Mathematical

Physics
LICENSE, DISCLAIMER OF LIABILITY, AND LIMITED WARRANTY

By purchasing or using this book (the “Work”), you agree that this license
grants permission to use the contents contained herein, but does not give
you the right of ownership to any of the textual content in the book or own-
ership to any of the information or products contained in it. This license
does not permit uploading of the Work onto the Internet or on a network
(of any kind) without the written consent of the Publisher. Duplication or
dissemination of any text, code, simulations, images, etc. contained herein
is limited to and subject to licensing terms for the respective products, and
permission must be obtained from the Publisher or the owner of the con-
tent, etc., in order to reproduce or network any portion of the textual mate-
rial (in any media) that is contained in the Work.
Mercury Learning and Information (“MLI” or “the Publisher”) and any-
one involved in the creation, writing, or production of the companion disc,
accompanying algorithms, code, or computer programs (“the software”),
and any accompanying Web site or software of the Work, cannot and do
not warrant the performance or results that might be obtained by using
the contents of the Work. The author, developers, and the Publisher have
used their best efforts to insure the accuracy and functionality of the textual
material and/or programs contained in this package; we, however, make
no warranty of any kind, express or implied, regarding the performance
of these contents or programs. The Work is sold “as is” without warranty
(except for defective materials used in manufacturing the book or due to
faulty workmanship).
The author, developers, and the publisher of any accompanying content,
and anyone involved in the composition, production, and manufacturing of
this work will not be liable for damages of any kind arising out of the use of
(or the inability to use) the algorithms, source code, computer programs,
or textual material contained in this publication. This includes, but is not
limited to, loss of revenue or profit, or other incidental, physical, or conse-
quential damages arising out of the use of this Work.
The sole remedy in the event of a claim of any kind is expressly limited to
replacement of the book, and only at the discretion of the Publisher. The
use of “implied warranty” and certain “exclusions” vary from state to state,
and might not apply to the purchaser of this product.
Mathematical
Physics
An Introduction

Derek Raine, PhD et al.

MERCURY LEARNING AND INFORMATION

Dulles, Virginia
Boston, Massachusetts
New Delhi
Copyright © 2019 by Mercury Learning and Information LLC. All rights
reserved.
Original title and copyright: Mathematical Methods for Physical Science by
Derek Raine et al. Copyright © 2018 Pantaneto Press. All rights reserved.

This publication, portions of it, or any accompanying software may not be

reproduced in any way, stored in a retrieval system of any type, or transmitted
by any means, media, electronic display or mechanical display, including, but not
limited to, photocopy, recording, Internet postings, or scanning, without prior
permission in writing from the publisher.

Publisher: David Pallai

Mercury Learning and Information
22841 Quicksilver Drive
Dulles, VA 20166
[email protected]
www.merclearning.com
1-(800)-232-0223
D. Raine et al. Mathematical Physics.
ISBN: 9781683922056
The publisher recognizes and respects all marks used by companies, manufacturers,
and developers as a means to distinguish their products. All brand names and
product names mentioned in this book are trademarks or service marks of their
respective companies. Any omission or misuse (of any kind) of service marks or
trademarks, etc. is not an attempt to infringe on the property of others.
Library of Congress Control Number: 2018949983
181920321 This book is printed on acid-free paper in the United States of America.
Our titles are available for adoption, license, or bulk purchase by institutions,
corporations, etc. For additional information, please contact the Customer
Service Dept. at (800) 232-0223 (toll free).

All of our titles are available in digital format at authorcloudware.com and other
digital vendors. The sole obligation of Mercury Learning and Information to
the purchaser is to replace the book, based on defective materials or faulty
workmanship, but not based on the operation or functionality of the product.
CONTENTS

Preface������������������������������������������������������������������������������������ xiii

1 Derivatives and Integrals............................................. 1

1.1 Definition of a derivative................................................ 1
1.2 Some basic derivatives.................................................... 4
1.3 “Function of a function” (chain rule)............................. 5
1.4 Product (and quotient) rule............................................ 8
1.5 Implicit differentiation.................................................. 11
1.6 Piecewise differentiable functions................................ 12
1.7 Higher order derivatives............................................... 14
1.8 Stationary points............................................................ 15
1.9 Integrals of elementary functions................................. 18
1.10 Integrals of combinations of functions......................... 20
1.11 Integration by substitution............................................ 21
1.12 Integration by parts....................................................... 25
1.13 Integration of rational functions................................... 27
1.14 Definite integrals........................................................... 28
1.15 Area under a graph........................................................ 29
1.16 Continuous and discontinuous functions..................... 30
1.17 Estimates of integrals.................................................... 32
1.18 Derivatives of integrals.................................................. 33
1.19 Reduction formulae....................................................... 35
1.20 Exercises........................................................................ 40
1.21 Problems........................................................................ 44
vi  •  Mathematical Physics

2 Elementary Functions................................................ 49
2.1 Binomial expansion....................................................... 50
2.2 Maclaurin series............................................................ 55
2.3 Taylor series................................................................... 58
2.4 Equilibrium points........................................................ 59
2.5 Definition of ex (the exponential function)................... 63
2.6 The inverse function of ex............................................. 63
2.7 Derivative of the exponential........................................ 64
2.8 Integration of exponentials........................................... 67
2.9 Hyperbolic functions..................................................... 68
2.10 Inverse functions........................................................... 69
2.11 Inverse hyperbolic functions......................................... 71
2.12 Exercises........................................................................ 72
2.13 Problems........................................................................ 75

3 Functions, Limits, and Series..................................... 77

3.1 Curve sketching: Quadratics......................................... 77
3.2 Curve sketching: General polynomials......................... 79
3.3 Curve sketching: Rational functions............................. 81
3.4 Graphical solution of inequalities................................. 83
3.5 The symmetry of functions........................................... 85
3.6 Limits............................................................................. 87
3.7 Indeterminate forms..................................................... 89
3.8 L’Hôpital’s rule.............................................................. 91
3.9 Limits of integrals.......................................................... 92
3.10 Approximation of functions........................................... 94
3.11 Sequences and series..................................................... 96
3.12 Infinite series................................................................. 97
3.13 Exercises...................................................................... 104
3.14 Problems...................................................................... 105

4 Vectors ��������������������������������������������������������������������� 111

4.1 Basic properties........................................................... 112
4.2 Linear dependence; basis vectors............................... 114
4.3 The scalar (dot) product.............................................. 118
4.4 The vector (cross) product.......................................... 121
4.5 Multiple products........................................................ 123
 Contents  •  vii

4.6 Equation of a line........................................................ 125

4.7 Equation of a plane..................................................... 127
4.8 Components of vectors............................................... 129
4.9 Scalar (dot) product (component form)..................... 131
4.10 Vector (cross) product (component form).................. 132
4.11 Scalar triple product.................................................... 133
4.12 Vector triple product................................................... 134
4.13 Vector identities........................................................... 135
4.14 Examples from physical science................................. 138
4.15 Extension: Index notation........................................... 140
4.16 Exercises...................................................................... 142
4.17 Problems...................................................................... 142

5 Matrices.................................................................... 147
5.1 Matrix representation.................................................. 149
5.2 Solution of systems of equations................................. 153
5.3 Products....................................................................... 156
5.4 The identity matrix...................................................... 159
5.5 Symmetric and antisymmetric matrices..................... 160
5.6 Inverses........................................................................ 161
5.7 The inverse of a 2 × 2 matrix...................................... 165
5.8 Determinants............................................................... 167
5.9 Properties of determinants......................................... 169
5.10 Solution of 2 × 2 linear systems.................................. 174
5.11 Solution of 3 × 3 systems............................................ 177
5.12 Homogeneous systems................................................ 179
5.13 A formula for the inverse matrix................................. 179
5.14 Eigenvalues and Eigenvectors.................................... 181
5.15 Matrices as transformations........................................ 186
5.16 Extension: Transformation groups.............................. 191
5.17 Exercises...................................................................... 192
5.18 Problems...................................................................... 195

6 Differential Equations 1.......................................... 199

6.1 What are differential equations?................................ 199
6.2 Solving differential equations..................................... 201
6.3 First order separable equations.................................. 203
viii  •  Mathematical Physics

6.4 Linearity versus non-linearity..................................... 205

6.5 Homogeneous versus inhomogeneous....................... 207
6.6 Finding solutions......................................................... 209
6.7 Homogeneous linear equations.................................. 211
6.8 Auxiliary equations with repeated roots..................... 215
6.9 Particular integrals...................................................... 217
6.10 Inhomogeneous linear equations................................ 222
6.11 Integrating factor method........................................... 226
6.12 Extension: Special functions....................................... 230

7 Complex Numbers.................................................... 235

7.1 Introduction of complex numbers.............................. 235
7.2 Operations on complex numbers................................ 239
7.3 Quadratic equations.................................................... 244
7.4 The Argand diagram.................................................... 247
7.5 Modulus and argument of a complex number........... 249
7.6 The complex exponential............................................ 252
7.7 De Moivre’s theorem.................................................. 256
7.8 The roots of unity........................................................ 258
7.9 Roots of real polynomials............................................ 258
7.10 Roots of complex polynomials.................................... 259
7.11 Extension: Complex variable...................................... 260
7.12 Exercises...................................................................... 263
7.13 Problems...................................................................... 264

8 Differential Equations 2.......................................... 267

8.1 Boundary and initial conditions.................................. 267
8.2 Auxiliary equations with complex roots...................... 277
8.3 Equations with complex coefficients.......................... 281
8.4 Complex inhomogeneous term................................... 282
8.5 Boundary (or initial) conditions.................................. 289
8.6 Systems of first order equations.................................. 291
8.7 Complex impedance.................................................... 292
8.8 Some examples from physics...................................... 294
8.9 Extension: Green functions........................................ 296
8.10 Exercises...................................................................... 298
8.11 Problems...................................................................... 300
 Contents  •  ix

9 Multiple Integrals..................................................... 303

9.1 Repeated integrals....................................................... 303
9.2 Integrals over rectangles in the plane......................... 307
9.3 Multiple integrals over irregular regions.................... 310
9.4 Change of order of integration................................... 313
9.5 Polar coordinates in two dimensions.......................... 316
9.6 Integrals over regions in the plane............................. 317
9.7 An important definite integral.................................... 319
9.8 Cylindrical polar coordinates...................................... 320
9.9 Volume integrals (cylindrical polar)............................ 322
9.10 Integrals over cylindrical surfaces.............................. 323
9.11 Spherical polar coordinates......................................... 325
9.12 Volume integrals (spherical polar).............................. 326
9.13 Integrals over spherical surfaces................................. 327
9.14 Solid angle................................................................... 329
9.15 Sketching surfaces....................................................... 333
9.16 Exercises...................................................................... 336
9.17 Problems...................................................................... 338
10 Partial Derivatives.................................................. 341
10.1 Functions of two variables.......................................... 341
10.2 Partial derivatives........................................................ 344
10.3 Directional derivatives................................................ 346
10.4 Functions of many variables....................................... 348
10.5 Higher derivatives....................................................... 349
10.6 Function of a function................................................. 350
10.7 Implicit differentiation................................................ 352
10.8 Derivative with respect to a parameter...................... 353
10.9 Taylor expansion about the origin............................... 355
10.10 Expansion about an arbitrary point............................ 357
10.11 Function of more than two variables.......................... 359
10.12 Stationary points.......................................................... 359
10.13 Definition of the total differential.............................. 365
10.14 Exact differentials........................................................ 366
10.15 The chain rule.............................................................. 367
10.16 Consistency with the chain rule.................................. 369
10.17 Exercises...................................................................... 372
10.18 Problems...................................................................... 375
x  •  Mathematical Physics

11 Partial Differential Equations............................... 379

11.1 Examples from physics................................................ 380
11.2 General solution of the wave equation....................... 383
11.3 Derivation of the general solution.............................. 386
11.4 A string initially at rest................................................ 388
11.5 A string given an initial velocity.................................. 391
11.6 A formula for general initial conditions...................... 393
11.7 Semi-infinite string...................................................... 395
11.8 Simple harmonic waves............................................... 397
11.9 Beats............................................................................ 397
11.10 Group velocity............................................................. 400
11.11 Separation of variables................................................ 402
11.12 Eigenvalues.................................................................. 406
11.13 Reflections at a discontinuity...................................... 408
11.14 Matter waves incident on a potential step.................. 412
11.15 Exercises...................................................................... 415
11.16 Problems...................................................................... 416

12 Fourier Series......................................................... 419

12.1 Introduction................................................................. 419
12.2 Fourier half range sine series...................................... 420
12.3 Fourier sine series coefficients................................... 421
12.4 Using the Fourier series results.................................. 422
12.5 Application to the wave equation............................... 425
12.6 Application to other differential equations................ 428
12.7 Fourier half range cosine series.................................. 430
12.8 Numerical series.......................................................... 432
12.9 Periodic extension of Fourier series........................... 433
12.10 Even and odd symmetry of periodic functions.......... 433
12.11 Fourier full range series.............................................. 435
12.12 Complex form of Fourier series.................................. 436
12.13 Properties of Fourier series........................................ 439
12.14 Fourier transforms and Fourier integrals................... 440
12.15 Fourier transforms...................................................... 441
12.16 Calculating Fourier transforms................................... 442
12.17 Convolutions................................................................ 443
12.18 Convolution theorem.................................................. 444
 Contents  •  xi

12.19 Fourier deconvolution................................................. 445

12.20 Extension: Laplace transform..................................... 446
12.A Appendix: Standard integrals...................................... 449
12.21 Exercises...................................................................... 451
12.22 Problems...................................................................... 452

13 Introduction to Vector Calculus............................ 455

13.1 Vector fields................................................................. 455
13.2 The gradient of a function.......................................... 458
13.3 Maximum rate of change............................................ 460
13.4 Level surfaces.............................................................. 462
13.5 Normals to a surface.................................................... 464
13.6 The divergence of a vector field................................. 467
13.7 Curl of a vector field................................................... 470
13.8 Flux of a vector field through a surface...................... 473
13.9 Circulation of a vector field........................................ 478
13.10 Divergence theorem (Gauss’s theorem)..................... 482
13.11 Stokes’s theorem.......................................................... 486
13.12 Extensions: Index notation.......................................... 490
13.13 Extensions: Other coordinates.................................... 491
13.14 Exercises...................................................................... 493
13.15 Problems...................................................................... 494

Appendix A Prerequisites............................................ 497

A.1. Numbers, Functions, and Proofs................................ 497
A.2. Necessary and Sufficient Conditions.......................... 499
A.3. Logarithms................................................................... 500
A.4. Trigonometric Functions............................................ 502
A.5. Division of Polynomials............................................... 507
A.6. Partial Fractions.......................................................... 508
A.7. Series............................................................................ 510

Appendix B The Greek Alphabet................................ 513

Index��������������������������������������������������������������������������� 515
PREFACE
The book differs from other introduction to mathematical methods
at this level in several important areas.
First, it does not follow the usual presentation of a description of
the theory followed by examples and exercises. Rather we use exam-
ples to introduce the theory. This approach is not new; it goes back
to the methods by which the scribes of Ancient Babylon learned
mathematics: by example problems and, to judge from the numbers
of surviving cuneiform tablets, lots of them!
Second, to help the reader digest the text, it is broken up into
quite short sections (often a page or so) followed by exercises. It may
be tempting to skip the exercises, (especially if one is used to doing
only a selection of “end-of-chapter” problems) on the grounds that
one can get through the book more quickly that way. This is true, in
the same sense that watching a film speeded up x8 will get to the end
more quickly, but it will be without much understanding of the plot.
There are optional additional exercises at the end of each chapter;
the ones in the text represent the minimum we think you need.
Third, however, we have tried to avoid too many “plug-and-
chug” exercises, that is, exercises which you solve by following the
text but substituting some different numbers (“pattern matching”).
These are useful to reinforce memory, but they are not very useful
to develop or test understanding. Rather, we have tried to make the
exercises diagnostic in the sense that they do test understanding,
that is, they test the ability to use what has been learned in a slightly
different context. An instructor can therefore use these to target
support for students.
xiv  •  Mathematical Physics

Finally, and related to the previous point, while we hope it is per-

fectly possible to use this book for self-study, it was not designed for
that purpose. It is intended for use as a course text. In this regard it
might be useful to say a little about the background to the writing of
the book - especially if you are intrigued to know why there are so
many authors. We would also like to thank contributions to various
versions of the text from Paul Abel, Mike Dampier, Andrew King, and
Tim Yeoman.
About forty years ago, it was agreed that our conventional
presentation of mathematical methods for our physics students -
lectures, marked homework,and examinations - was not as effective
as we might have hoped. So, instead of spending lecture time going
through theory and exercises on the board, we produced a text as, in
effect, the lecture notes, and refocused class time on weekly work-
shops and small group tutorials. Lectures were restricted to a weekly
one-hour introduction to the topics for that week. We made the exam-
inations harder, by requiring passes separately in the major topics
(calculus of one variable, many variables, linear algebra, differen-
tial equations, vector calculus) and the pass rates soared. The initial
text has been refined over the years (hence the number of authors)
and this book is another, more outward-facing version, which we are
pleased to have the opportunity to share with you.

D. J. Raine (lead author)

and the teaching team:
G. A. Wynn
S. Vaughan
M. Roy
R. O. Davies
E. J. Bunce

Leicester, 2017
 CHAPTER

1
DERIVATIVES AND
INTEGRALS

In this chapter we quickly cover the basic and most commonly

used rules for differentiation and integration of simple functions
and combinations of functions. We assume that you are familiar
with basic algebra, the concept of a function and the idea of a
limit. Much of the chapter will be familiar to you, but perhaps not
all of it. And it is important to practice these procedures as much
as possible – by working through the exercises, even if they look
simple – to prepare you for the more advanced material that will
build upon these ideas.

1.1.  DEFINITION OF A DERIVATIVE

The derivative of a function is the slope (or gradient) of the

tangent to the graph of the function at any point; i.e. it is the slope
approached by the line between two points on the graph as they
get closer (Figure 1.1). By the “slope” we mean (as usual) the
change in height divided by the horizontal displacement. To use
this definition to calculate the slopes of graphs we put it into more
formal language.
2  •  Mathematical Physics

y y

f (x+h)

f (x+h)
f (x) f (x)

x x
x x+h x x+h

(a) (b)

FIGURE 1.1:  The tangent at an arbitrary point x as the limit of line joining points (x, f(x)) and
(x + h, f(x + h)) as h gets smaller.

More formally, the derivative of a function y = f(x) is the func-

tion f ′(x) or dy/dx defined by

f ( x + h) − f ( x)
f ′ ( x ) = lim . (1.1)

h→ 0 h 
Equivalently, in a commonly used alternative notation, we write
dx in place of h to emphasize that we are making a small increment
in x. We also sometimes write y = y(x) instead of y = f(x) to indicate
that y is a function of x. The definition is then written

dy y( x + d x) − y( x)
= lim . (1.2)
dx d x → 0 dx

For a function of the form y = f(x) we use the notation f ′(x),
y′ and dy/dx interchangeably. The definition expresses in an exact
manner the fact that the derivative at x is the slope of the tangent at
x. In order to use it we need a way of taking the limit. The follow-
ing example shows how this is done for a simple function such as a
power of x.

Example 1.1 Use the definition from equation (1.1) to find

f ′(x) if f(x) = x3.
To use the definition we first need f(x + h).
3
f (x) = x 3 , and f ( x + h) = ( x + h) .
 Derivatives and Integrals  •  3

Next, expanding out the parenthetical, we have,

f(x + h) = x3 + 3x2h + 3xh2 + h3,

therefore
f ( x + h) − f ( x)
= 3 x2 + 3 xh + h2 ,
h
and

x ) lim ( 3 x2 + 3 xh + h2 ) ,
f ′ (=
h →0

= 3 x2 + 0 + 0,
= 3 x2 .

Exercise 1.1 Use the definition of the derivative to prove that if

f(x) = x4 then f ′(x) = 4x3.

Example 1.2 Use the definition of a derivative to find the

derivative of the function y = xn (with n a positive integer).
We begin with the function, and write out the necessary limit

dy ( x + h ) −x n n

=y x=
, and lim n
.
dx h→ 0 h

Then, using the binomial expansion (which you may be familiar

with and will find discussed in Chapter 2),

1
( x + h )n = x n + nx n−1 h + n ( n − 1 ) x n−2 h2 +  + h n ,
2
4  •  Mathematical Physics

and

1
( x + h )n − =
x n nx n−1 h + n ( n − 1 ) x n− 2 h2 +  + h n ,
2
and so

( x + h )n − x n 1 
= nx n−1 + h  n ( n − 1 ) x n− 2 +  + h n− 2  .
h 2 

Taking the limit as h → 0 gives

( x + h )n − x n
lim = nx n−1 ,
h →0 h

and therefore

dy
= nx n−1 .
dx

In principle, the derivative of any function can be worked out

from the definition. In practice, we memorize a few simple examples
(Section 1.2, Table 1.1) and we learn to build up more complicated
derivatives from these by a variety of methods (Sections 1.3 – 1.5).

1.2.  SOME BASIC DERIVATIVES

You should already know (or be willing to take on trust for the
present) the derivatives in Table 1.1.
Note: the trigonometric functions are introduced in section A.4
of Appendix A and the logarithm in Section A.3. Throughout this
course ln(x) will be written for loge(x), the “natural logarithm.” It
is common to use log(x) = log10(x), and the use of any other base is
indicated explicitly, as in log2(x). The derivatives in the table may be
taken as given.
 Derivatives and Integrals  •  5

TABLE 1.1:  Table of some standard derivatives.

Function y(x) Derivative dy/dx

a, a constant 0

x a axa−1

sin(x) cos(x)

cos(x) − sin(x)

tan(x) sec2(x)

ln(x) = loge(x) 1/x

eax aeax

The derivative of the sum of two functions is the sum of the deriv-
atives. This means that if f(x) and g(x) are both functions of x then

d df dg
( f + g) = +  (1.3)
dx dx dx

Exercise 1.2 What is dy/dx if

(i) y = xn + c where c is a constant and n an integer
(ii) y = x−3 + x4
(iii) y = 2sin(x) = sin(x) + sin(x)
(iv) y = sin(x) + cos(x).

We also learn rules for the derivative of a “function of a func-

tion” and of a product, which enable us to construct the derivatives
of quite complicated expressions from the table without having to
return to the basic definition.

1.3.  “FUNCTION OF A FUNCTION” (CHAIN RULE)

Let’s consider the function y = (1 + 2x)2. This can be thought

of as the composite of two functions: u = 1 + 2x and y = u2 because
6  •  Mathematical Physics

y(u(x)) = (1 + 2x)2. What is the relationship between the derivatives

of these three functions?
dy
y =(1 + 2 x ) =4 x2 + 4 x + 1, so
2
=8 x + 4,
dx
and, alternatively
dy
=
y u2 , so u 2 (1 + 2 x ) ,
= 2=
du
du
u=
1 + 2 x, so =
2.
dx

We can see that the following holds:

8x + 4 = 2(1 + 2x) × 2.

And so the relation in this case is

dy dy du (1.4)
= .
dx du dx

This is a general rule that can be proved from the definition of

a derivative: the derivative of a composite of two functions (a “func-
tion of a function”) is the product of their derivatives. In words: the
rate of change of y(u(x)) with respect to x is the rate of change of
y with respect to u times the rate of change of u with respect to x.
This is usually called the chain rule. So if y = f(u(x)) then we use
equation (1.4) to compute its derivative as the product of dy/du and
du/dx.

Example 1.3 If y = cos(x2) what is dy/dx?

We use the chain rule, choosing u to simplify the process. Let
u = x2 so y = cos(u) and

dy/du = − sin(u),
du/dx = 2x.
 Derivatives and Integrals  •  7

Then, using the chain rule,

dy dy du
= ,
dx du dx
= − sin ( u ) ⋅ 2 x,
= −2 x sin ( x2 ) .
Alternatively, and more economically, you can write down the
chain rule directly:

d cos ( x2 ) dx2
d
dx
{ }
cos ( x2 ) = 2
dx
⋅
dx
= ( )
− sin ( x2 ) ⋅ 2 x.

Exercise 1.3 What is dy/dx if

(i) y = sin(2x)
(ii) y = sin(x2)
(iii) y = cos3(x)
(iv) y = sin(ax)
(v) y = (1 + ax)n
(vi) y = (1 + cos(x))1/2?

Exercise 1.4 What is dy/dx if y = log10(x)? (Hint: see section A.3

to convert this to the derivative of ln(x) which you can get from
the table.)

Exercise 1.5 If y = ln[x + (1 + x2)1/2] show that dy/dx = (1 + x2)-1/2.

(Hint: look at what you are seeking to prove in order to see how to
simplify your initial expression.)

Exercise 1.6 If y = 1/f(x) show that dy/dx = − f ′/f 2 (where, as

usual, f ′ stands for df/dx).
Note that it is a matter of choice whether you commit the result
of this last exercise to memory and use it in specific cases, or use the
chain rule directly as needed.
8  •  Mathematical Physics

1.4.  PRODUCT (AND QUOTIENT) RULE

Suppose now that we want the derivative of a product of two

functions in 1.1, y(x) = u(x)v(x), say. We might guess that this is
du du
. You can readily check that this is wrong by applying it to
dx dx
a simple example, say y = x2 = x × x. The correct derivative of the
product of two functions, like y(x) = u(x)v(x), is the sum of two terms
as follows:
dy du du
=u + u , equivalently ( uu)¢ = uu ¢ + u u¢ . (1.5)
dx dx dx
This is known as the product rule or the Leibniz rule. An alter-
native way to remember the product rule is as follows:
d(uv) = udv + vdu.

Example 1.4 If y = x sin(x), what is dy/dx?

Let u = x, v = sin(x). Then we have that

du du
= 1,= cos ( x ) .
dx dx
Therefore, by equation (1.5),
dy
= x ⋅ cos ( x ) + sin ( x ) ⋅ 1 = x cos ( x ) + sin ( x ) .
dx

Example 1.5 If y = x2 cos3(2x), what is dy/dx ?

This combines the chain rule and product rule. First break it
down as a product of two terms y = uv:
u = x2, and v = cos3(2x).
We can see that 
du
= 2 x.
dx
 Derivatives and Integrals  •  9

As v is a function of a function we use the chain rule to find its

to find its derivative:

= u =
w 3
with w cos( 2 x ) ,
dw
= − 2sin( 2 x ) ,
dx
du du dw
= = ( 3 w2 ) ( − 2sin( 2 x ) ) ,
dx dw dx
= − 6 cos2 ( 2 x ) sin( 2 x ) .

Putting it all together

dy du du
= u +u ,
dx dx dx
−6 x2 cos2 ( 2 x ) sin ( 2 x ) + 2 x cos3 ( 2 x ) .
=
Eventually you should be able to do some of the intermediate
steps in your head. So your working might look something like
this:
(x2 cos3(2x))′ = x2(cos3(2x)3)′ + 2x cos3(2x),
= x2 · 3 cos2(2x)(cos(2x))′ + 2x cos3(2x),
= x2 · 3 cos2(2x) sin(2x)(−2) +2x cos3(2x),
= −6x2 cos2(2x) sin(2x) + 2x cos3(2x).
You should cultivate the habit of working like this, resorting to
introducing u and v explicitly only if you get stuck.

If you are familiar with proof by induction, you may like the fol-
lowing example.

Example 1.6 Assuming that dx = 1 prove by induction that

dx
n
dx
= nx n−1 when n is a positive integer.
dx
10  •  Mathematical Physics

We start by assuming that for integers up to some n we have

dx n
= nx n−1 . Then
dx
dx n+1 d n
= xx = 1 ⋅ x n + x ⋅ nx n−1 = ( n + 1 ) x n .
dx dx
Thus, if the result is true up to n it is true for n + 1. Since it is
true for n = 1 it is true for all n.

Exercise 1.7 What is dy/dx if

(i) y = x ln(x)
(ii) y = x2 sin(x)
sin( 2 x )
=
(iii) y ? (Treat this as x−3 sin(2x).)
x3

The derivative of the quotient of two functions, like

y(x) = u(x) / v(x), has its own rule:

du du
u− u 
dy dx
= dx , equivalently ( u / u )′ u′u − u ′u . (1.6)
dx u2 u2

Exercise 1.8
(i) Derive the quotient rule: By putting f = u, g = 1/v show that
if y = f g = u/v then dy/dx = (u′v′ − v′u)/ v2.
(ii) Use the quotient rule to find dy/dx if y = sin(2x)/x3 and check
your result with exercise 1.7(iii).
(iii) Use the quotient rule to find dy/dx if y = tan(x). (Recall that
tan(x) = sin(x)/ cos(x).)

Rather than learning the quotient rule it is often easier to cal-

culate (u/v)′ using the product rule for (uv−1)′. Alternatively, you can
memorize the rule as
vdu − udv
d ( u v) = .
v2
 Derivatives and Integrals  •  11

1.5.  IMPLICIT DIFFERENTIATION

Sometimes it is convenient to find the derivative dy/dx from an

equation like f (y) = g(x) without first solving for y as a function of x.
This is little more than an application of the chain rule.

Example 1.7 Find dy/dx if y = sin−1(x). (Note that sin−1(x) is the

inverse sine function, also called arcsin(x); it is not the same as
1/ sin(x).)
We begin by transforming to something more familiar. If
y = sin−1(x),
x = sin(y)
and differentiating both sides with respect to x and using the
chain rule we have
d dy d
=1 =
dx
( sin( y) )
dx dy
( sin( y) ) ,
dy
= cos( y) .
dx
Rearranging for dy/dx gives

dy 1
= .
dx cos( y)
We want the answer as a function of x so we have to get cos(y)
in terms of x given that sin(y) = x. We first substitute for cos(y)
using sin2(y) + cos2(y) = 1. Then,

dy 1 1
= = .
(1 − sin2 ( y) ) (1 − x )
12 2 12
dx

The method used in this example is referred to as implicit dif-

ferentiation because we differentiate the equation sin(y) = x which
gives y implicitly as a function of x, in contrast to y = sin−1(x) where y
is given explicitly.
12  •  Mathematical Physics

Exercise 1.9 Find dy/dx if

(i) y = cos−1(x)
(ii) y = tan−1(x)
(iii) y = xx. (Hint: begin by taking ln of both sides. The answer
is not x ⋅ xx−1 because the power to which x is raised is not a
constant!)

1.6.  PIECEWISE DIFFERENTIABLE FUNCTIONS

Many useful functions cannot be described fully using a single

formula, but can be described by different formulae that apply under
different circumstances. A piecewise function f (x) defines a function
in terms of formulae that apply for different ranges of x.
Two useful examples of piecewise functions are | x | (pronounced
“mod x”) and sgn(x) (pronounced “sign x” or “signum x”). We treat
them in turn.
First, f(x) = |x| = mod(x) is defined by

x if x≥0
f ( y) =  (1.7)
− x if x < 0.
This function is plotted in Figure 1.2(a).

Example 1.8 Evaluate

(i) | – 3|
(ii) | 7|
The function is defined separately for x ≥ 0 and x < 0. Since
– 3 is less than 0, look at the second part of the definition: for
x = – 3, y = –x gives y = – (– 3) = +3.
Next, we have 7 so we want the definition for x ≥ 0: for x = 7,
y = +x, so y = +7.
Obviously |x| is simply the absolute value of x (i.e. its magnitude
regardless of sign), but looking at the definition of a quantity
will often provide the first step in the answer to a question.
 Derivatives and Integrals  •  13

Second, f(x) = sgn(x). This is defined by

+1 if x > 0 

=f ( x ) =0 if x 0 (1.8)
−1 if x < 0.

This function is plotted in Figure 1.2(b).

y
y
y = sgn( x)
y = |x|

(a) (b)

FIGURE 1.2:   (a) The graph of the “modulus” function y = |x|. (b) The “sign” function y = sgn(x).

Exercise 1.10 What is

(i) sgn(1)
(ii) sgn(– 5)?

Example 1.9 If y = sgn(x) what is dy/dx?

Look at the definition. If x > 0, then y = +1, so dy/dx = 0. If
x < 0, then y = 1, so dy/dx = 0 (because y = constant there).
The derivative does not exist at x = 0. (The graph, shown
in Figure 1.2, obviously does not have a tangent there).
Looking at the graph confirms that the slope is indeed zero
everywhere else.

Notice that any real number can be expressed as the product of

its sign and its modulus
x = sgn(x) · |x|.
These two functions f(x) are piecewise linear functions because
each piece of the graph is a straight line. For mod(x), the graph is
14  •  Mathematical Physics

continuous at x = 0, but its derivative is discontinuous at this point:

there is an instantaneous change in the gradient. For sgn(x) the
graph has a discontinuity at x = 0. We shall discuss continuous and
discontinuous functions more in section 1.16.

Exercise 1.11 If y = |x| what is dy/dx? (Hint: start from the

definition of the mod function!)

1.7.  HIGHER ORDER DERIVATIVES

Until now we have concentrated on the first derivative of a

function. We can also define the second derivative, i.e. the gradient
of the gradient. We use the definition of equation (1.1) but replace
f(x) with f ′(x):
f ′( x + h) − f ′( x)
f ′′ ( x ) = lim , (1.9)
h →0 h
or, equivalently,

d2 y y′ ( x + d x ) − y′ ( x ) (1.10)
= lim .
dx 2 d x →0 dx
Similarly, we can define the third, fourth and higher order
derivatives. Often the nth derivative of a function y = f(x) is writ-
ten as f(n)(x), or dny/dxn. Note the parentheses around the super-
script (n) in f(n)(x) to distinguish it from (f(x))n. In practice we
continue to differentiate the result after each differentiating
(tidying up the expressions as we go) until we get to the required
derivative.

Example 1.10 If y = xn calculate d2y/dx2.

If y = xn then dy/dx = nxn–1. Next,

d 2 y d  dy  d
=
dx 2
=  
dx  dx  dx
( nx n−1 )

= n ( n − 1 ) x n−2 .
 Derivatives and Integrals  •  15

Exercise 1.12 Find all the derivatives (first, second etc.) of

the function f(x) = x3 + 6x2 − 9x + 12 until the derivative is
zero.

Exercise 1.13 Find derivatives of the first five orders of the func-
tion f(x) = sin(x).

1.8.  STATIONARY POINTS

A common application of differentiation is to find the maxima or

minima of functions.
Let y = f(x). A point at which dy/dx = 0 is called a stationary
point of the function f(x). Equivalently, at a stationary point of f(x),
we have f ′(x) = 0. See Figure 1.3.
• A stationary point of y = f(x) is a minimum if, at this point,
d2y/dx2 > 0;
• A stationary point of y = f(x) is a maximum if, at this point,
d2y/dx2 < 0;
• A stationary point of y = f(x) is (usually) a stationary point of
inflection if, at this point, d2y/dx2 = 0 (i.e. the derivative has a
maximum or minimum).

y y y f’ (x) > 0
f’ (x) = 0
f ’’(x) > 0
f ’’(x) = 0
f’ (x) < 0 f’ (x) > 0
x
f’ (x) = 0
f’ (x) = 0

f’ (x) < 0 f’ (x) > 0

x x
f’ (x) > 0
f ’’(x) < 0
(a) (b) (c)

FIGURE 1.3:  (a) At a minimum the slope increases as x increases, so d(y′)/dx = y″ > 0. (b) At a
maximum, the slope decreases as x increases, so y ″ < 0. (c) At a stationary point of inflection
the slope is zero and the rate of change of slope is zero.
16  •  Mathematical Physics

Example 1.11 Find the stationary points of the function

y = 2x3 + 3x2 − 36x + 24 and determine whether they are
maxima, minima, or points of inflection.
We begin by finding dy/dx:

dy/dx = 6x2 + 6x – 36,

= 6(x2 + x – 6),
= 6(x + 3)(x – 2).

Then we set dy/dx = 0 and solve for x, which gives x = –3

and x = 2. So the stationary points are x = –3, x = 2. Now we
calculate the second derivative,

d2y/dx2 = 6(2x + 1),

and evaluate it at the stationary points.

At x = – 3, d2y/dx2 < 0, so this is a maximum,

at x = 2, d2y/dx2 > 0, so this is a minimum.

Exercise 1.14 Find the stationary points of the following func-

tions, and determine whether they are maxima, minima, or points
of inflection.
(i) y = x5 − 5x
(ii) y = x3 − 3x2 + 3x + 3

In general, if y = f(x) has no stationary points in an interval

a < x < b then it is either increasing or decreasing. The following
are the mathematical definitions of increasing and decreasing as
applied to functions:
• A function is increasing on an interval if f (b) ≥ f (a) for all b > a
and a, b in the interval.
• A function is decreasing on an interval if f (b) ≤ f (a) for all b > a
and a, b in the interval.
 Derivatives and Integrals  •  17

Notice that we make no assumption that the functions are

continuous.
In the case of a continuous function, for which the derivative is
well defined at each point in an interval, we can use the first deriva-
tive to establish whether the function is increasing or decreasing.
• A function is increasing on an interval if f ′ (x) ≥ 0 for all x in
the interval.
• A function is decreasing on an interval if f ′(x) ≤ 0 for all x in
the interval.

Notice here that, mathematically speaking, “increasing” and

“decreasing” include being constant (i.e. “non-decreasing” and
“non-increasing,” respectively). If we want to exclude this possibility
we say a function is strictly increasing (f ′(x) > 0) or strictly decreas-
ing (f ′(x) < 0).

Example 1.12 Show that the function f(x) = x – x 2 is (strictly)

increasing in 0 < x < 12 . Hence show that x > x 2 for 0 < x < 12 .
To find out if the function is increasing (“going uphill”) we look
at the slope of the tangent: f ′(x) = 1 – 2x. Now we determine
where the slope is positive. In this case f ′(x) > 0 if 1 – 2x > 0,
hence the slope is positive over the stated region ( 0 < x < 12 ) . So
f (x) is increasing for 0 < x < 12 ; the function is indeed increasing
in this interval.
This means that f (b) > f (a) if b > a in this interval. Take a = 0
and b = x. Then f(x) > f (0) implies x – x2 > 0 – 0, which in turn
implies x > x2 for 0 < x < 12 .

Exercise 1.15 Show that x – sin(x) is (strictly) increasing for x > 0.

Deduce that sin(x) < x for x > 0.

Exercise 1.16 Show that y = ln(1 + x)– x is (strictly) decreasing for

x > 0. Deduce that ln(1 + x) < x in this range.
18  •  Mathematical Physics

1.9.  INTEGRALS OF ELEMENTARY FUNCTIONS

Integration is the inverse of differentiation. Thus, F(x) is the

indefinite integral of f(x) if f(x) is the derivative of F(x), i.e.

= dF ( x ) (1.11)
F ( x ) ∫=
f ( x ) dx if f ( x ) .
dx

In some books F(x) is called the anti-derivative of f(x).

Example 1.13 Find ∫ x3 dx

1 4
∫ x= x + c (where c is a constant) because
3
We see that dx
4
=x3 {
d 1 4
dx 4
x +c . }
Notice the “+c”, where c is any constant. There’s always a “plus c”
in an indefinite integral because anti-derivatives are defined only up
to an arbitrary additive constant.

Example 1.14 Find the indefinite integral ∫ bxm dx, where b

and m are constants.
Begin by writing down a relevant derivative. For y = axn we
have dy/dx = naxn−1. This gives an integral of the right form,

∫ nax
n −1
=
dx ax n + c.
Compare with the integral we want. Put na = b, n – 1 = m.
Hence n = m + 1 and a = b/(m + 1), provided m ≠ –1. Then,

b
∫ bx dx
= x m+ 1 + c. (for m ≠ −1)
m

( m + 1)
 Derivatives and Integrals  •  19

In practice we memorize this particular result, usually as “add one

to the power and divide by the new power.” But it is worth remember-
ing the method as well: to find an inverse we can guess a result of the
right form and then tidy up the details by differentiation and substitu-
tion. (If we guess wrong this won’t work, so it’s a fail-safe method.)
The case m = – 1 is special: we have

dx
∫x ∫=
−1
=
dx ln x + c.
x
Note the absolute value of x on the right.

Example 1.15 Find the indefinite integral ∫ cos(2x)dx.

Start from a relevant known integral. We know ∫ cos(x)dx =
sin(x) + c because d {sin ( x)} = cos ( x) . Then we check that the
dx
integral of cos (2x) is proportional to sin(2x). We have,

d
dx
{sin ( 2 x )} = 2 cos ( 2 x )
so that,

cos ( 2 x ) =
d 1
{
dx 2
sin ( 2 x )}
and therefore,
1
( 2 x ) dx
∫ cos= sin ( 2 x ) + c.
2

This shows you again how to go about obtaining a new integral

from a similar known one. But there is no need to go through all the
working once you know the result. The following exercise can be
done by comparison with the previous examples. Always check by
differentiating your answer.
Exercise 1.17 What are the indefinite integrals of
(i) sin(3x)
(ii) a/x?
20  •  Mathematical Physics

TABLE 1.2:   Table of some standard integrals; c is an arbitrary constant.

Function y(x) Indefinite integral ∫ ydx

xa
xa+1
+c ( a ≠ −1)
a+ 1

1/x In(x) + c
sin(x) – cos(x) + c

cos(x) sin(x) + c

tan(x) – ln | cos(x)| + c

ln(x) x ln(x) – x + c

eax a–1e ax + c (a ≠ 0)

f  ′(x)/f(x) ln(f (x)) + c

In order to integrate we need to have memorized a repertoire of

derivatives and to use systematic guesswork. In practice, the problem
is often to reduce an integral to some standard form which can be
looked up in tables of integrals (for example Table 1.2). The examples
in this chapter give some ways of doing this. The methods of Sections
1.11 and 1.12 are fundamental in their own right and must be learned.

1.10. INTEGRALS OF COMBINATIONS OF
FUNCTIONS

Sometimes integrals which look complex are actually easy to do,

as long as we know the correct rule or can apply some intuition. For
example, consider any function f(x). Then,
f ′( x)
∫= dx ln ( f ( x ) ) + c.
f ( x)
We can show this must be true by differentiating the right hand
side using the chain rule. If we let f(x) = u, then
 Derivatives and Integrals  •  21

d d
dx
{ln ( f ( x ) ) + c} = dx
( ln ( u ) ) ,
1 du
= ,
u dx
1
= f ′ ( x ).
f ( x)

Integration of functions like this – which can be written as a frac-

tion where the numerator is the derivative of the denominator – is
often called logarithmic integration.

1
x ) ( f ( x ) ) dx
( f ( x)) + c .
n +1
∫ f ′ (=
n
Exercise 1.18 Show that
n+1
(Remember that to prove a given integral you differentiate the
anti-derivative.)

1.11.  INTEGRATION BY SUBSTITUTION

It is often possible and useful to make a substitution that turns a

complicated integral into a simpler integral or a “standard” integral
that is given in charts like Table 1.2. Knowing which substitution to
use comes with experience and trial and error (and also guesswork!).

Example 1.16 By using a suitable substitution calculate

dx
∫ ax + b .
First, we look for an explicit change of variable such that

dx 1 dt
∫ ax + b = a ∫ t
.

We can then evaluate the integral, ∫ dt / t, using a standard result

from Table 1.2.
As we want to manipulate the integral into a form where the
new denominator involves t instead of ax + b, we start by trying
22  •  Mathematical Physics

t = ax + b. Then dt = adx so dx = dt/a. Next we substitute for x

and dx to give
dx 1 dt
∫ ax + b = a ∫ t
.

Finally, from the standard integral in Table 1.2, ∫dt/t = ln(t), so

that
dx 1
∫= ln ( t ) + c,
ax + b a
1
= ln ( ax + b ) + c.
a

Example 1.17 Find an explicit change of variable such that

dx dt
∫a 2
+x 2
= K∫
1 + t2

where K is a constant.
We want to replace the a2 with a 1 in the denominator, so we
take out factor a2 to get

dx 1 dx
∫a 2
+x 2
= 2∫
a 1 + x 2 a2
.

We now want t2 for x 2/a2 so we try t = x/a, and dt = dx/a or

dx = a dt. Then

dx 1 a dt
∫a 2
+x 2
= 2∫
a 1 + t2
,

1 dt
= ∫ .
a 1 + t2
 Derivatives and Integrals  •  23

Exercise 1.19
(i) Find an explicit change of variable that converts
dx 1 dt
∫ b − ( x − a)
2 2
into ∫
b 1 − t2
(ii) Similarly, simplify
dx
∫
(1 + b ( x + a ) )
2 2

In harder questions where we are not given the answer, the best
approach is often to try to make the integrand as simple as possible.
There is usually more than one way to start and, if you do not suc-
ceed the first time, try again!

dx
Example 1.18 Find I = ∫ .
(1 − x ) 2 12

It is tempting to try t2 = 1 – x2 to get rid of the square root.

Then, by differentiation 2tdt = –2xdx and therefore dx = – tdt/
(1 – t2)1/2. So
1 t dt dt
I=−∫ ⋅ = −∫ .
t (1 − t )
2 12
(1 − t 2 )
12

This is back where we started – so the substitution t2 = 1 – x2

does not work and we need some other way of removing the
square root.
If we had 1 – cos2(q) = sin2(q) in the denominator we could
extract the square root. So we try x = cos(q), then dx = – sin(q)
dq and we have
− sin (q ) dq
I = −∫ ,
(1 − cos2 (q ) )
12

− sin (q ) dq
=∫ ,
sin (q )
= − ∫ dq ,
=− q + c.
 24  •  Mathematical Physics

Finally, we must substitute back for x to give

I = – cos−1(x) + c

Note that we could have chosen x = sin(q) to get I = sin−1(x) + k

(where k is a different constant). These two results are the same
because sin−1(x) = p/2 – cos−1(x).

Exercise 1.20 Find

dx
∫ .
(9 − x )
2 12

Example 1.19 Find I = ∫ sin(x)cos(x) dx.

This will simplify by letting either t = sin(x), then dt = cos(x) dx,
or t = cos(x), then dt = – sin(x). We can choose either of these.
Using t = sin(x) we have
I = ∫ tdt,
1 2
=
t + c,
2
1 2
= sin ( x ) + c.
2
With practice we can set out our working more compactly as
follows:
1
I ∫ sin ( x )= d ( sin ( x ) )
cos ( x ) dx ∫ sin ( x )= ( sin ( x ) ) + c.
2
=
2
Note that in the final step the integral is of the form
1 2
∫ f=df
2
f + c , where f here happens to be sin(x).

Exercise 1.21 Find

(i) ∫ sin2(x) cos(x) dx
(ii) ∫ cosn(x) sin(x) dx (n > 0)
 Derivatives and Integrals  •  25

(iii) ∫ cot(x) dx
(iv) ∫ (1 + x2)5x dx

1.12.  INTEGRATION BY PARTS

This technique is one of the most important in mathematical

physics. You will probably have met the basic formula in one of two
equivalent guises, either
∫ u dv = uv – ∫ v du, (1.12)
or
∫ uv′ dx = uv – ∫ vu′ dx. (1.13)
The formula is derived by integrating the rule for differentiating
a product (Section 1.4). Use whichever form you are familiar with or
learn whichever you prefer.

Example 1.20 Find I = ∫ x sin(x) dx.

Solution 1: First using equation (1.12): we define u = x, and
dv = sin(x)dx. Then,
du = dx,
v = – cos(x).
Next, using equation (1.12), we find
I = ∫ x d(– cos(x)),
= x(– cos(x)) – ∫ (– cos(x)) dx,
= –x cos(x) + sin(x) + c.
Note: the integration constant in v = ∫ v′dx is arbitrary. It
will always cancel in the final answer, so we set it to zero for
convenience.
26  •  Mathematical Physics

Example 1.21 Find I=∫x sin(x) dx.

Solution 2: We could also use equation (1.13). First we define
u = x, and v′ = sin(x), then,

u′ = 1, and v′ = – cos(x),
I = uv–∫ vu′dx,
= x(– cos(x)) – (– cos(x)).1 dx,
= –x cos(x) + sin(x) + c.

Note that we chose u = x because differentiating removes the

awkward factor x from the integrand. The choice of the correct
approach is a matter of experience, and trial and error.
Finally, with some experience, we can set out the working more
compactly,

I= ∫ x sin(x) dx= – ∫ x d(cos(x))

= – x cos(x) + ∫ cos(x) dx =–x cos(x)+ sin (x) +c.

Example 1.22 Find ∫ x(1 + x)4 dx using integration by

parts.
We have
1
∫ x (1 + x ) ∫ x d (1 + x ) ,
4 5
dx=
5
1 1
x (1 + x ) − ∫ (1 + x ) dx,
5 5
=
5 5
1 1
x (1 + x ) − (1 + x ) + c.
5 6
=
5 30

Exercise 1.22 Find

(i) ∫ x cos(x) dx
(ii) ∫ ln(x) dx. (Hint: this is ∫ u dv where u = ln(x) and v = x.)
 Derivatives and Integrals  •  27

1.13.  INTEGRATION OF RATIONAL FUNCTIONS

A rational function is one of the form

polynomial in x
f ( x) =
another polynomial in x
You should review Section A.5 on how to divide polynomials and
Section A.6 on partial fractions.

Example 1.23 Find

2 x3 + 9 x2 + 4 x −20
I= ∫ 2x + 5
dx

Divide the numerator by the denominator so as to leave a

remainder of smaller degree than the denominator.
 5 
= ∫  x + 2x − 3 −
2
I  dx.
2x + 5 
Then, integrating each term,
x3 5
I= + x2 − 3 x − ln 2 x + 5 + c.
3 2

2 x2 − 5 x + 10
Exercise 1.23 Find ∫ x−4
dx

Exercise 1.24 Find the following integrals. (Hint: first express

the integrand in partial fractions.)
dx
(i)
∫ ( x + 2 )( x − 1)
x dx
(ii)
∫ ( x + 4 )( x + 1)
x2 + 6 x + 4
(iii)
∫ ( x + 4 )( x + 1)dx
28  •  Mathematical Physics

1.14.  DEFINITE INTEGRALS

The definite integral of a function f(x) between limits x = a and

x = b is evaluated from the indefinite integral F(x) at the limits of
integration
b
∫ f ( x ) dx = F ( b) − F ( a ) = [ F ]
b
(1.14) .
a  a

The square brackets on the right are simply a convenient nota-

tion for the difference between the value of the function (in this case
F(x)) evaluated at each of the two stated values (in this case x = b
and x = a).

1
Example 1.24 Find ∫−1
x2 dx (the definite integral of x2
−1
between – 1 and 1) and also find ∫
1
x2 dx .
1
1  x3  1  1 2
∫−1 =  3  = 3 −− 3  = 3.
2
x dx
  −1  

And similarly
−1
−1  x3  1 1 2
∫ x dx =  =− −   =− .
2
1
 3 1 3 3 3

In general
b a
∫ f ( x ) dx = −∫ f ( x ) dx.
a b

Exercise 1.25 Find

2
(i) ∫ x dx
3
1
p
(ii) ∫ sin 2 ( x ) dx. (Hint: sin 2 ( x) = 12 ëé1 - cos (2 x)ûù .)
0
It is useful to memorize how to find this last integral.
 Derivatives and Integrals  •  29

1.15.  AREA UNDER A GRAPH

The area “under” the graph of a function (i.e. the area between
the graph and the x-axis, with areas below the axis counted as nega-
tive) is given by the definite integral of the function.
y y = f (x)

FIGURE 1.4: The definite integral of a

function between limits a and b gives the
area “under” the graph of the function.

x
a b

Example 1.25 Find the area under the graph of sin(x)

(i) between 0 and p
(ii) between 0 and 2p
First
p p
∫ sin ( x ) dx = − cos ( x )  0 =− ( −1 ) − ( −1 ) =2.
0

And second, using different limits of integration,

2
sin( x ) dx =−cos( x )  0 =−1 − ( −1) =0.
2
∫ 0

Note that the area below the x-axis is counted as negative.

Exercise 1.26 What is the area under the graph of

(i) y = x2 between – 1 and 1
(ii) y = sin2(x) between 0 and p?

Exercise 1.27 Using the result of Example 1.20, find the area
under the graph of y = x sin(x) between 0 and p.
30  •  Mathematical Physics

1.16. CONTINUOUS AND DISCONTINUOUS

FUNCTIONS

A continuous function is one which has a graph with no breaks.

For example, all elementary functions (polynomials, logarithms,
trigonometric functions etc.) are continuous (at points where they
are finite). Formally, f (x) is continuous at x = a if f (x) approaches
f (a) as x approaches a.

Exercise 1.28
(i) At what point is sgn(x) discontinuous?
(ii) Draw the graph of the function y = sgn(x) between x = –1
and x = 2.
(iii) Find the area under the graph (as defined in Section 1.15).
y = f (x) for x < x1 y = g (x) for x ≥ x1 y = f (x) for x < x1 y = h (x) for x ≥ x1
y y

x x

x1 x1

(a) (b)
FIGURE 1.5: (a)  An example of a function defined piecewise that is continuous. In the region
x < x1, the function is y = f(x), and in the region x ≥ x1 the function is y = g(x). The two functions
are equal at x1, and continuous across x1. But notice the derivative is not continuous; there is
a sudden change in the gradient at x1. (b) Now with a discontinuity at the point x1. The graph
is no longer continuous and the derivative is not defined at x1.

The function sgn(x) (Section 1.6) is an example of a discontinu-

ous function. To integrate a discontinuous function you have to split
up the range into a union of intervals in which the function is con-
tinuous. This works because of a property of integrals
c b c
f ( x ) dx ∫ f ( x ) dx + ∫ f ( x ) dx
∫=
a a b

where b is any number. We can break up the range of integration

into subranges that might be easier to work with.
 Derivatives and Integrals  •  31

Example 1.26 Consider the following piecewise function.

C − t if t ≤ 3
f (t) = 
exp ( − t 5 ) if t > 3
Find the value of C that makes the function continuous over
the range t = 0 to t = 10, and then compute the definite integral
of f(t) over this range. See Figure 1.6.
The change in the definition of the function occurs at t = 3, so
we examine this point. Inserting this value into each formula
we get exp(–3/5) and C – 3. If the function is continuous over
this point, these two must be equal. Therefore
exp(– 3/5) = C – 3 ⇒ C = exp(– 3/5) + 3.
In order to integrate the above function we split the range
of integration (t = 0 − 10) into two parts, and integrate each
definition of function within its own range:

exp ( − t 5 ) dt.
10 3 10
∫ f ( t ) dt = ∫ ( C − t ) dt + ∫
0 0 3

We know how to integrate each piece.

2  0 = 3exp ( −3 5 ) + 9 2 = 6.1464...
3 3
∫ ( C − t ) dt = Ct − t
2
0

exp ( − t 5 ) dt =  −5exp ( − t 5 )  =5 exp ( −3 5 ) − exp ( −10 5 ) 

10 10
∫
3 3

= 2.0674...

The final result is the sum of these,

f ( t ) dt = ( 6.1464...) + ( 2.0674...) = 8.214...

10
∫0

Exercise 1.29 The function f(x) is defined by

 Ax + 1 for x≤1
f ( x) = 
2 x for x > 1.

(i) Show that the function is continuous at x = 1 if A = 1

32  •  Mathematical Physics

y
y = C–t

t=3 t = 10
2

y = exp(–t/5)
1
0

0 2 4 6 8 10 t
FIGURE 1.6:  The piecewise function y = f(t) of example 1.26. The range of integration is split
into two subranges t = 0 − 3 and t = 3 − 10 and the area under the graph over t = 0 − 10 is
the sum of these two areas.

(ii) If A = 1
(a) sketch the function f(x) between x = 0 and x = 2
(b) calculate the area under the graph between x = 0 and
x=2

Note that if A = 1 in the above exercise then f(x) is continuous at

x = 1, but its derivative is discontinuous at this point.

1.17.  ESTIMATES OF INTEGRALS

Sometimes we cannot evaluate a definite integral explicitly but

we can estimate it approximately. This is achieved by integrating an
inequality.

Example 1.27 Show that

2 ( 2p ) p+ 1
∫ x p sin 2 ( x ) dx ≤ for p > −1.
0 p+ 1
The difficult part is usually to find a suitable inequality that can
be integrated. We do this by trial and error in order to come up
with an approximate integrand that can be integrated and that
 Derivatives and Integrals  •  33

gives a useful bound. In the present example we could note

that sin2(x) ≤ 1 and, over the range 0 to 2p, xp ≤ (2p)p so
xp sin2(x) ≤ (2p)p.
Integrating both sides gives
2 p+1
ò 0
x p sin 2 ( x) ≤ (2π) .
A better approximation is obtained if we get rid of just the
troublesome sin2(x) from the integrand using the fact that
sin2(x) ≤ 1. We then have that
xp sin2(x) ≤ xp
for 0 ≤ x ≤ 2π since sin(x) ≤ 1. Then, integrating both sides,
2π 2π
∫ x p sin 2 ( x ) dx ≤ ∫ x p dx
0 0
2
 x p+ 1 
≤ 
 p + 1 0
( 2π ) p+ 1
≤ .
p+ 1

Note that the answer is not unique – the usefulness of the

inequality you obtain will depend on how you choose the inequality
to integrate.

Exercise 1.30 From sin(x) ≤ x for 0 ≤ x ≤ 1 show that

1
∫ ( sin ( x ) )
1 p
dx ≤ , for p ≥ 0.
0 p+1

1.18.  DERIVATIVES OF INTEGRALS

Let
x
F ( x ) = ∫ f ( t ) dt
a
34  •  Mathematical Physics

where a is a constant. Then, since integration is the inverse of dif-

ferentiation, we have
dF
= f ( x ).
dx
a
=
We can check this as follows. First note that F ( a) f ( t ) dt
∫= 0.
a

Then, changing the integration variable from x to t for clarity,

x dF x
∫ dt = F ( x ) − F ( a ) = F ( x ) = ∫ f ( t ) dt.
a dt a

Alternatively, it is sometimes useful to go back to the definition

of the derivative:
x+h
F ( x + h )=
− F ( x) ∫ f ( t ) dt ≈ hf ( x ) .
x

Dividing through by h and taking the limit as h → 0 gives the

result.

Example 1.28 Let f(t) = t2; then

x
x  t3 
x x3
=
F ( x) ∫ f (=
t ) dt ∫ =t dt = 2
 ,
0 0
 3 0 3
so

dF 1 d 3
=
dx 3 dx
{ }
x= x=
2
f ( x ).  

It should be obvious that in practice there is no need to carry out

the integration explicitly to obtain the answer. We’ve just done it to
show how it would work. Now suppose
g( x )
F ( x) = ∫ f ( t ) dt
a

where g(x) is some given function of x. Then,

dF dg
= f ( g ( x)) .  (1.15)
dx dx
 Derivatives and Integrals  •  35

The proof is as follows: F is a function of g(x), i.e. a function of

a function, so
dF dF dg dg
= = f ( g ( x)) .  (1.16)
dx dg dx dx

x2
Example 1.29 Calculate dF/dx if F ( x ) = ∫ e− t dt
0

Note that f(g(x)) is obtained by replacing t in f(t) by g(x) = x2.

We have g(x) = x2, and f(t) = e–t. So
dg /dx = 2x,
2
f (g (x)) = e- x .
Then, using equation (1.15),
dF 2
= e− x ◊ 2 x.
dx

Exercise 1.31
e2 x
∫ ( ln( t ) )
12
(i) If F ( x ) = dt what is dF/dx?
1

1
∫ ( ln( t ) )
12
(ii) If G( x ) = dt what is dG/dx?
e2 x
(Do not attempt to compute the integrals explicitly! For (ii) begin
by finding the relation between G and F.)

1.19.  REDUCTION FORMULAE

Sometimes an integration cannot be performed directly but can

be related to a simpler integral which can be evaluated explicitly.
Reducing the problem in this way leads to a reduction formula.
 36  •  Mathematical Physics

p2
Example 1.30 Find ∫ cos3 (q ) dq .
0
Reduction formulae are usually obtained by integration by
parts. The key is to write the integrand as a suitable product.
Here we write (with the benefit of hindsight) cos3(q) = cos2(q)
cos(q), then
p2 p2
=I cos (q ) dq ∫
∫=
3
cos2 (q ) cos (q ) dq .
0 0

Next, integrate by parts with u = cos2(q) and v′ = cos(q), to give

( 2 cos (q ) sin (q ) ) sin (q ) dq .

p2 p2
= cos2 (q ) sin (q )  0 + ∫
0

As cos(π /2) = 0 and sin(0) = 0 we have

p2
= 0+∫ 2 cos (q ) sin 2 (q ) dq .
0

Then, removing the sine terms using a standard identity,

2 cos (q ) (1 − cos2 (q ) ) dq
p2
= ∫
0
p2 p2
= 2∫ cos (q ) dq − 2 ∫ cos3 (q ) dq
0 0

Notice that the last term above is twice the integral we started
with, I. So,
p 2
=I 2∫ cos (q ) dq − 2 I.
0

Finally we can collect all the I terms on the left to give a

reduction formula expressing the integral of cos3(q) in terms
of the integral of a reduced power (here cos(q)) which we can
evaluate,
p 2 p 2 2
2 ∫ cos (q ) dq =
3I = 2, ⇒ ∫ cos3 (q ) dq = .
0 0 3

The reduction formula (sometimes also called a recurrence rela-

tion) expresses the required integral in terms of a similar integral,
but with a reduced value of some parameter. Reduction formulae
 Derivatives and Integrals  •  37

are very often obtained by integration by parts. You may have to

experiment a bit to get the integrand in a suitable form. Reduction
formulae of the form found in Example 1.31 occur in many other
contexts as a way of specifying an infinite number of terms.

