceleration of the noninertial frame
−ω̇
ω × (ω
−ω ω × r) = the centrifugal acceleration.

The last four terms in the equation above are referred to as “inertial forces” or fictitious
forces, because they arise from the acceleration of the noninertial frame and are not due
to the interaction between the particle and another body.

10.8 END-OF-CHAPTER PROBLEMS

Section 10.1: Motion in a Nonrotating Accelerating Reference Frame
1. A pendulum with mass m hangs from the ceiling of an accelerating car by a massless
rod. As the car accelerates forward, the pendulum tilts backwards. Find the pendu-
lum’s equilibrium angle.
2. Consider a helium-filled balloon attached to the floor of an accelerating car by a
massless string. As the car accelerates forward, in what direction does the balloon
tilt?

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 332 — #348

i i

332 Classical Mechanics: A Computational Approach

Section 10.2: Angular Velocity as a Vector

3. Show that relative angular velocities add in the same way as translational velocities.
For example, we know that if a particle is moving relative to a frame S with a velocity
v and that the frame S is moving with a velocity V relative to the inertial frame
S 0 , then the velocity of the particle relative to S 0 is v0 = V + v. Repeat this line of
reasoning with a body rotating with an angular velocity ω relative to a rotating frame
S that rotates with an angular velocity Ω relative to an inertial frame S 0 . Show that
the angular velocity of the body relative to S 0 is ω 0 = Ω + ω. Assume that the reference
frames S and S 0 have the same origin.
Section 10.4: Newton’s Second Law in a Rotating Frame
4. Consider a bullet fired horizontally towards the South with a velocity v0 = 500 m/s at
a latitude λ = 50◦ north. How does the Coriolis force acting on the bullet compare to
is weight?
5. A bucket of water is rotating about its vertical symmetry axis. What is the shape of
the water in the bucket?
6. Atomic clocks show that a modern day is longer by about 1.7 milliseconds than a
century ago; hence although small, ω̇ 6= 0 where ω is the angular velocity of the Earth.
Compute the acceleration due to the changing angular velocity ω̇ ω × r on a particle
fixed to the Earth’s surface located in Williamsburg, Virginia, USA.
7. Let S be a frame rotating with an angular velocity ω relative to an inertial frame S 0 .
Further, let the origins of both frames be the same. A particle has a potential energy
V . Find the Lagrangian in terms of the position and velocity coordinates r and ṙ in
S. Show that the Euler-Lagrange equations yield (10.4.15) with A = 0:

F0 = mA + ma + 2m (ω
ω × v) + m (ω̇
ω × r) + mω
ω × (ω
ω × r)

8. Plot the effective gravitational acceleration as a function of latitude.

9. A baseball is launched vertically upward from the ground with an initial velocity v0
at a latitude λ. Find the Coriolis deflection of the ball when it returns to the ground.
10. Find the deflection from the plumb line for a 1.0 kg particle falling from a height H =
100 m under the Earth’s gravity, experiencing linear air resistance. Assume that the
horizontal component of the particle’s velocity is small, so that the drag force acts
only in the vertical direction. Thus the drag force is Fd = −bż ẑ, where b = 0.1 Ns/m.
Compare the answer to the result if the particle fell from the same height in vacuum
at a latitude of 45 degrees.
11. Consider a particle of mass m confined to move in a vertical plane rotating with a
constant angular velocity ω about the y-axis. Find the the equations of motion for
the particle and describe the particle’s motion.
12. Repeat Problem 11, but now assume the plane’s angular velocity is increasing in time
such that its angular acceleration is α = αêy where α > 0.
13. A baseball is dropped from a height of 30 m at a latitude of 32° South. In what
direction and how far did the ball land from the vertical? How does the direction
compare to the result of Example 10.3? Explain any similarities or differences.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 333 — #349

i i

Motion in Noninertial Reference Frames 333

Section 10.5 The Foucault Pendulum

14. Show that (10.5.9)
q q
−iωz t
q(t) = e A exp it ωz + β + B exp −it ωz + β
2 2 2 2

is the solution to the second-order ODE (10.5.8) :

q̈ + 2iωz q̇ + β 2 q = 0.

15. What is the rate of rotation of Foucault pendulum’s plane of oscillation located at the
equator? How about the North Pole? Explain your answers.
16. What is the rate of rotation of a Foucault pendulum’s plane of oscillation if the
pendulum is located in Paris, France? How about Paris, Kentucky, USA. Plot the
Foucault pendulum’s rotation rate as a function of latitude. Where is its maximum?
What is the minimum value? Where does the minimum value occur?
17. Consider a Foucault pendulum experiencing a damping force F0 = −bvv̂. Is the pre-
cession frequency changed?
18. Is the Foucault pendulum’s precession frequency significantly changed if ω̇ ω 6= 0, where
ω is the angular velocity of the Earth about its rotation axis? To answer this question,
assume that the magnitude of the Earth’s angular velocity changes at a constant rate,
but its direction does not change (this is not true over long time periods). Furthermore,
assume that the change in the angular velocity is very small. For example, consider
the situation described in Problem 26 from Chapter 4 which states that the Earth’s
angular velocity changed from 7.6 × 10−5 rad/s to 7.3 × 10−5 rad/s during a time
period of 350 million years. Let the pendulum have a mass m and a length ` (assumed
to be very long) and to be located at a latitude λ in the Northern Hemisphere.
Section 10.6: Projectile Motion in a Noninertial Frame
19. Make the approximation suggested in the text after (10.6.4), then solve the resulting
system of two coupled ODEs in closed form. You will want to use a symbolic ODE
solver such as Mathematica’s DSolve. After finding the solutions, perform a series
expansion for x(t) and y(t) and interpret your results.
20. Using the same initial velocity and latitude used in Section 10.6, find the range and
Coriolis deflection of a 1.0 kg projectile experiencing linear air resistance of the form
F0 = −bvv̂ where b = 0.3 kg/s.
21. Repeat Problem 20, but this time, the projectile experiences quadratic air resistance
of the form F0 = −bv 2 v̂ where b = 0.3 kg/m.
22. A projectile is fired due east from a point on the Earth’s surface with a latitude of
λ (in the Northern Hemisphere). The projectile’s initial speed is v0 and is launched
at an angle α with respect to the horizontal. Calculate the lateral deflection of the
projectile as it hits the ground. Do this problem without the aid of a computer.
23. There is a legend that during World War I, the British navy consistently missed
German ships when fighting near the Falkland Islands because their ships did not
properly account for the Coriolis force. We won’t discuss the validity of this legend

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 334 — #350

i i

334 Classical Mechanics: A Computational Approach

here, but for a moment, let’s suppose that it is true and that the Coriolis force was
known, but the ship’s guns were set up to hit their targets for battles in the Northern
Hemisphere. The Falkland Islands are near 50◦ South latitude. Suppose the guns were
set to accurately hit targets at 50◦ North latitude, by how much did the British ships
miss their target during the Falkland Islands engagement? Assume that the German
ships were due east of the British ships and that the British ships’ guns had a muzzle
velocity of 500 m/s and the shells were shot at an angle of 20◦ with respect to the
horizontal.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 335 — #351

Example 11.1: Example of angular momentum and angular velocity for Figure
11.1
Evaluate the angular momentum and torque N acting on the mass m in the situation
shown in Figure 11.1. Assume that the mass m moves with a constant angular velocity Ω
around the z-axis.
Solution:
The Mathematica code is shown in Algorithm 10. We use cylindrical coordinates with
the angle increasing with time according to θ = Ωt. The program calculates the cross
product ω × r using the Cross command and verifies that this is indeed equal to the
velocity vector (dx/dt, dy/dt).
In this example, the angular momentum vector points always along the z-axis and its
magnitude is equal to `z = mR2 Ω.
The torque N = I Ω̇ = 0, since the mass moves with a constant angular velocity Ω
around the z-axis.

x = R ∗ Cos[Ω ∗ t];
y = R ∗ Sin[Ω ∗ t];
z = 0;
r = {x, y, z};
v = {D[x, t], D[y, t], D[z, t]}

OUTPUT: {−RΩSin[tΩ], RΩCos[tΩ], 0}

ω = Ω ∗ {0, 0, 1};
Cross[ω, r] == v

OUTPUT: True

L = m ∗ Cross[r, v]//Simplify

OUTPUT: 0, 0, mR2 Ω

Algorithm 10: The Mathematica code to solve Example 11.1.

Let us now consider a particle of mass m which rotates around the z-axis at a constant
angle θ and with an instantaneous angular velocity ω as shown in Figure 11.2. In this
situation, the angular momentum vector is perpendicular to the shaded plane in the figure,
which is defined by the position and velocity vectors. As a result, in the example of Figure
11.2 the angular momentum vector ` and the angular velocity vector ω do not point in the
same direction, and the direction of ` changes continuously in space during the rotational
motion.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 338 — #354

i i

338 Classical Mechanics: A Computational Approach

Figure 11.2: A particle of mass m revolves around the z-axis on a plane parallel to the
xy-plane with angular velocity vector ω. In this example, the angular momentum vector `
and angular velocity vectors ω are not pointing in the same direction. Compare this figure
with the situation in Figure 11.1.

In this situation the equation ` = Iω does not apply, because ` and ω do not point in
the same direction in Figure 11.1. We must redefine the moment of inertia concept for the
mass m. As we will see later in this chapter, we need to replace the scalar quantity I in this
equation with a more general physical quantity, the moment of inertia tensor I.

Example 11.2: Example of angular momentum and torque for Figure 11.2
Evaluate the angular momentum and torque for the rotating mass m shown in Figure
11.2, assuming that ω = |ω
ω | is constant.
Solution:
We follow the same method as in the previous example, using cylindrical coordinates.
The Mathematica code appears in Algorithm 11. Clearly in this situation the angular
momentum vector ` has nonzero components along the x-, y-, and z-axes. By inspection
of the calculated components `x and `y , we see that the vector ` rotates around the z-axis
with a constant angular speed Ω, so that the z-component `z = mR2 Ω sin2 θ stays constant
with time.

Before we discuss the rotational properties of solid bodies, let us review the center of
mass, and how it is used to describe the rotation of a system of particles.

11.2 REVIEW OF ROTATIONAL PROPERTIES FOR A SYSTEM

OF PARTICLES
In this section, we summarize some of the results we already saw in Chapters 4 and 5 about
the momentum, angular momentum, and energy of a system of particles. In particular, we
look again at the importance of the center of mass, and how it can help us simplify the
description of the motion of rigid bodies.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 339 — #355

i i

Rigid Body Motion 339

x = R ∗ Sin[θ] ∗ Cos[Ω ∗ t];

y = R ∗ Sin[θ] ∗ Sin[Ω ∗ t];
z = R ∗ Cos[θ];
r = {x, y, z};

v = {D[x, t], D[y, t], D[z, t]}

OUTPUT: {−RΩSin[θ]Sin[tΩ], RΩCos[tΩ]Sin[θ], 0}

ω = Ω ∗ {0, 0, 1};
Cross[ω, r] == v

OUTPUT: True

L = m ∗ Cross[r, v]//Simplify

OUTPUT: −mR2 ΩCos[θ]Cos[tΩ]Sin[θ], −mR2 ΩCos[θ]Sin[θ]Sin[tΩ], mR2 ΩSin[θ]2

Algorithm 11: The Mathematica code for Example 11.2.

11.2.1 The Center of Mass

Let us consider a collection of i = 1, 2, . . . , N discrete particles as in Figure 5.4, each with
mass mi and located at a position ri relative to the origin.

ri'=ri-R

mi
R

x
Figure 11.3: A collection of discrete particles (filled circles), with the empty circle repre-
senting their center of mass located at R.

Let R represent the location of the center of mass relative to the origin, shown as a
white circle in Figure 11.3, and r0i denote the positions of the particles relative to the center
of mass. Then ri = r0i + R and the position of the center of mass can be found from:

The Position of the Center of Mass for N Discrete Particles

N N N
1 X X X
R= mi ri M= mi mi r0i = 0 (11.2.1)
M
i=1 i=1 i=1

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 340 — #356

i i

340 Classical Mechanics: A Computational Approach

In the case of a continuous distribution of mass M , the summations are replaced by

integrals:

The Position of the Center of Mass for a Continuous Mass Distribution

1
Z Z Z
R= rdm M = dm r0 dm = 0 (11.2.2)
M

11.2.2 Momentum of a System of Particles

From Chapter 4, we recall that the total momentum of a system of particles is equal to
the product of the total mass M of the system and the velocity of the center of mass. We
can think of the system of particles as a single particle of mass M located at the system’s
center of mass, and of the net external forces Fext as acting on the center of mass of the
system. Newton’s second law can be written in terms of the motion of the center of mass
as follows:
Total Momentum of a System of Particles or Continuous Mass Distribution
N
X N
X
P= pi = mi ṙi = M Ṙ (11.2.3)
i=1 i=1

Ṗ = M R̈ = Fext (11.2.4)

As a consequence, if the total external force acting on a system of particles is zero, then
the center of mass is either at rest, or it moves with constant velocity and the total linear
momentum of the system is conserved.

11.2.3 Angular Momentum of a System of Particles

From Chapter 4, we also recall that P the total angular momentum of the system is the
sum of two terms. The first term (r0i × p0i ), represents the angular momentum of the
particles about the center of mass, and the second term (R × P), is the angular momentum
of the center of mass about the origin. In this situation Newton’s second law can be stated
as follows: if the internal forces between the particles in a system of particles are central
forces, then the total external torque N on this system is equal to the time rate of change
of the total angular momentum L̇ of the system.

Total Angular Momentum of a System of Particles

N
X
L= r0i × p0i + R × P (11.2.5)

i=1

L̇ = N (11.2.6)

As a consequence, if the total external torque acting on a system of particles is zero, then
the total angular momentum of the system is conserved.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 341 — #357

i i

Rigid Body Motion 341

11.2.4 Work and Kinetic Energy for a System of Particles

In Chapter 5, we saw that the total kinetic energy of a system of particles is also the sum of
two terms. The first term is the translational kinetic energy of the center of mass 12 M Ṙ2 ,
and the second term 0 2
2 mi (ṙi ) , represents the kinetic energy of each particle relative
P1
to the center of mass. We also saw that the total work W12 done in moving a system of
particles from a state of kinetic energy T1 to a state of kinetic energy T2 is equal to the
change in kinetic energy:

Kinetic Energy of a System of Particles

N
1 1

X 2
T = M Ṙ2 + mi ṙ0i (11.2.7)
2 2
i=1

W12 = T2 − T1 (11.2.8)

Furthermore, for a rigid body, the only motion relative to the center of mass is rotation.
Hence, the second term in (11.2.7) is the kinetic energy of rotation about the center of
mass. We can then think of equation (11.2.7) as stating that the total kinetic energy of a
rigid body is equal to the translational kinetic energy of the center of mass plus the kinetic
energy of rotation about the rigid body’s center of mass.
For the potential energy, we need to consider both external and internal forces, and their
corresponding potential energies. The potential energy of the ith mass is:
X
Vi = Viext + int
Vi,j (11.2.9)
j6=i

where Viext is the potential energy due to the external forces acting on mi , and Vi,j int is the

potential energy due to the internal interaction force between mi and mj . An example of
such an interaction force could be gravitational, electrostatic, or any other central force. In
int = V int r − r , i.e., the potential energy depends only on
general for central forces, Vi,j

i,j i j
the distance between the two particles. Therefore the total potential energy of the system
is:
XN N X
X
V = Viext + int
(11.2.10)

Vi,j ri − rj
i=1 i=1 j6=i

Rigid bodies are defined as solids in which the particles that make up the solid are at
fixed distances from each other. Therefore, the internal potential energy can be ignored
since it will be constant for each particle. Hence, for a rigid body we will only have to
worry about external forces; the internal central forces are irrelevant. When all internal and
external forces acting on a system of particles are conservative, the total mechanical energy
E = T + V is conserved.
In Chapters 4 and 5 we saw that the above results for linear momentum, angular
momentum, and kinetic energy hold for continuous mass distributions with the summa-
tions replaced by appropriate integrals.

11.3 THE MOMENT OF INERTIA TENSOR

In this section, we will develop a more general definition of the moment of inertia. Figure
11.4 shows a rigid body rotating around a fixed-axis AB, with an angular velocity vector ω .

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 342 — #358

i i

342 Classical Mechanics: A Computational Approach

We describe this rigid body as a collection of finite masses mi located at positions ri , with
respect to a coordinate system xyz which is fixed on the rotating body. The instantaneous
velocity vi of mass mi is given by the cross product of the angular velocity vector ω =
ωx î + ωy ĵ + ωz k̂ and the positions ri = xi î + yi ĵ + zi k̂:

vi = ω × ri (11.3.1)

Figure 11.4: A rigid body with a fixed axis of rotation. The coordinate system shown is
fixed on the rotating body.

The total angular momentum vector L relative to the center of the coordinate system
is given by the vector sum of the angular momenta of the masses mi :
X X
L= mi (ri × vi ) = mi [ri × (ω × ri )] (11.3.2)
where the summation i = 1 . . . n runs over the masses mi which make up the rigid body.
We now use the identity of the triple cross product A × (B × A) = A2 B − (A · B) A to
evaluate ri × (ω × ri ):
ri × (ω × ri ) = ri2 ω − ri (ri · ω) (11.3.3)

ri × (ω × ri ) = x2i + yi2 + zi2 ωx î + ωy ĵ + ωz k̂ − xi î + yi ĵ + zi k̂ (xi ωx + yi ωy + zi ωz )

(11.3.4)
Substituting (11.3.4) into (11.3.2) and collecting the terms with î, ĵ, and k̂:

L = Lx î + Ly ĵ + Lz k̂ (11.3.5)

L = (Ixx ωx + Ixy ωy + Ixz ωz ) î + (Iyx ωx + Iyy ωy + Iyz ωz ) ĵ + (Izx ωx + Izy ωy + Izz ωz ) k̂

(11.3.6)
where the scalar quantities Ikl (k, l = x, y, or z) are the components of the moment of
inertia tensor I, and are defined by:

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 343 — #359

i i

Rigid Body Motion 343

Moment of Inertia Tensor for Discrete Masses mi


XN XN 
Ixx = mi yi2 + zi2 = mi ri2 − x2i 




i=1 i=1 



XN XN 

Iyy = mi xi + zi =
2 2 2 2
(11.3.7)

mi ri − yi

i=1 i=1 


N N


X X 
Izz = m i xi + yi =
2 2 2 2 

mi ri − zi  

i=1 i=1

N N N
! ! !
X X X
Ixy = Iyx = − mi xi yi Ixz = Izx = − mi xi zi Iyz = Izy = − mi yi zi
i=1 i=1 i=1
(11.3.8)

From these equations it is clear that the moment of inertia matrix I is symmetric, i.e.,
Ixy = Iyx , Ixz = Izx , and Iyz = Izy .
The terms Ixx , Iyy , and Izz are referred to as moments of inertia. These are the moments
of inertia of the rigid body about each axis. The terms Ixy , Iyz , Ixz , . . . are called products of
inertia. As we will see later, the products of inertia are zero if the x, y, and z axes correspond
to the rigid body’s axes of symmetry. The products of inertia measure the symmetry of the
rigid body’s mass distribution about the axes, x, y, and z. To understand the importance
of the product of inertia, consider the following. Suppose you are interested in balancing an
automobile tire. In the case of the tire, the axis of symmetry is the axle passing through the
center of the tire and perpendicular to the plane of the tire. A proper tire rotates about the
axle, or in other words, its angular momentum vector points along the direction of the axle.
Suppose the axle points in the z-direction. We want L = Izz ω k̂ to be the angular momentum
of the wheel. However, suppose the mass of the wheel is not evenly distributed about the
axle, then the axle is no longer an axis of symmetry and there will be at least two nonzero
products of inertia. For simplicity, suppose that Izy 6= 0 and all other products of inertia
are equal to zero. Then in this case, the angular momentum of the wheel is

L = Iyz ωz ĵ + (Izy ωy + Izz ωz ) k̂ (11.3.9)

Because L is no longer parallel to the axle (which points in the z-direction), the wheel will
precess (wobble). Hence automotive technicians will add small weights to a tire to balance
it, such that the wheel’s angular momentum will be parallel to its axis of rotation.
Equations (11.3.7) can also be written as a single equation, as follows:
N 3
!
X X
Ikl = mi δkl 2
uis − uik uil (11.3.10)
i=1 s=1

where we use the compact notation uis (s = 1, 2, 3) to denote the three components of the
position vector ri = (xi , yi , zi ) of the mass mi . For example, u32 is the y-coordinate of
m3 . The symbol δkl in (11.3.10) is the Kronecker delta, which equals 1 when k = l and 0
otherwise.
Equation (11.3.6) can be written as the product of two matrices I and ω:

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 344 — #360

i i

344 Classical Mechanics: A Computational Approach

Angular Momentum and Moment of Inertia Tensor I

L = I·ω (11.3.11)
   
Ixx Ixy Ixz ωx
I = Iyx Iyy Iyz  and ω = ω y  (11.3.12)
Izx Izy Izz ωz

It is important to remember that the elements of the inertia tensor depend on the choice of
origin of the coordinate system. Example 11.3 shows how to calculate the elements of the
moment of inertia tensor for a single particle.

Example 11.3: Moment of inertia tensor for single particle

A particle of mass m is attached to a thin massless rod of length R and rotates around
the z-axis at a constant angle φ as shown in the figure below, with a constant angular
velocity ω . Calculate the moment of inertia tensor with respect to the XY Z coordinate
system. In which direction is the angular momentum vector ` pointing?

Solution:
The instantaneous position of the mass m is best described by cylindrical coordinates,
with the constant radius r = R sin φ, i.e.,

x = r cos θ = R sin φ cos θ y = r sin θ = R sin φ sin θ z = R cos φ.

For a constant angular velocity, the angle θ = ωt.
Substituting these values for x, y, z in the expression for inertia tensor we obtain:

Ixx = m r2 − x2 = m R2 sin2 φ − R2 sin2 φ cos2 θ = mR2 sin2 φ sin2 θ

Similarly

Iyy = m r2 − y 2 = mR2 sin2 φ cos2 θ Izz = m r2 − z 2 = mR2 sin2 φ − cos2 φ

Ixy = −mxy = −mR2 sin2 φ sin θ cos θ Ixz = −mxz = −mR2 sin φ cos φ cos θ
Iyz = −mxz = −mR2 sin φ cos φ sin θ

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 345 — #361

i i

Rigid Body Motion 345

The complete inertia tensor is then:

sin2 φ sin2 θ sin2 φ sin θ cos θ − sin φ cos φ cos θ

   
Ixx Ixy Ixz
I = Iyx Iyy Iyz = mR − sin2 φ sin θ cos θ
  2
sin2 φ cos2 θ − sin φ cos φ sin θ  .
Izx Izy Izz − sin φ cos φ cos θ − sin φ cos φ sin θ sin2 φ − cos2 φ

The angular momentum vector can now be found by multiplying this matrix I with
the angular momentum matrix ω = (0, 0, ω).