Exercise 1.32 Use a similar method to find a reduction formula

p 2
for ∫0 cos ( x ) and hence evaluate the integral.
5

Example 1.31 Let

p2
=Cn cos ( q ) dq, for n
∫=
2 n−1
1,2,3,...
0

Show that for n > 1,

( 2n − 2 )
Cn = Cn−1 .
2n − 1
We begin by splitting the integrand into a product cos2n–2 (q)
cos(q) so we can use integration by parts. Then we write
u = cos2 n-2(q ), and v¢ = cos (q )
Then

u=′ ( 2 n − 2 ) cos2 n−3 (q ) × ( − sin (q ) ) , and u= sin (q ) ,

π/2 π/2
=Cn cos2 n−2 ( q ) sin( q )  0 + ∫ ( 2 n −2 ) cos2 n−3 ( q ) sin( q ) sin( q ) dq.
0

Since Cn involves only cos(q) we next remove the sine terms

using sin2(q) = 1 – cos2(q). Then,

0 + ∫ ( 2 n −2 ) cos2 n−3 ( q ) ( 1 −cos2 ( q ) ) dq,

p2
Cn =
0

( 2 n −2 )  ∫0 cos2 n−3 ( q ) dq −∫0 cos2 n−1 ( q ) dq  .

p2 p2
=
 
The integrals can now all be written in terms of Cn and Cn–1,
where Cn–1 is obtained from Cn by replacing n by n – 1. We have
Cn = (2n – 2)Cn – 1 – (2n – 2)Cn.
38  •  Mathematical Physics

Finally, collecting all the Cn terms on the left, we obtain the

reduction formula or recurrence relation for Cn,
( 2n − 2 )
Cn = C .
( 2 n − 1 ) n −1

Example 1.32 Following from Example 1.31 show that

( 2 n −2 )( 2 n −4 ) ... ( 2 )
Cn = .
( 2 n −1)( 2 n −3) ... ( 3)
We begin with
( 2n − 2 )
Cn = C .
( 2 n − 1 ) n −1
Then, replacing n with n – 1,
( 2n − 4 )
Cn−1 = C ,
( 2 n − 3 ) n−2
and comparing these two equations we see that
( 2 n − 2 )( 2 n − 4 )
Cn = C .
( 2 n − 1)( 2 n − 3 ) n−2
If we continue until we get C1 on the right hand side, we find
that
Cn =
( 2 n − 2)( 2 n − 4) ... ( 2) C .
( 2 n −1)( 2 n − 3) ... ( 3) 1
At this point we can stop because we can integrate C1 explicitly
and also the reduction formula no longer applies. We have
p2
=C1 cos( q ) dq
∫= 1
0

and so, as required,

( 2 n −2 )( 2 n −4 ) ... ( 2 )
Cn = .
( 2 n −1)( 2 n −3) ... ( 3)
 Derivatives and Integrals  •  39

The result of example 1.32 is an explicit expression for the inte-

gral of an odd power of cos(q) which is sometimes useful. You should
remember that such a formula exists! The formula for the integral of
an even power in Exercise 1.34 is obtained in exactly the same way.
Expressions for the integrals of powers of sin(q) can also be found.
p2
Exercise 1.33 Find a reduction formula for ∫ cos7 ( x ) dx and
0
hence evaluate the integral.

Exercise 1.34 Let

p2
=Dn cos ( q ) dq, for n
∫=
2n
1,2,...
0

Show that for n > 1,

2n − 1
Dn = Dn−1
2n
and hence that
( 2 n − 1)( 2 n − 3 ) ... (1) p
Dn = .
( 2 n )( 2 n − 2 ) ... ( 2 ) 4
Exercise 1.35 The gamma function is defined for positive inte-
ger arguments (n ≥ 1) by

Γ(n) = (n – 1) Γ(n – 1),

with Γ(1) = 1. Show that Γ(n) = (n – 1)(n – 2)(n – 3). . . 1, and
hence that Γ(6) = 120.

Revision Notes

After completing this chapter you should be able to

• Know the derivatives of elementary functions including
powers as well as log and trigonometric functions
• Differentiate any combination of elementary functions
using the chain rule, product rule or by implicit
differentiation
40  •  Mathematical Physics

• Find and identify the stationary points of a given

function
• Recall the integrals of simple functions including
powers, sin and cos
• Know the following methods of integration: substitution,
use of trigonometric identities, partial fractions,
integration by parts, reduction formulae
• Calculate the definite integral of a discontinuous function
• Estimate the value of an integral by integrating an
inequality
• Differentiate an integral with respect to a limit
• Be able to obtain reduction formulae by integration by
parts

Further Questions
The sets of additional questions ending chapters divide into
two groups. Although they are not necessarily straightforward, the
first set comprises exercises that usually illustrate single concepts
or techniques from the text. The second set contains more diffi-
cult problems and requires a degree of insight usually drawing on
or illustrating a range of ideas, possibly using material from earlier
chapters.

1.20. EXERCISES

1. A continuous graph y = f(x) consists of the infinite straight

lines
3 x − 2 for x≥1
y=
−3 x + 4 for x ≤ 1.
Sketch the graph and find a formula, involving modulus
signs, for f(x).
2. Solve the equation 2x − 4 = |x − 3|.
 Derivatives and Integrals  •  41

( ) xx
3. Find the derivatives with respect to x of xx and x .
4. Simplify x + x + 5 x − 6 x − 10
5 3 2

x2 − 2
5. Verify that the derivative of the function
3t
g ( t ) tan −1 t −
=
3 + t2
is equal to 4t4/(1 +t2)(3 +t2)2, and deduce that for t > 0
(3 + t2)tan−1t > 3t.

6. Show that f(x) = 3x3 − 3x2 + x is an increasing function of x

and explain why the equation f(x) = 1/2 must have a unique
solution between x = 0.8 and x = 0.9.

7. Verify that d tan −1  sin x   = 2 cos x − 1 .

dx   2 − cos x   5 − 4 cos x
8. Show that the function f(x) = x3 − 6x2 + 12x − 6 is increasing
for all real x. For what values of x is its inverse function g (such
that g(f(x)) = x) defined?
9. Find the stationary points of y = 3x4 − 4x3 − 72x2 + 1 and
determine whether they are maxima, minima, or points of
inflection.
10. Verify that if the approximate integration formula
1 1
∫ f ( x ) dx ≈  f ( −1) + λf ( 0 ) + f ( 1) 
−1 m
gives the correct answer when f is a constant then
m = 1 + λ/2, and that if it also gives the correct answer when
f(x) = x2 then λ = 4 and m = 3.
11. Find the function the derivative of which is 2/(1 − 6x +
11x2–6x3) and which vanishes when x = 2/3.
12. Find ∫ 2 (1 + x2 ) xdx .
4

13. With the help of suitable changes of variables, find

∫ x (2 − 5x )
5 3 23
(i) dx
42  •  Mathematical Physics

∫ x (1 − x )
2 12
(ii) dx
sin x cos3 x
(iii)
∫ 1 + cos2 x dx
−1
(iv) tan x dx
∫ 1 + x2
x11
(v)
∫ x8 + 3 x4 + 2 dx
dx
14. Find ∫ 2 3 4 3 .
x −x
15. Indicating the method used, obtain the values A = 1, B = 2
for the constants in the partial fraction identity
A B 5 −x
f ( x) = + , where f ( x ) = .
6 −x 7 −x ( 6 −x)( 7 −x)
At which points P, Q does the curve y = f(x) meet the axes
0x, 0y? Show that the area of the region bounded by OP,
OQ and the arc PQ of the curve is ln(49/24).
16. In the partial fraction expression
2t A B
= + ,
(1 + t )(1 + 3t ) 1 + t 1 + 3t
you are given that A = 1; find the value of B. Hence show
that
1 2t2 p p
∫0 (1 + t 2 )(1 + 3t 2 ) dt= 4 − 3 3 .
17. Use integration by parts to show that

∫ ue du = ue u − e u + C,
u

(C an arbitrary constant) and with the help of a suitable

q
change of variable find ∫ esin( t ) sin ( 2 t ) dt .
0

18. With the help of an integration by parts, show that

p
∫ x cos ( x ) dx = −2.
0
 Derivatives and Integrals  •  43

19. Use integration by parts to show that

1
) ds ( ln ( s ) ) s
∫ 2 s ln ( s= − s2 + C 2

2
(C a constant) and with the help of a suitable change of
variable deduce that
1 2
sin ( 2 t ) ln ( sin ( t ) ) dt =
p2
∫p 4
ln   .
4 e
 y
ln   d ( y cos( y) ) = 1.
p
20. Show that ∫ p2  p
21. It is known that
4x A B C
= + +
( x + 1)( x − 1) 2
x + 1 ( x − 1) ( x − 1)
2

where A, B and C are constants. By multiplying both sides

by x + 1 and then putting x = −1 in the resulting equation
show that A = −1. Similarly, show that B = 2. Finally, by
putting x = 0 in the original equation, or otherwise, obtain
the value of C.

22. Find the range of values of x for which 1 1

≤ 1 − x2 and
1+ x 2
2
by integrating the inequality over this range deduce that
10
p≤ .
3
23. A continuous function f(x) is defined by
( x − 1 )
2
for x ≤ 1
f ( x) =  3
− x + 4 x − 5 x + C for x > 1.
2

Find the value of the constant C and show that f(x) is dif-
ferentiable at x = 1.
24. Given that the sequence an is generated by the recurrence
relation
( n + 1)( n − 3 )
an = a n −1
( n − 1)!
with a0 = 1, find a1, a2, a3, a4 and a2001. Note that 0! = 1 – see
the discussion of factorials in Section 2.1.
44  •  Mathematical Physics

1.21. PROBLEMS

1. A continuous graph y = f(x) consists of the straight line

segment joining the points (1, 8) and (2, 1) together with
infinite straight lines of respective slopes +5 and −5 over
the intervals ( – ∞ , 1) and (2, + ∞ ) . Find a formula for f(x)
of the form f(x) = ±|ax – b|±| cx – d |+h, where a, b, c, d, and
h are constants.
2. Divide the polynomial x3 –7x2 + 36 by x – 6.
Let f(x) = (7x2 – 36)1/3. Determine the ranges of values of x
for which the inequality f(x) > x is true, and state the values
of x for which f(x) > 6.
Sketch the graph of y = f(x) for x > 0, indicating in particu-
lar the points at which the graph crosses the x-axis and the
lines y = x and y = 6.
Show, on the basis of this graph, that if a sequence of num-
bers x1, x2, x3,... is defined (from a given x1) by xn+1 = f(xn) for
n = 1, 2,... then for x1 > 6 the sequence (xn) is decreasing,
with limit 6, while on the other hand if 0 < x1 < 3 then x2 is
further away from 3 than x1. What happens if 3 < x1 < 6?
3. For y = xex calculate y′, y″ and y(′″). Hence write down a
formula for the nth derivative y(n).
Deduce the value of A in the formula for the nth derivative
of a product
( uu ) ( n) =uu( n) + au′u( n−1) + ....

If z = x sin(x) what is the value of d8z/dx8 at x = 0?

∞
∫ (1 + 2 x )
2 k +1 2
4. Let=
Ik e− x − x dx . Find the value of I0, obtain a
0

reduction formula for Ik and deduce that I3 = 493.

1 dx
5. Evaluate
∫ 1 + x + x2
0
. By expanding the integrand, written

in the form (1 – x)(1 – x3) –1, in a series of powers of x, and

integrating term by term, show that
1 1 1 1 1 p
1− + − + − + ... = .
2 4 5 7 8 3 3
 Derivatives and Integrals  •  45

6. The potential energy of a system is given by

V (x) = x4 – 14x2 + 24x.
Find and classify the equilibrium points, i.e. stationary
points of V(x), and sketch the general shape of the graph;
show that the minimum at x = 2 is not global (i.e. that V(x)
takes values less than 8). Find an approximate expression
for V(x) near the point x = 2, as a quadratic in x – 2.
7. A function f(t) is defined by the formula
t2
f (t) = .
(1 + t )( 2 + t )2
(a) Obtain an expression for f(t) in the form
A B
(t)
f= +
1 + t ( 2 + t )2
where A, B are real constants, showing the method
used.
(b) Deduce that for x > –1
x x
ln (1 + x ) − ∫
= f ( t ) dt.
1+ x 2 0

(c) The left side is therefore positive for x > 0. Deduce

that ln(7/5) > 1/3, and hence that e < 2.744.
(d) By any method, and quoting any standard series you
may use, find the constant term and coefficients of
x, x2, x3 in the expansions in ascending powers of x
for

(i) 
f(x), (ii) g(x) = ln(1 + x) – x/(1 + x/2), and state
for what values of x these expansions are valid.
8. Let f(x) be a differentiable function of x increasing on the
interval a < x < b. What condition is satisfied by the deriva-
tive f ′(x)? If f(x) is an increasing function show that 1/f(x)
is decreasing, except at points where f(x) = 0. Show that
46  •  Mathematical Physics

x2 − 1 is increasing for x ≠ 0 and deduce that x

f ( x) =
x x2 − 1
x
is decreasing except at x = ±1. Sketch a graph of 2 .
x −1

Deduce the range (or ranges) of x for which x

> 3x .
x −1
2

p2
9. Let Jm = ∫0 cos (q ) dq . Calculate the value of J0
m

p2
and J1. Writing Jm = ∫ cos m −1 ( q ) d sin( q ) , show that
0

m −1
Jm = Jm − 2 where m is a positive integer. Deduce that
m
2 n −1 2 n −3 1 2 n 2 n −2 2
=J2 n ⋅⋅⋅ J0 =
and J2 n+ 1 ⋅⋅⋅ J1
2 n 2 n −2 2 2 n + 1 2 n −1 3
where n is a positive integer, and hence that
2
p  2.4.6...2 n  1 J2 n
=  .
2  3.5... ( 2 n − 1 )  2 n + 1 J2 n+1
Show that J2n–1 ≥ J2n ≥ J2n+1 for any n, and use the relation

J2 n −1 2 n + 1
=
J2 n + 1 2n

to deduce that
J2 n −1 J2 n
=
lim =
lim 1.
n →∞J2 n + 1 n →∞ J2 n + 1

Hence show that

p  2 2 4 4 6 2n − 2 2n 2n 
= lim  ⋅ ⋅ ⋅ ⋅ ⋅⋅⋅ ⋅ ⋅ .
2 n→∞  1 3 3 5 5 2 n − 1 2 n − 1 2 n + 1 

10. Let
1
∫ x n (1 − x ) dx
m
=
I n, m
0
 Derivatives and Integrals  •  47

where n and m are positive integers. Show that

n
I n, m = In−1, m +1
m +1
and hence that
n! m !
I n, m = .
( n + m + 1)!
Verify this formula by direct integration in the case n = 1,
m = 2.
 CHAPTER

2
ELEMENTARY
FUNCTIONS

In the first part of this chapter we look at the representation of an

arbitrary function f(x) as a power series, f(x) ≈ a0 + a1x + a2x2 + . . .
where a0, a1, a2 . . . are constants. We call this representation the
“expansion” of the function (or more precisely, the power series
expansion of the function about the point x = 0). The expansion is
intended to be exact if we include the infinite number of terms of
the whole series or an approximation if we cut off the series after a
finite number of terms. Power series expansions are very useful if
we want to know the form of f(x) for small x, and this is often the
case in physical problems where systems operate close to equilib-
rium. We shall find that power series expansions are particularly
useful near to a stationary point or where the solution of a prob-
lem depends on a small parameter. Later on (in Fourier series) we
shall meet other types of expansion which will be useful for other
purposes.

In the second part of the chapter we introduce the exponential

and related functions and their power series expansions. Together
with polynomials, the trigonometric functions and their inverses
and combinations, these make up what are called the elementary
functions.
50  •  Mathematical Physics

2.1.  BINOMIAL EXPANSION

You are probably familiar with the fact that a function like
(1 + x)n, where n is a positive integer, can be expanded in powers of x.
For example, for n = 1, 2, 3,

(1 + x)1 = 1 + x,
(1 + x)2 = (1 + x)(1 + x) = 1 + 2 x + x 2,
(1 + x)3 = (1 + x)(1 + 2x + x 2) = 1 + 3x + 3x 2 + x 3.

We can see there is a pattern in the terms: there are n + 1 terms

in the expansion, and there is a symmetry to the coefficients in front
of each term (e.g. 1, 3, 3, 1 for n = 3).
In fact, the general expression can be written as a sum of n + 1
terms called the binomial expansion,

n ( n − 1 ) 2 n ( n − 1 )( n − 2 ) 3
(1 + x )n =1 + nx + x + x + . . . + xn  (2.1)
2 ⋅1 3 ⋅ 2 ⋅1

Example 2.1 Expand (1 + x)5.

We apply the definition of equation (2.1),
5 ⋅ 4 2 5 ⋅ 4 ⋅ 3 2 5 ⋅ 4 ⋅ 3 ⋅ 2 4 5 ⋅ 4 ⋅ 3 ⋅ 2 ⋅1 5
(1 + x )5 =1 + 5 x + x + x + x + x ,
2 ⋅1 3 ⋅ 2 ⋅1 4 ⋅ 3 ⋅ 2 ⋅1 5 ⋅ 4 ⋅ 3 ⋅ 2 ⋅1
=1 + 5 x + 10 x2 + 10 x3 + 5 x 4 + x 5 .

Exercise 2.1
(i) Expand (1 + x)7
(ii) Write out the three lowest order terms in the expansion of
(1 + x)12

Example 2.2 Expand (2 + x)5.

To use the definition from equation (2.1) we need to put this
in the form (1 + . . .)n. We therefore extract the factor 2 to give
 Elementary Functions  •  51

(2 + x)5 = 25(1 + x/2)5.

We can now use the binomial expansion to write
(2 + x)5 = 25 (1 + 5(x/2) + 10(x/2)2 + 10(x/2)3 + 5(x/2)4 + (x/2)5)
= 32 + 80x + 80x2 + 40x3 + 10x4 + x5.

Exercise 2.2 Expand

(i) (3 + 2x)4
(ii) (1 + x2)3
(iii) (a + x)n

Factorials
For n a positive integer, we define n! (pronounced “n factorial”)
as the product

n! = n × (n – 1) × (n – 2) × · · · × 2 × 1. (2.2)

For n = 0 we define 0! = 1. This definition has no deep meaning;

0! is simply defined this way for convenience.

Example 2.3 Find 5!.

5! = 5 × 4 × 3 × 2 × 1 = 120.

Factorials are common in mathematical physics; for example,

they are heavily used in statistical mechanics. Returning to equation
(2.1), we can rewrite the denominator of each coefficient: r ⋅ (r – 1) · · ·
3⋅2⋅1 = r!. Can we find a more compact way to write the numerator?

Example 2.4 Verify that

n!
n ( n − 1). . .( n − r + 1) =
( n − r )!
52  •  Mathematical Physics

From equation (2.2) we have

n! = n(n – 1) · · · 3 ⋅ 2 ⋅ 1,
and

(n – r)! = (n – r)( n – r – 1) · · · 3 ⋅ 2 ⋅ 1.
Taking the ratio of these two we have
n! n ( n − 1). . . 3 ⋅ 2 ⋅ 1
= .
( n − r )! ( n − r )( n − r − 1 ) . . . 3 ⋅ 2 ⋅ 1
We can now cancel terms from the right end of the numerator
and denominator. The terms in the denominator cancel with
the terms (n – r)·( n – r – 1) · · · 3 ⋅ 2 ⋅ 1 in the numerator,
leaving (n – r + 1) and larger terms on the top,
n!
= n ( n − 1). . .( n − r + 1) .
( n − r )!

We now use factorials to write the binomial expansion

(Equation 2.1) in a more compact way,
n
n!
∑
(1 + x ) n =
r =0 r ! ( n − r ) !
xr . (2.3)

Here the “big sigma” symbol Σ indicates that we sum over all
terms with an index r running from 0 to n (i.e. we put r = 0, r = 1, . . . ,
r = n and add the results).
Exercise 2.3 Using the definition from equation (2.3) verify that
(1 + x)3 = 1 + 3x + 3x2 + x3.

5!
Exercise 2.4 Evaluate for r = 0, 1, 3, 4, 5. Hence write
r !( 5 − r )!
out equation (2.3) for n = 5 and compare with Example 2.1.

The coefficients in front of the xr terms, i.e. n!/r!(n – r)!, are the
binomial coefficients. Sometimes n!/r!(n – r)! is considered as a func-
tion of n and r and is called “binomial n, r” or “n choose r.” There
 Elementary Functions  •  53

are also other notations that are commonly used for this expression;
 n
for example,   or nCr mean the same thing. So you might see the
r 
binomial series written as
n n
 n
= (1 + x )n ∑ n Cr x r , =
or (1 + x ) ∑   x r .
n

= r 0= r 0r 

Binomial Series
What happens if n is not a positive integer? We first consider
the case n = –1. Writing out the binomial series, equation (2.1), with
n = –1 gives the infinite series:
( −1)( −2 ) ( −1)( −2 )( −3 )
(1 + x )−1 = 1 − x + x2 + x3 + . . . (2.4)
2 3⋅2 
= 1 – x + x2 – x3 + . . . .  (2.5)
This is correct provided |x| < 1 because, then, the right hand side
is just a geometric series (section A.7) with ratio (–x). If |x| < 1, then
the left hand side is the correct expression for the sum of an infinite
geometric series. If |x| > 1 the RHS is the sum (or difference) of ever
larger terms which grows arbitrarily large in magnitude and so cannot
equal the LHS, which is a finite number.
In general, the binomial expansion of (1 + x)n is valid when n is
not a positive integer, and |x| < 1, but will be an infinite series of terms,
not the finite series of equation (2.1).
n ( n − 1 ) 2 n ( n − 1 )( n − 2 ) 3
(1 + x )n =1 + nx + x + x + . . . . (2.6)
2 ⋅1 3 ⋅ 2 ⋅1 

Example 2.5 Write down the first three terms of the binomial
expansion of (1 + 2x)1/2 and state the range of x for which the
expansion is valid.
Using equation (2.6) we have
11 
1  − 1
1 + (2x) +   2 x 2 + . . . (2.7)
2 2
(1 + 2 x )1 2 = ( )
2 2! 
54  •  Mathematical Physics

1 2
=1 + x − x +.... (2.8)
2 
The expansion is valid for |2x| < 1, i.e. for –1/2 < x < 1/2.

Exercise 2.5 Write down the binomial expansions of the follow-

ing, stating the range of values of x for which each is valid:
(i) (1 – x)–1
(ii) (1 + 3x)–1/2

Exercise 2.6 Use your answer from Exercise 2.5 (i) to obtain the
expansion
a a ax ax2
= + 2 + 3 +... ,
b− x b b b
valid for |x| < |b|. Use this expansion to show that a radio fre-
quency of 198 (= 200 – 2) kHz corresponds to a wavelength of
approximately 1515m. (Wavelength=wavespeed/frequency, and
the speed of radio waves is approximately 3 × 108 m s–1.)

Exercise 2.7 By considering the binomial series expansion of

(1 + δ)1/3 and then putting δ = x – 1, show that

5 5 1
x1 3 ≈ + x − x2
9 9 9
near x = 1. Verify that the value obtained when x = 1000/729 dif-
fers from the true value of x1/3 by less than 0.25%.

Exercise 2.8 If d is small compared with |x| and 1, find the expan-
sion (in powers of d) of
(3x + d )
(1 + d ) ( x − 3d )
2

to first order and to second order in d.

Use of the binomial expansion to approximate rational expres-

sions containing small quantities is an important tool that is commonly
used in scientific work.
 Elementary Functions  •  55

2.2.  MACLAURIN SERIES

The binomial expansion is a special case of a more general result

that enables us to represent any well-behaved function as a power
series.
If f(x) is a well-behaved function, the Maclaurin series of f(x)
gives an expansion for f(x) as a power series valid for small x,
1 2 1
f ( x )= f ( 0 ) + xf ′ ( 0 ) + x f ′′ ( 0 ) + . . . + x p f ( ) ( 0 ) + . . . .
p
(2.9)
2! p! 
Here, f ¢(0) stands for f ¢(x) = df/dx evaluated at x = 0; f ″(0) stands
for f ″(x) = d2f/dx2 evaluated at x = 0; f(p)(0) stands for f(p)(x) = dpf/dxp
evaluated at x = 0. The series can be written very compactly as
∞
x p ( p)
f ( x) = ∑ f ( 0 ). (2.10)
p = 0 p!

For sufficiently small x the sum of all the terms after the nth
is negligibly small compared with the sum of the first n terms. For
such x we can approximate a function by the first few terms of its
Maclaurin series.
Exercise 2.9 For the function f(x) = (1 + x)1/2, evaluate f(0), f ¢(0),
5 5 5
f≤(0), f   , f ′   and f ′′   .
4 4 4

Example 2.6 Find the first three terms of the Maclaurin series
of (1 + 2x)1/2.
We have
f ( x ) =+ (1 2 x ) ⇒ f ( 0 ) =
12
1,
1
f ′ ( x= ) (1 + 2 x )−1 2 ⋅ 2 ⇒ f ′ ( 0=) 1,
2
1
f ′′ ( x ) =− (1 + 2 x ) ⋅ 2 ⇒ f ′′ ( 0 ) =−1.
−3 2

2
56  •  Mathematical Physics

Inserting these into equation (2.9) we get

x2
(1 + 2 x )1 2 = 1 + x −+....
2
Notice that this result agrees with equation (2.8) where we
expanded (1 + 2x)1/2 using the binomial expansion. Any function
has only one series expansion about zero. The Maclaurin series
is just a general way of finding this.

Example 2.7 Find the Maclaurin series for sin(x). Write down
the general term.
We have
f ( x ) = sin ( x ) ⇒ f ( 0 ) = 0,
f ′ ( x ) = cos ( x ) ⇒ f ′ ( 0 ) = 1,
f ′′ ( x ) =− sin ( x ) ⇒ f ′′ ( 0 ) = 0,
f ′′′ ( x ) =
− cos ( x ) ⇒ f ′′′ ( 0 ) =
−1.

Therefore, for the Maclaurin series (Equation 2.9) we get

x3 x3
sin ( x ) =0 + x + 0 − + . . . =x − + . . . . (2.11)
3! 6 
The general term is simply the expression for an arbitrary
power of x in the expansion. From equation (2.11) we can
see that the sequence of terms in the expansion will look like
x − x3/3! + x5/5! − x7/7!+. . .
Each successive term in the series changes sign, and only odd
powers of x appear in the expansion. If p is an integer that labels
each term in the series (see Equation 2.10), then we can make
each successive term change sign by including a factor (−1)p,
and 2p + 1 will count all the odd numbers. This means that the
general term is (−1)px2p+1/(2p + 1)! where p is an integer. Hence
x3 ( −1) p x2 p+1
sin ( x ) =x − + . . . + +... (2.12)

6 ( 2 p + 1)!

∞
( −1) p x2 p+1
=∑ . (2.13)
p= 0 ( 2 p + 1)! 
 Elementary Functions  •  57

Exercise 2.10 Show that the next non-zero term of equation

(2.11) is x5/120.

When we only require the first few terms of the expansion of a

function of a function it is often easier to combine Maclaurin series
than to start afresh. In Example 2.8 we find the expansion of a func-
tion, f(x), where this is the case. If we were to try to work out the
Maclaurin series of this function directly we would, of course, obtain
exactly the same answer, but it would take much more work.

Example 2.8 Find the Maclaurin series for (1 + sin(x))−1/2 up

to terms of order x2.
Begin by writing out the inner Maclaurin series for sin(x)
keeping only relevant terms.

(1 + sin(x))−1/2 = [1 + (x − x3/6 + . . .)]−1/2 .

Expand the outer series as far as necessary using the binomial
form for (1 + x)−1/2
3
=1 − ( x − x3 6 + . . .) 2 + ( x − x 3 6 + . . .) + . . .
2

8
=−
1 x 2 + 3 x2 8 + . . . .

Exercise 2.11 Obtain the following expansions as Maclaurin

series and in (i)–(iii) write down the general term
(i) (1 + x)–1 = 1 – x + x2 – . . .,
(ii) cos(x) = 1 – x2/2 + x4/24 + . . .,
(iii) ln(1 + x2) = x2 – x4/2 + x6/3 + . . .,
(iv) (1 + ln(cos(x)))–1 = 1 + x2/2 + . . ..

In Exercise 2.11 the series (i) is correct only for –1 < x < 1,
whereas the cosine series (ii) (and the sine series also) are valid for all
values of x (though not very useful for numerical computation if x is
large). The series (iii) is valid for –1 ≤ x ≤ 1. It is beyond the scope of
this course to explain how this can be worked out.
58  •  Mathematical Physics

2.3.  TAYLOR SERIES

Maclaurin series are valid near x = 0, and enable us to expand a

function as a series for x near to 0. Suppose we want an approxima-
tion for a function near some other point. For example, we know that
sin(p/2) = 1 but what is sin(p/2 – d) approximately equal to for small
d? If we want an expansion near a point x = x0 we use the Taylor series
1 1
x ) f ( x0 ) + ( x − x0 ) f ′ ( x0 ) +
f (= ( x − x0 ) f ′′ ( x0 ) + . . . + ( x − x0 ) f ( p) ( x0 ) + . . . .
2 p

2! p!
(2.14)
Here f ¢(x0) stands for f ¢(x) evaluated at x = x0; f ≤(x0) stands for
f ≤(x) evaluated at x = x0 etc. Note that the Taylor series gives us f(x) as
a power series in (x − x0). The Maclaurin series is a special case of the
Taylor series with x0 = 0. By putting x = x0 + d we obtain an equivalent
form of the Taylor series
1 2 1
f ( x0 + d=
) f ( x0 ) + d f ′ ( x0 ) + d f ′′ ( x0 ) + . . . + d p f ( ) ( x0 ) + . . . (2.15)
p

2! p! 
which resembles the Maclaurin series with x0 replacing 0, and d
replacing x. The series can be written compactly as
∞
d p ( p)
∑
f ( x0 + d ) =
p = 0 p!
f ( x0 ) .

Example 2.9 Find the Taylor series of the function x1/3 about
the point x0 = 1 up to terms of order (x – 1)2.
We have
f ( x) = x1 3 and x0 = 1 ⇒ f ( x0 ) = 1,
1 1
f ′ ( x ) = x −2 3 ⇒ f ′ (1 ) = ,
3 3
2 1 −5 3 2
f ′′ ( x ) =− ⋅ x ⇒ f ′′ (1 ) =− .
3 3 9
Therefore, for the Taylor series (Equation 2.14) we get
1 2
1 + ( x − 1) − ( x − 1) + . . . .
2
x1 3 =
3 9
This is just the binomial expansion of (1 + (x – 1))1/3.
 Elementary Functions  •  59

Exercise 2.12 Show that the Taylor series expansion of sin(q)

about q = p/2 is
d2
sin ( π 2 + d ) =1 − +... ,
2
1
and hence that sin ( 85 ) ≈ 1 − ( π 36 ) . Verify that the error is
 2

2
less than 0.00025%.

Taylor’s Theorem: Loosely speaking, Taylor’s theorem says that

we can approximate a smooth function (which is infinitely differen-
tiable) at a given point, as an nth order polynomial function of the
displacement from that point (a Taylor series), and that the remainder
goes to zero (the approximation converges on the true function) as n
increases to infinity.

2.4.  EQUILIBRIUM POINTS

One of the major applications of the Taylor series is in the study

of physical systems near points of equilibrium. An equilibrium point
of a system characterized by a potential energy V(x) is a point at which
V′(x) = 0. An equilibrium point is stable if it gives a minimum of V(x)
and unstable if it gives a maximum.

Example 2.10 Suppose the potential energy of a certain

system with one degree of freedom is given as

V(x) = x4 – 8x2.
Sketch a graph of V(x), find and classify the equilibrium points,
and obtain an approximate expression for V(x) near to the point
of equilibrium where x > 0.
At equilibrium, by definition V′(x) = 0. We have V(x) = x4 – 8x2,
so V′(x) = 4x3 – 16x = 0. This has solutions x = 0 and x = ±2 and
these are the points of equilibrium. Now V″(x) = 12x2 – 16.
60  •  Mathematical Physics

V(x)
10

−4 −2 2 4 x The function V(x) = x4 – 8x2.

−10
−30

• At x = 0, V″(0) = –16 < 0, so this is a maximum, hence

unstable.
• At x = +2, V″(2) = 32 > 0, so this is a minimum, hence
stable.
• At x = –2, V″(–2) = 32 > 0, so this is another minimum,
hence stable.
Near x = +2 the Taylor series for V(x) is

( x − 2 )2 ′′
V ( x=
) V (2) + ( x − 2) V′(2) + V (2) + . . .
2
( x − 2) 2

=−16 + ( x − 2 ) ⋅ 0 + 32 + . . .
2
≈ −16 + 16 ( x − 2 ) .
2

Note how we find all the stationary (equilibrium) points and then
consider them one by one. Some may be maxima, hence unstable
equilibria; some may be minima, hence stable equilibria.

Exercise 2.13 Let the potential V(x) be given by

x 4 x3 8
V ( x) = + − x2 + .
4 3 3
Find and classify the equilibrium points of this potential. By con-
sidering the shape of the graph, show that x = −2 gives a global
minimum (this means V(x) ≥ V(−2) for every x, not just near
x = −2) and find an approximate expression for V(x) near this
point.
 Elementary Functions  •  61

Harmonic Behavior Close to An Equilibrium Point

Near to equilibrium every physical system behaves in an analo-
gous way to a mass on a spring.
For example, for any potential energy V(x), we can show that the
corresponding force near to a minimum, F = dV/dx, is proportional to
δ, the distance away from the minimum, i.e. F = −kδ, the same as for
a mass on a spring.
Close to a minimum, the Taylor series of V(x) cannot contain any
linear terms, because dV/dx = 0 at the minimum. From the general
Taylor series,
∞
d p ( p)
∑
V ( x0 + d ) =
p = 0 p!
f ( x0 ) ,

V ′′ ( x0 )
≈ V ( x0 ) + d 2,
2
where we have neglected higher order terms because these will
be small if the distance from the minimum, d, is small. Then, the
potential
1
=
V (d ) V ( x0 ) + kd 2 ,
2

where V(x0) and k = V″(x0) are constants. For any physical system
close to a minimum V(d) ∝ δ2 and the force F = −dV/dx is proportional
to the distance, F = −kδ.
This means that, close to a minimum in the potential energy, every
physical system behaves in an analogous way to a harmonic oscillator.

Example 2.11 The Leonard-Jones potential,

 x 12 x  
6

V ( x ) V0  0  − 2  0   ,
=
 x   x  
describes the potential energy of two atoms separated by
a distance, x. This has a minimum value of −V0 when x = x0.
Show that, near to x = x0, the restoring force felt by the atoms is
proportional to the distance, d, from the minimum.
62  •  Mathematical Physics

V(x)

0.5V0

x0 1.5x0 2x0
0
x

–0.5V0

–V0
δ

The Leonard-Jones potential (blue line above) has a minimum

when the atomic separation is x = x0. If atoms are too close
together (x < x0), or too far apart (x > x0), their potential energy
is large (V(x) > −V0).
Close to x = x0 we can expand V(x) as a Taylor series,

V ( x0 ) = − V0
V′( x) = −12 V0 x012 x −13 + 12 V0 x06 x −7 ⇒ V ′ ( x0 ) =
0
( x ) 156V0 x0 x − 84V0 x0 x ⇒ V ′′ (=
V ′′= 12 −14 6 −8
x0 ) 72 V0 x0−2 .

Then
1  72 V  2
V ( x0 + d ) =−V0 +  2 0 d ,
2  x0 

and the potential near to the minimum varies quadratically

(dashed curve above).
The restoring force is F = −dV/dx = −dV/dδ, so

 72 V 
−  2 0 d =
F= − kd .
 x0 
Thus, the force is proportional to the distance from the
minimum and, close to their equilibrium separation, a pair of
atoms will behave like a mass on a spring.
 Elementary Functions  •  63

2.5. DEFINITION OF e x (THE EXPONENTIAL

FUNCTION)

The number e is defined by the formula

1 1 1 1
e =1 + + + + +...
1! 2! 3! 4! (2.16)
= 1 + 1 + 0.5 + 0.166. . . + 0.04166. . . + . . .

It can be calculated to any number of decimal places by adding
together a sufficient number of terms of this infinite series. In fact
e = 2.7182818284590452353602874. . . .
The exponential function is defined as ex (i.e. the number e raised
to the power x). Often we write exp(x) for ex. Figure 2.1 illustrates the
fact that ex is an increasing function of x and ex → +∞ as x → +∞, while
e−x → 0 as x →+∞.

y
15

y
1.0
10

0.6

y = ex
y = e−x
5

0.2

x x
−4 −2 0 2 4 1 2 3 4 5
(a) (b)
FIGURE 2.1:  The functions (a) ex and (b) e −x .

2.6.  THE INVERSE FUNCTION OF e x

Using logarithms to the base e (see section A.3) we have

(i) eln(x) = x (by the definition of ln(x) as the power to which e
must be raised to obtain x)
(ii) ln(ex) = x ln(e) = x (by taking logs) since ln(e) = 1. Therefore
the function x → ln(x) is the inverse function to x → ex (be-
cause the application of one mapping followed by the other
brings us back to the starting value – see Section 2.10).
64  •  Mathematical Physics

y y = ln(x)

1
x
1 2 3

−1
−3
−5

FIGURE 2.2:  Graph of ln(x). Note that log(x) (for any base) is only defined for x > 0.

2.7.  DERIVATIVE OF THE EXPONENTIAL

If we assume that the properties of ln(x) are given we can obtain

the derivative of ex by implicit differentiation (see Section 1.5). Let
y = ex, then
ln(y) = x
and so differentiating with respect to x gives
1 dy
= 1.
y dx
Therefore dy/dx = y and

d x
dx
{ e } = ex . (2.17)

In words, the rate of change of ex is ex. This is a fundamental
result that distinguishes the number e; it is not true if we replace e by
any other number. (See Exercise 2.17.)

Example 2.12 Find the Maclaurin series for ex.

Let f(x) = ex. We begin by computing f′(x), f″(x) etc. as needed
for the Maclaurin series.
f(x) = ex ⇒ f (0) = e0 = 1,
f ′(x) = ex ⇒ f ′(0) = e0 = 1,
f ″(x) = ex ⇒ f ″(0) = e0 = 1.
And so the Maclaurin series is
 Elementary Functions  •  65

x2
f ( x ) =f ( 0 ) + xf ′ ( 0 ) + f ′′ ( 0 ) + . . . .
2!
Therefore
x2 x3 xn
ex = 1 + x + + +...+ +.... (2.18)
2! 3! n! 

Equation (2.18) may also be taken as the definition of the

function ex.

Exercise 2.14 Find e3/2 to two decimal places by summing suf-

ficiently many terms of the Maclaurin series. Why would it be
impractical to calculate e10 or e−10 by the same method?

Exercise 2.15 From the Maclaurin series for ex verify that

d x
{e } = e x .
dx

Example 2.13 Find d {e5 x } .

dx
We use the chain rule. Let y = 5x, then
de y d y dy
=
dx dy
e × { }
dx
= e y ⋅ 5 = 5 e5 x .

Exercise 2.16 Use the chain rule to find dy/dx if

(i) y = eax,
x2
(ii) y = e ,
(iii) y = ef(x).

Exercise 2.17 By putting y = ax, taking logs and then differentiat-

ing show that
d x
dx
{ a } = a x ln ( a ) .

So in general
d x
dx
{ a } ∝ ax

with equality only if a = e.

66  •  Mathematical Physics

Example 2.14 Calculate the Taylor expansion of exp(−x2)

about x = 1 to first and second order.

f ( x) = 1 ⇒ f ( x0 ) =
2
e− x and x0 = e−1 ,
f ′( x) =
−2 xe− x ⇒ f ′ (1 ) =
2
−2 e−1 ,
f ′′ ( x ) =
−2 e x + 4 x2 e− x ⇒ f ′′ (1 ) =
2 2
2 e−1 .

Therefore, to first order, the Taylor series for exp(−x2) is

exp(−x2) = e−1 − 2(x − 1)e−1 = e−1(3 − 2x).

To second order, the Taylor series expansion is

exp(−x2) = e−1(1 − 2(x − 1) + (x − 1)2) = e−1(4 − 4x + x2).

A sketch of the function together with its Taylor expansion to

various orders is shown in Figure 2.3.

y
1.5

2nd 3rd
1.0
0.5

4th 9th
0.0

1st
−0.5

x
0.0 1.0 2.0

FIGURE 2.3:  The function exp(−x2) (solid blue line) together with its Taylor expansion to
various orders (see Example 2.14). The black dashed line shows the first order
approximation, and the long dashed grey line shows the Taylor expansion to second order;
3rd, 4th and 9th order approximations are also shown. Close to x = 1 the expansion is very
good with just a few terms. For example, the first order (linear) approximation is accurate
to within 2% for 0.82 < x < 1.12. Further from x = 1 we need to use a higher order expansion
to give a good approximation to the function.

Exercise 2.18 Calculate the Taylor series expansion of exp(−x) to

second order about the point x = 1.
 Elementary Functions  •  67

2.8.  INTEGRATION OF EXPONENTIALS

We now know how to find the derivatives of exponential func-

tions. The next step is to use this knowledge to find the integrals of
exponential functions.
Since
d x
ex =
dx
{e } ,
integration of both sides gives

∫ e dx= e x + constant.
x

Example 2.15 Find I = Ú eax dx where a is a constant.

Put u = ax. Then du = adx and
1 u 1 ax
a∫
=I e=
du e + C,
a
where C is a constant.

∫ xe
− ax2
Exercise 2.19 Find (Hint: put u = −ax2.)

∫e
− x2
Note that the indefinite integral dx (without the additional
factor of x in the integrand) cannot be expressed in terms of elementary
functions. Do not waste time trying! This is known as the Gaussian inte-
gral and is very important in probability theory and quantum mechanics.
y y
4

y = sinh(x)
4

x
y = cosh(x)
−3 −2 −1 1 2 3
2

−2
−4

x
−2 0 2

(a) (b)
FIGURE 2.4:  Graphs of the hyperbolic functions (a) cosh(x) and (b) sinh(x).
68  •  Mathematical Physics

2.9.  HYPERBOLIC FUNCTIONS

We define new functions, called hyperbolic functions, as combi-

nations of ex and e−x:
1
( x ) ( e x + e− x ) ,
cosh=
2
and
1 x
( x)
sinh=
2
( e − e− x ) ,

where sinh is pronounced either “shine” or “sinsh.”

Example 2.16 Show that cosh2(x) − sinh2(x) = 1.

We have
2
1  1 2x
x )  ( e x + e− x )=
cosh 2 (=  ( e + 2 + e−2 x ) ,
2  4
2
1  1 2x
x )  ( e x − e− x )=
sinh 2 (=  ( e − 2 + e−2 x ) .
2  4
Hence
1 1
cosh 2 ( x ) − sinh 2 ( x=
)
( 2 ) − ( −2=) 1.
4 4
(This result should be memorized and can be used without
proof.)

The hyperbolic functions are useful combinations of exponen-

tials which frequently appear in mathematical physics, for example in
quantum mechanics and relativity.
The functions cosh(x) and sinh(x) are analogous to cos(x) and
sin(x) and they satisfy similar identities, which sometimes (but not
always!) differ by a sign, as in Example 2.16. When we come to discuss
complex numbers we shall discuss the connection between the trigo-
nometric and hyperbolic functions more thoroughly.
We complete the analogy with trigonometric functions by
defining
 Elementary Functions  •  69

sinh ( x )
tanh ( x ) = ,
cosh ( x )
1
coth ( x ) = ,
tanh ( x )
1
cosech ( x ) = ,
sinh ( x )
1
sech ( x ) = .
cosh ( x )

tanh(x) is often pronounced as “tansh,” cosech(x) as “coshek” and

sech(x) as “shek.”

Exercise 2.20 Show that 1 − tanh2(x) = sech2(x) and that

coth2(x) − 1 = cosech2(x).

d
Example 2.17 Show that cosh ( x ) = sinh ( x ) .
dx

d d  e x + e− x  e x − e− x
dx
{ =
cosh ( x )}  =
dx  2 
 = sinh ( x ) .
2

Exercise 2.21 Show that d {sinh ( x )} = cosh ( x ) .

dx

Exercise 2.22 Show that sinh(x) has the same sign as x for all x
and hence that cosh(x) has a global minimum at x = 0.

2.10.  INVERSE FUNCTIONS

Example 2.18 If y = x2 then x = y1/2 is the inverse function.

However, because we usually write an independent variable as x,

we usually say that y = x1/2 is the inverse function to y = x2. (Alterna-
tively, x → x1/2 is the function inverse to x → x2.)
70  •  Mathematical Physics

Example 2.19 If y = sin(x) then x = sin−1(y); so y = sin−1(x) is

the inverse function to y = sin(x). The notation arcsin(x) is also
used for sin−1(x).

Note again that sin−1 (x) is not the same as (sin(x))−1.

If y = f(x) then x = f−1(y) is the function inverse to f that associates
(one or more) values of x with any given y (in the range of f).
y y y=x
3

3
y = f (x)
2

2
–1
y=f (x)
1

1
x x
–3 –2 –1 1 2 3 –3 –2 –1 1 2 3
–1

–1
–2

–2
–3

–3

(a) (b)
FIGURE 2.5:  (a) A graph of a function, y = f(x). (b) The graph of the inverse function y = f−1(x)
is obtained by reflection in the line y = x.

TABLE 2.1:  Table of some standard functions and their inverse functions. The
third and fourth rows give restrictions on the x and y values for which the inverse
function y = f −1(x) is defined.

Function x2 ex sin(x) cos(x) tan(x)

Inverse x1/2 ln(x) sin−1(x) cos−1(x) tan−1(x)
Range of x x≥0 x>0 −1 ≤ x ≤ 1 −1 ≤ x ≤ 1 −∞ < x < ∞
Range of y y≥0 (none) −π/2 ≤ y ≤ π/2 0≤y≤π −π /2 < y < π /2

In both the examples the inverse function is many-valued: for

y = x2, both y = +x1/2 and y = −x1/2 are inverse functions (for x ≥ 0).
Often the non-negative square root is intended if we write, say, 2 .
In example 2.19, for y = sin−1(x) (where x is between −1 and 1), the
value in the range −p/2 ≤ y ≤ p/2 is often intended, but any of the
functions sin−1(x) + 2np or (2n + 1)p − sin−1(x), with n an integer, is
also an inverse to y = sin(x). The other inverse trigonometric functions
cos−1(x) and tan−1(x) are similarly many-valued. See Table 2.1.
 Elementary Functions  •  71

Exercise 2.23 Find all distinct real solutions of the equation

tan(3q) = 1 in the range 0 < q < p.

2.11.  INVERSE HYPERBOLIC FUNCTIONS

Since the sinh and cosh functions are combinations of exponen-

tials, their inverses involve the logarithm.

(
Example 2.20 Show that sinh −1 ( x ) = ln x + x2 + 1 . )
We need to turn this into a problem about the sinh
function whose properties we know. So to begin we invert
y = sinh−1(x). Now we can use the definition of sinh,
x = sinh(y) = (ey − e−y )/2.
We try to turn this into an equation we can solve for y. To begin,
multiply through by 2ey and rearrange to give
e2y − 2xey − 1 = 0.
This is a quadratic in ey. Let u = ey (which means we are seeking
y = ln(u)); then,
u2 − 2xu − 1 = 0.
This is a simple quadratic equation that has solutions,

u=
x ± x2 + 1.
We choose the + sign because otherwise u < 0 and ln(u) does
not exist. Then, finally,

(
y = ln x + x2 + 1 . )

( )
Exercise 2.24 Show that cosh −1 ( x ) = ln x ± x2 − 1 and hence,
or otherwise, show that
d 1
dx
{cosh −1 ( x )} = ± .
x −1
2

Explain the ± signs, with reference to a graph.

72  •  Mathematical Physics

Revision Notes

After completing this chapter you should be able to

• Compute n!
• Write down the general coefficient in the binomial
expansion and appreciate the use of this expansion
• Compute Maclaurin and Taylor series for combinations
of elementary functions
• Define the exponential function ex and know its
properties
• Define the hyperbolic functions sinh and cosh and
derive their properties
• Find the inverses of simple trigonometric and hyperbolic
functions

2.12. EXERCISES

1. (a) Use Taylor series to show that

1 p 3
cos ( 61o ) ≈ − .
2 360
(b) By regarding 10 as (10.24 + d)1/2 where d = −0.24,
and using the fact that (10.24)1/2 = 3.2, calculate the ap-
proximate value 3.1625 for 10 . (The correct value is
3.1622778.)
(c) By a similar method, show that (2.2)1=3 ª 1.30059.
2. Show that if an angle A is measured in degrees as 90 − d
where d is small, then
p 2d 2
sin ( A ) ≈ 1 − .
64800
 Elementary Functions  •  73

p  1  d2 d3 
3. Show that sin  + d= 1 + d − − + . . . .
4  2 2 6 
4. By combining known series, find the Taylor series of
1+ x
  about x = 0, and use it to show that ln(1.25) is ap-
1− x
proximately equal to 488/2187.
5. Write down the first three terms in the binomial expansion
of 4 + x . Use this truncated series to estimate 4.5 and
calculate the percentage error in your result.
6. Find the first four non-zero terms in the Taylor series of
(ln x)2 about x = 1.
7. Find the first non-vanishing term of the Taylor series for
y = sin4(x) near x = 0.
8. Use a Taylor series to show that sin(32o) is approximately
1 2p
equal to + 3.
2 360
9. Write down the series for sin(q) in terms of q and deduce
the approximate formula
1
q ≈ sin(q ) + q 3 .
6
1
By considering the case q = p/6 obtain the estimate p = 3 .
8
10. By equating coefficients of xn on both sides of
(1+ x)n(1+ x)n = (1+ x)2n
deduce the identity
2
n
 n   2n 
∑   =  n .
r =0  r   
11. Sketch the graph of y = 1 + sin(x) for − p/2 £ x £ p/2 and
hence that of its inverse function. For what values of the
variable is the latter defined?
74  •  Mathematical Physics

12. By writing 10x in the form ey evaluate the indefinite

integral ∫ 10 x dx.
13. From first principles show that 1 − tanh2(x) = sech2(x) and
deduce that
coth2(x) − 1 = cosech2(x).
What is the value of cosh(x) when 3 cosh(x) = 5 sinh(x)?
14. Write down the Maclaurin series for sin(x) and obtain the
series for cos(x) by differentiation.
Verify that when the square
x3 x 5 x3 x 5
(x − + − . . .) × ( x − + − . . .)
3! 5! 3! 5!
is formed and terms of the same degree collected to-
gether, the coefficients of x2 and x4 agree with those
obtained by using the series for cos(2x) in the formula
1
sin 2 (=
x)
2
(1 − cos ( 2 x ) ) .
sinh ( x )
15. Sketch the graph of f ( x ) = . Has f(x) any maxima
or minima? cosh ( x )

16. Define the functions cosh(x) and sinh(x), and deduce that
each is the derivative of the other and that cosh(0) = 1,
sinh(0) = 0. Hence obtain the Maclaurin series for cosh(x)
as
x2 x 4
1+ +
2! 4!
and show that

e0.1 + e−0.1 ≈ 2.010008333....

 Elementary Functions  •  75

2.13. PROBLEMS
1 1+ x
1. From the formula tanh −1 ( x ) = ln( ) deduce that
2 1− x
d 1 and hence show (with the help of an
dx
( tanh −1 ( x ) ) =
1 − x2
integration by parts) that ∫ tanh −1 ( x ) dx can be written as
1
x tanh −1 ( x ) + ln (1 − x2 ) + C.
2
2. (a) S
 how that the function tanh−1(x) may be written as
ln 1 + x− ln 1 − x and that

d d
( 1 − x2 )
−1
tanh −1 ( x ) = and tanh ( x ) =
1 − tanh 2 ( x ) .
dx dx
(b) Sketch the curves y = tanh(x), y = tanh−1(x),
y = 1 − tanh−2(x) and y = (1 − x2)−1, indicating any axis
crossings and the asymptotic behavior for each.
(c) Show that the curves y = tanh(x) and y = 1 − tanh2(x)

cross at the point (( ln ( 2 + 5 )) / 2, ( ) )

5 − 1 / 2 and by
tanh −1
( x)
examining the limit lim or otherwise, that the
( x →1
1 − x )
2 −1

curve y = tanh−1(x) never becomes as large as

y = (1 − x2)−1.

3. Show that sinh −1 ( x ) = ln ( x + x2 + 1 ) and hence that

d 1
sinh −1 ( x ) = .
dx x +1
2

Use an integration by parts to show that

( x ) dx
∫ sinh =
−1
x sinh ( x ) − x2 + 1+ C.
76  •  Mathematical Physics

4. Verify that the following three methods for expanding

f ( t ) = e3t + t in a power series
2

a0 + a1t + a2t2+ . . .
all yield the same values for a0, a1 and a2.
(i) Substitute 3t+t2 for x in the series for ex and collect
powers of t.
2
t
(ii) Multiply together the power series for e3t and e .
1
Obtain the Taylor coefficients f ( 0 ) , f ' ( 0 ) ,
(iii)  f '' ( 0 )
2
1
f (0), f '(0), f '' ( 0 ) using
2
f′(t) = (3 + 2t)f(t),  and  f″(t) = (3 + 2t)f′(t) + 2f (t).
 CHAPTER

3
FUNCTIONS, LIMITS,
AND SERIES

In physics we will often describe physical processes or quantities

in terms of mathematical functions. Understanding the behavior of
functions is therefore an important part of understanding the phys-
ics. In this chapter we will look at curve sketching, at the symmetry
of functions, at approximations to functions and, importantly, at
the asymptotic behavior of a function as some variable tends to a
limit. We will also introduce the ideas of convergent and divergent
series.

3.1.  CURVE SKETCHING: QUADRATICS

It is often very useful to visualize a function to get an overview of

its behavior. This is particularly the case when a function f(x) is used
to describe a physical quantity. One simple way to visualize f(x) is to
sketch a graph of f(x) against x.
To sketch a quadratic function we need to find its maximum or
minimum, where it crosses the axes and how it goes to infinity.
78  •  Mathematical Physics

Example 3.1 Find the stationary point and zeros of the

quadratic y = x2 – 4x + 3 and sketch its graph.
We could find the stationary points by setting dy/dx = 0, and
find the zeros from the formula x =  − b ± ( b2 − 4 ac )  / 2 a .
 
Alternatively we can use the following procedure which begins
by “completing the square”:

y = x2 – 4x + 3 = (x – 2)2 – 1.
This has its smallest value (of −1) at x = 2 (since the smallest
value of the squared term is 0). So x = 2 gives a minimum. Next
determine the axis crossings – first the x-axis:

y
4
3

(x − 2)2 − 1 = 0,
2

⇒ (x − 2)2 = 1,
1

⇒ x − 2 = ±1.
x
1 2 3 4
−1

Hence x = 3 or 1. The graph crosses the y axis when x = 0,

which gives y = 3. Finally, y → +∞ (where “→” means “tends
to”) for both large positive and large negative values of x. The
graph is therefore as shown in the figure.

Exercise 3.1 By following the indicated procedure, sketch the

graphs of the following, labeling the maximum or minimum and
axes crossings.
(i) y = −4x2 + 12x − 9
(ii) y = 4x2 − 12x + 10
 Functions, Limits, and Series  •  79

3.2. CURVE SKETCHING: GENERAL

POLYNOMIALS

A polynomial function of degree n is a function of the form

y = anxn + an−1xn−1 + · · · + a3x3 + a2x2 + a1x + a0,

where an, an−1, . . ., a1, a0 are constant coefficients. An nth degree

polynomial has at most n roots (values of x where y = 0). Finding
these can be difficult for n = 3, 4, and there is no general method for
finding the roots of higher degree polynomials.
Nevertheless, in sketching graphs of any function we usually
look for:
•• behavior at + ∞ and −∞ (this gives us the asymptotes),
•• crossing of axes (y = 0, x = 0),
•• maxima and minima (dy/dx = 0; d2y/dx2 < 0 or > 0),
•• points of inflection (d2y/dx2 = 0).

Example 3.2 Sketch the graph of the cubic polynomial

y = 4x3 − 3x.
First, the asymptotes: when |x| is large 4x3 >> 3x so we can
neglect the 3x in comparison with 4x3. Then, for large |x|, y
varies like 4x3. This means as x → +∞ , y also tends to +∞, and
as x → −∞ , y also tends to − ∞ .
Now for the x-axis crossing. We see y = 0 when x = 0 or
x= ± 3/2 .
Next to find the maxima and minima:

dy/dx = 12x2 − 3 = 0
80  •  Mathematical Physics

when x = ±1/2.
To join smoothly onto the curve for x → ±∞, x = −1/2 must
give a maximum and x = 1/2 a minimum. This can also be
determined from the sign of d2y/dx2.
Also,

d2y/dx2 = 24x = 0

when x = 0, hence when y = 0. So the origin is a point of

inflection. Approaching the origin from x < 0 the curve is
getting steeper; beyond the origin it becomes less steep, i.e. the
slope gets less negative.

y
3
2
1

x
−1.5 −0.5 0.5 1.0 1.5
−1
−2
−3

Exercise 3.2 Sketch the graphs of the cubic polynomials

(i) y = x3 − 3x2,
(ii) y = x3 − 3x2 + 2.

Note that a quadratic has at most two zeros, while a cubic has at
most three and at least one zero. In general, a polynomial of degree
n (i.e. having highest order term xn) has at most n zeros; if n is odd
 Functions, Limits, and Series  •  81

it has at least one zero (because a polynomial of odd degree must

range between −∞ and +∞ and hence must cross zero).

3.3.  CURVE SKETCHING: RATIONAL FUNCTIONS

A rational function is a polynomial divided by another polyno-

mial, i.e. a fraction with polynomials as numerator and denominator.
See Section A.6.

Example 3.3 Sketch the graph of y = 1/x.

First, the behavior at infinity. For large |x|, y tends to zero. For
small positive x, y → + ∞ , and for small negative x, y → −∞.
Also, y ≠ 0 for any finite x.
Next, stationary points.

dy/dx = −1/x2 ≠ 0,
for any finite x. And
d2y/dx2 = 2/x3 ≠ 0,

meaning there are no points of inflection. So the graph is as

shown below.
y
2
1

x
−1 1 2 3
−1
82  •  Mathematical Physics

Example 3.4 Sketch the graph of y = x/(x – 2).

Asymptotes: since we want to look at large x, write y in a form
that involves 1/x for large x (which we can then neglect).

−1
1  2
=
y = 1 − 
1−2 / x  x
2 4
y =1 + + 2 +...
x x

(using the binomial series; see section 2.1). So y → 1 as |x| → ∞.

If x is near to 2 then |y| will get very large. But the sign of y
depends on whether x > 2 or < 2. For x – 2 small and positive,
y → + ∞; for x – 2 small and negative, y → –∞.
We can see that y = 0 for x = 0, and dy/dx ≠ 0 for any finite x. We
can also find that d2y/dx2 ≠ 0. So the graph is as shown below.

y
5
1

x
−1 2 3 4 5
−5

Exercise 3.3 Sketch the graphs of

(i) y = (x – 3)–1,
(ii) y = x + (2/x).
 Functions, Limits, and Series  •  83

3.4.  GRAPHICAL SOLUTION OF INEQUALITIES

A good starting point when solving an inequality is to sketch the

form of the curves involved. As we shall show below, the graphi-
cal method is foolproof while, in contrast, there are some pitfalls to
algebraic methods.

Example 3.5 What is the region determined by the condition

y > 0, y < 3 – x, y < x?
First we draw the lines y = 0, y = 3 – x, y = x and then shade the
disallowed regions.

y
y=x
3
2

y = 3–x
1

y=0 x
1 2 3

It is then easy to see that the permitted region is the interior

of the triangle with vertices at the points (0, 0),  3 , 3  ,
2 2
(3, 0).

Exercise 3.4 What is the region determined by the conditions

y < 4, y > –x, y > 2x + 1?
84  •  Mathematical Physics

Example 3.6 For what range(s) of x is 1/(3 – x) < 1/(x – 1)?

First, we let
y1 = 1/(3 – x),
y2 = 1/(x – 1),
then sketch the graphs of y1 and y2 .
y
3
2
1

y = 1/(3−x) xp y = 1/(x−1)
x
1 2 3 4
−1
−2

From the graph y1 < y2 for x > 3 and between xp and 1. The
crossover point is xp, where 1/(3 – xp) = 1/(xp – 1), i.e. xp – 1 = 3 – xp
so xp = 2. So 1/(3 – x) < 1/(x – 1) for 1 < x < 2 and x > 3.

The graphical method is useful because there are some common

pitfalls associated with trying to solve an inequality algebraically.
If we were to try to solve Example 3.6 algebraically it would
be tempting to proceed via the following incorrect procedure. We
require 1/(3 – x) < 1/(x – 1), hence, multiplying through by (x – 1)
(3 – x), we need (x – 1) < (3 – x) or 2x < 4, and so x < 2.
This is wrong because, for x < 1, multiplication by (x – 1)
multiplies the inequality by a negative number. This reverses the
sign of the inequality, a fact that has been ignored. This can be
 Functions, Limits, and Series  •  85

remedied by squaring the multiplier (i.e. multiply through by

(x – 1)2(3 – x)2), but we recommend that you stick to the graphical
method.

Exercise 3.5 Determine where

x 1
< .
x−6 x−4

3.5.  THE SYMMETRY OF FUNCTIONS

Symmetry is a very important idea in theoretical physics. For

example, the symmetries of the laws of physics determine the ele-
mentary particles found in nature. Here we will introduce the idea
of even and odd functions. As we shall see, we can often use the
symmetry of a function to make mathematical operations, like inte-
gration, much easier.

Even and Odd Functions

If f(–x) = f(x), the function, f(x), is even. This means that the
graph of f(x) is symmetric about the y-axis; f(x) = x2 and f(x) = cos(x)
are examples of even functions.
If g(–x) = –g(x), the function, g(x), is odd. This means that g(x) is
antisymmetric about the y-axis; g(x) = x3 and g(x)= sin(x) are exam-
ples of odd functions. Notice that the graph of an odd function must
pass through the origin: g(0) = 0.
A function may be neither even nor odd. For example, both
h(x) = exp(x) and h(x) = x + cos(x) are functions that are neither
even nor odd.
It is always possible to write a function as a sum of even and odd
functions,
h(x) = f(x) + g(x),
86  •  Mathematical Physics

where
1
( x)
f=
2
( h ( x ) + h ( − x ) ) is even,
and
1
( x)
g=
2
( h ( x ) − h ( − x ) ) is odd.

Exercise 3.6 Write ex as the sum of an even and an odd function.

Products of Even and Edd Functions

The product of two even functions is always even. So, because
f1(x) = x2 and f2(x) = cos(x) are both even, then h(x) = f1(x)f2(x)
= x2 cos(x) is even.
The product of two odd functions is always even. So, because
g1(x) = x and g2(x) = sin(x) are both odd, then h(x) = g1(x)g2(x) = x sin(x)
is even.
The product of an even function and an odd function is always
odd. So, because f(x) = x2 is even and g(x) = sin(x) is odd, then
h(x) = f(x)g(x) = x2 sin(x) is odd.

Integrals of Even and Odd Functions

We can make use of symmetry to simplify the integrals of even
and odd functions between limits that are symmetric about the ori-
gin. The integral of any odd function g(x) between −a and a is simply
zero. We can see this graphically in Figure 3.1: the area under the

y y y

–a a x x

–a a x

(a) (b) (c)

FIGURE 3.1:  (a) An example of an odd function, f(x) = sin 2x. (b) An example of an even
function, f(x) = x2. (c) Two functions, h1(x) = exp(x) (solid line) and h2(x) = x + cos x (dashed
line) that are neither odd nor even.
 Functions, Limits, and Series  •  87

curve between −a and 0 is equal and opposite to the area under the
curve between 0 and a. For any odd function,
a
∫ g ( x ) dx = 0.
−a

For any even function we have

0 a
∫ f ( x ) dx = ∫ f ( x ) dx.
−a 0

Again, this is illustrated graphically in Figure 3.1.

In general,

a
0 if h ( x ) isodd
∫− a h ( x ) dx =  a (3.1)
2 ∫0 h ( x ) dx if h ( x ) iseven. 

Exercise 3.7 Examine the following integrals. Decide whether

the integrand is even or odd and use this to evaluate the definite
integral.
1
(i) ∫ sinh ( x ) dx ,
−1

1
(ii) ∫ cosh ( x ) dx .
−1

3.6. LIMITS

In Section 3.1 we looked at the asymptotic limits of some sim-

ple functions and introduced the notation “x → a” which is read as
“x tends to a.” Here we will show how to find the limits of a function
in some more complicated cases.

Limits by Substitution
Consider the function y = x2. As x tends to 4, for example, then
y gets and remains arbitrarily close to 16, the value of y when x = 4.
88  •  Mathematical Physics

We write that as x → 4, y → 16. Equivalently we say that 16 is the

limit of y as x tends to 4 and write

= =
lim y 16 or lim x2 16.
x→4 x→4

Exercise 3.8 What is lim sin ( x ) ?

x → p /2

Example 3.7 If y = 1/x, what is limx → +∞ y?

As x → +∞, y → 0 because as x increases y becomes and
remains arbitrarily close to 0.

Example 3.8 Find the limit of y = e2+(1/x) as x → +∞.

We have 2+1/x → 2 as x → +∞ since 1/x → 0 from Example 3.7.
So y → e2 as x → +∞.

Example 3.9 Find the limit of y = 1/x as x → 0.

Here we need to be careful. As x tends to zero through positive
values (which we write as x → 0+), y increases arbitrarily
through positive values so lim x→0+ y = +∞. But as x tends to
zero through negative values (which we write as x → 0−), y gets
increasingly large and negative: so lim x→0− y = −∞.

Exercise 3.9 Find the limits of

y = (x − 3)−1 as x → 3 (consider the two cases x → 3+ and
(i) 
x → 3−)
(ii) y = xp as x → +∞ (consider the cases p > 0, p = 0, p < 0
separately)
(iii) y = x/(x − 3) as x → +∞.
 Functions, Limits, and Series  •  89

3.7.  INDETERMINATE FORMS

In many cases we cannot evaluate limits by simple substitution

as we did previously. For example, consider the following limits:

sin ( x )
x
 1
( a ) lim , ( b ) xlim x ln ( x ) , ( c ) lim x1/ x , ( d ) lim  1 +  .
x →0 x +
→0 x →+∞ 
x →+∞ x
Direct substitution gives the meaningless results 0/0, 0 × (−∞),
∞0, 1∞, respectively. It is hopeless to try to proceed further like this.
In particular 0/0 is not 1, but meaningless. And 0 × ∞ is not 0, or ∞,
or any other number, but meaningless.
Such expressions may nevertheless approach definite limits
(finite or infinite) − for example, 1, 0, 1, e, respectively, in the
above four cases. But we must use a different method to evaluate
these indeterminate limits. One method that sometimes works is
expansion in series. Often limits can be related to those already
known. Alternatively, a method that usually works is l’Hôpital’s
rule.

sin ( x )
Example 3.10 Find lim .
x x →0
Substituting x = 0 we get sin(0)/0 which is indeterminate. To
work out the limit we expand sin(x) as a series
x3
sin ( x ) x − + ...
x2
= 6 =1 − + ...
x x 6
Putting x = 0 in the resulting series leads to a determinate
result.
So
sin ( x )
lim = 1.
x →0 x
90  •  Mathematical Physics

ex
Example 3.11 Find lim where n is a positive integer.
x →+∞ x n

This is of the form ∞ /∞, so indeterminate. To work out the

limit, we expand ex as a series.
x2 x n −1 xn x n +1
1+ x+ + ... + + + + ...
ex 2! ( n − 1 )! n! ( n + 1 )!
=
xn xn
x2 − n x −1 1 x
= x − n + x1 − n + + ... + + + + ...
2! ( n − 1 )! n! ( n + 1 ) !
Now take the limit as x → +∞
ex 1
n
→ 0 + 0 + 0 + ... + 0 + + ∞ + ...
x n!
The negative powers of x become arbitrarily small, but the
terms involving positive powers of x become arbitrarily large.
Therefore the sum tends to
ex
lim = +∞.
x →+∞ x n

This is an important result which should be remembered (and

can generally be quoted without proof): as x → +∞, ex grows faster
than any power of x. As a corollary to this result (exercise 3.10) we
have that, as x → +∞ , e−x decreases faster than any power of x.

Exercise 3.10 Show that lim x→+∞ x n e− x = 0. (Hint: this is

(ex/xn)−1.)

Example 3.12 Find lim+ x ln ( x ) .

x →0

This is indeterminate of the form 0×(−∞). Since we cannot

expand ln(x) as a series about 0 we try to relate this limit to one
we know. This we can do by substituting x = e−u
x ln(x) = −e−uu.
 Functions, Limits, and Series  •  91

Then u → +∞ as x → 0+. Therefore

lim x ln ( x ) = − lim ue− u
x →0 + u →+∞

= 0.
And we used the result of Exercise 3.10 in the last line.

Example 3.13 Find lim x1/ x.

x →+∞
This has the indeterminate form ∞0. Since we cannot expand as
a series we relate this to a limit we already know. This we can
do by taking logs.
L = lim x1/ x
x →+∞

ln ( x )
ln ( L ) = lim .
x →+∞ x
To relate this to example 3.12 let u = 1/x. So u → 0+ as x →+∞.

= − lim+ u ln ( u )
u→0

= 0.
Now we see that ln(L) = 0 which implies L = 1.

3.8.  L’HÔPITAL’S RULE

An alternative method for finding limits that usually works is

l’Hôpital’s rule:
f ( x) f '( x)
lim = lim (3.2)
x → x0 g ( x ) x→ x0 g' ( x ) 

This is only valid when the limit (on the left) is indetermi-
nate, and both f(x0) and g(x0) are zero, or both are ±∞. Notice
that to apply the rule we differentiate both f(x) and g(x) and then
divide f ′(x) by g′(x). This is not the derivative of (f/g). Also, if the
92  •  Mathematical Physics

limit on the right is also indeterminate, you may apply the rule
again.

Example 3.14 Find lim+ x ln ( x ) using l’Hôpital’s rule.

x →0

To apply l’Hôpital’s rule we have to state the problem in the

form lim+ ( f g ) :
x →0

ln ( x ) −∞
lim+ = which is indeterminate
x →0 1/ x ∞
Using l’Hôpital’s rule with f(x) = ln(x) and g(x) = 1/x we get

= lim
=
( ln ( x ) )' lim 1 / x
x →0 (1 / x ) ' x → 0 + −1 / x
+ 2

= lim+ ( − x=
) 0.
x →0

Exercise 3.11 Find

1 − cos ( x )
(i) lim ,
x →0 x2
x
 a
(ii) lim  1 +  ( a =
constant ) ,
x →+∞  x
sin ( x ) − x
(iii) lim .
x →+∞ x3

3.9.  LIMITS OF INTEGRALS

Integrals with infinite limits: By definition

∞ b
∫ f ( x ) dx = lim ∫ f ( x ) dx.
a b→∞ a
 Functions, Limits, and Series  •  93

An integral that gives a finite answer is said to converge; when the

integral is not convergent it is said to diverge.

Exercise 3.12 For which values of the real number a does the
∞
integral ∫
1
x a dx converge?

Integrals with infinite integrands: If f(x) becomes infinite at

x = a we define
b b
∫ f ( x ) dx = lim+ ∫ f ( x ) dx.
a l→a l

1
Example 3.15 Evaluate ∫ 0
x −1/ 3 dx .
We have
1 1
∫0
x −1/ 3 dx = lim+ ∫ x −1/ 3 dx
l →0 l

3 1
= lim+  x2 / 3 
l →0  2 l
3 3 
= lim+  − l 2 / 3 
l →0  2 2 
3
= .
2

1
Exercise 3.13 For what values of a does the integral ∫0
xa dx
converge?

Exercise 3.14 By use of integration by parts evaluate

∞
I = ∫ xe− x dx.
0
94  •  Mathematical Physics

Exercise 3.15 Let

∞
=Γ ( n) ∫ t n−=
1 −t
e dt, for n 1,2,3,...
0

Use integration by parts to show that Γ(n) = (n − 1) Γ(n − 1) for

n > 1 and hence deduce that Γ(n) = (n − 1)!.

Γ(x), with n replaced by a real variable x, is called the gamma

function and satisfies the same reduction formula (but the integral
cannot be evaluated in an elementary manner except when x is an
integer).

3.10.  APPROXIMATION OF FUNCTIONS

1 x
( x)
Consider the function cosh=
2
( e + e− x ) . As x → +∞, e−x
becomes negligible compared to ex. Thus cosh(x) behaves like ex for
large x. Neglecting parts of complicated expressions under certain
conditions can often simplify calculations enormously, but we need a
precise rule for what we are allowed to do. To emphasize that a rule
is involved mathematicians use a special terminology: we say a func-
tion f(x) is asymptotically equivalent to a function g(x) as x → x0, and
write f ~ g as x → x0, if f and g behave similarly in the following sense:
f ( x)
Definition: f ( x )  g ( x ) as x → x0 if → 1as x → x0 .
g ( x)
“f ~ g” is read “f is asymptotically equivalent to g,” but scientists often
say “f is approximately equal to g” while using the same ~ notation
and meaning the same thing. This definition encapsulates the idea
that the difference between the two functions, when x is near x0, is
vanishingly small, relative to the size of the functions themselves.
Note that this is not the same as taking a limit. For example, as
x → 0 we have sin(x) → 0 but (as the next example shows) sin(x) ~ x.
To say this again: “→” and “~” are not the same because the former
gives a number and the latter a function.
 Functions, Limits, and Series  •  95

Example 3.16 Show that sin(x) ~ x as x → 0.

First we look at the definition. We need to verify that

sin ( x )
→ 1as x → 0.
x
This will be true if
sin ( x )
lim → 1as x → 0.
x →0 x
This was proved earlier in Example 3.10.

More generally we may write down formulae such as

sin(x) ~ x − x3/6 as x → 0. What is intended is that x − x3/6 is a more
accurate approximation for sin(x) near x = 0 than x is. More precisely
sin(x) − x ~ −x3/6 as x → 0.

Exercise 3.16 Verify that sin(x)− x ~ −x3/6.

Exercise 3.17 Use the definition of asymptotic equivalence to

show that
(i) (ex − 1)−1 ~1/x as x → 0,
(ii) (ex + 1)−1~ e−x as x → +∞,
(iii) (ex + 1)−1− 1 ~ −ex as x → −∞,
(iv) (cosh(x)/ sinh(x)) − 1 ~ 2e−2x as x →+∞.

An approximation frequently useful in applications of math-

ematics is Stirling’s formula
n
 n
n!∼   2p n ,  (3.3)
e
from which follow, with increasing degrees of accuracy,
(i) ln(n!) ~ n ln(n),
(ii) ln(n!) ~ n ln(n) − n,
96  •  Mathematical Physics

1
(iii) ln ( n!)  n ln ( n ) − n + ln ( n ) ,
2
1 1
(iv) ln ( n!)  n ln ( n ) − n + ln ( n ) + ln 2p.
2 2
The second one is usually sufficient in physics and should be memo-
rized. Note that you cannot get back to equation (3.3) by exponentiating
(ii), although you can do so from the more accurate form (iv). In general
asymptotic equivalences do not remain valid when exponentiated.
Exercise 3.18 Use trial and error on a pocket calculator to deter-
mine how large n must be for formula (ii) to give an answer accurate
to within 10%. For this value of n, how accurate an estimate does this
imply for n! itself, and how accurate an estimate does (iv) give for n!?

If we are not given the asymptotic form we proceed by trying to

identify the leading term, as in the following example.

Example 3.17 Find the asymptotic form for cosh(x) as

x → +∞.
Identify the dominant term
cosh(x) = (ex + e−x)/2 ~ ex/2 as x → +∞,
since e−x → 0. Check by using the definition

cosh ( x )
=
(e x
+ x− x )
= 1 + e−2 x → 1as x → ∞.
e x /2 ex

Exercise 3.19 Find the asymptotic form for sinh2(x) as x → +∞.

3.11.  SEQUENCES AND SERIES

A sequence is simply an ordered set of numbers, such as

{2, 4, 6, 8}, which may be finite or infinite. In this example, 6 is
the third element of the sequence. A series is an unevaluated sum
 Functions, Limits, and Series  •  97

of a sequence, in this case 2 + 4 + 6 + 8. The sum of this series is

20. (See section A.7 for a reminder.) We can find the limit of an
infinite sequence of numbers from the limit of a “corresponding”
function.

Example 3.18 Find the limit of the sequence 1/n as n → +∞

through positive integer values.
The function 1/x is equal to the sequence for x = 1, 2, . . . . So
the limit of the sequence is the same as limx→+∞(1/x) = 0. This
can also be seen directly because 1/n becomes and remains
arbitrarily close to 0 as n increases through positive integers.
The next example is less obvious.

Example 3.19 Find the limit of the sequence 1, 21/2, 31/3, . . .

The general term is n1/n. The function x1/x is equal to the
sequence for x = 1, 2, . . . . So the limit of the sequence is the
same as lim x1 x = 1 from Example 3.13.
x →+∞

Exercise 3.20 Find the limits as n → +∞ of

n ( n − 1)
(i)
n2 − 1
n
 a
(ii)  1 +  (Hint: use the result of exercise 3.11(ii).)
 n

3.12.  INFINITE SERIES

An important application of limits is to series (sums of sequences).

Given a sequence of numbers, such as {a1, a2, a3, . . . , an, . . .}, the
new sequence defined by
98  •  Mathematical Physics

S1 = a1
S=
2 a1 + a2
S3 = a1 + a2 + a3

n
Sn = ∑ ai
i =1

is the sequence of partial sums. If the sequence of partial sums con-

verges to some limit L, then we say the series converges with sum
L. If the series of partial sums does not converge, we say the series
is divergent.

Example 3.20 What is the sum of the series S = 1 + 1/2 + 1/4

+ 1/8 + . . .?
Imagine trying to construct the sum in an obvious way: we
start with 1; then add 1/2 to get 1 + 1/2 = 3/2; then add 1/4 to
1 1
get 1 + + = 7 / 4 and so on. The sequence S1 = 1, S2 = 3/2,
2 4
S3 = 7/4 . . . has a limit which is the sum of the infinite series. To
obtain this limit we construct the general term of the sequence
Sn. This is the sum of the first n terms
1
1− n
1 1 1
Sn =1 + + + ... + n−1 = 2 .
2 4 2 1
1−
2
The expression on the right comes from the formula for the
sum of a geometric progression with ratio 1/2 (see Section A.7).
Now to find the limit of the sequence
1
1− n
= =
S lim Sn lim = 2 2,
n →∞ n →∞ 1
1−
2
 Functions, Limits, and Series  •  99

(since 1/2n → 0 as n → ∞). Thus

1 + 1/2 + 1/4 + . . . = 2.

There are very few series that can be summed in this way
because we cannot usually find a simple closed form1 for the nth
partial sum, Sn, from which to obtain the limit. (Note that this
is not the same thing as a formula for the nth term of the series
which is often easy!) Instead, we might try to get an approxima-
tion to the sum by summing a sufficiently large finite number
of terms. For example, an approximate value for the sum of the
series 1 + 1/2 + 1/4 + 1/8 + . . . obtained by summing the first four
terms is S  S4 = 15/8 = 1.875 (which is in fact in error by 6.25 %).
If you add enough terms on your calculator you will end up with
the value of 2.000. . . . This is not a proof that the sum of the
series is 2. You have run out of decimal places on your calcula-
tor and you have no proof that the sum of the infinite number of
small terms you have not yet considered is less than the accuracy
of your calculator.
Here is another example where your calculator would mani-
festly be useless. Consider the series S = 1 + 1/2 + 1/3 + 1/4 + . . . .
The first 4 terms give S4 = 2.08. Is this a good approximation?
The next 4 terms give S8 = 2.7145. . . . What about 12 terms? Or
100? In fact, the sum creeps up and up very slowly. (The sum of
the first 100 million terms is less than 20.) We can guess that even
though each additional term is getting smaller there are so many
of them that the sum is always growing toward infinity. This is
true, but you cannot prove it on your calculator. (This will only
show that the sum is beyond the maximum number your calcula-
tor will store, not that it is beyond any number any calculator can
store!) The following is a proper proof that the series does not
have a finite sum.

1
A “closed form” expression is one involving a finite number of mathe-
matical operations on the standard functions. This definition is intentionally
a little vague as the set of allowed functions may vary depending on the
context.
100  •  Mathematical Physics

The essence of the proof is to show that the given series is greater
than a series which is known not to have a finite sum. There is no
rule for how to do this – just educated guesswork. Here we use the
trick of bracketing terms together. With a suitable choice of how to
bracket the terms we have

S = 1 + 1/2 + 1/3 + 1/4 + 1/5 + 1/6 + 1/7 + 1/8 + . . .

= 1 + 1/2 + (1/3 + 1/4) + (1/5 + 1/6 + 1/7 + 1/8) + . . .
≥ 1 + 1/2 + (1/4 + 1/4) + (1/8 + 1/8 + 1/8 + 1/8) + . . . ,

where the inequality arises because each term in the series has been
replaced by something equal or smaller. But for the new series we
can form the partial sums:
1 1 1
S >1+ + + +,
2 2 2
which clearly goes on getting bigger and bigger, i.e. the series
diverges to +∞.
Thus, before we try to approximate a series by summing a
number of terms, we need to know whether the series does have a
finite sum (in which case we say the series converges) or whether it
diverges. Courses on pure mathematics and mathematical reference
books give a large number of methods for doing this, based essen-
tially on comparison with known series. (The large number of meth-
ods has arisen because often a particular test will be inconclusive;
one goes through them in order until one finds a test that works.) We
give an example of comparison in which the bracketing technique
given above is used to establish convergence.

Example 3.21 Show that S = 1 + 1/22 + 1/32 + 1/42 + . . . has a

finite sum.
We are given some help in the statement of the question that
the series is in fact convergent. We therefore want to bracket
the terms in such a way that they are smaller than some known
convergent series. There is no rule for doing this – you just
have to spot the possibilities. (There is usually more than one.)
 Functions, Limits, and Series  •  101

S = 1 + (1/22 + 1/32) + (1/42 + 1/52 + 1/62 + 1/72) + . . .

< 1 + (1/22 + 1/22) + (1/42 + 1/42 + 1/42 + 1/42) + . . .
= 1 + 2/22 + 4/42 + . . .
= 1 + 1/2 + 1/4 + . . . = 2.

Therefore the series has a finite sum. In fact S = p2/6 as can be

shown by the use of Fourier series.

The same method works for the series 1 + 1/2k + 1/3k + . . .

for any constant k > 1. Henceforth you may generally assume that
∑
∞
n =1
1 / nk is convergent for k > 1 and divergent for k ≤ 1. Conver-
gence or divergence of other series can often be settled by compari-
son with one of these.

Example 3.22 Determine whether the following two series

are convergent or divergent:

2 3 4
(i) S = + + + ...
1×1 2 × 2 3× 3

32 − 2 2 4 2 − 2 2 52 − 4 2
(ii) S = + + + ...
13 23 33
(i) 
The terms in the numerator are growing linearly,
those in the denominator quadratically, so the series
looks something like 1/n. So try to prove it diverges by
comparing with 1/n.

2 3 4
S= + + + ...
1×1 2 × 2 3× 3
1 2 3
> + + + ...
1×1 2 × 2 3× 3
1 1 1
= + + ...
1 2 3
102  •  Mathematical Physics

The given series is therefore greater term by term than a known

divergent series and hence divergent.
(ii) The numerator is screaming out to be factorized as a square
minus a square. So we guess it grows linearly, whereas the
denominator is cubed. So overall we expect this to look like
1/n2, hence to converge. However, it’s difficult to see what is
happening term by term, so we write out the general term.

( n + 2 )2 − ( n + 1 )2
an =
n3
2n + 3
=
n3
2 3
= 2 + 3.
n n

S is the sum of two series each of which is convergent (from

the discussion following Example 3.21, and since a constant
multiple of a convergent series is convergent)
1 1
=S ∑=
a 2∑
n n
n 2
+ 3∑
n3
.

Hence it is convergent.

Direct comparison of the above kind works only for series of

positive terms. A series containing both infinitely many positive and
infinitely many negative terms will certainly be convergent if the
positive and negative parts are separately convergent. Other series
such as 1 − 1/2 + 1/3 − 1/4 + . . . , or more exotic examples such as
2 1 4 3 5
3 2 5 4 6
  −   +   −   +   + . . . , in which both positive and
2 1 4 3 5
negative parts are divergent, cannot be tested in this way. (Although
they do converge to ln 2 and 0.4456 . . . , respectively.)
 Functions, Limits, and Series  •  103

Exercise 3.21 Show that the series

1 1 1
S= + + + ...
1× 2 2 × 3 3× 4
has a finite sum, but the series
1 1 1
S=+ + + ...
1 2 3
is divergent. Does the series
1 1 1
S =1 − 2
+ 2 − 2 + ...
2 3 4
converge? (Hint: consider the positive and negative parts sepa-
rately.)

Revision Notes

After completing this chapter you should be able to

• Sketch graphs of polynomials and rational functions
• Solve inequalities graphically
• Write down conditions for a function to be continuous
• Identify indeterminate limits and use l’Hôpital’s rule or
series methods to find them
• Define what is meant by convergence of a series
• Define asymptotic approximations of functions and
verify given examples
• Calculate integrals over an infinite range as limits
104  •  Mathematical Physics

3.13. EXERCISES

1. Sketch the graph of y = x3 − 13x2 + 40x, indicating its inter-

sections with the axes and any stationary points and their
nature.
2. Show that the derivative of g(x) = x4 − 4x3 + 4x2 can be writ-
ten as 4x(x − 1)(x − 2), and hence sketch the graph of
y = g(x), indicating the positions of any maxima or minima.
3. State why it is obvious that x > 3(x − 2)/(x − 4) for all large
positive values of x. Determine the precise ranges of x
for which the inequality is true by rewriting it in the form
(x − A)(x − B)/(x − 4) > 0 and noting where the left side
changes sign, or otherwise.
4. Sketch the region consisting of the points in the x, y plane
for which the inequalities |x| ≤ 1 and x ≤ y ≤ 1 are
simultaneously satisfied, describe its shape and determine
the vertices.
1
5. Verify that = n + 1 − n , and hence write
n + n+1
down a formula for the sum of the first n terms of the series
1 1 1
+ + +  and deduce that this
1+ 2 2+ 3 3+ 4
series is divergent. How many terms are needed to give a
partial sum greater than 10?
6. Use l’Hôpital’s rule or the series for sin(x) to evaluate the
sin ( x )
limit of as x → 0.
x
7. Write down a short formula for the sum of the first N terms
of the infinite series
∞   p   p 
∑  n sin  2 n  − ( n + 1) sin  2 ( n + 1)  
n =1    
 Functions, Limits, and Series  •  105

and hence show that this series is convergent, with

p
sum 1 − .
2
1 + cos ( 2 x ) − 2 cos ( x )
8. Show that lim = −1 using l’Hôpital’s
x →0 sin 2 ( x )
rule.
e x − e− x − 2 x
9. Use l’Hôpitals rule to find the limit of
x − sin ( x )
as x → 0.
10. Show that
sin ( x ) − cos ( x ) 2
lim = .
x →p 4 16 x − p
2 2
8p
∞
 1 1 
11. Prove that the series ∑  2 n − 1 ⋅ 2 n + 1  is convergent.
n =1

∑
∞
12. Evaluate the sum n =1
2 − n. Hence show that the series

∑
∞ 2

n =1
2 − n converges.
13. Evaluate the limit of (ex − 1)/x as x → 0. Deduce the asymp-
totic form of ex − 1 as x → 0 . Show that (ea/x − 1)−1 ~ x/a as
x → ∞.

3.14. PROBLEMS

1. The function
x3 + 4 x2 + Ax + 1 x3 + x2 − 4
=f ( x) −
x2 + 2 x x2 − 4

tends to a finite limit L as x → −2. Show that A = 9/2, find

L, and also discuss the behavior of f(x) as x → 0+, x → 0−,
x → 1 and x → ±∞.
106  •  Mathematical Physics

2. Discuss the behavior of

x3 + x2 − 4 x − 2 x3 + 3 x2 + 4 x
−
x2 − 3 x + 2 x2 − 1
as x → ±1, x → 2 and x → ∞.
3. A function f (z) is defined by the formula
3 z4 − 9 z3 + 10 z − 2
f ( z) = .  (3.4)
z3 − 3 z2 + 2 z
(a) Obtain the expression (polynomial plus partial
fractions)
1 2 3
f ( z) = 3z − − − (3.5)
z z −1 z − 2

showing the method used. (Do not merely verify that

(3.4) reduces to (3.5).)
(b)  Assuming that for |z| < 1 (and only then)
1
= 1 + z + z2 + ...;
1− z
obtain the terms a−1, a0, a1 in the expression, valid for
0 < |z| < 1,
∞
f ( z ) = ∑ an z n .
n = −1

(c)  Show that if 0 < a < b < 1 then

1
1− b
3 2 b 1 − b
( b − a2 ) − ln( a ) − 2 ln( 1 − a ) − 3 ln( 21 ),
b
∫ f ( z ) dz=
a 2 1− a
2

and state to what extent the result would differ if

0 < a < 1 < b or 1 < a < b < 2.
4. From the fact that (cos(x))′ = − sin(x) obtain the
derivative of sec(x) as sec(x) tan(x), and also write down the
derivative of tan(x). Derive expressions (in terms of sec(x)
 Functions, Limits, and Series  •  107

and tan(x)) for the second, third, and fourth derivatives

of sec(x). Hence show that all non-zero coefficients in the
Maclaurin series for sec(x) are positive, and that this series
begins
x2 5 x 4
1+ + + ....
2 24

Deduce a formula of the form

cos(x) + sec(x) − 2 ~ Axp as x → 0.

where A and p are constants. Show also that the left side
can be written as
4sin 4 ( 2x )
cos ( x )

and use this to verify your asymptotic formula.

5. It is known that (for x ≠ −1)
3x − 2 A Bx + C
= + 2 (3.6)
( x + 1) ( x + 4 ) x + 1 x + 4 
2

where A, B, C are constants. Obtain the values of A, B, C by

the following method.
(i) Multiply both sides by x+1; show that substitution of
x = 1 in the resulting identity gives A = 1.
(ii) If f(x) denotes the right hand side of (3.6) show that
x f(x) → A + B as x → +∞, and hence that A + B = 0.
(iii) Finally, show that C can now be found by putting
x = 0 in (3.6).
6. From the identity
∞
xn ∞ ∞
x n 1 ∞ x n +1
=
∑
n 1 n ( n=
=∑ n n+=
+ 1) n 1
( 1
− 1
1 ) x n
=∑ − ∑
n x n 1 n+1
n 1=

(valid for −1 ≤ x < 1), deduce the formula

1
1 + ( − 1)ln (1 − x )
x
108  •  Mathematical Physics

for the sum of the series on the left. (You may quote the
Maclaurin series for ln(1 − x).) What is the limit of this
expression as x → 1 from below?
7. Write down the expansion of cos(q) in powers of q (valid for
all real q).
(a) By considering the identity cos(q) sec(q) = 1, show that
the first four coefficients in the expansion
sec(q) = a0 + a2q 2 + a4q 4 + a6q 6 + . . .

are given by a0 = 1, a2 = 1/2, a4 = 5/24, a6 = 61/720.

d2k
(b) Hence write down the value of
dq 2 k
( sec (q ) ) at
q = 0, for k = 0,1,2,3.
(c) Calculate the limit of (sec(q) − 1)/ q 2 as q → 0 both
from the series and also with the help of l’Hôpital’s
rule.
8. Write down the power series for ex and deduce the series
for cosh(x). Hence show that

2 cosh ( x) −1 < e x
for x > 0,
and
2 cosh ( x)−2  e x
− 1 for x → 0 +.

Show also that

( x ) dx < 12 ( e + X −1) for

X
∫ X > 0.
x
cosh
0

Are these two quantities asymptotically equal as X → 0+?

9. (i) Write down the expansion of 1/(1 − x) in an infinite
series of powers of x, and by integration obtain the
expansion
x2 x3 x 4
− ln (1 − x ) =x + + + + ....  (3.7)
2 3 4
 Functions, Limits, and Series  •  109

(ii)  Derive the inequalities

x2 x2 x3
x+ + < − ln (1 − x ) < x + + .
2 2 3 (1 − x )

By choosing a suitable x, show that the value of ln(10/9)

1
differs from 21/200 by less than %.
3
(iii)  Verify that the values for the limit
− ln (1 − x ) − x
lim( )
x →0 x
obtained from (3.7) and from l’Hôpital’s rule are equal.
10. (i) It is known that if ∑a n is a convergent series of
positive numbers and bn ~ kan as n → ∞ (where k is
a constant) then ∑b n is also convergent. Given that
∑
∞
n =1
n− p (where p is a constant) is convergent if and
only if p > 1, examine for convergence the series
1
(a) ∑
n ( n + 1)
1
(b) ∑
( n + n+1 )
(ii) Find the first four terms in the Maclaurin expansion of
(1 − x)−1/2, and show that the coefficient an of xn can be
written as
1 × 3 × 5 × ... × ( 2 n − 1 )
2 × 4 × 6 × ... × 2 n
and also show (by multiplying numerator and
denominator by 2 × 4 × 6 × . . . × 2n) that this is equal
to (2n)!/22n(n!)2.
110  •  Mathematical Physics

(iii) With the help of Stirling’s formula deduce that

∞
an  1 / ( p n ) as n → ∞ and hence the series ∑a x
n=0
n
n

is divergent when x = 1.

(iv) This series is, however, convergent when x = −1; what

should be the value S of its sum? Show that one would

expect over N = 3000 terms to be needed, before the

N −1

∑ ( −1) a
n
partial sum n together with all subsequent
n=0
partial sums, is within 0.01 of S.

∫ ( x − 1) e
−x
11. (a) Show that xe x + C and also find
dx =

∫ ( x − 1) e
−x
dx .
(b)  The function f(x) is defined by

( x − 1 ) e
−x
for x > 2,
f ( x) = 
 A ( x + 1 ) e x<2
x
for
1 −4
and is continuous at x = 2. Show that A = e .
3
X
(c) Evaluate, for X > 2, the integral ∫ f ( x ) dx . What is the
2
∞
meaning of ∫ f ( x ) dx and what is its value?
2
2
(d) Also find the values of ∫ f ( x ) dx ( Y < 2) and
Y

2 ∞ 8 −2
∫ f ( x ) dx and deduce that ∫ f ( x ) dx = e .
−∞ −∞ 3
12. (a) Find the values of the function (sin(x))/x and its first
two derivatives at x = 0. Hence or otherwise write down
the Maclaurin series for (sin(x))/x up to terms
in x2.
(b) What is the ratio of the values of x which make the
truncated Maclaurin series in part (a) and the original
function (sin( x)) /x zero? (Use the smallest possible
absolute value of x in the latter case.)
 CHAPTER

4
VECTORS

The physical world (as we perceive it) exists in three spatial dimen-
sions. Physical quantities such as velocities – which have both
magnitude and direction – need three components for their specifi-
cation. We call these vectors. (In fact, even more than this, a chemi-
cal system, or an economic model may have, say, 20 components;
such a system can often be represented conveniently by a vector
in the appropriate number of dimensions (20 in this example): the
algebra is the same.) You are going to learn to do algebra with vec-
tors (add and multiply them) in a geometrically meaningful way.
The first part of the chapter deals with the algebra and the sec-
ond with the geometrical picture. You can think of this in two ways:
either that the algebra enables you to do complicated geometrical
things rather mechanically (by following the algebraic rules), or that
the geometry enables you to get a picture of the algebra. In either

B B B
C

A
A A
D

(a) (b)
 
FIGURE 4.1:  (a) The vectors AB and BA . (b) The parallelogram law of vector addition.
112  •  Mathematical Physics

case you have to learn to switch smoothly between the two methods,
and this requires practice. Your reward will be command of a math-
ematical tool that pervades physics.

4.1.  BASIC PROPERTIES

Definition: A vector is an object with a magnitude and a direction.

Therefore any vector can be represented by a line segment with
the specified length (or magnitude) and direction. Note that this pic-
ture provides a representation of any physical quantity, for example
a force, that can be described by a magnitude and direction in space.
It does not say that a vector is an arrow, only that they behave geo-
metrically in the same way. (Bodies are accelerated by forces, not by
arrows on a page.) This is important: many physical quantities are
not represented by spatial arrows (velocities and forces in relativity
theory, for example).

A line segment
 with end-points A and B can be written AB . The
line segment BA is a vector of the same magnitude but opposite
direction (indicated by the direction of the arrowhead in Figure 4.1).
Note that vectors as defined here
 (“freevectors”)
 do not have
a specific point of application. So AB = − BA in Figure 4.1 even
though the line segments do not lie on one another; indeed all line
segments of the same magnitude and direction represent the same
vector.
Conversely, we can place the line segment representing a given
vector wherever is convenient; often this isfrom
 a single point, O,
which is then called the origin. The vector OP is said to be the posi-
tion vector of the point P. (In practical applications one may use
vectors and their points of application (“fixed vectors”) but these are
not dealt with here.) Line segments can be added: A to B followed
by B to C is the same as A to C (Figure 4.1). Therefore vectors can
be added in the same way to give a resultant vector:
  
AB + BC = AC. (4.1)
 Vectors  •  113

The sum is in the same plane as the summands. Equation (4.1)

is usually referred to as the parallelogram law. Figure 4.1 shows also
that vectors can be added in any order: we have also
    
AB + BC = AD + DC = AC.

Line segments can be multiplied by real numbers to give fur-

ther line segments; for example in Figure 4.2, A to C is one and
a half times A to B. Therefore vectors can be multiplied in the
same way:
 
AC = 1.5 AB.  (4.2)
 
Here AC is the vector in the same direction as AB , 1.5 times
as long. (In this context real numbers are also called scalars, so in
equation (4.2) we say that the vector is multiplied by the scalar 1.5.)
We can do both multiplication by scalars and addition, so, if l and m
are scalars, expressions like
 
l AB + m CD

are
vectors.
 Such expressions are referred to as linear combinations

of AB and CD .

C v

µv w
B

u
O λu
A
(a) (b)
 
FIGURE 4.2: (a) AC = 1.5 AB . (b) Linear dependence, with w = lu + mv.
114  •  Mathematical Physics

 
Exercise 4.1If ABis 2 units due north and AC is 1 unit due
east, what is AB + 2 AC ?
Exercise 4.2 A person swims across a river with a speed rela-
tive to the water of 4 ft s−1at right angles to the current which is
flowing at 3 ft s−1. Draw a figure to show the swimmer’s resultant
motion and check that it accords qualitatively with your intuition.
Exercise 4.3 Rain falls vertically at 10 m s−1; draw a vector diagram
to show how it appears to a person running at 5 m s−1.
Exercise 4.4 A, B, C, D are four points in space; and the seg-
ments AB, BC, CD, DA have mid-points P, Q, R, S, respectively.
Show that
 
(i) 2AP = AB
 
(ii) 2PQ = AC
(iii) PQRS is a parallelogram.

Usually, instead of writing AB , say, for the line segment represent-
ing a vector, we give a vector a name, v say. The bold font indicates that

this is the name of a vector; in hand written work we use v or v or v.

•• The magnitude of a vector v is written as |v| or v.
• A linear combination of two vectors u and v would be written
lu + mv.
• The zero vector is printed as 0 (and written 0 ), so that, for
example, u − u = 0.

4.2.  LINEAR DEPENDENCE; BASIS VECTORS

Linear Dependence
We have seen that two vectors may be added to make a third in
the same plane; e.g.

w = lu + mv.  (4.3)
 Vectors  •  115

If a relation of the form (4.3) holds we say that u, v, and w are

linearly dependent.
Conversely, if u, v, and w are vectors lying in a plane, with u and
v not parallel, then we can always find l and m such that equation
(4.3) holds by completing the parallelogram of vectors as in Figure
4.2(b). Thus, in the plane any two (non-parallel) vectors may be given
independently but any third vector must be expressible in terms of
these two as in equation (4.3). Thus, any three (non-parallel) vectors
in the plane are linearly dependent and any one can be expressed in
terms of the other two.
A set of vectors is said to be linearly independent if no vector in
the set can be decomposed into a linear sum of the other vectors. If we
rearrange equation (4.3) we get the criterion for linear dependence
lu + mv + nw= 0.  (4.4)

If this can only be satisfied when the coefficients are all zero
(i.e. l = m = n = 0) then the vectors u, v, w are linearly indepen-
dent. There is a close connection between the maximum number
of linearly independent vectors and the dimension of a given space:
the dimension of a “vector space” is the largest number of linearly
independent vectors we can choose.
In three-dimensional space we can specify three non-coplanar
vectors arbitrarily (say u, v, w) and construct any other vector (say x)
as a linear combination of these:
x = lu + mv + nw.

We say that x, u, v, and w are linearly dependent. It follows that

any four (or more) vectors in three-dimensional space are linearly
dependent and any one can be expressed in terms of the other three.

Basis Vectors
We can use these observations to provide a useful way of specify-
ing vectors. We make a convenient choice of the maximum number
of independent vectors (two non-parallel vectors in two dimensions,
three non-coplanar vectors in three dimensions). These we call our
basis vectors. Any other vector in that space can then expressed as a
linear combination of these basis vectors.
116  •  Mathematical Physics

It is often most convenient to choose the basis vectors to be of

unit length and mutually perpendicular. Two such vectors in the
plane are usually called i and j. In three-dimensional space, three
such vectors are labelled i, j, k (see Figure 4.3). These are said to
form an orthonormal basis: the ortho- indicates they are mutually
orthogonal, the normal means they are unit vectors (normalized to
have unit magnitude, i.e. of length 1) and basis means that they are
linearly independent of each other, and every other vector is linearly
dependent on them (can be made from a linear combination of the
basis vector).

j
k C
v
i
3
k
O j
1
j A B
2
i i
(a) (b)
FIGURE 4.3:  (a) Basis vectors in 2D and 3D. (b) Diagram for example 4.1.

Unless stated otherwise, i, j, k are taken to be oriented to form a

right–handed set; i.e. a rotation from i to j is clockwise when viewed
in the direction of k. Sometimes the fact that these are vectors of
unit length (“unit vectors”) is symbolized by a carat symbol (a “hat”)–
e.g. iˆ , ˆj, kˆ . However, we shall take it that the notation i, j, k itself
signifies unit vectors.

Example 4.1 Find the magnitude and direction of the vector

v = i + 2j + 3k.
The vector v is obtained by starting at the origin O and moving
1 unit along the i direction (call this point A), then 2 along the j
3 along the k direction (call this
direction (call this point B), and
point C). The length of v (line OC ) can be calculated using
 Vectors  •  117

Pythagoras’s theorem on the triangles OAB (to give the length

 
of OB ) and OBC (to give the length of OC ).
The length of the hypotenuse of OAB is

OB = 12 + 2 2 = 5 . The length of the hypotenuse of OBC

is OC = V = 12 + 2 2 + 32 = 14 units.
By considering the triangle OBC we see the angle between v
( )
and k is cos−1 3 14 ; by considering the triangle OAB we see
the angle between the plane containing v and k and the vector
(
i is cos−1 1 5 . )
Exercise 4.5 Draw a figure showing the magnitude and direction
of the vector v = i + j + 2 k.

In general, the magnitude of a vector v = v1i + v2j + v3k is

(v + v22 + v32 ) .
12
v = 2
1

You should remember this formula.

Example 4.2 If a = i – 3j – 6k and b = 7i + 2k, find 3a + 2b.

To get 3a we multiply the coefficients in front of each of
the basis vectors by 3; similarly to get 2b. Then we add the
coefficients for each of the basis vectors.

3a + 2b = 3(i – 3j – 6k) + 2(7i + 2k)

= 3i – 9j – 18k + 14i + 4k
= 17i – 9j – 14k.

Exercise 4.6 If u = 3i + j and v = i – 2j + 0.5k find expressions,

in terms of the basis vectors i, j, k, for
(i) u + v,
(ii) 2v – u.
118  •  Mathematical Physics

Example 4.3 Find the unit vector in the direction of the vector
v = i + 2j + 3k.
To obtain the unit vector we must divide each component
by the magnitude of v. We know from example 4.1 that
v= 12 + 2 2 + 33 = 14 , so the unit vector is

v 1
vˆ = = ( i + 2 j + 3k ) .
v 14

The introduction of a set of basis vectors provides the cru-

cial link between the geometrical picture of a vector (an object
with magnitude and direction) and the algebra of vectors (objects
formed from linear combinations of the basis vectors). By passing
between the two approaches we can formulate a problem in geo-
metric terms, solve it using algebra and then return to a geometric
picture, if that is appropriate. We shall see many examples of this
in what follows: it is one of the reasons why vectors are so useful in
physical science.

4.3.  THE SCALAR (DOT) PRODUCT

We have seen how to add vectors both geometrically and in

terms of a basis, and we have seen how to multiply a vector by a
scalar but so far we have given no meaning to multiplying vectors.
We are free to choose a definition for the product of two vectors,
but if the resulting construct is to have any use it must correspond
to some geometrical operation on directed line segments. It turns
out that there are two such operations. The first of these is called
the scalar or “dot” product and is defined here. Later we shall
meet another product called the vector product. These are the
only two meaningful ways of multiplying vectors. (Actually there
is another, called the tensor product, but that will not concern
us here.)
 Vectors  •  119

Definition: Let q be the angle between the two vectors a and b. The
scalar product (or “dot product”) of two vectors a and b, written a · b,
is defined to be the number |a||b| cos(q):

a · b = |a||b| cos(q)  (4.5)

Note that the definition implies the important result
a · a = |a|2.
Notice that the scalar product takes two vectors and produces one
scalar (a number). Also notice, from the definition (Equation 4.5),
that the scalar product is commutative:

a·b=b·a (4.6)
(the order of the two vectors in the scalar product does not affect
the result).
If n̂ is a vector of unit length making an angle q with a vector
v, then
= cos (q ) v cos (q )
nˆ ⋅ v nˆ v=

is the (orthogonal) projection of v on n̂ (see Figure 4.4). In Section 4.9

we shall find an alternative expression for the projection of one vec-
tor on another that does not explicitly involve the angle q.

b
c
b+c

θ b
a a

|b|cos(θ) a·c
a·b

(a) (b)
FIGURE 4.4:   (a) Projection of b and a. (b) The sum of projections onto vector a (which can
be assumed to be a unit vector).
120  •  Mathematical Physics

Example 4.4 If a is a vector of magnitude 3, and b a vector

of magnitude 1/2, and if the angle between a and b is 60°,
what is a · b?
1 1 1 3
a ⋅ b = a b cos q = 3 × × cos(60°) = 3 × × = .
2 2 2 4

Exercise 4.7 If a · b = 0 and |a| ≠ 0, |b| ≠ 0, show that a and b are

orthogonal (i.e. perpendicular to one another).
Exercise 4.8
(i) What is a · b if a and b are parallel? When is a · b = –|a||b|?
(ii) Show that i · i = j · j = k · k = 1 and that i · j = j · k = k · i = 0.

Exercise 4.9 If a ⋅ b =
1 2 and |a| = |b| = 1, what is the angle
between a and b?
Exercise 4.10 Show that the scalar product may be taken in
either order; i.e. prove equation (4.6).
We shall need the distributive law for the scalar product
a · (b + c) = a · b + a · c.  (4.7)
To establish equation (4.7) assume first that a is a unit vector.
Then equation (4.7) says that the orthogonal projection of b + c on
a is the sum of the projections of b and of c, a fact which is obvious
geometrically (see Figure 4.4(b)). For a general a (not a unit vector)
divide (4.7) through by |a| since then a /|a| is a unit vector. Note that
equation (4.7) also implies that
(b + c) · a = a · b + a · c.
Exercise 4.11 If a = 7i + 0.2j + 3k, find a · k.
Exercise 4.12 By expanding (a + b) · (a + b) derive the formula
|a + b|2 =|a|2 +|b|2 + 2|a||b| cos(q),
where q is the angle between a and b. Draw a figure to demon-
strate that this is the usual cosine rule of trigonometry. Also deduce
that |a + b| ≤ |a| + |b| (which is called the “triangle inequality”).
 Vectors  •  121

Here we summarize the most important algebraic properties of

the scalar product
a · a = |a|2
a·b=b·a (commutative)
a · (b + c) = a · b + a · c (distributive)

4.4.  THE VECTOR (CROSS) PRODUCT

Our second way of multiplying vectors is the vector or “cross” product.

Definition: Let q be the angle between the two vectors a and b. The
vector product (also called the “cross product”) of a and b, written
a × b (or sometimes a ∧ b) and read as “a cross b”, is defined to be
the vector of magnitude |a||b| sin(q) in a direction n̂ perpendicular
to both a and b such that a, b and n̂ form a right-handed set. Thus
rotation from a to b when viewed in the direction of n̂ is clockwise
(see Figure 4.5) and

a×b =a b sin (q ) nˆ . (4.8)

Notice the n̂ in the definition: the vector product gives a vector;
the scalar product gives a scalar!

Example 4.5 Show that if a is parallel (or anti-parallel) to b,

then a × b = 0 and conversely.
|a × b| = |a||b| sin(q) = 0 if and only if sin(q) = 0 or a = 0 or
b = 0. So, |a × b| = 0 if and only if q = 0 (or π), i.e. if and only
if a and b are parallel (or anti-parallel). If a = 0 or b = 0 it is
regarded as parallel to every vector.

Exercise 4.13 What is a × b if a and b are perpendicular?

Exercise 4.14 Show that i × j = k. What are j × k and k × i?

Exercise 4.15 Show that |a × b| is the area of the parallelogram

formed from a and b in Figure 4.5(b).
122  •  Mathematical Physics

area=|a×b|=|a||b|sin(θ)
|a||b|sin(θ)
a×b
b
^
n
|b|sin(θ)
b a
θ θ
a

(a) (b)
FIGURE 4.5:  (a) The vector product a × b. (b) The vector product a × b interpreted as a
vector with magnitude equal to the area of the parallelogram formed by a and b.

Exercise 4.16 Show that a × b = – (b × a).

Unlike the scalar product, the order in which we take the vec-
tor product does matter. We need the distributive law for vector
multiplication:
a × (b + c) = a × b + a × c.

We won’t prove this relation here (the proof is quite tedious) but
you can look it up if you are interested in the details.

Exercise 4.17 If v = 3i + 2j − k, show that v × k = 2i − 3j. What

is k × v?
Here we summarize the most important algebraic properties of
the vector product:
a × a = 0
a × b = − (b × a) (anti-commutative)
a × (b + c) = (a × b) + (a × c) (distributive)
Note that
a × (b × c) ≠ (a × b) × c (not associative)
in general.
 Vectors  •  123

4.5.  MULTIPLE PRODUCTS

From three vectors a, b, c we can form the scalar (i.e. number)

(a×b) · c, by taking the scalar product of c with the vector product of
a and b. This is called the scalar triple product of a, b, and c
a × b is the vector with magnitude equal to the area of the base
of the parallelepiped formed from a, b, and c (Exercise 4.15), in the
direction perpendicular to the plane of a and b (Figure 4.6). Therefore
(a × b) · c = (area of base) × (perpendicular height)
= volume V of the parallelepiped.
This assumes that a, b, c are a right-handed set, by which we mean
that a × b makes an acute angle with c; otherwise (a × b) · c = −V
(the sign is reversed).
Exercise 4.18 By considering the volume represented in each
case show that
(a × b) · c = (c × a) · b = (b × c) · a.  (4.9)
Deduce that (a × b) · c = a · (b × c).
The scalar triple product is usually written as [a, b, c]. From exer-
cise 4.18, [a, b, c] may be taken to mean (a × b) · c or a · (b × c) (or
any of the permutations of Exercise 4.18); i.e. we can interchange the
dot and cross, as well as permute the vectors cyclically. If we inter-
change the vectors in the vector product then this will change the sign
of the result. Thus, for example (b × a) · c = − (a × b) · c.

volume = (a×b)·c
a×b
c

a
FIGURE 4.6:   The scalar triple product and the volume of a parallelepiped.
124  •  Mathematical Physics

In terms of the notation [a, b, c] for the scalar triple product the
above properties can be written

[a, b, c] = [c, a, b] = [b, c, a] = − [a, c, b] = −[c, b, a] = − [b, a, c].

Exercise 4.19 Show that [a, b, c] =0 if and only if a, b, c lie in

a plane.

Exercise 4.20 Show that

b×c
A=
[ a, b, c ]
satisfies A · a = 1, A · b = 0, A · c = 0 (if [a, b, c] ≠ 0).

Exercise 4.21 Let a, b, c be three non-coplanar vectors and sup-

pose that v = v1a + v2b + v3c. By taking the scalar product of v
with the vector 1
= A b × c show that
[ a, b, c ]

v1 =
[ v, b, c ] ,
[ a, b, c ]
and find similar expressions for v2 and v3.

The vector A in Exercise 4.20 is called the reciprocal vector of a.

A set of three vectors a, b, c has a reciprocal set of vectors A, B, C as
follows:

b×c
A= ,
[ a, b, c ]
c×a
B= ,
[ a, b, c ]
a×b
C= .
[ a, b, c ] (4.10)
 Vectors  •  125

These exist if [a, b, c] ≠ 0, which means that a, b, c are not

coplanar (Exercise 4.19). Reciprocal vectors are particularly useful
in the study of crystal structure.
From three vectors a, b, c we can also form the vector (a × b) × c,
called the vector triple product. We shall later establish a formula,
equation (4.22), for this product in terms of the operations we have
already met, so no special symbol is needed.
Exercise 4.22 In the plane of which two of the vectors a, b, c
does (a × b) × c lie?

4.6.  EQUATION OF A LINE

In this section and the next we shall treat some geometrical

properties of lines and planes directly in terms of vectors. Only after
we have demonstrated how this can be done do we introduce coor-
dinates and components which enable us (if we wish) to do these
calculations algebraically.
Let r be the position vector of a general point P on the straight
line l in the direction of the unit vector t̂ passing through the point

A having position vector a (Figure 4.7), and let AP have length l.
Then

t
A
r
l a

O
FIGURE 4.7:   Equation of a line: r= a + l tˆ .

r= a + l tˆ . (4.11)
126  •  Mathematical Physics

As l varies between −∞ and +∞ the points P (reached by the

position vector r) describe the whole of the line l. Thus equation
(4.11) is the equation of the line l.

Example 4.6 Let r1 = a + lt, r2 = b + ms be the equations of two

non-parallel straight lines as l and m vary (with |s| = |t| = 1). Show
that the two lines intersect if and only if (a − b) · (s × t) = 0.
r1 Since the lines are not parallel they
r2
intersect if, and only if, (a − b), s and
t lie in a plane (see figure). Three
s vectors lie in a plane if their scalar triple
(a-b) product vanishes (exercise 4.19). That is,
t
b
a (a − b) · (s × t) = 0.
(Alternative solution) If the
lines intersect r1 and r2 have a common value at a point of
intersection. So, the two lines intersect where

a + lt = b + ms,  (4.12)

for some l and m, i.e. if

a − b = ms − lt.

Since s · (s × t) = 0 and t · (s × t) = 0, we take the scalar product

of both sides with s × t to get rid of the unknowns l and m.
Hence
(a − b) · (s × t) = 0.

If (a − b) · (s × t) = 0 then (a − b) lies in the plane of s and

t, so equation (4.12) holds and there is a common point of
intersection.

Exercise 4.23 If r1 = a + lt and r2 = b + ms are the equations of

two non-parallel straight lines, show that
(i) a − b is a vector joining the lines,
s× t
(ii) is a unit vector perpendicular to both lines, and hence
s× t
 Vectors  •  127

(iii) the shortest distance p between the two lines satisfies

|[a, s, t] − [b, s, t]| = p |s × t|.
(Hint: the shortest distance is the length of the projection of any vec-
tor joining the lines on the direction perpendicular to both lines.)

4.7.  EQUATION OF A PLANE

Let a be the position vector of a point A on the plane ∑ which

contains the unit vectors s and t. Let r be the position vector of the
point R obtained by moving from A a distance l in the direction s
and m in the direction t (Figure 4.8); then
r = a + ls + mt. (4.13)
As l and m vary, R describes the whole plane ∑. Thus equation
(4.13) is the equation of ∑.

Example 4.7 Find a vector normal to the plane ∑ given by

equation (4.13).
s × t is normal to s and to t by the properties of the vector
product. You can see that s × t is normal to any vector in the
plane by drawing s × t in the diagram. Formally, any vector in
the plane can be written as r − a = ls + mt for some l and m and
hence (r − a) · s × t = 0 (see Example 4.6). So, the vector s × t
is normal to both s and t and hence it must also be normal to ∑.

Example 4.8 Show that as r ranges over points in ∑ (equation 4.13)

r · (s × t) = constant.
Since the equation we are aiming at does not involve l or m we
try to eliminate these terms. This can always be done by taking
the scalar product of the equation with a vector perpendicular
to both, as in Example 4.6. Let

r = a + ls + mt,
128  •  Mathematical Physics

Σ t R
s
A

a
r

O
FIGURE 4.8:  Equation of a plane: r = a + ls + mt.

so,
r · (s × t) = a · (s × t) + ls · (s × t) + mt · (s × t)
= a · (s × t) + 0 + 0
= constant.

s× t
We let nˆ = ; then it follows from Example 4.8 that we can
s× t
write, for any r in the plane ∑,
r ⋅ nˆ =p, (4.14)

where p = a ⋅ nˆ = constant (by Example 4.8). Since equation (4.14)
holds if and only if r is in ∑, this is also (an alternative form of) the
equation of the plane ∑. In applications we use whichever of equa-
tion (4.13) and equation (4.14) is more convenient.

Example 4.9 Show that p= a ⋅ nˆ is the shortest distance to

the plane ∑ from the origin O.
The shortest distance occurs along
the direction from the origin
perpendicular to the plane – i.e.
r in the direction of the normal nˆ .
^
n This distance is the projection of
r (any vector to the plane) on nˆ .
The distance is r · nˆ . =p.
 Vectors  •  129

Exercise 4.24 Show that two distinct planes with unit normals
n̂1 and n̂ 2 intersect if and only if nˆ 1 × nˆ 2 ≠ 0 .

4.8.  COMPONENTS OF VECTORS

We turn now to the algebraic approach to vectors. We have seen

that any vector a (in three dimensions) may be represented by intro-
ducing unit basis vectors i, j, k:
a = a1i + a2j + a3k.  (4.15)
Thus, we can refer to a in terms of its components a1, a2, a3 with-
out writing out equation (4.15) in full. We say that a is represented
by the row vector (a1, a2, a3) in this basis. Often one writes this as
a = (a1, a2, a3),
with the meaning of equation (4.15). Once we are given a basis,
any vector in three dimensions can be represented by an (ordered)
triplet of numbers like this. (A vector in two dimensions obviously
requires only two numbers and two basis vectors.) The same vector
would be represented by a different triplet of numbers by someone
using a

a = (a1, a2, a3)

z

a3
k
a1 j
y
i O
a2
x
FIGURE 4.9:  Vector components.
130  •  Mathematical Physics

different basis. Thus, a triplet of numbers is not a vector and

a vector is not another name for triplet of numbers – a vector is a
triplet of numbers in a specified basis.(In more sophisticated math-
ematical treatments, a vector is the set of triplets by which it is rep-
resented in all bases.)
If axes are labeled such that i, j, k are unit vectors along the
x, y and z axes, respectively, from some origin O, then a1, a2, a3
are, respectively, the x-component of a, the y-component and the
z-component. So sometimes one writes (ax, ay, az ) for a.
Thus, a is represented by the point with coordinates (a1, a2, a3)
relative to O. When dealing with the position vector r of a point rela-
tive to O, one usually writes (x, y, z) for the components of r (instead
of (r1, r2, r3) or (rx, ry, rz)).

Example 4.10 What are the components of a + b in terms of

the components of a and b?
Write each vector as a linear combination of the basis vectors
(equation 4.15) and rearrange
a + b = (a1i + a2j + a3k) + (b1i + b2j + b3k)
= (a1 + b1)i + (a2 + b2 )j + (a3 +b3)k,
so the components of (a + b) are (a1 + b1, a2 + b2, a3 + b3).

Exercise 4.25 What are the components of la (l a scalar)?

Exercise 4.26 If i, j, k are three mutually perpendicular unit

vectors and a = (a1, a2, a3), show that a · i = a1, a · j = a2 and
a · k = a 3.

Exercise 4.27 With r = (x, y, z), a = (1, −2, 0) and t = (−1, 3, 0),
write the three coordinate equations for the line r = a + lt.
Deduce that this line is in the (x, y) plane and find m and c such
that it can be put into the form y = mx + c.
 Vectors  •  131

4.9. SCALAR (DOT) PRODUCT

(COMPONENT FORM)

In terms of components we have the important expression for

the scalar product of vectors a and b:
a · b = a1b1 + a2b2 + a3b3. (4.16)
This follows because
a · b = (a1i + a2j + a3k) · (b1i+ b2j + b3k) (4.17)
= a1b1 + a2b2 + a3b3.  (4.18)
Note that all the cross terms (e.g. a1i · b2j etc.) are zero since
i · j = 0 etc. Only the terms (a1i · b1i), (a2j · b2j), (a3k · b3k) remain, and
i · i =1 etc.
Thus we can work out a · b in either of two ways, whichever is
most convenient:
a · b = |a||b| cos(q)
= a1b1 + a2b2 + a3b3. (4.19)
Note particularly the following important application of equa-
tion (4.19) to finding the angle between two vectors given in com-
ponent form.

Example 4.11 Find (in radians as a number between 0 and π)

( )
the angle between the vector a = −1, 3,1 and b = (2, 0, −1).
Since we are given a and b in component form we can use the
components to work out the scalar product.

a1 b1 + a2 b2 + a3 b3 =−1 × 2 + 3 × 0 + 1 × −1 =−3.

Also

( a ⋅ a )1 2 = ( −1 ⋅ −1 + 3 ⋅ 3 + 1 ⋅ 1 ) =
12
a = 5,
b= ( b ⋅ b )1 2 = ( 2 ⋅ 2 + 0 ⋅ 0 + −1 ⋅ −1)1 2 = 5.
132  •  Mathematical Physics

But we also have expression (4.19) for the scalar product

involving the angle we are trying to find. Hence

a⋅b −3 3
cos (q ) = = = − ,
a b 5 5 5

Take the inverse of cos in the correct quadrant to find the angle q:
q = cos−1 (−3/5) = π − cos−1 (3/5) = 2.214 ... radians

Exercise 4.28 Find in radians the angle between the vectors

a = (−2, −2, 1) and b = (2, 0, −1).
Exercise 4.29 Find the projection of the vector a = (3, 5, 1) on
1
the direction nˆ = (1,1,1 ) .
3

Exercise 4.30 Find a vector N and a scalar P such that the equa-
tion x − y + 2z = 3 can be written in the form r · N = P (where
r = (x, y, z) as usual). Hence write the equation in the form r ⋅ nˆ =p
where n̂ is a unit vector. Deduce that the equation represents a
plane and obtain the shortest distance from the origin to the plane.
For readers familiar with matrix algebra, we can write the scalar
product of two vectors as a row vector multiplied by a column vector

  1 
b
⋅ b ( a1 , a2 , a3 )  b2=
a=  ↓ a1 b1 + a2 b2 + a3 b3 .  (4.20)
b 
 3

4.10. VECTOR (CROSS) PRODUCT

(COMPONENT FORM)

In terms of the components of two vectors a and b their vector

product is
a × b = (a2b3 – a3b2, a3b1 − a1b3, a1b2 – a2b1). (4.21)
 Vectors  •  133

Proof:
a × b = (a1i + a2j + a3k) × (b1i + b2j + b3k)
= a1i × (b2j + b3k) + a2j × (b1i + b3k) + a3k × (b1i + b2j),
expanding out using the distributive law and remembering that
i × i = j × j = k × k = 0. Then making use of i × j = −j × i = k etc,
we have that
= a1b2k − a1b3j – a2b1k + a2b3i + a3b1j – a3b2i
=( a2b3 – a3b2)i + (a3b1 – a1b3)j + (a1b2 – a2b1)k.
We can use either equation (4.21) or equation (4.8) to evaluate a
vector product – depending on the information available. It is important
to know both expressions and to be able to spot which is appropriate.
Exercise 4.31 If a = (−2, −2, 1) and b = (2, 0, −1), find a × b.
Show by explicit calculation that
(i) a × b is orthogonal to a and b,
(ii) (a × b) × a lies in the same plane as a and b. (Express it as
a linear combination of these two vectors.)
Exercise 4.32 Find the unit normal to the plane containing the
vectors s = (1, 1, 2) and t = (−1, −3, 0) and hence write the equa-
tion of the plane containing these vectors which passes through
the point (0, 1, −1) in the form ax + by + cz = d.
For readers familiar with matrix algebra, we can write the vector
product of two vectors as a determinant as follows:

i j k
a×b =a1 a2 a3
b1 b2 b3
= ( a2 b3 − a3 b2 ) i + ( a3 b1 − a1 b3 ) j + ( a1 b2 − a2 b1 ) k.

4.11.  SCALAR TRIPLE PRODUCT

Now we have expressions for both scalar and vector products in

component form it is possible to deduce the component form of the
triple products.
134  •  Mathematical Physics

Example 4.12 Find an expression for the scalar triple product

[a, b, c] = a·(b × c) in terms of the components of the three
vectors a, b and c.
We find b × c from the components of b and c using formula (4.21).
a · (b × c) = (a1, a2, a3) · (b2c3 − b3c2, b3c1 − b1c3, b1c2 − b2c1).
Now find the scalar product of the result with a using
equation (4.16):
a · (b × c) = a1b2c3 − a1b3c2 + a2b3c1 – a2b1c3 + a3b1c2 – a3b2c1.