` = Iω
After the matrix multiplication and using the trig identities cos2 φ − sin2 φ = cos (2φ)
and 2 cos φ sin φ = sin (2φ), we obtain:

`x = −mR2 ω sin (2φ) cos (ωt) `y = −mR2 ω sin (2φ) sin (ωt) `z = −mR2 ω cos (2φ)

The physical interpretation of these equations is that as the mass m rotates around
the z-axis at the fixed angle φ, the angular momentum vector ` rotates around the z-axis
also, but with a different angle equal to 2φ. The z-component of the angular momentum
vector stays constant in time, and is equal to `z = −mR2 ω cos (2φ).
If we are dealing with a continuous uniform distribution of particles instead of a collection
of discrete masses mi , the summations in the above equations become integrals over the
continuous variables (x, y, z), and the density of the material ρ(x, y, z) must be included
inside the integrals. For solids in three dimensions, the components of the inertia tensor
become:
Inertia Tensor for a Continuous Mass Distribution
ZZZ 
Ixx = y 2 + z 2 ρdxdydz 





V


ZZZ 


Iyy = x + z ρdxdydz
2 2

(11.3.13)
V


ZZZ 


Izz = x + y ρdxdydz 
2 2





V

ZZZ ZZZ ZZZ

Ixy = − xyρdxdydz Ixz = − xzρdxdydz Iyz = − yzρdxdydz
V V V
(11.3.14)

Example 11.4 shows how to evaluate the elements of the inertia tensor for a triangular
pyramid with a variable density, by using numerical integration in Mathematica.
Example 11.4: The moment of inertia of a triangular pyramid
Consider the triangular pyramid with vertices at the origin, (1, 0, 0), (0,1, 0), and (0, 0, 1)
shown in the figure below. The pyramid has a density ρ = y 2 ex sin x2 (note that this
density is not physical since it can have the value of zero). Find the moment of inertia
tensor with respect to the Cartesian coordinate system with the origin at (0, 0, 0).

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 346 — #362

i i

346 Classical Mechanics: A Computational Approach

(0,0,1)

dV
y
(0,1,0)
(1,0,0)

Solution:
In order to find the moment of inertia tensor, we first need to choose a coordinate
system and identify the limits of integration. We will use Cartesian coordinates for the
calculation because it will be easiest to describe the pyramid in Cartesian coordinates.
Next, we need to figure out the limits of integration that we will need to evaluate (11.3.13)
and (11.3.14).
The volume element dV appears in the figure. Notice that its height is determined by
its x and y coordinates. The “front” face of the pyramid in the figure is described by the
equation x + y + z = 1. Therefore, the limits for the z integral will range from 0 to 1 − x − y.
Next, the distance the volume element can be translated along the y-axis in the xy-plane
is determined by its x-coordinate. Therefore, the limits for the y integral will be from 0
to 1 − x (the equation of the line that forms the base of the pyramid in the xy-plane).
Finally, x can range from 0 to 1.
Now that we have all of the pieces, we can perform the integrals. We will evaluate the
integrals in Mathematica using the algorithm below. Notice that the limits of the integral
that we would do last by hand, are the first limits in the NIntegrate command.
Rho[x , y , z ]:=y ∧ 2*Exp[x]*Sin[x];

Ixx = NIntegrate[Rho[x, y, z] ∗ (y ∧ 2 + z ∧ 2), {x, 0, 1}, {y, 0, 1 − x}, {z, 0, 1 − x − y}];

Iyy = NIntegrate[Rho[x, y, z] ∗ (x∧ 2 + z ∧ 2), {x, 0, 1}, {y, 0, 1 − x}, {z, 0, 1 − x − y}];
Izz = NIntegrate[Rho[x, y, z] ∗ (x∧ 2 + y ∧ 2), {x, 0, 1}, {y, 0, 1 − x}, {z, 0, 1 − x − y}];

Ixy = NIntegrate[Rho[x, y, z] ∗ x ∗ y, {x, 0, 1}, {y, 0, 1 − x}, {z, 0, 1 − x, y}];

Iyz = NIntegrate[Rho[x, y, z] ∗ y ∗ z, {x, 0, 1}, {y, 0, 1 − x}, {z, 0, 1 − x, y}];
Ixz = NIntegrate[Rho[x, y, z] ∗ x ∗ z, {x, 0, 1}, {y, 0, 1 − x}, {z, 0, 1 − x, y}];

tensor = {{Ixx, Ixy, Ixz}, {Ixy, Iyy, Iyz}, {Ixz, Iyz, Izz}};
MatrixForm[tensor]

0.00493672 0.00163878 0.000819388

 

OUTPUT:  0.00163878 0.00258742 0.00185127 

0.000819388 0.00185127 0.00505578

11.4 KINETIC ENERGY AND THE INERTIA TENSOR

We now proceed to evaluate the kinetic energy of a rigid body and demonstrate how it
relates to the moment of inertia tensor. The total kinetic energy is the sum of the kinetic

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 347 — #363

i i

Rigid Body Motion 347

energies for the masses mi moving with velocities vi :

X1
1X
T= mi vi · vi = (mi [(ω × ri ) · vi ]) (11.4.1)
2 2
where we used vi = ω × ri , and we have dropped the limits of the summation for simplicity.
Note that we used ω i = ω for all values of i because for a rigid body, each particle has the
same angular velocity. The fact that ω is the same for each particle is due to the fact that
the distances between particles are fixed in a rigid body. We now use the property of the
scalar triple product (A × B) · C = A · (B × C) in order to simplify T :
1X 1 X 1
T= (mi ω· [(ri × vi )]) = ω· mi [(ri × vi )] = ω· L (11.4.2)
2 2 2
where we used the definition of the angular momentum of a system of particles L =
mi [(ri × vi )] from (11.3.2). By writing out the components we obtain:
P

1 1 1 1
T = ω · L= Ixx ωx2 + Iyy ωy2 + Izz ωz2 +Ixy ω x ω y +Iyz ω y ω z +Izx ω z ω x (11.4.3)
2 2 2 2
By using (11.3.11), this equation can be written in a compact form as a matrix equa-
tion:
Kinetic Energy of a Rigid Body
1 1
T = (ω)T · L = (ω)T · I · ω (11.4.4)
2 2

Here the dot product indicates the multiplication of the 3 matrices, and (ω)T indicates the
transpose of the column matrix representing the angular velocity vector ω.
If the angular velocity vector ω points in the direction of the unit vector n̂, we can write
ω = ωn̂, where ω is the magnitude of the angular frequency. Therefore the above matrix
equation becomes:
1
T = ω 2 (n̂)T · I · n̂ (11.4.5)
2
From introductory physics we recall that the kinetic energy for rotation around a fixed axis
is:
1
T = In ω 2 (11.4.6)
2
where In is the scalar moment of inertia around this axis. By comparing the last two
equations, we obtain the general expression:

Moment of Inertia for Rotation Around an Axis in the Unit Vector

Direction n̂

In = (n̂)T · I · n̂ (11.4.7)

The products of inertia can be found using (11.4.7). For example, Ixy is obtained by calcu-
lating the quantity:
T
Ixy = bi · I · bj (11.4.8)

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 348 — #364

i i

348 Classical Mechanics: A Computational Approach

Example 11.5: Moment of inertia of a cube

Find the moment of inertia of a cube when it is rotated around
(a) the main diagonal
(b) a face diagonal on the xy-plane.
(c) the edge that runs along the x-axis.
Solution:
We first evaluate the inertia tensor with respect to a coordinate system centered at one
of the vertices, as in the previous example. Once the tensor is evaluated, the moment of
inertia I is obtained using (11.4.7) and the appropriate unit vector n̂ pointing along the
direction of rotation.
For the main diagonal we have n̂ = √13 î + ĵ + k̂ , for the face diagonal on the xy-plane

n̂ = √12 î + ĵ , and for the edge that runs along the x-axis, n̂ = î.
The Mathematica code below gives I = ma2 /6, I = 2ma2 /3 and I = 5ma2 /12, corre-
spondingly. As we may have expected from geometrical considerations, the largest value
of I is obtained for the face diagonal of the cube.
ρ = m/a∧ 3;
Ixx = Iyy = Izz = Integrate[ρ ∗ (y ∧ 2 + z ∧ 2), {x, 0, a}, {y, 0, a}, {z, 0, a}];
Ixy = Iyz = Ixz = −Integrate[ρ ∗ (x ∗ y), {x, 0, a}, {y, 0, a}, {z, 0, a}];

Ixx Ixy Ixz

 

iCube =  Ixy Iyy Iyz  ;

Ixz Iyz Izz
(*rotate around main diagonal*)
ω = 1/Sqrt[3] ∗ {1, 1, 1};
ω.iCube.ω

a2 m
OUTPUT: 6

(rotate around edge of cube)

ω = {1, 0, 0};
ω.iCube.ω

2a2 m
OUTPUT: 3

(rotate around face diagonal)

ω = 1/Sqrt[2] ∗ {1, 1, 0};
ω.iCube.ω//Simplify

5a2 m
OUTPUT: 12

11.5 INERTIA TENSOR IN DIFFERENT COORDINATE SYSTEMS—THE

PARALLEL AXIS THEOREM
Two important theorems which simplify the calculations of the moments of inertia of a rigid
body, are the parallel axis theorem and the perpendicular axes theorem.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 349 — #365

i i

Rigid Body Motion 349

Figure 11.5: Two coordinate systems with their axes parallel to each other, are shifted in
space relative to each other by a vector a.

In its simplest form, the parallel axis theorem states that if I is the scalar moment of
inertia of a body around an axis AB, and ICM is the corresponding moment of inertia about
a second axis parallel to AB and passing through the center of mass, then I and ICM are
related by:

A Simple Form of the Parallel Axis Theorem

I = ICM + md2 (11.5.1)

where d is the shortest distance between the two parallel axes.

The perpendicular axes theorem refers to a thin lamina mass, which is distributed on the
xy-plane of the Cartesian coordinate system. If Ix , Iy , and Iz are the moments of inertia of
this distribution around the x, y, and z axes respectively, then these are related by:

The Perpendicular Axis Theorem for a Lamina

Iz = Ix + Iy (11.5.2)

More generally, it is convenient and useful to know the relationship between inertia tensors
expressed in different coordinate systems.
We now prove the parallel axis theorem. Figure 11.5 shows two Cartesian coordinate
systems with origins O and OCM and with their axes parallel to each other. Note that OCM
is located at the object’s center of mass. If a is the vector connecting the two origins, the
relationship of any position vectors ri and r0i in the two coordinate system is:

ri = r0i + a (11.5.3)

A very useful tool for describing the motion of rigid bodies is the concept of principal axes
of rotation. This is best described within the general theory of linear algebra, and is closely
related to the concepts of eigenvalues and eigenvectors of matrices.
Our discussion of principal axes of rotation begins with a review of the inertia tensor
itself. First, it is important to remember that the elements of the inertia tensor depend on
the axes and choice of origin for the system. Second, recall that the inertia tensor is a real
symmetric matrix, i.e., Ikl = Ilk (k, l = 1, 2, 3). There is a theorem in linear algebra that
states that any real symmetric matrix can be diagonalized. In other words, the matrix can
be rewritten in a form which has zeroes in all of the off-diagonal elements. Physically, this
means that for any choice of the origin for any rigid body, there always exists a set of axes
in which the inertia tensor is diagonal, i.e., it is of the form:

λ1 0 0
 

I =  0 λ2 0  (11.6.1)
0 0 λ3
This special set of axes is known as the principal axes of the rigid body. According to the
same theorem from linear algebra, there are three principal axes of rotation and they are
mutually orthogonal. In addition, any axis of symmetry through the origin is a principal
axis.
The main property of the principal axes is that when rotation takes place around one
of these three axes, the angular momentum L and angular velocity vectors ω are parallel
to each other. Therefore, if ω points along the direction of a principal axis, then L = λωω
(where λ is a scalar), in order for L and ω to point in the same direction. However, recall
that in general, L = I · ω . Hence:
I · ω = λω (11.6.2)
The scalar λ is called a principal moment of inertia, and it is the moment of inertia of the
object when it is rotating about the principal axis ω . Before discussing principal moments
and principal axes further, it is important to pause and discuss some general properties of
(11.6.2).
The equation I · ω = λω ω is an eigenvalue problem in the theory of linear algebra. In
general, in an eigenvalue problem we are given a square matrix B, and we are looking for
two quantities, a vector a and a scalar quantity λ, such that:

Ba = λa (11.6.3)

The vector a is called an eigenvector of the square matrix B, corresponding to the eigenvalue
λ. Hence, according to (11.6.2), the principal axes are the eigenvectors of the inertia tensor,
and principal moments are the eigenvalues associated with each principal axis.We see from
(11.6.3), that the matrix B scales the vector a by a value λ.

i i

Ixz Iyz Izz
Eigenvalues[iCube]
n 2 2 2
o
OUTPUT: 11a12 m , 11a12 m , a 6m

Eigenvectors[iCube]

OUTPUT: {{−1, 0, 1}, {−1, 1, 0}, {1, 1, 1}}

vec1 = Eigenvectors[iCube][[1]];
vec2 = Eigenvectors[iCube][[2]];
vec3 = Eigenvectors[iCube][[3]];

Graphics3D[{Arrow[{{0, 0, 0}, vec1}], Arrow[{{0, 0, 0}, vec2}],

Arrow[{{0, 0, 0}, vec3}], FaceForm[{Opacity[.2]}], Cuboid[{0, 0, 0},
{1, 1, 1}], Dashed, Line[{{−1, 0, 0}, {1, 0, 0}}],
Line[{{0, −1, 0}, {0, 1, 0}}], Line[{{0, 0, −1}, {0, 0, 1}}]},
Axes → True, BaseStyle → 14, AxesLabel → {x, y, z},
PlotRange → {{−1, 1}, {−1, 1}, {−1, 1}}]

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 354 — #370

i i

354 Classical Mechanics: A Computational Approach

Example 11.7: Inertia tensor of a cube with respect to center of mass

Find the inertia tensor and the principal axes for a cube with uniform density, mass
m, and side a, with respect to the center of mass of the cube.
Solution
The Mathematica code below calculates the eigenvalues and plots the eigenvectors of
the moment of inertia matrix with respect to the center of mass of the cube. Note that we
continue to use a = 1 in our code for plotting purposes.
The code gives I1 = I2 = I3 = ma2 /6 and the three eigenvectors are the unit vectors
(1, 0, 0), (0, 1, 0) and (0, 0, 1). The principal moments of inertia with respect to the center
of the cube are equal to each other, and the corresponding moment of inertia matrix is
diagonal. This means that if the cube is rotated around any axis through its center, the
moment of inertia will always be the same and equal to I = ma2 /6.
ρ = m/a∧ 3;
Ixx = Iyy = Izz = Integrate[ρ ∗ (y ∧ 2 + z ∧ 2), {x, −a/2, a/2}, {y, −a/2, a/2}, {z, −a/2, a/2}]

a2 m
OUTPUT: 6

Ixy = Iyz = Ixz = −Integrate[ρ ∗ (x ∗ y), {x, −a/2, a/2}, {y, −a/2, a/2}, {z, −a/2, a/2}]

OUTPUT: 0

Ixx Ixy Ixz

 

iCube =  Ixy Iyy Iyz  ;

Ixz Iyz Izz
iCube//MatrixForm
 2 
a m
6 0 0
OUTPUT:  a2 m
0 0
 
6

2
0 0 a m
6
Eigenvalues[iCube]
n 2 2 2
o
OUTPUT: a 6m , a 6m , a 6m

Eigenvectors[iCube]

OUTPUT: {{0, 0, 1}, {0, 1, 0}, {1, 0, 0}}

vec1 = Eigenvectors[iCube][[1]];
vec2 = Eigenvectors[iCube][[2]];
vec3 = Eigenvectors[iCube][[3]];

Graphics3D[{Arrow[{{0, 0, 0}, vec1}], Arrow[{{0, 0, 0}, vec2}],

Arrow[{{0, 0, 0}, vec3}], FaceForm[{Opacity[.2]}], Cuboid[{−0.5, −0.5, −0.5},
{0.5, 0.5, 0.5}], Dashed, Line[{{−1, 0, 0}, {1, 0, 0}}],
Line[{{0, −1, 0}, {0, 1, 0}}], Line[{{0, 0, −1}, {0, 0, 1}}]},
Axes → True, BaseStyle → 14, AxesLabel → {x, y, z},
PlotRange → {{−1, 1}, {−1, 1}, {−1, 1}}]

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 355 — #371

i i

Rigid Body Motion 355

11.7 PRECESSION OF A SYMMETRIC SPINNING TOP WITH ONE POINT

FIXED AND EXPERIENCING A WEAK TORQUE
The motion of spinning tops is a fascinating example of the rotational motion of rigid bodies
around an axis, and has important practical applications in gyroscopes. It will also serve
as an example of how the concept of principal axes can be applied to solve problems in
rotational mechanics.
Figure 11.6 shows a symmetric spinning top of mass m, whose vertex is fixed at the
origin, placed on a frictionless surface, with its rotational axis at an angle θ with respect to
the vertical. If the top is spinning rapidly, the top will not fall as we might expect due to
the effect of gravity, but its symmetry axis describes a cone in space at the constant angle
θ, as shown in Figure 11.6. This type of motion is called precession and is due to the torque
acting on the fast spinning object.

dL⟂

r CM

mg
x

y
Figure 11.6: The spinning top precesses at a fixed angle θ due to the gravitational torque.
The top is rotating about its symmetry axis which points in the same direction as L.

i i

Finally by combining (11.7.8) and (11.7.7), we obtain:

rmg
ΩP = (11.7.9)
Iω
According to this expression, the precessional frequency for a fast spinning top precessing
at a constant angle θ will depend on these factors: the weight of the top mg, the distance
of the center of mass r from the contact point, the spinning angular frequency ω, and the
moment of inertia I around the symmetry axis.
As discussed in Chapter 10, the Earth has an equatorial bulge, therefore, it experiences
a gravitational torque due to the Sun and the Moon. These torques cause Earth’s axis of
rotation to precess slowly, completing one revolution every 26,000 years. The phenomenon
is called the precession of the equinoxes. One of the results is that Polaris is not always
the North Star, and the star that the Earth’s North Pole points towards changes as the
Earth precesses. Another interesting effect of precession is that precession of the equinoxes
changes our astrological signs! Traditionally, an astrological sign is assigned to a person
based on the zodiac constellation the Sun was in on the day of said person’s birth. However,
those assignments were made millennia ago. Since then, the Earth’s axis has precessed and
the Sun was in a different zodiac constellation when said person was born. For example, a
Libra is born between September 23 and October 22. However, during that time period, the
Sun was actually in the constellation Virgo, not Libra. In general, each person’s astrological
sign is one earlier than the one traditionally assigned. While we are speaking of assigning
astrological signs, there is another constellation along the ecliptic, Ophiuchus, which was
never accounted for in astrology, so someone might actually be an Ophiuchus and not know
it! Someone born between November 29 and December 17 is an Ophiuchus. Of course, the
location of the Sun on a person’s birthday has no effect on the individual; that is well-
proven. However, if nothing else, the effects of precession on astrological signs may provide
for interesting cocktail conversations!

11.8 RIGID BODY MOTION IN THREE DIMENSIONS AND EULER’S

EQUATIONS
In this section, we will derive the equations of motion for a rotating rigid body in the form
known as Euler’s equations.
Let us consider a rigid body which rotates about an axis AB that passes through a
fixed point. An example of this could be a spinning top shown in Figure 11.6. In order to
specify the motion of the rigid body in space, we need 6 coordinates, so there are 6 degrees
of freedom for this system. These coordinates are usually taken to be the 3 coordinates of
the center of mass of the body, plus three more coordinates which are usually taken to be
angular velocities expressing the rotation around the center of mass. If there is a fixed point
on the rigid body, then we need only three coordinates to describe the rotational motion.
Since the rigid body is rotating, the principal axes are fixed on the body and therefore
also rotate in space, so they are not stationary with respect to the inertial (fixed) X 0 Y 0 Z 0 -
frame shown in Figure 11.7. This means that when we use the principal axes to describe
the motion, we are by necessity dealing with non-inertial rotating reference frames, and we
can use the techniques from Chapter 10 to describe the motion of the rotating body.
The specific non-inertial frame defined by the principal axes of the rotating body is
called the body coordinate frame, which is the XY Z coordinate system in Figure 11.7, and
is usually taken to be the principal axes of the object. The inertial X 0 Y 0 Z 0 -frame is called
the space coordinate frame.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 358 — #374

i i

358 Classical Mechanics: A Computational Approach

Figure 11.7: The motion of a rotating rigid body rotating about the axis AB can be described
in two coordinate systems. The space coordinate frame X 0 Y 0 Z 0 is an inertial coordinate
system. The body coordinate frame XY Z is fixed on, and rotates with the rigid body, and
is therefore a non-inertial frame. The body coordinate frame usually corresponds to the
principal axes of the rotating rigid body.

Newton’s second law in an inertial frame of reference says that the time derivative of
the angular momentum L in the inertial frame equals the applied torque N:

dL d
= (Ispace ω) = N (11.8.1)
dt space dt
where Ispace is the moment of inertia tensor calculated in the space (inertial) frame, and ω =
(ω1 , ω2 , ω3 ) is the angular velocity. This equation does not lend itself easily to calculations
because both Ispace and ω change during the motion. However, the calculation can be
simplified using the body frame.
Let ê1 , ê2 , and ê3 be the unit vectors of the body frame pointing along the principal
axes of the rotating body, i.e., the axes X,Y, and Z in Figure 11.7. In the body frame, the
moment of inertia tensor is constant and diagonal:

I1 0 0
 

I =  0 I2 0  (11.8.2)
0 0 I3
where I1 , I2 ,and I3 are the object’s principal moments of inertia. The angular momentum
in the body frame L can be written as:

L = L1 ê1 + L2 ê2 + L3 ê3 = I1 ω1 ê1 + I2 ω2 ê2 + I3 ω3 ê3 (11.8.3)

where ωk are the components of the angular velocity.
In Chapter 10, we saw that in a rotating reference frame, the time derivative in (11.8.1)
must be replaced with the following time derivative in the body frame:

dL dL
= +ω×L = N (11.8.4)
dt space dt body

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 359 — #375

i i

Rigid Body Motion 359

By substituting L = Iω in this equation, and since the inertia tensor and the principal
moments Ik do not depend on time, we have:

dL
= Iω̇
ω (11.8.5)
dt body
Therefore, we arrive at the following vector form of Euler’s equations:

Iω̇ + ω × (Iω) = N (11.8.6)

For the principal axes coordinates, we have Lk = Ik ωk . Finally, by writing out the com-
ponents of the cross product in (11.8.6), the components of this vector equation become:

Euler’s Equations

I1 ω̇1 + (I3 − I2 )ω2 ω3 = N1

I2 ω̇2 + (I1 − I3 )ω3 ω1 = N2 (11.8.7)
I3 ω̇3 + (I2 − I1 )ω1 ω2 = N3

where Nk (k = 1, 2, 3) are the components of the applied torque. This system of three cou-
pled differential equations for (ω1 , ω2 , ω3 ) is known as Euler’s equations for a rigid body. In
special cases they can be solved analytically, however in most cases they must be integrated
numerically, in order to obtain the time dependent functions (ω1 (t) , ω2 (t) , ω3 (t)).