Exercise 4.33 If a = (a, 0, 0), b = (0, b, 0) and c = (0, 0, c) com-

pute (a × b) · c and explain the result geometrically.
For readers familiar with matrix algebra, we can write the scalar
triple product as a determinant

a1 a2 a3
a ⋅ (b × c) =b1 b2 b3
c1 c2 c3
= a1 b2 c3 − a1 b3 c2 + a2 b3 c1 − a2 b1 c3 + a3 b1 c2 − a3 b2 c1 .

4.12.  VECTOR TRIPLE PRODUCT

The following important result should be memorized (but not

the proof):

a × (b × c) = b(a · c) − c(a · b).  (4.22)

This is sometimes known as the “bac minus cab” rule, the name
chosen to help recall the order of the vectors on the right side of the
equation.
Proof: Since no axes have been specified we may choose them to
simplify the problem. Choose the i-axis perpendicular to b and c, so
that b1 = c1 = 0. Then
 Vectors  •  135

a × (b × c) = (a1,a2, a3) × (b2c3 − b3c2, 0, 0)

= (0, a3(b2c3 − b3c2), − a2(b2b3 − b3c2)),
while
b(a · c) − c(a · b) = (0, b2, b3)( a2c2 + a3b3) − (0, c2, c3) ( a2b2 + a3b3)
= (0, a3c3b2 − a3b3c2, a2c2b3 – a2b2c3)

which is the same.

4.13.  VECTOR IDENTITIES

We can use equation (4.22) together with the properties of the

scalar triple product (section 4.5) to simplify various complicated
products of vectors.

Example 4.13 Show that if v = w × r then |v|2 = |w|2|r|2 – (w∙r)2.

From the definition of |v|:
|v|2 = v · v.
The expression involves a double vector product so we try to
put it in a form where we can use the expansion (4.22). Using
v=w×r
v · v = (w × r) · v.
Interchanging scalar and vector products (dot and cross)
= w · (r × v) = w · (r × (w × r)).
We now have a triple vector product so use equation (4.22)
= w · ((r · r)w – (r · w)r).
Expanding out the parenthetical by the distributive law
= (r · r)(w · w) – (r · w)(w · r).
Finally using the definition of “| |”
= |w|2|r|2 – (w· r)2,
as required.
136  •  Mathematical Physics

Exercise 4.34 Show that (a × b) · (a × c) = (b · c)(a · a) – (a · c)(b · a).

Exercise 4.35 Show that

(a × b) × (c × d) = [a, b, d]c – [a, b, c]d.

Find a similar expression for (a × b) × (c × d) as a linear combina-

tion of a and b, and hence deduce that

[a, b, c]d = [d, b, c]a + [a, d, c]b + [a, b, d]c.

(Compare Exercise 4.21)

Let a, b, c be the (unit)

vectors from the center O to
points on a unit sphere cen-
tered on O. A spherical trian-
gle is made from arcs of great
circles joining the points at the
A β ends of a, b and c. Let a, a, b
a
and g denote the sides of the
triangle; because the sphere is
γ
C of unit radius the length α is
c
the same as the angle between
O
b and c etc. Since a × b is nor-
α mal to the plane containing a
b B and b, a × c is normal to the
plane containing a and c, and
A is the angle between these
planes, we have

(a × b) · (a × c) = |a × b||a × c| cos(A) = sin(g) sin(b) cos(A). (4.23)

Exercise 4.36 Deduce from equation (4.23) and Exercise 4.34 that

cos(a) = cos(b) cos(g) + sin(b) sin(g) cos(A).

This is the cosine formula of spherical trigonometry.

 Vectors  •  137

Example 4.14 If a, b and l > 0 are given, solve for x the

equation
x × a + lx = b.  (4.24)
Take scalar and vector products with the constant vectors in the
problem and attempt to simplify the result. Here the possible
independent constant vectors we might want to use are a, b
and a × b. Taking the scalar product with a on both sides gives
0 + la · x = a · b.  (4.25)
Therefore
a · x = l–1 (a · b).  (4.26)
Now taking the vector product of a with both sides gives
another equation:
a × (x × a) + la × x = a × b.  (4.27)
And note that, by equation (4.24),
a × x = lx – b.  (4.28)
Now, expand the vector triple product equation (4.27) using
equation (4.22). Then substitute for a · x using equation (4.26),
and substitute for a × x using equation (4.28). The result should
be an equation that is linear in x and can be solved with a little
algebra. See Exercise 4.37. The solution is

1
=X 2
a +l 2
(l b + l −1
( a ⋅ b ) a + a × b ).

Example 4.15 An alternative method for solving

x × a + lx = b.

Express the unknown x as a linear combination of the

constant vectors of the problem with coefficients to be
determined. Let
138  •  Mathematical Physics

x = aa + b b + g a × b.
Substitute in (4.24)

(aa + b b + ga × b) × a + (aa + b b + g a × b) l = b.  (4.29)

Multiplying out and comparing coefficients of a, b and a × b

will yield three equations for the three unknowns a, b and g.

Exercise 4.37
(i) Complete the missing algebra steps in example 4.14.
(ii) Complete the missing algebra steps in example 4.15 to
obtain the same solution.
Exercise 4.38 Given vectors a and b such that a · b = 0, solve
the equation x × a = b subject to the condition x · a = k (a given
constant).

4.14.  EXAMPLES FROM PHYSICAL SCIENCE

Work is done by a force when the force acts on a body that

moves, and the force has a component along the line of motion.

Example 4.16 A particle acted on by a force F = 4i − 2j N is

given a displacement l = 2i + 8j + 2k m. How much work is
done by the force?
The work done is the component of force in the direction of
motion times the distance. Since we are given the component
form of the force and the displacement we use the component
form of the scalar product. Thus

Work done = F · l = Fxlx + Fyly + Fzlz

=4×2−2×8+0×2
= −8 J.
 Vectors  •  139

Exercise 4.39 How much work is done by a force F = 8i + 2k N

acting on a particle which is displaced through a distance
l = −2i + 4j + 10k m?
Exercise 4.40 A particle displaced through l = 0.5i + 0.2j m is
acted on by two forces, F1 = 3i + 3j N and F2 = 4i + 3k N.
(i) What is the work done by F1 and F2?
(ii) What is the net force F1 + F2?
The power supplied by a force is equal to the rate at which the
force does work. This can be written as P = F · v, where v is the
velocity of a particle acted on by a force F.
Exercise 4.41 What is the power input of a force 2i + 3j + 2k
N that acts on a particle moving with velocity 3i + 2j − 2k m s−1?
A constant electric field E passes through a plane surface of area
A with unit normal n̂ . The electric flux φ through A is defined to be
φ = E · n̂ A.
Exercise 4.42 A surface of area 2 cm2 has a unit normal
=nˆ (2 i + j) / 5. The surface is located in a region with a uni-
form electric field E = 4i + 2j + 2k V m−1. What is the electric
flux through the surface?
The torque t about the origin exerted by a force F acting through
a point with position vector r is defined as t = r × F. Clearly t is
orthogonal to both r and F.
Exercise 4.43
(i) A force F = 150i + 200j − 150k N acts on a body through a
point at r = 2i − 3j m relative to the origin O. Calculate the
torque exerted by this force about O.
(ii) Verify that the torque is orthogonal to the applied force.
The angular momentum L of a particle with linear momentum
p at a position r relative to some origin is L = r × p. As with torque,
angular momentum is defined relative to an origin. Torque is related
to angular momentum: the net torque acting on a body equals the rate
of change of angular momentum. If the angular momentum of a body
is constant then there is no net torque on the body (and conversely).
140  •  Mathematical Physics

Exercise 4.44 At some instant a particle has velocity v = −3i + 6j

m s−1 and is located at a point with position vector r = −4i + 2j m.
If the particle has a mass 2.5 kg, what is its angular momentum
about the origin?

4.15.  EXTENSION: INDEX NOTATION

Many vector equations involve the x, y and z components in

equivalent ways. For example, the scalar product of A and B is
AxBx + AyBy + AzBz where each term in the sum is identical in form
but with a different index. If we label the coordinates as x1, x2 and
x3 instead of x, y and z we can write the scalar product as
3

∑ Ai Bi .
A ⋅B=
i =1

We can even abbreviate this if we use the convention (called

the Einstein summation convention) that a repeated index is to be
summed over. With this understanding we simply drop the explicit
summation sign and write the expression for the scalar product as
A · B = AiBi.
A useful quantity is the Kronecker delta defined by

 1 if i= j
d ij = 
 0 if i ≠ j.

So, for example, d11 = 1 and d13 = 0. A second useful quantity is

the permutation symbol defined by

1 if i, j, k = 1,2,3 or 2,3,1or 3,1,2


Πijk =
 −1 if i, j, k = 2,1,3 or 1,3,2 or 3,2,1
 0 otherwise.


So, for example, Œ132 = −1 and Œ112 = 0. We can write the cross
product using Œijk and the summation convention in the compact form
(A × B)i = Œijk AjBk.
 Vectors  •  141

So, for example, we deduce that (A × B)1 = Œ123A2B3 + Œ132A3B2

= A2B3 − A3B2 because all the other potential terms in the sum over
j and k vanish.
A useful identity is ŒijkŒilm = djldkm − d jmdlk from which we
can deduce directly the formula for the triple vector product in
Section 4.12.

Revision Notes

After completing this chapter you should be able to

• Calculate linear combinations of vectors
• Calculate the magnitude and direction of a vector given
in component form
• Calculate the components of a vector given as a
magnitude and direction
• Calculate the scalar product of two vectors from
magnitudes and directions or from components
• Calculate the vector product of two vectors
• State the properties of the scalar triple product
• Give a formula for the vector triple product
• Derive vector identities for more complex products
• State the equation of a line in vector form
• State the equation of a plane in two different forms
• Find the unit normal to a plane and its distance from the
origin
• Compute the work done by a force, the flux of a constant
field through a plane surface, the torque exerted by
a force about an axis and the angular momentum of a
particle of a given velocity
142  •  Mathematical Physics

4.16. EXERCISES

1. If a = i − j and b = i + j find a unit vector perpendicular to

both b and a × b.
2. Find the angle between the vectors (1, −1, −4) and (2, 2, 1).
3. Find the angle between the vectors (3, 2, 4) and (1, 0, 2).
4. Write down the components of a unit vector which is
normal to both the vectors (1, −1, 3) and (2, 2, −2).
5. Find the components of a unit vector which is perpendicu-
lar to both (1, −1, −1) and (0, 2, 3).
6. Two sides of a triangle are formed by the vectors i + 2j + k
and 3i + 2j − k. What is the area of the triangle and the unit
normal to the triangle?
7. (i) If vectors a and c are given such that a ∙ c = 0, find all
solutions to the following equation for the vector x:
a × x = c.

(ii) Find an expression for the vector X in terms of A, B and

C given that X · A + C = 0 and X × A + B = 0.

8. If A and B are orthogonal show that the length of the vector

2
CA+A × B is A ( B + C2 .
9. Find the unit normal to the plane x + 2y − 2z = 15. What is
the distance of the plane from the origin?

4.17. PROBLEMS

1. S
 how that the equation of the straight line passing through
the points with position vectors c and d can be written as

r = c + lt,
 Vectors  •  143

where t = d − c, and l is a parameter. Show on a diagram

which parts of the line correspond to (i) l < 0, (ii) 0 ≤ l ≤ 1,
(iii) l > 1. Show that the equation of the plane that passes
through the point with position vector a, and that has nor-
mal n, can be written as

r · n = a · n.

Show that, if t · n ≠ 0, the straight line intersects the plane

in the point with position vector

(a − c) ⋅ n
c+{ }t.
t⋅n

What does this point become if a = d?

2. T
 he points A, B, C, D and E have rectangular Cartesian
coordinates (3, 1, 2), (−2, 7, 1), (4, −1, 3), (1, 1, −4), (2, 0, 1),
respectively. Find the equation of the straight line, L,
through D and E. Find an equation for the plane, P, through
A, B and C. Find where L intersects P.
3. F
 or a = (3, 1, 2), b = (−2, 7, 1), c = (4, −1, 3) verify the
identities
(i) a × (b × c) = (a · c)b − (a · b)c,
(ii) (a × b) · c = a · (b × c).
Now, letting a, b, c, d be any vectors and assuming that (i)
and (ii) are always true, show that

(a × b) · (c × d) = (a · c)(b · d) − (b · c)(a · d),

and also simplify(a × b) × (c × d).

4. T hree points A, B, C have coordinates (1, 1, 2), (2, −1, −2),
(3, 2,1):
(i) Calculate the perimeter of the triangle ABC.
 
(ii) Find the vectors AB and AC .
144  •  Mathematical Physics

 
(iii) Find a vector perpendicular to both AB and AC .
(iv) Find the area of the triangle ABC.
(v) Find the equation of the plane through ABC in the
form ax + by + cz = d.
(vi) Find the perpendicular distance of this plane from the
origin.
5. T
 he plane which has vector equation k · r = 1 (with k a
unit vector) intersects the sphere which has vector equa-
tion |r| = 2. If a and b are the position vectors of the two
diametrically opposite points on the circle of intersection
of the plane and the sphere show that (a − b) = 2a −2k.
Hence find the diameter of the circle.
6. T
 he angle between two non-parallel unit vectors a and b is
q, and two further vectors A and B are defined by

A = a − b cos(q) and B = b − a cos(q).

Show that

A · B = −sin2(q)a · b and A × B = sin2(q)a × b.

An arbitrary vector d is to be expressed as a linear

combination
d = la + mb + na × b;

show that

l sin2(q) = d · A, m sin2(q) = d · B, n sin2(q) = d · (a × b).

A point d lies on the intersection of the two planes r · a = a

and r · b = b, and satisfies the condition that d · (a × b) = 0.
Show that d sin2(q) = aA + bB.
7. (i) Write down the vector equation for the plane through
the point c containing the vectors a and b.
 Vectors  •  145

(ii) If a = (1, 1, 4), b = (2, 3, 1) and c = (1, 2, 1) write down

the equation of the plane in the form

lx + my + nz = p.

(iii) Write down the vector equation of the line through c

in the direction of the normal to the plane, and find the
point d where it cuts the (x, y) plane.
(iv) Find the points m1, m2, and m3 at which this plane cuts
the x-, y- and z-axes, respectively.
(v) What is the volume of the tetrahedron with vertices at
d, m1, m2, and m3?
8. Find the solution for x for the following vector equation:

a × x + bx + c = 0.

Show also that x and c are equal in length if a and c are

orthogonal vectors and a2 + b2 = 1.
9. F
 ind the angle between the vector a = (1, 0, 2) and the
plane 3x − y + z = 1.
A line is drawn parallel to a through the origin. Find the co-
ordinates of the point where this line cuts the above plane.
10. Find the equation of the line l1 through the point (5, 2, 0) in
the direction of (2, 1, 2) and of the line l2 through (−1, 1, −1)
in the direction of (4, 0, −1). Find the point of intersection
A of l1 and l2. Hence determine the equation of the plane
through A containing l1 and l2. What is the distance of this
plane from the origin?
11. Let a = (0, 0, 3 /2) and t = (1, 1, 1) 3. Find the points on
the line, l1, r = a + lt that lie a unit distance from the origin.
Find the equation of the line, l2, joining the origin to one of
these points. Find the equation of the plane containing the
lines l1 and l2.
 CHAPTER

5
MATRICES

You should be familiar with the solution of a pair of simultaneous

linear equations in two unknowns, by the elimination of one vari-
able. You might think that more equations in more unknowns could
be solved by a similar process of successive eliminations. If all you
ever want to do is to solve specific equations then this is correct
(but not very efficient; imaging trying this for 100 equations in 100
unknowns typical of large engineering projects).
However, if you want to understand, for example, the circum-
stances under which a system of equations has a solution (since not
all of them do), and to apply important tools for analyzing them,
you will need to be familiar with the matrix representation of such
systems of equations. The theory leads to the use of matrices as rep-
resentations of transformations (for example a rotation of coordinate
axes) and the construction of the “eigenvectors” and “eigenvalues”
that turn out to play a crucial role in almost all branches of physical
science (wherever small oscillations are important, e.g. in quantum
mechanics and in numerical data analysis). These topics (matrices,
eigenvectors etc.) are sometimes gathered under the umbrella term
linear algebra.
Let’s begin with what matrices are and some motivation for their
introduction. Suppose we want to solve the pair of simultaneous
equations

3x + 2y = 7
6x − y = 4
148  •  Mathematical Physics

We start by (for example) multiplying the second equation by 2:

3x + 2y = 7
12x − 2y = 8

Adding the equations and multiplying the first by 5 gives

15x + 10y = 35
15x + 0y = 15

and subtraction gives

0x + 10y = 20
15x + 0y = 15

from which we can read off x = 1 and y = 2. Notice, however, that

the x and y are acting merely as placeholders: if we keep every-
thing aligned we do not have to write out the x and y in every
line. For example, the first two steps of the calculation could be
written:

 3 2 7 
 12 −1  =  4 
   

and
 3 2 7 
 12 −2  =  8 
   

And so on, provided we have a rule that enables us to put back the x
and y. This rule must be such that

 3 2  x  3 x + 27
=  . (5.1)
 6 −1  y  6 x − y 

With this rule in place we can manipulate the set of coefficients

as an entity, as long as the rules are compatible with the rules for
 Matrices  •  149

manipulating simultaneous equations. The set of coefficients is then

called a matrix. Note that a matrix is not just an array of numbers;
it is the array plus the rules for manipulation. Let us now formalize
these rules.

5.1.  MATRIX REPRESENTATION

Let a = (a1, a2, a3) and x = (x1, x2, x3) be vectors. Consider the
scalar product a · x = a1x1 + a2x2 + a3x3 from Section 4.9. We can
write this product as

  1 
x
( a1 , a2 , a3 )  x2  = a ⋅ x = a1 x1 + a2 x2 + a3 x3 (5.2)
x 
 3

where corresponding terms reading across the row and down the
column are to be multiplied together (a1 × x1, a2 × x2, a3 × x3) and
the results added. (The arrows are there for illustration only.) This
can be extended thus:


 a1 a2 a3   x1   a ⋅ x   a1 x1 + a2 x2 + a3 x3 
       
 b1 b2 b3   x2  =  b ⋅ x  =  b1 x1 + b2 x2 + b3 x3  (5.3)
       
 c1 c2 c3   x3   x ⋅ x   c1 x1 + c2 x2 + c3 x3 

where the result of going across each row in turn on the left is placed
at the corresponding level on the right. The first array on the left
is called a matrix. More specifically it is a 3 × 3 matrix (there are 3
rows running across the page, 3 columns running down the page).
A vector is a special case: it can be written as a single column, in
which case it is known as a “column vector” or 3 × 1 matrix, or as a
single row, hence known as a “row vector” or 1 × 3 matrix. Both are
used in equation (5.2).
150  •  Mathematical Physics

We can use matrices to represent systems of linear equations, as

in the following example.

Example 5.1 Represent the system of equations

(i) x + y − z = −3
( ii ) 2x + y − z = 0 (5.4)
( iii ) 4 x − 2 y − 3 z = 4 
in matrix form.
Write out the coefficients as a matrix, and the set of unknowns
and the constants on the right as column vectors:

1 1 −1   x   −3 
     (5.5)
2 1 −1   y  =
 0
 4 −2 −3   z   4 
      

Using the rule shown in equation (5.3) on the left side of

equation (5.5) gives us (5.4) as required.

However, if we want this representation to be useful, we must

establish rules for manipulating the matrix of coefficients that cor-
respond to manipulation of the equations. In particular, writing the
system (5.5) in the form

Ax = h,

where A stands for the matrix on the left side and h for the column
vector on the right, we should like to “divide through by A” and
obtain the matrix solution

x = A−1h.  (5.6)

In order to attach a meaning to the matrix A−1 we must develop

the rules of algebra for matrices.
 Matrices  •  151

1. Multiplication by a scalar. We can multiply the system (5.4) by

any non-zero number to obtain an equivalent system: e.g. multiplying
(5.4) by 2 we get

2x + 2y − 2z = −6
4x + 2y − 2z = 0
8x − 4y − 6z = 8.

In matrix form this can be written as

 2 2 −2  x  −6 
     
 4 2 −2  ⋅  y  =  0 ⋅
 8 −4 −6  z   8
     
Thus we have shown that

 1 1 −1   2 2 −2 
   
2×2 1 − =1  4 2 −2  ,
 4 −2 −3   8 −4 −6 
   

i.e. to multiply a matrix by a scalar we multiply all the entries by the

scalar value.
2. Addition (or subtraction). Equation (5.5) can be written
(in infinitely many ways) as a sum (or difference) of subsystems: e.g.

0x + 0y − 2z + x + y + z = −3
0x + y + 0z + 2x + 0y − z = 0
x − 5y − 6z + 3x + 3y + 3z = 4.

The choice of coefficients is quite arbitrary provided they sum to

those in equation (5.5). In matrix notation this is

 0 0 −2   1 1 1    x   −3 
       
 0 1 0  +  2 0 −1    y  =
 0 .
 1 −5 −6   3 3 3    z   4 
     
 
152  •  Mathematical Physics

Thus to add two matrices, we add the corresponding entries.

In order to add or subtract two matrices, they must have the same
dimensions (in the above example we add two 3 × 3 matrices to get
another 3 × 3 matrix).

Exercise 5.1 Let

 3 0 −1   8 1 1  1  x
       
A=  0 1 5  , B=  5 2 −3=, h =
 2  , and x  y .
 3 −2 3   0 1 −1   1  z
       

(i) Write out the system of equations Ax = h explicitly.

(ii) What is the matrix 3A − B?
Notation: It is sometimes more convenient to use a more systematic
notation for the elements of A. If we write

 a11 a12  ← row1

A= 
 a21 a22  ← row2
↑ ↑
col1 col 2

(the first index labels the row, the second the column) then aij is the
element of A in the ith row and jth column. An M × N matrix has M
rows and N columns. A square matrix has the same number of rows
as columns, i.e. has dimensions N × N.
Using this notation we can write down the rules above. For mul-
tiplication by a scalar B = λA we multiply all the elements of the
matrix by the scalar:

bij= λaij,

where subscripts i, j run over the dimensions of the matrix. And

for the sum of two matrices (of the same dimensions) C = A + B,
we compute the elements of C as the sum of the corresponding
elements of A and B:
cij = aij + bij.
 Matrices  •  153

5.2.  SOLUTION OF SYSTEMS OF EQUATIONS

Systems of linear equations can be solved by operations equiva-

lent to those we have just introduced. To solve equations (5.4) we
can begin by subtracting 2×(i) from (ii), keeping the other equations
untouched.

(x + y − z ) + ( +0 x + 0 y + 0z ) =−3
( 2x + y − z ) + ( −2 x − 2 y + 2z ) = 0 − ( −6 ) = 6.
( 4 x − 2 y − 3z ) + ( +0 x + 0 y + 0z ) =4

In matrix terms this can be written as

 1 1 −1   0 0 0    x   −3 
       
 2 1 −1  +  −2 −2 2    y  =  6 ,
    0 0    z   4 
  
 4 −2 −3   0

or

1 1 −1   x   −3 
     
 0 −1 1   y  = 6 . (5.7)
 4 −2 −3   z   4
      

Equation (5.7) is an equally valid representation of the original

system (1.3). We can continue this procedure, but to simplify mat-
ters we simply carry out the operations directly on the matrix entries.
Thus, the operation that produces (5.7) from (5.5) amounts to sub-
tracting twice the first row from the second in the matrix on the left
and in the column vector on the right:

1 1 −1   x   −3   1 1 −1   x   −3 
    R 2 → R2 − ( 2 × R1) 
    
=
2 1 −1   y   0 =  0 −1 1   y   6 .
 4 −2 −3   z   4  4 −2 −3   z   4
         
154  •  Mathematical Physics

(The notation above the arrow “R2 → R2 − (2 × R1)” means

“replace the second row with the original second row minus twice
the first row.”)
A possible further series of equivalent representations, obtained
by the indicated operations on rows, is

1 1 −1   x   −3   1 1 −1   x   −3 
    R 3 → R3 − ( 4 × R1) 
    
=0 −1 1   y   6 =0 −1 1   y   6 
 4 −2 −3   z   4  0 −6 −1   z   16 
         
 1 1 −1  x   −3 
R 3 → R 3 − ( 6 × R 2 )
  0 −1 1  y  =  6 
    
 0 0 −5  z   20 
    

Now, the equivalent system is easier to solve. Writing it out

we have

x + y − z = −3
−y + z = 6 (5.8)
−5z = −20

from which z = 4, y = −2, x = 3.

Note that the final form of the matrix representation involves
only zero entries below the diagonal. Such a matrix is said to be
upper-triangular (meaning the non-zero entries are all in the top
right triangle, and on the leading diagonal). This enables us to
solve (5.8) easily for z, then y, then x working backwards up the
system. In fact, we say that the system of equations (5.8) can be
solved by back-substitution. The aim of the series of equivalent
representations was to achieve this triangular form. Note: reduc-
tion to a lower triangular form by a similar method would be
equally effective.
Rather than use back-substitution we can use row operations to
reduce the matrix still further. Continuing, we have
 Matrices  •  155

 1 1 −1   x   −3 
    
 0 −1 1   y  =
 6 
 0 0 −5   z   −20 
    

R2 → ( 5 × R2 ) + R3
 1 1 −1   x   −3 
R → −
  0 −5 0   y  =
3 R 3 / 5  
     10 
0 0 1   z   4 


R1 → R1 + ( R2 / 5 ) + R3
 1 0 0  x  3
R 2 → − R 2 / 5
  0 1 0   y  =  
    −2 
 0 0 1  z   4
    

from which the solution can be read off directly.

It is clearly unnecessary to write out the (x, y, z) column and
the equals sign at every step, and the right hand side can be car-
ried along as part of the array, as we will see in the following
example.

Example 5.2 Solve the system

x + 2y + 3z = 10 
2x + 5y + 8z = 24 (5.9)
3x + 8y + 14z = 39

by reduction to triangular form and back-substitution.

We work through a series of equivalent representations to
obtain a triangular form:
R2 → R2 − ( 2 × R1 )
 1 2 3 10   1 2 3 10 
 R → − ( × )
  0 1 2 4  .
3 R 3 3 R1
 2 5 8 24   
 3 8 14 39  0 2 5 9 
   
156  •  Mathematical Physics

Start from the left hand column using either of the rows above.

 1 2 3 10   1 2 3 10 
 R 3 → R 3 − ( 2 × R 2 )  
 0 1 2 4    0 1 2 4  .
0 2 5 9  0 0 1 1 
   
From the upper triangular form write out the equivalent
system and solve by back-substitution.
x + 2y + 3z = 10
y + 2z = 4
z=1
from which we see z = 1, and so y = 4 − 2 = 2 and x = 10 − (2 × 2)
− 3 × 1 = 3. Or,
z = 1, y = 2, x = 3.

This method is a useful way of organizing the working of the

solution of linear systems by hand. But, like the simultaneous equa-
tions method, it is not a practical algorithm for really large systems
or for implementation as a computer program. (For larger systems
special methods have been developed.)

Exercise 5.2 Solve the system

x + y + 3z = 5
x + 6y + 10z = 17
−x + 4y − 5z = 8.

5.3. PRODUCTS

In Section 5.1, we learned how to multiply a matrix and a col-

umn vector. We multiply two matrices by treating the second one
as a set of column vectors. For example, multiplying the same 3 × 3
matrix by different column vectors we get
 Matrices  •  157

 3 0 −1   8   24 
    
0 1 5   5  =  5 ,
 1 −2 3   0   −2 
    
 3 0 −1   1   2 
    
0 1 5   2  =  7 ,
 1 −2 3   1   0 
    
 3 0 −1   1   4 
    
0 1 5   −3  = −8  .
 1 −2 3   −1   4 
    

So, putting these together

 3 0 −1   8 1 1   24 2 4
     
0 1 5  5 2 =−3   5 7 −8  .
 1 −2 3   0 1 −1   −2 0 4 
    

Exercise 5.3 Let

 3 0 −1   8 1 1
   
=A 0 1 5 =
and B  5 2 −3  .
 1 −2 3   0 1 −1 
   

Find
(i) BA (the product of B and A in the indicated order)
(ii) AB
(iii) Does AB = BA in this case?

If AB ≠ BA for matrices A and B, we say that A and B do not

commute (with respect to multiplication). In general, matrix mul-
tiplication is not commutative: the result depends on the order of
multiplication. Therefore, if we are multiplying a matrix equation
158  •  Mathematical Physics

B = C by a matrix A we must specify whether this is on the left or

the right, giving AB = AC or BA = CA. (Both are true but they are
different equations if A and B do not commute.)

Exercise 5.4 If A and B are as in Exercise 5.3 and

 3 0 1
 
C  1 −1 1  ,
=
 0 2 1
 
find
(i) C(AB),
(ii) (CA)B.
(iii) Verify that C(AB) = (CA)B in this case.

The equality C(AB) = (CA)B is called the associative law and is

true for all matrices A, B, C. Therefore we can write CAB for this
product without ambiguity. Multiplication by a scalar can be done at
any stage of taking a matrix product. For example,

λ (ABC) = ( λ A)BC = A( λ (BC)) etc.

Exercise 5.5 The following 2×2 matrices are important in quan-

tum mechanics ( i= −1 as usual):

 0 1  0 −i   1 0
=s 1 =  , s2  =  , s3  
 1 0  i 0  0 −1 
Find a relation between s1s2 – s2s1 and s3.

We can also write the product in terms of matrix elements.

If A is an M × K matrix and B is a K × N matrix, we can compute
the elements of the product C = AB, which is an M × N matrix, as
follows
 Matrices  •  159

K
c ij = ∑ aik bkj
k =1

where i = 1, 2, . . . , M and j = 1, 2, . . . , N . For example for

the product of a 2 × 3 matrix with a 3 × 2 matrix we have

b b12 
 a11 a12 a13   11   c 11 c 12 
a  b21 b22  = 
 21 a22 
a23     c 21 c 22 
 b31 b32 

where c11 = a11b11 + a12b21 + a13b31 (formed from the 1st row of A and
the 1st column of B as marked).

5.4.  THE IDENTITY MATRIX

Definition: the identity matrix. The matrix

 1 0 0
 
I = 0 1 0
 0 0 1
 

is called the (3×3) identity matrix. Sometimes it is called the

unit matrix. The N × N identity matrix (always square) has ones
on the leading diagonal (top left to bottom right) and zeros
elsewhere.

 1 0  0
0 1  
IN = 
  0
 
 0  0 1
160  •  Mathematical Physics

Exercise 5.6
(i) For a general 3×3 matrix

 a11 a12 a13 

 
A =  a21 a22 a23  ,
a a32 a33 
 31

show that IA = AI = A.
(ii) What is SA where S is the matrix

7 0 0
 
S = 0 7 0  ?
0 0 7
 

5.5. SYMMETRIC AND ANTISYMMETRIC MATRICES

Definition: transpose of a matrix. If A is the matrix

 a11 a12 a13 

 
A =  a21 a22 a23  ,
a a32 a33 
 31

the transpose of A is the matrix AT:

 a11 a21 a31 

T  
A =  a12 a22 a32  ,
a a23 a33 
 13
obtained by interchanging rows and columns. We call A symmetric if
AT = A and antisymmetric if AT = −A.
 Matrices  •  161

Exercise 5.7 Let A and B be the matrices defined in Exercise 5.3.

Find AT and BT and verify that

ATBT = (BA) T.

Note the order in which the matrices are multiplied. This result is
true for general matrices A and B.

Exercise 5.8
(i) The matrix

⋅ a b 
 
⋅ ⋅ c 
 ⋅ ⋅ ⋅
 
is antisymmetric. Fill in the missing elements.
(ii) For any matrices A and B, what are (AT)T and (A + B)T ?
(iii) Verify that if M is an arbitrary matrix, the matrices
1 1
2
( M + MT ) and
2
( M − MT ) are respectively symmetric
and antisymmetric. Deduce that every matrix is the sum of a
symmetric and an antisymmetric matrix.

Part (i) of Exercise 5.8 gives another way of looking at an anti-

symmetric matrix: the terms below the diagonal are the negatives of
the corresponding terms above the diagonal and the diagonal terms
are zero.

5.6. INVERSES

Definition: Given a matrix A, if a matrix M exists with AM = MA = I

then M is the inverse of A. We write M = A−1. In words: the prod-
uct of a matrix and its inverse is the identity matrix (A−1A = I). (Of
course, if M is the inverse of A then A is the inverse of M.)
162  •  Mathematical Physics

When we work with scalars we consider ab = c and a = c/b as

equivalent (b ≠ 0). But for matrices “division” does not have such
a clear meaning. For matrices we instead work with AB = C and
A = CB−1.

Exercise 5.9 Are the following statements (a) true or (b) false?

1 
2 0 0
  2 0 0
(i)  0 − 1 0  is the inverse of   ,
 3   0 −3 0 
0  0 0 1
0 1  
 
 

 5 0 −5   1 −1 2 
(ii) 1     ,
 −6 10 1  is the inverse of  1 2 1
25    −4 −1 2 
 7 5 3  

 1 −1 2   5 0 −5 
   
(iii) 25  1 2 1  is the inverse of 1 .
 −6 10
 −4 −1 2   7 5 3
   

In fact if AM = I it can be shown that automatically MA = I, so

there is no need to check both conditions for an inverse. We now
look at a way to calculate the inverse.
Let A be the general 3 × 3 matrix

 a11 a12 a13 

 
A =  a21 a22 a23 
a a32 a33 
 31
and let its inverse be
 x1 x2 x3 
 
A −1 =  y1 y2 y3  .
z z2 z3 
 1
 Matrices  •  163

To find A−1 explicitly in terms of the components of A we have

to solve

 a11 a12 a13   x1 x2 x3   1 0 0


−1     
AA =  a21 a22 a23   y1 y2 y3  = I =  0 1 0 
a a32 a33  z z2 z3   0 0 1
 31  1  
i.e. we must solve the equations

 a11 a12 a13   x1   1 

    
 a21 a22 a23   y1  =  0  ,
a a32 a33   z1   0 
 31
 a11 a12 a13   x2   0 
    
 a21 a22 a23   y2  =  1 
a a32 a33   z2   0 
 31
 a11 a12 a13   x3   0 
    
 a21 a22 a23   y3  =  0 
a a32 a33   z3   1 
 31
for (x1, y1, z1), (x2, y2, z2) and (x3, y3, z3), respectively. We can do this
using the method of Section 5.2 three times, or all in one go as in the
following example.

 1 2 3
Example 5.3 Find the inverse of A=  2 5 8  .
 3 8 14 
 
We have to solve

 1 2 3   x1   1 
    
 2 5 8   y1  =  0  etc.
 3 8 14   z   0 
  1  
To do this we construct a sequence of equivalent representations
that reduce the matrix to the identity matrix.
164  •  Mathematical Physics

R2 − ( 2 × R1 )
 1 2 3 1 0 0 1 2 3 1 0 0
  R3 − ( 3 × R1 )  
 2 5 8 0 1 0    0 1 2 −2 1 0 
 3 8 14 0 0 1  0 2 5 −3 0 1 
  
1 2 3 1 0 0
R 3 − ( 2 × R2) 
 
0 1 2 −2 1 0
0 0 1 1 −2 1 


R1 − ( 2 × R2 )
 1 0 −1 5 −2 0
R 2 − ( 2 × R 3 ) 
 0 1 0 −4 
 5 −2  ,
0 0 1 1 −2 1 

and so finally we have

 1 0 0 6 −4 1  
R 1 + R3  0 1 0 −4
 
5 −2  . (5.10)

 0 0 1 1 −2 1 

The equivalent system is now

 x1   6   x2   −4   x3   1 
           
I  y1  =
 −4  , I  y2  =
 5  , I  y3  =
 −2  ,
 z   1  z   −2   z   1
 1    2    3  

from which we read off (x1, y1, z1) etc. It is clear that the matrix
to the right of the vertical bar in the final representation (5.10)
is the required inverse, i.e.

 6 −4 1
−1  
A =
 −4 5 −2  .
 1 −2 1 

 Matrices  •  165

The rule for finding A−1 is therefore: convert the matrix (A|I)
by row operations of the above type to (I|B); if this can be done then
A−1 exists and is equal to B.

 1 1 3
 
Exercise 5.10 Find the inverse B of A =  1 6 10  by the
 −2 8 10 
 
method indicated. Check that AB = I.

In this section, we have found a method by which we can con-

struct the inverse of any given matrix for which an inverse exists. In
the next sections we shall find a general formula for the inverse of a
square matrix of any size. (Note that matrix inverses are only defined
for square matrices.) We begin with 2 × 2 matrices.

5.7.  THE INVERSE OF A 2 × 2 MATRIX

We solve the general 2 × 2 system

ax + by = h  (5.11)
cx + dy = k  (5.12)

which can be written in matrix form as

 a b   x   h  
    =  . (5.13)
c d  y   k 

Example 5.4 Show that if ad − bc ≠ 0 then the solution of

equations (5.11) and (5.12) is

dh − bk ak − ch
=x = , y . 
(5.14)
ad − bc ad − bc
166  •  Mathematical Physics

To eliminate y between the pair of simultaneous equations

(5.11) and (5.12), we multiply equation (5.11) by d, multiply
equation (5.12) by b, and then subtract.

dax + dby = dh,

bcx + bdy = bk.

So
dax − bcx = dh − bk,

which leads to an equation for x

dh − bk
x= ,
ad − bc
as required. Similarly, to eliminate x multiply equation (5.11)
by c, multiply equation (5.12) by a, and then subtract.

Alternatively, we can start from the matrix form and manipulate

the array of coefficients as in Sections 5.2 or 5.6.
We can write equations (5.14) in a matrix form as

1 h 1 h
x= ( d , −b )   , y = ( −c , a )  
ad − bc k  ad − bc k 
or, combining these, as

 x 1  d −b   h 
 y =  a   k 
(5.15)
  ad − bc  − c 

Let
 a b  x h
A =  , x =   and h =   .
 c d  y  k

Then equation (5.13) can be written as Ax = h and therefore

equation (5.15) is x = A−1h. Thus, if we have a matrix
 Matrices  •  167

 a b 1  d −b 
A =  , then we can write A −1 =  .
 c d ad − bc  −c a
This is the general formula for the inverse of a 2 × 2 matrix.
Note that from this general form we can see that not all 2 × 2
matrices have an inverse. Namely, if ad − bc = 0, then A−1 will not
exist. (This is analogous to the fact that not all numbers have inverses;
namely 0 does not have an inverse because there is no number x
such that 0x = 1.)
To extend this to larger matrices we need to introduce the idea
of a determinant, which we shall do in the next section.

5.8. DETERMINANTS

Definition: The determinant, written det A or |A|, of a 2 × 2 matrix

A is defined by

 a b
det  =
 ad − bc.
 c d
Or, using more systematic notation,

 a11 a12 
A=  → det A = a11a22 − a12 a21 .  (5.16)
 a21 a22 

Exercise 5.11 Find the values of

 1 3
(i) det  
 −1 2 
 3 1
(ii) det
 2 1

 −3 −2 
(iii) det  
 −4 −3 
168  •  Mathematical Physics

Definition: The determinant, det A or |A|, of a 3 × 3 matrix A is

defined by

 a11 a12 a13 

    a22 a23   a21 a23 
det =
 a21 a22 a23  a11 det  a  − a12 det 
a   32 a33   a31 a33 
 31 a32 a33 
 a21 a22 
+ a13 det   (5.17)
 a31 a32 

Note the alternation of signs: in general the sign of the coeffi-

cient of aij is (−1)i+j. Note also how the 2 × 2 determinants in (5.17)
are formed: for example, we take a11, cross out the row and column
that it is in, and this leaves the first 2 × 2 determinant on the right
of equation (5.17).

 a11 a12 a13 

   a22 a23 
 a21 a22 a23  → a11 det 
a  a32 a33 
 31 a32 a33 

Then we do likewise with a12 and a13. The determinants of the

(2 × 2) submatrices, obtained by taking out row i and column j, are
called the minors, Mij, of A.

 1 −1 0 
 
5.5 If A  2
Example= 4 −3  , find det A.
 −2 1 1 

Expanding by the first row as in the definition (5.17) gives

 4 −3   2 −3   2 4
det A =1 × det   − ( −1 ) × det   + 0 × det  
 1 1  −2 1  −2 1 
= (1 × 7 ) + (1 × −4 ) + 0= 3.
 Matrices  •  169

Exercise 5.12
 0 1 2
 
(i) Let the matrix B =  −7 3 −4  . Verify that det B = −17.
 1 2 3 

(ii) For A as defined in Example 5.5, and B defined here, find
AB and verify that
det AB = det A · det B.

Exercise 5.13 By induction, generalize the definition of a deter-

minant to N × N matrices.

Another way to remember how to construct the determinant of

a 3×3 matrix is to write the first two columns on the right, and then
draw diagonal lines as follows:

a11 a12 a13 a11 a12

a21 a22 a23 a21 a22
a31 a32 a33 a31 a32
– – – + + +
This is called Sarrus’s rule. We sum the products along the
southeast (top left to bottom right) lines, and subtract the products
along the southwest (top right to bottom left) lines. It is a useful way
to remember the terms of the determinant of a 3 × 3 matrix, but it
does not extend to larger matrices.
det A
= a11a22a33 − a11a23a32 + a12a23a31 − a12a21a33+ a13a21a32 − a13a22a31

5.9.  PROPERTIES OF DETERMINANTS

The following properties are valid for determinants of N × N

matrices. (Note that determinants are only defined for square matri-
ces.) We consider proofs only in the 2 × 2 case, where they involve
only elementary algebra. In the following, A is a general N × N matrix.
170  •  Mathematical Physics

Property 1: If B is obtained from A by multiplying any

one row or column by λ, then det B = λ det A. It follows that
det(λA) = λ N det A.

Example 5.6 Prove that det(λA) = λ N det A for N = 2.

 l a11 l a12 
l A = .
 l a21 l a22 

So,

det(λA) = λa11λa22 − λa12λa21

= λ2(a11a22 − a12a21)
= λ2 det A.

Property 2: If à is obtained from A by interchanging a pair of rows

or a pair of columns, then det à = − det A.

Exercise 5.14 Show that

 a11 a12   a12 a11   a21 a22 

det   =
−det   =
−det  .
 a21 a22   a22 a21   a11 a12 

In our definition of the determinant we singled out the first

row of the matrix. Property 2 shows that there is nothing special
about the first row: by Property 2 we can use any row to frame a
definition and hence to calculate a determinant. This can be used
to simplify the calculation, e.g. by selecting a row containing one
or more zeros.

 2 4 −3 
 
Example 5.7 Find det  1 −1 0  .
2 1 1 

First, interchange the first and second rows; then expand
by the top row. The zero means there are only two minors
 Matrices  •  171

(2×2 determinants) to evaluate. We have already evaluated this

determinant in Example 5.5.

 2 4 −3   1 −1 0 
   
det  1 −1 0  =−det  2 4 −3  =−3.
2 1 1  
 2 1 1

Alternative solution Expand by the second row. This is

exactly the same calculation, but arrived at in a different way.

 2 4 −3 
   4 −3 
det  1 −1 0  =( −1)2 +1 ⋅ (1) ⋅ det 
2   1 1 
 1 1
 2 −3 
+ ( −1 ) ⋅ ( −1 ) ⋅ det 
2+2

2 1 
=−7 + 4 =−3.

Note the signs here: recall from the definition that the
coefficient of aij has a factor (−1)i+j.

Property 3: If A has two rows or columns proportional then det

A = 0. (In particular det A = 0 if two rows or columns are the same.)

 a11 l a11 
Exercise 5.15 Show that det  =0 and that
 a12 l a12 
 a11 a12 
det  =0.
 l a11 l a12 

Property 4: A square matrix A and its transpose AT have the same

determinant: det AT = det A.

Exercise 5.16 Show that

 a11 a12   a11 a21 

det   = det  .
 a21 a22   a12 a22 
172  •  Mathematical Physics

Since transposition interchanges rows and columns, this implies

that any statement of a property of the determinant under a row
operation must also be valid under the equivalent column operation.
Thus we do not have to prove row and column properties separately;
e.g. the second result in Exercise 5.15 follows from the first as a
consequence of Exercise 5.16. We can also expand a determinant by
columns instead of rows.
Property 5: Adding a multiple of one row (or column) to another
leaves the determinant unaltered.

Exercise 5.17 Show that

 a11 + l a21 a12 + l a22   a11 a12 

det   = det  .
 a21 a22   a21 a22 

Property 6: If the rows (or columns) of A are linearly dependent

then det A = 0–i.e. if a, b, c, . . . are the rows of A, regarded as row vec-
tors, and l a + m b + n c + . . . = 0 (with at least one of l , m , n ,. . . ≠ 0 ,
then det A = 0. (Recall the discussion of linear dependence for vec-
tors in Section 4.2.)

Example 5.8 If two vectors are linearly dependent then

 a11   a12   0 
l +m =  ,
 a21   a22   0 
and we have that

λa11 + μa21 = 0 and λa12 + μa22 = 0 (5.18)

 a11 a12 
with λ, μ not both 0. Show that det  =0.
 a21 a22 
Assume λ ≠ 0. Add (μ/λ) × row 2 to row 1. By Property 5 this
does not change the determinant:

 m m 
a + a a12 + a22 
det A =det  11 l 21 l .
 
 a21 a22 
 Matrices  •  173

m m
But we know from equations (5.18) that a11 + a21 =
0 =+
a12 a22 .
l l
Therefore
 0 0
det A =   = 0.
 a21 a22 
If μ ≠ 0 we instead add λ/μ×row 1 to row 2.

Property 7: If A and B are square matrices of the same size, then the
determinant of the product is the product of the determinants: det(AB)
= det A det B.

Exercise 5.18 For general matrices

 a11 a12   b11 b12 

A=   , B= ,
 a21 a22   b21 b 22 
show by explicit calculation of both sides that det AB = det A det B.

Property 8: A matrix A has an inverse if and only if det A ≠ 0. (A is

then said to be non-singular.)
Here we summarize some of the important properties of
determinants.

1. If every element of one row (or column) is multiplied by a scalar

λ, so is the determinant. If every element of the matrix A (N×N)
is multiplied by a scalar λ, the determinant is multiplied by λN:
det(λA) = λn det A.
2. If two rows (or columns) are interchanged, the determinant
changes sign (but not magnitude).
3. If two rows (or columns) of A are proportional, then det
A = 0.
4. If the rows (or columns) are linearly dependent, then det A = 0.
5. The determinant is unchanged if a constant multiple of one
row (or column) is added to another row (or column).
174  •  Mathematical Physics

6. The transpose AT has the same determinant as A: det AT = det A.

7. If A and B are square matrices (of the same size) then
det(AB) = det A det
8. A matrix A has an inverse A−1 if and only if det A ≠ 0 (mean-
ing it is non-singular).
9. If a row or column is all zeros, then det A = 0.

5.10.  SOLUTION OF 2 × 2 LINEAR SYSTEMS

In section 5.7, we found an expression for the solution of the

general 2 × 2 system:

ax + by =
h  a b  x h
⇔   =  .
cx + dy =
k  c d   y  k 

Using determinants we can write the solution (Equation 5.14) as

h b  a h
det   det  
 k d  c k
=x = , y . (5.19)
a b  a b 
det   det  
c d   c d  

This alternative form is called Cramer’s rule.

a b 
• If det   ≠ 0 , the system is said to be non-singular.
c d 
In this case Cramer’s rule gives the unique solution of the
system of equations.
a b 
• If det   = 0 , the system is said to be singular. There
c d 
are then two possibilities:
 Matrices  •  175

1. The system has no solution, because the contradiction 0 = 1

can be derived.
2. Otherwise there are infinitely many solutions.

Example 5.9 Determine whether the following system has a

solution, and if so find every solution:
3x + 4y = 5.
6x + 9y = 9
First check if the system of equations has a solution:

3 4
det  = 3 ≠ 0,
6 9
so the system has a unique solution. By Cramer’s rule this is

5 4  3 5
det   det 
=9 9 9  6=9  −3
=x = = 3, y = −1.
3 4 3 3 4 3
det   det  
6 9 6 9
The two simultaneous equations could also be solved by the
elementary method of elimination.

Example 5.10 Determine whether the following system has a

solution, and if so find every solution:

3x + 4y = 5
.
6x + 8y = 9

First check if the system of equations has a solution. A singular

system has either no solution or infinitely many.
3 4
det   = 0,
6 8
176  •  Mathematical Physics

so the system is singular. Here we shall use standard row and

column operations on the coefficients to attempt to solve the
system. For the 2 × 2 case we could also use the elementary
method of elimination

 3 4 5  R2 − (2 × R1)  3 4 5 
    
 6 8 8   0 0 −2 
from which 0 = −2, a contradiction. The system therefore has
no solution.
Take another look at the two equations we began with. The left
sides are the same except for a factor 2, but the right sides are
not; hence they contradict one another.

Example 5.11 Determine whether the following system has a

solution, and if so find every solution:

3x + 4y = 5
.
6x + 8y = 10
We have that

3 4
det   = 0,
6 8
so the system is singular. Using row reduction:

3 4 5  R 2 − (2 × R1)  3 4 5 
    
 6 8 10  0 0 0

so 0 = 0 and 3x + 4y = 5; thus the system is equivalent to the

one equation, 3x + 4y = 5. There are infinitely many solutions.
1
For example, let x = t (an arbitrary number) and= y ( 5 − 3t ) .
4
Take another look at the two equations we began with. The
second equation is just twice the first equation, and therefore
adds no more information than one equation alone (which then
cannot be solved uniquely for two unknowns).
 Matrices  •  177

Exercise 5.19 For each of the following systems, determine

whether it has a solution and if so find every solution:
(i) x + 2y = 4
2x + 2y = 6
(ii) x + 2y = 4
2x + 4y = 8
(iii) x + 2y = 4
2x + 4y = 3

5.11.  SOLUTION OF 3 × 3 SYSTEMS

Cramer’s rule: The solution of the general 3 × 3 linear system of

equations

 a11 a12 a13   x   h1 

      
 a21 a22 a33   y  =  h2  (5.20)
a a32 a33   z   h3 
 31
is
 h1 a12 a13   a11 h1 a13   a11 a12 h1 

det  h2 a22 a23 
  
det  a21 h2 a23 

det  a21 a22 h3 

h a a33  a  a h3 
=x =  3 32
, y =  31 h3 a33  , z  31 a32 .
det A det A det A
(5.21)
Proof: We can write the first of equations (5.20) as

h1 − a11x − a12y − a13z = 0.  (5.22)

We verify that (5.21) satisfies this by evaluating the left hand

side of (5.22). We get (using property 2 from Section 5.9 to exchange
columns, changing signs)
178  •  Mathematical Physics

det A ( h1 − a11 x − a12 y − a13 z )

 h1 a12 a13   h1 a11 a13 
   
h1 det A − a11 det  h2 a22
= a23  + a12 det  h2 a21 a23 
h a a33  h a31 a33 
 3 32  3
 h1 a11 a12 
 
− a13 det  h2 a21 a22 
h a a32 
 3 31

 h1 a11 a12 a13 

h a11 a12 a13 
= det 
1

 h2 a21 a22 a33 

 
 h3 a31 a32 a33 

by expanding the determinant by the first row,

= 0,
since two rows of the determinant are equal. The other two equa-
tions in (5.20) can be verified similarly. This completes the proof.
• If det A ≠ 0, the system is said to be non-singular. In this case
Cramer’s rule gives the unique solution of the system of equations.
• I f det A = 0, the system is said to be singular. There are then
two possibilities:
1. The system has no solution, because the contradiction
0 = 1 can be derived.
2. Otherwise there are infinitely many solutions.

Exercise 5.20
(i) Determine which of the following systems has a unique solution
2x − 3y + 3z = −2
2x − 2y + 6z = 1
x + λz = k
for (a) λ = 6, k = 1, (b) λ = 3, k = 3, (c) λ = 6, k = 7/2.
(ii) Use Cramer’s rule to find the unique solution.
(iii) Find all the solutions in case (c).
 Matrices  •  179

5.12.  HOMOGENEOUS SYSTEMS

If (h1, h2, h3) = (0, 0, 0) then (5.20) can be written

Ax = 0.
We refer to this as a homogeneous linear system.
•  If a homogeneous system is non-singular (i.e. if det A ≠ 0)
Cramer’s rule gives x = 0 (i.e. x = 0, y = 0, z = 0). We say that
the system has only the trivial solution.
• If a homogeneous system is singular (i.e. if det A = 0) then there
will be infinitely many solutions. (One might guess this since
Cramer’s rule gives 0/0 for each of the x, y, and z. A proper proof
can be constructed from the converse of property 6 of determi-
nants: that the rows of a singular matrix are linearly dependent;
but we shall not stop to do this). We shall see an important
application of this in Section 5.14.

We have stated the theory in terms of 3×3 systems, but the argu-
ment applies to any n×n homogeneous system (i.e. any system of
the form Ax = 0 with A an n × n matrix. In particular if A is a num-
ber, a say (a 1 × 1 matrix) then ax = 0 has only the trivial solution
x = 0 if a ≠ 0 and infinitely many solutions if a = 0. The statements
above are just the (important!) generalizations of this result to simul-
taneous equations.

5.13.  A FORMULA FOR THE INVERSE MATRIX

Let
 a11 a12 a13 
 
A =  a21 a22 a23  .
a a32 a33 
 31
In Section 5.8, we defined the minor of an element of A. For
example,
 a22 a23 
• the minor of a11 is det  a ,
 32 a33 
180  •  Mathematical Physics

 a21 a23 
• the minor of a12 is det 
 a31 a33 

and so on. In general, to obtain the minor of aij we cross out the
ith row and jth column from A and take the determinant of what
remains. We now define the cofactor of aij, call it cij:
cij = (−1)i+j × Mij where Mij is the (i, j) minor of A.
So
 a22 a23   a21 a23 
c 11 = det   , c 12 = −det  .
 a32 a33   a31 a33 
and so on.
Using cofactors the determinant for a 3 × 3 matrix can be written

det A = a11c11 + a12c12 + a13c13

where
 a21 a23 
( −1)1+ 2 M 12 with minor M 12 =
c 12 = det  
 a31 a33 
and so on. The determinant of any N × N matrix can be written
similarly as the sum of the products of the elements of any row (or
column) with their corresponding cofactors.
Finally, we define the adjoint matrix to A, adj A, as the matrix
constructed from the cofactors of A:

 c11 c21 c31 

 
adj A=  c12 c22 c32  .
c c23 c33 
 13
Note the arrangement of rows and columns – it is the transpose
of what you might guess.

Exercise 5.21 Verify that if A is any 2×2 matrix whatsoever, then

A(adj A) = (det A)I.
 Matrices  •  181

In fact this result holds for any square matrix. From it we get a
formula for A−1:
1
A −1 = adj A. (5.23)
det A 

It is clear from this that A−1 exists provided A is non-singular

(det A ≠ 0) (property 8).
Consider now the linear system
Ax= h.
If det A ≠ 0 we can form A . Multiplying through on the left by A−1
−1

we get
A−1Ax = A−1h.

Since A−1Ax = |x = x, it follows that

x = A−1h
as the solution of the system. Thus, we see again that the system has
a unique solution if det A ≠ 0. This is what we set out to understand
in Section 5.1 (Equation 5.6).

5.14.  EIGENVALUES AND EIGENVECTORS

Consider the linear system of equations for x

Ax = λx  (5.24)
where A is a given matrix and λ some real number. For example, if
A is a 2 × 2 matrix, this system would be explicitly

a11x + a12y = λx
a21x + a22y = λy  (5.25)
or equivalently
a11 x + a12y − λx + 0y = 0
a21 x + a22y + 0x − λy = 0.  (5.26)
182  •  Mathematical Physics

Putting equations (5.26) back in matrix form gives

 a11 a12  x   −l 0  x   0 
a + =
 21 a22  y   0 −l 
   
 y   0 

which we can write as

Ax − λIx = 0

or

(A − λI)x = 0,  (5.27)

where I is the identity matrix. Equation (5.24) can be written as equa-

tion (5.27) for any N × N matrix A and the N × N identity matrix I.
From Section 5.12, we know that if det(A − λI) = 0 then equa-
tion (5.27) has a unique solution, namely x = 0. This is not of any
interest. To obtain non-trivial solutions of equation (5.24) we must
impose the condition
det(A − λI) = 0.  (5.28)
Since A is supposed given, equation (5.28) can be satisfied only
if λ takes appropriate values.
Definition: A value of λ that satisfies equation (5.28) is called an
eigenvalue of the matrix A. If λ is an eigenvalue, equation (5.24) has
infinitely many non-zero solutions for x. These solutions are called
eigenvectors of the matrix A (corresponding to the eigenvalue λ).
Not all of the (infinite number of) eigenvectors will be indepen-
dent. In fact, if A is an n×n matrix it will have at most n independent
eigenvectors.

Example 5.12 Find the eigenvalues of the matrix

 0 1 −1 
 
A=  1 2 1 .
 −1 1 2 
 
 Matrices  •  183

λy
Ax = λx

y
x

x λx
FIGURE 5.1:  Illustration of an eigenvector. The vector x is (left) multiplied by the matrix A
and the result is another vector in the same direction as x but scaled in magnitude by a factor
λ (its eigenvalue). If x is an eigenvector of A (with eigenvalue λ), then so is any multiple of this,
i.e. αx is also an eigenvector (and it has the same eigenvalue). We often use unit eigenvectors,
which are scaled to have unit length (x • x = 1).

The eigenvalues are solutions of det(A − λI) = 0:

  0 1 −1   l 0 0  
   
det   1 2 1 −  0 l 0 
  −1 1 2   0 0 l  
   
 −l 1 −1 
 
=det  1 (2 − l ) 1  =0.
 −1 1 ( 2 − l ) 


Write out the determinant by expanding out the top row, i.e.

0 = (−λ)[(2 − λ)2 − 1] − (1)[(2 − λ) + 1] + (−1)[1 + (2 −λ)];

(2 − λ)2 − 1 is a “square minus a square” so can be factorized:

0 = −λ[(2 − λ) + 1][(2 − λ) − 1] − [(2 − λ) + 1] − [1 + (2 − λ)];

(2 − λ + 1) = (3 − λ) is a common factor, so we take it outside:

0 = − λ (3 − λ)(1 − λ) − (3 − λ) − (3 − λ) = − (3 − λ)[λ (1 − λ) + 2].
184  •  Mathematical Physics

Do not expand out the cubic – you will not be able to factorize
it (unless you are given or can guess one solution). Thus,

0 = (3 − λ)(λ2 − λ − 2) = (3 − λ)( λ + 1)( λ − 2).

So the eigenvalues are λ = 3, −1 and 2.

Example 5.13 Find the unit eigenvectors corresponding to

the eigenvalue λ = 3 in example 5.12.
The eigenvectors (x, y, z) are found by solving equation (5.27):
(A − λI)x = 0, or

 −l 1 −1   x   0 
    
 1 (2 − l ) 1  y =  0 .
 −1
 1 ( 2 − l )   z   0 
The eigenvectors corresponding to λ = 3 therefore satisfy

 −3 1 −1   x   0 
    
 1 −1 1   y  = 0 ,
 −1 1 −1   z   0 
    
which we can write as
−3x + y − z = 0,
x − y + z = 0,
−x + y − z = 0.

The last two equations are the same, so rearranging we get

y = x + z.
Substitute for y in the first equation, and we obtain:
−3x + (x + z) − z = 0,
from which x = 0, and hence y = z. A vector (x, y, z) with x = 0
and y = z must be of the form (0, k, k), so the eigenvectors are
(0, k, k), k ≠ 0.
 Matrices  •  185

To find the unit eigenvectors we must chose k such that the

vector length (magnitude) is unity. The length of the vector
(0, k, k) is 2 k2 ; therefore choosing k = ± 1 2 will give a vector
of unit length. The unit eigenvectors are ± 1 ( )
2 (0, 1, 1).

Note what is going on here: We have just chosen λ such that

the equations do not have a unique solution. Therefore there is
no point in trying to find one. A non-unique solution must contain
an arbitrary parameter! Usually one chooses this to be one of x, y
or z (whichever is most convenient) but any combination will do.
Whatever choice is made, the same pair of equal and opposite unit
eigenvectors will be obtained.

Example 5.14 Find the remaining eigenvectors following

from example 5.12.
For λ = 2 we have similarly

( A − I) x= ( A − 2I) x=0,
i.e.
 −2 1 −1   x   0 
    
 1 0 1  y =  0 ,
 −1 1 0   z   0 
    
which gives
–2x + y – z = 0,
x + z = 0,
–x + y = 0,

from which z = –x, y = x. Thus, the eigenvectors are of the form

( )
(k, k, –k), k ≠ 0 . The unit eigenvectors are ± 1 3 (1,1, − 1 ) .
For λ = –1, similarly, the corresponding eigenvectors are
found to be of the form (2k, –k, k). The unit eigenvectors are
( )
± 1 6 ( 2, − 1,1 ) .
186  •  Mathematical Physics

Note: an N × N real symmetric matrix has N eigenvalues which

are all real numbers (but not necessarily distinct if the eigenvalue
equation has repeated roots). This can provide a useful check on
your work.

Exercise 5.22 Show that λ = 1 is an eigenvalue of the matrix

 1 −4 4 
 
 −4 −1 0  .
 4 0 3 

Find the other eigenvalues and the corresponding eigenvectors.

5.15.  MATRICES AS TRANSFORMATIONS

We can view the application of a matrix A to a vector x as either

effecting a transformation from the vector x to the vector x¢ = Ax (an
active transformation) or as giving the coordinates x¢ = Ax of a fixed
vector x relative to a new coordinate system (a passive transforma-
tion). Depending on the context, different choices of A correspond to
different active transformations or to different changes of coordinate
system.

Rotation in the Plane

Suppose that (x, y) axes in the plane are rotated through
an angle q to new axes (x¢, y¢). A point P acquires new coor-
dinates (x¢, y¢) which, by Figure 5.2, are related to its original
coordinates (x, y) by
x¢ = OC = OD + DC = OD + AQ = x cos(q) + y sin(q)
y¢ = OE = PQ − QC = PQ − AD = y cos(q) − x sin(q)

Then the change of coordinates from (x, y) to (x¢, y¢) is

x¢ = x cos(q) + y sin(q)
y¢ = −x sin(q) + y cos(q)
 Matrices  •  187

y´ P

θ x´
C
D
E
Q
θ
O x
A
FIGURE 5.2:  Rotation of coordinates in the plane. A point P given in coordinates (x, y). The
same point is at coordinates (x¢, y¢) in axes that are rotated (counterclockwise) by an angle q
from the original axes.

and the change of coordinates can be written in matrix form as

x¢ = R x,  (5.29)
where R is the matrix

 cos (q ) sin (q ) 
R= .
 − sin (q ) cos (q ) 
Alternatively, we may think of this as a transformation of the
point P with coordinates (x, y) to a new point P¢ with coordinates
(x¢, y¢) in a fixed coordinate system by rotation about the origin
through the angle −q:
 x  x′ 
 y  →  y′  .
   

Exercise 5.23
(i) Show that RRT = I and RTR = I, i.e. R−1 = RT.
(ii) What matrix would represent the transformation of a point
by rotation about the origin through an angle +q in a fixed
coordinate system?
188  •  Mathematical Physics

A square matrix satisfying R−1 = RT is said to be orthonormal or

unitary. Rotations are represented by orthonormal matrices. The con-
verse is slightly more complicated: any transformation effected by an
orthonormal matrix with positive determinant is a rotation. (Negative
determinant gives a rotation followed, or preceded, by a reflection.)
Note that if the rows of a 3 × 3 matrix are r1, r2 and r3 then the
requirement of orthogonormality can be seen equivalently to be that
ri · r j = 0 for i ≠ j (when r1, r2 and r3 are considered as vectors) and
that r1, r2, and r3 are unit vectors (r1 · r1 = 1 etc.).

Exercise 5.24 For the matrix R below, compute

(i) RT and
(ii) RRT.
(iii) Explain why the matrix is orthogonal.
(iv) Find the corresponding angle of rotation if this represents a
rotation of coordinate axes:

 12 3 2
R= .
− 3 2 1 2 


Rotation in Three Dimensions

A 3×3 orthonormal matrix (i.e. a matrix R such that RRT = I)
with determinant +1 effects a rotation in 3D. The axis of rotation
is the only real eigenvector of the matrix. The angle of rotation q is
obtained from 1 + 2 cos(q) = R11 + R22 + R33 (although we shall not
prove this here).

Exercise 5.25 Verify that the matrix

 2 3 −1 3 2 3 
 
 2 3 2 3 −1 3 
 −1 3 2 3 2 3 
 

is orthogonal. Show that λ = 1 is an eigenvalue of the matrix and

hence find the axis and angle of rotation.
 Matrices  •  189

Projection. The matrix transformation x¢ = Px where

1 0 0 
 
P = 0 1 0
0 0 0
 
effects a projection onto the (x, y) plane, since it gives

x¢ = x, y¢ = y, z¢ = 0.
So any point (x, y, z) becomes

 x  x′ = x 
   
 y  →  y′ = y .
   
z  z′ = 0 
Note that det P = 0. A transformation represented by a singular
matrix P is a projection if P2 = P (i.e. repetition of the transformation
has no further effect).
Figure 5.3 shows an example using the projection matrix

1 0 0 
 
P = 0 1 0 .
0 0 0
 

x
FIGURE 5.3:  Left: an example of orthogonal projection. Points are projected onto the (x, y)
plane using the matrix P.
190  •  Mathematical Physics

Figure 5.4 shows an example using the projection matrix

 1 −1 
Q= .
0 0
Reflection. Reflection about a plane is another transformation
that can be achieved by application of a matrix. In the case of reflec-
tion, the determinant of the corresponding matrix is always −1. For
example, reflection about the line y = x is achieved by the matrix

0 1 
R= 
1 0 

(x, y)

x
(x-y, 0)

FIGURE 5.4:  In general a projection will not be orthogonal. For example the matrix Q repre-
sents the projection in two dimensions as shown in the figure.

since then
 x′   0 1   x   y 
= y′  =    
  1 0   y  x 
so the transformation of a point (x, y) is

 x   y
 y →  x  .
   

Exercise 5.26 Verify that (in three dimensions) reflection in the

(x, y) plane is achieved by

 x′   1 0 0   x 
 ′   
 y  =  0 1 0   y .
 z′   0 0 −1   z 
    
 Matrices  •  191

5.16.  EXTENSION: TRANSFORMATION GROUPS

Consider again the rotations in the plane from section (5.15).

Rotation clockwise through the angle q (an active transformation) is
effected by the matrix
 cos (q ) − sin (q ) 
 
 sin (q ) cos (q ) 
acting on the vector (x, y) (since an active transformation clockwise
through an angle q is equivalent to a passive transformation counter-
clockwise through q.) A further rotation through q¢ will be given by
the product
 cos (q ′ ) − sin (q ′ )  cos (q ) − sin (q )   cos (q + q ′ ) − sin (q + q ′ ) 
  = 
 sin (q ′ ) cos (q ′ )  sin (q ) cos (q )   sin (q + q ′ ) cos (q + q ′ ) 
acting on (x, y), which, as we expect, is a rotation through an angle
q + q¢. The product of two rotation matrices is therefore always
another rotation matrix with a different parameter (q+q¢ in this
example). The set of rotation matrices also contains an identity
element which leaves a vector unchanged, namely the rotation
matrix with parameter q= 0, and a matrix corresponding to the
inverse of any rotation, through q say, that returns us to the origi-
nal vector, namely the matrix with parameter −q .
These conditions, together with the fact that matrix multplication
is associative (i.e. A(BC) = (AB)C), define the conditions for the set of
matrices to form a group. We can now work backwards. Any set of matri-
ces that have the property of constituting a group will effect a rotation
in some space. Thus we can use a group of transformations to define
“vectors” in the corresponding space. The simplest example would be to
go from three dimensions to n-dimensional vectors transforming under
the n × n rotation matrices. Another example is provided by four-vectors
in relativity, which transform under a set of matrices representing the
Lorentz group. (These matrices are different from four-dimensional rota-
tions, so relativistic 4-vectors are not the same as “ordinary” vectors in four
dimensions.) We can also define objects representing transformations in
some kind of internal spaces, such as the two (complex) dimensional spin
space of the electron, or the three-dimensional color space of quarks.
192  •  Mathematical Physics

The components of the transformation matrices we have consid-

ered so far are differentiable functions of the one or more param-
eters (representing what are called Lie groups). We can also extend
groups of transformations, such the reflections we considered above,
which have finite numbers of elements. These represent discrete
symmetries such as one finds, for example, in crystal structures.

Revision Notes

After completing this chapter you should be able to

• Solve a system of linear equations by reducing the matrix
representation to triangular form
• Find the inverse of a matrix by row or column operations.
Perform algebraic operations on matrices
• Find the determinant of a matrix, using the properties of
determinants to simplify the calculation
• Use det AB = det A det B to find the determinant of a
product
• Use the determinant of coefficients to determine whether
a system of equations has one, many or no solutions
• Define and compute eigenvalues and eigenvectors
• Construct matrix representations of simple transformations
• Determine whether a given matrix is orthogonal

5.17. EXERCISES
1 0 2
1. If 0 1 3  find det A, and hence (or otherwise) show

 
 
 −1 1 1 
that the equations
x + 2z = 0
y + 3z = 0
−x + y + z = 0
have non-trivial solutions.
 Matrices  •  193

2. Solve the equations

2x + 3y + 5z = 7
x + 7y + z = 6
3x + y + 2z = 9.

3. Find the values of λ and μ for which the equations

x + 2y = μ
2x + λy = 2
have (i) a unique solution, (ii) no solution, (iii) an infinite
number of solutions. Give expressions for the solutions
where possible.
4. For what value(s) of λ does the system
x − 2y + z = 0
x+λz=0
3x + y + 2z = 0
have non-trivial solutions?  2 0 −1 
5. Show that the product of the two matrices  0 2 1 
 2 −1 0 
 
1 1 2 
 
and 2
 2 −2  is 6I, where I is the unit matrix. Hence
 −4 2 4 
 
(or otherwise) solve the set of linear equations

x + y + 2z = 1
2x + 2y − 2z = 3
−4x + 2y + 4z = 5.

1 0 −1   3 −2 1 
   
6. Given A =  2 1 0  and B =  3 0 1  find
 −1 2 −1   2 0 1
   
det(AB).What is det(BA)?
194  •  Mathematical Physics

0 −1 2 0
7. Evaluate the determinant 4 0 0 1 .
0 2 2 −1
−1 −1 1 1

 1 0 −1  0 1 3
   
8. If A =  2 2 5  and B =  −1 1 1  what are AB, AT,
 3 −1 2   −2 1 0 
   
(ATBT)T where AT denotes the transpose of A?
1 0 4   1 −1 −2 
   
9. If A =  −1 1 0  and B =  0 0 1  find AB and
 3 1 −1  4 0 1 
   
verify by explicit calculation of both sides that det AB = det A
det B.
1 2 
10. What are the eigenvalues of the matrix   ?
2 3
1 2 3
11. Verify that   is an eigenvector of   and find a
1 3 2
second linearly independent eigenvector.
12. The matrix A and its eigenvalues, λ, and eigenvectors, x,
satisfy Ax = λx. Deduce the eigenvalues of A−1.
13. A linear transformation is represented by
 x′   −1 1   x 
 y′  =  2 3   y  .
    
 ind the inverse transformation and hence the equation of
F
the line 2y + x = 0 in the (x¢, y¢) coordinate system.

1 0 −1 
 
14. Find the eigenvalues of the matrix Q =  0 2 0  and find
 −1 0 −1 
 
the eigenvector corresponding to the eigenvalue λ = 2.
 Matrices  •  195

15. Show that λ = 1 is an eigenvalue of the matrix

1 0 −1 
 
Q =  −2 1 0  . Find the corresponding eigenvector and
2 0 4 

the other eigenvalues.

5.18. PROBLEMS

1. Show that the eigenvalues of the matrix

1 3 0 
 
A =  3 −2 −1 
 0 −1 1 
 
are 1, 3, and −4, and find corresponding eigenvectors.
Verify that these eigenvectors are mutually orthogonal.
If the eigenvectors are (x1, x2, x3), (y1, y2, y3), (z1, z2, z3) and
a matrix P is defined as

 x1 x2 x3 
 
P =  y1 y2 y3 
z z2 z 3 
 1
show that PAPT is a diagonal matrix.
2. Reduce to triangular form the linear equations
4x + 3y + 2z = 1
w + 4y + 3z = 2
2w + x + 4z = 3
3w + 2x + y + λz = u
where λ, u are any real numbers. For what values of λ, u
will the equations have a solution? When is the solution
unique? Taking u = 4, what is the solution for the cases
λ = 0 and λ = 5?
196  •  Mathematical Physics

3. Solve the linear equations

x + 3y + 2z = 4
2x + y + 3z = 5
3x + 2y + (1 + μ)z = 0
where μ is a constant, and hence show that there is a single
value of μ for which the equations have no solution. Find
the value of μ which makes y = 0, and find the correspond-
ing values for x and z.
4. The coordinates (x, y, z) of a point, P, satisfy the determi-
nantal equation

x y z 1
1 0 1 1
= 0.
2 1 0 1
1 3 2 1
By working out the determinant to express the equation
in a simpler form, show that P must lie on a certain plane.
Verify that the point with coordinates (1, 3, 2) lies on this
plane, and explain how this could be shown directly using
the determinantal form of the equation.
1 0 2 
 
A  0 −1 0  satisfies the equation
5. Show that the matrix=
3 0 1 
 

( A − l1I)( A − l2I) ( A − l3I) =

0

where λ1, λ2 and λ3 are the eigenvalues of A and I is the

unit matrix.
By expanding the equation for A above into a polynomial of
the form A3 + aA2 + bA + cI = 0, find the inverse of A.
 Matrices  •  197

6. Show explicitly that the matrix

2 1 1 
 
A = 0 2 3
0 0 5
 
satisfies the matrix equation

A3 − 9A2 + 24A − 20I = 0.

Hence or otherwise find A−1.

 CHAPTER

6
DIFFERENTIAL
EQUATIONS 1

Differential equations are fundamental to physical science. The

behavior of any system continuously evolving in time (the motion
of a body subject to a force, a chemical reaction, ...) is governed by
one or more differential equations. So is the behavior of continu-
ous systems in space. Solving the equations means finding out how
the body moves given the forces acting, how the reaction proceeds
given the reagents etc. Only the simplest of equations can be solved
exactly. But these are important in developing a physical intuition for
the sort of behavior to expect. (Conversely a good physical insight
will provide a basis for the mathematical solution.) They are also
used later in the derivation of approximations to the solutions of
more complex equations. You need to develop a good understanding
of the equations in this chapter.

6.1.  WHAT ARE DIFFERENTIAL EQUATIONS?

A differential equation is an equation for an unknown function,

y(t) say, of an independent variable (in this case t) that involves one or
more derivatives of y with respect to t.
200  •  Mathematical Physics

Example 6.1 Here are some examples of differential equations,

from different fields.
A simple model for population growth (with λ > 0), or radioactive
decay (with λ < 0)
dy
= l y. (6.1)
dt 

Simple harmonic motion, w a constant

d2 y
2
= −w 2 y. (6.2)
dt 

Governor of steam turbine

d3 y d2 y dy
3
− 3 2
+ 4 + 2y =
40. (6.3)
dt dt dt 

Motion of a planet in general relativity

d2 u
+ u = k + 3 u2 . (6.4)
dq 2


Economics growth model, A, B constants

dr
dt
( )(
=Ar 1 2 + 1 Br 1 2 + 1 . ) (6.5)


Geophysics, f(x) a given function of x

d 5
dx
{
x f ( x) 1 + y = }
5 f ( x ) x4 .

(6.6)

The order of a differential equation is the number of the highest

derivative appearing in it.
 Differential Equations 1  •  201

Example 6.2 What are the orders of equations (6.1) and (6.3)?
Equation (6.1) is first order because it contains only first
derivatives (i.e. dy/dt).
Equation (6.3) is third order because the number of the highest
derivative is 3 (i.e. d3y/dt3).

Exercise 6.1 What are the orders of the differential equations

(6.2) and (6.4) to (6.6)?

It might be helpful to clarify the notation that is used for differ-

ential equations. If a function y depends on time t we often write y
instead of dy/dt, ÿ for d2y/dt2 etc. If the independent variable is not
time but position, x, we may use “prime” symbols like y¢, y¢¢, . . . yn
for dy/dx, d2y/dx2, . . . , dny/dxn. For example, equation (6.4) can be
written as

u¢¢ + u = k + 3u2.

Yet another notation is to use superscript numbers to signify the

order of the derivative; e.g. f (3)(x) means d3f/dx3 and so on. In treat-
ing the dynamics of a point particle, we often use x(t) to denote the
position as a function of time, with x here as the dependent vari-
able. For most of this chapter we shall use x or t as the independent
variable.

6.2.  SOLVING DIFFERENTIAL EQUATIONS

So the simplest differential equations are those of the form

y¢ = f(x), or y¢¢ = f(x), . . . , y(n) = f(x). To solve the equation means
to determine the unknown function y(x) which will turn the equa-
tion into an identity upon substitution. All we need to do to solve
these types of differential equations is integrate the appropriate
number of times (not forgetting the correct number of constants
of integration).
202  •  Mathematical Physics

Definition: We call the solution of a differential equation that

contains the correct number of arbitrary constants the general
solution, y(x).

dy
Example 6.3 Solve for y(x) the differential equation = x2
dx
Step 1: In this simple case we can integrate both sides
directly to find y(x)
x3
y ( x=
) ∫ x dx=
2
+C
3 
where C is a constant.
Step 2: We can then write our solution to the differential
equation as
x3
y ( x=
) +C
3

Do not forget the constants when solving differential

equations.
Step 3: We can check that this is indeed a solution by
back-substitution:

dy ( x ) d  x 3 
=  + C=
 x
2

dx dx  3 

Do not omit step 3 to save time. On the contrary, it can save a lot of
wasted work if you pick up a mistake at this point.
Exercise 6.2 Solve the following, and check your solutions:
dy x
(i) =e ,
dx
dy
(ii) = sin ( t ) .
dt
 Differential Equations 1  •  203

6.3.  FIRST ORDER SEPARABLE EQUATIONS

A first order differential equation is called separable if it can be

written in the form

dy
= f ( x ) g ( y) , (6.7)
dx 

where f(x) is a function of x only and g(y) is a function of y only, and

this includes cases where f or g are simply constants. Equations of this
form can be rearranged so that all terms depending on x and all terms
depending on y appear on opposite sides of the equation (this is the
separation of the variables), and then integrated

dy

∫ g ( y) = ∫ f ( x ) dx.  (6.8)

It should then be possible to find the solution y(x) that satisfies the
differential equation.
Note that we cannot write equation (6.7) as y = g(y) ∫ f(x)dx as
many begining students are tempted to do, because y and hence g(y)
is a function of x.

Example 6.4 Solve for y(x) the differential equation yy¢ = x.

Step 1: We separate the variables in the equation in such a way
that the left hand side becomes a function of y only, and the
right side a function only of x.

dy
∫ y dx dx = ∫ x dx ,
or

∫ y dy = ∫ x dx .
204  •  Mathematical Physics

In this form both sides can be directly integrated.

Step 2: There is no need to include two constants of integration
here: any constant on the left can be taken over to the right and
included in C.

y2/2 = x2/2 + C.

Step 3: Write the general solution. Hence

y(x) = (x2 + 2C)1/2.

As noted in the example above, constants should be combined

where possible so that the total number of arbitrary constants in
the solution is immediately visible. The solution here contains
only one arbitrary constant, however arrived at. In fact, the fac-
tor 2 in 2C is of no significance (2C, like C, can be any arbitrary
number), so we would normally write the solution to example 6.4
just as
y(x) = (x2 + C )1/2.

The only difference between Example 6.4 and those in Section

y2
6.2 is that on integration we obtain a function of y ( here) on the
2
left hand side rather than y itself. So equations that can be expressed
in the form we are considering are known as first order separable
equations.
Key point: Rearrange first order separable equations so that you
have
Function of y, y¢ only = Function of x only
(x does not appear explicitly)  (y does not appear explicitly)
i.e. the LHS is a function of y only, and the RHS is a function of x
only. Some manipulation may be required to put the equation in this
form. For example dy/dx = x/y is equivalent to Example 6.4. To see
this, just multiply through by y.
 Differential Equations 1  •  205

Exercise 6.3 Solve the following first order separable equations;

(i) yy¢ = a (a given constant) for y(x),
(ii) y¢ = (1 − y2)1/2ξ1/2 for y(ξ),
(iii) t( y ) + 2 = y for y(t).

Note that each of the solutions in exercise (6.3) contains one

arbitrary constant (and no more). This is a general rule: the solution
to a first order equation contains at most one arbitrary constant.
Thus a first order equation is solved by some formula y = f(x, C)
involving an arbitrary constant C. For all possible values of C we get
all of the possible particular solutions. We will see later that solu-
tions to second order equations have at most two arbitrary constants.
Third order equations require three arbitrary constants, and so on.
Key point: The general solution to an nth order equation must
contain n arbitrary constants.
We will come back to particular solutions later on, in Section 6.9,
and in Chapter 8.

6.4.  LINEARITY VERSUS NON-LINEARITY

Suppose that an nth order differential equation has x as the

independent variable and y as the unknown function. Recall that
the nth derivative of y can be written as y(n). If the equation can be
written in the form

A0 y(n) + A1 y(n−1) + ... + An−1 y + An = 0,  (6.9)

where A0, A1, . . . , An may be constants or functions of x only, then it

is called a linear differential equation. The left side of equation (6.9)
is a linear combination of y, y(1), y(2), . . . , y(n).
However, if the A0, A1 , . . . , An terms themselves contain any of
y, y(1), y(2) or higher derivatives, products of y (e.g. y2), powers of y
206  •  Mathematical Physics

(e.g. y1/2) or any functions of y (e.g. sin(y) or ey), then the equation
is non-linear.

Example 6.5 State if the following equations are linear or

nonlinear, and why
(i) y = λt, (ii) y¢¢ = −(sin(x))y, and (iii) yy¢ = x.
(i) y = λt is linear because the unknown function y
appears linearly.
(ii) y″ = − (sin(x))y is linear because the unknown func-
tions (y(x) and their derivatives) appear linearly, even
though x does not.
(iii) yy′ = x is non-linear because it involves a product of
unknown functions, y and y′.

Exercise 6.4 Classify the equations shown in Example 6.1 as lin-

ear or non-linear.

Linear equations are important examples of differential equa-

tions because:

(i) They can often be solved.

(ii) Approximations to interesting physical problems can often
be obtained as linear equations.
(iii) When An = 0 in equation (6.9) solutions can be added to give
new solutions (see Section 6.5).

When faced with a linear equation in the form of equation (6.9) we

generally take An over to the right hand side and also divide through
by the coefficient A0 of the highest order derivative y(n). Thus first and
second order equations in this form can alternatively be written as

y¢ + P (x)y = Q(x),(6.10)
 Differential Equations 1  •  207

and

y¢¢ + F(x)y¢ + G(x)y = H(x), (6.11)

where P(x), Q(x), F(x), G(x) and H(x) are referred to as the coeffi-
cients of y and its derivatives, and are functions of x only (or in some
cases are replaced by constants; see Section 6.7). For most of the
rest of this chapter the equations we shall look at will be linear and
usually of first or second order.

6.5.  HOMOGENEOUS VERSUS INHOMOGENEOUS

The linear differential equation y¢ + P(x)y = Q(x) is called

homogeneous if Q(x) = 0 and inhomogeneous otherwise. Similarly,
the linear differential equation y¢¢ + F(x)y¢ + G(x)y = H(x) is called
homogeneous if H(x) = 0 and inhomogeneous otherwise. Some
authors call these unforced (homogeneous) and forced (inhomoge-
neous) because the term Q(x) is often thought of as a forcing term
for a physical system.
For example,

y¢¢ + w2y = 0 is homogeneous,  (6.12)

y¢¢ + w y = 2x is inhomogeneous.
2
(6.13)

If two solutions y = y1(x) and y = y2(x) of a homogeneous linear

equation are known then any linear combination y = Ay1(x) + By2(x)
(with A and B constants) is itself another solution. This is not true for
inhomogeneous equations.
To see how this works, let us work through an example.

Example 6.6 Verify that y = sin(wx) and y = cos(wx) are

solutions of equation (6.12) and that, if A and B are arbitrary
constants, then y = A sin(wx) + B cos(wx) is also a solution.
 208  •  Mathematical Physics

We verify that something is a solution by substitution in the

given equation. If y = sin(wx) then
y¢ = w cos(wx),

y¢¢ = − w2 sin(wx).

Substituting into equation (6.12) we see that −w2 sin(wx) +

w2 sin(wx) = 0, and we have indeed verified that y = sin(wx) is
a valid solution.
Following a similar method we can easily show that y = cos(wx)
is also a solution.
Finally, we can show that the sum of the two solutions
y = A sin(wx) + B cos(wx) is also a solution by direct substitution
into equation (6.12)

d2
2 {
A sin (w x ) + B cos (w x )} + w 2 ( A sin (w x ) + B cos (w x ) )
dx
 d2   d2 
= A  2 {sin (w x )} + w 2 sin (w x )  + B  2 {cos (w x )} + w 2 cos (w x ) 
 dx   dx 
= A ( −w 2 sin (w x ) + w 2 sin (w x ) ) + B ( −w 2 cos (w x ) + w 2 cos (w x ) )
=0

Therefore y = A sin(wx) + B cos(wx) is also a solution.

Exercise 6.5
(i) Verify that y = 1 and y = t are two solutions of d2y/dt2 = 0 and
that y = A + Bt is a solution (A and B constants).
(ii) Let y = y1(x) and y = y2(x) be two solutions of y¢¢ + F(x)y¢ +
G(x)y = 0. Show that Ay1 + By2 is also a solution for arbitrary
constants A, B.
(iii) Given that ex and e−x are solutions of d2y/dx2 − y = 0,
explain why it follows that sinh(x) and cosh(x) are also
solutions.
 Differential Equations 1  •  209

(iv) Verify that y = 2x/w2 is a solution of equation (6.13) but that

y = 2x/w2 + 2x/w2 = 4x/w2 is not.
(v) If y = y1(x) and y = y2(x) are two solutions of the inhomoge-
neous linear equation y¢¢ + F(x)y¢ + G(x)y = H(x), what dif-
ferential equation is satisfied by Ay1 + By2, where A and B are
constants?

6.6.  FINDING SOLUTIONS

We now return to Example 6.3: dy/dx = x2. The solution is the

integral of x2. Suppose that we remember the integral of x2 is some-
thing like x3 but have forgotten whether it is 3x3 or x3/3 (this example
is simple, but if it were more complex we might not remember the
correct factor). We can guess a more general form, y = Ax3, and then
solve for A. Substitution in the differential equation gives

d
dx
{ Ax3 } = 3 Ax2 . (6.14)


This is a solution if 3Ax2 = x2, i.e. if A = 1/3. (Of course, the gen-
eral solution still involves a constant of integration C.) Similarly, if
you guess that the integral is some power of x but cannot remember
which one, you could try y = Axn where both A and n are to be deter-
mined by substitution.

Example 6.7 Solve dy/dx = x2 by trying a solution of the form

y = Axn + C, with A, n and C constants.
Step 1: We let y be of the proposed form and substitute it into
the given differential equation.

y¢ = nAxn−1 = x2.

Step 2: Work out the conditions under which the proposed

form leads to a solution. If this is not possible, then there is
210  •  Mathematical Physics

no solution of the proposed form, and an alternative guess is

required. The above condition is true if both

A = 1/n and n − 1 = 2,

that is, when

n = 3, and a = 1/3

Step 3: Finally we can write down our general solution as

y = x3/3 + C.

We can use this same approach not only where we have forgotten
an exact integral but where we perhaps never knew one, as long as
we can guess the correct form of the solution. This may seem rather
mysterious the first time you see it, but it is not really mystery – it
is experience. You will gain this experience by working through this
chapter. If you are mathematically minded you will soon remem-
ber what sort of functions satisfy what sort of equations. If you are
physically minded you will use physical intuition to guess what sort
of behavior to expect from a given system (as represented by a dif-
ferential equation).
But what happens if we make the wrong guess? Let us see.
Suppose that, instead of trying y = Ax3 in equation (6.14), we guess
y = Ax4 with A a constant. Then dy/dx = 4Ax3 = x2 if A = 1/4x. This
is impossible. (We differentiated y assuming A was a constant and it
has turned out not to be.) So the guess is wrong and we stop here.
The failure tells us that we must start again with a better guess.
The final problem is that the “guesswork” does not feel like “real
math” and we would naturally prefer to be given an algorithm to fol-
low in order to solve these types of equation (i.e. a set of instructions
that always works). There is no algorithm. That is okay; after all inte-
gration is a process of educated guesswork too, and we have learned
to cope with that. Solving a differential equation involves a similar
process of educated guesswork, based on a correct identification of
the type of equation to be solved.
 Differential Equations 1  •  211

Exercise 6.6
(i) By trying a solution of the form y = A cos(px) solve
y¢ = sin(2x) for y(x).
(ii) By trying a solution of the form y = (Ax+B)e2x solve y¢ = 3e2x
for y(x).

However, suppose we try to solve y′ = e− x ln ( x ) . We want to be

2

able to solve the indefinite integral ∫ e− x ln ( x ) dx – but this cannot

2

be expressed in terms of known functions. Therefore not all differ-

ential equations have solutions expressible in terms of known func-
tions. In such cases information about the solution must be obtained
by either numerical or approximate methods, but we will not tackle
that here.

6.7.  HOMOGENEOUS LINEAR EQUATIONS

The following equation is similar to equation (6.10):

y¢ + py = 0,

but here we have a constant coefficient of y, p = constant (instead

of a known function of x, P(x)). In addition, it is first order (it con-
tains y differentiated just once), and is linear (the left hand side is
a linear combination of y and y¢). Also, unlike equation (6.10) it is
homogeneous (there is no term independent of y, i.e. Q(x) = 0). It
is also separable (it can be written as dy/y = pdx). We can therefore
integrate it directly to obtain the general solution, as in Section 6.3.
Alternatively, we can solve it by using physical or mathematical
intuition as the basis of a guessed solution. Let us rewrite the equa-
tion as y¢ = −py. In this form you may recognize that the equation
looks like the radioactive decay law. So we could reasonably expect
an exponential type solution, say y = Aeλx, where A is a constant.
Alternatively, the equation looks something like y′ = y which
we know is satisfied if y = ex. So again we are led to try y = Aeλx.
212  •  Mathematical Physics

Differentiating, we find y′ = λAeλx and therefore, to satisfy the

equation, we need λAeλx + pAeλx = 0. This is true if λ = −p but
with no restriction on A. Thus the general solution in this example,
containing one arbitrary constant, is y = Ae−px.

Let us take a look at a complete worked example.

Example 6.8 Solve y¢¢ − 4y = 0.

Step 1: This equation is second order, linear, homogeneous and
has constant coefficients. But it is not separable. We therefore
have to call on some other experience to solve it. Since it looks
similar to the example we discussed above we can try an initial
guess

y = Aeλx,

and it follows that

y¢ = Aλeλx and y¢¢ = Aλ2eλx.

Step 2: To test if our trial function really is a solution we

substitute y and y¢¢ above into the original differential equation,
which gives

y¢¢ − 4y = (Aλ2eλx) − 4(Aeλx) = 0.

Step 3: Divide through by Aeλx to get λ2 − 4 = 0, from which we

can see that λ2 = 4 and λ = ±2. We have therefore found two
solutions

y1 = Ae2x and y2 = Ae−2x.

Step 4: The general solution must contain two independent

arbitrary constants though, A and B say, and therefore the
general solution is written down as the sum of the two solutions
as follows:

y = Ae2x + Be−2x,
 Differential Equations 1  •  213

where A and B are arbitrary constants. This follows because

a linear combination of solutions of a linear equation gives
another solution. This can be checked by substituting back into
the differential equation.

We will now look at another example where we will solve a simi-

lar second order differential equation.

Example 6.9 Solve y¢¢ − y¢ − 6y = 0.

Step 1: The equation is second order, linear, homogeneous
with constant coefficients, so our experience so far tells us what
form to try. So we try as our initial guess

y = Aeλx,

and therefore

y¢ = Aλeλx and y¢¢ = Aλ2eλx.

Step 2: To test the trial function substitute into the differential

equation to be satisfied. Then cancel Aeλx to get the auxiliary
equation.

y¢¢ − y¢ − 6y = Aλ2eλx − Aλeλx − 6Aeλx = 0,

λ2 − λ − 6 = 0.

This removes the x dependence so our guess was a good one.

The resulting equation in λ is called the auxiliary equation.
Step 3: Solve the auxiliary equation for λ

( λ − 3) ( λ + 2) =0,

λ = −2 or λ = 3.
214  •  Mathematical Physics

Step 4: Substitute the values of λ to form the solutions for y.

The two solutions are

y1 = Ae−2x and y2 = Ae3x.

Remember, the general solution involves two arbitrary

constants since we are solving a second order equation. Hence
the general solution is written as the sum of y1 and y2, but with
two independent arbitrary constants

y = Ae−2x + Be3x,

where A and B are arbitrary constants.

It is very important to remember this method of solving differ-

ential equations with constant coefficients. Let us work through one
more example, following the same steps to first form the auxiliary
equation. We can then solve for λ and write the general solution.

Example 6.10 Solve y¢¢ − 3y¢ − 4y = 0.

Step 1: First check if the equation is linear (no products of
y, y¢, . . .)? Is it homogeneous (there is no term independent of
y)? Does it have constant coefficients (1, −3, −4)? This method
works only if you answer yes to all three questions. With
experience you can do the differentiation and substitution in
your head. For example, we try as our initial guess

y = Aeλx,

y¢ = Aλeλx and y¢¢ = Aλ2eλx.

Step 2: Form the auxiliary equation by substituting our initial

guess into the differential equation and canceling Aeλx.

λ2 − 3λ − 4 = 0.
 Differential Equations 1  •  215

Step 3: Solve the auxiliary equation. And hence

(λ − 4)(λ + 1)=0,

λ= 4 or −1

Step 4: The general solution is the sum

y = Ae4x + Be−x

where A and B are independent arbitrary constants.

Note that in these cases where one of the roots of the auxiliary
equation is positive the general solution is unstable since the solu-
tion will tend to infinity as x → ∞.
Exercise 6.7
(i) Show that the equation y¢¢ + 7y¢ + 12y = 0 has the general
solution y = Ae–4x + Be–3x.
(ii) Show that this can also be written as y = e –7x/2(P cosh(x/2) +
Q sinh(x/2)) with P = A + B, Q = B – A.

Exercise 6.8 Assuming b2 – 4ac > 0, find the general solution of

ay¢¢ + by¢ + cy = 0 (a, b, c constants).

Provided the differential equation is linear, homogeneous and

with constant coefficients, that is all there is to finding the general
solution. Do not try to use this method for equations which are non-
linear, inhomogeneous or which do not have constant coefficients.

6.8. AUXILIARY EQUATIONS WITH REPEATED

ROOTS

In the previous section, we looked at solving the auxiliary equa-

tion for examples where λ has two different values, λ1 and λ2 say.
216  •  Mathematical Physics

Here we are going to look at the special case where we find λ has two
equal values when we solve the auxiliary equation, or repeated roots.
Let us look at an example, following the same steps as previously.

Example 6.11 Solve y¢¢ + 4y¢ + 4y = 0.

Step 1: Is the equation linear, homogeneous and does it
have constant coefficients? We can answer yes to all of these
questions and thus we begin by trying

y = Aeλx.

Step 2: Substitute y, y¢ and y¢¢ in the differential equation, and

form the auxiliary equation

λ2 + 4λ + 4 = 0.

Step 3: Solve the auxiliary equation

(λ + 2)(λ + 2) = 0 and thus λ = −2, −2.

Step 4: Write the solution

y = Ae−2x + Be−2x

which just combines to give y = Ce−2x.

But this cannot be the general solution because it involves only
one arbitrary constant.

We know that every second order differential equation must

have a solution which contains two arbitrary constants, so there must
be another term that contains the second constant. If we proceed
systematically with this example we can try to find a second solution
of the differential equation when one solution is known. We will see
how to solve these repeated roots examples in general.
The trick is to try y = (unknown function) × (known solution).
We try
y = u(x)e−2x,
 Differential Equations 1  •  217

and then calculate y¢ and y¢¢

y¢ =u¢(x)e−2x − 2u(x)e−2x,

y¢¢ =u¢¢(x)e−2x − 2u¢(x)e−2x − 2u¢(x)e−2x + 4u(x)e−2x.

Now we have to find u(x). By substituting y, y¢ and y¢¢ into the

differential equation and canceling e−2x we can find u(x):

y¢¢+4y¢+4y = [u¢¢(x)−4u¢(x)+4u(x)]+4[u¢(x)−2u(x)]+4[u(x)] = 0.

Simplifying we are left with u¢¢(x) = 0, so that u¢(x) = A, and

u(x) = Ax + B where A and B constants. Hence the general solution
is

y = (Ax + B)e−2x.

In fact, you do not need to go through the whole process each

time, if you can remember the form of the solution; you can just use
it directly.
Key point: If the auxiliary equation has λ as a real and repeated
root, we can try a solution of the form y = (Ax + B)eλx.

Exercise 6.9 Find the general solution of y¢¢ + 6y¢ + 9y = 0.

6.9.  PARTICULAR INTEGRALS

Having looked at how to solve homogeneous differential equa-

tions (with constant coefficients), we are now going to see what hap-
pens when the right hand side of equation (6.9) is not zero. Consider
a linear inhomogeneous equation of the form

y¢¢ + fy¢ + gy = H (x),(6.15)

where H(x) is a known function of x (similar to Equation 6.11, but here

we have constant coefficients f and g instead of known functions of
x, F(x) and G(x)). If we can find a solution of equation (6.15) which
has no arbitrary constants, this is called a particular integral (PI). Sys-
tematic methods do exist for finding PIs. However, in simple cases,
218  •  Mathematical Physics

the best method is again to use educated guesswork. Roughly speak-

ing, we are going to try a PI of similar form to H(x). The cases that
follow are those that occur most commonly. Remember, we assume
throughout that f and g are constants. Thus this section explains how
to find a PI for inhomogeneous linear equations with constant coef-
ficients. The examples below will all be for second order equations,
but the method works similarly if the order is higher.
Case 1 Find a PI for y¢¢ + f y¢ + gy = H(x) where H(x) is a poly-
nomial in x.
Method Try a PI of the form yPI = polynomial of the same
degree.

Example 6.12 Find a PI for

y¢¢ + 7y¢ + 12y = 24x + 26.

Step 1: First we try a polynomial of the same form as H(x)

yPI = Cx + D,

y¢PI = C and y¢¢PI = 0.

Step 2: Substitute in the differential equations

0 + 7C + 12(Cx + D) = 24x + 26.

Step 3: Since this must hold for all x we can compare coefficients
of x0 (= 1) and coefficients of x1 (= x). So we have that
12C = 24, and

7C + 12D = 26.

Solving these we find

C = 2, and D = 1.

Step 4: Write down the particular integral solution

yPI = 2x + 1.
 Differential Equations 1  •  219

The method fails if g = 0 in equation (6.15). In this case substi-

tuting yʹʹPI = 0 and yʹPI = A in the left hand side in the example above
would not work. The way to proceed under these circumstances is
to multiply through by x in the original trial yPI and try again. The
correct guess in this case is yPI = Cx2 + Dx.

Exercise 6.10 Find a PI of the differential equations

(i) y¢¢ – 3y¢ – 4y = –16x – 40,
(ii) y¢¢ – 2y¢ + y = x2 – 3.

Case 2 Find a PI for y¢¢ + f y¢ + gy = H(x) where H(x) = Cemx

(C and m constants).

Method Try a PI of the form yPI = Demx.

Example 6.13 Find a particular integral for

y¢¢ + 7y¢ + 12y = 5e2x.

Step 1: Guess a PI of the same form as H(x) in the differential

equation.

yPI = De2x.

Step 2: Differentiate yPI twice

y¢PI = 2De2x and y¢¢PI = 4De2x.

Step 3: Substitute into the differential equation and solve for D.

(4D + 14D + 12D)e2x = 5e2x.

Because we have guessed a solution of the right form the factor

e2x cancels, leaving an equation for D. Therefore

1
=
30D 5=
or D
6
220  •  Mathematical Physics

Step 4: Write out the PI

1
y PI = e 2 x .
6

As with Case 1, sometimes this method will fail. The left hand
side equals zero if emx is a solution of the homogeneous equation (i.e.
the solution to y¢¢ + 7y¢ + 12y = 0, found by solving the auxiliary
equation), in which case we cannot solve for D.
Key point: If using yPI = Demx makes the left hand side = 0, then
the next guess is yPI = Dxemx. If this fails in the same way, the correct
approach is to multiply through by x and try again. So, next we would
try yPI = Dx2emx, and so on until the method works.

Exercise 6.11 Find a PI for each of the differential equations

(i) y¢¢ – 2y¢ + 3y = e2x,
(ii) y¢¢ + 2y¢ – 3y = ex.

Case 3 Find a PI for y¢¢+fy¢+gy = H(x) where H(x) = (a + bx)emx,

and a, b and m are constants.
Method Try a PI of the form yPI = (C + Dx)emx.
Again the method may fail (if substituting yPI = Cemx into the LHS
gives 0). The correct guess in that case is then to try yPI = (Cx + Dx2)emx.
Case 4 Find a PI for y¢¢+ fy¢ + gy = H(x) where H(x) = a cos(mx) +
b sin(mx), and a, b and m are constants.
Method Try a PI yPI = C cos(mx) + D sin(mx). (In Chapter 8 we
shall look at a more efficient method.)

Example 6.14 Find a PI for yʹʹ + 7yʹ + 12y = −3 sin(3x).

Step 1: Choose an appropriate form for yPI based on the form
of H (x).
yPI = C cos(3x) + D sin(3x).
 Differential Equations 1  •  221

Step 2: Differentiate yPI twice and substitute into the differential

equation
yʹPI = −3C sin(3x) + 3D cos(3x),
yʹʹPI = −9C cos(3x) − 9D sin(3x).

Substituting into the differential equation we get

(−9C cos(3x) − 9D sin(3x)),

+7(3D cos(3x) − 3C sin(3x)),
+12(C cos(3x) + D sin(3x)) = −3sin(3x).

Step 3: Since the equation holds for all x we can equate

coefficients of cos(x) and sin(x)
−9C + 21D + 12C = 0,
−9D − 21C + 12D = −3.

Therefore C = −7D from the first equation, so 150D = −3 from

the second. From which
1 7
=D = , C .
50 50

Step 4: Write out the PI

7 1
=y cos ( 3 x ) − sin ( 3 x ) .
50 50

Note that even though H(x) in the example above involves only
sin(3x) (and with integer coefficient −3) we find that the PI involves
both cos(3x) and sin(3x) (and unwieldy fractions).

Exercise 6.12 Find a PI of the differential equations

(i) yʹʹ + 7yʹ + 12y = −150 cos(3x)
(ii) yʹʹ − 2yʹ + 3y = 34 sin(2x).
222  •  Mathematical Physics

Case 5 Find a PI for y¢¢ + fy¢ + gy = H(x) where H(x) =(sum of

functions of different types).
Method We solve the functions separately (as per the examples
above) and add the solutions together.

Example 6.15 Find a PI for

y¢¢ + 7y¢ + 12y = 5e2x + 24x + 26.

We have already found PIs for when the right hand side is
24x + 26 and 5e2x. The solutions have been found in Examples
6.12 and 6.13. The required PI is therefore the sum of the two
yPI solutions for those examples
1 2x
yPI = 2 x + 1 + e
6

Exercise 6.13 Find a particular integral for each of the following:

(i) y¢¢ + y = 2ex + x + 14
(ii) y¢¢ + 3y¢ + 2y = sin(4x)
(iii) y¢¢ − 2y¢ − 3y = 2e−x.

It is often useful to think of the above differential equations as

determining the response of a physical system, described by y(x) (or
y(t)), to a “driving force” given by the known functions of x or t on
the right hand side. The rules show that an oscillating force produces
an oscillating response, an exponential force causes an exponential
response, and so on.

6.10.  INHOMOGENEOUS LINEAR EQUATIONS

In Section 6.9, we found how to obtain a PI of an equation of

the form
y¢¢ + fy¢ + gy = H(x),  (6.16)
 Differential Equations 1  •  223

with f and g constants and H(x) a known function of x. In Sections

6.7 and 6.8, we saw how to obtain the general solution of equations
of the form
y¢ + gy = 0,  (6.17)
and
y¢¢ + fy¢ + gy = 0.  (6.18)

Key point: The general solution of an inhomogeneous differen-

tial equation has two components:

(i) Solving the homogeneous part (i.e. H(x) = 0) gives the

complementary function (CF), yCF (x).
(ii) Solving the inhomogeneous term (i.e. H(x) ≠ 0) gives the
particular integral (PI), yPI (x).
The full general solution of the equation

y¢¢ + fy¢ + gy = H(x)

is given by the sum of the complementary function and the particu-

lar integral
yGS = yCF + yPI.

Let us now take a look at some examples of how to do this, i.e.

solving the homogeneous part of the differential equation and add-
ing it to the particular integral to find the general solution.

Example 6.16 Find the general solution of

y¢¢ + 7y¢ + 12y = 24x + 26

Step 1: Find the complementary function (from Exercise 6.7)

yCF = Ae−4x + Be−3x.

Step 2: Find the particular integral (from Example 6.12)

yPI = 2x + 1.
224  •  Mathematical Physics

Step 3: Write down the general solution. The general solution

is the sum of the CF and the PI, and hence we can write yGS as

yGS = yCF + yPI = Ae−4x + Be−3x + 2x + 1.

Here is a more detailed example.

Example 6.17 Find the general solution (CF+PI) of

y¢¢ + 4y¢ + 4y = 2e−2x.
Step 1: Find the CF by solving the auxiliary equation. Begin
with a guess
y = Aeλx,
∴y¢ = Aλeλx and y¢¢ = Aλ2eλx,
λ2 + 4λ + 4 = 0.
Now we have the auxiliary equation we solve it
(λ + 2)2 = 0,
λ = −2.
The auxiliary equation has repeated roots. We know that for the
case of repeated roots we can write down the solution directly
as yCF = (Ax + B)eλx. This gives us the complementary function

yCF = (Ax + B)e−2x.

Step 2: Find the PI by guessing a solution of the same form as

the RHS. If we try y = Ce−2x for a PI we will get zero when we
substitute yPI, y¢PI and y¢¢PI back into the differential equation.
This arises because Ce−2x is part of the CF so it cannot be a PI.
As we have a repeated roots example, this will also be true if we
multiply by x, as Cxe−2x is also part of the CF. So we multiply by
x again and try
yPI = Cx2e−2x,
y¢PI = 2Cxe−2x − 2Cx2e−2x,
y¢¢PI = 2Ce−2x − 4Cxe−2x − 4Cxe−2x + 4Cx2e−2x.
 Differential Equations 1  •  225

We now need to find the constant, C. We substitute yPI y¢PI and

y¢¢PI into the equation to be solved, collect the coefficients of x
and solve for C

y¢¢ + 4y¢ + 4y = 2Ce−2x − 8Cxe−2x + 4Cx2e−2x

+ 4(2Cxe−2x − 2Cx2e−2x) + 4Cx2e−2x,
= 2e−2x.
Collecting terms, first coefficients of x2, then x1, then x0

4C − 8C + 4C = 0,
−8C + 8C = 0,
2C = 2.

Therefore C = 1, and we have the particular integral

yPI = x2e−2x.

Step 3: Write down the general solution, which is the sum of the
CF and PI

yGS = (x2 + Ax + B) e−2x.

Note the strategy used above: we proceed systematically until we find

something that works (i.e. C ≠ 0) for our particular integral. But be sen-
sible: if things start to get really complicated it is probably due to having
made a mistake early on.
Here is an example from electrical circuit theory:

Example 6.18 The equation

dq
R +q C =
V0
dt
226  •  Mathematical Physics

describes the time dependent response of the current and

the charge in an electical circuit comprising a resistance R,
a capacitor C and a battery, voltage V0. Since this is a linear
equation with constant coefficients we look for a CF of the
form qCF(t) = aept with a a constant. This gives

Rp + 1/C = 0

hence p = −1/RC. For the PI we try qPI (t) = A. This gives

A = CV0. Putting this together, the general solution is
q(t) = ae−t/RC +V0 /C.

Exercise 6.14 Find the general solution of each of the following:

(i) y¢¢ − y = x
(ii) y¢¢ + 4y¢ + 3y = e−x
(iii) y¢¢ + 2y¢ + y = xe−x.

Exercise 6.15 This is another example from electrical circuit the-

ory. If we apply an alternating voltage to a circuit with a resistance
R and capacitance C the charge on the capacitor is governed by
dq
R V0 cos (w t ) .
+q C= (6.19)
dt 

Find the general solution for q(t).

6.11.  INTEGRATING FACTOR METHOD

In this section, we will introduce the integrating factor method

for solving any first order linear differential equations.
We start with any first order differential equation of the form
y¢ + P(x)y = Q(x),  (6.20)

where P(x) and Q(x) are known functions of x as in equation (6.10).

We define the integrating factor as R(x) = e∫ P(x) dx, and will use this to
 Differential Equations 1  •  227

solve our differential equation. First we will look at a worked exam-

ple, and then discuss how the method works.

Example 6.19 The equation xy¢ − 2y = x2 is first order and

linear. Find its general solution.
Step 1: First we divide through by the coefficient of y¢ so that
we have our equation in the correct form
2 2
y′ − x , so P ( x ) =
y= − .
x x

Step 2: Next, we use P(x) to find the integrating factor, R(x)

2 1
∫ − x dx e −2 ∫ x dx ,
( x ) e ∫= e=
P ( x )dx
=
R

therefore
ln ( x −2 ) 1
=
R ( x ) e=
−2ln ( x )
e= .
x2

Step 3: Now, we multiply through by the integrating factor R(x)

1 2 1
2
y′ − 3 y = .
x x x

Step 4: We note that the left hand side is the derivative of a

d  1 
product, namely  y  , and hence we can write
dx  x 2 
d  1  1
 y= .
dx  x 2  x

Step 5: Finally, we integrate both sides to obtain the general

solution. Do not forget the constant of integration, and then
rearrange for yGS
1 1
2
y = ∫ dx,
x x
= ln ( x ) + C,

and therefore yGS = x2 ln(x) + Cx2.

228  •  Mathematical Physics

The method works for every first order linear equation. The key to
the method is remembering how to find the integrating factor.
Proof: We need to verify that multiplying through by the integrating
factor R(x) leads to a left hand side that can be put in the form of the
derivative of a product. According to the rule to be proved, we first
take the integrating factor

R(x) = e ∫ P(x) dx.

We then need to take the derivative of R(x). However, R(x) is a

composite of two functions h(x) and g(x), say, so we need to use the
chain rule to find R¢(x). So if we let

h(g(x)) = eg(x) where g(x) = ∫ P(x) dx,

then we can write the derivative of R(x) = h(g(x)) as

dh dg
R′( x ) = ,
dg dx
 ∫ P ( x )dx  ⋅  d P x dx  =
R′( x ) =
e   ∫ ( )  R ( x ) P ( x ).
   dx 
So we have found that the derivative of the integrating factor is
simply the product of the integrating factor itself and the coefficient
of y in equation (6.20).
Now let us multiply equation (6.20) by R(x).

R(x)y¢ + R(x)P(x)y = R(x)Q(x).

We notice that (using the product rule)

d
dx
(R=
( x ) y ) R ( x ) y ′ + R′( x ) y,
= R ( x ) y ′ + R ( x ) P ( x ) y.

This is just the left hand side of equation (6.20) multiplied by

the integrating factor R(x), as required. Note that the left hand side
d
becomes
dx
( R ( x) y) every time, so there is no need to work this
 Differential Equations 1  •  229

out each time. You can use the result directly as in the following
example.

Example 6.20 Find the general solution of x2y¢ + xy = 1/x.

Step 1: Divide by the coefficient of y¢ (which in this case is x2)
1 1
y′ + y= .
x x3
Step 2: Find the integrating factor
1
∫x dx
( x ) e=
R= ln ( x )
e= x.

Step 3: Multiply through by R(x)

1
xy′ + y = .
x2
Step 4: Recall that the left side is the derivative of the product
R(x)y
d 1
( xy) = .
dx x2

Step 5: Integrate both sides, and rearrange to find the general

solution yGS
1
xy =− + c,
x
1 c
∴ yGS = − + .
x2 x

Exercise 6.16 Solve by this method

(i) y¢ + y = 2
(ii) x−1y¢ − 2y = 1
 230  •  Mathematical Physics

6.12.  EXTENSION: SPECIAL FUNCTIONS

In this section, we look briefly at a way to derive solutions of gen-

eral linear equations with coefficients that are not constant, but instead
functions of the independent variable. An example would be
d2 y dy
x2
dx 2
dx
( )
+ x + x2 − a 2 y =
0

with a a constant. This introduces the topic of special functions.

We know that the general solution to the equation

y¢¢ + y = 0  (6.21)

is y = a cos(x) + b sin(x). Expanding sin and cos in a Taylor series, we

can express the general solution as a power series:
 x2 x 4   x3 x 5 
y= a1− + + ...  + b  x − + + ...  .
 2 4!   3! 5! 
   
Now, if we had not already known about trigonometric func-
tions, could we have derived these series solutions? To this end, let
us suppose that we seek a solution of (6.21) of the form
∞
y = xn ∑ an x n
n=0

where we have introduced the factor xn because we are assuming

that we do not yet know that the solution can be expanded in integer
powers. By differentiating the series we must have that
∞ ∞

=n 0=
n
n +n − 2
y″=
+y
n 0
∑ ( n + n )( n + n − 1) a x + ∑ a=
nx
n
0.

To find n and the coefficients an we compare powers of x. From

xn−2 we get
v (v − 1)a0 = 0

from which n = 0 or n = 1, since we assume a0 ≠ 0. We take n = 0;

taking n = 1 turns out to give nothing more. With n = 0 we compare
in turn the powers of x0, x1 and xn giving
 Differential Equations 1  •  231

x0 : 2a2 + a0 = 0
x1 : 3 × 2a3 + a1 = 0
xn : (n + 2)(n + 1)an+2 + an = 0
Thus
an an − 2
an + 2 =
− = ...
=
(n + 2)(n + 1) (n + 2)(n + 1)n(n − 1)
 (−1)n / 2 a0
 (n + 2)! n even

= ( n −1 ) / 2
 (−1) a1
n odd
 (n + 2)!

We deduce that

 x2 x 4   x3 x 5 
y= a0 1 − + + ... − a1  x − + + ... ,
 2! 4!   3! 5! 

which is of the form y = a cos(x) + b sin(x), where we could now define

sin(x) and cos(x) by the series.
Of course, defining sin and cos in this way we now have to derive
their other properties, which is a major undertaking. However, for
more general linear differential equations, where we do not already
have known solutions to hand, the approach via solution in series gives
us a way of defining new functions and determining their properties.
The most common functions studied in this way are called special
functions. This characterization clearly leaves open which functions are
included, but we’ll give a snapshot of some of the most common ones.

Bessel functions: The defining equation is

d 2 y dy
x2 2
+ x + ( x 2 − n2 ) y =
0
dx dx
which is satisfied by
∞
( −1)m  x
2m+n

y ( x ) J=
= n ( x) ∑  
m =0 Γ ( m + 1) Γ ( m + n + 1)  2 
.
232  •  Mathematical Physics

If n is not an integer, then J−n is a second independent solution.

If n is an integer, then the second solution involves ln(x) so cannot be
obtained by the series method directly.

A special case occurs when the series terminates (because a coef-

ficient – and hence all succeeding ones – equals zero). In that case we
obtain a polynomial solution rather than a new function, but because
of the importance of these cases, the specific polynomials are given
names.
Hermite polynomials, Hn(x): The defining differential equation is

d 2 Hn
+ (2 n + 1 − x2 )Hn =0
dx 2

which, for integer n, has the polynomial solutions

H0(x) = 1, H1(x) = 2x, Hn+1(x) = 2xHn(x) − 2nHn-1(x).

Laguerre polynomials, Ln(x): The defining differential equation is

d 2 Ln dL
x 2
+ (1 − x ) n + nLn =
0
dx dx
which, for integer n, has the polynomial solutions

2n + 1 − x n
L0 =
1, L1 ( x) =
1 − x, Ln+1 ( x) = Ln ( x ) − Ln−1 ( x ) .
(n + 1) n+1

Legendre polynomials, Pl(x): The defining differential equation is

d 2 Pl d 2 Pl
(1 − x2 ) − 2 x + l (l + 1)Pl ( x) =
0,
dx2 dx
where l is an integer. (The use of l instead of n is conventional, arising
from the association of Legendre polynomials with angular momen-
tum, which is customarily denoted by l in quantum mechanics.) The
polynomial solutions are
 Differential Equations 1  •  233

(2 l − 1) (l − 1)
=
P0 ( x) 1,=
P1 ( x) x=
, Pl ( x) xPl −1 ( x) − Pl − 2 ( x).
l l

Before the advent of computers series solutions were used exten-

sively to elucidate the properties of solutions of differential equations
and to compute numerical solutions. Nowadays, this aspect has been
superseded by more direct numerical methods. Nevertheless, many
of these examples (and others) are of theoretical importance in quan-
tum mechanics and other areas of physics that involve linear differen-
tial equations.

Revision Notes

After completing this chapter you should be able to

• Define the terms order, linear, homogeneous, constant
coefficients with reference to a differential equation
• Explain the terms complementary function (CF) and
particular integral (PI)
• State the number of constants in the general solution
• Find a PI and CF for a linear equation with constant
coefficients by trial solutions of the appropriate form,
and hence find the general solution as PI + CF
• Solve a linear equation of the first order by the
integrating factor method

Exercises and problems on differential equations can be found

at the end of Chapter 8.
 CHAPTER

7
COMPLEX NUMBERS

Complex numbers, which involve the square root of −1, are an

essential tool in physics and mathematics. It is necessary to use com-
plex numbers in many areas of study, from electronics, to quantum
mechanics, to economics. They arise most directly from studying the
solution to polynomial equations, but they allow us to see deep con-
nections between algebra and geometry, and between the exponen-
tial, trigonometric and hyperbolic functions. In this chapter, we will
introduce you to some of the fundamental results and applications
of complex numbers.

7.1.  INTRODUCTION OF COMPLEX NUMBERS

Within the system of integers it is not possible to divide certain

(in fact, most) pairs of numbers (because the result m/n is usually not
an integer for m and n integers). We therefore introduce the system
of rational numbers (fractions). In the system of rational numbers
the quadratic equation x2 = 2 does not have a solution. We therefore
introduce some extra numbers, the irrational numbers, which can-
not be expressed as fractions. The number 2 is defined such that
( )
2
2 = 2 (You may be able to write down any number of more or
less accurate approximations to 2 , as fractions or finite decimals,
but you can’t say what 2 really is except through this definition.)
The two solutions of x2 = 2 are then ± 2 . The irrational numbers
236  •  Mathematical Physics

(including those that, like π but unlike 2, are not solutions of alge-
braic equations with integer coefficients), together with the ratio-
nals, make up the real number system.
Now to the main point. The real number system is not large
enough to allow us to solve every quadratic equation. For example,
the quadratic equation
z2 + 1 = 0
has no real roots (see Figure 7.1). This is the same as saying the
equation
z2 = −1  (7.1)
cannot be solved by a real number. We therefore introduce another
new type of number, call it −1 or i.
The “number” i is defined such that i2 = −1.
You can’t say what i really is except through this definition. The
two solutions of equation (7.1) are then z = ±i. It turns out that the
introduction of i enables us to take the square root of any number,
hence to solve any quadratic equation. (In fact, the introduction of
i is sufficient to solve any polynomial equation, not just quadratics,
as we shall see later.)

f (z)
6
4
2

–2 –1 1 2 z

FIGURE 7.1:  Graph of the quadratic function f (z) = z2 + 1. As this does not cross the z-axis,
there are no real z values for which f (z) = 0; the quadratic equation z2 + 1 = 0 has no real roots.
 Complex Numbers  •  237

Example 7.1 Find the solutions of z2 = −4.

Write z2 as the product of two numbers with known square
roots.

z2 = 4 × (−1).
So,
z =± 4 × ( −1 ) =± 2 × i =± 2 i.

Example 7.2 Find the solutions of z2 − 2z + 5 = 0.

Complete the square to get

z2− 2z + 5 = (z − 1)2 +4 =0,

which is true if

( Z − 1 )2 =
−4
Z − 1 =± 2 i
z= 1 ± 2 i.

An alternative method is, of course, to use the formula for the

roots of a quadratic equation: i.e. if

z =  − ( −2 ) ± ( 4 − 20 )  / 2
 
= 2 ± ( −1 ) × 16  / 2
 
= 1 ± 2i
238  •  Mathematical Physics

Exercise 7.1 Find, in the form x + yi, all solutions of the quadratic
equations
(i) z2 − 6z + 25 = 0
(ii) 2z2 + 2z + 5 = 0.

An expression of the form x + yi is called a complex number. The

complex number x + i0 is the same as the real number x. The com-
plex number 0 + yi is called pure imaginary and is written as yi. For
a complex number z = x + yi, the real number x is called the “real
part of z,” while the real number y is called the “imaginary part of z.”
Note that we refer to the imaginary part as y and not iy. This way of
writing complex numbers – as x + yi – is called the standard form.
If z = x + yi we write
x = Re(z),

(or sometimes x = ℜ ( z ) , with an “R” in the Fraktur typeface), the

real part of z. And we write
y = Im(z)

(or sometimes y = ℑ ( z ) , with an “I” in the Fraktur typeface), the

imaginary part of z.
Two complex numbers z1 = x1 + y1i and z2 = x2 + y2i are equal if,
and only if, their real and imaginary parts are equal (i.e. if x1 = x2 and
y1 = y2). This is called equating real and imaginary parts.

Example 7.3 What are u and v (assumed real) if u2 + 2uvi =

25 + 30i?
Equate the real and imaginary parts

u2 = 25 ⇒ u = ±5
2uv = 30 ⇒ v = ±3.
Put only the corresponding values together

u = 5, v = 3  or  u = −5, v = −3.

 Complex Numbers  •  239

7.2.  OPERATIONS ON COMPLEX NUMBERS

The result of any combination of algebraic operations on com-

plex numbers is a complex number because it can always be reduced
to an expression of the form x + yi.

Arithmetic
Let z1 = x1 + y1i and z2 = x2 + y2i be two complex numbers. In the
following list we put the result of an algebraic operation on z1 and z2
in standard form (x + yi).

1. Addition: z1 + z2 = (x1 + x2) + (y1 + y2)i

2. Subtraction: z1 − z2 = (x1 − x2) + (y1 − y2)i
3. Negative: −z1 = −x1 − y1i = −x1 + (−y1)i
4. Multiplication: z1 × z2 = (x1 + y1i)(x2 + y2i) = (x1x2 − y1y2) +
(x1y2 + y1x2)i

The basic rule of complex algebra is: treat complex numbers as if

real but replace i2 by −1.
There is no need to memorize the formula for (x1 + y1i)(x2 +
y2i), since it comes immediately from multiplying out (to give x1x2 +
x1y2i + y1x2i + y1y2i2) putting i2 = −1, and collecting real and imagi-
nary parts.

Exercise 7.2 Write in standard form (x + yi) each of

(i) (2 + 3i) + (7 − 6i),
(ii) (9 − 3i) − (7 − 6i),
(iii) (2 + 3i)(7 − 6i).
Exercise 7.3 Work out [z − (2 + 3i)] × [z − (2 − 3i)] and hence
find the solutions of z2 − 4z + 13 = 0.

Example 7.4 By working out the real and imaginary parts of

both sides, show that z1z2 = z2z1 (so complex numbers may be
multiplied in any order). We say that multiplication of complex
numbers is “commutative.”
240  •  Mathematical Physics

Let z1 = x1 + y1i , z2 = x2 + y2i. Then

z1z2 = (x1 + y1i)(x2 + y2i).

Multiply out and put i2 = −1 to get

= x1x2 – y1y2 + (x1y2 + y1x2)i.

Interchange the order, since reals can be multiplied in any

order.
= x2 x1 − y2y1 + (x2y1 + y2x1)i

= z2z1.

The final equality follows by symmetry: there is no need to

multiply out z2z1 explicitly because it can be obtained from z1z2
by interchanging 1 and 2 throughout.