11.9 THE FORCE-FREE SYMMETRIC TOP

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 362 — #378

i i

362 Classical Mechanics: A Computational Approach

with an angular velocity vector ω = (ωx , ωy , ωz ). The tangential velocity vi of mass mi

is given by the cross product:
vi = ω × ri
The angular momentum vector L of a system of particles with masses mi relative to the
center of the mass coordinate system is given by the vector sum of the angular momenta
of the masses:
X X
L= mi (ri × vi ) = mi [ri × (ω × ri )]
In cases where the vectors L and ω do not point in the same direction, the equation L = Iω
does not apply, and we replace the scalar quantity I with the symmetric inertia tensor I
defined by:

L = I·ω
   
Ixx Ixy Ixz ωx
I = Iyx Iyy Iyz  and ω = ω y 
Izx Izy Izz ωz
where the scalars Ikl (k, l = 1, 2, 3) are the components of the inertia tensor I, defined by:
X X X
Ixx = mi yi2 + zi2 = mi ri2 − x2i Ixy = Iyx = − etc.

mi xi yi

The elements of the inertia tensor depend on the choice of origin for the system. For
a mass m with density ρ these elements are given by:
ZZZ ZZZ
Ixx = ρ y + z dxdydz
2 2
Ixy = Iyx = − etc.

ρxydxdydz
V V

The kinetic energy and moment of inertia tensor are related by the matrix multiplication
equation:
1 1
T = ω· L = (ω)T ·I · ω
2 2
where (ω)T indicates the transpose of the column matrix ω.
The scalar moment of inertia I for rotation around an axis in the direction n̂ of any
unit vector, is found from the inertia tensor using the matrix multiplication equation:

In = (n̂)T · I · n̂

The parallel axis theorem for scalar moments of inertia states: if I is the scalar moment
of inertia of a body around an axis AB, and ICM is the corresponding moment of inertia
about a second axis parallel to AB and passing through the center of mass, then I and
ICM are related by:
I = ICM + md2
The generalized parallel axis theorem between moment of inertia tensors is as follows: If I
is the inertia tensor defined in a coordinate system with the origin fixed at point O, and
I0 is the inertia tensor defined in a center-of mass coordinate system with its origin at the

where I1 is the moment of inertia of the attached object around the rotational axis,
and A is the torsion coefficient describing the stiffness of the wire. Consider the torsion
pendulum shown in Figure 11.12, which consists of a thin disk of radius R and mass
M with a cylindrical mass M of radius R/4 placed on top of the disk. Find the ratio of
the periods of this torsion pendulum, with and without the presence of the cylindrical
mass on top of the thin disk.

Figure 11.12: Problem 11.6: A torsion pendulum.

7. A bowling ball of uniform density is thrown along a horizontal alley with initial velocity
vo in such a way that it initially slides without rolling. The ball has mass m, the
coefficient of static friction with the floor is µs , and the coefficient of sliding friction
with the floor is µk . Compute how far the ball will slide before it starts rolling. You
can ignore the effect of air resistance.
8. A coin is spinning about its axis of symmetry through its center with angular frequency
ω as shown in Figure 11.13A below. The coin is placed down on a horizontal surface
as shown in Figure 11.13B. The coin stops slipping and starts rolling away. What is
the velocity of the coin when it rolls away?

(A) (B)

Figure 11.13: Problem 11.8: A spinning coin.

9. A wheel of mass M and radius R is projected along a horizontal surface with an initial
linear velocity vo and an initial angular velocity ωo , as shown in Figure 11.14. The
wheel starts sliding along the surface. Let the coefficient of friction between the wheel
and the surface be µ.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 366 — #382

i i

366 Classical Mechanics: A Computational Approach

(a) How long does it take for the wheel to stop sliding?
(b) What is the velocity of the center of mass of the wheel when the slipping stops?

Figure 11.14: Problem 11.9: A wheel projected along a horizontal surface.

10. A solid uniform cylinder of mass m and radius R is placed on a plane inclined at angle
θ with respect to the horizontal, as shown in Figure 11.15. Let a be the acceleration
of the axis of the cylinder along the incline. The coefficient of friction between the
cylinder and plane is µ. For θ less than some critical angle θc , the cylinder will roll
down the incline without slipping.
(a) What is the critical angle θc ?
(b) What is the acceleration a for angles less than the critical angle?

Figure 11.15: Problem 11.10: A cylinder rolling down an inclined plane.

11. A wheel of radius R and moment of inertia I is mounted on a frictionless axle through
its center. A flexible, weightless cord is wrapped around the rim of the wheel and
carries a body of mass M which begins descending as shown in Figure 11.16. What is
the tension in the cord?

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 367 — #383

i i

Rigid Body Motion 367

Figure 11.16: Problem 11.11: A mass hanging from a wheel.

12. A thin uniform stick of mass m is resting with its bottom end on a frictionless table.
The stick is released from rest at an angle θo with respect to the vertical, as shown in
Figure 11.17. Find the force exerted by the table upon the stick at an infinitesimally
small time after its release.

Figure 11.17: Problem 11.12: A thin uniform stick in the process of falling over.

Section 11.2: Review of Rotational Properties for a System of Particles

i i

372 Classical Mechanics: A Computational Approach

27. A torsion pendulum consists of a vertical wire attached to a cube, which may rotate
about the vertical. The cube is hung from a corner, then is hung from midway along
an edge, and finally it is hung from the middle of a face. Which one of the three
periods of oscillation of the cube is largest?

Section 11.7–11.8: Rigid Body Motion in 3D and Euler’s Equations—The

Motion of a Spinning Top-Precession

28. Consider a rotating object which is not experiencing any external torques. Using
Euler’s equations, show that the magnitude of the angular momentum is constant.
Also show that the rotational kinetic energy is constant.
29. A thin rectangular plate has dimensions a, and b. Find the torque required to rotate
this plate with constant angular velocity ω around one of its diagonals.
30. A uniform rigid wheel is located on the xy-plane, and has principal moments of inertia
I1 = I2 6= I3 about its body-fixed principal axes x̂1 , x̂2 and x̂3 as shown in Figure 11.25.
The wheel is attached at its center of mass to a bearing which allows frictionless
rotation about one space-fixed axis. The wheel is “dynamically balanced,” i.e., it can
rotate at constant ω, so that it can exert no torque on its bearing.
(a) Examine and integrate the Euler equations.
(b) What conditions must the components of ω satisfy for this dynamically balanced
system? Discuss the motion

Figure 11.25: Problem 11.30: A uniform rigid wheel rotating with along the axis ω .

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 373 — #389

i i

CHAPTER 12

Coupled Oscillations

In this chapter, we will explore the properties of coupled harmonic oscillators. These systems
can be analyzed by using either the Lagrangian approach of Chapter 8, or alternatively
using Newton’s second law. The simplest form of these systems in mechanics contains two
masses connected by springs to each other. A second simple example of coupled mechanical
oscillators is the double pendulum, which also exhibits a wide range of interesting behaviors.
We will see that these simple oscillating systems can exhibit normal modes of oscillation,
which are patterns of motion in which all parts of the system move sinusoidally with the
same frequency. The frequencies of the normal modes of a system are known as its natural
frequencies of oscillation. We will find that any motion exhibited by the system can be
expressed as a linear combination of these normal modes.
The discussion of the two-mass system will lead us to a more general description of
linearly coupled harmonic systems, and how their equations of motion can be written in
matrix form. The best way to obtain solutions to the equations of motions for coupled oscil-
lations is by using standard techniques from linear algebra, in order to find the eigenvalues
and eigenvectors of a matrix. The eigenvectors and eigenvalues of the matrix characterizing
the oscillating system are closely related to its normal modes.
This chapter will conclude with a general treatment of coupled oscillations and a dis-
cussion of normal coordinates.

12.1 COUPLED OSCILLATIONS OF A TWO-MASS THREE-SPRING SYSTEM

12.1.1 The Equations of Motion—Numerical Solution
Consider two masses m1 and m2 attached to three springs with spring constants
k1 , k2 , and k3 and to the two fixed walls, as shown in Figure 12.1.

Figure 12.1: A system of two coupled harmonic oscillators consisting of two masses m1 and
m2 connected with three springs with constants k1 , k2 , and k3 .

373

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 374 — #390

i i

374 Classical Mechanics: A Computational Approach

Let us denote by x1 (t) and x2 (t) the horizontal displacements of the two masses from
their respective equilibrium points. The force on the first mass due to the first spring is
−k1 x1 . The middle spring will be stretched by a distance (x1 − x2 ), and the force on the
first mass due to this middle spring will be −k2 (x1 − x2 ). The total force on the first mass
must then be F1 = −k1 x1 − k2 (x1 − x2 ). Similarly, the force on the second mass due to the
middle spring is −k2 (x2 − x1 ), and the force on the second mass due to this third spring
will be −k3 x2 .
Using the notation ẍ for acceleration, the equations of motion from Newton’s second
law F = ma for the two masses are:
)
m1 ẍ1 = −k1 x1 − k2 (x1 − x2 )
(12.1.1)
m2 ẍ2 = −k2 (x2 − x1 ) − k3 x2
In a more general way, we can obtain the same equations by starting with the Lagrangian
formulation of Chapter 8. The potential energies of the two end springs are V1 = k1 x21 /2
and V3 = k3 x22 /2, while the potential energy for the middle spring is V2 = k2 (x1 − x2 )2 /2,
so that the Lagrangian is equal to:
1 1 1 1 1
L = T − VT otal = m1 x˙1 2 + m2 x˙2 2 − k1 x21 − k3 x22 − k2 (x1 − x2 )2 (12.1.2)
2 2 2 2 2
The Euler-Lagrange equations are:

d ∂L ∂L
− =0 → m1 ẍ1 = −k1 x1 − k2 (x1 − x2 ) (12.1.3)
dt ∂ ẋ1 ∂x1

d ∂L ∂L
− =0 → m2 ẍ2 = −k2 (x2 − x1 ) − k3 x2 (12.1.4)
dt ∂ ẋ2 ∂x2
These are of course the same equations as in (12.1.1).
In general, it is not possible to obtain the solutions x1 (t) and x2 (t) of the system of
equations (12.1.1) analytically, and they must be obtained by numerically integrating the
equations for given initial conditions of the system. The initial conditions are usually given
as the initial positions and initial speeds of the two masses.
Example 12.1 shows how to obtain and plot the numerical solutions x1 (t) and x2 (t),
with the initial conditions x1 (0) = 0, x2 (0) = a, ẋ1 (0) = a, and ẋ2 (0) = 0.

Example 12.1: Numerical solution for the general case of two coupled oscil-
lating masses
Integrate (12.1.1) for k1 = 1 N/m, k2 = 2 N/m, k3 = 3 N/m, m1 = 1 kg, m2 = 2 kg and
plot the numerical solutions x1 (t) and x2 (t), with the initial conditions x1 (0) = 0, x2 (0) =
1m, ẋ1 (0) = 0 and ẋ2 (0) = 0. This situation corresponds to the case where the first mass m1
is initially at rest at its equilibrium position (x1 (0) = 0 and ẋ1 (0) = 0), and the second mass
is pulled a distance a = 1 meter from its equilibrium and released from rest (ẋ2 (0) = 0).
Solution:
In this example we can use either the NDSolve command in Mathematica to numerically
solve the differential equations, or alternatively use the DSolve command to obtain the
analytical solutions x1 (t) and x2 (t). Although Mathematica can obtain analytical expres-
sions for x1 (t) and x2 (t), they are not listed here because they are algebraically very
complex, and therefore the output is suppressed in this example by using the semicolon
(;) at the end of the command line.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 375 — #391

i i

Coupled Oscillations 375

The parameter numValues in the code contains the numerical values for the parameters
m1 , m2 , k1 , k2 , and k3 in the form of a rule (→). These numerical values are needed in
order to plot the solutions using the Plot and GraphicsGrid commands.
sol = DSolve[{m1 ∗ x1”[t] == −k1 ∗ x1[t] + k2 ∗ (x2[t] − x1[t]), m2 ∗ x2”[t] == −k3 ∗ x2[t] +
k2 ∗ (x1[t] − x2[t]), x1[0] == 0, x10 [0] == 0, x2[0] == a, x20 [0] == 0}, {x1[t], x2[t]}, t];

numValues = {a → 1, k1 → 1, k2 → 2, k3 → 3, m1 → 1, m2 → 2};

gr1 = Plot[x1[t]/.sol/.numValues, {t, 0, 20}, PlotLabel → “(A) x1[t]”];

gr2 = Plot[x2[t]/.sol/.numValues, {t, 0, 20}, PlotStyle → Dashed,
PlotLabel → “(B) x2[t]”];

GraphicsGrid[{{gr1, gr2}}, Frame → True, ImageSize → Large]

(A) x1[t] (B) x2[t]

1.5 1.0
1.0
0.5
Out[ ]=
0.5

5 10 15 20 5 10 15 20
-0.5
-0.5
-1.0
-1.5 -1.0

The plots of x1 (t) and x2 (t) in Example 12.1 are obviously complex, and it is not
possible to give a simple physical description of the motion of the two masses. The key
physical component which creates this complex behavior is the middle spring in Figure
12.1, since this is the component that couples the motion of the two masses.

12.1.2 Equal Masses and Identical Springs: The Normal Modes

In Examples 12.2 and 12.3, we look at the special case of equal masses (m1 = m2 ), and
springs with equal spring constants (k1 = k2 = k3 ). This is an interesting physical situation,
in which the analytical solutions are simple, and it may be easy to understand the physics
of the situation. These two examples introduce the concept of normal modes in a simple
and clear manner.
Example 12.2: Equal masses and identical springs: The symmetric oscillation
mode
Integrate (12.1.1) and plot the numerical solutions x1 (t) and x2 (t) for the special case
of identical masses and identical springs, m1 = m2 and k1 = k2 = k3 . Consider the physical
situation with the initial conditions x1 (0) = a, x2 (0) = a, ẋ1 (0) = 0 and ẋ2 (0) = 0. Plot
the solutions with Mathematica, by using the numerical values a = 1.0 m, k = 1.0 N/m,
m = 1.0 kg.
Solution:
This situation corresponds to the case where the two masses are pulled the same
distance a from their corresponding equilibrium, and are released from rest (ẋ1 (0) = 0

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 376 — #392

i i

376 Classical Mechanics: A Computational Approach

and ẋ2 (0) = 0). The Mathematica code uses the DSolve command to obtain analytical
expression for x1 (t) and x2 (t), and the ExpToTrig command is used to convert the solutions
x1 (t) and x2 (t) into a form containing trigonometric functions, instead of exponential
functions. The result of DSolve appears above the graph. p
The analytical solutions are x1 (t) = x2 (t) = a cos k/m t . This tells us that if the
two masses are initially displaced from equilibrium by the same distance and released
from rest, the two masses will move together with the same speed and in phase as if the
middle spring was not present. This makes physical sense, since in this situation the middle
spring will be unstretched from its natural length, and will remain unstretched during the
motionp of the two masses. The frequency of oscillation for both masses in this situation is
ω1 = k/m, i.e., the same frequency as if only one of the two masses were attached to a
single spring with a spring constant k.
sol = ExpToTrig[DSolve[{m ∗ x1”[t] == −k ∗ x1[t] + k ∗ (x2[t] − x1[t]), m ∗ x2”[t] ==
−k ∗ x2[t] + k ∗ (x1[t] − x2[t]), x1[0] == a, x10 [0] == 0, x2[0] == a, x20 [0] == 0},
{x1[t], x2[t]}, t]]//Simplify

numValues = {a → 1, k → 1, m → 1};

gr1 = Plot[x1[t]/.sol/.numValues, {t, 0, 10}, PlotLabel → “x1[t]”];

gr2 = Plot[x2[t]/.sol/.numValues, {t, 0, 10}, PlotLabel → “x2[t]”, PlotStyle → Dashed];

GraphicsGrid[{{gr1, gr2}}, Frame → True, ImageSize → Large]

OUTPUT:
nn h√ i h √ ioo
x1[t] → aCos √kt
m
, x2[t] → aCos √kt
m

x1[t] x2[t]
1.0 1.0

0.5 0.5
Out[ ]=

2 4 6 8 10 2 4 6 8 10
-0.5 -0.5

-1.0 -1.0

Example 12.3: Equal masses and identical springs: The antisymmetric oscil-
lation
Repeat Example 12.2, by using a different set of initial conditions x1 (0) = a, x2 (0) = −a,
ẋ1 (0) = 0 and ẋ2 (0) = 0. Plot the solutions with Mathematica by using the numerical values
a = 1.0 m, k = 1.0 N/m, m = 1.0 kg.
Solution:
In this situation, the two masses are initially displaced from their equilibrium positions
by equal and opposite distances a, and are then released from rest.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 377 — #393

i i

Coupled Oscillations 377

The analytical solutions(appearing above the graph) in this case from Mathematica
are x1 (t) = a cos 3k/m t and x2 (t) = −a cos 3k/m t . This tells us that the two
p p

masses will move together with the same speed, but they will be completely out of phase
as shown in the output of the code. The frequency of oscillation for both masses in this
situation is ω2 = 3k/m.
p

sol = ExpToTrig[DSolve[{m ∗ x1”[t] == −k ∗ x1[t] + k ∗ (x2[t] − x1[t]), m ∗ x2”[t] ==

−k ∗ x2[t] + k ∗ (x1[t] − x2[t]), x1[0] == a, x10 [0] == 0, x2[0] == −a, x20 [0] == 0},
{x1[t], x2[t]}, t]]//Simplify

numValues = {a → 1, k → 1, m → 1};

gr1 = Plot[x1[t]/.sol/.numValues, {t, 0, 10}, PlotLabel → “x1[t]”];

gr2 = Plot[x2[t]/.sol/.numValues, {t, 0, 10}, PlotLabel → “x2[t]”, PlotStyle → Dashed];

GraphicsGrid[{{gr1, gr2}}, Frame → True, ImageSize → Large]

OUTPUT:
nn h√ √ i h √ √ ioo
x1[t] → aCos √3 kt
m
, x2[t] → −aCos 3 kt
√
m

x1[t] x2[t]
1.0 1.0

0.5 0.5
Out[ ]=

2 4 6 8 10 2 4 6 8 10
-0.5 -0.5

-1.0 -1.0

Examples 12.2 and 12.3 show that the system of two equal masses and three p
identical
springs in Figure 12.1 has two natural frequencies given by ω1 = k/m and ω2 = 3k/m.
p

By properly choosing the initial conditions in the system as in the two examples, we can
force both masses to oscillate with a single frequency, either ω1 or ω2 . In these special
situations, the two natural frequencies are uncoupled from each other, and we say that
these are the normal modes of the oscillating system.
Figure 12.2 shows schematically thep motion of the two masses in either the symmetric
oscillation pattern with frequency ω 1 = k/m (left panel), or an antisymmetric oscillation
with frequency ω2 = 3k/m (right panel).
p

i i

Coupled Oscillations 379

sol = ExpToTrig[DSolve[{m ∗ x1”[t] == −k ∗ x1[t] + k ∗ (x2[t] − x1[t]), m ∗ x2”[t] ==

−k ∗ x2[t] + k ∗ (x1[t] − x2[t]), x1[0] == 0, x10 [0] == 0, x2[0] == a, x20 [0] ==
0}, {x1[t], x2[t]}, t]]//Simplify

numValues = {a → 1, k → 1, m → 1};

gr1 = Plot[x1[t]/.sol/.numValues, {t, 0, 10}, BaseStyle → {Bold, FontSize → 14}, PlotLabel

→ “(A) x1[t]”];
gr2 = Plot[x2[t]/.sol/.numValues, {t, 0, 10}, BaseStyle → {Bold, FontSize → 14}, PlotStyle
→ Dashed, PlotLabel → “(B) x2[t]”];
gr3 = Plot[(x1[t] + x2[t])/.sol/.numValues, {t, 0, 10}, BaseStyle → {Bold, FontSize → 14},
PlotLabel → “(C) x1[t]+x2[t]”];
gr4 = Plot[(x1[t] − x2[t])/.sol/.numValues, {t, 0, 10}, BaseStyle → {Bold, FontSize → 14},
PlotStyle → Dashed,
PlotLabel → “(D) x1[t]-x2[t]”];

test = GraphicsGrid[{{gr1, gr3}, {gr2, gr4}}, Frame → True, ImageSize → Large]

OUTPUT:
nn h√ i h √ √ i h√ i h √ √ ioo
x1[t] → 12 a Cos √kt
m
− Cos 3 kt
√
m
, x2[t] → 1
2 a Cos √kt + Cos
m
3 kt
√
m

(A) x1[t] (C) x1[t]+x2[t]

1.0
0.5 0.5

2 4 6 8 10 2 4 6 8 10
-0.5 -0.5

-1.0 -1.0

(B) x2[t] (D) x1[t]-x2[t]

1.0 1.0
0.5 0.5

2 4 6 8 10 2 4 6 8 10
-0.5 -0.5

-1.0 -1.0

In Example 12.5, we examine one more interesting behavior of the two-mass system,
the case of weakly coupled oscillators. In this example, the weak coupling is established by
choosing a middle spring with smaller spring constant than the two end springs (k1 = k3 = k
and k2 << k).

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 380 — #396

i i

380 Classical Mechanics: A Computational Approach

Example 12.5: Weakly coupled oscillators

Integrate (12.1.1) and plot the numerical solutions x1 (t) and x2 (t) for the special case of
identical masses (m1 = m2 = 1 kg), identical end springs (k1 = k3 = 1 N/m), and a smaller
middle spring constant (k2 = 0.2 N/m). Consider the physical situation with the initial
conditions x1 (0) = 1 m, x2 (0) = 0, ẋ1 (0) = 0, and ẋ2 (0) = 0.
Solution:
The Mathematica code below shows the solutions x1 (t) and x2 (t) in panels (A) and
(B). The motion of the two masses shows a clear beat pattern, in which mass m1 reaches a
maximum amplitude of oscillation at the same time that m2 reaches a minimum amplitude,
and vice versa.
In this example, it is again possible to uncouple the motions of the two masses, by
plotting the sum x1 (t)+x2 (t) and the difference x1 (t)−x2 (t) of the two functions x1 (t) and
x2 (t). This is shown in panels (C) and (D) of the figure, where the two linear combinations
x1 (t) + x2 (t) and x1 (t) − x2 (t) can be seen to have pure harmonic oscillations with different
frequencies, corresponding to the two normal modes of the system. We will see a more
general method of uncoupling the normal modes of the system later in this chapter, when
we discuss the concept of normal coordinates.
sol = ExpToTrig[DSolve[{m ∗ x1”[t] == −k1 ∗ x1[t] + k2 ∗ (x2[t] − x1[t]), m ∗ x2”[t] ==
− k3 ∗ x2[t] + k2 ∗ (x1[t] − x2[t]), x1[0] == a, x10 [0] == 0, x2[0] == 0, x20 [0] == 0}, {x1[t],
x2[t]}, t]]//Simplify;

numValues = {a → 1, k1 → 1, k2 → .2, k3 → 1, m → 1};

gr1 = Plot[x1[t]/.sol/.numValues, {t, 0, 90}, BaseStyle → {Bold, FontSize → 14}, PlotLabel

→ “(A) x1[t]”];
gr2 = Plot[x2[t]/.sol/.numValues, {t, 0, 90}, BaseStyle → {Bold, FontSize → 14}, PlotStyle
→ Dashed, PlotLabel → “(B) x2[t]”];
gr3 = Plot[(x1[t] + x2[t])/.sol/.numValues, {t, 0, 90},
BaseStyle → {Bold, FontSize → 14}, PlotLabel → “(C) x1[t]+x2[t]”];
gr4 = Plot[(x1[t] − x2[t])/.sol/.numValues, {t, 0, 90},
BaseStyle → {Bold, FontSize → 14}, PlotStyle → Dashed,
PlotLabel → “(D) x1[t]-x2[t]”];

GraphicsGrid[{{gr1, gr3}, {gr2, gr4}}, Frame → True, ImageSize → Large]

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 381 — #397

i i

Coupled Oscillations 381

(A) x1[t] (C) x1[t]+x2[t]

1.0 1.0

0.5 0.5

20 40 60 80 20 40 60 80
-0.5 -0.5

-1.0 -1.0
Out[ ]=

(B) x2[t] (D) x1[t]-x2[t]

1.0 1.0

0.5 0.5

20 40 60 80 20 40 60 80
-0.5 -0.5

-1.0 -1.0

In the next two sections, we develop a more formal mathematical analysis of the normal
modes for the system in Figure 12.1 by using the techniques of linear algebra.

12.2 NORMAL MODE ANALYSIS OF THE TWO-MASS THREE-SPRING

SYSTEM
We now proceed to analyze the system of equations (12.1.1) in two different cases. In
Subsection 12.2.1, we show how to obtain the analytical solution for the two-mass three-
spring system by using the standard matrix techniques of linear algebra. In Subsection
12.2.2, we show how to solve the same problem by turning it into an eigenvalue/eigenvector
type of problem, which can be easily analyzed using the commands available in Mathematica
and Python.

12.2.1 Equal Masses and Identical Springs—Analytical Solution

In the case of equal masses m1 = m2 = m, and identical spring constants k1 = k2 = k3 = k,
the equations of motion (12.1.1) become:

mẍ1 = −kx1 − k(x1 − x2 ) mẍ2 = −k(x2 − x1 ) − kx2 (12.2.1)

This system of equations can be written in a compact matrix form:

m 0

ẍ1 −2k k x1
= (12.2.2)
0 m ẍ2 k −2k x2
We can now solve this matrix equation by using the standard methods of linear algebra,
and by following the same eigenvalue problem procedure we used in Chapter 11 for the
principal moments of inertia.
We proceed in two steps, first we find the natural frequencies ω of the system, and second
we find the positions x1 (t) and x2 (t), as follows.
Since we expect oscillatory motion, we try solutions of the form:
x1 (t) = A1 eiωt x2 (t) = A2 eiωt (12.2.3)

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 382 — #398

i i

382 Classical Mechanics: A Computational Approach

where A1 and A2 are the unknown amplitudes of oscillation for the two masses, and ω is
the unknown frequency of oscillation. Substituting these into the matrix equation (12.2.2):

m 0 −ω 2 A1 eiωt A1 eiωt

−2k k
= (12.2.4)
0 m −ω 2 A2 eiωt k −2k A2 eiωt
By canceling the exponential factor eiωt which is common to all terms, and combining the
matrices, we obtain:

−ω 2 m + 2k 0

−k A1
= (12.2.5)
−k −ω 2 m + 2k A2 0
This matrix equation represents a system of equations. A theorem from linear algebra says
that if the determinant of the matrix is nonzero, then there is a unique solution, which in
this case is the trivial solution A1 = A2 = 0. However, in order for multiple solutions to
exist, the determinant of the matrix must be zero. We are interested in a nontrivial solution
A1 , A2 6= 0, so we solve for the values of ω which cause the determinant to be zero.
We set the determinant of the matrix equal to zero:

−ω 2 m + 2k

−k
det =0 (12.2.6)
−k −ω 2 m + 2k
ω 2 m − 2k ω 2 m − 2k − k 2 = 0 (12.2.7)

Solving for ω, we obtain four possible solutions, only two of which are positive:
r r
k 3k
ω1 = , ω2 = (12.2.8)
m m
This completes the first part of the analysis, where we determined the two natural frequen-
cies ω1 and ω2 . In the previous section, we found that oscillatory solutions to our system of
equations can have one of these two frequencies.
In order to complete the description of the system, we mustq also find the two unknown
amplitudes of oscillation A1 and A2 . If we substitute ω1 = k
m into the matrix equation
(12.2.5), we obtain:

q 2 
k
m m − 2k k
 A1 −k k A1
 A2 = k =0 (12.2.9)

 q 2 
 −k A2
k k
m m − 2k

By multiplying the matrices, we obtain these two equations:

− kA1 + kA2 = 0 (12.2.10)

kA1 − kA2 = 0 (12.2.11)

It is clear that these two equations are identical, and that A1 = A2 . Note that this always
happens when we are finding the eigenvectors of a 2 × 2 matrix, one p of the equations will be
redundant and can just be ignored. We conclude that when ω1 = k/m, the two amplitudes
are equal A1 = A2 , and therefore x1 (t) = x2 (t) = A1 eiωt . We can now write the first solution
for the motion of the two masses, which corresponds to the first normal model ω1 = k/m:
p

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 383 — #399

i i

Coupled Oscillations 383

x1 (t) = x2 (t) = A1 eiω1 t (12.2.12)

Since A1 = A2 and x1 (t) = x2 (t), this type of motion corresponds to both masses moving in
the same direction and in phase at all times, as we saw previously in Figure 12.2(a). This
type of motion is known as the first normal mode or the symmetric mode of oscillation,
and the general motion of the system in this mode can be written in terms of trigonometric
functions:
x1 (t) = x2 (t) = D1 cos (ω1 t − φ1 ) (12.2.13)
In matrix notation, the first normal mode can be written as:

x1 (t) 1

= D1 cos (ω1 t − φ1 ) (12.2.14)
x2 (t) 1
By working in a similar fashion for the second natural frequency of the system, we substitute
ω2 = 3k/m into (12.1.1), and obtain A1 = −A2 . Since A1 = −A2 , this type of motion
p

corresponds to the two masses moving in opposite directions, while the center of mass
remains stationary, as shown in Figure 12.2b. This type of motion is known as the second
normal mode or the antisymmetric mode of oscillation.
We can then write the second possible solution corresponding to ω2 = 3k/m as:
p

x1 (t) 1

= E1 cos (ω2 t − φ2 ) (12.2.15)
x2 (t) −1

In general, the motion of the system will

p be a linear combination of the two normal modes,
corresponding to the frequencies ω1 = k/m and ω2 = 3k/m.
p

A faster method of obtaining the normal mode frequencies ω1 and ω2 and the amplitudes
A1 and A2 , is by solving the matrix equation (12.2.5) using the symbolic capabilities of
Mathematica. Example 12.6 shows how to use the Solve command in Mathematica, to
obtain the general solution for the two-mass three-spring system.

Coupled Oscillations 385

The Two-Mass System as an Eigenvalue Problem

GA = ω 2 A (12.2.20)
 
k1 +k2 −k2
m1 m1

A1
A= G=
 
A2

−k2 k2 +k3
m2 m2

We are looking for the eigenvalues λ = ω 2 of the square matrix G, which will give us the
natural frequencies of oscillation. We are also looking for the corresponding eigenvectors A,
which will give us the normal modes of oscillation corresponding to each natural frequency.
In order for the eigenvalue equation(12.2.20) to have a nontrivial solution (A 6= 0), the
determinant of the matrix G − ω 2 1 must be zero, where 1 is the 2 × 2 identity matrix.
This gives:  
k1 +k2 2 −k2
− ω
 m1 m1
det  =0 (12.2.21)

−k2 k2 +k3 2
m2 m2 − ω
This is the characteristic equation of our eigenvalue problem. From this point on, we proceed
by following the same two-step method used in the previous section. First, we must solve
the characteristic equation (12.2.21) in order to find the frequencies ω1 and ω2 . These
frequencies will depend on k1 , k2 , k3 , m1 , and m2 . In the next step, we substitute the first
natural frequency ω1 into (12.2.19), in order to find (A1 , A2 ), the first normal mode. Finally,
we repeat the previous steps using the second natural frequency ω2 , in order to find (A1 , A2 )
for the second normal mode.
The Mathematica and Python codes in Example 12.7 show how to find the eigenvectors
and eigenvalues for the two-mass system in Figure 12.1.
Example 12.7: The two-mass system as an eigenvalue/eigenvector problem
Solve the two oscillating mass system in Figure 12.1 as an eigenvalue/eigenvector prob-
lem, in these two cases:
(a) Identical springs and identical masses.
(b) Identical springs and different masses.
Solution:
(a) In this case the matrix,
2k −k
 
m m
G= 
−k 2k
m m
and the Mathematica code below uses the commandspEigenvalues and Eigenvectors. The
code produces the first eigenvalue ω12 = k/m or ω1 = k/m, and the corresponding eigen-
vector,

1

A1
=
A2 1
This is exactly what we obtained in Section 12.2 for the firstpnormal mode of the
oscillating two-mass system. Similarly, the second eigenvalue ω2 = 3k/m, and the cor-
responding eigenvector

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 386 — #402

i i

386 Classical Mechanics: A Computational Approach

A1 −1
=
A2 1
This is again the same result we obtained in Section 12.2 for the second normal mode.
2∗k −k
A= m
k
m
2∗k ;
−m m
Eigenvalues[A]//Simplify

OUTPUT: k
3k
m,m

Eigenvectors[A]//Simplify

OUTPUT: {{−1, 1}, {1, 1}}

Here is the corresponding Python code, which produces the same eigenvalues and eigen-
vectors, each with a multiplicity of 1. For example, the output (k/m, 1, [M atrix([[1], [1]])]
is interpreted as the ω1 = k/m with a multiplicity of 1, and the Matrix result indicates
p

the corresponding eigenvector

1

A1
=
A2 1

from sympy import ∗

init printing ()
k = Symbol ( ” k ” , p o s i t i v e =”True ” )
m = Symbol ( ”m” , p o s i t i v e =”True ” )

G=Matrix ( [ [ 2 ∗ k/m,−k/m] , [ − k/m, 2 ∗ k/m] ] )

a=G. e i g e n v a l s ( )
p r i n t ( ” E i g e n v a l u e s =” , l i s t ( a . k e y s ( ) ) )
p r i n t ( ” E i g e n v e c t o r s =” ,G. e i g e n v e c t s ( ) )

OUTPUT
E i g e n v a l u e s= [ 3 ∗ k/m, k/m]
E i g e n v a l u e s= [ ( k/m, 1 , [ Matrix ( [ [ 1 ] , [ 1 ] ] ) ] ) ,
( 3 ∗ k/m, 1 , [ Matrix ( [ [ − 1 ] , [ 1 ] ] ) ] ) ]

(b) In this case the matrix ,

 2k −k 
m1 m1
G= 
−k 2k
m2 m2

and the Mathematica code produces the two eigenvalues and eigenvectors:
   m1 −m2 +z 
A1
s
k (m1 + m2 − z) m1
ω1 =  = 
m1 m2
A2 1

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 387 — #403

i i

Coupled Oscillations 387

   −m1 +m2 +z 
A1
s
k (m1 + m2 + z) m1
ω2 =  = 
m1 m2
A2 1
q
where z = m21 − m1 m2 + m22 . In this two-mass three-spring system with unequal masses,
the ratio of the amplitudes A1 /A2 depends in a complicated manner on the masses
(m1 , m2 ), but does not depend on the spring constant k.
−k
2∗k
A= m1
k
m1
2∗k ;
− m2 m2
Eigenvalues[A]//Simplify
√ √
k m1+m2− m12 −m1m2+m22 k m1+m2+ m12 −m1m2+m22
OUTPUT: m1m2 , m1m2

Eigenvectors[A]//Simplify
√ √
m1−m2+ m12 −m1m2+m22 −m1+m2+ m12 −m1m2+m22
OUTPUT: m1 ,1 , − m1 ,1

12.3 THE DOUBLE PENDULUM

Figure 12.3 shows a double pendulum, consisting of two masses m1 and m2 attached to
massless rigid rods of lengths L1 and L2 . We can treat this problem using the same meth-
ods as for the two-mass oscillating system, by developing the equations of motion and
evaluating the natural frequencies. In Subsection 12.3.1, we will develop the Lagrangian
and the equations of motion. In Subsection 12.3.2, we will find the analytical solution for
the special case of two identical coupled pendula. In Subsection 12.3.3, we will treat the
double pendulum as an eigenvalue problem, and show how to obtain the natural frequencies
and the amplitudes of the normal modes.

12.3.1 The Lagrangian and Equations of Motion—Numerical Solutions

The position (x1 , y1 ) of the first mass is (x1 , y1 ) = (L1 sin θ1 , −L1 cos θ1 ), so that the kinetic
energy of mass m1 is:

1 1 1
T1 = m1 ẋ21 + ẏ12 = m1 L21 cos2 θ1 θ̇12 + L21 sin2 θ1 θ̇12 = m1 L21 θ̇12 (12.3.1)
2 2 2
The location (x2 , y2 ) of the second mass is shifted with respect to the first mass by (x1 , y1 ),
so that (x2 , y2 ) = (L1 sin θ1 + L2 sin θ2 , −L1 cos θ1 − L2 cos θ2 ). Therefore the kinetic energy
of the second mass is:

1
T2 = m2 ẋ22 + ẏ22 (12.3.2)

2
1

d 2 d 2
= m2 (L1 sin θ1 + L2 sin θ2 ) + (L1 cos θ1 + L2 cos θ2 ) (12.3.3)
2 dt dt

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 388 — #404

i i

388 Classical Mechanics: A Computational Approach

After differentiating and collecting terms, we find:

1
T2 = m2 L21 θ̇12 + 2L1 L2 θ̇1 θ̇2 cos (θ1 − θ2 ) + L22 θ̇22 (12.3.4)

2

+y L1
θ1

L2
θ2

Figure 12.3: The double pendulum oscillator is characterized by the two angles (θ1 (t), θ2 (t)).

The total kinetic energy is then:

1 1
T= (m1 + m2 ) L21 θ̇12 + m2 L1 L2 θ̇1 θ̇2 cos (θ1 − θ2 ) + m2 L22 θ̇22 (12.3.5)
2 2
For amplitudes of small oscillations we use the approximation cos (θ1 − θ2 ) ' 1, so that:
1 1
T=
(m1 + m2 ) L21 θ̇12 + m2 L1 L2 θ̇1 θ̇2 + m2 L22 θ̇22 (12.3.6)
2 2
The total potential energy V is the sum of potential energies for each pendulum:

V = m1 gy1 + m2 gy2 = (m1 + m2 ) gL1 cos θ1 + m2 gL2 cos θ2 (12.3.7)

θ12
For amplitudes of small oscillations, we use the approximation cos θ1 ' 1 − 2 and cos θ2 '
θ22
1− 2 so that:
θ2 θ22

V = − (m1 + m2 ) gL1 1− 1 − m2 gL2 1 − (12.3.8)
2 2
The Lagrangian L = T − V of the system is:
1 1
L= (m1 + m2 ) L21 θ̇12 + m2 L1 L2 θ̇1 θ̇2 + m2 L22 θ̇22
2 2
θ12 θ2

+ (m1 + m2 ) gL1 1 − + m2 gL2 1 − 2 (12.3.9)
2 2

We can now find the Euler-Lagrange equations:

d ∂L ∂L ∂L ∂L
− =0 − =0 (12.3.10)
dt ∂ θ̇1 ∂θ1 ∂ θ̇2 ∂θ2
By evaluating the derivatives and simplifying, we obtain:

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 389 — #405

i i

Coupled Oscillations 389

(m1 + m2 ) L1 θ̈1 + m2 L2 θ̈2 = − (m1 + m2 ) gθ1 (12.3.11)

m2 L1 θ̈1 + m2 L2 θ̈2 = −m2 gθ2 (12.3.12)
If we desire to do so, we can use the commands in Python and Mathematica to carry
out a numerical integration of this system of differential equations, and plot the functions
θ1 (t) and θ2 (t). This is left as a problem at the end of the chapter. Next, we will find
analytical solutions for θ1 (t) and θ2 (t), for the special case of two coupled identical pendula.

12.3.2 Identical Masses and Lengths—Analytical Solutions

We now use a trial solution in the form of exponential functions θ1 (t) = A1 eiωt and θ2 (t) =
A2 eiωt . Substituting and canceling the common exponential term eiωt , we find:

− (m1 + m2 ) L1 A1 ω 2 − m2 L2 A2 ω 2 = − (m1 + m2 ) gA1 (12.3.13)

− m2 L1 A1 ω 2 − m2 L2 A2 ω 2 = −m2 gA2 (12.3.14)

After collecting the terms:

(m1 + m2 ) −L1 ω 2 + g A1 − m2 L2 ω 2 A2 = 0 (12.3.15)

−m2 L1 ω 2 A1 − m2 L2 ω 2 − g A2 = 0 (12.3.16)

This system of linear equations for A1 and A2 will have a nontrivial solution, only if the
determinant is zero:

(m1 + m2 ) −L1 ω 2 + g 2

−m 2 L2 ω
det =0 (12.3.17)
−m2 L1 ω 2 −m2 L2 ω 2 − g

In principle, we can obtain the solution to this characteristic equation, however the resulting
algebraic expressions are very complicated. So instead of looking at the completely general
solution, let us look at the special case of equal lengths (L1 = L2 = L), and identical masses
(m1 = m2 = m). In this special case, (12.3.17) becomes:

2m −Lω 2 + g 2

−mLω
det =0 (12.3.18)
−mLω 2 −m Lω 2 − g

This gives the possible solutions:

√ g √ g
r r
ω1 = 2+ 2 ω2 = 2− 2 (12.3.19)
L L
As in the case of the two-mass three-spring oscillators, this completes the first part of the
analysis, where we determined the frequencies ω1 and ω2 of the normal modes. In order to
complete the description of the system, we must alsoqfind the two unknown amplitudes of
√
oscillation A1 and A2 . Substituting the value of ω1 = 2 − 2 g/L into (12.3.15), we find:
n √ g o √ g
2m −L 2 − 2 + g A1 − mL 2 − 2 A2 = 0 (12.3.20)
L L
Simplifying and canceling the common factors:
√
A1 2− 2 1
= √ =√ (12.3.21)
A2 2 −1 + 2 2

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 390 — #406

i i

390 Classical Mechanics: A Computational Approach

(a) (b)

Figure 12.4: The symmetric and antisymmetric normal modes of oscillation for the double
pendulum with equal masses and equal lengths. The two natural frequencies are ω1 =
q √ √
2 ± 2 g/L , and the amplitudes are related by A2 = ± 2A1 .

q √ √
The solution for ω1 = 2 − 2 g/L is then A2 = 2A1 , and we can write the first possible
solution corresponding to the first normal mode :
√ √
x1 (t) = A1 eiω1 t x2 (t) = 2x1 (t) = 2A1 eiω1 t (12.3.22)
√ √
Since A2 = 2A1 and θ2 (t) = 2θ1 (t), this type of motion corresponds to both masses
moving in the same direction and in phase at all times,√ as shown in Figure 12.4a. The
amplitude of the second pendulum must be equal to 2 times larger than the amplitude
of the first pendulum. This is the first normal mode or the symmetric mode of oscillation,
and the general motion of the system in this mode can be written as a linear combination
of trigonometric functions:
√
θ2 (t) = 2θ1 (t) cos (ω1 t − φ1 ) (12.3.23)
In matrix notation, the first normal mode can be written as:

θ1 (t) 1

= D1 √ cos (ω1 t − φ1 ) (12.3.24)
θ2 (t) 2
q √
By working in a similar manner with the second frequency ω2 = 2 + 2 g/L, we find:

θ1 (t)

−1
= D1 √ cos (ω1 t − φ1 )
θ2 (t) 2
This is the second normal mode or the antisymmetric mode of oscillation, in which both
masses moving in the same direction completely out of phase
√ at all times, as shown in Figure
12.4b. The amplitude of the second pendulum is again 2 times larger than the amplitude
of the first pendulum.

The above results can be obtained using the symbolic capabilities of Mathematica, by
using the Solve and Simplify commands. This is shown in Example 12.8.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 391 — #407

i i

Coupled Oscillations 391

Example 12.8: The natural frequencies and normal modes of the double pen-
dulum
Find the natural frequencies and normal modes of the double pendulum with equal
masses and equal lengths.
Solution:
When m1 = m2 = m and L1 = L2 = L, the system of equations (12.3.15) and (12.3.16)
can be written in matrix form as:
−2m Lω 2 − g 2 0

−mLω A1
=
−mLω 2 −m Lω 2 − g 0

A2

Once more,qwe accept only the two positive values of the natural frequency ω. The solu-
√ √ √
tion for ω1 = 2 − 2 g/L has amplitudes related by A2 = 2A1 and θ2 (t) = 2θ1 (t).
q √ √
The solution for ω2 = 2 + 2 g/L has amplitudes related by A2 = − 2A1 and
√
θ2 (t) = − 2θ1 (t). These results are identical to the ones obtained using the analytical
approach.
2 ∗ m ∗ (−L ∗ ω ∧ 2 + g) −m ∗ L ∗ ω ∧ 2

A= ;
−m ∗ L ∗ ω ∧ 2 −m ∗ (L ∗ ω ∧ 2 − g)

A1 0

Solve A. == , {ω, A1, A2} //Simplify
A2 0

OUTPUT:
√
(−2+ 2)g √
q
{A1 → 0, A2 → 0}, ω → − − L , A2 → 2A1 ,
√
(−2+ 2)g √
q
ω→ − L , A2 → 2A1 ,
n p √ qg √ o n p √ qg √ oo
ω → − 2+ 2 L , A2 → − 2A1 , ω → 2 + 2 L , A2 → − 2A1

12.3.3 The Double Pendulum as an Eigenvector/Eigenvalue Problem

We already saw that it is possible to solve the two-mass three-spring system as an eigenval-
ue/eigenvector problem. Let’s see how we can do that for the double pendulum problem.
One problem we encounter is that the equations of motion (12.3.11) and (12.3.12) have
terms containing both θ̈1 and θ̈2 . Let us then first isolate the θ̈1 and θ̈2 terms into separate
equations, so that the eigenvalue problem becomes more clear. By subtracting (12.3.11) and
(12.3.12) we obtain the equation for for θ̈1 :

m1 L1 θ̈1 = − (m1 + m2 ) gθ1 + m2 gθ2 (12.3.25)

By substituting θ̈1 from this expression in (12.3.11), we obtain the equation for θ̈2 :

m1 L2 θ̈2 = − (m1 + m2 ) gθ1 + (m1 + m2 ) gθ2 (12.3.26)

As previously, we substitute the exponential solutions θ1 (t) = A1 eiωt and θ2 (t) = A2 eiωt
into these two equations to obtain:

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 392 — #408

i i

392 Classical Mechanics: A Computational Approach

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 394 — #410

i i

394 Classical Mechanics: A Computational Approach

s s
1 X 1 X
L = T −V = ajk q̇j q̇k − Kjk qj qk (12.4.13)
2 2
j,k=1 j,k=1

Example 12.9 shows how to calculate the coefficients ajk and Kjk for the double pendulum
system we studied in the previous section.