Exercise 7.4 By working out the real and imaginary parts of both
sides, show that z3(z1 + z2) = z3z1 + z3z2 (multiplication of complex
numbers is “distributive”).

Complex Conjugate
If z = x + yi we define the complex conjugate of z to be

z* = x − yi.  (7.2)

Thus, to obtain the complex conjugate of a complex number,

simply replace i by −i. Other notations you may come across are z
and z†. The complex conjugate of a complex number is an important
concept that occurs frequently.
 Complex Numbers  •  241

Example 7.5 Verify that if z = x + yi, then zz* = x2 + y2. Use

equation (7.2) to write the complex conjugate, and then
multiply out, putting i2 = −1.

zz* = (x + yi)(x − yi)

= x2 − (yi)2 + xyi − xyi

= x2 − (y)2(−1)

= x2 + y2.

We see that a complex number multiplied by its complex conju-

gate gives a (positive) real number. The complex conjugate of a real
number x is simply x (real).
Exercise 7.5 By considering the real and imaginary parts show
that
(i) ( z1 + z2 ) =
∗
z1∗ + z2∗ ,
(ii) ( z1 z2 ) = z1∗ z2∗ .
∗

These are useful results: the conjugate of a sum of two complex

numbers is the sum of their conjugates; the conjugate of the prod-
uct of two complex numbers is the product of the two conjugates.

Exercise 7.6 Write each of the following in standard form

(x + yi), and find their complex conjugates.
(i) (1−i)4
3
 1 3 
(ii)  − + i
 2 2 
3
 1 3 
(iii)  − + i
 2 2 
242  •  Mathematical Physics

Division
The procedure for division of two complex numbers is similar to
that for multiplication. We write the complex numbers in standard
form, z1 = x1 + y1i and z2 = x2 + y2i, take the ratio and manipulate the
result into a simpler form.

z1 x1 + y1 i x1 + y1 i x2 − y2 i
= = × .
z2 x2 + y2 i x2 + y2 i x2 − y2 i

∗
The trick is to multiply top and bottom by z= 2 x2 − iy2 , as in the
expression on the right, which will give an expression with a real num-
ber as the denominator

x1 x2 + y1 y2 y1 x2 − x1 y2
= + 2 2 i
x22 + y22 x2 + y2

provided z2 ≠ 0. The method is important. It enables you to write a

complex fraction in the standard form x+yi. Rather than remember-
ing this formula it is better to derive the result directly in each particu-
lar case as shown next.

5+ i
Example 7.6 Express in the standard form x + yi.
3 + 2i
We multiply top and bottom by the complex conjugate of the
bottom

5 + i 5 + i 3 − 2i
= ×
3 + 2i 3 + 2i 3 − 2i
15 + 3 i − 10 i − 2 i2
=
32 + 2 2
17 − 7 i 17 7
= = − i.
13 13 13
 Complex Numbers  •  243

2 − 5i
Exercise 7.7 Write in the standard form x + yi.
2 + 3i

( z1 / z2 ) z1* / z2* if z2 ≠ 0.
*
Exercise 7.8 Show that=

Now we know how to multiply and divide complex numbers we

can compute the inverse of a complex number. If z ≠ 0, then

1 1 z* x  −y 
= = +  2 2  i. (7.3)
*
z z z x +y  x +y 
2 2

Hence every complex number z (≠ 0) has an inverse. From this
it follows that if z1z2 = 0, then z1 = 0 or z2 = 0. (Proof: suppose z1 ≠ 0;
then we can multiply through by 1/z1 to get z2 = 0.)

Exercise 7.9 Verify that the product of a complex number with

its inverse is unity: (1/z) × z = 1.

Now we can summarize the rules of algebra for complex

numbers.

1. Addition: z1 + z2 = (x1 + x2) + (y1 + y2)i

2. Subtraction: z1 − z2 = (x1 − x2) + (y1 − y2)i
3. Multiplication by real number: az1 = ax1 + ay1i
4. Multiplication: z1 × z2 = (x1x2 − y1y2) + (x1y2 + y1x2)i
5. Conjugation: z* = x − yi.

z1 z1 z2* x1 x2 + y1 y2  y1 x2 − x1 y2 
6. Division:= = + 2 2  i
z2 z2 z2* x22 + y22  x2 + y2 

1 1 z* x  −y 
7. Inverse:= = + 2 i .
z z z x + y  x + y2 
* 2 2
244  •  Mathematical Physics

7.3.  QUADRATIC EQUATIONS

We have seen how to solve any quadratic equation, az2 + bz + c = 0,

with real coefficients, a, b, and c, by means of the formula

−b ± (b 2
− 4 ac )
(7.4)
z= , 
2a

and if 4ac > b2 you use i to take the square root of the negative number.
Suppose, however, that one or more of the coefficients a, b or
c is complex. We can still use this formula. But is the square root of
a complex number a complex number, or do we have to introduce
further new numbers into our scheme? Let us start with the simplest
example, the square root of i itself.

Example 7.7 Solve z2 = i (i.e. find the square root of i).

Let z = x + yi with x and y real (assuming that the result will be
a complex number)

( x + yi )
2
z=
2

= x2 − y2 + 2 xyi = i

Equate real and imaginary parts

x2 − y2 = 0,  and  2xy = 1

These simultaneous equations have two real solutions,

x=
+ 1 / 2, and y =
+ 1 / 2, or
x=
− 1 / 2, and y =
− 1 / 2.
Therefore
1 i 1 i
z =+ or z=
− −
2 2 2 2
 Complex Numbers  •  245

We have therefore shown that the square root of i is another

complex number, not something new. Taking the square root of a
general complex number is more difficult. The following shows two
ways of doing it; we shall give a systematic method for finding all
roots later on (and they will all turn out to be complex numbers).

Example 7.8 Solve z2 = −5 + 12i (i.e. find the square root of

−5 + 12i).
Let z = x + yi, with x and y real

z2 = (x + yi)2
= x2 − y2 + 2xyi = −5 + 12i
Equate real and imaginary parts and solve for x and y

x2 − y2 = −5  and  2xy = 12

∴ y = 6/x and x2 − 36/x2 = −5

Multiply through by x2 and factorize the resulting quadratic for x2

x4 + 5x2 − 36 = 0,
(x2 + 9)(x2 − 4) = 0.
∴ x2 = −9 or x2 = 4
x2 = −9 is impossible for x real. Hence x = +2, and y = 6/x = +3,
or x = −2 and y = 6/x = −3. Therefore

z = 2 + 3i or z = −2 − 3i.

Alternative Solution: Proceed as before to find x2 − y2 and xy

by equating real and imaginary parts.
x2 − y2 = −5 and 2xy = 12
We can solve these by a trick which avoids the need to solve
a quadratic equation. The trick is to find x2 + y2 in terms of the
already known x2 − y2 and xy.
246  •  Mathematical Physics

(x + y2 ) =( x2 − y2 ) + ( 2 xy )
2 2 2 2

( −5 ) + (12 ) =
=
2
169
2

(x 2
+ y2 )= 169= 13

Now we can solve for x2 or y2. Adding the above equation to

(x2 − y2) = −5 we get
2x2 = 13 − 5 = 8, ∴ x = ±2
And so, using 2xy = 12 we get
y = 6/x = ±3
Putting this all together
z = 2 + 3i or z = −2 − 3i.

Exercise 7.10 Use one of these methods to find the square root
of 5 + 12i.
Since we now know how to find the square roots of a complex
number we can now solve any quadratic by using the standard for-
mula. We know az2 + bz + c = 0 has roots

−b ± (b2
− 4 ac )
z= .  (7.5)
2a

Example 7.9 Solve z2 − (2 − i)z + (2 − 4i) = 0.

The solution of the quadratic equation is

(2 − i) ± (2 − i) − 4 (1 ) ( 2 − 4 i )
2

z=
2
( 2 − i ) ± ( −5 + 12 i )
1/ 2

= .
2
 Complex Numbers  •  247

To obtain (−5 + 12i)1/2 we have to solve w2 = −5 + 12i (see

example 7.8):
(−5 + 12i)1/2 = ±(2 + 3i).

Therefore
( 2 − i) ± ( 2 + 3i)
z=
2
z=
2 + i or z=
− 2 i.

Exercise 7.11 Deduce from the solution of Example 7.9 that

z2 − (2 − i)z + (2 − 4i) = (z − (2 + i))(z + 2i)
and verify the result by multiplying out the right side.

Exercise 7.12 Solve the quadratic equations

(i) z2 − (2 + 2i)z + (2i − 16) = 0
(ii) z2 − (4 + i)z + (5 + 5i) = 0.

7.4.  THE ARGAND DIAGRAM

The complex number z = x+yi can be represented by the point

(x, y) in the plane. This graphical representation of complex num-
bers is called the Argand diagram. Essentially, we use the real and
imaginary parts of a complex number as coordinates. See Figure 7.2.

Example 7.10 Indicate the position of 3 − 4i in the Argand

diagram.
This will be the point (3, −4) shown in Figure 7.2.
248  •  Mathematical Physics

y y
z1+ z2

4
z1
2

2
–4 –2 2 4 x
z2
–2

2 4 x
–4

z = 3 – 4i

(a) (b)
FIGURE 7.2:  (a) Argand diagram for the complex number z = 3 − 4i. (b) Argand diagram
showing the addition of two complex numbers z1 + z2. Note that the real part is along the
horizontal axis (x = Re(z)), the imaginary part is along the vertical axis (y = Im(z)).

Figure 7.2(b) shows an Argand diagram for the sum of two com-
plex numbers. The real and imaginary parts each add to form a third
complex number, just as the components of (Cartesian) vectors add
in two-dimensional space. Figure 7.3(a) shows an Argand diagram
for a complex number and its conjugate. When we form the con-
jugate we swap the sign of the imaginary part, which is a reflec-
tion about the x-axis in the Argand diagram. Figure 7.3(b) shows
an Argand diagram for a complex number z and the same number
multiplied by i repeatedly: z, zi, −z and −zi. Each multiplication by i
rotates the point in the Argand diagram by 90 degrees (π/2 radians)
about the origin, in the counterclockwise direction. Clearly there is
a close connection between complex numbers and two-dimensional
geometry, especially rotations and reflections.

Exercise 7.13 Indicate the positions in the Argand diagram of the

following complex numbers. (You will need to rewrite the complex
numbers in the standard form x + yi; refer to Exercise 7.6.)
(i) 2 + 5i
(ii) (1 − i)4
2
(iii)  − 1 + 3 i 
 2 2 
3
 
(iv)  − 1 + 3 i 
 2 2 
 Complex Numbers  •  249

y y
z z

4
iz
2

2
−−4 –2 2 4 x −−4 –2 2 4 x
–2

–2
–iz
−−4

−−4
z* –z

(a) (b)
FIGURE 7.3:  (a) Argand diagram for the complex number z = 2 + 4i and its conjugate z* =
2- 4i. (b) Argand diagram showing the complex number z = 2 + 4i and the complex numbers
obtained by repeated multiplication by i: z, zi, −z and −zi.

7.5. MODULUS AND ARGUMENT

OF A COMPLEX NUMBER

The modulus of the complex number z = (x + yi) is

|z| = (x2 + y2)1/2.  (7.6)

The modulus of z is the distance of (x, y) from the origin in the
Argand diagram. Alternatively, we have

|z| =(z*z)1/2,  (7.7)

from Example 7.5, and hence

|z|2 = z*z.  (7.8)

This can be very useful in manipulations.

Example 7.11 Show that |z1z2| = |z1||z2|.

This can be shown by putting z1 = x1 +y1i, z2 = x2+y2i. However,
the result involves only the complex numbers and not their
250  •  Mathematical Physics

real and imaginary parts. Therefore we use our alternative

expression for |z|, namely |z|2 = zz*. (In this form we avoid the
awkward square root.)
( z1 z2 )( z1 z2 )
2 *
z1 z2 =
= ( z1 z2 ) ( z1* z2* )
= z1 z1* z2 z2*
2 2
= z1 z2 .

Therefore

|z1z2| = |z1||z2|.

Exercise 7.14 Show that, for any complex numbers z1 and z2,
z1 z
= 1 .
z2 z2

Exercise 7.15 Find | z | if

(i) z =
− 3 + i,

( )
2
(ii) z =− 3 + i ,
1 + 3i
(iii) z = .
1+ i

If the complex number z ≠ 0 is represented by the point

P = (x, y) in the Argand diagram, then the argument of z, arg(z),
is the angle from the x-axis to the line OP joining point P with the
origin O. See Figure 7.4:
 y
arg ( z ) = tan −1   . (7.9)
 x  
 Complex Numbers  •  251

y
z
y z

π/4 + 2π
r π/4
x
θ −3π/4

x
−z

(a) (b)
FIGURE 7.4:   (a) A complex number z in the Argand diagram, showing its modulus and
argument (r, q) and its real and imaginary parts (x, y), which are polar coordinates and Car-
tesian coordinates in the Argand diagram, respectively. (b) The Argand diagram for z = 1 + i
and − z = −1 − i. The argument is defined up to multiples of 2π. The argument of z is π/4 or
π/4 + 2π = 9π/4 etc.; the argument of −z is − 3π/4 or −3π/4 + 2π = 5π/4 etc.

This formula needs to be used carefully to obtain a correct angle

from the inverse tan. For example, arg(1 + i) = tan–1(1) = π/4, but
arg(−1 − i) = tan–1(1) = −3π/4. (See Figure 7.4(b).) Note also that
arg(z) is many-valued: we can add arbitrary multiples of 2π and get
the same argument. If z = 0 then arg(z) is not defined.

Exercise 7.16 Find the modulus and argument of 1 − i and

− 3 − i.

From Figure 7.4(a) we see that (|z| , arg(z)) are the polar coor-
dinates of z in the Argand diagram. So write r = |z|, 𝜃 = arg(z). Then
x = r cos(𝜃), y = r sin(𝜃) and we get the polar representation of a
complex number

z = r(cos q + i sin q).  (7.10)

This is important. We can use the polar (r, q) or Cartesian (x, y)

representation, or both, as convenient.
252  •  Mathematical Physics

Example 7.12 Write z = 1 + i in the form r(cos θ + i sin θ).

We know that r = | z |, so compute this first.

(1 + 1 )
z=
1/ 2
=2.

Then write the given z in the form 2 × () ; the expression in

parentheses must be (cos(q) + i sin(q)). Alternatively use the
formula for the argument (equation 7.9) q = tan-1(1).

 1 i 
=z 2 + ,
 2 2
 p  p 
= 2  cos   + i sin    .
 4  4 

Exercise 7.17 Indicate the positions of the complex numbers i,

1 − i 3 , −1 − i 3 on the Argand diagram and write each of them
in the form r(cos(q) + i sin(q)).

7.6.  THE COMPLEX EXPONENTIAL

We shall show that for real q:

ei𝜃 = cos(q)+ i sin(q).  (7.11)

This result relates the exponential and trigonometric functions

through complex numbers, and is enormously useful in physics and
mathematics. It is known as Euler’s equation. (The equation is also
true for complex q.)

Proof: We have (from Chapter 2), that

x2 x3 x 4
ex = 1 + x + + + + ....
2! 3! 4!
So, putting x = iq,
q 2 iq 3 q 4
e iq =1 + iq − − + + ....
2! 3! 4!
 Complex Numbers  •  253

We collect the real terms and the imaginary terms together

 q2   q3 
= 1 − + ...  + i  q − + ...  ,
 2!   3 
and notice these are the series expansions of cos(𝜃) and sin(𝜃).

= cos(𝜃) + i sin(𝜃).

1 iq
(q )
Example 7.13 Show that cos=
2
( e + e− iq ) .
e−i𝜃 is obtained from ei𝜃 either by complex conjugation
(replacing i by −i) or by letting q → −q.

eiq = cos(q) + i sin(q),

e−ie = cos(q) − i sin(q).

Hence
eiq + e−ie = 2cos(q)

1 iq
(q )
∴ cos=
2
( e + e− iq )

Exercise 7.18 Show that sin(q) = (eiq − e −iq)/2i.

We therefore have expressions for sin and cos in terms of com-
plex exponentials
1 iq
(q )
cos=
2
( e + e− iq )
1 iq
=
sin (q )
2i
( e − e− iq ) . (7.12)

Exercise 7.19 What, in their simplest forms, are
(i) eiπ/2,
(ii) eiπ,
(iii) e2πni (n = 0, ±1, ±2,...),
(iv) |ei𝜃 |.
254  •  Mathematical Physics

Exercise 7.20 Show that ei𝜃1 = ei𝜃2 if (and only if) q2 = q11+2nnπ
(n = 0, ±1, ±2 ...).

Exercise 7.21 Show that

cosh(q) = cos(iq), sinh(q) = −i sin(iq),

and that
cos(q) = cosh(i), sin(q) = −i sinh(iq).

This explains the similarities (and differences) between the

identities involving the trigonometric and hyperbolic func-
tions.

Comparing equations (7.11) and (7.10) we see that any complex

number can be written in the form

z = rei𝜃  (7.13)

where r = |z| and q = arg(z). This form is very useful for taking
products and roots of complex numbers as the following examples
show.

Example 7.14 If z = reiq (z ≠ 0), find z −1:

z−1 = (reiq)−1 = r−1e−iq.

Example 7.15 If z1 = r1 eiq1 and z2 = r2eiq2 find z1z2, |z1 z2| and
arg(z1z2).

z1z2 = r1r2eiq1 eiq2 = r1r2eiq1+ iq2= r1r2ei(q1+q2)

So |z1z2| = r1r2 and arg(z1z2) = q1 + q2.

 Complex Numbers  •  255

We see from this example that to form the product of two com-
plex numbers we multiply the moduli and add the arguments.

iq z z
Exercise 7.22 If z1 = r1 e iq and z2 = r2 e 2 (z2 ≠ 0) find 1 , 1
1

  z2 z2
and arg  z1 
 z2 

Example 7.16 If z = reiq find z1/2 and indicate the positions of

the square roots of z on the Argand diagram.

z1/2 = (reiq )1/2 = ±r1/2eiq/2.

See Figure 7.5. (Remember the ±: there are two square roots!)

y z
2

r
1

θ z1/2
θ/2 x
–2 –1 1 2

z1/2 r1/2
–1
–2

FIGURE 7.5:   The Argand diagram showing a complex number z and its two square roots, z1/2.

Exercise 7.23 Let z= 1 + 3 i . By expressing z in the form reiq,

find z1/2 and show your result on the Argand diagram.
256  •  Mathematical Physics

7.7.  DE MOIVRE’S THEOREM

Since (cos(q) + i sin(q))n = (eiq)n = einq = cos(nq) + i sin(nq), we

have another fundamental result, known as de Moivre’s theorem,

(cos(q) + i sin(q))n = cos(nq) + i sin(nq).  (7.14)

valid for n = 0, ±1, ±2, …. This can be used to give a proof of various
complicated-looking identities.

Example 7.17 Show that 4 cos3(q) = cos(3q) + 3 cos(q).

Since the left hand side involves cos3(q) we try putting n = 3 in
de Moivre’s theorem.

(cos(q) + i sin(q))3 = cos(3q) + i sin(3q).

Then multiply out the left side

cos3(q) + 3i cos2(q) sin(q) + 3i2 cos(q) sin2 (q) + i3 sin3 (q)

= cos(3q)+ i sin(3q).(7.5)

Equating real parts (we have no further use for the imaginary
parts)
cos3 (𝜃) − 3cos(q) sin2(q) =cos(3q).
Replace sin2(q) by 1 − cos2(q) and we have

4cos3(q) − 3cos(q) = cos(3q).

Alternative method: We can use the exponential form of
complex numbers to obtain this identity more directly.
From equations (7.12)
3
1 
=
4 cos3
(q ) 4   e iq + e− iq  
2 
1 iq
( e + e− iq ) .
3
=
2
 Complex Numbers  •  257

Now expand this using the binomial theorem

1
4 cos3 (q ) = ( e iq ) + 3 ( e iq ) ( e− iq ) + 3 ( e iq )( e− iq ) + ( e− iq ) 
3 2 2 3

2  
1 3 iq
=  e + 3 e iq + 3 e− iq + e−3 iq 
2
1 3 iq 3
=  e + e−3 iq  +  e iq + e− iq 
2 2
= cos ( 3q ) + 3cos (q ) .

Exercise 7.24 Show that sin5(q) = (sin(5q) − 5sin(3q) + 10 sin(𝜃))/16.

Exercise 7.25 Let Cn stand for the sum 1 + cos(q) + cos(2q) + ...
+ cos[(n − 1) q]. The aim of this question is to find a closed-form
expression for the sum of this series to n terms.
(i) Use de Moivre’s theorem to show that Cn is the real part of
1 + e iq + ( e iq ) +  + ( e iq )
2 n −1
.

(ii) This is a geometric progression. Identify the common ratio

and hence obtain its sum.
 e inq − 1 
(iii) Deduce that Cn is the real part of  iq  . Show that
 e −1 

e inq − 1 e
=
inq / 2
( einq / 2 − e− inq / 2 )
e iq − 1 e iq / 2 ( e iq / 2 − e− iq / 2 )
and hence that
 ( n − 1 )  sin ( nq 2 )
1 + cos (q ) + cos ( 2q ) +  + cos ( ( n − 1 )q ) =
cos  q
 2  sin (q 2 )

This section has shown that one of the many uses of complex
numbers is in obtaining real trigonometrical identities. These can
also be established without the use of complex numbers, but often
not so simply.
258  •  Mathematical Physics

7.8.  THE ROOTS OF UNITY

We are familiar with the idea that the solution to z2 = 1 is z = ±1

so 1 has two square roots. Now that we know about complex numbers
we can solve the more general equation zn = 1. The complete set of n
complex solutions are the nth roots of unity.

Example 7.18 Show that z3 = 1 is solved by z = 1, w, w2 where

w = e2πi/3. (These are the cube roots of unity.)
Begin by noting that z3 = 1 = e2πi = e4πi are three equivalent
ways of writing 1. They give different cube roots. Any other way
(e.g. e6πi) will repeat one of these. So

z = 1, e2πi/3, e4πi/3.

The cube roots are written as 1, 𝜔 and 𝜔2 where w = e2πi/3, and

𝜔2 = e4πi/3.

Exercise 7.26 Show that z4 = 1 is solved by z = 1, 𝜔, 𝜔2, 𝜔3,

where 𝜔 = eπi/2, and that zn = 1 is solved by z = 1, 𝜔, 𝜔2,..., 𝜔n–1
where 𝜔 = e2πi/n.
Plot the cube roots of 1 and the fourth roots of 1 in the Argand dia-
gram and describe where the nth roots lie.

7.9.  ROOTS OF REAL POLYNOMIALS

If P(z) is a polynomial with real coefficients and if z0 is a root, i.e. if

P(z0) = 0, then z0∗ is also a root, i.e. P ( z0∗ ) = 0 . Thus:
The roots of a real polynomial are either real or occur in complex
conjugate pairs

Proof:
Let
P (z) = anzn + an−1zn−1 + ... + a 0,
 Complex Numbers  •  259

where a0, a 1 ,... an are real coefficients. Taking the complex con-
jugate of this equation gives

( P ( z=
)) a∗n ( zn ) + a∗n−1 ( zn−1 ) +  + a0∗ ,
∗ ∗ ∗

since the conjugate of a sum of complex numbers is the sum of

the conjugates of the summands (see Exercise 7.5). Now, noting that
(qm)* = (q*)m for integer m we can write this as
= a∗n ( z∗ ) + a∗n−1 ( z∗ )
n n −1
+  + a0∗ .

And since all the am’s are real ( a∗m = am ) we have

= an(z*)n + an−1(z*)n−1 + ... + a 0

= P (z*)
But P(z0) = 0, so (P(z0))* = 0, and hence P ( z0∗ ) = 0 This means
that z0∗ is a root if z0 is a root.

Exercise 7.27 Why does this proof fail if an, an−1,..., a0 are not
real?

7.10.  ROOTS OF COMPLEX POLYNOMIALS

Remainder theorem: Let Q(z) be a polynomial with complex

coefficients. (Of course, this includes the case when the coefficients
are in fact real.) Then
z0 is a root of Q(z) = 0 if and only if z − z0 is a factor of Q(z).

Proof: If (z − z0) is a factor of Q(z) then Q(z) = (z – z0)P(z) for some

polynomial P(z) and hence Q(z0) = 0. Conversely, if Q(z0) = 0 then
Q(z) = Q(z) − Q(z0) and each pair of terms of the same degree is
divisible by (z − z0).
Fundamental theorem of algebra: There is at least one complex
number, z0 say, such that Q(z0) = 0; i.e. every complex polynomial
has a root.
260  •  Mathematical Physics

This is a remarkable result (and not easy to prove rigorously).

Turn back to example 7.1 to remind yourself that it is not true for
real numbers. But once we have introduced the solution, i, of one
real quadratic equation, z2+1 = 0, we can, in principle, solve all poly-
nomial equations. The “in principle” is important − no explicit for-
mulae exist for general equations higher than quartic, even though
we know the roots always exist.
Exercise 7.28 Deduce that every polynomial has a complete
factorization into linear factors,

Q(z) = an(z − z1) (z − z2) ... (z − zn)

and that every real polynomial has a complete factorization into

some product of real quadratic factors and real linear factors.
Exercise 7.29 Given that z = 1 + i is a root of the equation

z4 − 8z3 + 24z2 − 32z + 20 = 0

find the other three roots.

7.11.  EXTENSION: COMPLEX VARIABLE

We have already met some very simple examples of functions of

a complex variable; z2 and z are examples. If z = x + iy as usual,
we can write z2 = (x2 − y2) + 2xyi, that is in the form f(z) = u(x, y) +
iv(x, y) where u and v are real functions of the real variables x and
y. To extend the calculus from real to complex functions, we need to
be able to differentiate f(z) with respect to z. To see that this restricts
the possible choices of u and v consider the function F obtained by
putting u = x and v = 0. If we now let z vary keeping x constant, we
get dF /= dz lim ( F ( z + d z ) − F ( z ) ) /=
d z 0 if instead we let z vary
d z→0
keeping y constant, we obtain, similarly, dF/dz = 1. To ensure that
the limit is independent of how we approach it, it can be shown that
u and v must satisfy
∂u ∂v ∂u ∂v
= and = − .
∂x ∂y ∂y ∂x
 Complex Numbers  •  261

These are called the Cauchy-Riemann relations. You can check

that they are satisfied by f (z) = z2 but not by F(z) = x. (Informally,
we can see that F is not a function of z only, but of both z and z*
since x = (z + z*)/2.)
Differentiable functions of a complex variable are called analytic
functions. They have some remarkable properties. For example, if
f (z) is an analytic fucntion in some region Ω, it satisfies Cauchy’s
theorem

∫ f ( z ) dz = 0
where the integral is taken around a circle in Ω with center z0.
(To do the integral around a circle of radius R put z = z0 + Reiq, dz =
iReiq dq and integrate with respect to q from 0 to 2π. The theorem
is true for any closed curve, not just circles, but the technique for
evaluating the integral is then more complicated.)
The theorem can be used to prove Cauchy’s integral formula for
an analytic function:

1 f ( z)
f ( z0 ) = ∫
 dz.
2p i z − z0

A function analytic in an annulus,1 A, centered on a point z0, can

be expanded as a Laurent series (the generalization of Taylor series
for analytic functions):
b1 b2
f ( z) =a0 + a1 ( z − z0 ) + a1 ( z − z0 ) +  +
2
+ +
z − z0 ( z − z0 )2

where z lies within the annulus. (If f(z) is analytic within a disc
centered on z0, that includes z0, then the bn will all be zero.) Cauchy’s
formula can be used to show that
1 f ( z) 1 f ( z)
=an = ∫

2p i ( z − z0 ) n +1
, bn ∫

2p i ( z − z0 )− n+1
,

1
An annulus is the region between two concentric circles.
262  •  Mathematical Physics

where the integrals are taken around closed curves within the
annulus A, and that

∫ f ( z ) dz = 2p ib . 
1
(7.16)

The quantity b1 is called the residue of f(z) at z0. Equation (7.16)

can be used to evaluate integrals which are often difficult to obtain
by other means. (The technique is called contour integration.)

Revision Notes

After completing this chapter you should be able to

• Add, subtract, multiply and divide complex numbers
• Find the complex conjugate of an algebraic expression
• Find the real and imaginary parts of a complex algebraic
expression
• Solve quadratic equations with real or complex
coefficients
• Find the argument and modulus of a complex number
and indicate it on the Argand diagram
• Express a complex number in the form
z = r(cos(q) + i sin(q)) = reiq
1 iq
• Recall cos (q ) =
2
( e + e− iq ) , sin (q ) =( eiq − e− iq ) / 2 i
• Quote de Moivre’s theorem and use it to derive
trigonometric identities
• Find the nth roots of unity
 Complex Numbers  •  263

7.12. EXERCISES

1. Write in the form x + iy, where x and y are real, the com-
plex number u/v, where u = 1 + 7i, v = 1+ i, and verify that
(i) |x + iy| = |u| / |v|, (ii) x – iy = u* / v*.
2. Write in the form x + iy, where x and y are real, each of the
complex numbers w = (2 + 3i) / (1 + i), z = 3eiπ/4, and their
conjugates w*, z*.
3. Write the complex number w =−1 + i 3 in the form rei𝜃
where r > 0 and −π < q < π, and indicate its position on the
Argand diagram.
4. Show the positions of the circle |z| = 5 and of the particu-
∗
lar complex number z0 = −3 + 4i and of z0 in the Argand
diagram, and also indicate the approximate positions of the
two square roots of z0.
5. Show the positions of the circle |z| = 5 and of the particular
complex number z0 = 5 − 12i in the Argand diagram, write
down the values of |z0| and tan(arg z0), and indicate the
approximate positions of the two square roots of z0.
6. Indicate the positions of the complex numbers z1 = 1 + i
and z2 = 2 i in the Argand diagram, verify that they are at
an equal distance from the origin, and explain why
arg(z1 + z2) must be 3π/8.
7. Write the complex number z = − 1 + 2i in the form rei𝜃
where r = |z| > 0 and −π < q < π, expressing q in terms of
the number a = tan–1 2. Show that z* = 5 / z. Also indicate
the positions of z, z* and z2 on the Argand diagram.
8. Express each of the complex numbers z1 = 1 − i and z2 = i
in the form rei𝜃 where r > 0,−π < q < π, and indicate their
positions in the Argand diagram. What is z3, if 0, z1, z2, z3
form a parallelogram?
9. Express the complex number (1 + i) / (3 + 4i) in the form
x + iy where x and y are real. Show that if it is written in the
form reiq where r > 0 and−π < q ≤ π then r = 2 / 5 , and
determine q correct to two decimal places.
264  •  Mathematical Physics

10. Given that z = x + iy (where x and y are real) and

z2 = 3 + 4i, find the values of x2 − y2 and xy. Hence show
that X = x2 satisfies the quadratic equation X2 − 3X − 4 = 0
and find the two possible values for x.
11. Find the three cube roots of 2 and indicate their positions
in the Argand diagram.
12. From the binomial expansion of (eiq + e−iq)4 deduce that

8 cos4(q) = cos(4q) + 4 cos(2q) + 3.

7.13. PROBLEMS

1. Solve the quadratic equation z2 − 5iz − 7 + i = 0, expressing

each of the two roots in the form x + iy with real x and y,
and hence factorize the left hand side of the equation.
2. (i) Describe how the successive coefficients in the binomi-
al expansion of (1 + x)n can be calculated; in particular
write down the expansion of (1 + x)5 and hence that of
(u + v)5.
(ii) Derive the formula eiq= cos(q) + i sin(q) from the power
series expansions of eiq, cos(q) and sin(q). Hence show
that (eiq + e−iq) / 2 = cos(q).
(iii)  From the last formula, or otherwise, deduce that
1
cos5 (q )=
16
( cos ( 5q ) + 5cos ( 3q ) + 10 cos (q ) ) .
3. Show that if z = reiq = r cos(q) + ir sin(q) then

Im(ez) = ercos(q) sin(r sin(q))  (7.17)

and also give the real part of ez. By taking the imaginary
part of the expansion of ez in ascending powers of z, deduce
from (7.17) that
r 2 sin ( 2q ) r 3 sin ( 3q )
er cos(q ) sin(r sin (q ) =
r sin (q ) + + + ....
2! 3!
(7.18)
 Complex Numbers  •  265

Regarding q as constant, use (7.18) to evaluate the limit

1
lim+ er cos(q ) sin ( r sin (q ) ) ,
r →0 r

and also verify that the same answer is obtained when

l‘Hôpital’s rule is applied instead.
4. S
 how that if (a + bi)2 = 5 − 12i, where a and b are real,
then a2 − b2 = 5 and a2 + b2 = 13, and find the two square
roots of 5 − 12i. Hence solve the quadratic equation
z2 − 3z + 1 + 3i = 0,
and also factorize the left side. How could it have been eas-
ily noticed that z − i is a factor?
5. Let 𝜃 be an arbitrary real number, and 𝜙 = 𝜃/2.
(i) Verify that 1 + eiq can be written as 2ei𝜙 cos(𝜙).
(ii) By treating the expression
 n  n
Cn = 1 + n cos (q ) +   cos ( 2q ) +   cos ( 3q )
2 3
+ + cos ( nq )

where n is a positive integer, as the real part of a binomial

expansion, show that

Cn = (2 cos(ø))n cos(nø).
2p 4p
(iii) If <q < what is the behavior of Cn as n → ∞?
3 3
6. Verify that if x and y are real numbers with y > 0, and
1  1/ 2
( ) 1  1/ 2
p=
2 
x + x 2
+ y 2 1/ 2 

and =
q
2 
− x + ( x 2
+ y )
2 1/ 2 


then p2 − q2 = x and 2pq = y and therefore p + iq is a square

root of x + iy. Hence show that the two square roots of
266  •  Mathematical Physics

( )
z =−2 + 2 i 3 are ± 1 + i 3 . Verify this by writing z in the
form reiq and finding r1/2eiq/2.
 3 
Solve the quadratic equation z2 + 2 z +  1 − i = 0.
 2 
7. Verify that z = i is a root of the equation
z3 + (2 + i)z2 − (1 + 4i)z − 2 + 3i = 0,
and hence find the other roots. Indicate the positions of the
roots in the Argand diagram.
N −1
8. Evaluate the sum ∑ exp ( 2p ipn / N ) where p and N are
n=0
integers (≥ 1), for the cases when p ≠ 0 and p = 0. By put-
ting p ≠ k – k ′and p = k + k′ where k, k′ are integers and
k ≠ 0, k′ ≠ 0, show that

N
N −1
 2p kn   2p k′n   , k = k′,
 ∑
n=0
cos  cos
  =
 2
 N   N   0 otherwise
(7.19)

N
N −1
 2p kn   2p k′n   , k = k′,
 ∑
n=0
sin   sin  =2
 N   N   0 otherwise
(7.20)

N −1
 2p kn   2p k′n 
∑ sin 
n=0
 cos 
N   N 
 = 0. (7.21)

What are the values of these sums when k = k′ = 0?

∞
zn
9. Given that exp ( z ) = ∑ , where z is a complex num
n= 0 n!
ber, write down the series expansion of exp(xeiq) where x
∞
xn
and q are real. Hence sum the series ∑ cos ( nq ) and
∞ n= 0 n!
xn
∑
n= 0 n!
sin ( nq ) . By equating the coefficients of a suitable
power of x, use your results to express cos (4q) and sin(4q)
as sums of powers of sin(q) and cos(q).
 CHAPTER

8
DIFFERENTIAL
EQUATIONS 2

In Chapter 6, we learned different methods for solving some first

order and second order differential equations. Having worked
through the topic of complex numbers in Chapter 7 we can now see
what happens if we are faced with complex roots in an auxiliary equa-
tion. We will also demonstrate the use of complex numbers to simplify
the solution of certain types of differential equations. First though,
we will show how to constrain the physical problem which a given
differential equation represents, by setting initial and/ or boundary
conditions. By doing this we fix the value of the arbitrary constants
that appear in the general solution of a differential equation.

8.1.  BOUNDARY AND INITIAL CONDITIONS

We have learned in Chapter 6 that the general solution to an nth

order differential equation contains n arbitrary constants. In physi-
cal applications of the differential equation these constants are then
fixed (and found) by imposing realistic conditions on the solution
and / or its derivatives at specified values of the independent variable
(say position x or time t). This can be done if we understand how the
system that is represented by the differential equation behaves at a
given time or position, e.g. at t = 0 or x = 0.
268  •  Mathematical Physics

Example 8.1 Find the solution for y(t) of  y + w2 y =0 such

that at t = 0, y(0) = 0 and y (0) = 0 , where y means dy/ dt and ÿ
means d2y / dt2 and so on. (In a physical application t might be
time and the conditions would represent the given position and
speed at an initial instant.)
First we need to find the general solution. In this particular
case we have found the solution in Section 6.5 of Chapter 6,
with t here replacing x.

yGS = A sin(wt) + B cos(wt).

(This is the equation for simple harmonic motion.) Now we can

apply the given initial conditions. These are specified at t = 0,
so we must find the value of our general solution yGS evaluated
at t = 0.

yGS(0) = A sin(0) + B cos(0) = A × 0 + B × 1 = B.

We are given y(0) = 1 and therefore, B = 1. Now we must

differentiate the general solution to get y GS,

y GS ( t ) Aw cos (w t ) − Bw sin (w t ) .
=

We are given y (0) = 0; therefore at t = 0

y GS ( 0 )= Aw × 1 − B × 0 = Aw= 0,

hence,
A = 0.
Finally, we substitute A = 0 and B = 1 in the general solution.
This is then the required solution which satisfies the differential
equation and also the extra conditions imposed at t = 0 (it is
often useful to check this). Thus
y= A cos(wt).
 Differential Equations 2  •  269

Note that the first step is to obtain the general solution, by any
of the methods that you have learned so far (or will learn in future).
Once you have done this, you can then apply the initial or boundary
conditions.
Key point: When the conditions imposed are values of the solu-
tion and its derivatives for the same value of the independent variable
then these are called initial conditions. This is the case in examples
8.1–8.3 and in Exercise 8.1. The independent variable is often time,
and the conditions are often imposed at t = 0, as in Example 8.1
above. When the conditions are values of the solution for different
values of the independent variable, as in Exercise 8.2, then these are
called boundary conditions.
Let us now look at another example of how to apply initial con-
ditions where we have already previously obtained the general solu-
tion in the exercises in Chapter 6.

Example 8.2 Solve y″ + 7y′ + 12y = 19 + 12x subject to the

initial conditions y(0) = 1 and y′(0) = 2, given the general
solution yGS (x) = Ae−4x + Be−3x + x + 1.
For this example we are able to start from the given general
solution. This is the same as the example already solved in
Section 6.7.
The general solution, yGS, is the sum of the CF and the PI,
  yGS(x) = yCF+ yPI,
  yGS(x) = Ae− 4x+ Be−3x+ x + 1.
We are going to need y′GS, to get y′GS(0), and

y′GS = −4Ae− 4x− 3Be− 3x+ 1.

Now we can apply the initial conditions. We substitute x = 0 in

the general solution and set yGS(0) = 1,

yGS(0) = (A × 1) + (B × 1) + 0 + 1 = A + B + 1.
270  •  Mathematical Physics

Then,

yGS(0) = A + B + 1 = 1 or A = −B.

Now apply the condition that y′GS(0) = 2 in a similar way, and

solve for A and B:

y′GS(0) = (−4A × 1) − (3B × 1) + 1,

= −4A − 3B + 1 = 2.

This gives

−4A − 3B = 1.

From above, A = −B, so B = 1, and A = −1. Finally, we

substitute the values of A and B into the general solution. Thus
our solution is

y = −e−4x+ e−3x+ x + 1.

We are now going to look at a complete example, where we will

begin by finding the general solution of a second order, linear, inho-
mogeneous equation.

Example 8.3 Solve y″− y = ex subject to the conditions

yGS(0) = 0 and y′GS(0) = 1.
First we need to find the general solution. We do this by first
finding the CF when
y″ − y = 0.
As this is linear and homogeneous with constant coefficients
we try

yCF = Aeλx,
 Differential Equations 2  •  271

then

y′CF = λAeλx,
  y″CF = λ2Aeλx.

Next we substitute back into the differential equation, and

form the auxiliary equation

λ2Aeλx − Aeλx = 0.

Then, canceling Aeλx,

λ2 − 1 = 0,
⇒ λ= ±1.

Finally we can write the complementary function

yCF = Aex+ Be−x.

Now, to find a PI our first thought is to try yPI = Cex, but

unfortunately this will not work because it is part of the CF (try
it and see what happens). Recall from Section 6.9 that when
this happens we multiply by x and try again.

yPI = Cxex,
  y′PI = Cex + Cxex,
   y″PI = 2Cex + Cxex.

Substituting this into the differential equation we can solve for

C, and hence find yPI. As y″− y = ex, we have

(2Cex+ Cxex) − (Cxex) = ex,

 2Cex = ex.
272  •  Mathematical Physics

Therefore we find that C = 1/2, and we can write the PI as

follows:
1
yPI = xe x .
2

We can write the general solution as the sum of yCF and yPI:

1 x
yGS =Ae x + Be− x + xe .
2
To apply the initial conditions we also need y′GS so let us derive
that next:

1 x 1 x
′ = Ae x − Be− x +
yGS e + xe .
2 2
We are given that y(0) = 0 and y′ (0) = 1. Substitute x = 0 in yGS,
and then in y′GS:

  yGS(0) = (A × 1) + (B × 1) = A + B = 0,
1 1
′ ( 0 ) = ( A × 1 ) − ( B × 1 ) + × 1 = A − B + = 1.
yGS
2 2

Finally we can solve for A and B and substitute them back into
the general solution, yGS. From above we have A + B = 0 and
A − B = 1/2. Adding these together we get

1 1 1
2A = , leading to A = and= B -=
A - .
2 4 4
Finally, substituting into the GS gives
1 x 1 -x 1 x
y= e - e + xe .
4 4 2
 Differential Equations 2  •  273

Key point: Note that the initial or boundary conditions are

applied to the general solution as a whole, and not just to the CF
or PI separately. The first step in solving a differential equation is
always to find the general solution.
Since the general solution to an nth order equation contains n
arbitrary constants, n initial or boundary conditions are required to
find the solution.

Exercise 8.1 Solve the equation y″ − y= x for y(x), subject to the

initial conditions y= 0 and y″ = 0 at x = 0.

Exercise 8.2 Solve the equation z = -g (where g is a constant)

subject to the following boundary conditions:
(i) z = 5 at t = 1 and z = 4 at t = 2,
(ii) z = 0 at t = 2 and z = 0 at t = 3.

Exercise 8.3 Solve the following for y(x) subject to the specified
conditions. (See Exercise 6.14).
(i)  y″ − y = x subject to y(0) = 1, y′(0) = 2,
(ii)  y″ + 4y′ + 3y = e−x subject to y(0) = 0, y′(0) = 2,
(iii)  y″+ 2y′ + y = xe−x subject to y(0) = 0, y(1) = 1/e.

In the examples that we have looked at so far, we have applied

initial or boundary conditions to second order differential equations.
The method we have discussed here however applies to any type
of differential equation, and is the method to follow in general. In
the next example, we shall see how to apply the conditions to a first
order differential equation, where we will begin by using the inte-
grating factor method to find the general solution.
274  •  Mathematical Physics

2
Example 8.4 Solve y¢ + y = 1 subject to y= 0 at x = 1.
x
The equation is first order but not separable, and does not
have constant coefficients, so we will use the integrating factor
method here to find the general solution. We begin by writing
down the integrating factor, R(x).
æ 2 ö
R ( x) = exp ççç ò dx)÷÷÷,
è x ø
2 ln( x) ln( x2 )
= e= e= x2 .

We then multiply the differential equation by the integrating

factor, R(x),
2
x 2 y¢ + x 2 y = x2 ,
x
and recognize that the left hand side is equivalent to the
derivative of the product (R(x) × y) (see Section 6.11):

(x2y)′= x2.
We now integrate both sides to obtain our general solution,
yGS(x). We have
x3
x2 y = + C,
3
so that the general solution, which contains one arbitrary con-
stant, C, is
x C
y= + 2 .
3 x
We can now impose the boundary condition: at x = 1,
yGS (1) = 0, so
1 C
GS (1)
y= =
+ 0
3 12

1
and C = - .
and
3
 Differential Equations 2  •  275

The solution to this differential equation for the given boundary

condition is therefore
x 1 x3 - 1
=y =- 2 .
3 3x 3 x2

Exercise 8.4 Solve y′ − 2xy = x subject to y= 0 at x = 0.

We are often interested in how physical systems behave for very

large values of the independent variable. In such circumstances we
might set the boundary conditions at ±∞, where our general solu-
tion approaches a given value(its asymptotic value). Let us look at an
example of how this is done.

y 1
Example 8.5 Find the solution of y¢ - = 2 such that
y(x) → 0 as x → ∞. x x

The equation is first order but not separable, and does not have
constant coefficients, so we will again use the integrating factor
method to solve the equation. We begin by writing down the
integrating factor, R(x).
æ -1 ÷ö
R ( x) = exp ççç ò dx÷÷,
è x ø
- ln( x) ln( x-1 ) 1
= e= e= .
x
We then multiply through by the integrating factor

1 y 1
y¢ - 2 = 3 ,
x x x
which can be written as

æ y ö÷¢ 1
çç ÷ = .
çè x ø÷ x3
276  •  Mathematical Physics

Integrating both sides and rearranging gives

1
yGS = - + Cx.
2x

In order that y → 0 as x → ∞ we must have C = 0, so the

required solution is
1
y=- .
2x

Let us try one more example of this before we move on.

Example 8.6 Find the general solution of ÿ − y = 0 such

that y → 0 as t → +∞ and hence find the solution satisfying
y(0) = 1 that remains bounded (i.e. does not take arbitrarily
large values) for all t ≥ 0.
The equation is linear, homogeneous and has constant
coefficients so we try y = Aeλt. Then

y = l Ael t ,
y = l 2 Ael t .
Next we substitute into the differential equation and form the
auxiliary equation

λ2Aeλt − Aeλt = 0.

Canceling Aeλt we have

λ2 − 1 = 0, so λ = ±1.

Hence we can write down the general solution, which has two
arbitrary constants.
yGS = Aet + Be−t.
 Differential Equations 2  •  277

But we have been asked to write down the general solution

such that y → 0 as t → + ∞. For yGS above, we can see that
y → ± ∞ as t → + ∞ unless A = 0. If A = 0 the condition is
satisfied, as e–t → 0 as t → + ∞. Thus,
yGS = Be− t.

Now we can apply the boundary condition y(0) = 1. At t = 0,

y(0) = Be0 = B = 1. Hence
y = e− t.

Exercise 8.5 Find the solution of y″ + 2y′ − 3y = e−x satisfying

y(0) = 0 that tends to zero as x → +∞.

8.2. AUXILIARY EQUATIONS WITH COMPLEX

ROOTS

The second part of this chapter is going to address solving dif-

ferential equations which involve complex numbers in one way or
another. Here you will need to use the knowledge gained from work-
ing through Chapter 7. We are going to begin by continuing the
work started in Section 6.7 on homogeneous linear equations, and
have a look at what happens if the roots of the auxiliary equation are
complex numbers.

Example 8.7 Find the general solution of y″ + y = 0.

This is a second order, linear equation with constant
coefficients; hence we try

y = Aeλx,
            
y′ = λAeλx, and y″ = λ2Aeλx.

Substituting into the differential equation, we form the

auxiliary equation
278  •  Mathematical Physics

λ2Aeλx + Aeλx = 0,
λ2 + 1 = 0,
λ2 = −1 thus λ = ±i.

The general solution contains two arbitrary constants

yGS = Aeix + Be− ix.

Recall from Euler’s equation (see Section 7.6) that

     eiq = cos(q) + i sin(q) and,

e−iq = cos(q) − i sin(q).

Hence we can rewrite our general solution in terms of sin and

cos as follows:
yGS = A(cos(x) + i sin(x)) + B(cos(x) − i sin(x)),
= (A + B) cos(x) + i(A − B) sin(x),
= a cos(x) + b sin(x).
Here we have written a = A + B and b = i(A − B). If y is
required to be real (which it often is) then a and b will be real.
Solving for A and B we get

1 1
A = ( a - ib), B = ( a + ib).
2 2

There is no need to go through all of the details every time we

use this technique. The following example is similar but shows you
how to set out the solution correctly and succinctly.

Example 8.8 Find the general solution of y″ + 4y = 0.

The equation is linear, homogeneous and with constant
coefficients so we try y = Aeλx (and hence y″ = λ2Aeλx as in
previous examples). Substituting in the results for y and y″,
then canceling factors of Aeλx, we have the auxiliary equation
 Differential Equations 2  •  279

λ2Aeλx + 4Aeλx = 0,
      ⇒λ2 = −4,
     So λ = ±2i.
We can write the general solution in explicitly real form with a
and b as the arbitrary real constants, as in Example 8.7 above.
yGS = Ae2ix + Be− 2ix,
yGS = a cos(2x) + b sin(2x).

Note that in the examples above we have found that the complex
roots of our auxiliary equations are both of the form ±bi. This has
allowed us to write the general solution yGS directly as a real function
of cos and sin, instead of a function of complex exponentials. In the
next example we will see what happens when the complex roots are
of the form a ± bi.

Example 8.9 Find the general solution of y″ + 8y′ + 20y = 0.

The equation is linear, homogeneous and with constant
coefficients so we try y = Aeλx (and hence y′ = λAeλx, and
y″ = λ2Aeλx as in previous examples).
l 2 Ael x + 8l Ael x + 20 Ael x =
0,
            l + 8l + 20 =
2
0,

    
( )
l = −8 ± 64 − 80 / 2 =−4 ± 2 i.

We can then write the general solution, with A and B as two

arbitrary constants

yGS = Ae(− 4 + 2i)x + Be(− 4 − 2i)x.

This is the same as

yGS = Ae−4xe2ix + Be−4xe−2ix,
= e−4x[Ae2ix + Be−2ix].
280  •  Mathematical Physics

We can again write the complex exponential terms in real form

using Euler’s equation, with arbitrary constants a and b
yGS = e−4x[a cos(2x) + b sin(2x)].

Key point: If the auxiliary equation has real coefficients but com-
plex roots, the roots will be of the form λ = a ± ib with a and b real. We
can then directly write the general solution of the form yGS = Ae(a+bi)x
+ Be(a−bi)x in terms of real functions as yGS = eax[a cos(bx)+ b sin(bx)].
Here a and b are related to A and B as a = A + B and b = i(A − B).
Here is an example from electric circuit theory.

Example 8.10 The charge q(t) on the capacitor in a circuit

with a resistor, capacitor and inductor in series is governed by

d2 q dq
L 2
+ R + q C = 0. (8.1)
dt dt
where R, C and L are positive constants and R2 < 4L/C. Find
the charge as a function of time.
For a trial solution q = Aeλt the auxilliary equation is

Lλ2 + Rλ + 1/C = 0.
This is a quadratic equation that can be solved to give
1/ 2
R 1 çæ 2 4 L ö÷
l =- ± çR - ø÷÷ . (8.2)
2 L 2 L çè C
If (R2 − 4L/C) < 0, then the roots are complex numbers, and
we can write this as
1/ 2
R 1  4L 
l=
− ± iwc , where wc =  − R2 
2L 2L  C 
1/ 2
 1 R2 
=  − 2 .
 LC 4 L 
(8.3)
 Differential Equations 2  •  281

Then the solutions have the form e(-R /2L±iwc)t, and the general
solution is
q(t) = e− Rt/2L (Feiwct + Ge− iwct).
Or, using the relation that eiwct= cos(wct) + i sin(wct), we can
write the equivalent form
q(t) = e− Rt/2L (Acos(wct) + B sin(wct)),  (8.4)
where the constants A and B (or F and G) are determined by
the initial values of the charge and the current in the circuit.

Exercise 8.6 Show that the general solution of y″ + 2y′ + 5y = 0

is y = e−x(a cos(2x) + b sin(2x)).

5
Exercise 8.7 Find the general solution of y″ + y′ + y = 0.
2

Exercise 8.8 Find the general solution of y″′ − 2y″ + 3y′ − 2y = 0

given that one solution is y = ex.

8.3.  EQUATIONS WITH COMPLEX COEFFICIENTS

We can use the same approach to solve a linear, homogeneous

equation with constant complex coefficients. In this case the solu-
tion will turn out to be complex. When we are dealing with a com-
plex function we use z instead of y.

Example 8.11 Find the general solution of z″ − iz = 0.

The equation is linear, homogeneous and with constant
coefficients so in the usual way we try z = Aeλx (and hence
z′ = λAeλx, and z″ = λ2Aeλx as in previous examples).
λ2Aeλx − iAeλx = 0,
   or λ2 − i = 0,

     so l = ± i .
282  •  Mathematical Physics

Recall that the square root of i is another complex number, and

was found in Example 7.7:
1+ i
i= ± ;
2

=so zGS Ae(1+ i) x / 2

+ Be−(1+ i) x / 2 .
    

Here the constants A and B are complex. This is as far as we need

to go since we do not require that z be real, so we are not required
to write this in terms of real functions. The complex constants A
and B would be fixed by the boundary conditions in an application.

Exercise 8.9 Find the general solution of z″ + iz′ + 2z = 0.

8.4.  COMPLEX INHOMOGENEOUS TERM

Examples of the type shown in Section 8.3 do not commonly

occur in physical sciences, and so we will not take this any further.
However, the following type of differential equation is very com-
mon–that is, a differential equation with real coefficients and a com-
plex inhomogeneous term. Let us look at some examples of how to
solve these types of equation.

Example 8.12 Find a particular integral of z + z = 3 e .

2 it

Since the right hand side of this second order, inhomogeneous

complex equation contains an exponential we will try a
particular integral solution of a similar form zPI = Ae2it (see
Section 6.9). It then follows that
z = 2iAe2it,
2
z = (2 i) Ae2 it ,
 Differential Equations 2  •  283

where the coefficient A is to be determined. We substitute

these results into the differential equation to obtain
(2i)2Ae2it + Ae2it = 3e2it.
Next we divide through by e2it and replace i2 by −1.
−4A + A = 3,
hence −3A = 3, and A = −1.
Finally we can write the PI as
zPI = −e2it.

Example 8.13 Find a particular integral of z″ + 2z′ + 3z = 6e3ix.

Again, the right hand side of this second order, inhomogeneous
complex equation contains an exponential. We will therefore
try a particular integral solution of a similar form for zPI = Ae3ix.
z′ = 3iAe3ix and z″ = (3i)2Ae3ix.
To determine the coefficient, A, we substitute into the
differential equation,
(3i)2Ae3ix + 6iAe3ix + 3Ae3ix = 6e3ix.
Next we divide through by e3ix and replace i2 by −1.
−9A + 6iA + 3A = 6,
    −6A + 6iA = 6,
6 1
=
  hence
hence A = .
−6 + 6 i −1 + i
Before we write down the solution for the PI, we can tidy up
A by multiplying through by the complex conjugate of the
denominator top and bottom (see Section 7.2)
( −1 − i ) 1
A= , or − (1 + i ) .
A=
( −1 − i )( −1 + i ) 2
Finally we can write the PI as
1
− (1 + i ) e3 ix .
zPI =
2
284  •  Mathematical Physics

Exercise 8.10 Find a PI of z + 2 z - 3 z = 2e .

it

Now let us consider an example where we are being asked to

find not just a particular integral, but the general solution.

Example 8.14 Find the general solution of z + 2 z + 3 z = 6 e3 it .

This equation is a second order, linear, inhomogeneous
equation with constant coefficients. Therefore we know that the
general solution will be a combination of the complementary
function, CF, and a particular integral, PI. First, let us solve the
homogeneous equation for the CF. The homogeneous part is

z + 2 z + 3 z = 0

We will try zCF = Aeλt hence zCF = l Ael t , and zCF = l 2 Ael t .
We substitute into the differential equations and divide through
by Aeλt to obtain the auxiliary equation

l 2 + 2l + 3 =0,

−2 ± 4 − 12
l= ,
2
−8 4 2
=−1 ± =−1 ± i =−1 ± i 2.
2 2
Then,
( −1+ i 2 ) t ( −1− i 2 ) t
=zCF Ae + Be ,
=zCF e− t Ae i ( 2t
+ Be− i 2t
).
The general solution is
zGS = zCF + zPI.
We found zPI in example 8.13 above, so

= (
zGS e− t Ae i 2t
+ Be− i 2t
) − 12 (1 + i ) e 3 it
.
 Differential Equations 2  •  285

Note that since the differential equation is complex, so is the

solution (thus A and B are two arbitrary complex numbers).

Exercise 8.11 Find the general solution of z″ + 2z′ − 3z = 20eix.

We now come to one of the most useful applications of com-

plex numbers to solving differential equations. This is where we use
our knowledge of complex numbers to solve real equations with real
coefficients and a harmonic driving term on the right hand side, i.e.
equations of the form

y + ay + by = c cos (w t ) + d sin (w t ). (8.5)

Example 8.15 Find a PI for y″ + 2y′ + 3y = 6 cos(3x).

Compare this with the complex equation that we solved in
Examples 8.13 and 8.14:
z″ + 2z′ + 3z = 6e3ix
If we use Euler’s equation, eiq = cos q + i sin q, we can rewrite
the inhomogeneous term (i.e. the RHS) as
z″ + 2z′ + 3z = 6 cos(3x) + 6i sin(3x).
Now we can see that the equation that we have been asked to
find a PI for (in y) is just the real part of the complex equation
above (in z).
So the solution to the equation, y″ + 2y′ + 3y = 6 cos(3x), is
just the real part of the solution to z″ + 2z′ + 3z = 6 cos(3x) + 6i
sin(3x).
Similarly, a PI solution for the real equation in y is just the real
part of the PI solution for the complex equation in z, found in
Example 8.13:

{1
}
yPI = Re - (1 + i) e3 ix ,
2
or, using e iq = cos(q ) + i sin(q ) , we have
286  •  Mathematical Physics

{1
yPI = Re - (1 + i)( cos(3 x) + i sin(3 x)) ,
2 }
{1
}
= Re - ( cos(3 x) + i sin(3 x) + i cos(3 x) + i 2 sin(3 x)) .
2

But i2 = −1, so
1 1
yPI = - cos(3 x) + sin(3 x) .
2 2

This solution can be checked by substituting back into the

equation that we have been asked to solve.

Similarly, if we wanted to solve the equation in Example 8.15

with 6 sin(3x) instead of 6 cos(3x) as the inhomogeneous term, we
should take the imaginary part of a PI to the complex equation
z″ + 2z′ + 3z = 6 cos(3x) + 6i sin(3x). Let us look at another example.

Example 8.16 Find the general solution of

y″ + 8y′ + 20y = 8 cos(4x)

To find the general solution we need to find the CF and a PI

for this equation, and add them together. We have previously
found the CF for this equation in Example 8.9, so we can use
that directly here:

yCF = e− 4x(a cos(2x) + b sin(2x)).

Next we need to find a PI. To do this we are going to first of all

write down a complex equation whose PI has a real part which
is equal to 8cos(4x). We do this because it will be quicker to
solve the complex equation and take the real part than it is to
solve for the PI in real form (i.e. in terms of cos and sin). We
write the following complex equation:
 Differential Equations 2  •  287

z″ + 8z′ + 20z = 8e4ix.

Now we try a PI of the form zPI = Ae4ix.

z = Ae4ix,
z′ = 4iAe4ix, and z″ = −16Ae4ix.
           

We substitute this into the complex differential equation, and

divide through by e4ix.
−16A + 32iA + 20A = 8,

2
   A = (1 + 8 i).

Then, multiplying the top and bottom by the complex conjugate

of the denominator, (1 − 8i), we have

2
   A = (1 - 8 i) .
65
2
Hence zPI = (1 - 8 i) e4 ix .
65

To find the real PI we take the real part of the complex PI,
yPI = Re(zPI):

 
yPI = Re {
2
65
(1 - 8 i) e4 ix , }
2
hence yPI =
hence
      Re{(1 - 8 i}( cos( 4 x) + i sin( 4 x))}
65
2
= ( cos( 4 x) - 8 i 2 sin( 4 x))
65
2
= ( cos( 4 x) + 8 sin( 4 x)) ,
65
2 16
hence yPI =
hence cos( 4 x) + sin( 4 x) .
65 65
288  •  Mathematical Physics

Finally we can write the general solution as the sum of the CF

and PI:

yGS = CF + PI

2 16
= e-4 x ( a cos (2 x) + b sin (2 x)) + cos (4 x) + sin (4 x).
65 65

Note that the solution contains both cos(4x) and sin(4x) terms
even though the inhomogeneous term in the differential equation
only has cos(4x). Physically this means that the response of the sys-
tem is out of phase with the driving force, 8 cos(4x). As discussed
above we could find a PI by guessing yPI = c cos(4x) + d sin(4x), with
c and d to be found, but this is usually much more long-winded and
the chances of making an error are higher.

Exercise 8.12 Use the method above to find a PI for each of the
following equations:
(i) y″ + 2y′ + 9y = 2cos(3x);
(ii) y″ + 7y′ + 12y = −150 sin(3x).

Exercise 8.13 A series LCR circuit driven with voltage

V = V0 cos(wt) satisfies the differential equation
d2 q dq
L 2
+ R + q / C = V0 cos (w t )
dt dt
where L, R and C are positive constants with (R2 − 4L/C) < 0.
If the steady state response (i.e. the PI) is q(t) = q0 cos(wt − d)
show that
V0
q0 =
2
w (1 / wC - w L) + R 2

and the phase difference d is given by

æ R ö
d = tan-1 ççç ÷÷÷.
è 1 / wC - w L ø
 Differential Equations 2  •  289

Show that the steady state current dq/dt has its maximum
amplitude when w2 = 1/LC. (This value of w is called the resonant
frequency of the circuit.)

Exercise 8.14 Find the general solution (CF + PI) of

y″ + 2y′ + 10y = 13 cos(2x).

8.5.  BOUNDARY (OR INITIAL) CONDITIONS

The following example puts together all we have learned in this

chapter. We can now use our knowledge of complex numbers to
solve complex differential equations, and we can apply boundary or
initial conditions to those solutions in exactly the same way that we
did at the start of the chapter.