Example 12.9: Finding the generalized Lagrangian for the double-pendulum

Calculate the coefficients ajk and Kjk for the double pendulum.

Solution:
The kinetic energy of the double pendulum for small oscillations was found in the
previous section to be:
1 1
T= (m1 + m2 ) L21 θ̇12 + m2 L1 L2 θ̇1 θ̇2 + m2 L22 θ̇22
2 2
1 1 1 1
= (m1 + m2 ) L1 θ̇1 + m2 L1 L2 θ̇1 θ̇2 + m2 L1 L2 θ̇1 θ̇2 + m2 L22 θ̇22
2 2
2 2 2 2
The coefficients ajk are the coefficients appearing in (12.4.4) for the kinetic energy
T . Since our generalized coordinates are q1 = θ1 and q2 = θ2 , then ajk are the coeffi-
cients of the θ̇12 /2, θ̇22 /2, and θ̇1 θ̇2 /2 terms in this expression for T . Therefore a11 =
coefficient of θ̇12 /2 = (m1 + m2 ) L21 , a12 = a21 = coefficient of θ̇1 θ̇2 /2 = m2 L1 L2 , and

a22 = coefficient of θ̇22 /2 = m2 L22 .

The potential energy of the double pendulum was found in the previous section to be:

θ12 θ22

V = − (m1 + m2 ) gL1 1 − − m2 gL2 1 −
2 2

The coefficients Kjk are the coefficients appearing in (12.4.9) for the potential energy
V , for the θ12 , θ22 , and θ1 θ2 terms. Therefore K11 = coefficient of θ12 = (m1 + m2 )gL1 ,
K12 = K21 = coefficient of (θ1 θ2 ) = 0, and K22 = coefficient of θ22 = m2 gL2 .
We can then use the constant coefficients ajk and Kjk as two 2 × 2 matrices. The
coefficients ajk are used to create the matrix M and the coefficients Kjk are used to
create the matrix K:

(m1 + m2 ) L21 (m1 + m2 ) gL1 0

m2 L1 L2
M= K=
m2 L1 L2 m2 L22 0 m2 gL2

As we will see next, the matrix M can be thought of as a generalized “mass matrix”
and the matrix K as a generalized “spring constant” matrix K.

12.4.2 The Equations of Motion for Small Oscillations Around an Equilibrium Point
We can now proceed to derive the general equations of motion, by evaluating the Euler-
Lagrange expressions:
s
∂L ∂V X
=− =− Kjk qk (12.4.14)
∂qj ∂qj
k=1

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 395 — #411

i i

Coupled Oscillations 395

s n
!
d ∂L d ∂T d X X
= = ajk q̇k = ajk q̈k (12.4.15)
dt ∂ q˙j dt ∂ q˙j dt
k=1 k=1

The Euler-Lagrange equations then become:

s
X s
X
ajk q̈k = − Kjk qk (12.4.16)
k=1 k=1

Finally, we can write these equations of motion in compact form as a single matrix equation.

General Equations of Motion for Small Oscillations Around Equilibrium

Mq̈ = −Kq (12.4.17)

     
a11 a12 ··· a1s q̈1 K11 K12 ··· K1s q1
 a21 a22 ··· a2s  q̈2   K21 K22 ··· K2s  q2 
.. .. .. .. .. = .. .. .. .. .. (12.4.18)
     
. . . . . . . . . .
   
     
as1 as2 ··· ass q̈s Ks1 Ks2 ··· Kss qs

with the s × 1 column matrix q denoting the generalized coordinates of the system, M rep-
resenting the symmetric s × s generalized “mass matrix”, and K the symmetric generalized
“spring constant” matrix.
     
q1 a11 . . . a1s K11 . . . K1s
q =  ...  M =  ... ..
.
..  K =  ..
.   . ..
.
.. 
.  (12.4.19)
  

qs as1 ... ass Ks1 ... Kss

Since we expect oscillatory motion, we try solutions of the form qj (t) = Aj eiωt , where Aj are
the amplitudes of oscillation and ω is the natural frequency of the system. By substituting
into (12.4.17) and canceling the exponential terms as usual, we obtain:

Generalized Equations of Motion for Small Oscillations Around Equilibrium

ω 2 MA = KA (12.4.20)

with the s × 1 column matrix A (whose elements are the Aj from the trial solution qj )
denoting the unknown amplitude matrix.
Equation (12.4.20) is the generalized matrix form of the familiar equation mω 2 = k,
which describes the oscillation frequency ω for a mass m attached to a spring with spring
constant k. This equation can be solved in principle by using any of the linear algebra
techniques we saw previously in this book.
Example 12.10 shows one method of solving (12.4.20) for the double pendulum.

Example 12.10: Finding the natural frequencies and normal modes of the dou-
ble pendulum, again
Solve ( (12.4.20) for the double pendulum with equal masses and equal lengths, and
find the natural frequencies and the normal modes.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 396 — #412

i i

396 Classical Mechanics: A Computational Approach

Solution:
In Example 12.9, we calculated the coefficients ajk and Kjk of the matrices M and K
for the double pendulum. Equation (12.4.20) then becomes:

(m1 + m2 ) L21 (m1 + m2 ) gL1 0

m2 L1 L2 A1 A1
ω2 =
m2 L1 L2 m2 L22 A2 0 m2 gL2 A2

By collecting terms, i.e., writing (12.4.20) as ω 2 M − K A = 0 we obtain:

(m1 + m2 ) L21 ω 2 − gL1 m2 L1 L2 ω 2 0

A1
=
m2 L1 L2 ω 2 m2 L22 ω 2 − gL2 A2 0
Simplifying and setting L1 = L2 = L and m1 = m2 = m:

2mL Lω 2 − g mL2 ω 2 0

A1
=
mL2 ω 2 mL Lω 2 − g A2 0
This, of course, is the same equation we obtained and solved in Example 12.8 (to see
this, divide the above by −L), and we can reuse the code for Example 12.8 to find ω and
A.
In the next subsection, we proceed to show how the above general matrix formalism of the
equations of motion leads to the concept of normal coordinates.

12.4.3 Normal Coordinates

Let us consider the general matrix equation (12.4.20), and rewrite it in the form:
 
A1
 A2 
K − ωj2 M Aj = 0 where Aj =  (j = 1 . . . s) (12.4.21)


 ... 
As j

KAj = ωj2 MAj (12.4.22)

where s is equal to the number of degrees of freedom of the coupled oscillator. Equation
(12.4.22) tells us that the column vector Aj can
be characterized
as a generalized eigenvector
of the real-valued symmetric square matrix K − ωj2 M .
A general theorem of linear algebra for real-valued symmetric square matrices, proves
that these s generalized eigenvectors Aj are linearly independent (s × 1) column vectors,
and that they form a set of complete set of basis vectors for the space of all (s × 1) column
vectors. Therefore, we can write any solution q of the equations of motion (12.4.17) as a
linear combination of these eigenvectors:
s
X
q= cj (t)Aj (12.4.23)
j=1
where cj (t) are time dependent coefficients. Our goal is to show that these coefficients cj (t)
oscillate independently of each other, i.e., that each coefficient corresponds to a different
frequency ωm .
Substituting (12.4.23) into (12.4.17) and using (12.4.22), we obtain:

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 397 — #413

i i

Coupled Oscillations 397

Mq̈ = −Kq (12.4.24)

X X
c̈j (t)MAj = − cj (t)KAj (12.4.25)
X X
c̈j (t)MAj = − cj (t)ωj2 MAj (12.4.26)

Since the vectors are linearly independent, the last equation can hold only if the coeffi-
cients of the vector MAj are equal:

c̈j (t) = −cj (t)ωj2 (12.4.27)

The solutions of these s equations are the sinusoidal functions cj (t) given by:

cj (t) = bj sin (ωj t − φj ) (12.4.28)

where φj is a phase angle, and bj is the amplitude; both parameters (φj , bj ) are determined
by the initial conditions of the system. This equation establishes that each coefficient cj (t)
oscillates with a different frequency ωj . By substituting (12.4.28) into (12.4.23), we obtain
the desired general result:
s
X
q= bj sin (ωj t − φj ) Aj (12.4.29)
j=1

In words, this equation tells that any solution q of the equations of motion, can be
written as the linear combination of s independent oscillating terms, each of which has its
own characteristic natural frequency ωj . These oscillating terms are known as the normal
coordinates cj (t) of the system. Hence, all motions of a linear coupled oscillator can be
expressed as a linear combination of normal coordinates.
The material in this section is a formal way of demonstrating the result of Example 12.4.
In Example 12.4 we found that the analytical solution, of the case
equal masses
p andequal
springs, could be written as the sum of two terms, cos k/mt and cos 3k/mt . For
p

example, the
p solution x1 (t) could be written in the form (12.4.29) by writing ω1 = k/m
p

and ω2 = 3k/m as the frequencies, φ1 = φ2 = π/2 as the phases, and

0

a/2
A1 = A2 = (12.4.30)
0 −a/2
as the amplitudes.

12.5 CHAPTER SUMMARY

Systems of coupled harmonic oscillators can be analyzed by using either the Lagrangian
approach, or alternatively using Newton’s second law. The simplest form of these systems
consists of two masses connected by springs to each other, with the Lagrangian:
1 1 1 1 1
L = T − V = m1 x˙1 2 + m2 x˙2 2 − k1 x21 − k3 x22 − k2 (x1 − x2 )2
2 2 2 2 2
This leads to the system of coupled differential equations:

m1 ẍ1 = −k1 x1 − k2 (x1 − x2 )

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 398 — #414

i i

398 Classical Mechanics: A Computational Approach

m2 ẍ2 = −k2 (x2 − x1 ) − k3 x2

The solution x1 (t) and x2 (t) of such systems are in general complex, and it is not
possible to give a simple physical description of the motion of the masses. However, in the
special case of equal masses and equal spring constants, we obtain the analytical solution
by using the standard matrix techniques of linear algebra.
The first possible
p solution for the motion of the two masses corresponds to the natural
frequency ω1 = k/m, and both masses move in the same direction and in phase at all
times.
This type of motion is known as the first normal mode or the symmetric mode of
oscillation, and can be written as:

x1 (t) 1

= D1 cos (ω1 t − φ1 )
x2 (t) 1

The second possible p solution for the motion of the two masses corresponds to the
natural frequency ω2 = 3k/m, and in this type of motion the two masses moves in
opposite directions, while the center of mass remains stationary.
This type of motion is known as the second normal mode or the antisymmetric mode
of oscillation, and can be written as:

x1 (t) 1

= E1 cos (ω2 t − φ2 )
x2 (t) −1

In general, the motion of the system will be a linear p combination ofpthe two possible
normal modes, corresponding to the frequencies ω1 = k/m and ω2 = 3k/m.
This system can also be reduced into an eigenvalue problem, by substituting a trial
solution of the form x1 (t) = A1 eiωt and x2 (t) = A2 eiωt in the system of coupled equations.
In a second well-known example of coupled harmonic oscillators, we examined the
double pendulum, which also has two normal modes. The frequencies ofqthe symmetric
q √ g √ g
and antisymmetric modes of oscillation are ω1 = 2− 2 L and ω2 = 2+ 2 L .
In this chapter, we also looked at the general mathematical formalism for the normal
modes of coupled oscillators. Here we use generalized coordinates qj (j = 1, 2, . . . , s) and
the corresponding generalized velocities q̇j , where s is the number of degrees of freedom
of the system.
The Lagrangian for small oscillations around an equilibrium position can be reduced
into a quadratic function of the generalized coordinates and generalized velocities of the
system in the form:
n
1X 1 X
L = T −V = ajk q̇j q̇k − Kjk qj qk
2 2
j,k j,k=1

where ajk and Kjk are constants.

For such a system, the Euler-Lagrange equations become:
n
X n
X
ajk q̈k = − Kjk qk
k=1 k=1

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 399 — #415

i i

Coupled Oscillations 399

This system of coupled differential equations can be solved using standard techniques
of linear algebra.
Any solution q = qj (j = 1, 2, . . . , s) of these equations of motion can be written as the
linear combination of s independent oscillating terms, each of which has its own charac-
teristic natural frequency ωj :
s
X
q= bj sin (ωj t − φj ) Aj
j=1

These oscillating terms are known as the normal coordinates of the system.

12.6 END-OF-CHAPTER PROBLEMS

Section 12.1–12.2: Coupled Oscillations of Spring-Mass Systems—Normal
Mode Analysis
1. Two masses m1 and m2 are connected to each other with a spring of spring constant k,
and they are placed on
pa horizontal frictionless table. Show that the natural frequency
of the system is ω = k/µ, where µ = m1 m2 /(m1 + m2 ) is the reduced mass of the
system.

2. Consider the two-mass three-spring system studied in this chapter, with different
masses m1 , and m2 and identical spring constants k1 = k2 = k3 .
(a) Transform the Lagrangian by introducing two new variables q1 = x1 +x2 and q2 =
x1 − x2 . Obtain new equations of motion by using the Euler-Lagrange equations
for this transformed Lagrangian.
(b) Solve the new equations and obtain q1 (t) and q2 (t), and the natural frequencies
of the system. This is an example where by using a new set of coordinates, we
can uncouple the differential equations of motion. These new coordinates q1 (t),
q2 (t) represent the normal coordinates of the system.
3. In the two-mass three-spring system studied in this chapter, obtain the analytical
solution x1 (t) and x2 (t), by using the following initial conditions: the first mass is
at rest in the equilibrium position, and the second mass is moved a distance a from
equilibrium, and released from rest.
4. Obtain an approximate analytical solution for the case of two weakly coupled oscilla-
tors described in Example 12.5. This system is weakly coupled if the spring constant
k2 of the middle spring is much smaller than the spring constants of the two end
springs (k2 << k1 = k3 ). Use m1 = m2 and the following initial conditions: the first
mass is at rest in the equilibrium position, and the second mass is moved a distance
a from equilibrium, and released from rest. Find the frequency ω at which energy is
transferred back and forth between the two coupled oscillators.

5. Consider two equal masses m attached to one wall and to each other with springs of
spring constant √
k, as shown in Figure 12.5. Show that the natural frequencies of the
system are ω = 5±1
p
2 k/m.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 400 — #416

i i

400 Classical Mechanics: A Computational Approach

Figure 12.5: Problem 12.5: Two equal masses attached by a spring.

6. Consider the three different situations shown in Figure 12.6, where three identical
masses m are attached to springs with the same spring constant k. Write the equations
of motion for these three cases.

Figure 12.6: Problem 12.6: Three identical masses connected by springs.

7. Three equal masses m are connected to each other with identical springs of constant
k, as shown in Figure 12.7. Find and describe the normal modes of oscillation (assume
no friction), by using appropriate programming code.

Figure 12.7: Problem 12.7: Three identical masses connected by two identical springs.

12. Four identical masses are connected by four identical springs, and are constrained to
move on a frictionless circle of radius R. How many normal modes of small oscillations
are there and what are the frequencies of small oscillations?
13. Three bodies of equal mass m are connected with springs of constant k, and are placed
in an equilateral shape as shown in Figure 12.11. The masses are constrained to move
on the xy-plane. Find the natural frequencies of this system.

Figure 12.11: Problem 12.13: Three masses connected in a triangle.

14. A simple classical model of the CO2 molecule would be a linear structure of three
masses, with the electrical forces between the ions represented by two identical springs
of equilibrium length L and force constant k. Assume that only motion along the
original equilibrium line is possible, i.e., ignore rotations. Let m be the mass of each
oxygen and M be the mass of the carbon atom
(a) How many vibrational degrees of freedom does this system have?
(b) Find the normal modes and calculate the natural frequencies.
15. Two equal masses m are suspended from a ceiling by springs with constants k, as
shown in Figure 12.12.
(c) Find the equations of motion of the two masses by using a coordinate system
(y1 , y2 ) as measured from the ceiling.
(d) Transform these equations of motion by making a substitution y1 = x1 + u1 and
y2 = x2 + u2 , where u1 , and u2 are the equilibrium positions of the two masses.
Solve the new system of equations and described the normal modes of oscillations
of this system in the vertical direction.

Figure 12.12: Problem 12.15: Two masses hanging by springs from a vertical support.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 403 — #419

i i

Coupled Oscillations 403

Section 12.3: The Double Pendulum

16. A double pendulum consists of two equal masses m, and the upper and lower strings
have lengths L1 = 3L and L2 = 4L. Find the normal modes for small oscillations of
this system.
17. Two unequal masses M and m (with M > m) hang from a support by strings of equal
lengths l. The masses are coupled by a spring of spring constant K and of unstretched
length equal to the distance L between the support points as shown in Figure 12.13.
Find the normal mode frequencies for the small oscillations along the line between the
two masses, and describe the motion of the two masses in each mode. Write down the
most general solution.

Figure 12.13: Problem 12.17: Two unequal mass pendulums connected by a spring.

Now, let us change the differential equation to the form: ẍ + x2 = 0, and say we find two
solutions to this equation that we will also call x1 and x2 . Is it true that c1 x1 (t) + c2 x2 (t)
is still a solution? We insert the superposition into the nonlinear differential equation:

(c1 x¨1 + c2 ẍ2 ) + (c1 x1 + c2 x2 )2 = c1 x¨1 + x21 + c2 ẍ2 + x22 + 2c1 c2 x1 x2 = 2c1 c2 x1 x2

Notice that now, we do not get zero. In other words, the differential equation no longer
satisfies superposition. The result is that unlike linear systems, nonlinear systems cannot be
broken into parts, solved individually, and then have those solutions combined for a solution
to the whole system. Recall the topic of the driven damped harmonic oscillator. If there are
multiple sinusoidal drive terms on the righthand side of the equation, we could solve the
differential equation for one right-hand term at a time, then add the individual solutions to
get a particular solution. That cannot be done for nonlinear systems. This makes nonlinear
systems more difficult to solve in closed form. In fact, closed form solutions for nonlinear
differential equations often cannot be found.
At this point, we might be wondering if any of this matters. Are nonlinear systems
just mathematical curiosities? The answer is no. In fact, many systems from the physical,
biological, and social sciences are nonlinear, and linear descriptions of behavior are often
approximations. For example, Hooke’s law, F (x) = −kx, holds for an oscillator undergoing
small amplitude oscillations. Once the amplitude of the oscillations is large enough, then
the restoring force is no longer linear, and nonlinear terms need to be included in order for
the equation of motion to accurately describe the oscillation. One of the results is that the
period of oscillation is no longer independent of the amplitude of oscillation.
As we will see, nonlinear systems can display a much more diverse set of behaviors than
linear systems, which tend towards one of four basic types of behavior: exponential growth,
exponential decay, oscillation, and a combination of the previous three. The limited types of
behavior and mathematical simplicity makes linear systems easier to work with and allows
us to predict future states of the system with a high degree of accuracy at any point in
the future. Nonlinear systems, however, are often not exactly solvable and display complex
behaviors such as bifurcations and chaos, which can make long term predictions of behavior
impossible.
So, what do we do if we encounter a nonlinear system? Do we attempt to make a linear
approximation and hope that the approximation is good enough? Do we just give up because
we cannot solve the problem? First, we do not have to give up because we cannot solve a
problem in closed form. The numerical methods described in this book so far are powerful
tools and can provide insight into the behavior of a nonlinear system. In this chapter, we
will expand upon those numerical methods, learning how to interpret these solutions in new
ways. Second, linear approximations can work well and can shed some light on the solution

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 409 — #425

i i

Nonlinear Systems 409

of a nonlinear system. These linear approximations, often done using Taylor expansion, will
allow us to use some of the tools for linear differential equations in order to provide some
understanding of the nonlinear system. Ultimately, one of the most powerful tools we have
in our arsenal is the phase space plot mentioned in Chapter 6. The phase space plots will
provide graphical methods of understanding the long-term behavior of a nonlinear system.
Recall, that for the damped driven harmonic oscillator, we were mostly interested in the
steady-state solution and less interested in the transient behavior. The same will generally
be true for nonlinear systems, and phase space plots will allow us to identify the nature of
the steady-state solution.
Beyond physics, nonlinear terms appear in equations that govern systems which have
interacting components. For example, equations that govern population size, predator-prey
interactions, certain chemical reactions, arms races, and the spreading of disease (just to
name a few) are nonlinear. Therefore, the mathematical techniques we will learn in this
chapter are applicable to systems from a wide range of natural and social sciences. Some
of the most exciting developments in science is the application of “traditional physics tech-
niques” to problems in the biological and social sciences.

13.2 THE DAMPED HARMONIC OSCILLATOR, REVISITED

To develop an understanding of one of the primary tools for nonlinear systems, phase space
diagrams, we will study a familiar system, the damped harmonic oscillator. In mechanical
systems, phase space diagrams plot the momentum p versus the position x of a given
system. When working with problems from the biological and social sciences, the concept
of momentum may not make sense, and therefore velocity (i.e., rate of change) and position
(i.e., the value of a quantity) is most commonly used for those systems. For our mechanical
systems p = mẋ, our phase space diagrams will plot velocity vs. position the mass would
only affect the scale of the vertical axis. In this section, we will reproduce the phase plot
for the damped harmonic oscillator:

mẍ + 2γ ẋ + ω02 x = 0 (13.2.1)

with m = 1, ω0 = 1, and x(0) = 1, ẋ(0) = 0, and all quantities are in SI units. We will
generate the phase portrait for γ = 0 and γ = 0.1, also in SI units. In order to produce a
phase plot, we will perform the following steps:
1. Break up (13.2.1) into two first-order ordinary differential equations (ODEs) with
v ≡ ẋ.
2. Solve the two first-order differential equations with a numerical solver using a step
size dt. In this case, we will use odeint from the Python SciPy library.
3. Create a list of ordered pairs, {x(idt), v(idt)}, for i = 0, . . . , n, where n is the number
of steps used to solve the ODEs.
4. Plot the list of ordered pairs to create the phase portrait.
We begin by breaking up (13.2.1) into two first-order equations and substituting m = 1 and
ω0 = 1:
)
ẋ =v
(13.2.2)
v̇ = − 2γv − x

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 410 — #426

i i

410 Classical Mechanics: A Computational Approach

In Algorithm 12, we numerically integrate (13.2.2) using odeint command in Python’s

SciPy library. Of course, we could use Euler’s method, fourth-order Runge-Kutta, or any
other method to integrate (13.2.2). We should always use caution when using code written
by someone else to make sure that it does what is expected. Before using such code, we
should test out the code on problems whose results we already know. SciPy is open source,
so you have access to the source code and can even make changes to it! If you choose to
use a command, such as odeint, in a library like SciPy, it is important to know what it is
doing. At the time of this writing, the odeint command uses the LSODA routine from the
FORTRAN 77 (as in 1977!) library ODEPACK. So in this case, a Python “command” is
actually calling a FORTRAN 77 library. This is done to leverage the speed advantage of a
compiled language such as FORTRAN 77. However, one does not need to know FORTRAN
to use the Python command odeint. The LSODA routine uses something called Adams
Methods in the case of equations like (13.2.2). The details of Adams methods is beyond the
scope of the text, the point is that it is important to research the algorithms used in any
libraries when writing our own code, in order to make sure that the algorithm we intend
to use is appropriate for the problem at hand. In this case, odeint is an appropriate choice
to numerically solve (13.2.2). Note the phrase “At the time of this writing...” used earlier
in this paragraph. It is important to note that libraries get updated and changed. Always
consult the library’s online help/manual files for the most current information.

import numpy a s np
from s c i p y . i n t e g r a t e import o d e i n t
import m a t p l o t l i b . p y p l o t a s p l t

d e f dho ( y , t ) : #d e f i n e t h e ODEs t o be s o l v e d
x, v = y
gamma = 0 . 1
d e r i v = [ v , −2.0∗gamma∗v − x ] #RHS o f ODEs
return deriv

x v e c = [ 1 . 0 , 0 . 0 ] #x ( 0 ) = 1 and #v ( 0 ) = 0
t = np . l i n s p a c e ( 0 , 1 0 0 , 1 0 0 1 ) #time r a n g e f o r t h e s o l u t i o n

s o l u t i o n = o d e i n t ( dho , x ve c , t )
plt . plot ( solution [ : , 0 ] , solution [ : , 1 ] )

Algorithm 12: The Python code used to generate the phase space diagram for (13.2.2).

The results of this code for γ = 0.1 and γ = 0 are shown in Figure 13.1. Pretend, for a
moment, that we did not know how to solve this problem in closed form, but instead all
we had were the phase space plots that were generated numerically; this is often true for
nonlinear systems. What can we learn from Figure 13.1 about the behavior of (13.2.1)? Let
us first consider the left graph in Figure 13.1, where γ = 0.1 and, hence, there is a small
amount of damping in the system. To help visualize the behavior of the system, we will
think of (13.2.1) as describing a mass on a spring with damping. The path in phase space,
called the phase space trajectory, or simply trajectory, describes the behavior of the system,
and represents the solution to (13.2.2) and therefore, also to (13.2.1). To understand the
behavior of the system, follow the trajectory in phase space! As we follow the trajectory from
its starting point we see that the position decreases towards zero while the velocity becomes

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 411 — #427

i i

Nonlinear Systems 411

γ = 0.1 γ=0
1.00
1.0
0.75

0.50
0.5
0.25
velocity

velocity
0.00 0.0

−0.25
−0.5
−0.50

−0.75
−1.0
−1.00
−1.0 −0.5 0.0 0.5 1.0 −1 0 1
position position

Figure 13.1: The phase space plot for (13.2.2) with γ = 0.1 (left) and γ = 0 (right).

which is the eigenvector equation for the matrix A. Hence, in our trial solution, u = weλt , λ
is the eigenvalue of A associated with the eigenvector, w. We have previously studied eigen-
values and eigenvectors in Chapters 11 and 12. Here we will study them from a geometric
perspective.
Returning to our analysis, we can rewrite (13.3.9) in the form,

(A − λ1) w = 0 (13.3.10)
where 1 is the 2 × 2 identity matrix (with 1’s on the diagonal and 0’s otherwise) and 0 is the
zero vector. The field of Linear Algebra tells us that if the matrix, A − λ1, is invertible, then
the solution to (13.3.10) is w = 0, which is the trivial solution. For a nontrivial (and therefore
interesting!) solution, we need the matrix A − λ1 to be singular (i.e., not invertible). Linear
Algebra also tells us that a matrix is singular if its determinant is zero. Hence, in order
to get a nontrivial solution, det (A − λ1) = 0 must be true. The condition det (A − λ1) = 0
produces a polynomial in λ, called the characteristic polynomial, which can be solved to
obtain the eigenvalues of A. Because the matrix, A, is a 2 × 2 matrix, it will produce
two eigenvalues and each eigenvalue can be plugged into (13.3.10) to get its associated
eigenvector. Using the principle of linear superposition, the solution to (13.3.8) is therefore,

u = c1 w1 eλ1 t + c2 w2 eλ2 t (13.3.11)

where λi is the eigenvalue of A associated with the eigenvector, wi and ci are constants
determined by initial conditions.
Before moving on with examples, it is important to pause and try to understand the
implications of (13.3.9) and (13.3.11). These next two paragraphs are difficult, but worth the
effort needed in understanding them. When a vector is multiplied by a matrix, the result is
a vector. Consider the matrix equation, Mv1 = v2 where M is a N × N matrix and v1 and
v2 are N -dimensional vectors (N × 1 matrices). We can think of the equation Mv1 = v2
as saying that when multiplied by the matrix, M, the vector v1 is transformed into the
vector v2 . The new vector, v2 , may have a different magnitude and direction compared to
v1 . The point here is that matrices change vectors. Now, let us look at (13.3.9). Notice that
in (13.3.9), the vector w, when multiplied by the matrix A, essentially does not change
direction. The length of w is changed by a factor of λ, but the new vector λw still lies along
the same line as w (if λ < 0, the new vector λw now points in the opposite direction of w).
This makes eigenvectors special, they essentially define directions that are unchanged by
the transformation A.
Why do we care about how matrices change vectors? Recall that the proposed solution
to (13.3.8) is u = weλt and u is a point in the phase space near the fixed point. The mag-
nitude u grows or decays exponentially in time depending on the sign of λ. The geometric
interpretation of eigenvalues and eigenvectors tells us that a solution that starts near the
fixed point along the direction of w will either move towards the fixed point along the
direction of w if λ < 0, or away from the fixed point along the direction of w if λ > 0. At
the end of this section, we will briefly address the case of λ = 0. The sign of the eigenvalue λ
gives the stability of the fixed point along the direction of w. Because we are working with
problems in two dimensions, (13.3.11) tells us that there are two directions of stability for
the fixed point. One direction is along w1 and the other is along w2 , these directions are
eigenvectors of A. Each direction has its own stability determined by λ1 and λ2 , respectively.
Hence, if λ1 < 0 and λ2 > 0, then a point u starting near the fixed point will initially move
towards the fixed point along the direction of w1 but, as time progresses, it will eventually
start to move away from the fixed point along the direction of w2 . The fixed point in this

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 416 — #432

i i

416 Classical Mechanics: A Computational Approach

case represents an unstable equilibrium since perturbations from the equilibrium (the fixed
point) grow exponentially as t → ∞.
At this point, the math might be overwhelming. The procedure is not difficult (we
promise!) and can be performed to solve problems even if the linear algebra isn’t fully
understood (although you should study linear algebra, it is really quite interesting). We
will illustrate this procedure using example problems in order to clarify the steps outlined
above.
Example 13.1: The damped harmonic oscillator, revisited, again
Consider the equations of motion for the damped harmonic oscillator with m = 1:

ẋ =y
ẏ = − 2γy − ω02 x
Find the Jacobian matrix, compute its eigenvalues and its eigenvectors.
Solution:
We have already noted that f = y and g = −2γy − x and that the fixed point is located
at the origin. We can use (13.3.7) to get the Jacobian,
 
∂f ∂f
0 1
 
∂x ∂y
A= =
  

∂g ∂g
−ω 2 −2γ
∂x ∂y (0,0) 0

Next, we begin the process of computing the eigenvalues,

0 1 1 0 1

−λ
det(A − λ1) = det = =0

−λ
−ω02 −2γ 0 1 −ω02 −2γ − λ

Computing the determinant gives:

−λ(−2γ − λ) + ω02 = λ2 + 2γλ + ω02 = 0

which is solved for λ:
q
λ = −γ ± γ 2 − ω02
Does this look familiar?
q It should! Next, we will use (13.3.10) to find the eigenvector
associated with λ = −γ + γ 2 − ω02 , (we will let the reader do the other one):

0 1 1 0 0

w1
q
(A − λ1) w = − −γ + γ − ω0
2 2 =
−ω02 −2γ 0 1 w2 0

which simplifies to:

 q 
γ− γ 2 − ω02 1 0

 w1 =
0
 q
−ω02 2 2
−γ − γ − ω0 w 2

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 417 — #433

i i

Nonlinear Systems 417

q
γ − γ 2 − ω02 w1 + w2 =0
q
−ω02 w1 + −γ − γ 2 − ω02 w2 =0

resulting in  
q
−1 γ + γ 2 − ω02 

w1
=  ω02
w2
1
What does all of this say about the stability of the equilibrium at the origin? We will
return to that later in this section, for now, this example is intended to simply demonstrate
the procedure outlined so far in this section.

This next example is not physically motivated. In other words, the equations do not
describe a particular physical system. However, it is a simple system of equations that allows
for a direct demonstration of the procedure and is more easily extended to a discussion of
stability than the example above.

Example 13.2: A system of two first-order ordinary differential equations

Consider the following system of ordinary differential equations:

ẋ =y(x + 1)
ẏ =x(y + 1)
Find the fixed points and solve the corresponding linearized system of equations.
Solution:
First, we need to solve for y(x + 1) = 0 and x(y + 1) = 0. It should be easy to verify for
this problem that there are two fixed points, (x∗ = 0, y ∗ = 0) and (x∗ = −1, y ∗ = −1). For
this example problem, we will study the fixed point at the origin and leave the analysis of
the other fixed point for the reader.
Following the procedure outlined above, we would perturb our solution using, x∗ = 0+v
and y ∗ = 0 + v, plug into our system of equations and Taylor expand to get the linearized
system. Fortunately, however, we have already done all of that work above and can simply
plug into (13.3.7):

x+1

u̇ y u
=
v̇ y+1 x v

(x=0,y=0)
0 1

u̇ u
=
v̇ 1 0 v
Now that we have our Jacobian, we can compute the eigenvalues by finding the char-
acteristic polynomial using, det(A − λ1) = 0:
0−λ 1

1 = λ2 − 1 = 0
0−λ

Hence the characteristic polynomial, λ2 − 1 = 0, gives eigenvalues λ = ±1. Next, we

find the eigenvectors for the eigenvalue λ = 1 using (13.3.10):

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 418 — #434

i i

418 Classical Mechanics: A Computational Approach

0 1 1 0 −1 1

w1 w1
− = =0
1 0 0 1 w2 1 −1 w2
The above equation corresponds to the system of equations:

−w1 + w2 =0
w1 − w2 =0
which has the solution, w1 = 1 and w2 = 1. It is√common √ to normalize the eigenvector,
hence the eigenvector associated with λ = 1 is (1/ 2, 1/ 2). Repeating
√ the
√ procedure, we
should verify that the eigenvector associated with λ = −1 is (−1/ 2, 1/ 2). Finally, the
solution for the system linearized about the origin is:
√ √
1/√2 −1/√ 2

u
= c1 et + c2 e−t
v 1/ 2 1/ 2
where c1 and c2 are constants that cannot be found without initial conditions for our
linearized system.

tmax = 6;

soln1 = NDSolve[{x0 [t] == y[t] ∗ (x[t] + 1), y 0 [t] == x[t] ∗ (y[t] + 1), x[0] == −0.1, y[0] ==
0.099}, {x, y}, {t, 0, tmax}];

soln2 = NDSolve[{x0 [t] == y[t] ∗ (x[t] + 1), y 0 [t] == x[t] ∗ (y[t] + 1), x[0] == 0.1, y[0] ==
−0.101}, {x, y}, {t, 0, tmax}];

soln3 = NDSolve[{x0 [t] == y[t] ∗ (x[t] + 1), y 0 [t] == x[t] ∗ (y[t] + 1), x[0] == −0.1, y[0] ==
0.11}, {x, y}, {t, 0, tmax}];

soln4 = NDSolve[{x0 [t] == y[t] ∗ (x[t] + 1), y 0 [t] == x[t] ∗ (y[t] + 1), x[0] == 0.1, y[0] ==
−0.09}, {x, y}, {t, 0, tmax}];

stable = Plot[−x, {x, −0.1, 0.1}, PlotStyle → Black];

unstable = Plot[x, {x, −0.1, 0.1}, PlotStyle → Black];
stream = StreamPlot[{y(x + 1), x(y + 1)}, {x, −0.1, 0.1}, {y, −0.1, 0.1}];
trajectory1 = ParametricPlot[Evaluate[{x[t], y[t]}/.soln1], {t, 0, tmax}, PlotStyle → Red];
trajectory2 = ParametricPlot[Evaluate[{x[t], y[t]}/.soln2], {t, 0, tmax}, PlotStyle → Red];
trajectory3 = ParametricPlot[Evaluate[{x[t], y[t]}/.soln3], {t, 0, tmax}, PlotStyle → Red];
trajectory4 = ParametricPlot[Evaluate[{x[t], y[t]}/.soln4], {t, 0, tmax}, PlotStyle → Red];

Show[stream, stable, unstable, trajectory1, trajectory2, trajectory3, trajectory4,

Frame->True, Axes → False, BaseStyle → {FontSize → 18},
FrameLabel → {x, y}, ImageSize → Large]
Algorithm 13: The Mathematica code for Example 13.3.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 421 — #437

i i

Nonlinear Systems 421

• An unstable spiral occurs when Re(λ) > 0 and the trajectory spirals away from the
fixed point. This would correspond to the damped harmonic oscillator with a negative
damping.

• A center occurs when Re(λ) = 0 and there is neither exponential growth away from the
fixed point, nor is there exponential decay onto the fixed point. The point is neutrally
stable and the trajectory circles the center. This corresponds to the undamped simple
harmonic oscillator.
What happens when there is a zero eigenvalue? One typically uses numerical methods to
analyze the system. Finally, it should be mentioned that there are cases that are called
degenerate because they have only one eigenvalue and/or one eigenvector. These lead to
other types of fixed points which are beyond the scope of this book. The interested reader
is directed to [Strogatz(2014)] for a thorough and readable review of even more types of
fixed points. Next, we will use our newfound knowledge of fixed points to analyze the simple
plane pendulum.

13.3.1 The Simple Plane Pendulum, Revisited

Consider a simple plane pendulum shown below in Figure 13.2. The pendulum consists of
a massless rod of length, `, bob of mass m, and makes an angle with the vertical θ.

θ ℓ

Figure 13.2: A simple plane pendulum.

The equation of motion of the pendulum has been found many times in this book,

θ̈ + ω02 sin θ = 0 (13.3.12)

where the natural frequency = g/`. Equation (13.3.12) can be solved in closed form,
ω02
however those solutions are not very insightful. Let us instead proceed with the analysis
developed in Section 13.3 by breaking up (13.3.12) into two first-order equations:

θ̇ =y
(13.3.13)
ẏ = − ω02 sin θ
where the variable y is the angular velocity. Notice that fixed points occur at (θ∗ = 0, y = 0)
and (θ∗ = ±π, y = 0), these make sense physically because they correspond to the very
bottom and top of the pendulum’s path. Based on the way we have defined our coordinate
system, θ = ±π are the same point, the very top of the pendulum’s path. Note that θ = +π
when the pendulum approaches the top moving counterclockwise, and θ = −π when it
approaches the top moving clockwise. The Jacobian of the system is:

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 422 — #438

i i

422 Classical Mechanics: A Computational Approach

y 0

-1

-2

-3
-3 -2 -1 0 1 2 3
θ
Figure 13.3: The phase portrait for (13.3.12) with ω0 = 1.

0 1

A= (13.3.14)
−ω02 cos θ 0
At the origin, the Jacobian (13.3.14) becomes:

0 1

(13.3.15)
−ω02 0
with eigenvalues λ = ±iω0 . The origin, corresponding to the bottom of the pendulum’s
path, is a center. The center is representing the oscillation of the pendulum about the
stable equilibrium position.
At (±π, 0),the Jacobian (13.3.14) becomes

0.1

0.0

V(x)
Out[ ]=
-0.1

-0.2

-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5

Figure 13.5: The double well potential (13.3.17) with k = = 1.

mẍ = kx − x3 (13.3.18)

Next, using m = k = = 1, we will find the fixed points, identify their stability, and plot
the phase portrait. In order to plot the phase portrait we will need to recast (13.3.18) into
a two-dimensional system,

ẋ =y
)
(13.3.19)
ẏ =x − x3
We see from (13.3.19) that fixed points exist at (0, 0), (1, 0), and (−1, 0). The Jacobian
matrix is,

0 1

A= (13.3.20)
1 − 3x2 0
√
which provides eigenvalues of ±1 for (0, 0) and ±i 2 for the points (1, 0) and (−1, 0).
Hence, the origin is a saddle and the other two points are centers. The phase portrait
can be plotted using a technique similar to that done in Example 13.3 and is shown in
Figure 13.6. Notice the two homoclinic orbits, one in red (in e-book) and the other in green
(in e-book), which surround the two centers. Inside each homoclinic orbit are closed-loop
trajectories corresponding to oscillations about each stable equilibrium point. Outside of
the homoclinic orbit are larger closed-loop trajectories which correspond to an oscillation
that passes through each equilibrium point.
The Mathematica code used to create Figure 13.6 is shown in Algorithm 14. The variable
names identify which part of the plot the particular line creates. The overlay of the stream
plot and the phase portrait trajectories can be very helpful when visualizing regions of the
phase portrait where we have not created trajectories. However, as the next subsection will
show, the overlay can make the phase portrait too “busy” and therefore difficult to read.

13.3.3 Damped Double-Well

As another example of a phase portrait, we will examine a damped-double well. The force
acting on the particle is,

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 425 — #441

i i

Nonlinear Systems 425

0
y

Out[ ]=

-1

-2
-2 -1 0 1 2
x
Figure 13.6: The phase portrait for the double-well potential.

sp = StreamPlot[{y, x − x∧ 3}, {x, −2, 2}, {y, −2, 2}];

leftHomoclinicSolution = NDSolve[{x0 [t] == y[t], y 0 [t] == x[t] − x[t]∧ 3, x[0] == 0.01,

y[0] == 0}, {x, y}, {t, 0, 100}];
leftHomoclinicPlot = ParametricPlot[Evaluate[{x[t], y[t]}/.leftHomoclinicSolution],
{t, 0, 100}, PlotStyle → Green];

rightHomoclinicSolution = NDSolve[{x0 [t] == y[t], y 0 [t] == x[t] − x[t]∧ 3, x[0] == −0.01,

y[0] == 0}, {x, y}, {t, 0, 100}];
rightHomoclinicPlot = ParametricPlot[Evaluate[{x[t], y[t]}/.rightHomoclinicSolution],
{t, 0, 100}, PlotStyle → Red];

Show[sp, rightHomoclinicPlot, leftHomoclinicPlot]

Algorithm 14: The Mathematica code used to create Figure 13.6.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 426 — #442

i i

426 Classical Mechanics: A Computational Approach

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 428 — #444

i i

428 Classical Mechanics: A Computational Approach

Figure 13.7: The phase portrait of the damped double-well (13.3.22).

Figure 13.8: The stream plot created by Algorithm 16.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 429 — #445

i i

Nonlinear Systems 429

13.3.4 Bifurcations of Fixed Points

An interesting phenomenon that can occur in nonlinear systems is the so-called bifurcation
of fixed points. If we think of the parameters of a nonlinear system as being variables, we
see that the location and stability of fixed points (and therefore equilibria) can change.
Consider again the damped double-well potential,

ẋ =y
)
(13.3.24)
ẏ =kx − x3 − by
The fixed points in (13.3.24) occur at the origin and at,
r !
k
(x∗ , y ∗ ) = ± ,0 (13.3.25)

If we think of k and as parameters that can be varied, then we see that the location
of the fixed points in (13.3.25) can change. For example, as k → 0 the fixed points (13.3.25)
collide at the origin. This collision is an example of a bifurcation and it tells us the condition
of which two stable equilibria can exist in the system. In this case, we use the term stable
equilibrium in the sense of Chapter 5, a local minimum in the potential. Furthermore, the
eigenvalues of the Jacobian matrix evaluated at (13.3.25) are,
1 p
λ= −b ± b2 − 8k (13.3.26)
2
When b > 0, we see that the fixed points located at (13.3.25) are stable spirals. However,
if b → 0, the previously fixed points become centers. This is an example of another type of
bifurcation where the fixed point changes stability. There are many types of bifurcations
between fixed points and a complete overview of them is beyond the scope of this book.
However, the interested reader should consult [Strogatz(2014), Hilborn(2001), Enns(2011)].
In the next paragraph, we will briefly mention two important bifurcations.
One of the most common bifurcations is the so-called saddle-node bifurcation. In a
saddle-node bifurcation, a saddle point and a stable node collide and both points disappear
from the phase portrait. Physically speaking, the system goes from having two equilibria,
one stable and one unstable, to having no equilibrium points (sometimes there is another
attracting structure elsewhere in the phase portrait). Another common bifurcation of fixed
points is the so-called transcritical bifurcation. In a transcritical bifurcation, a saddle and a
stable node collide, and pass through each other. However, as the collision occurs, the fixed
points exchange stabilities, so the saddle becomes a stable node and the stable node becomes
a saddle. Physically speaking, the equilibria exchange stabilities, but both equilibrium states
continue to exist after the bifurcation. Of course, for each bifurcation, the equilibrium state
is changing because the fixed point is moving, therefore changing the position and speed of
the equilibrium state.
We might imagine, that if we were designing a system we would want to know of possible
bifurcations in your system in order to prevent problems such as a desirable stable equi-
librium suddenly becoming unstable. For example, we might want to engineer an airplane
wing that stays relatively flat during flight. All airplane wings move a little during flight.
However, we would want to avoid the flat state from becoming unstable and the system
undergoing galloping, where the amplitude of oscillation of the wing gets large. A bifurca-
tion in which the galloping state becomes stable would be a dangerous situation, as can be
seen in videos of the famous Tacoma Narrows Bridge collapse.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 430 — #446

i i

430 Classical Mechanics: A Computational Approach

Next, we are going to study more complicated behaviors unique to nonlinear systems,
namely limit cycles and chaos.