Example 8.17 Find the solution of ÿ+ 2 y + 3y = 6 cos(3t)

subject to y(0) = 0, y (0) = 1.
We will begin by finding the general solution. First we find the
CF, by guessing yCF = Aeλt, and hence we can form the auxiliary
equation and solve for λ,
l 2 + 2l + 3 =0,
−2 ± ( 4 − 12 )
l= ,
2
hence l =−1 ± 2i

So a = −1, b = 2 . Having solved the auxiliary equation we

can write down yCF directly in real form as

(
yCF = e-t a cos( 2t ) + b sin( 2t ) . )
We can use the result of example 8.15,
1 1
yPI = - cos(3 t ) + sin(3 t ) .
2 2
290  •  Mathematical Physics

We add the CF and the PI to form the general solution

1 1
( )
yGS = e-t a cos ( 2 t ) + b sin ( 2 t ) - cos (3 t ) + sin (3 t ).
2 2
Then, differentiating, we can also find y GS,

(
y GS = -e-t a cos( 2t ) + b sin( 2t ) )
(
+e-t - 2a sin( 2t ) + 2 b cos( 2t ) )
3 3
+ sin(3 t ) + cos(3 t ) .
2 2

Now we impose the initial conditions. At t = 0,

1 1 1
yGS ( 0 ) =e0 ( a cos ( 0 ) + b sin ( 0 ) ) − cos ( 0 ) + sin ( 0 ) =a − =0.
2 2 2
And so a = 1/2. We also have y GS (0) = 1, so

y GS ( 0 ) (
− e0 ( a cos ( 0 ) + b sin ( 0 ) ) + e0 − 2 a sin ( 0 ) + 2 b cos ( 0 )
= )
3 3
+ sin ( 0 ) + cos ( 0 ) =
1,
2 2
3
⇒ − a + 2 b + =1.
2

And using a = 1/2 we find b = 0. Therefore, the solution is

=y
1 −t
2
e cos ( 2t ) − 12 cos ( 3t ) + 12 sin ( 3t ).

Exercise 8.15 Find the solution of y″ + 2y′ + 10y = 13 cos(2x)

subject to the initial conditions y(0) = 0, y′(0) = 2.
 Differential Equations 2  •  291

8.6.  SYSTEMS OF FIRST ORDER EQUATIONS

Just as we can have simultaneous linear equations, we can have

simultaneous differential equations. The simplest case is a set of
homogeneous first order linear equations with constant coefficients.
The following example looks at two such equations and illustrates
how these can sometimes be solved. (The example can also be solved
more simply using addition and subtraction to derive two indepen-
dent equations, but we use it here to illustrate the more general
method.)

Example 8.18 Find the general solution of the system for u(t),
v(t):
u + v =
0,
v − u =
0.
The system is linear with constant coefficients so we guess
exponential forms for both u and v with A ≠ 0, B ≠ 0.
=
Try u Ae =lt
and v Belt ,
=
then u l=
Aelt and v l Belt .
Substitute and divide through by eλt. This is possible since we
have chosen u and v to have the same exponential dependence.
   λA + B = 0,
and λB − A = 0.
Eliminate A and B to get the auxiliary equation for λ, and solve
for λ.
    B = A/λ,
λ2A + A = 0.
Since A ≠ 0
λ2 + 1 = 0,
  so λ = ±i.
292  •  Mathematical Physics

The general solution is a sum of the two possible solutions with

arbitrary coefficients.
uGS = Ceit +De– it.
C and D are arbitrary complex constants. The general solution
can be rewritten in explicitly real form, with a = 0 and b = 1
(see Example 8.8)
uGS = a cos(t) + b sin(t).
a and b are arbitrary real constants. The easiest way to obtain v
is to return to the original system
vGS = − u ,
vGS a sin ( t ) − b cos ( t ) .
=

Notice how this works: we add together the possible solutions

for one function (u here), choosing the constants to make the solu-
tion real, and the general solution for the other function (v here)
then follows.

Exercise 8.16 Find the general solution of the following system

of first order equations.
u − v − u =
0,
v + 4 v + 6 u =0.
What is the solution satisfying u(0) = 1, v(0) = 0?

8.7.  COMPLEX IMPEDANCE

We have seen that a PI of a linear differential equation with

constant coefficients subject to a harmonic forcing term (the right
hand side) is a harmonic at the same frequency as the forcing term
but a different amplitude and phase. Since the equation is lin-
ear the amplitude of the force and the response must be linearly
related. We represent the driving term as the real part of a com-
plex exponential and we write the constant of proportionality as Z,
 Differential Equations 2  •  293

called the impedance of the system. For a resistive mechanical sys-

tem the impedance relates the amplitude of the steady state veloc-
ity v to the amplitude of the applied harmonic force by v = F/Z;
for electrical circuits the impedance relates the current and the
applied voltage: I = V/Z. Note that this gives only the steady state
behavior and only for a harmonic driving term. Note also that to
get the physically measured real quantities we take the real parts:
for example, the applied voltage is Re{V eiwt} and the response is
Re{(V/Z)eiwt}. The impedance Z is constant in time but a function
of the frequency w (and the parameters of the system). Once we
know the impedance Z, solving for the steady state of a resistive
system subject to a harmonic force becomes an exercise in complex
algebra.

Exercise 8.17 The complex impedance Z of a circuit (an inductor,

capacitor and resistor in series) is given as
1
Z =+
R + iw L,
iwC
where R, C, and L are real positive constants. Show that Z can be
written in the form
2
 1  id
Z = R +  wL −
2
 e ,
 wC 

where
 1 
 wL −
d = tan − 1  wC  .

 R 
 
Hence show that an applied voltage Re(V0eiwt), with V0 real, gives
rise to a current
V0
= I cos (w t − d ) .
2
 1 
R2 +  wL − 
 wC 
294  •  Mathematical Physics

8.8.  SOME EXAMPLES FROM PHYSICS

We have studied numerous examples of solutions of equations

for some abstract functions y(x) or y(t). In real physical problems
the variables represent physical quantities and the notation will
reflect this. The major difference between the problems studied so
far and the examples in the following set is just the notation. The
procedures are the same as those you have applied in this chapter
and in Chapter 6.

Exercise 8.18 By seeking a solution of the form x = constant × emt

(and not otherwise), find the general solutions of the equation

x + rx + l=
x 0, r > 0

(where x = dx / dt ) in the cases

(i) λ = r2
(ii) λ = r2/4
(iii) λ = −r2.

Exercise 8.19 By seeking a solution of the form y = constant × emx

(for suitable m), find a particular integral of

y″ + 2y′ + 4y = 3 cos(wx)

and hence find the general solution. At what driving frequency, w,

is the amplitude of the forced oscillation, yPI, a maximum? (The
amplitude of an oscillation a cos(λx) + b sin(λx), where a, b and λ
are real constants, is (a2 + b2)1/2. Equivalently, the amplitude of an
oscillation Re(Aeiλx) is |A|.)

Exercise 8.20 The equation

dI
+ kI =j
ds
 Differential Equations 2  •  295

determines the specific intensity I(s) of radiation propagating in

a medium. (The positive functions k(s) and j(s) characterize the
absorption and emission properties of the medium and s mea-
sures distance along the ray path.) Find the solution I(s) satisfying
I(0) = 0 in the cases that (i) j, k are constants, (ii) j = as, k = bs,
with a, b constants. Show that in both cases I(s) can never exceed
j(s)/k(s) for s > 0.

Exercise 8.21 The angular velocity W of a rapidly rotating shaft

can be shown to obey an equation of the form
d 3W d 2 W
− 2 + 2W =
10
dt 3 dt
under certain conditions (t is time).
1. Find a particular integral of this equation, and hence
the general solution. (Hint: m3 − m2 + 2 = (m + 1)
(m2 − 2m + 2).)
2. Show that, except for certain special initial conditions, the
angular velocity will vary periodically, with an amplitude
increasing with time. (This type of unstable motion is called
“hunting.”)

Exercise 8.22 A “critically damped” oscillator obeys the equation

d2 y dy r 2
+ r + y=
0,
dt 2 dt 4
where y is the displacement, t is the time and r is a non-zero
constant. Find the solution obeying the initial conditions y = 0 at
t = 0 and dy/dt = v > 0 at t = 0. Show that the displacement is a
maximum at a time tmax which is independent of v and show that
the maximum displacement is ymax = 2v/re where e is the base of
natural logarithms.
296  •  Mathematical Physics

8.9.  EXTENSION: GREEN FUNCTIONS

So far we have used guesswork to obtain a particular integral of a

differential equation. In fact, if we know two independent solutions
of a homogeneous equation, then we can construct a formula for a
particular integral. In this section, we will outline how this is done,
leaving the details to more advanced texts.
Suppose we seek a PI of the equation y″ + y = f(x) where f(x) is
given. We know that cos(x) and sin(x) are two linearly independent
solutions of the homogeneous equation. (By linearly independent
we mean that there are no constants a and b such that a cos(x) +
b sin(x) = 0.) We construct the function

( cos ( x ) sin ( x′ ) − sin ( x ) cos ( x′ ) ) / W ( x′ ) x′ ≤ x

G ( x, x′ ) = 
0 x' > x

where W(x) = sin(x)(cos(x))′ − (sin(x))′cos(x) = −1. We now assert

that a particular integral of y″ + y = f(x) is

yPI ( x ) = ∫ G ( x, x′ ) f ( x′ ) dx′

∫ ( cos ( x ) sin ( x′) − sin ( x ) cos ( x′) ) / W ( x′) f ( x′) dx′
x
=
x0

with x0 < x an arbitrary constant. It is an exercise in differentiating

an integral with respect to a parameter (x here) to verify this result
(see section 10.8). Different choices of the lower limit x0 and different
linear combinations of cos(x) and sin(x) as the independent solutions
correspond to different PIs satisfying different boundary conditions.
The function G(x, x′) is called a Green’s function (sometimes a
Green function). If we think of the differential equation as a differen-
tial operator on y giving us f, then the Green’s function is the inverse
integral operator on f giving us back y.
 Differential Equations 2  •  297

In general, if we have two independent solutions u(x) and v(x)

of the equation y″ + a(x)y′ + b(x)y = 0, then a PI of the equation
y″ + a(x)y′ + b(x)y = f(x) is

y ( x ) = ∫ G ( x, x′ ) f ( x′ ) dx′
x

x0

Where

 ( u ( x ) v ( x′ ) − u ( x′ ) v ( x ) ) / ( W ( x′ ) x′ ≤ x
G ( x, x′ ) = 
 0 x' > x

and where W(x) = u′(x)v(x) − u(x)v′(x). This again can be verified

by back-substitution. In practice, there are only a limited number of
choices of the coefficients a(x) and b(x) for which we can find explicit
solutions of the homogeneous equation. Nevertheless, Green func-
tions have found widespread applications in theoretical physics.

Revision Notes

After completing this chapter you should be able to

• Apply boundary and initial conditions to the general
solution of a differential equation
• Solve differential equations involving complex roots in the
auxiliary equation, and write the solution in explicitly real
terms
• Solve differential equations with complex coefficients
• Solve differential equations with a complex inhomogeneous
term, and use this to simplify solving real equations with
harmonic inhomogeneous terms (i.e. a function of cos
and/or sin)
• Solve a simple system of first order differential equations
298  •  Mathematical Physics

8.10. EXERCISES

1. Solve the differential equation

dy
− 3 x2 y =
0
dx
subject to the condition y(0) = 1.
2. Solve the differential equation
dy
+ 3 x2 y2 =
0
dx
subject to the condition y(0) = 1.
3. Solve the differential equation
dy
+ xy2 =
0
dx
subject to the condition y(0) = 1.
dy
4. Integrate the differential equation = x 2 sec( y) .
dx
5. Solve the differential equation
dy
ey = x2
dx
by separation of variables for y as a function of x.
2
6. Find the solution of y′ + y = 1 that approaches a finite
value as x → 0. x

7. Solve the equation

dy 2 1
+ y=
dx x x3
subject to the condition y(1) = 0.
8. Solve the differential equation
dy y
+ =1
dx x
subject to the condition y(1) = 1.
 Differential Equations 2  •  299

9. Solve the differential equation

d2 y dy
2
−2 − 3y =
0
dx dx
subject to the conditions y → 0 as x → +∞, y(0) = 1.
10. Solve the differential equation
d 2 y dy
+ − 2y =
0
dx2 dx
subject to the conditions y → 0 as x →∞, y(0) = 3.
11. Solve the differential equation
d2 y dy
2
−3 − 4y =
0
dx dx
subject to the conditions y → 0 as x → +∞ and y(0) = 3.
 y direct substitution show that y = sin(wt + p/6) is a solu-
12. B
tion of the differential equation
d2 y
+ w2 y =
0.
dx2
Write down one other independent solution.
13. Find the general solution of y″+ 10y′ + 34y = 0.
14. F
 ind a particular integral of the equation y″ + 2y′ + 4y =
sin(3x).
15. Find a particular integral of the differential equation
d2 y dy
2
+2 − 4y =
− 4 x2 .
dx dx
16. Find a particular integral of the equation
d3 y d2 y dy
3
+ 4 2
+3 + y=− 2 e− x .
dx dx dx

d 2 y dy
17. Find a particular integral of the equation 2
+ + y=
e2 x.
dx dx
300  •  Mathematical Physics

18. F
 ind a particular integral of the differential equation
y″+ 7y′ + 4y = 8x2.
19. F
 ind the general solution of the differential equation
y″ − y′ + 3y = 0.
20. The general solution of the equation ÿ − 2y + 5y = 0 is

y = et(A cos(2t) + B sin(2t)).

What is the solution satisfying y = 0 and y = 1 at t = 0?

21. F
 ind the most general solution of the differential equation
y″ −y′ −2y = 0 such that y(x) → 0 as x → ∞.

8.11. PROBLEMS

1. Find the solution of the differential equation

d2 y dy
2
+3 cos ( x )
+ 2y =
dx dx
subject to the conditions y = 11/10, dy/dx = −7/10 at x = 0.
2. Find the solution of the differential equation
d2 y dy
2
+2 + 5y =
5x
dx dx
subject to the conditions y = −2/5 and dy/dx = 0 at x = 0.
3. Solve the differential equation
d 2 y dy
− − 2y =
0
dx2 dx
subject to the conditions y → 0 as x → +∞, and y(0) = 1.
4. Find a particular integral of the equation
d 2 y dy
− − 6y =
13.
dx2 dx
Hence (or otherwise) find the solution of the equation
satisfying y(0) = 0 and dy/dx = 0 at x = 0.
 Differential Equations 2  •  301

5. The complex function z(t) satisfies the differential equation

z − 2 z + 2 z = 2 (1 − 5 t ) e2 it .

Find the general solution of this equation.

What is the solution which satisfies z(0) = 0 and z (0) = 0?
6. Solve the differential equation
d2 y dy
2
+2 + 2y = 5cos ( 3 x )
dx dx
subject to the conditions y = dy/dx = 0 for x = 0.
7. Solve the differential equation
d2 y dy
2
+2 sin ( 2 x )
+ 2y =
dx dx
subject to the conditions y(0) = y(π/2) = 0.
8. Solve the differential equation
d2 y dy
2
−4 cos ( x )
+ 8y =
dx dx
subject to the conditions y = dy/dx = 0 for x = 0.
9. Find the general solution of the equation
d 2 y dy
2
+ e− t .
− 12 y =
dt dt
For what values of the arbitrary constants is the solution
bounded (i.e. remains finite) as t → +∞? Determine the
relation that must be satisfied between y(0) and y′(0) in
order that the solution be bounded as t → +∞. (Such a
system is physically unstable since for the smallest change
in the initial conditions y → ±∞.)
10. F
 ind the solution of y″ − 4y′ + 5y = 65 cos(x) − 5 subject to
the conditions y(0) = 0, y′(0) = 1.
11. F
 ind the general solutions of the following differential
equations:
(i) y′ = (1 − y2)1/2x3
(ii) y′ + ycot(x) = 2cos(x)
302  •  Mathematical Physics

(iii) y″ − 4y′ + 3y = 0
(iv) y″ − 4y′ + 4y = 0.
12. Find the general solutions of the following differential
equations:
(i) y2y′ − 1 = cos(x)
(ii) x3y′ + x2y = 1
(iii) y″ − 3y′ − 4y = 0
(iv) y″− 6y′ + 9y = 0.
13. F
 ind the general solutions of the following differential
equations:
(i) y′ = 3x2y
2x
(ii) cot ( x ) y' − y =
sin ( x )
(iii) y″ −10y′ + 25y = 0
(iv) y″− 7y′ + 12y = 6ex.
14. (i) Find the general solution of the differential equation

y″ + 6y′ + 13y = 0.

Express your solution in terms of real functions of the

variable x.
(ii) The function y satisfies the differential equation

y″ − 5y′ + 4y = 8x − 6.

Find the general solution of this equation. What is the

solution which satisfies y(0) = 1 and y′(0) = −1?
 CHAPTER

9
MULTIPLE INTEGRALS

In this chapter, we learn how to integrate functions of two or three

variables over two- or three-dimensional regions with simple shapes.
As well as the more obvious applications – for example calculating the
volume of a given body or the rate of energy emission from a given
surface – there are important but less obvious ones – for example find-
ing the mean speed of stars in a galaxy. By using suitably chosen coor-
dinates such integrals can often be evaluated relatively simply. One
very important application is to the definition of solid angle. As we
shall see, the concept of solid angle extends the notion of an angle
subtended at a point by an arc to two-dimensional extended objects.

9.1.  REPEATED INTEGRALS

Given a function f(x, y) of two variables (x, y), we can obtain the
integral
f ( x, y) dy.
y2
F ( x) = ∫ y1

by holding x temporarily constant. Given some limits y1 and y2 the

result is a function F(x) of just one variable (x). We can then inte-
grate the result with respect to x:

∫ (∫ )
f ( x, y) dy dx. (9.1)
x2 x2 y2
=I F ( x ) dx
∫=
x1 x1 y1
304  •  Mathematical Physics

We can write this with fewer parentheticals as

f ( x, y) dydx (9.2)
x2 y2
I= ∫ ∫x1 y1

with the understanding that equation (9.2) means “do the integrals
in the order indicated in equation (9.1),” i.e. “from the inside, work-
ing out.” We call this a repeated integral (also an iterated integral or
a double integral). Note that the order indicated by the bracketing
in equation (9.1) is one convention. Some authors use an alternative
convention in which the first integral sign is associated with the first
differential and so on in sequence.

1 1
Example 9.1 Evaluate ∫ ∫ −1 0
x2 y dydx .

1
The inner integral is ∫0
x2 y dy . For this inner integral with
respect to y, x is a constant, thus
1
1  y2  x2 1
∫0
= ∫0
= =
2 2 2
x y dy x y dy x   .
 2 0 2
So
1
1 1 1 x2  x3  1
∫ ∫
−1 0
x2 y=
dydx ∫−1 2
= dx = 
  −1 3
6
.

1 ∞
Example 9.2 Evaluate ∫∫ 0 a
xe− xy dydx (where a > 0 is a given
constant).
∞
The inner integral is ∫a xe dy . Here x is treated as a constant
− xy

parameter for the integration with respect to y:

y= ∞
∞ ∞  e− xy 
∫ x∫ e
− xy − xy
xe dy = dy =
x −  = e− ax .
a a
 x  y= a
 Multiple Integrals  •  305

So
x =1
 e− ax  1
− ( e− a − 1 ) .
1 ∞ 1
∫∫ ∫0
− xy − ax
xe dydx = e dx =
− a  =
0 a
  x=0 a

2 1
Exercise 9.1 Evaluate ∫∫
0 0
xy3 dydx .

The extension to more than two integrals is straightforward, as

in the following example.

1 p /2 2
Example 9.3 Evaluate ∫∫ ∫ zx2 sin( y) dzdydx .
0 0 0

The innermost integral is

2 2
∫ zx2 sin( y) dz = x2 sin( y) ∫ z dz
0 0

= 2 x sin( y) .
2

So now
p /2 2 p /2
∫ ∫ zx2 sin( y) dzdy = ∫ 2 x2 sin( y) dy
0 0 0
p /2
= 2 x2 ∫ sin( y) dy
0

= 2x . 2

And finally
1 p /2 2 1 2
∫∫ ∫ zx2 sin( y) =
dzdydx ∫=
2 x dx 2
.
0 0 0 0 3

1 p /2 1
Exercise 9.2 Find ∫∫ ∫ xy sin( zy) dzdydx .
0 0 0

If, as in all these examples, the limits of integration are constants,

then (for the integrands that are likely to arise in practice in physical
sciences) the integrals can be evaluated in any order.
306  •  Mathematical Physics

This is obvious for cases like Example 9.1 because the double
integral is a product of single integrals:

∫ (∫
1

−1
1

0 )
x2 ydy dx = ∫ x2 ( ∫ ydy) dx
1

−1
1

= ( ∫ x dx ) ( ∫ y dy )
1 1
2
−1 0

 product 
 
=  of single  .
 integrals 
 
which, reversing the sequence of steps,

= ( ∫ y dy) ( ∫ x dx)
1

0
1

−1
2

= ∫ y( ∫ x dx ) dy
1 1
2
0 −1

= ∫ ( ∫ yx dx ) dy.
1 1
2
0 −1

In fact, it is generally true that when the integrand can be fac-

tored into the product of two functions, f(x,y) = g(x)h(y), then the
repeated integral over x and y can be factored into the product of
two single integrals

∫ ∫ f ( x, y) dxdy g( x ) h( y) dxdy
y1 x1 y1 x1

y0 x0
= ∫ ∫ y0 x0

= (∫ x0
x1
g( x ) dx )( ∫ y1

y0 )
h( y) dy .

We may also change the order of integration in cases, such as

Example 9.2, where the double integral is not equal to a product
of single integrals. (In these cases one choice of order may result
in an easier calculation than another, but the final result will be the
same.)
But if the limits of integration are not constants (e.g. if a in
Example 9.2 were a function of x), then the order cannot simply be
reversed. This is dealt with in Section 9.3.
 Multiple Integrals  •  307

Exercise 9.3 Evaluate the integral

∫ (∫ )
∞ 1
xe− xy dx dy,
a 0

and check that the result is the same as obtained by integrating in

the reverse order in Example 9.2. (Hint: the inner integral can be
evaluated by integration by parts; also

 e− y e− y  e− y
∫  y y2 
 +  dy =
−
y
.

9.2.  INTEGRALS OVER RECTANGLES IN THE PLANE

The area under the graph of a function y = f(x) is given by the

definite integral of f. We therefore expect that the volume under a
surface, z = f(x,y), is the double integral of f(x,y). To demonstrate
this we shall adapt the proof from one dimension to two.
y First, consider a curve in the
y = f (x)
plane. The area Ak of the rect-
angle (shaded) of width dx at xk
Ak ≈ height of curve × base
y = f (xk)
≈ f(xk)dx
while we have that the area
under the curve between a and
b is
δx
x
A≈ ∑ f ( x ) d x.
k
k

a xk b

In the limit dx → 0, k → ∞ we get the exact result:

 Area under the curve 
  b

 in the interval  = ∫a f ( x ) dx.

 between a and b 
 
308  •  Mathematical Physics

Now, for the analogous vol-

z ume under a surface. Looking at
z = f (x, y)
the figure we have the volume
Vk of the box above the point
(xk ,yk) with base dx × dy:
Vk ≈ height of surface × area
of base
(xk, yk)
≈ f(xk, yk) dxdy
y
Meanwhile, the volume under
x δx
δy
the surface is
V≈ ∑ f ( x , y ) d xd y.
k k
k

In the limit dx, dy → 0, k → ∞ we get the exact result

 Volume under the surface 

  y2

 above the plane rectangle  = ∫y1

 x ≤ x ≤ x ,y ≤ y ≤ y 
(∫x1
x2
)
f ( x, y ) dx dy.
 1 2 1 2 

This volume is also referred to as the “integral of the function

f(x, y) over an area,” the “area” being in this case the region of the
plane rectangle. Such integrals often arise in physical science as
“integrals of functions over plane regions.” Note that the inner inte-
gration is over a slab of width dy at some general value of y and then
the outer integral adds all the slabs together.
As in the next example, the result of a repeated integration
is not necessarily a physical volume (with units of m3), just as
the result of a single integration is not necessarily a physical
area (with units of m 2). When we talk about the “volume” of a
double integral, we mean the volume under the surface in the
graph of f(x, y). If f(x, y) and x and y all have units of length,
then the double integral does indeed give a physical volume. But
if f(x, y) is a charge density over a surface then its integral with
respect to its two spatial dimensions (x, y) will be a charge, not a
volume.
 Multiple Integrals  •  309

Example 9.4 A square plate with sides of unit length has a

surface mass density

s(x, y) = 1 + cos(px) cos(py),

where x and y are measured from the center of the plate
parallel to its sides. Find the mass M of the plate.
The mass of an element dxdy of surface is the mass per unit
area (i.e. the surface density) × area, hence sdxdy. The mass
M of the plate is the integral of this, i.e. of the density per unit
area over the region of the plate, thus:
1/ 2 1/ 2
M= ∫ ∫ s ( x, y) dx dy.
−1/ 2 −1/ 2

Notice the limits of each integration. We are told in the

question that the plate is square, has sides of length 1, and x
and y are measured from the center. So x runs from −1/2 to
+1/2, and so does y.
Evaluate the inner integral:

12
 1 
∫1 2 ( 1 + cos( p x) cos( py) ) dx =
12

 x + p sin( p x ) cos( p y) 
−1 2

2
= 1+ cos( p y) .
p
Now evaluate the outer integral:

1/ 2 2 
=
M ∫  1 + cos( p y)  dy
1/ 2  p 
1/ 2
 2 
=  y + 2 sin( p y) 
 p −1/ 2
4
= 1+ .
p2
310  •  Mathematical Physics

Exercise 9.4 The electromagnetic radiation emitted between fre-

quencies ν and ν + dν from the annular region of a disc between
radii r and r + dr has a power

Cr−a exp(−bνr1/4)dνdr

where C, a and b are positive constants and a > 3/4. If the inner
radius of the disc is r0 and the outer radius is taken to be infi-
nite, write down and evaluate the double integral giving the total
power radiated by the disc over all (positive) frequencies.

9.3. MULTIPLE INTEGRALS OVER IRREGULAR

REGIONS

In all the examples above, we were given the coordinates with

which to integrate (dx, dy) and the region we integrated over was
rectangular in these coordinates, defined by lines of constant x and y.
Suppose we have some function f(x, y) and we wish to integrate over
some non-rectangular region – for example to calculate the charge
on a disc if f(x, y) is the surface density of the charge. To integrate a
function over an irregular region in the plane we divide the region
into tiny area elements, each of area dAk (each given a number, k).
Then we sum dIk = f(xk, yk)dAk over all the area elements that fall
entirely within the region.

=I ∑
= dI ∑ f ( x , y ) dA
k in R
k
k in R
k k k (9.3)

In the limit of dA → 0 and k → ∞ we get the double integral over

the region

I= ∫∫ f ( x, y) dA (9.4)
R

This is subtly different from the repeated integrals we exam-

ined previously. Here we evaluate over area elements dA – we did
not specify which coordinates to use to integrate over. In practice,
double or multiple integrals can be evaluated as repeated integrals.
 Multiple Integrals  •  311

(This result is known as Fubini’s theorem and holds for almost all
integrals you are likely to encounter.) So we choose coordinates that
make the integrand or the region simple to deal with, and evaluate
the repeated integral.
y Left: A bounded, irregular
(xk, yk)
region in the (x, y) plane, par-
titioned into rectangular cells.
R In order to define the double
δy
integral of a function f(x, y)
over a non-rectangular region
R, we imagine the region parti-
tioned into rectangular cells of
area dA = dx × dy. And for each
δx x cell (numbered k) we compute
the volume over that cell,
Vk = f (xk, yk)dx dy, just as in the case of the rectangular region. We
then sum these small volumes over only those cells that fall entirely
within the region R:
V≈ ∑ f ( x , y ) d xd y.
k in R
k k

As before, as the grid becomes finer (dx, dy → 0) the difference

between the grid and the integration region tends to zero and we have

=I f ( x, y) dA ∫∫ f ( x, y) dx dy. (9.5)
∫∫=
R R

So we can write the double integral over an area as a repeated

integral over x and y. But we must be careful with the limits as the
following example shows.

Example 9.5 Evaluate I = ∫∫∆ x y dydx where D is the triangle

2

(Figure 9.1) bounded by the lines x = 1, y = 0, y = 2x.

We begin by putting in the limits of integration. In the inner
integral with respect to y, x has a general value somewhere
in D. Draw the dotted line (Figure 9.1) parallel to the y-axis
at some general value of x in D. From this we see that, within
312  •  Mathematical Physics

D, y runs from 0 to 2x. These are the limits for the inner (y)
integral. Thus, the inner integral is
2 y= 2 x
2 y 
2x 2x
∫= x y dy x= ∫0 y dy x =
2 2
  2 x4 .
0 2
  y= 0
Now, to cover D we must let x range between 0 and 1. These
are the required limits for the outer integral, thus

∫ (∫ )
1
1 2x 1  x5  2
= x2 y = ∫0 = =
4
I dy dx 2 x dx 2 5 .
0 0
 0 5

Exercise 9.5 Evaluate I = ∫∫

∆
x2 y dydx where D is the triangular
region in Figure 9.2(a), bounded by the lines x = 0, y = 0, y = 1 − x.

Exercise 9.6 Evaluate I = ∫∫ D

x2 y dydx over the quarter of the
unit disc in Figure 9.2(b).

Note that in these examples we have drawn the diagram for you.
In general you will have to begin by sketching your own figure (see
Exercise 9.7).
y
z
2
2
y = 2x x=1
y
2
1
1
1

y = 2x
0
0 x
0
0 y=0 1 1 x

(a) (b)
FIGURE 9.1: (a) The region bounded by y = 2x, x = 1 and y = 0. (b) The function z = x2y with
the triangular region indicated on the (x, y) plane.
 Multiple Integrals  •  313

y
x=0
y
x=0
1 x2+ y2 = 1
1
y = 1–x

0 0
0 1 x 0 1 x
y=0 y=0

(a) (b)
FIGURE 9.2: (a) The region for Exercise 9.5 bounded by y = 1 − x, x = 0 and y = 0. (b) The
region for Exercise 9.6 bounded by x2 + y2 = 1, x = 0 and y = 0.

9.4.  CHANGE OF ORDER OF INTEGRATION

As stated, for most functions order of integration does not mat-

ter if the limits of integration are simple constants, e.g. integrating
over a rectangular region (with x1 ≤ x ≤ x2 and y1 ≤ y ≤ y2). When
the limits of the integrals are themselves functions, we can still
change the order of integration if we are careful with the limits, as in
Example 9.6. (There are some cases, with functions that are not of
constant sign or involving infinities or discontinuities, where we can-
not change the order of integration. But these are seldom encoun-
tered in undergraduate physics.)

Example 9.6 Evaluate I = ∫∫ x2 y dxdy where D is the region

∆
outlined in the figure (below). This is the same as in example 9.5
but with the order of the integrations reversed.
For the inner integration with respect to x, y has a general
value in D. Draw the dotted line at a general y within D
314  •  Mathematical Physics

y
parallel to the x-axis. We see
2
that, within D, x runs from
x = y/2 x=1
x = y/2 to x = 1. For the outer
integral the range y = 0 to y = 2
1 covers the triangle so these are
the limits for the y integration.

0 x
0 y=0 1

I =∫
0
2
(∫y/2
1
)
x2 y dx dy
x =1
2  x3  2 y y3 
= ∫
0
y  =
 3  x= y / 2
dy ∫ 0
1 −
3 8 
 dy

2
1  y2 y5  1 4 32 
= −  =  − 
3  2 8 × 5 0 3 2 8 × 5 
1 4 2
=  2 −  = .
3 5 5

The function to be integrated and the region of integration are

the same as in the example of Section 9.3, so the answer is, of course,
the same. But note that you have to work out the limits (e.g. from a
diagram) each time. This requires thought and care.
Sometimes integrals that arise in one order are best evaluated in
another. The following is an artificial example.

∫ (∫ )
1 p /2
Example 9.7 Evaluate cosec ( x ) dx dy by
0 sin − 1 ( y )

changing the order of integration (assuming that sin–1(y) is in

the range 0 to π/2).
 Multiple Integrals  •  315

y First we draw a sketch. The

1 y = sin (x )
1 1
inner integral is an integral
with respect to x, where
the limits are x = sin–1(y) to
x = p/2. So we plot y = sin(x) and
examine the horizontal extent
of the region between this
x1 = sin–1 (y1) curve and x = p/2. Looking at
0
x the outer integral tells us that y
0 x1 π/2 π ranges between y = 0 and y = 1.
This gives us the shaded area in
the figure.
Now we can safely change the order of integration. Now the
inner integration is with respect to y. For a general value
of x (draw a vertical line at some x), y ranges between
0 and sin(x). So these are now the limits for the inner y
integral.
Within the shaded region x ranges over 0 to p/2, so
these are the limits for the outer x integral, so we now have
that

0
1
∫ ∫
 p /2

sin −1
( y)
cosec( x ) dx dy =
 ∫
0
p/2
( ∫ cosec( x) dy) dx
0
sin( x )

= ∫
0
p /2
( cosec( x) ∫ dy) dx
sin( x )

0
p /2 sin( x )
=∫ cosec( x )  y 0 dx
0
p /2
=∫ cosec( x ) sin( x ) dx
0
p /2 p
= ∫=
0
dx
2
.
316  •  Mathematical Physics

Exercise 9.7 Sketch the regions of integration and insert the cor-
rect limits on the right hand sides of the following equalities:

∫ ( ∫ f ( x, y) dx ) dy = ∫ ( ∫ f ( x, y) dy) dx ,
1 y
(i)  0 0

∫ ( ∫ f ( x, y) dy) dx .
∞
 ∞
f ( x, y) dx dy =
(ii)  ∫ ∫
0 y 

P(x, y)
y y

dr
r

dA
r rdq

q dq

0
0 x x

(a) (b)
FIGURE 9.3: (a) Polar coordinates (r, q) and Cartesian coordinates (x, y). (b) The area ele-
ment in terms of polar coordinates: dA = rdrdq .

9.5.  POLAR COORDINATES IN TWO DIMENSIONS

Instead of the Cartesian labels (x, y) for points in the plane, we

can use a distance r from the origin and an angle q (from the x-axis)
as in Figure 9.3(a). From the figure we deduce the relation between
the (x, y) and (r, q) labels of the point P:
x = r cos(q),
     y = r sin(q).(9.6)

The inverse relations are

  r = (x2 + y2)1/2,
    q = tan–1(y/x).  (9.7)
 Multiple Integrals  •  317

Functions in the plane may take on a simpler form in polar

coordinates.

Example 9.8 What is the function f(x, y) = (x2 + y2)3/2 in polar

coordinates?
We substitute x = r cos(q), y = r sin(q) in f(x, y) to get
f(r, q) = (r2 cos2(q) + r2 sin2(q))3/2 = r3

− 1/ 2
 y2 
Exercise 9.8 What is the function f ( x,=
y) x  1 + 2  in
 x 
polar coordinates (r, q)?

Exercise 9.9 What are the curves

(i)  r = constant
(ii)  q = constant?

9.6.  INTEGRALS OVER REGIONS IN THE PLANE

In polar coordinates the element of area is dA = dr × rdq (see

Figure 9.3b). We can use this to integrate a function over a plane
region. Instead of

∫ f dA = ∫∫ f ( x, y) dxdy
we can write

∫ f dA = ∫∫ f ( r , q ) rdrdq .
This is useful if the limits of integration are more readily
expressed in (r, q) coordinates. The temptation to write drdq (instead
of rdrdq) must be resisted; dxdy is an area, but drdq is not – they
cannot be equal.
318  •  Mathematical Physics

Example 9.9 Find the integral of the function f(x, y) = xy over

the region x2 + y2 ≤ 1, x ≥ 0, y ≥ 0.
Let

I= ∫∫ R
xy dxdy

where R is the given region. Because the area is a quadrant of

a circular disc, we transform the problem to polar coordinates.
So, first transform the integrand:

f ( x, y) = xy
= ( r cos( q ) ) ( r sin( q ) )
= r 2 cos( q ) sin( q )
Next transform the element of area:

dxdy = rdqdr

Finally, we transform the limits: the area of integration is the

first quadrant for which 0 ≤ r ≤ 1,0 ≤ q ≤ p/2, and so now the
integrand is a product of a function of r and a function of q with
constant limits:

p/ 2 1
I = ∫ ∫ r cos( q ) sin( q ) rdq dr
2
0 0
p/ 2 1
= ∫ sin( q ) d ( sin q ) ∫ r dr 3
0 0
p/ 2 1
1  1 
=  sin 2 ( q )   r 4 
2 0  4 0
1 1 1
= × = .
2 4 8

Note how the integral in q is evaluated using d(sin(q)) = cos(q) dq.

This is equivalent to changing the variable to u = sin(q). It is
not necessary to change the limits for sin(q) explicitly.
 Multiple Integrals  •  319

Exercise 9.10
(i) Find the integral of the function f(x, y) = x2 + y2 over the
disc x2 + y2 ≤ 4.
(ii) A disc of radius 2m centered on the origin has a variable
surface density given by s(r) = 0.1r2 kg m–1. What is the
mass of the disc?

9.7.  AN IMPORTANT DEFINITE INTEGRAL

As a further example of integration over a region in polar coordi-

∞
∫
2
nates we look at the following trick to evaluate e− x dx .
0

Example 9.10 Evaluate the integral of the function

2 2
f(x, y) = e−(x +y ) over the whole plane. In Cartesian coordinates
this is
∞ ∞
I =∫ ∫
2 2
e− x e− y dxdy
−∞ −∞

= (∫ −∞
∞
e− x dx
2

)( ∫ ∞

−∞
2
e− y dy )
= (∫ dx ) .
∞ 2
2
e− x
−∞

∞
But there are no elementary methods to evaluate ∫ e− x dx .
2

0
We can, however, transform the problem to polar coordinates.
The problem becomes: evaluate the integral of the function
2
f(r, q) = e− r over the whole plane. This is
2p ∞
∫ ∫
2
I = e− r rdrdq
0 0

= (∫ 0
2p
dq )( ∫ 0
∞ 2
e− r rdr )
∞
 1 2
= 2p × − e− r  = p
 2 0
320  •  Mathematical Physics

Exercise 9.11 Deduce from the example that

∞
∫
2
e− x dx = p
−∞

and hence that

∞ 1
∫
2
e− x dx = p
0 2

The example is important because (i) this integral appears in

various branches of physics and mathematics, and (ii) it cannot be
evaluated by elementary methods (do not waste time trying!).

9.8.  CYLINDRICAL POLAR COORDINATES

We can extend two-dimensional polar coordinates to three dimen-

sions by adding the z-coordinate. Thus we label points in space by two
distances and an angle as in Figure 9.4(a). Note that the usual conven-
tion is to label the radial coordinate r here (in place of r in section 9.5)
and the angular coordinate φ (in place of q) to avoid confusion with the
spherical polar coordinates introduced in Section 9.11 below.1
z

z ρ dz dV
P

O y y
x dφ
dρ
x ρdφ
φ

(a) (b)
FIGURE 9.4: (a) Cylindrical polar coordinates (r, φ, z) and Cartesian coordinates (x, y, z).
(b) The volume element in terms of cylindrical polar coordinates dV = pdpdφdz.

1
This notation is common in physics, but not universal. When working
on a particular problem, always check the definitions.
 Multiple Integrals  •  321

From Figure 9.4(a) we deduce the relation between the (x, y, z)

and (r, φ, z) labels of the point P:
x = r cos (f ) ,
y = r sin (f ) , (9.8)

z = z. 
The inverse relation is

(x + y2 )
1/ 2
r
= 2
,
f = tan −1 ( y / x ) , (9.9)
z = z. 

Exercise 9.12 What are the surfaces

(i)  r = constant,
(ii)  z = constant,
(iii)  φ = constant?

Exercise 9.13 What does the function

f(x, y, z) = zn ln(1+ x2 + y2)

become as a function of cylindrical polar coordinates (r, φ, z)?

Note that in Exercise 9.13 the function is independent of φ; we

say that it is axially symmetric. If n = 0 the function is independent
of both φ and z; we say that the function is cylindrically symmetric.

Example 9.11 Show that a function that is cylindrically

symmetric is constant on cylinders centered on the z-axis.
If f is cylindrically symmetric it is independent of both φ and z,
i.e. it is a function of r only, f(r). Thus f is constant on surfaces r
= constant, i.e. on cylinders.
322  •  Mathematical Physics

9.9.  VOLUME INTEGRALS (CYLINDRICAL POLAR)

In cylindrical polar coordinates (r, φ, z) the element of volume

is (from Figure 9.4b)

dV = dr × rdφ × dz = pdpdφdz.

We can use this to integrate a function over a region of three-

dimensional space (solid body). Instead of

∫ f dV = ∫∫∫ f ( x, y, z) dx dy dz
we can write

∫ f dV = ∫∫∫ F ( r, f, z) rdr df dz,

where F(r, φ, z) is the function f(x, y, z) expressed in terms of r, φ
and z. This is useful if the limits of integration are readily expressed
in (r, φ, z) coordinates.

Example 9.12 Find the integral of the function x2 + y2 over

the cylinder 0 ≤ r ≤ 1, 0 ≤ z ≤ 1.
First we transform the function to cylindrical coordinates

x2 + y2 = (r cos(φ))2 + (r sin(φ))2 = r2.

The correct element of volume is

dV = rdrdφdz,
while the limits of integration are 0 ≤ r ≤ 1, 0 ≤ z ≤ 1, 0 ≤ q ≤ 2p.
Put these together and evaluate the integral and we have that
1
r4 
( x2 + y2 ) dV =
1 2p 1
∫ ∫0 ∫0 ∫0  4  [f ]0 [ z]0
2p 1
r 2
× r d r df dz=
V
 0
1 p
= × 2p × 1 = .
4 2
 Multiple Integrals  •  323

dz
dA

r
y
a dφ
x adφ

FIGURE 9.5: The cylindrical element of area for a cylinder of radius r = a: dA = adφdz.

Exercise 9.14
(i) Find the integral of the function f(x, y, z) = xyz over the
wedge defined by 0 ≤ r ≤ 1, 0 ≤ φ ≤ p/2, 0 ≤ z ≤ 2.
(ii) The number per unit volume of bacterial fossils in a
cylindrical wedge of rock, 0 ≤ φ ≤ p/2, of radius 1m and
height 2m is 105 xyz m–3. How many fossils are in the rock?

9.10.  INTEGRALS OVER CYLINDRICAL SURFACES

Suppose we are given a function f(φ, z) at each point on some part

of the surface of a cylinder r = a = constant. The function might repre-
sent the surface mass density of the cylinder, or the surface density of
electric charge, or the flux of energy through the surface, and so on. We
might then wish to evaluate the integral of f over the surface (to get the
total mass, charge etc.). From Figure 9.5 we have dA = adφdz.
So

∫ f dA = ∫∫ f (f , z ) adf dz
the surface integral of f over the cylindrical surface.
324  •  Mathematical Physics

Example 9.13 Find the integral of the function z sin(φ/2) over

the cylindrical surface r = 4, 1 ≤ z ≤ 2.
We write down the element of area on the surface:

dA = 4dφdz.

The limits of integration required to cover the portion of the

cylinder once are 0 ≤ φ ≤ 2 p and 1 ≤ z ≤ 2. Finally, we write
down and evaluate the integral:
2
2p  z2 
z sin (f 2 ) 4= 4    − 2 cos (f 2 )  0
2 2p
∫∫
1 0
df dz
 2 1
= ( 8 − 2 ) ( − 2 × − 1 + 2 × 1 ) = 6 × 4 = 24.

cos (f )
Exercise 9.15 Find the integral of the function over the
(1 + z )
2

part of the surface of the cylinder r = 2, −p/2 ≤ φ ≤ p/2, z ≥ 0.

z
z r sin(θ)dφ
P dV

θ
θ r dr

O y dθ
φ y
x
r
dφ
x

(a) (b)

FIGURE 9.6: (a) Spherical polar coordinates (r, φ, q) and Cartesian coordinates (x , y, z).
(b) The volume element in terms of spherical polar coordinates dV = r2 sin(q)drdqdφ.
 Multiple Integrals  •  325

9.11.  SPHERICAL POLAR COORDINATES

We can label points in space by a distance and two angles as

in Figure 9.6(a). The coordinates (r, q, φ) are the spherical polar
coordinates of the point P. From the figure we deduce the relation
between the (x, y, z) and (r, q, φ) labels of P:
x = r sin (q ) cos (f ) ,
y = r sin (q ) sin (f ) ,

z = r cos (q ) . (9.10)

The inverse relation is

(x + y2 + z 2 )
1/ 2
r= 2
,

= (
q cos−1 z / ( x2 + y2 + z2 )
1/ 2
),
f = tan −1 ( y / x ) .
(9.11)
Note carefully (and remember) how the angles q and φ are defined:
this is the standard convention and you need to know it − q is the
co-latitude of P (not the latitude; i.e. it is measured down from the z-axis,
not up from the equator). The angle φ is measured in the (x, y) plane
from the x-axis.

Exercise 9.16 What are the surfaces

(i) r = constant,
(ii) q = constant,
(iii) φ = constant?

Exercise 9.17 What does the function f(x, y, z) = zn ln(1+ x2 + y2)

become in spherical polar coordinates?

Note that the function in Exercise 9.17 is independent of φ: it is

axially symmetric. A function that is independent of both q and φ is
said to be spherically symmetric.
326  •  Mathematical Physics

Exercise 9.18
(i) Show that f(x, y, z) = ln(1 + x2 + y2 + z2) is spherically
symmetric.
(ii) Show that a spherically symmetric function is constant on
spheres centered on the origin.

9.12.  VOLUME INTEGRALS (SPHERICAL POLAR)

In spherical polar coordinates (r, q, φ) the element of volume is

(Figure 9.6b):

dV = dr × rdq × r sin(q)dφ = r2 sin(q) drdqdφ.

We can use this to integrate a function over the volume of a solid

body. Instead of

∫ f dV = ∫∫∫ f ( x, y, z) dxdydz.
we can write

∫ f dV = ∫∫∫ f ( r, q, f ) r sin( q ) drdq df.

2

This is useful if the limits of integration are readily expressed in

(r, q, φ) coordinates.

Example 9.14 Find the integral of the function f(x, y, z) = z

over the volume of the hemisphere 0 ≤ r ≤ 1, 0 ≤ q ≤ p/2,
0 ≤ φ ≤ 2p.
For a volume that is part of a sphere we choose to work in
spherical polar coordinates. Write the function in spherical
polars:

f(x, y, z) = z = r cos(q).
 Multiple Integrals  •  327

Next, write down the limits for the integration: these are given as

0 ≤ r ≤ 1, 0 ≤ q ≤ p/2, 0 ≤ φ < 2p.

We can now write down the integral

∫ ∫ ∫ ( r cos( q ) ) ( r sin( q ) drdq df )

2p p /2 1
∫ z dV = 2
0 0 0
1 q= p / 2 2p
= ∫ r dr ∫ sin( q ) d ( sin( q ) ) ∫ df.
3
0 q= 0 0
1 p /2
 r  1
4

=    sin 2 ( q )  [ f] 0
2p

 0 
4 2 0
1 1 p
= × × 2p = .
4 2 4

Note the range of q required to cover the hemisphere. If we

choose points in the quadrant of the (x-z) plane, φ = 0, 0 ≤ q ≤ p/2
and swing this around through 2p we cover the complete hemisphere.
Rotating the arc φ = 0, 0 ≤ q ≤ p gives the complete sphere.

Exercise 9.19 What range of coordinates covers the whole

sphere of radius a?

Exercise 9.20 Find the integral of the function f(x, y, z) = z2 over

the sphere r ≤ 1.

9.13.  INTEGRALS OVER SPHERICAL SURFACES

Suppose we are given a function f(q, φ) at each point on some

part of the surface of a sphere, r = a = constant. We might need
to evaluate the integral of f over the surface. From Figure 9.7
the element of area is dA = a sin(q)dφ × adq = a2 sin(q) dqdφ. So

∫ f dA = ∫∫ f (q ,f ) a sin (q ) dq df ,
2
328  •  Mathematical Physics

z a sin(θ)dφ

adθ

dA
dθ
y
a
dφ
x
FIGURE 9.7: The spherical element of area for a sphere of radius r = a.

the surface integral of f over the spherical surface. Note the factor of
sin(q) in the integrand. The area is not a2dqdφ. You can see from the
figure that the same dqdφ near the poles has a smaller area than near
the equator. The factor of sin(q) takes this into account.

Example 9.15 Find the integral of the function f(x, y, z) = z

over the surface of the hemisphere r = a, 0 ≤ q ≤ p/2, 0 ≤ φ
≤ 2p.
Because the surface is part of a sphere it is convenient to use
spherical polar coordinates. Write the function in spherical
polars:

z = a cos(q).

Next, write down the element of area

dA = a2 sin(q) dqdφ.

Write down the limits of integration for each of the variables:

0 ≤ q ≤ p/2, 0 ≤ φ ≤ 2p.
 Multiple Integrals  •  329

Now we write down and evaluate the integral:

2p p /2
∫ z dA = ∫ ∫ ( a cos( q ) ) a sin( q ) dqdf
2
0 0
q= p / 2 2p
=a ∫3
sin( q ) d ( sin( q ) ) ∫ df
q= 0 0
p /2
1  1
= a3  sin 2 ( q )  [ f ] 0 = a3 × × 2p = p a3 .
2p

2 0 2

Exercise 9.21
(i) Find the integral of f(x, y, z) = 1 − z2 over the surface of the
sphere r = 1.
(ii) The moment of inertia of a uniform sphere of radius a
and unit surface density about an axis through its cen-
ter at the origin is given by the integral of a2 − z2 over
the surface of the sphere. Find the moment of inertia of
such a sphere.

9.14.  SOLID ANGLE

The angle, measured in radians, subtended by an arc of the unit

circle at its center is the length of the arc (Figure 9.8a). The angle, in
radians, subtended at a point O by an element of length dl, distance
r from O, is

d l cos( i)
dq =
r

where i is the angle between the normal to the element and the line
from O to the element (Figure 9.8b).
330  •  Mathematical Physics

i
q P
dl
1 r

q
dq
O O

(a) (b)
FIGURE 9.8: (a) The length of an arc of the unit circle about point O is equal to the angle
(in radians) at O subtended by the arc. (b) A tiny length element dl viewed at a distance r
from point O is inclined by an angle i from the circle of radius r centered on O. The angle is
dq = (dl cos i)/r.

z z

dA dA
P
i
r
O O
y y

x x

(a) (b)

FIGURE 9.9: (a) The area element dA at point P, a distance r from point O, with angle i be-
®
tween the normal to the area element and the line OP. (b) The solid angle subtended by dA
from O can be thought of as the area of the image of dA projected onto a unit sphere from
point O.

The solid angle, measured in steradians (sr), subtended by a cap

of the unit sphere at its center is the area of the cap (Figure 9.10).
For the purpose of visualization the solid angle subtended at O
by some surface S can be thought of as the area of the shadow
that S casts (when lit from O) on the surface of a large sphere,
center O, regarded as of unit radius. Equivalently, for the purpose
 Multiple Integrals  •  331

of calculation: the solid angle, in steradians, subtended at a point O

by an element of area, A, distance r from O is
d A cos( i)
dW = (9.12)
r2
where i is the angle between the normal to the element and the
line from O to the element (Figure 9.9a,b). The solid angle, Ω, sub-
tended by a surface is the integral of cos(i)/r2 over the surface, i.e.
cos( i)
=W ∫=
dW ∫
r2
dA. (9.13)

The concept of solid angle is important in many areas of physical

science. To evaluate the integral for solid angle in any particular case
you will need i and r as functions of q and φ. This will either be given
or deducible for the given surface.

Example 9.16 Find the solid angle subtended by a sphere at

its center.
We use the formula (9.13). We write down the element of area
at P = (r, q, φ) on the sphere assuming a radius r:

dA = r2 sin(q) dqdφ.

Now work out the angle i at P, the angle between the radius
vector and the normal to the surface: here i = 0, because the
normal is along the radius vector, and so cos(i) = 1. Write down
the limits of integration, chosen here to cover the whole sphere:

0 ≤ q ≤ p, 0 ≤ φ ≤ 2p.

Substitute in equation (9.13):

2p p r 2 sin (q )
=
W ∫=
dW ∫ ∫
0 0 r2
dq df
q =p 2p q =p
∫ sin (q ) dq ∫ df =
=
q =0
2p  − cos (q ) q = 0 =
0
4p.

Note that the result is independent of r.

332  •  Mathematical Physics

q P

O
y

x
FIGURE 9.10: The solid angle on a sphere.

In fact the sphere subtends solid angle 4p steradians at every

interior point, not just the center. (You need to know this.)

Exercise 9.22 What is the solid angle subtended by the sector

of a sphere 0 < φ < p/4, 0 < q < p/4 at its center? See Figure
9.11(a).

z
q=0
A
q = p/4

φ=0 φ = π/4

(a) (b)
FIGURE 9.11: (a) Exercise 9.22. (b) Exercise 9.24.
 Multiple Integrals  •  333

Exercise 9.23 What is the solid angle subtended by a portion

of a cylindrical surface of radius a, 0 ≤ φ ≤ p/2, −a ≤ z ≤ a at
the origin?

Exercise 9.24 A small hole of area A is cut in a sphere containing

a uniform radiation field which carries an energy per unit time
per unit normal area per unit solid angle I. Therefore the rate at
which energy is emitted into the surrounding space is

∫ IA cos (q ) dW
hemisphere

integrated over the hemisphere 0 ≤ q ≤ p/2. Show that equation

(24) gives pAI. See Figure 9.11(b).

Exercise 9.25 The pressure exerted by a gas of kinetic energy

density u is
1
u cos2 (q ) dW;
p ∫hemisphere
P=

assuming that u is independent of angle, perform the integration

to find P.

Exercise 9.26 Show that the solid angle subtended by a plane

rectangle of sides 2a, 2b at a point distance l along the normal
through its center can be written
b a l dxdy
W=∫ ∫ .
(x + y2 + l 2 )
−b a 2 32

9.15.  SKETCHING SURFACES

Sketching three-dimensional surfaces from a given formula is

hard: it can be simplified if the surface has some symmetry (e.g.
axial, cylindrical, spherical). In such cases the formula will simplify
in the appropriate coordinate system (found by trial and error).
334  •  Mathematical Physics

Example 9.17 Sketch the surface z = x2 + y2.

In cylindrical polar coordinates this simplifies to z = r2, i.e. z
is independent of φ, so the surface is axially symmetric (i.e. a
surface of revolution about the z-axis). We need therefore only
plot the curve of z against r for fixed φ and rotate the result.
In Figure 9.12(a), z has been plotted against r in the plane
φ = 0 (i.e. the (z - x) plane, so r = x there) giving a parabola. This
parabola has then been rotated about the z-axis to generate the
paraboloid represented by the given equation.

Example 9.18 Sketch the surface z = +(1 − x2 − y2)1/2 + 2.

z z

y y
x x

(a) (b)
FIGURE 9.12: (a) Example 9.17: z = x + y . (b) Example 9.18: z = (1 − x − y2)1/2 + 2. In both cases
2 2 2

the functions have axial symmetry (because they depend on x and y only through r2 = x2 + y2),
so we can draw the graph of the function z = f(0, y) on the (y, z) plane, and then rotate about
the z-axis.
 Multiple Integrals  •  335

Write Z = (z − 2); then the surface becomes

Z2 = 1 − x2 − y2, Z > 0.

In spherical polar coordinates based on the Z-axis (i.e. Z2 = r2

cos2(q) with Z > 0), this becomes r2 = 1; Z > 0. The surface
r2 = 1 is a sphere and the condition Z > 0 gives the hemisphere
in Figure 9.12(b).

Exercise 9.27 Sketch the surfaces

(i) z2 = x2 + y2,
(ii) z2 = 1 + x2 + y2.

Revision Notes
After completing this chapter you should be able to
• Compute a given multiple integral
• Insert limits for an integral over a given plane region
• Insert the limits corresponding to a change in the order of
integration over a plane region
• Define cylindrical and spherical polar coordinates
• W
 rite down the transformations between Cartesian
coordinates and cylindrical coordinates and spherical
polar coordinates and their inverses
• Write down the elements of area and volume in these
coordinate systems
• Use the appropriate elements of area and volume to
integrate over parts of spheres and cylinders and their
surfaces
• Define and compute a solid angle
• Recognize axial, cylindrical, and spherical symmetry
336  •  Mathematical Physics

9.16. EXERCISES

∫  ∫
x2 ydx  dy .
1 y
1. Evaluate 0 
0

2. Write down an expression for the integral

∫ ∫ f ( x, y) dx dy
2 y

1 1

with the order of integration reversed, after sketching the

area of integration in the (x, y) plane.
3. Draw a diagram to show the area of integration in
1
 x
f ( x , y) dy dx ,
∫ ∫
0 x 

and hence rewrite the integral in the form

∫ ∫ f ( x, y) dx dy with the correct limits of integration
inserted.
4. W
 rite down the following integral with the order of inte-
gration reversed (including the correct limits):
1 12
∫∫ f ( x , y) dxdy .
0 y 2

5. S
 how that the integral of the function f(x, y) = xy2 over the
region of the (x, y) plane bounded by the lines y = 0, x = 1
and the curve y2 = x is 2/21.
6. Let R be the tetrahedron given by x ≥ 0, y ≥ 0, z ≥ 0 and
x + y + z ≤ 1. Show that
1
∫∫∫ R
y dx dy dz =
24
where the integration is over the volume of the tetrahe-
dron.
7. D
 raw a diagram to show the relation between Cartesian
coordinates (x, y, z) and standard spherical polar coor-
dinates (r, q, φ). What does the function f(x, y, z) = (x2 +
y2)1/2/z become in spherical polar coordinates?
 Multiple Integrals  •  337

8. D
 raw a diagram showing the relation of standard spherical
polar coordinates (r, q, φ) to Cartesian coordinates (x, y, z).
Indicate on your diagram the standard cylindrical polar co-
ordinates (r, φ, z) and obtain the relation between (r, φ, z)
and (r, q, φ).
9. F
 ind the integral of the function cos2(q) cos(φ/4) over the
surface of the unit sphere centered at the origin.
10. F
 ind the integral of cos(2q) over the surface of the unit
sphere. (q is the angle between the radius vector and a
fixed direction.)
11. I f f = (x2+y2)z calculate the integral of f over the solid
cylindrical wedge of unit radius, unit height in x ≥ 0, y ≥ 0,
1 ≥ z ≥ 0.
 spherical cap of angle a is the surface of a sphere, center
12. A
O, cut off by a right circular cone of semi-angle a with
vertex at O. Find the solid angle subtended at O by such a
spherical cap. Hence show that a circular disc of angular
diameter 2a at a very large distance (compared with its
radius) subtends a solid angle at a point on its axis of
approximately pa2.
13. F
 ind the solid angle subtended by a circular disc of radius
3 cm at a point a distance 4 cm along the axis perpendicu-
lar to the disc through its center.
14. W
 rite down an expression for the solid angle subtended at
the origin by an element of surface dS with normal n at a
point with position vector r. Use this expression to estimate
the solid angle subtended by a coin of radius 1.295 cm (or
approximately 5/4 cm) at a distance of 1/2 m.
338  •  Mathematical Physics

9.17. PROBLEMS

1. W
 rite down the element of volume dV in spherical polar
coordinates (r, q , φ). Find
exp ( ir cos (q ) − r )
I=∫ dV ,
V r

where ∈ is a constant, i= − 1 , and the integral is taken

over all space. Hence show that lim I = 4p .
→0
2. A
 hemisphere of radius 2 is drilled from the pole to
the center to remove material within the surface
z = x2 + y2. Show that the volume remaining is that of a
unit sphere.
3. S
 how that the solid angle subtended at the origin by the
portion of the surface of a right circular cylinder of radius a
with axis along the z-axis between the planes x = z and
x = −z, x > 0 can be written
p2
 a cos(f ) a2 
W = ∫ ∫ dz  df .
− p 2  − a cos( f )
 ( a + z ) 
2 2 32

4. ( a) Show that the solid angle subtended by a circular disc

of unit area at a point on its axis unit distance from its cen-
ter is
1 p r dr
WD = 2p ∫ .
0
( 1+ r )2 3 2

(b) Show that the solid angle subtended at the origin by a

triangle with vertices at (0, 0, 1), (a, 0, 1), (a, aa, 1), (a > 0,
a > 0) is

a aa dy dx
W∆ = ∫∫ .
( 1+ x + y2 )
0 0 2 3 2
 Multiple Integrals  •  339

(c) From (b) deduce an expression for the solid angle, Ωn,
subtended by an n-sided regular polygon of unit area
at a point on its axis unit distance from its center, and
show that this can be written as

1 p px n dy
Wn = 2 n∫ ∫ 32
dx.
0 0
 y2 
(1 + x )
2 32
 1 + 
1 + x2 

By approximating the inner integral in the expression for
Ωn show that Ωn ≤ ΩD, and hence that a regular polygon
subtends a smaller solid angle than a disc of the same
area.
 CHAPTER

10
PARTIAL DERIVATIVES

In this unit we extend the rules for differentiating functions of a

single variable to functions of several variables. We shall then be
able to deal with functions of the three space coordinates and func-
tions of space and time. This step is essential if we are to be able to
formulate the laws of the physical sciences in more than one space
dimension in a general way. As an example, you are familiar with
how to find a maximum or minimum of a function of one variable,
but how do you find these for a function of two variables? Pictori-
ally this is equivalent to asking how to find the summit of a hill; you
might also need to find the steepest way down. We also consider the
multi-dimensional analogues of the chain rule for differentiation and
Taylor series for the expansion of a function about a given point, and
we introduce the notion of a differential.

10.1.  FUNCTIONS OF TWO VARIABLES

Let f(x, y) = x2 + xy2. This means that for each x and y (chosen inde-
pendently) there corresponds a value for f given by this expression.

Example 10.1 At (x, y) = (1, 0), f(x, y) = x2 + xy2 has the value
12 + 1 × 02 = 1; at (x, y) = (1, 2), f has the value 12 + 1 × 22 = 5.
342  •  Mathematical Physics

Exercise 10.1
(i) What value does the function f(x, y) = x3y + y2 have at
(x, y) = (−1, 3)?
(ii) What value does g(u, v) = u3v + v2 have at (u, v) = (−1, 3)?
(iii) And at (u, v) = (2, 4)?

y
z

(a) (b)

FIGURE 10.1:  The function z = (2 + x2 − y2 ) e1− x − y 4. (a) A surface plot. (b) A contour(level
2 2

curves) plot. The axes have been offset from the origin in (a) to avoid cluttering the plot.