13.4 LIMIT CYCLES

We are now going to eliminate the restriction of small amplitude oscillations and allow
for larger amplitudes. So far we have learned that nonlinear oscillators can have simple,
sinusoidal-like oscillations, just like the harmonic oscillator. However, as the amplitude of
oscillation gets larger, the period begins to depend on the amplitude. In some nonlinear sys-
tems, the oscillations can get complicated and look nothing like a sinusoid. These nonlinear
oscillations appear in the phase portrait as a limit cycle, an isolated closed-loop trajectory.
The limit cycle’s isolation makes them different from centers, which have multiple closed
orbit trajectories around the fixed point. Like fixed points, limit cycles can be stable, unsta-
ble, and even take on saddle-like stability. Unfortunately, the stability of limit cycles is
beyond the scope of this text, but [Strogatz(2014)] provides a good introduction and we
will provide some other references at the end of this chapter.

13.4.1 The Duffing Equation

Recall that in Chapter 5, we made some assumptions about the potential associated with
a restoring force. Namely, the potential needed to have a local minimum at the equilibrium
position x0 , which in this Chapter we will set to 0. We then found that if we considered
small amplitude deviations from the equilibrium, then we could approximate the potential
energy using a parabola, V = 1/2kx2 , which results in Hooke’s Law, F = −kx. All of this
was done by Taylor expanding the potential energy about the equilibrium. If necessary, go
back to Chapter 5 and reread that section. Because we are now looking at larger amplitude
oscillations, we will need to include more terms in the Taylor expansion of the potential
energy V (x) about the equilibrium at the origin,

1 d2 V 1 d3 V 1 d4 V

dV
V (x) ≈ V (0) + x+ x +
2
x +
3
x4 + · · · (13.4.1)
dx x=0 2! dx2 x=0 3! dx3 x=0 4! dx4 x=0

We have already made the argument that dV dx = 0 at the equilibrium, and that argument
continues to hold for the large amplitude case. Next, we know that the derivatives produce
a constant value when evaluated at the origin, so we will rewrite (13.4.1) in the form:
1 2 1 3 1 4
V (x) ≈ V (0) + kx + αx + δx + · · · (13.4.2)
2! 3! 4!
Next, we apply a restriction to the force which is common in physics, even in nonlinear
systems, that V (x) is symmetric about the equilibrium. Hence, the potential energy of the
object at a position x is the same as that at a position −x. In other words, we would not
expect the oscillator to have an energy that is dependent upon which side of equilibrium it
is on. In order to satisfy this restriction, we must require that α = 0. We can then find F
by calculating F = −dV /dx to get the force:
δ
F (x) = −kx − x3 (13.4.3)
6
We can then define a new parameter, = δ/6. The force becomes:

F (x) = − kx − x3 (13.4.4)

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 431 — #447

i i

Nonlinear Systems 431

0.8 1.4
d = 0.68
0.6 1.2

0.4 1.0

0.2 0.8
x'

x
Out[ ]=
0.0 0.6

-0.2 0.4

-0.4 0.2

-0.6 0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 4500 4520 4540 4560 4580 4600
x t

Figure 13.9: The phase portrait (left) and x(t) (right) of (13.4.6) with ci = 1, for all i, ω = 1,
and d = 0.68.

The resulting equation of motion is,

mẍ + kx + x3 = 0 (13.4.5)

using F (x) = mẍ. We see that the result of examining large amplitude oscillations is that
we must include a cubic nonlinearity in our equation of motion. If we include a damping
term and a drive term, the new net force acting on the system is mẍ = Fnet (x) = −kx −
x3 − β ẋ + f cos(ωt), and our equation of motion becomes:

ẍ + c1 ẋ − c2 x + c3 x3 = d cos(ωt) (13.4.6)
where c1 = β/m, c2 = k/m, c3 = /m, and d = f /m. Equation (13.4.6) is called the Duffing
equation and is a standard model equation used in nonlinear dynamics to study a wide range
of nonlinear phenomena. With the right choice of parameters, ci and d, the Duffing equation
can be made to display a large variety of nonlinear behaviors. Because the behavior of the
Duffing equation is so well known, it is often used as a model system to test new algorithms
for analyzing nonlinear systems. In the next subsections, we will use the Duffing equation
to demonstrate limit cycles, period-doubling bifurcations, and chaos.

13.4.2 Limit Cycles and Period Doubling Bifurcations

To demonstrate limit cycles and what is called period-doubling bifurcations, we will use
equation (13.4.6) with ci = 1, for all i, ω = 1, and vary the value of d. In each case we will
solve the Duffing equation with initial conditions x(0) = 0 and ẋ(0) = 0 and plot the phase
portrait along with a short segment of x(t). All units can be considered to be SI units.
Figures 13.9, 13.10, and 13.11 show some sample solutions of (13.4.6) for various values
of d, after the transients have decayed and the system settled onto its attracting limit cycle.
Notice that in Figure 13.9 when d = 0.68, the Duffing equation produces a limit cycle, an
isolated closed trajectory corresponding to an oscillation. The solution, x(t) (top right),
may look sinusoidal, but it is not, notice how narrow the peaks and troughs are compared
to a sinusoid. Furthermore, the limit cycle is not circular, and sinusoidal oscillations are the
result of circular trajectories in the phase portrait.
The graphs of Figure 13.10 shows the phase portrait (left) and solution (right) when d
is increased to 0.69. The limit cycle appears to intersect itself, however it does not. The
apparent intersection is due to a three-dimensional phase portrait being projected into two

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 432 — #448

i i

432 Classical Mechanics: A Computational Approach

0.8 1.4
d = 0.69
0.6 1.2

0.4 1.0

0.2 0.8
x'

x
Out[ ]=
0.0 0.6

-0.2 0.4

-0.4 0.2

-0.6 0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 4500 4520 4540 4560 4580 4600
x t

Figure 13.10: The phase portrait (left) and x(t) (right) of (13.4.6) with ci = 1, for all i,
ω = 1, and d = 0.69.

1.5
0.8
d = 0.75
0.6
1.0
0.4
0.2
x'

Out[ ]= 0.0 0.5

-0.2
-0.4
0.0
-0.6
0.0 0.5 1.0 1.5
x 4500 4520 4540 4560 4580 4600
t

Figure 13.11: The phase portrait (left) and x(t) (right) of (13.4.6) with ci = 1, for all i,
ω = 1, and d = 0.75.

dimensions. Although beyond the scope of this Chapter, the Duffing equation is actually a
three-dimensional system due to the explicit dependence of time in this system. The solution
for d = 0.69 is called a period-2 solution because, loosely speaking, the oscillator has to go
through two oscillations before returning to its initial state, once the oscillator has settled
onto the limit cycle behavior.
Figure 13.11 shows the phase portrait (left) and solution (right) when d is increased
to 0.75. There are even more apparent intersections in the phase portrait and the solution
x(t) appears even more complicated. Figure 13.11 illustrated a period-4 solution to (13.4.6).
We are seeing an interesting phenomenon that can occur in nonlinear oscillators called a
period-doubling bifurcation, whereas the parameter d in (13.4.6) varies, the period of the
oscillator (and therefore the limit cycle) doubles. One method of illustrating period-doubling
bifurcations is by using something called a Poincaré section, which we will discuss below.
The Poincaré section samples the solution once a period and then plots the phase space.
Since the drive frequency is ω = 1, we can sample x(t) and ẋ(t) at times t = 2nπ, where
n is an integer. The code for producing Poincaré sections is shown below in Algorithm 17.
The variable, psdata, stores a table of sampled solutions, every n2π timesteps. The First
command is used because Mathematica introduces some extra curly brackets when this
method of solution sampling is used.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 433 — #449

i i

Nonlinear Systems 433

psdata = Table[{First[x[t]/.soln], First[x0 [t]/.soln]}, {t, 700, 50000, 2π}];

poinecareSection = ListPlot[psdata, Frame → True, Axes → False, FrameLabel →

{x, “x’”}, BaseStyle → {FontSize → 18}, ImageSize → img, PlotLabel →
“Poincaré Section”];
Algorithm 17: Mathematica code for creating a Poincaré section for the Duffing equa-
tion.

0.75 0.75
d = 0.68 d = 0.69
0.5 0.5

0.25 0.25

x'
x'

0 0
-0.25 -0.25
-0.5 ● -0.5 ●
●
0 0.3 0.6 0.9 1.2 0 0.3 0.6 0.9 1.2
x x
Out[ ]=

0.75
d = 0.75
0.5
0.25
x'

●
0
-0.25 ●
●
-0.5 ●
0 0.3 0.6 0.9 1.2
x

Figure 13.12: The Poincaré sections for the solutions shown in Figures 13.9 through 13.11.

The Poincaré sections for the solutions in Figures 13.9 through 13.11 are shown in Figure
13.12.
Notice how in Figure 13.12, for d = 0.68, there is only one point. That is because, after
settling onto the limit cycle, the solution repeats its state once every period. Notice that we
are not sampling the solution at its maximum; the x-coordinate of the point in the Poincaré
section does not correspond to the maximum value of the top right graph in Figure 13.9.
The fact that the point in the Poincaré section does not match either the maximum or
minimum value of the solution does not matter. What matters is that every t = 2nπ, the
system returns to the same state. However, we see that when d = 0.69, the oscillator appears
to jump between two points in its Poincaré section. It takes 4π = 2 × 2π, or twice the period
to repeat its state. Continuing this reasoning, we see that when d = 0.75, the system is
exhibiting period-4 behavior.
As mentioned previously, Figures 13.9 through 13.12 are demonstrating period-doubling
bifurcations. We see a period-doubling bifurcation in the limit cycle between d = 0.68 and
d = 0.69. We see another period-doubling bifurcation between d = 0.69 and d = 0.75. As we
continue to increase d, we would see several more period doubling bifurcations, the series
of which is called a period-doubling cascade, which will eventually lead to chaotic behavior

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 434 — #450

i i

434 Classical Mechanics: A Computational Approach

Phase Space
x(t)
1.0
1.5
0.5 1.0
0.5
x'

0.0 0.0

x
Out[ ]=

-0.5
-0.5 -1.0
-1.5
-1.0
4500 4600 4700 4800 4900 5000
-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 t
x
Poincaré Section

0.5

0.0
x'

Out[ ]=

-0.5

-1.0
-1.0 -0.5 0.0 0.5 1.0 1.5
x

Figure 13.13: Plots of the phase portrait, solution, and Poincaré section of (13.5.1).

once d = 0.8. However, to illustrate chaos, we will change the parameters used in (13.4.6)
in order to better visualize the attractor corresponding to chaotic behavior.

13.5 CHAOS
In addition to interesting oscillations, nonlinear systems can also display a behavior called
chaos. To explore the phenomenon of chaos, we will use the Duffing equation, but with a
specific set of parameter values:

ẍ + 0.15ẋ − x + x3 = 0.3 cos(t) (13.5.1)

It is worth mentioning that there are many different values of ci and d from (13.4.6) that
will produce chaotic behavior in the Duffing equation, including those used in Section 13.4.2
with d = 0.8. However we use the ones used in (13.5.1) because they provide a good view
of the attractor associated with the chaotic dynamics. let us start by plotting the solution,
phase portrait, and Poincaré section of the solution to (13.5.1) with initial conditions of
x(0) = 0 and ẋ(0) = 0 and allow transients to decay. The results are shown in Figure 13.13.
Notice how in Figure 13.13, the phase portrait does not contain a limit cycle, but some
unusual trajectory that seems to oscillate about each of the stable equilibrium in the double-
well potential. Those oscillations are clear in the graph of x(t), where the solution oscillates
about the equilibrium at x = 1, before switching and oscillating about x = −1. Notice
that the duration of time spent near each equilibrium varies and the switching between
one equilibrium to the next seems unpredictable. This is strange behavior! Like in Section
13.4.2, we can use a Poincaré section to understand the structure of the attractor. Notice

i i

SHO, Difference = 0.1 Duffing Equation, Difference = 0.1

1.0 1.5
1.0
0.5
0.5
0.0 0.0
x

x
-0.5
-0.5
-1.0

-1.0 -1.5
0 10 20 30 40 50 0 20 40 60 80 100
t t
Out[ ]=

Duffing Equation, Difference = 0.01 Duffing Equation, Difference = 0.00001

1.5 1.5
1.0 1.0
0.5 0.5
0.0 0.0
x

x
-0.5 -0.5
-1.0 -1.0
-1.5 -1.5
0 20 40 60 80 100 0 20 40 60 80 100
t t

Figure 13.14: Plots comparing similar initial conditions for the simple harmonic oscillator
(SHO) and the Duffing equation. The value of Difference in the title of each plot descibes
the difference in the inital positions of the two oscillators. The red (in e-book) curve has
an initial position of x(0) = 0 and the blue (in e-book) curve has an initial position of
x(0) = 0 + Difference. In all cases ẋ(0) = 0.

In this Chapter, we discussed that nonlinear systems can take on a wide variety of
behaviors. To be honest, we have not even scratched the surface on the variety of interest-
ing phenomena that can occur in nonlinear systems. As promised, we will include a few extra
suggestions on how to extend the reader’s study of nonlinear systems. We recommend start-
ing with [Strogatz(2014)], as the book is easily understandable and enjoyable to read. From
there, one can take many pathways to learning nonlinear systems. A book we frequently
recommend to students is [Enns(2011)], which contains some very interesting applications of
nonlinear systems to a wide variety of fields in the natural and social sciences. Another book
commonly used in nonlinear dynamics courses is [Hilborn(2001)]. If the reader is interested
in analyzing data measured from nonlinear systems, both [Kantz and Schreiber(2004)] and
[Abarbanel(1996)] are recommended. Finally, if the rader is interested in developing models
of systems (especially complex systems), we recommend [Wilensky and Rand(2015)].
If one spends enough time working with (and learning about) nonlinear systems, even-
tually, one will come to start thinking of problems not as linear versus nonlinear, or even as
physics versus not physics. Instead, one will likely just see the system you are working with
as just that, an interesting system to be analyzed with one of many tools the reader will
have in his/her mathematical and computational toolbox. In the end, breaking up problems
in the natural world by discipline is a human-imposed construct. Real systems are neither
biology versus physics versus sociology, they are just systems to be studied. While the con-
struct of silos for problems is useful for the purposes of education, once one no longers need
it, we recommend discarding it and focus on studying problems that are interesting to you.

13.7 CHAPTER SUMMARY

In this chapter, we studied systems whose equations of motion have nonlinear terms. These
equations have the form

ẋ =f (x, y)
ẏ =g(x, y)
Although the equations governing nonlinear systems are not usually solvable in closed
form, we can use tools such as phase portraits to understand possible steady state behaviors
of the system as well as identify the stability of equilibria.
The fixed points of the system can be found by solving

f (x∗ , y ∗ ) =0
g(x∗ , y ∗ ) =0
Although we cannot solve the whole system analytically, we can linearize the system
about the fixed points, which results in an easily solvable system of equations:
   ∂f ∂f   
u̇ ∂x ∂y u
 =  
 
v̇ ∂g ∂g v
∂x ∂y

where the eigenvalues of the Jacobian matrix A give us information about the stability of
the equilibrium. The eigenvalues can be found using the characteristic polynomial,

det (A − λ1) = 0

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 439 — #455

i i

Nonlinear Systems 439

In nonlinear systems, the equilibrium states can change (both location in phase space and
stability).
Finally, we also explored a behavior called chaos, which is unique to nonlinear systems.
A hallmark of chaos is that the solutions of a chaotic system display sensitive dependence
on initial conditions. In other words, two initially similar states will evolve to very different
behaviors, thus inhibiting one’s ability to make long-term predictions on chaotic systems.

13.8 END-OF-CHAPTER PROBLEMS

Section 13.1: Linear versus Nonlinear Systems
1. Consider the differential equation

d3 x
+x = 0
dt3

Is this equation a linear or a nonlinear differential equation? Justify your answer. Solve
the equation for x(t).
2. Consider the differential equation

d3 x
+ t3 x = 0
dt3

Is this equation a linear or a nonlinear differential equation? Justify your answer.

3. Consider the differential equation

ẍ = t2 x − tx2

Suppose you have found two solutions to this equation, x1 (t) and x2 (t). Is it true that
c1 x1 + c2 x2 is also a solution? Prove your answer.
Section 13.2: The Damped Harmonic Oscillator, Revisited
4. Using a computer, plot the phase portrait for the damped harmonic oscillator with
γ < 0 and ω02 = 1. You may use x(0) = 1 and ẋ(0) = 0. Do you get what you expect?
Why or why not? What does it mean for γ < 0?
5. Using a computer, plot the phase portrait for the damped pendulum, ẍ + ω02 sin x +
γ ẋ = 0 for ω02 = 1 and three values of γ, less than, equal to, and greater than ω02 . For
each case use x(0) = 1 and ẋ(0) = 0. Discuss your results.
6. Repeat Problem 5 using x(0) = 3 and ẋ(0) = 0. Discuss your results and compare them
to that of Problem 5.
Section 13.3: Fixed Points and Phase Portraits

7. Consider a one-dimensional system, ẋ = f (x). We can assign a potential function to

i i

Nonlinear Systems 441

The nonlinear term J(1 − |J|) is sometimes called a repair nonlinearity [Sprott(2004)]
and the number 1 in the nonlinear term is a measure of when Juliet’s love becomes
counter productive. Plot the phase portrait for this system and discuss the nature of
the relationship between Romeo and Juliet.
15. Find the Jacobian matrix for (13.3.24) and compute its eigenvalues for each of the
fixed points.
16. The Lotka-Volterra model of competition is a mathematical model for two species X
and Y who compete for the same resources. Let x be the population size of species X
and y be the population size of species Y , then one choice of a Lotka-Volterra model
parameters would be:

ẋ =4x − 3x2 − xy
ẏ =2y − 4y 2 − 3xy

For this model, identify the fixed points and their stability. Can the two species ever
coexist? Note that this kind of equation-based model tends to work well for large
populations. For small populations, agent-based models tend to be more useful. For a
discussion of agent-based versus equation-based modeling, the interested reader should
consult [Wilensky and Rand(2015)]
17. Plot the phase portrait for the Lotka-Volterra competition model in Problem 16.
Discuss the final state of the system for various initial conditions.
18. Lotka-Volterra equations can also be used to describe predator-prey models where
species Y (predator) eats species X (prey). Consider the predator-prey model

ẋ =3x − 7xy
ẏ = − 2y + 5xy

where like the competition model, we will assume x(t) and y(t) measure the number
of each species in the hundreds. Find the fixed points of this predator-prey model and
identify their stability. Comment on the nature of the equilibrium state corresponding
to each fixed point.
19. Draw the phase portrait for the predator-prey model in Problem 18.

20. The SIS (susceptible-infected-susceptible) disease transmission model can be used for
diseases that do not remove individuals from the population, either through immunity
or death. SIS models are sometimes appropriate for some bacterial diseases such as
streptococcal pharyngitis (also known as sore throat). Consider two populations of
individuals s(t) and i(t), representing the number of susceptible and infected individ-
uals respectively. The SIS model is:

ṡ =µn − µs + γi − βsi
i̇ = − µi − γi + βsi

where the coefficients represent rates and n = s + i is the total population. Comment
on the physical meaning of each coefficient: µ,γ, and β. Find the fixed points for the
system and identify their stability. Note that negative values of s and i are meaningless.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 442 — #458

i i

442 Classical Mechanics: A Computational Approach

Section 13.4: Limit Cycles

21. Show that the Duffing equation, (13.5.1), undergoes a series of period doubling bifur-
cations by changing the amplitude of the cosine term on the right-hand side of the
equation.
22. The logistic map, xn+1 = rxn (1 − xn ), is a discrete form of the logistic equation (see
Problem 9). After choosing an initial value x0 and parameter r, it is easy to compute
successive values. For example, x1 = rx0 (1 − x0 ) and x2 = rx1 (1 − x1 ), and so on.
Using x0 = 0.1, plot xn vs. n for r = 2.9, 3.2, 3.5, and 3.55. Identify the periodicity of
each solution. You can produce a Poincaré section by plotting xn+1 vs. xn . Finally,
try r = 3.91. What kind of behavior is the logistic map exhibiting when r = 3.91?
23. Weakly driven nonlinear oscillators can be excited to large amplitudes through a
phenomenon called autoresonance. The oscillator is driven by a periodic forcing
term whose amplitude is small, but its frequency changes (typically decreasing) with
time. As the drive frequency decreases through the oscillator’s resonant frequency,
the oscillator becomes phase-locked with the driving force, resulting in the oscillator
maintaining resonance even though the drive frequency is no longer the oscillator’s
natural frequency. Consider the following driven pendulum:

ẍ + ω02 sin(x) = cos ω0 t − 1/2αt2

Notice that the drive frequency (the time derivative of the drive phase) is fd = ω0 − αt,
so that at t = 0, the drive frequency is equal to the natural frequency of the pendulum.
Starting with initial conditions, x(−1000) = 0 and ẋ(−1000) = 0, compute x(t) for
= 0.0459 and = 0.0461 for the pendulum with ω0 = 2π, α = 0.001. For which value
of is the pendulum displaying autoresonance? For more information on autoresonance
and the pendulum, check out [Fajans and Friédland(2001)].
24. Electrical circuits provide excellent examples of nonlinear systems. One of the most
famous circuit-inspired equations is the so-called Van der Pol equation

ẍ − 1 − x2 ẋ + x = 0

which is a simple harmonic oscillator with a nonlinear damping term. Notice that the
sign of the damping actually changes with the value of x. For example if > 0, then
the damping is negative for x < 1 and is positive for x > 1. Thus small oscillations
will grow due to negative damping, but once the oscillations are large enough, they
will dampen out as the damping term becomes positive. The Van der Pol oscillator
can also display so-called relaxation oscillations, a common form of oscillations in
nonlinear systems where rapid changes of x are followed by slower variations. To see
relaxation oscillations in action, solve the Van der Pol oscillator for = 8.0. Plot both
x(t) and the phase portrait. Describe the solution.
25. It can sometimes be convenient to describe limit cycles using polar coordinates.
Consider the following system in polar coordinates:

ṙ =r(r − 1)(r − 2)
θ̇ =1

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 443 — #459

i i

Nonlinear Systems 443

where r and θ are the usual polar coordinates with r ≥ 0. Notice that one can find
values of r that make ṙ = 0, but there are no values of θ or r for which θ̇ = 0. That
is OK! Give a description of the phase portrait in this case. There are limit cycles in
this system. Where are they and what is their stability? The stability can be found
by using the method outlined in Problem 9.
Section 13.5: Chaos
26. The Lorenz equations:

ẋ =σ(y − x)
ẏ =rx − y − xz
ż =xy − bz
display chaotic behavior with σ = 10, r = 28, and b = 8/3. Plot the phase portrait (x,
y, z) for the Lorenz equations.
27. The Lorenz equations in Problem 26 are a chaotic system. Show that they demonstrate
sensitive dependence on initial conditions.
28. Chua’s circuit is a electrical circuit with a piecewise linear negative resistance NR . It
is one of the simplest circuits to exhibit a variety of nonlinear behaviors including a
period-doubling route to chaos. The circuit schematic is shown in Figure 13.15. Let
x, y, and z be proportional to the voltages across the capacitors C1 and C2 , and the
current in the inductor L, respectively. The equations governing the circuit can be
obtained from Kirchoff’s rules and simplified to be:
ẋ =α(y − x − g(x))
ẏ =x − y + z
ż = − βy

where g(x) = m1 x + 12 (m0 − m1 ) (|x + 1| − |x − 1|) with α = C2 /C1 , β = R2 C2 /L,

and m0 and m1 being the slopes of the inner and outer segments of the voltage-
current curve, respectively. For more information the interested student should con-
sult [Enns(2011)]. Create a phase portrait (x-y-z space) for the equations of motion
using α = 15.6, β = 25.58, m0 = −8/7, m1 = −5/7. The result will be the famous
double-scroll attractor.

R
NR
L C1 C2

Figure 13.15: Problem 13.28: Chuas’s circuit.

29. A period-doubling route to chaos can be observed in the equations for Chua’s circuit
(see Problem 28) by using the same parameters as used in Problem 28, but with
β = 50, 35, 33.8, etc. For each value of β, create a graph of x(t) and identify the
periodicity of each solution. For what value of β is there a period-8 solution?

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 444 — #460

[MW()] Merriam-Webster online dictionary. https://www.merriam-webster.com/.

Accessed: 2018-02-18.
[MIT(2018)] Massachutsetts institute of technology open course ware. https://ocw.mit.
edu/index.htm, 2018. Accessed: 2018-05-31.
[MMA(2018)] Wolfram language and system documentation center. https://reference.
wolfram.com, 2018. Accessed: 2018-05-31.
[Abarbanel(1996)] HDI Abarbanel. Analysis of Observed Chaotic Data. Springer, New
York, 1996.
[Arnold(1997)] V.I Arnold. Mathematical Methods of Classical Mechanics. Springer, New
York, 2nd edition, 1997.
[de Lange and Pierrus(2010)] OL de Lange and J Pierrus. Solved Problems in Classical
Mechanics. Oxford, New York, 2010.
[Ellington(2003)] AJ Ellington. A meta-analysis of the effects of calculators on students’
achievement and attitude levels in precollege mathematics classes. Journal for Research
in Mathematics Education, 34(5):433–463, 2003.
[Enns(2011)] RH Enns. It’s a Nonlinear World. Springer, New York, 2011.
[Fajans and Friédland(2001)] J Fajans and L Friédland. Autoresonant (nonstationary) exci-
tation of pendulums, plutinos, plasmas, and other nonlinear oscillators. American
Journal of Physics, 69:1096–1102, 2001.
[Goldstein et al. (2001)] H Goldstein, CP Poole, and JL Safko. Classical Mechanics. Pear-
son, London, 3rd edition, 2001.
[Hadhazy(2001)] A Hadhazy. Fact or fiction: The days (and nights) are getting longer.
Scientific American, 6 2001.
[Hamilton(1834)] WR Hamilton. On a general method in dynamics; by which the study of
the motions of all free systems of attracting or repelling points is reduced to the search
and differentiation of one central relation, or characteristic function. Philosophical
Transactions of the Royal Society, Part 2, 247–308, 1834.
[Hamilton(1835)] WR Hamilton. Second essay on a general method in dynamics. Philo-
sophical Transactions of the Royal Society, Part 1, 95–144, 1835.
[Hilborn(2001)] R Hilborn. Chaos and Nonlinear Dynamics: An Introduction for Scientists
and Engineers. Oxford University Press, New York, 2nd edition, 2001.

445

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 446 — #462

i i

446 Bibliography

[Kantz and Schreiber(2004)] H Kantz and T Schreiber. Nonlinear Time Series Analysis.
Cambridge University Press, Cambridge, UK, 2nd edition, 2004.
[Kinder and Nelson(2015)] JM Kinder and P Nelson. A Student’s Guide to Python for
Physical Modeling. Princeton University Press, Princeton, NJ, 2015.
[Lorenz(1963)] E Lorenz. Deterministic nonperiodic flow. Journal of Atmospheric Sciences,
20:130 – 141, 1963.
[Morin(2008)] D Morin. Introduction to Classical Mechanics with Problems and Solutions.
Cambridge, Cambridge, 2008.
[Noether(1918)] E Noether. Invariante variationsprobleme. Nachrichten von der
Gesellschaft der Wissenschaften zu GÃűttingen, Mathematisch-Physikalische Klasse,
1918:235–257, 1918.
[Piessens et al. (1983)] R Piessens, E. de Doncker-Kapenga, C.W. Überhuber, and D.K.
Kahaner. QUADPACK: A Subroutine Package for Automatic Integration. Springer,
New York, 1983.
[Press et al. (2007)] WH Press, SA Teukolsky, WT Vetterling, and BP Flannery. Numerical
Recipes, 3rd Edition. Cambridge University Press, Cambridge, UK, 2007.
[Sprott(2004)] JC Sprott. Dynamical models of love. Nonlinear Dynamics, Psychology, and
Life Sciences, 8(8):303 – 313, 2004.
[Strogatz(2014)] S Strogatz. Nonlinear Dynamics and Chaos: With Applications to Physics,
Biology, Chemistry, and Engineering. Westview Press, Cambridge, UK, 2 edition, 2014.
[Supiano(2018)] B. Supiano. So what are you going to do with that degree? physics majors
get that question, too. The Chronicle of Higher Education, 9 2018.
[Taylor(2005)] JR Taylor. Classical Mechanics. University Science Books, Sausalito, CA,
2005.
[Thornton and Marion(2004)] ST Thornton and JB Marion. Classical Dynamics of Parti-
cles and Systems. Thomson-Brooks/Cole, Belmont, CA, 5 edition, 2004.
[Tokpah(2008)] C.L. Tokpah. The Effects of Computer Algebra Systems on Students’
Achievement in Mathemaics. dissertation, Kent State University, 2008.
[Wilensky and Rand(2015)] U Wilensky and W Rand. An Introduction to Agent-Based
Modeling. MIT Press, Cambridge, MA, 2015.

[Worthington(2012)] J Worthington. A Study of the Planar Circular Restricted Three Body

Problem and the Vanishing Twist. PhD thesis, University of Sydney, 10 2012.

i i

i i
i i

“master˙book˙file” — 2020/9/5 — 9:32 — page 447 — #463

i i

Index

6-12 potential, 275 boundary conditions, 32

boundary value problem, 32
acceleration, 3 brachistochrone, 216
definition of, 5 Bulirsch-Stoer method, 244
acceleration due to gravity, 34 burn rate, 112
effective, 323
acceleration vector C, 20
Cartesian coordinates, 65 C++, 20
cylindrical coordinates, 83 calculus of variations, 209, 210
polar coordinates, 81 canonical equations of motion, 259
spherical coordinates, 85 canonical momentum, 256
action integral, 229 CAS, 16, 20
Adams Methods, 410 center of mass, 123, 277, 338, 339
adaptive step size control, 248 continuous mass distribution, 114
air resistance, 39 point particles, 113
angular acceleration, 3, 81, 336 central force, 10, 100, 273
angular momentum, 125, 335 and torque, 128
conservation, 129 definition of, 273
conservation of for central forces, 275 centrifugal force, 286, 302, 311, 320
magnitude for two-body problem, 279 direction of, 321
of a rigid body, 342 centrifugal potential energy, 286
of the two-body problem, 278 Chandler, Seth, 361
particle off xy-plane, 337 chaos, 1, 407, 408, 434
system of particles, 126, 340 chaotic systems, 2
angular position, 3 definition of, 437
angular velocity, 3, 78, 314, 335 characteristic equation, 352, 385, 389
vector, 313 characteristic polynomial, 415
antisymmetric mode of oscillation, 377, 383 Chua’s circuit, 443
double pendulum, 390 classical mechanics, 1
aphelion, 288 fundamental question, 2
arc length, 64 classical physics, 1
array, 18 closed form solution, 14
Atwood machine, 236 Code Academy, 21
autoresonance, 442 cofactor expansion, 73
azimuthal angle, 84 collision
elastic, 109
basic descriptors of motion, 3, 11 inelastic, 108
basin of attraction, 419, 426 compiled language, 20
Bernoulli, Jakob, 209 examples of, 20
Bernoulli, Johann, 209 compiler, 20
bifurcation, 407, 408, 429, 444 complex conjugate, 166
body cone, 360 complex systems, 22
body coordinate frame, 357 computation, 13

Cartesian coordinates, 63 weakly coupled oscillators, 379
cylindrical coordinates, 83 work, 135, 146
polar coordinates, 77, 81 Work-Kinetic Energy Theorem, 136, 150
spherical coordinates, 85
Virial Theorem, 161, 168 Yukawa potential, 275
volume density
mass integral, 114 zodiac constellation, 357

i i

Kinetic AppStudioExtensionsUserGuide
No ratings yet
Kinetic AppStudioExtensionsUserGuide
144 pages
Statistical Physics
From Everand
Statistical Physics
Gregory H. Wannier
3.5/5 (3)
Mathematics 9 Curriculum Mapdocx
No ratings yet
Mathematics 9 Curriculum Mapdocx
12 pages
MPLS TP Overview
100% (1)
MPLS TP Overview
30 pages
Special Functions of Mathematical (Geo-) Physics - Gutting
100% (1)
Special Functions of Mathematical (Geo-) Physics - Gutting
505 pages
Gill
No ratings yet
Gill
474 pages
Introduction To C Programming Course Materail
100% (1)
Introduction To C Programming Course Materail
161 pages
Formulation and Evaluation of Topical Herbal Gel For The Treatment
No ratings yet
Formulation and Evaluation of Topical Herbal Gel For The Treatment
16 pages
Introduction To SR and GR by Hans Stephani PDF
80% (5)
Introduction To SR and GR by Hans Stephani PDF
418 pages
Wave Motion As Inquiry: Fernando Espinoza
No ratings yet
Wave Motion As Inquiry: Fernando Espinoza
237 pages
Numerical Methods Scientists Engineers: For and
100% (1)
Numerical Methods Scientists Engineers: For and
11 pages
Types of Machine Learning
No ratings yet
Types of Machine Learning
24 pages
(MA 262) Differential Equations and Linear Algebra 3e - Goode Annin PDF
100% (1)
(MA 262) Differential Equations and Linear Algebra 3e - Goode Annin PDF
805 pages
Bartels2015 Book NumericalMethodsForNonlinearPa
100% (2)
Bartels2015 Book NumericalMethodsForNonlinearPa
394 pages
Ordinary Dierential Equations Principles and Applications
100% (1)
Ordinary Dierential Equations Principles and Applications
349 pages
Differential Equations A Linear Algebra Approach (Anindya Dey)
100% (2)
Differential Equations A Linear Algebra Approach (Anindya Dey)
626 pages
Office Automation
No ratings yet
Office Automation
14 pages
Mathematical Methods in Physics Samuel D. Lindenbaum PDF
100% (1)
Mathematical Methods in Physics Samuel D. Lindenbaum PDF
475 pages
Topological Vector Spaces and Distributions
From Everand
Topological Vector Spaces and Distributions
John Horvath
3/5 (2)
Partial Differential Equations Analytical and Numerical Methods
80% (5)
Partial Differential Equations Analytical and Numerical Methods
638 pages
Formwork (Shuttering) For Different Structural Members - Beams, Slabs Etc
No ratings yet
Formwork (Shuttering) For Different Structural Members - Beams, Slabs Etc
6 pages
Lagrangian and Hamiltonian Mechanics - M. G. Calkin PDF
78% (9)
Lagrangian and Hamiltonian Mechanics - M. G. Calkin PDF
221 pages
Abell - Differential Equations With Mathematica
100% (2)
Abell - Differential Equations With Mathematica
641 pages
(Cambridge IISc Series) A. K. Nandakumaran, P. S. Datti - Partial Differential Equations - Classical Theory With A Modern Touch (Cambridge IISc Series) - Cambridge University Press (2020)
No ratings yet
(Cambridge IISc Series) A. K. Nandakumaran, P. S. Datti - Partial Differential Equations - Classical Theory With A Modern Touch (Cambridge IISc Series) - Cambridge University Press (2020)
377 pages
(Advances in Applied Mathematics) Chow, Pao Liu-Stochastic Partial Differential Equations-CRC Press (2014)
100% (1)
(Advances in Applied Mathematics) Chow, Pao Liu-Stochastic Partial Differential Equations-CRC Press (2014)
333 pages
A. Wipf - Path Integrals
No ratings yet
A. Wipf - Path Integrals
164 pages
Classical Mechanics MIT
100% (1)
Classical Mechanics MIT
199 pages
Quantum Theory of Many-Particle Systems
From Everand
Quantum Theory of Many-Particle Systems
Alexander L. Fetter
4.5/5 (2)
Alexander Altland, Jan Von Delft - Mathematics For Physicists - Introductory Concepts and Methods (2019, Cambridge University Press)
100% (2)
Alexander Altland, Jan Von Delft - Mathematics For Physicists - Introductory Concepts and Methods (2019, Cambridge University Press)
721 pages
Derivations of Applied Mathematics
No ratings yet
Derivations of Applied Mathematics
664 pages
Logan Applied Mathematics Solution Manual
100% (2)
Logan Applied Mathematics Solution Manual
63 pages
Encyclopedia of Physical Geography - Congo Democratic Republic - India. Vol 2-U X L (2003)
100% (1)
Encyclopedia of Physical Geography - Congo Democratic Republic - India. Vol 2-U X L (2003)
242 pages
Hi Ac 8011 Liquid Particle Counting System Manual
No ratings yet
Hi Ac 8011 Liquid Particle Counting System Manual
36 pages
Ricci-Calculus - An Introduction To Tensor Analysis and Its Geometrical Applications
100% (5)
Ricci-Calculus - An Introduction To Tensor Analysis and Its Geometrical Applications
535 pages
gnuplot Cookbook
From Everand
gnuplot Cookbook
Lee Phillips
No ratings yet
Partial Differential Equations PDF
No ratings yet
Partial Differential Equations PDF
135 pages
VU21997 - Expose Website Security Vulnerabilities - Class 4 SQLMap Final
No ratings yet
VU21997 - Expose Website Security Vulnerabilities - Class 4 SQLMap Final
21 pages
Analytic Methods in Physics by Charlie Harper PDF
67% (3)
Analytic Methods in Physics by Charlie Harper PDF
316 pages
Second Order Partial Differential Equations: An Introduction To
No ratings yet
Second Order Partial Differential Equations: An Introduction To
301 pages
Bermant, Aramanovich - Mathematical Analysis - A Brief Course For Engineering Students
100% (1)
Bermant, Aramanovich - Mathematical Analysis - A Brief Course For Engineering Students
809 pages
Aidco 450E BR
No ratings yet
Aidco 450E BR
4 pages
15.18 Auxiliary Power Units (APUs)
No ratings yet
15.18 Auxiliary Power Units (APUs)
24 pages
Chemical Resistance Guide
No ratings yet
Chemical Resistance Guide
20 pages
Bouzidi, Rabah - Le Van, Anh - Lagrangian Mechanics - An Advanced Analytical approach-ISTE (2019) PDF
No ratings yet
Bouzidi, Rabah - Le Van, Anh - Lagrangian Mechanics - An Advanced Analytical approach-ISTE (2019) PDF
331 pages
An Introduction Into The Feynman Path Integral PDF
No ratings yet
An Introduction Into The Feynman Path Integral PDF
94 pages
32 Unit Wise Maths Formulas
No ratings yet
32 Unit Wise Maths Formulas
10 pages
(De Gruyter Stem) Dingyu Xue - Differential Equation Solutions With MATLAB - Fundamentals and Numerical Implementations (De Gruyter STEM) (2020, de Gruyter)
100% (1)
(De Gruyter Stem) Dingyu Xue - Differential Equation Solutions With MATLAB - Fundamentals and Numerical Implementations (De Gruyter STEM) (2020, de Gruyter)
454 pages
Level Iii Ut Specific Examination
No ratings yet
Level Iii Ut Specific Examination
8 pages
Lecturas No 1 Role and Scope of Industrial Engineers (1) - Recognized
No ratings yet
Lecturas No 1 Role and Scope of Industrial Engineers (1) - Recognized
14 pages
Ic Engine and Turbo Machinery
No ratings yet
Ic Engine and Turbo Machinery
25 pages
Shape Vocabulary Word Mat
No ratings yet
Shape Vocabulary Word Mat
4 pages
Programming Language - Common Lisp 8. Structures
No ratings yet
Programming Language - Common Lisp 8. Structures
10 pages
Refrigeration Unit Datasheet
No ratings yet
Refrigeration Unit Datasheet
8 pages
QFT Lecture Notes
100% (1)
QFT Lecture Notes
175 pages
Asymptotic Methods in Analysis
From Everand
Asymptotic Methods in Analysis
N. G. De Bruijn
3.5/5 (2)
Complex Variables
From Everand
Complex Variables
Francis J. Flanigan
No ratings yet
Matrix Theory and Applications for Scientists and Engineers
From Everand
Matrix Theory and Applications for Scientists and Engineers
Alexander Graham
No ratings yet
Rocker Arm Installation & Removal 0 Deutz 0312 4232 2011
No ratings yet
Rocker Arm Installation & Removal 0 Deutz 0312 4232 2011
4 pages
JAROL Assumes The Promotion of Energy-Saving Technology As Its Own Task! 1. PREFACE NOTICE
50% (2)
JAROL Assumes The Promotion of Energy-Saving Technology As Its Own Task! 1. PREFACE NOTICE
182 pages
2018 Book MethodsForPartialDifferentialE PDF
100% (4)
2018 Book MethodsForPartialDifferentialE PDF
473 pages
Classical Mechanics
No ratings yet
Classical Mechanics
427 pages
Lukong Cornelius Fai - Feynman Path Integrals in Quantum Mechanics and Statistical Physics-CRC Press (2021)
100% (1)
Lukong Cornelius Fai - Feynman Path Integrals in Quantum Mechanics and Statistical Physics-CRC Press (2021)
415 pages
Quantum Mechanics in Simple Matrix Form
From Everand
Quantum Mechanics in Simple Matrix Form
Thomas F. Jordan
3.5/5 (3)
Grade - X: English
No ratings yet
Grade - X: English
4 pages
1 Plant Nutrition
No ratings yet
1 Plant Nutrition
35 pages
Linear and Nonlinear Waves Textbook)
100% (1)
Linear and Nonlinear Waves Textbook)
334 pages
(UNITEXT For Physics) Franco Strocchi - A Primer of Analytical Mechanics-Springer International Publishing (2018)
100% (2)
(UNITEXT For Physics) Franco Strocchi - A Primer of Analytical Mechanics-Springer International Publishing (2018)
120 pages
Elementary Functional Analysis
From Everand
Elementary Functional Analysis
Georgi E. Shilov
4/5 (1)
Gregor Skačej - Primož Ziherl - Solved Problems in Thermodynamics and Statistical Physics (2019, Springer) PDF
No ratings yet
Gregor Skačej - Primož Ziherl - Solved Problems in Thermodynamics and Statistical Physics (2019, Springer) PDF
292 pages
DSPDF Formulae
No ratings yet
DSPDF Formulae
3 pages
Quantum Chemistry: The Development of Ab Initio Methods in Molecular Electronic Structure Theory
From Everand
Quantum Chemistry: The Development of Ab Initio Methods in Molecular Electronic Structure Theory
Henry F. Schaefer III
3/5 (1)
1 - Assignment - PH 401 (EE) - MODULE - 6 (Statistical Mechanics)
No ratings yet
1 - Assignment - PH 401 (EE) - MODULE - 6 (Statistical Mechanics)
2 pages
Axiomatics of Classical Statistical Mechanics
From Everand
Axiomatics of Classical Statistical Mechanics
Rudolf Kurth
5/5 (1)
Introduction to Differential Geometry for Engineers
From Everand
Introduction to Differential Geometry for Engineers
Brian F. Doolin
No ratings yet
Topological Methods in Euclidean Spaces
From Everand
Topological Methods in Euclidean Spaces
Gregory L. Naber
No ratings yet
Techniques and Applications of Path Integration
From Everand
Techniques and Applications of Path Integration
L. S. Schulman
No ratings yet
Molecular Quantum Electrodynamics
From Everand
Molecular Quantum Electrodynamics
D. P. Craig
4/5 (2)
Variational Methods for Eigenvalue Problems: An Introduction to the Methods of Rayleigh, Ritz, Weinstein, and Aronszajn
From Everand
Variational Methods for Eigenvalue Problems: An Introduction to the Methods of Rayleigh, Ritz, Weinstein, and Aronszajn
S. H. Gould
5/5 (1)
Foundations of Electrodynamics
From Everand
Foundations of Electrodynamics
Parry Moon
No ratings yet
An Introduction to the Calculus of Variations
From Everand
An Introduction to the Calculus of Variations
L.A. Pars
No ratings yet
Introduction To Ordinary Differential Equations With: Mathematica®
No ratings yet
Introduction To Ordinary Differential Equations With: Mathematica®
9 pages
Applied Functional Analysis
From Everand
Applied Functional Analysis
D.H. Griffel
No ratings yet
Cinematography: Lighting
88% (24)
Cinematography: Lighting
77 pages
Matrices and Linear Transformations: Second Edition
From Everand
Matrices and Linear Transformations: Second Edition
Charles G. Cullen
3.5/5 (2)
Barry Spain - Ordinary Differential Equations-Van Nostrand Reinhold (1969)
No ratings yet
Barry Spain - Ordinary Differential Equations-Van Nostrand Reinhold (1969)
152 pages
The Quantum Mechanics of Many-Body Systems: Second Edition
From Everand
The Quantum Mechanics of Many-Body Systems: Second Edition
D. J. Thouless
No ratings yet
International GCSE Biology (4BI1) - Grade Characteristics: Holistic Approach To Grades
No ratings yet
International GCSE Biology (4BI1) - Grade Characteristics: Holistic Approach To Grades
7 pages
Nmrws2: H O That Are Aldehydes
No ratings yet
Nmrws2: H O That Are Aldehydes
4 pages
Physical Mathematics
No ratings yet
Physical Mathematics
2 pages
Solved Problems in Classical Electromagnetism
From Everand
Solved Problems in Classical Electromagnetism
Jerrold Franklin
No ratings yet