The graph of a function of two variables f(x, y) is the set of all

points (x, y, z) for which z = f(x, y). We can think of this as a surface,
with z being the height of the surface that varies with x and y. Using
a computer we can render such a surface, as in Figures 10.1(a) and
10.2(a). Another way to visualize a function of two variables is to
plot the level curves, the curves of constant z = f(x, y). These are
just contour lines of the surface, as shown in Figures 10.1(b) and
10.2(b). It is usually best to plot contours for values of z that are
evenly spaced. Where the contour lines are close together, the func-
tion must be changing rapidly (increasing or decreasing over short
distances in the (x, y) plane in a direction perpendicular to the con-
tours); where the contours are widely separated the function must
be varying slowly (staying nearly constant).
 Partial Derivatives  •  343

y
z

(a) (b)
FIGURE 10.2:  The function z = sin(x) sin(y). (a) A surface plot. (b) A contour (level curves)
plot. The axes have been offset from the origin in (a) to avoid cluttering the plot.

Let f(x, y) = x2y + y2. Keeping y constant temporarily we find

df by the standard rules of differentiation:
dx
df
= 2 xy + 0 ( treating y as constant temporarily ) .
dx
df
If instead we keep x constant temporarily we can find by the
dy
standard rules:
df
= x2 + 2 y ( treating x as constant temporarily ).
dy

df
Example 10.2 If f(x, y) = x3/y what is (i) if y is temporarily
dx
df
kept constant and (ii) if x is temporarily kept constant?
dy
df 3 x2
(i) = (keeping y constant).
dx y
df x3
(ii) = − 2 (keeping x constant).
dy y
344  •  Mathematical Physics

Exercise 10.2 If f(x, y) = x2y, what is

df
(i) if y is temporarily kept constant,
dx
df
(ii) if x is temporarily kept constant?
dy

10.2.  PARTIAL DERIVATIVES

Notation: Instead of writing “df/dx if y is temporarily kept con-

stant” we use the abbreviated notation
 ∂f 
  .
 ∂x  y
We call this the partial derivative of f with respect to x (with y
held constant). The expression is read as “partial df by dx.” Similarly
“df /dy (keeping x temporarily constant)” becomes
 ∂f 
  ,
 ∂y  x
the partial derivative of f with respect to y, and is read as “partial df
by dy.” If f is already understood to be a function of x and y we can
abbreviate these further to
∂f ∂f
and .
∂x ∂y
The “curly” ∂/∂x signifies that everything other than x is to be
kept constant temporarily (i.e. during the process of differentiation).
There are a number of alternative notations; the following mean
exactly the same thing:
 ∂f  ∂f
  , , f, x , fx , ∂ x f .
 ∂x  y ∂x
Interpretation: The function f(x, y) can be represented by plot-
ting its value at (x, y) as the height of a surface above the point (x, y)
in the (x, y) plane, as in Figure 10.3. If we keep y constant, f is then
 Partial Derivatives  •  345

Exercise 10.3 If f(x, y) = x sin(y), what is

∂f
(i)
∂x
∂f
(ii)
∂y
∂f ∂f
(iii) (1,0 ) –i.e. evaluated at (x, y) = (1, 0)?
∂y ∂y

Exercise 10.4 If z(u, V) = u2 + V2, what is zu?

a function of x only (for that value of y); e.g. if y = 1 and f(x, y) =

x2+y2 then f = x2 + 1. This is represented as the cross-section of f in
the plane y = 1, i.e. the curve z = x2 + 1. (Figure 10.3 shows a similar
function.)

(x0, y0, z0) (x0, y0, z0)

z z

y y

x x
plane: y = y0 plane: x = x0

(a) (b)

FIGURE 10.3:  The partial derivative as the slope of a cross-section. (a) A function z = f(x, y)
plotted as a surface. A plane of constant y (= y0) is shown. The curve z = f(x, y0) is the
intersection of the surface with this plane. The gradient of the tangent of this curve at some
point P = (x0, y0) is the partial derivative of z with respect to x at P. (b) The same surface, plotted
with a plane of constant x (= x0). The curve z = f(x0, y) is the intersection of the surface with this
plane. The gradient of the tangent of this curve at P is the partial derivative of z with respect
to y at P.
346  •  Mathematical Physics

The function obtained from f(x, y) by keeping y = y0 constant

is the cross-section of the surface z = f(x, y) in the plane y = y0.
The partial derivative fx(x0, y0) is the slope of this curve at the point
(x0, y0).

10.3.  DIRECTIONAL DERIVATIVES

We have seen that (∂f/∂x)y is the slope of the surface z = f(x, y) in

the x-direction, i.e. parallel to the x-axis (see Figure 10.3). Similarly,
(∂f/∂x)x is the slope of the surface in the y-direction. The slope of the
surface in the direction making angle q (counterclockwise) with the
x-axis is
∂f ∂f
∇ û f cos (q )
= + sin (q ) . (10.1)
∂x ∂y
We call this expression the directional derivative of f. It uses a
special symbol, ∇ (the “del” operator) which is used for derivatives
and gradients in more than one dimension. Here, û is a unit vector
making an angle q with the x-axis. So ∇ûf is the gradient of f in the
direction of the vector û.
z

FIGURE 10.4:  Illustration of the directional derivative. A function z = f ( x , y ) is plotted as

a surface. At a point P the tangent plane along with tangents parallel to the x- and y-axes are
marked. Also marked is the tangent of the function z in some arbitrary direction following
a curve along the surface. The gradient of this is the directional derivative, which is a linear
combination of the gradients along each of the axes.
 Partial Derivatives  •  347

Example 10.3 Find the derivative of f = 1 − x2 − y2 in the

direction making angle p/4 (counterclockwise) with the x-axis
at the point (1, 2).
The derivative in a particular direction is given by equation
(10.1) so we first need the partial derivatives of f:
∂f ∂f
=
−2 x, =
−2 y.
∂x ∂y
Inserting these values into (10.1) we get
directional derivative = cos(p/4)(−2x) + sin(p/4)(−2y).
Since we want the directional derivative at the point (1, 2) we
let x = 1 and y = 2:

2 2
directional derivative = ( −2 ) + ( −4 ) =−3 2.
2 2

Exercise 10.5 Find the derivative of f = x3 + y2 in the direc-

tion making an angle 60° (counterclockwise) with the x-axis at the
(
point −1, 3 . )
There is a connection here to vectors that we will mention in
passing, but we will return to this idea when we study vector calcu-
lus. If û is a unit vector at an angle q with the x-axis then û = cos(q)i +
sin(q)j, and its components are (cos q, sin q). We can also define a
vector whose components are the partial derivatives of f. We give
this the special symbol ∇f, so
∂f ∂f  ∂f ∂f 
∇f = i+ j=  ,  .
∂x ∂y  ∂x ∂y 
Now, the dot product of û and ∇f gives the directional derivative
∂f ∂f
∇ û f = û ⋅ ∇f = cos (q ) +sin (q ) ,
∂x ∂y
which is the same as we had before.
348  •  Mathematical Physics

10.4.  FUNCTIONS OF MANY VARIABLES

The notion of a partial derivative extends to functions of more

than two variables (except that we cannot now draw a picture).

Example 10.4 Let (x, y, z) be independent variables and let

r(x, y, z) = (x2 + y2 + z2)1/2. Find (∂r/dx)y,z, the derivative of r
w.r.t. x keeping y and z constant. (We usually write this as just
∂r/dx if it is understood that r is a function of x, y, z.) Note
that, by assertion, z is here an independent variable, on the
same footing as x and y, whereas in Section 10.2, z was used
for a function of x and y. Care is needed to distinguish the
dependent and independent variables in a given problem.
We are to differentiate r with respect to the variable x.
Thus we shall treat the other variables as constants for now.
If this is not clear consider the following: y2 and z2 are to be
treated as constant so use a notation that makes this clear, say
y2 + z2 = a2. Then r = (x2 + a2)1/2 which can be differentiated
with respect to x, as a function of a function, in the usual way
(i.e. by the chain rule):
dr 1 2
= ( x + a2 ) ⋅ 2 x.
−1 2

dx 2

Now replace a2 by y2 + z2 and we have that

∂r 1 2x x
= = .
∂x 2 ( x + y + z )
2 2 2 12
r
With more practice you can simply imagine y and z to be
constants and write the calculation directly.

Exercise 10.6 If f(x, y, z) = (x2 + y2 + z2)−1/2 write down

(i) ∂f/∂x,   (ii) ∂f/∂y,   (iii) ∂f/∂z.

∂g
Exercise 10.7 If g(t, u, v, w) = t + u2v3w4, what is ?
∂v
 Partial Derivatives  •  349

10.5.  HIGHER DERIVATIVES

Let f(x, y) = 3x3y2. We have ∂f/∂x = 9x2y2 and ∂f/∂y = 6x3y. We

can now differentiate again, apparently in four ways. We have
∂  ∂f  ∂
=   =
∂x  ∂x  ∂x
( 9 x2 y2 ) 18 xy2 .
Similarly
∂  ∂f  ∂
=   =
∂y  ∂x  ∂y
( 9 x2 y2 ) 18 x2 y.

We also have that

∂  ∂f  ∂
=   =
∂x  ∂y  ∂x
( 6 x3 y) 18 x2 y,
and
∂  ∂f  ∂
=   =
∂y  ∂y  ∂y
( 6 x3 y ) 6 x3 .
We write these second derivatives more compactly as
∂ f 2
∂2 f ∂2 f ∂2 f
= =
18 xy 2
, =
18 x 2
y, 18 x 2
y,=and 6 x3 .
∂x2 ∂x∂y ∂y∂x ∂y2
If we define the mixed second derivative fxy to mean (fx)y
(i.e. ∂/∂y(∂f/∂x)) we can write even more succinctly fxx = 18xy2,
fxy = 18x2y etc.
Exercise 10.8 If f(x, y, z) = (x2 + y2 + z2)1/2, find
(i) fx,   (ii) fy,   (iii) fxx,   (iv)fyx,   (v) fxy,   (vi) fyy.

Note in the above example and in Exercise 10.8 we have that

∂2 f ∂2 f
=
= equivalently f yx f xy .  (10.2)
∂x∂y ∂y∂x
This can be shown to be the case for reasonably smooth func-
tions (those normally occurring in physics): the order in which par-
tial derivatives are taken does not matter.
350  •  Mathematical Physics

Exercise 10.9 Let f = ex log(x + y). Find fx and fy and verify that
fxy = fyx.

10.6.  FUNCTION OF A FUNCTION

We now examine the rule for differentiating a “function of a

function” (recall Section 1.3) for functions of many variables. If we
have some function f(u) where u(x, y) is itself a function of (x, y) we
wish to find the rule that gives us ∂f/∂x etc. Here, f can be written
either as a function of just u, or as a function of both x and y together.
If u were a function only of x the chain rule would give
df df du
= .
dx du dx
But u(x, y) is a function of x and y. However, we hold y constant
when we compute the partial derivative with respect to x. So

∂f df ∂u
= .  (10.3)
∂x du ∂x
where the curly ∂’s remind us that y is being held constant. Other-
wise there is nothing new here. (Notice we do not need the curly ∂’s
for df/du since f can be written as a function of u only.)

Example 10.5 Let r(x, y, z) = (x2 + y2 + z2)1/2, and let f = 1/r.

Find ∂f/∂x.
To obtain ∂f/∂x, we treat y and z as constants. We apply
∂f df ∂r
= .
∂x dr ∂x
We have from Example 10.4 that
∂r x
= ,
∂x r
 Partial Derivatives  •  351

and we can see that

df 1
= − 2.
dr r
So, these give
∂f 1 x x
=− 2⋅ =− 3.
∂x r r r

This example is important in calculations that involve the inverse

square law.
Exercise 10.10
∂r
(i) Let r(x, y, z) = (x2 + y2 + z2)1/2 and let f = 1/rn. Find .
∂x
∂2 f ∂2 f
(ii) Let f = 1/r; find and write down expressions for
∂x2 ∂y2
∂2 f
and .
∂z2

Let f(u) be a given function of u and let u(x, y, z) be a given func-

tion of (x, y, z). If u = u(x, y, z) is substituted for the variable in f(u),
then we have seen that
∂f df ∂u
= ,
∂x du ∂x
but in general
∂2 f d2 f ∂2 u
≠ × ;  (10.4)
∂x2 du2 ∂x2
i.e. we cannot simply change the first derivatives for second deriva-
tives and expect the equation to be valid.
Exercise 10.11 Demonstrate in equation (10.4) by means of
an example (of your choice) for the functions f(u) and u(x, y, z).
Establish the correct formula
2
∂ 2 f d 2 f  ∂u  df ∂ 2 u
= ×  + × ,
∂x2 du2  ∂x  du ∂x2
and verify that this is correct for your example.
352  •  Mathematical Physics

10.7.  IMPLICIT DIFFERENTIATION

We can partially differentiate both sides of an equation using the

“function of a function” rule, and then rearrange to find the partial
derivative that we seek (recall Section 1.5). Apart from the use of
curly ∂ there is nothing new here.

Example 10.6 Let r(x, y, z) be given by r2 = x2 + y2 + z2.

∂r
Find ∂x .

First, differentiate the left side, which is a function (r2) of a

function r(x, y, z), with respect to x:

∂ 2 d ( r 2 ) ∂r ∂r
=
∂x
( r ) = 2r .
dr ∂x ∂x

On the right side y and z are constants when differentiating

partially with respect to x; thus

∂ 2
∂x
( r ) = 2 x + 0 + 0.
Comparing these two we see that
∂r x
= .
∂x r

Exercise 10.12 Let sin2(w) = x2 + y2 + z2. Show that

∂w 2x
= .
∂x sin ( 2w )
 Partial Derivatives  •  353

10.8.  DERIVATIVE WITH RESPECT TO A PARAMETER

The following is a simple but important application of partial dif-

ferentiation. We have a function of two variables, f(x, t), and we wish to
know the derivative with respect to x of its integral with respect to t. To
differentiate an integral with respect to a parameter the rule is

d
dx (∫
a
b
)
f ( x, t ) dt = ∫
a
b ∂f ( x, t )
∂x
dt (10.5)

On the left side we imagine performing the integral to obtain a

function of x which we then differentiate (giving another function
of x).On the right side we imagine differentiating the integrand first
with respect to x and then integrating with respect to t to obtain a
function of x. Equation (10.5) says that the order in which we do the
differentiation does not matter: we get the same result either way.
(The partial derivative is required on the right of equation (10.5) to
indicate that t is held temporarily constant for the differentiation.
On the left the integral is a function only of x anyway, so we write this
as an ordinary derivative.)
Given two ways of evaluating the same thing we are free to
choose the easier. This enables us to find quickly some useful inte-
grals that would otherwise be difficult.

d 1 sin ( tx )
dx ∫0
Example 10.7 Find dt .
t
Using equation (10.5) we have that

d 1 sin ( tx ) 1 ∂  sin ( tx ) 
∫
dx 0 t
dt = ∫
0 ∂x

 t 
 dt
1
= ∫ cos ( tx ) dt
0

1 t =1
= sin ( tx )  t = 0
x
sin ( x )
= .
x
354  •  Mathematical Physics

Example 10.8 From the equation

1 1
∫ t dt
= x > −1,
x
, for
0 x +1
deduce that
( −1)m m!
∫ t ( log ( t ) ) dt
1 m
= x
= , for m 1,2,.
0
( x + 1 ) m +1
First, differentiate the right side of the given equation:

d 1 x d 1  1
dx ∫0
t dt =  = − .
dx  x + 1  ( x + 1 )2
Next we differentiate the left side (with respect to x) using
equation (10.5). (Recall that you learned how to differentiate
ax in Section 2.7.)
d 1 x 1 ∂
∫
dx 0
t dt = ∫
0 ∂x
( t x ) dt
= ∫ t x ( log ( t ) ) dt.
1

Therefore
1
∫ t ( log ( x ) ) dt = − ( x + 1)
1
x
2
.
0

If we differentiate both sides again we get another factor of

log(t) from the differentiation of tx with respect to x:
( −1)( −2 )
∫ t ( log ( t ) )
1 2
x
dt = − ,
0
( x + 1 )3
and hence repeating this for m applications gives the desired
result

( −1)m m!
∫ t ( log ( t ) ) dt
1 m
= x
= , for m 1,2,.
0
( x + 1 ) m +1
 Partial Derivatives  •  355

p2 d p2
∫ t sin ( tx ) dt = − cos ( tx ) dt .
dx ∫0
Exercise 10.13 Show that
0

p2 sin ( p x 2 ) p2
Given that ∫ cos ( tx ) dt = t sin ( t ) dt . , find ∫
0 x 0

(This is an alternative to integration by parts for evaluating this

integral.)

∞ 1
Exercise 10.14 Given that ∫ 0
e−λt dt =
λ
, use the method of
∞
Example 10.8 to find ∫0
t n e−λt dt , for n = 1,2,. . . .
∞
Hence show that ∫0
x n e− x dx = n! and show that 0! = 1 using this
formula. (Thus, this formula agrees with the usual definition of n!
for n a positive integer, and gives a meaning to cases when n is not
a positive integer.)

10.9.  TAYLOR EXPANSION ABOUT THE ORIGIN

One of the most important applications of partial derivatives in

physics is the approximation of a function by a simple formula near a
given point. Suppose we are given the function f(x,y). We can evalu-
ate f and its partial derivatives at the point P = (0, 0). Let
∂f  ∂f 
=
( 0,0 ) , f x ( P )  =
f ( P ) f= ( 0,0 ) , fy ( P )   ( 0,0 ) ,
 ∂x   ∂y 

 ∂2 f 
f xx ( P ) =  2  ( 0,0 ) ,
 ∂x 
(10.6)
and so on. The multiple variable form of Taylor’s theorem says that
we can approximate the function f(x, y) near the point (0, 0) by a
finite number of terms from the following series:
f ( x, y) = f ( P ) + xf x ( P ) + yf y ( P )

1
 +  x2 f xx ( P ) + 2 xyf xy ( P ) + y2 f yy ( P ) 
2!
1
+  x3 f xxx ( P ) + 3 x2 yf xxy ( P ) + 3 xy2 f xyy ( P ) + y3 f yyy ( P )  + … .
3!
(10.7)
356  •  Mathematical Physics

Equation (10.7) is the Taylor expansion of f(x, y) about (0, 0).

Taylor series centered on the origin, like these, are often called
Maclaurin series. We discussed both Maclaurin and Taylor series for
functions of a single variable in Sections 2.2 and 2.3.
The coefficients in each group of terms follow the binomial pat-
tern (Section 2.1), although in practice we almost never need the
higher order ones. Note that however complicated our initial func-
tion f, the Taylor series involves powers of x and y only (in practice
usually up to quadratic or cubic at most). This is the point of the
calculation. If you end up with anything more complicated you have
made a mistake.

Example 10.9 Find the Taylor series expansion up to second

order of the function f(x,y) = sin(xy) about the point P = (0, 0).
We evaluate each of the required quantities at the given point
(0, 0) in turn, starting with the function itself:
f(P) = f(0, 0)=sin(0) = 0.
Next, each of the first derivatives at P:
fx = y cos(xy), so fx(P) = 0 × cos(0) = 0,
fy = x cos(xy), so fy(P) = 0 × cos(0) = 0.
Then each of the second derivatives. Note that we compute all
the derivatives and only then substitute values for x and y at P:
fxx = −y2 sin(xy), so fxx(P) = 0 × sin(0) = 0,
fyy = −x2 sin(xy), so fyy(P) = 0 × sin(0) = 0.

We can compute either of fxy or fyx (i.e. differentiate either fx or

fy) since these are equal:
fxy = cos(xy) − xy sin(xy), so fxy(P) = cos(0) − 0 × sin(0) = 1.
Finally, using the definition of the Taylor series (10.7)
1 2
sin ( xy ) ≈ 0 + x ⋅ 0 + y ⋅ 0 +  x ⋅ 0 + 2 xy ⋅ 1 + y2 ⋅ 0  + 
2! 
and so, sin(xy) ≈ xy + . . . to second order near (0,0).
 Partial Derivatives  •  357

The result agrees with the first term of the series expansion for
sin(u) with u = xy, as, of course, it must. You will find it extremely
helpful if you set out your work systematically like this.
Exercise 10.15 Let f(x,y) = 1 + exey and P = (0, 0). Find fx(P),
fy(P), fxx(P), fxy(P), fyy(P), and hence obtain the Taylor expansion of
f about the origin to second order.

10.10.  EXPANSION ABOUT AN ARBITRARY POINT

Suppose we are given the function f(x, y). We can evaluate f and
its partial derivatives at the point P = (x0, y0). Let
∂f
= ( x0 , y0 ) , f x ( P )
f ( P ) f= ( x0 , y0 ) ,
∂x
and so on. Taylor’s theorem says we can approximate the function
f(x,y) near the point (x0,y0) by a finite number of terms from the fol-
lowing series:

f ( x, y=
) f ( P ) + ( x − x0 ) f x ( P ) + ( y − y0 ) fy ( P )
1
+ ( x − x0 ) f xx ( P ) + 2 ( x − x0 ) ( y − y0 ) f xy ( P ) + ( y − y0 ) f yy ( P ) 
2 2
 2!  
+… (10.8)

Equivalently, we can write this as

f ( x0 + h, y0=
+ k) f ( x0 , y0 ) + hf x ( x0 , y0 ) + kf y ( x0 , y0 )
1 2
 + h f xx ( x0 , y0 ) + 2 hkf xy ( x0 , y0 ) + k2 f yy ( x0 , y0 )  + ….
2!  
(10.9)
where h = x − x0 and k = y − y0 are the displacements from P. Equa-
tion (10.8) (or (10.9)) is the Taylor expansion of f(x, y) about the
point P = (x0, y0). It is easier to see the form of the series if we leave
out explicit reference to the point P.
f ( x0 + h, y0 + k) = f + hf x + kf y
1 
+ h2 f xx + 2 hkf xy + k2 f yy 
2! (10.10)
1
+ h3 f xxx + 3h2 kf xxy + 3hk2 f xyy + k3 f yyy  + .
3!
358  •  Mathematical Physics

Example 10.10 Find the Taylor expansion of f = sin(xy) about

the point P = (p/2, 1).
Evaluate each of the coefficients in the Taylor series in turn at
the point P:
f(P) = sin(p/2) = 1,
fx = y cos(xy), so fx(P) = cos(p/2) = 0,
fy = x cos(xy), so fy(P) = (p/2) cos(p/2) = 0,
fxx = −y2 sin(xy), so fxx(P) = − sin(p/2) = −1,
fxy = cos(xy) − xy sin(xy),
p
so f xy ( P ) =cos ( p 2 ) − ⋅ 1 ⋅ sin ( p 2 ) =−p 2 ,
2
fyy = −x sin(xy), so fyy(P) = −(p/2)2 sin(p/2) = −p2/4.
2

Substitute into the general form of the Taylor series (10.8):

sin ( xy ) = 1 + ( x − p 2 ) × 0 + ( y − 1 ) × 0
1
+ ( x − p 2 ) ( −1 ) + 2 ( x − p 2 )( y − 1 )( − p 2 )
2

2! 
+ ( y − 1 ) ( − p 2 4 )  + ,
2


and so
1 p2 2
sin ( xy ) =1 − ( x − p 2 ) + p ( x − p 2 )( y − 1 ) + ( y − 1 )  + .
2

2 4 
The result is a power series (or a quadratic if we stop after the
second order terms). Near the point (p/2, 1) we can put X = x − p/2
and Y = y − 1 and write the series more succinctly as
1 p p2
sin ( xy ) =
1 − X 2 − XY − Y 2 +.
2 2 8

Exercise 10.16 Let f(x,y) = x3y4 + x4y3. Find the partial deriva-
tives of f up to second order at (1,1) and hence obtain the Taylor
expansion of f about the point (1, 1) to second order.
 Partial Derivatives  •  359

10.11.  FUNCTION OF MORE THAN TWO VARIABLES

Suppose we are given the function f(x, y, z). We can evaluate f

and its partial derivatives at the point P = (x0, y0, z0). Write
∂f
= ( x0 , y0 , z0 ) , f x ( P )
f ( P ) f= ( x0 , y0 , z0 ) etc.
∂x
Taylor’s theorem says that we can approximate the function
f(x, y, z) near the point (x0, y0, z0) by a finite number of terms from
the following series:
f ( x, y, z=
) f ( P ) + ( x − x0 ) f x ( P ) + ( y − y0 ) fy ( P ) + ( z − z0 ) fz ( P )
1
+ ( x − x0 ) f xx ( P ) + 2 ( x − x0 ) ( y − y0 ) f xy ( P )
2

 2! 
+2 ( x − x0 )( z − z0 ) f xz ( P ) + ( y − y0 ) f yy ( P )
2

+2 ( y − y0 ) ( z − z0 ) f yz ( P ) + ( z − z0 ) f zz ( P )  + .
2

(10.11)
Extensions to more variables follow a similar pattern, but you
are unlikely to meet them.
To sum up then, if we need to find the behavior of a function
near a given point we can approximate the function by a polyno-
mial, usually a quadratic, which is much simpler than the original
function.

10.12.  STATIONARY POINTS

We return to the problem of finding stationary points (from

Section 1.8), now considering functions of more than one variable.
In two dimensions a stationary point can be a maximum, mini-
mum or saddle point. At a stationary point, by definition, f(x, y) is
approximately constant (to first order) (Figure 10.5). A necessary
condition for P = (x0, y0) to be a stationary point is that
∂f ∂f
= ( x0 , y0 ) 0,=
and ( x0 , y0 ) 0,  (10.12)
∂x ∂y
360  •  Mathematical Physics

or, in more compact notation, fx(P) = fy(P) = 0. If the derivatives

along the two directions are both zero, then the derivative in any
direction is zero (by Equation 10.1); i.e. the tangent plane at P is
horizontal.
(x0, y0)
z

x
FIGURE 10.5:  Near a stationary point the function f(x, y) is constant to first order (in all
directions).

Example 10.11 Find the stationary points of the function

f(x, y) = x2y + 2xy2 + 6xy.
First, calculate fx and fy and set them equal to 0:
fx = 2xy + 2y2 + 6y = y(2x + 2y + 6) = 0 and
fy = x2 + 4xy + 6x = x(x + 4y + 6) = 0
for a stationary point. Now solve the two equations for x and y.
One solution is (x, y) = (0, 0). If this is not the case then
(i) y ≠ 0 and so 2x + 2y + 6 = 0, or
(ii) x ≠ 0 and so x + 4y + 6 = 0.
If y = 0 then (ii) gives x = −6, while if x = 0 then (i) gives
y = −3. If x ≠ 0, and y ≠ 0 then (i) and (ii) are two simultaneous
equations for x and y which can be solved to give x = −2, y = −1.
So we have found four stationary points
(x, y) = (0, 0), (−6, 0), (0, −3), and (−2, −1).
These are the stationary points of f.
 Partial Derivatives  •  361

Exercise 10.17 Find the stationary points of the functions

(i) f(x, y) = x3 − 2x2 − 4x − y2 + 6y,
(ii) f(x, y) = xy2 − 2y − x.

In order to identify the type of stationary point we use a second

derivative test (compare with the test used for functions of a single
variable in Section 1.8). We calculate the three second derivatives at P:
∂2 f ∂2 f ∂2 f
=
a f=
xx , =
b f=xy , =
c f=yy
∂x2 ∂x∂y ∂y2
(There is no need to compute fyx because it is the same as fxy.)
We then compute

D = ac − b2 = fxxfyy − (fxy)2.  (10.13)

If D > 0 then the point is a maximum or minimum; if not then

it is a saddle point or requires more detailed investigation to iden-
tify its nature (see Figure 10.6). The second derivative test works as
follows:
• If D > 0 and fxx > 0 then P is a minimum (fyy will also be > 0).
• If D > 0 and fxx < 0 then P is a maximum (fyy will also be < 0).
• If D < 0 then P is a saddle point (a maximum in one direc-
tion and a minimum in another, like the top of a mountain
pass).
• If D = 0 then further investigation is needed.

maximum saddle point

minimum

FIGURE 10.6:  Three different types of stationary point for a function of two variables.
Maximum (left), minimum (center) and saddle point (right). Near a stationary point the
function f(x, y) is constant to first order. As you can see from the figures, stationary points
are important for the analysis of stability.
362  •  Mathematical Physics

Example 10.12 Find the stationary point of the function

f(x, y) = 3x2 − 2xy + y2 and determine its nature.
First we compute fx and fy to find the stationary point(s)

fx = 6x − 2y and fy = −2x + 2y.

At a stationary point fx = 0 and fy = 0 so from the second

equation x = y and from the first 6x − 2x = 4x = 0 gives x = y = 0;
hence P = (0, 0) is the only stationary point. To determine
whether this is a maximum or minimum or saddle point we
compute the second derivatives and D.

fxx = 6; fxy = −2; fyy = 2;

fxx(P) = 6 > 0; and fyy(P) = 2 > 0;
D = f xx ( P ) f yy ( P ) − f xy2 ( P ) = 12 − 4 > 0.

So P = (0, 0) is a minimum.

Example 10.13 Find and determine the nature of the

stationary points of the function

f(x, y) = x2y + 2y2x + 6xy.

The stationary points have been found in Example 10.11; they

were (0, 0), (−6, 0), (0, −3), (−2, −1). So next we compute the
second partial derivatives:

 fx = 2xy + 2y2 + 6y, fy = x2 + 4xy + 6x,

fxx = 2y,  fxy = 2x + 4y + 6,  fyy = 4x.

Now evaluate them at the stationary points:

P = (0, 0): fxx(P) = 0, fxy(P) = 6, fyy(P) = 0.
Therefore D = fxx(P)fyy(P) − fxy(P)2 < 0 and (0, 0) is a saddle
point.
 Partial Derivatives  •  363

P = (−6, 0): fxx(P) = 0, fxy(P) = −6, fyy(P) = −24.

Therefore D = fxx(P)fyy(P) − fxy(P)2 < 0 and (−6, 0) is a saddle
point.
P = (0, −3): fxx(P) = −6, fxy(P) = −6, fyy(P) = 0.
Therefore fxx(P)fyy(P) − fxy(P)2 < 0 and (0, −3) is a saddle point.
P = (−2, −1): fxx(P) = −2, fxy(P) = −2, fyy(P) = −8.
Therefore D = fxx(P)fyy(P) − fxy(P)2 > 0 and so P is either a
maximum or a minimum. Since fxx(P) and fyy(P) are negative,
(−2, −1) is a maximum.

Exercise 10.18
(i) Find and determine the nature of the stationary points of
the function

f(x,y) = x2 − 2x − y3 + y2 + 8y.

(ii) Show that the function

f(x,y) = −[x2y − (x + 1)]2 − (x2 − 1)2

has maxima at (1, 2) and (−1, 0).

(The function in Exercise 10.18(ii) has two peaks and no passes.

This is a counter-example to the intuitive but incorrect idea that two
mountain peaks must be separated by a pass. In effect the pass here
has been shifted to infinity.)

Examination Using Taylor Series

Now we have seen how to find and test stationary points we can
examine how this works using Taylor’s theorem. Suppose we are given
a function f(x, y). We can use Taylor’s theorem to find the station-
ary points (maxima and minima) of f as follows. Suppose P = (x0, y0)
is a stationary point and expand f as a Taylor series about (x0, y0):

f(x,y) ≈ f(P) + (x − x0)fx(P) + (y − y0)fy(P) + ...

364  •  Mathematical Physics

At a stationary point f is constant to first order; hence (x − x0)

fx(P) + (y − y0)fy(P) = 0 for all (x, y) near (x0, y0). Therefore P is a
stationary point if fx = 0 and fy = 0 at P.
Just as was the case for functions of a single variable
(Section 1.8) we can classify stationary points of a function of two
variables as follows:
• (x0, y0) is a maximum if f(x, y) < f(x0, y0) for all (x, y) near
(x0, y0).
• (x0, y0) is a minimum if f(x, y) > f(x0, y0) for all (x, y) near
(x0, y0).

If neither condition holds, i.e. if for some (x, y), f(x, y) > f(x0, y0)
and for some (x, y), f(x, y) < f(x0, y0), then
• (x0, y0) is a saddle point (see Figure 10.6).

To translate these conditions into a usable test we use Taylor’s

theorem to tell us what f looks like near P = (x0, y0). This is clearer if
we use h = x − x0, k = y − y0 to represent small displacements from
x0, y0 and let a = fxx(P), b = fxy(P), c = fyy(P). Then Taylor’s theorem
up to second order is
1
f ( x, y ) = f ( P ) +  h2 f xx ( P ) + 2 hkf xy ( P ) + k2 f yy ( P )  + 
2 
1
= f ( P ) +  ah2 + 2 bhk + ck2  (10.14)
2
The linear terms vanish by equation (10.12). We can rewrite
this as

1 1
f ( x, y ) − f ( P )=  ah2 + 2 bhk + ck2 = Q (10.15)
2 2
with Q = ah2 + 2bhk + ck2.
Now, if P is a maximum this means f(x, y) < f(P) near P, and so
Q < 0; or if P is a minimum then f(x, y) > f(P) near P and so Q > 0,
for all values of the displacements h and k. In both these cases Q ≠ 0
which means the quadratic equation

Q = ah2 + 2bhk + ck2 = 0

 Partial Derivatives  •  365

can have no real solution except h = k = 0. This is the case if b2 − ac < 0,

or equivalently, D = ac − b2 > 0. This is therefore the condition for a
maximum or a minimum.
Hence the stationary point P = (x0, y0) is

1. A minimum if
fxx(P) > 0, and fyy(P) > 0, and f xx ( P ) f yy ( P ) − f xy2 ( P ) > 0

2. A maximum if
fxx(P) < 0, and fyy(P) < 0, and f xx ( P ) f yy ( P ) − f xy2 ( P ) > 0

3. A saddle point if
f xx ( P ) f yy ( P ) − f xy2 ( P ) < 0 (if fxx(P), fyy(P) have opposite signs)

If the second derivatives, fxx, fyy and fxy, are all zero one would
need to investigate the behavior of the function in more detail to
correctly identify the type of stationary point.

10.13.  DEFINITION OF THE TOTAL DIFFERENTIAL

Given a function f(x, y), its total differential, df, is defined by

∂f ∂f
df = dx + dy equivalently df =
f x dx + f y dy.  (10.16)
∂x ∂y
Since this is a definition it is not necessary to ask what it means 
we shall simply show that it is useful.

Example 10.14 Find the total differential of the function

f(x, y) = xy3.
We calculate fx = y3 and fy = 3xy2. So from the definition (10.16)
we have
df = y3dx + 3xy2dy.
366  •  Mathematical Physics

Exercise 10.19 Find the total differential of

(i) f(x, y) = x3y + x/y,
(ii) f(u, v) = u2/v.

10.14.  EXACT DIFFERENTIALS

Now consider the converse: given a(x, y) and b(x, y) can we write
adx + bdy = df for some f ? If such an f exists we say that adx + bdy is
an exact differential (it is also called a perfect differential). It is often
useful to know if adx + bdy is an exact differential, because, if it is,
then there exists a curve f(x, y) = constant, such that (a, b) is normal
to the curve at each point. As an example, let a = 3x2y + 1/y and
b = x3 − x/y2; does there exist an f such that

df = (3x2y + 1/y)dx + (x3 − x/y2)dy?

If you are good at this sort of thing you can guess (with the help
of Exercise 10.19) that f = x3y + x/y will do. Can we find a more
mechanical method?
Suppose we can find an f such that

df = adx + bdy.

But then, given this f, we have

df = fxdx + fydy,

by equation (10.16). Thus, if both are to be valid, a = fx, b = fy. Hence,

ay = fxy, bx = fyx. Since fxy = fyx, this means that if f exists then

ay = bx.

Although we shall not prove it, this is in fact a necessary and suf-
ficient condition for the differential to be exact.
 Partial Derivatives  •  367

Example 10.15 Suppose a = (3x2y + 1/y) and b = (x3 − x/y2).

Show that adx + bdy is indeed an exact differential.
We check if ay=bx. We have ay = 3x2 − 1/y2, bx = 3x2 − 1/y2 = ay,
so adx + bdy is an exact differential.

A differential adx + bdy is said to be inexact if this is not true,

i.e. there exists no function f(x, y) with this differential. For example,
2ydx + 3xdy is an inexact differential. If it were an exact differential
there would be some function f(x, y) whose partial derivatives are
∂f/∂x = 2y and ∂f/∂y = 3x. But we can integrate each of these to find
f(x, y). If we integrate the first with respect to x we get f(x, y) = 2xy +
g(y), where g(y) is some function of y only; if we integrate the second
with respect to y we get f(x, y) = 3xy + h(x), where h(x) is some func-
tion of x only. These two are inconsistent (for any choice of g(y) and
h(x)); therefore no such f(x, y) exists. You can prove to yourself that
this differential does not meet the criterion above (i.e. that ay ≠ bx).
The difference between exact and inexact differentials is important
in several areas of physics, most notably thermodynamics.

Exercise 10.20 If a = x, b = xy, show that adx + bdy is not an

exact differential.

Exercise 10.21 If a = ln(yx3/2) and b = x/y, show that adx + bdy is

an exact differential. (This arises in thermodynamics, with, in the
standard thermodynamic notation, x = T, y = V, a = S and b = P.)

10.15.  THE CHAIN RULE

Recall (Section 1.3) the rule for differentiating a “function of a

function” of a single variable: given f(u) and u = g(x) we find df/dx
by the chain rule:
df df du
= . (10.17)
dx du dx 
368  •  Mathematical Physics

Now we consider the case for a function of many variables.

Given f(u, v) and u = u(x, y), v = v(x, y) we find ∂f/∂x and ∂f/∂y by
the chain rule:
∂f ∂f ∂u ∂f ∂v
= + ,  (10.18)
∂x ∂u ∂x ∂v ∂x
∂f ∂f ∂u ∂f ∂v
= + .  (10.19)
∂y ∂u ∂y ∂v ∂y
(You can remember this by thinking of it as applying equation
(10.17) to each variable in turn and summing the results.) Note that,
for example, in equation (10.18), ∂f/∂x is found keeping y constant
(so f is considered as a function of x and y) and ∂f/∂u is found keep-
ing v constant (so f is considered as a function of u and v).

Example 10.16 Given f(u, v) with u = xy and v = (x − y) find fx

and fy in terms of fu and fv.
By the chain rule we have
∂f ∂f ∂u ∂f ∂v
= + .
∂x ∂u ∂x ∂v ∂x
However
∂u ∂v ∂f ∂f ∂f
= y and = 1, so = y + .
∂x ∂x ∂x ∂u ∂v
Similarly
∂u ∂v ∂f ∂f ∂f
=
x, and =
−1, so =
x − .
∂y ∂y ∂y ∂u ∂v

Example 10.17 Given f(x, y) with x = r cos(q) and y = r sin(q)

find fr in terms of fx and fy.
By the chain rule we have
∂f ∂f ∂x ∂f ∂y
= + .
∂r ∂x ∂r ∂y ∂r
 Partial Derivatives  •  369

But
∂x ∂y
= cos (q ) , and
= sin (q ) ,
∂r ∂r
and so we have that
∂f ∂f ∂f
= cos (q ) + sin (q ) .
∂r ∂x ∂y

You will find that the chain rule is important wherever it is neces-
sary to use different coordinate systems (e.g. Cartesians and polars).
Exercise 10.22 Complete Example 10.17 by finding fq in terms
of fx and fy. Hence obtain expressions for fx and fy in terms of fr
and fq . Deduce that if f = rq then
x  x y
( x 2 + y2 ) .
12
q cos (q ) − sin (q ) =cos−1   − , where r =
fx =
r r r

10.16.  CONSISTENCY WITH THE CHAIN RULE

Given f(u, v) and u = u(x,y), v = v(x,y) we can form the following

differentials:

df = fudu + fvdv  (10.20)

du = uxdx + uydy  (10.21)
dv = vxdx + vydy.  (10.22)

Substituting equations (10.21) and (10.22) into equation (10.20)

we obtain

df = fu(uxdx + uydy) + fv(vxdx + vydy)

= (fuux + fvvx)dx + (fuuy + fvvy) dy.

But equation (10.16) says

df = fxdx + fydy.
370  •  Mathematical Physics

Therefore, comparing coefficients of dx and dy in the last two

equations,
fx = fuux + fvvx and fy = fuuy + fvvy  (10.23)

which are exactly the chain rule equations (10.18) and (10.19).

Example 10.18 Given f(x, v) and v = v(x, y) find (∂f/∂x)y in

terms of (∂f/∂x)v and (∂f/∂v)x.
Given f(x, v) we would get f(x, y) by substituting for v(x, y). For
differentials, the strategy is the same: compute df in terms of
dx and dy by substituting for dv. So, first we compute df (x, v):

df = (fx)vdx + (fv)xdv.

Next, compute dv:

dv = (vx)ydx + (vy)xdy.

Substituting for dv

df = (fx)vdx + (fv)x[(vx)ydx + (vy)xdy]

= [(fx)v + (fv)x(vx)y]dx + (fv)x(vy)xdy.

Write df in the form “(· · ·)dx + (· · ·)dy”:

df = (fx)ydx + (fy)xdy.

Finally, compare coefficients of dx:

(fx)y = (fx)vdx + (fv)x(vx)y.

i.e.
 ∂f   ∂f   ∂f   ∂v 
=    +    .
 ∂x  y  ∂x  v  ∂v  x  ∂x  y
 Partial Derivatives  •  371

Exercise 10.23 Suppose we are given a function f(x, y) with

x = x(t) and y = y(t). Show that

 dx dy 
=df  f x + fy dt,
 dt dt 
and hence that
df dx dy
= fx + fy .
dt dt dt
Verify that this result can be obtained directly from the chain rule.

If we are given that x is some function of a variable u, x = x(u),

and also u as some function of x, u = u(x), then we can find dx/du if
we know du/dx. In fact
−1
dx  du 
=  .
du  dx 
Suppose now that we are given x = x(u, v) and y = y(u, v). Then we
must have some inverse relation u = u(x, y) and v = v(x, y). How can we
find (∂x/∂u)v if we know (∂u/∂x)y, (∂v/∂x)y, (∂u/∂y)x and (∂v/∂y)x? The
next exercise shows how. (The answer is not ( ∂x ∂u )v =( ∂u ∂x ) y
−1

because different variables are being held constant on the two sides.)

Exercise 10.24 If x = x(u, v), y = y(u, v) and the inverse relations

u = u(x, y) and v = v(x, y) are given, write down the definitions of
the differentials du, dv and dx. Using the fact that also

dx = 1 · dx + 0 · dy,

show that
x u uy
xv = − ,
uy
and that

xu = vy/(uxvy − uyvx).

This last relation gives us the required expression for (∂x/∂u)v.

372  •  Mathematical Physics

Revision Notes

After completing this chapter you should be able to

• Calculate the partial derivatives of a given function
f(x, y,…)
• Calculate the partial derivatives with respect to x, y, z of
a function f(r) where r = (x2 + y2 + z2)1/2
• Differentiate an integral w.r.t. a parameter
• Calculate a finite number of terms of the Taylor series of
a function of several variables
• Calculate the stationary points of a given function and
determine whether they are maxima, minima or saddle
points
• Calculate the total differential of a given function
• Determine whether a given expression is an exact
differential
• Use the chain rule to relate partial derivatives with
respect to different coordinates

10.17. EXERCISES

1. By explicit calculation verify that if f(x, y) = ln(cos(x/y))

∂2 f ∂2 f
then = .
∂x∂y ∂y∂x
2. Write down the chain rule for the partial derivative ∂f/∂u of
a function f(u, v) in terms of ∂f/∂x and ∂f/∂y where
x = x(u, v) and y = y(u, v). If f(x, y) = (ln(x))(ln(y)) and
x = ue−v, y = uv show that ∂f/∂u = (ln(u2v) − v)/u.
3. If f(x, y, z) = z/r2 where r = (x2 + y2 + z2)1/2, what are
∂f ∂f
(i) and (ii) ?
∂x ∂z
 Partial Derivatives  •  373

∂r ∂ 1
4. Let r2 = x2 + y2 + z2; find and ( ) , and show that
∂x ∂x r
∂2  1  1 3 x2
  =
− + .
∂x2  r  r3 r5
5. Let r2 = x2 + y2 + z2; if f(r) is a function of r only show that
∂f df x ∂ 2 f x2 d 2 f 1 df x3 df
= and = + −
∂x dr r ∂x2 r 2 dr 2 r dr r 3 dr
∂2 f xy d  1 df 
and =  .
∂x∂y r dr  r dr 
∂f ∂f
6. Let f(x, y) = xy2 sin(y/x); show that x +y = 3f .
∂x ∂y
7. Let r = (x2 + y2 + z2)1/2; show that
∂ ∂ ∂
( x + y + z )r n = nr n for any n ≠ 0.
dx dy dz
∂f
8. Let r = (x2 + y2 + z2)1/2 and f = 1/r. Find , write down
∂x
∂f ∂f
expressions for and and hence show that f satisfies
∂y ∂z
∂f ∂f ∂f
x +y +z = − f.
∂x ∂y ∂z

9. Let u = x2 − y2 and v = y. For a function f(x, y) find

∂f ∂f ∂f ∂f
and in terms of and . Hence find the
∂x ∂y ∂u ∂v
general solution of the partial differential equation
∂f ∂f
y +x = 2 xy .
∂x ∂y
10. Show that
d ∞
exp(=
−λx2 ) dx x2 exp( −λx2 ) dx( λ > 0 ) ;
∞
−
dl ∫0 ∫ 0
374  •  Mathematical Physics

∞
∫
2
hence (or otherwise) evaluate x2 e− x dx given that
0

∞ p
∫
2
e− x dx = .
0 2
2
d ∞ e− x t ∞ p
∫ ∫
2
11. Evaluate 2
dx (t > 0)given that e− x dx = .
dt 0 x 0 2
12. Find the Taylor series of f(x, y) = x/y about the point
x = 1, y = 1 up to and including terms of the third order (i.e.
as a cubic polynomial).
13. Given that f(x, y) is stationary at the origin and its second
derivatives are fxx = 2, fyy = −2, fxy = 3, determine the nature
of the stationary point.
14. Write down the conditions for a(x, y)dx+b(x, y)dy to be a
perfect differential, and deduce that
1 x
dx − 2 dy
y y

is a perfect differential.
15. For what value of n is the expression xn[(6x5 + 3y2)dx − 2xydy] a
perfect differential?
2 2 x2 x3 2
16. Is e x y (1 + )dx − e x y dy a perfect differential? (Show
y y
your reasoning.)
17. Write down the condition for p(x, y)dx + q(x, y)dy to be a
perfect differential. Is
2x x2
dx − dy
y + x2 y2 + x 2 y

a perfect differential?
 Partial Derivatives  •  375

10.18. PROBLEMS

1. Show that if x = r cos(q) and y = r sin(q) then ∂r = x and

∂r y ∂x r
= . Suppose now that the function g(r, q) has the
∂y r
form g(r, q) = f(r) sin(q), where f(r) is a function of r only.
By writing g(r, q) in the form yF(r), where F(r) = f(r)/r, or
otherwise, show that
∂g xy dF ∂g y2 dF
(i) = , (ii) = F +
∂x r dr ∂y r dr
∂2 g ∂2 g  d 2 F 3 dF 
and (iii) + = y  2 + 
∂x2 ∂y2  dr r dr 
Hence prove that if g(r, q) satisfies the partial differential
∂2 g ∂2 g
equation 2 + 2 = 0 , then f(r) = Ar + B/r where A and
∂x ∂y
B are constants.
2. Given u(r, q) with x = r cos(q), y = r sin(q), show that
∂u ∂u sin (q ) ∂u
= cos (q ) − ,
∂x ∂r r ∂q
∂u ∂u cos (q ) ∂u
= sin (q ) + .
∂y ∂r r ∂q
Now consider f(r, q) with x = r cos(q), y = r sin(q). By set-

∂f ∂2 f
ting u = use the above results to express in terms
∂x ∂x2
∂ 2 f ∂ 2 f ∂ 2 f ∂f ∂f
of , , 2, and .
∂r ∂r ∂q ∂q ∂r
2
∂q
∂2 f
Obtain the corresponding expression for and show
that ∂y 2

∂ 2 f ∂ 2 f ∂ 2 f 1 ∂f ∂ 2 f
+ = + + .
∂x2 ∂y2 ∂r 2 r ∂r ∂q 2
376  •  Mathematical Physics

3. Find the stationary point of the function

f(x, y) = x4 + 2x2 + 3xy + 3y.

Write down the Taylor series about the stationary point

up to terms of second order, and hence (or otherwise)
show that this point is neither a maximum nor a mini-
mum.
4. (i) S
 how that the quadratic form ax2 + 2bxy + cy2 is positive
for all x, y ≠ (0, 0) if a > 0 and ac − b2 > 0.
(ii) Show that the function

f(x, y) = x4 + 3x2 + 4xy + 2y2 − 4x − 4y + 2

has a single stationary point at (0, 1) which is a minimum.

(iii) Write down the Taylor expansion of f(x, y) about the
stationary point and hence show that the minimum
is a global minimum, i.e. f nowhere takes a smaller
value.
5. Find the stationary points of the function

f(x, y) = x4 − 2x2 + y3 − 3y

and for each determine whether it is a minimum, maximum

or saddle point. Suppose the true values of x and y are x = 1,
y = 1 and that they are measured each with an error of at
most 10%. What is the corresponding maximum percentage
error in f? If the true values are x = 2, y = 2 what is the maxi-
mum percentage error in f corresponding to an error of 1% in
x and y? Briefly explain your result.
6. Find the stationary points of the function
5 2
f ( x, y ) =x 4 + x3 y + xy
8
 Partial Derivatives  •  377

and determine whether the stationary point not at the ori-

gin is a maximum, minimum or saddle point.
7. Find the stationary point of the function f(x, y) = xy+1/x+1/y,
(x ≠ 0, y ≠ 0) and determine whether it is a maximum, mini-
mum or saddle point.
8. Obtain the Taylor expansion to second order of the func-
tion f(x, y) = (xy)1/2 about the point x = 1, y = 1. Hence show
1
that ( xy ) < ( x + y ) near (1, 1).
12

2
9. Find the stationary point of the function ln(x) − x/y2 − 2y
and determine its nature.
1
10. Show that f ( x, y ) = 3 x2 + 2 xy + y2 − 2 x + has one station-
2
ary point. Find the position (x0, y0) of the stationary point
and show that it is a minimum. Deduce that f(x, y) > 0 for
all x, y. Show that g(x, y) = exp(a(x + y))f(x, y) has a mini-
mum at (x0, y0) for all values of a. Show further that when a
≠ 0, g(x, y) has a second stationary point and determine its
nature.
 CHAPTER

11
PARTIAL DIFFERENTIAL
EQUATIONS

Many situations in physical science are governed by the relations

involving partial derivatives of an unknown function, hence by
partial differential equations (PDEs). Examples include the gravita-
tional field of a distribution of mass, the electromagnetic field gener-
ated by a charge distribution, the diffusion of heat in a reactor, the
probability density of an electron in an atom and the flow of a fluid.
For example, the rate at which the temperature in a bar is changing
at any point (partial derivative of temperature with respect to time)
is related to the spatial gradient of temperature at that point (partial
derivative with respect to spatial coordinate).
In many real situations the governing partial differential equa-
tions are complicated and cannot be solved using analytical meth-
ods; they must be solved by numerical methods. But in simple cases
there do exist methods of explicit solution, and these provide insight
into the behavior of the system and assist the development of physi-
cal intuition. In this and the next chapter we concentrate mainly
on one particular equation, namely the wave equation in one space
dimension that governs, for example, the propagation of disturbances
on strings and in tubes of fluid. We consider two methods of solu-
tion: the general solution involving arbitrary functions introduced
in Section 11.2 and the sum of separable solutions, involving arbi-
trary coefficients, which is taken up in Section 11.11 and Chapter 12.
The former is a special feature of the wave equation, the latter a
rather general approach for linear equations.
380  •  Mathematical Physics

11.1.  EXAMPLES FROM PHYSICS

Let y(x, t) = sin(x + ct), with c a constant, and let u = x + ct.

Then, writing yx for ∂y/∂x etc., we have, by the chain rule,

∂y ∂u
yx = × = cos ( u ) × 1 = cos ( x + ct ) .
∂u ∂x

Similarly yt = c cos(x + ct) and

− sin ( x + ct ) , ytt =
yxx = − c2 sin ( x + ct ) .

Therefore
1
ytt = yxx
c2

or
1 ∂2 y ∂2 y
= . (11.1)
c2 ∂t 2 ∂x2 

We say that the function y(x, t) = sin(x + ct) satisfies the partial
differential equation (11.1). But this may not be − in fact, is not − the
only solution. To find other solutions of equation (11.1) we must per-
form a reverse process: start from equation (11.1) and find the pos-
sible forms for y(x, t). In practice, solutions are usually required to
satisfy further subsidiary conditions (boundary conditions or initial
conditions, or a combination), as will be explained later.
Equation (11.1) is called the wave equation (in one space dimen-
sion). It describes the propagation, for example, of a wave on a uni-
form string under tension (y is the transverse displacement of the
string from its equilibrium position, x is the distance along the string
and t the time; compare Figure 11.1) or of a sound wave in a uniform
tube (y is the pressure in the tube) etc. The constant c is the speed
of the wave. In the case of the string, the physical interpretation of
this equation is that the transverse acceleration (perpendicular to the
string) of an element of string, ∂2y/∂t2, is a result of the net tension
acting on the element arising from the distortion of the string ∂2y/∂x2.
The following are four more examples of partial differential
equations in physics.
 Partial Differential Equations  •  381

The “wave equation” for f = f(x, y, z, t) in three space dimensions:

∂2 f ∂2 f ∂2 f 1 ∂2 f
+ + = . (11.2)
∂ x 2 ∂ y2 ∂ z 2 c 2 ∂ t 2

Laplace’s equation:

∂2 f ∂2 f ∂2 f
+ + =0, (11.3)
∂x2 ∂y2 ∂z2

where f = f(x, y, z) might be, for example, the electrostatic poten-
tial, or the gravitational potential in empty space, or the velocity
potential for an inviscid fluid.
The “diffusion equation”:

∂ 2 y 1 ∂y
= , (11.4)
∂x2 D ∂t 
where y = y(x,t), D = constant, describing, for example, the diffusion
of heat in a bar (y is then the temperature).
Schrödinger’s equation:

∂2y ∂y
= −i (11.5)
∂x 2
∂t 
for the wave function y of a free particle.
These equations are linear (i.e. the unknown function and its
derivatives occur only linearly) and homogeneous (there is no term
involving only the independent variables and not the unknown func-
tion). They are also of second order (i.e. second derivatives occur but
not higher ones). A method of solution will be presented in section
11.11. However, in Sections 11.2–11.9 we explore an exceptional
feature of the wave equation (11.1): the existence of an explicit form
for the general solution.

Exercise 11.1 For any one of the equations (11.1)–(11.5), show

by substitution that if two functions, g and h say, of the indepen-
dent variables are known to be solutions, then Ag + Bh is also a
solution where A, B are arbitrary constants. (This is true because
the equations are linear and homogeneous.)
382  •  Mathematical Physics

Example 11.1 Find the general solution for h(x,y), a function

of x and y, of the equation ∂h/∂x = 0.
The obvious approach is to integrate. If
∂h
=0
∂x
then h is a constant as far as differentiation with respect to x is
concerned, i.e. h is an unknown function of y only, say g(y) and
so h(x,y) = g(y). It is independent of x. Differentiate g(y) with
respect to x to check, remembering that partial differentiation
with respect to one variable means that the others are held
constant, and so
∂h ∂g ( y )
= = 0.
∂x ∂x

This is the general solution. Note that it contains one arbitrary

function (of one variable). This is an obvious extension of the
corresponding result for a function of a single variable, namely
that the solution to a first order ordinary differential equation
contains one arbitrary constant. For example, if the function of
a single variable h(x) satisfies the ordinary differential equation
dh/dx = 0, then f (x) is independent of x, i.e. f (x) = c, a constant.

Exercise 11.2 Write down an extension of the above example to

a function of three variables h(x, y, z).
y(x, t)

x
FIGURE 11.1:  Wave on a string at some instant of time t.
 Partial Differential Equations  •  383

11.2. GENERAL SOLUTION OF THE WAVE

EQUATION

The wave equation in one space dimension for the unknown

function y(x, t) is

∂2 y 1 ∂2 y
= . (11.6)
∂x2 c2 ∂t 2 

For a concrete picture we shall think of y(x, t) as the transverse

displacement of a uniform stretched string at position x at time t.
(The string is undergoing transverse motion of small amplitude in a
horizontal plane, so gravitys can be neglected.)
We now come to a fundamental result. Take any functions f (x),
g(x) of a single variable (with the property that they vary smoothly
enough to have second derivatives at each point). A solution of equa-
tion (11.6) is
y(x,t) = f (x − ct) + g(x + ct)(11.7)

Proof: We calculate fxx and ftt and show they satisfy equation (11.6).
Let u = x − ct, so f = f(u(x, t)), i.e. f(x − ct) is a function of a function
u = (x − ct). Differentiate f (u(x, t)) by the chain rule:

∂f df ∂u df
= =
∂x du ∂x du

since ∂u/∂x = 1. Note that we write df/du (not curly ∂) because f is a

function of a single variable, u. Differentiate again to get fxx,

∂2 f d  df  ∂u d 2 f
= =   .
∂x2 du  du  ∂x du2

Using the fact that ∂u/∂t = −c,

∂f df ∂u df
= = ( −c ) ,
∂t du ∂t du
384  •  Mathematical Physics

and so
∂2 f d  df  ∂u 2
2 d f
= − c  = c .
∂t 2 du  du  ∂t du2

Hence fxx = (1/c2)ftt, so y = f (u); i.e. y = f (x − ct) satisfies equa-

tion (11.6).

Exercise 11.3 Complete the proof by showing also that g(x + ct)
satisfies equation (11.6).

In fact, equation (11.7) is the general solution of equation (11.6);

i.e. every solution can be written in the form equation (11.7) for
some functions f and g. This is an important result. (For a proof, see
Section 11.3.)

Example 11.2 Is the function y(x, t) = x2 − c2t2 a solution of the

wave equation?
The function can be written as (x − ct)(x + ct) which is not
of the form “a function of (x − ct) plus a function of (x + ct).”
Therefore this is not a solution of the wave equation. (Of
course, this can be checked by direct substitution, but this
usually requires more effort.)

Exercise 11.4 If f (u) = u2, g(u) = cos(u), what does equation

(11.7) give for y(x, t)?

Exercise 11.5 Which of the following are not solutions of the

wave equation?
(i) (x − ct)2,
(ii) xt,
(iii) x2 + c2t2,
(iv) sin(x) sin(ct).

The function y = f(x − ct), where c is a positive constant, rep-

resents a wave. (The shape of this wave will depend on the choice
of the function f, which can be anything, not just a sine or cosine
 Partial Differential Equations  •  385

y(x, t)

x
y(x, t)

FIGURE 11.2:  A wave moving to the right. The circles in the top figure indicate the displace-
ment of a point at x = 1 with time t: y(1, t). The circles in the bottom figure indicate the
movement with time of a point of constant phase.

(Exercise 11.5).) The quantity x − ct is the phase of the wave at posi-

tion x and time t.
In which direction does the wave y = f(x − ct) move? To answer
this consider the point with zero phase at any time. This is the point
x − ct = 0. So at t = 0 the point with zero phase is at the origin x = 0.
At later times (t > 0) it is at x = ct along the positive x-axis. There-
fore the wave moves to the right with wavespeed (or phase speed) c.
(The choice of zero phase is merely for convenience; any other
choice gives the same result.)
Another way to remember the direction of propagation is that
for the crest to stay at the constant height y = f (x0) (i.e. to move with
the wave) as t increases, x must increase so as to keep x − ct constant:
thus the wave moves to the right with speed c.
386  •  Mathematical Physics

The phase speed and direction of propagation define the phase

velocity.

Exercise 11.6 What is the phase velocity of the wave

y = g(x + ct)?

The phase speed, c, of the wave (to the left or right) should not
be confused with the transverse (“up and down”) velocity, ∂y/∂t, of a
point of the string. If the displacement of the string is given by equa-
tion (11.7), the transverse velocity of a point of the string is

∂y ∂
= f ( x − ct ) + g ( x + ct )  =
− cf ′ ( x − ct ) + cg′ ( x + ct ) ,
∂t ∂t 

where the prime denotes differentiation of a function w.r.t. its argu-

ment, e.g.
df ( u )
f ′ ( x − ct ) = .
du u= x − ct

Exercise 11.7 What is the transverse velocity of the string in

exercise 11.4 at the point x = 5p/8, at time t = 5p/8c?

11.3.  DERIVATION OF THE GENERAL SOLUTION

To derive the general solution we make a transformation from

(x, t) to new variables (u, v) defined by u = x − ct, v = x + ct. We have,
by the chain rule,
∂y ∂y ∂u ∂y ∂v ∂y ∂y ∂u ∂v
= + = + , since = 1and = 1.
∂x ∂u ∂x ∂v ∂x ∂u ∂v ∂x ∂x
Hence
∂y  ∂ ∂
=  + y
∂x  ∂u ∂v 
and

∂2 y  ∂ ∂  ∂y  ∂ ∂  ∂ ∂ ∂2 y ∂2 y ∂2 y
= + = + + y= 2 +2 + .
∂x2  ∂u ∂v  ∂x  ∂u ∂v   ∂u ∂v  ∂u ∂u∂v ∂v2
 Partial Differential Equations  •  387

Exercise 11.8 Show similarly that

∂2 y 2∂ y ∂2 y ∂2 y 
2
= c  2 − 2 + .
∂t 2  ∂u ∂u∂v ∂v2 

Putting these together, the wave equation (11.6) becomes

yuu + 2 yuv + yvv = yuu − 2 yuv + yvu .

Now we rearrange to see that

∂2 y ∂2 y
=
4 0,=
or 0.
∂u∂v ∂u∂v
This is now in a form which can be solved. We can write this as
∂  ∂y 
= 0.
∂u  ∂v 
This implies that ∂y/∂v is independent of u, i.e. an (arbitrary)
function of v only:
∂y
= h ( v) ,
∂v
where h is an arbitrary function of v. Integrating this equation gives

y = ∫ h(v) dv + f (u),

where f(u) is independent of v, i.e. a function of u only. Since the

integral of an arbitrary function is also an arbitrary function, we can
write this as
y = g(v) + f (u)

where g and f are both arbitrary functions. This is the general solu-
tion – it involves two arbitrary functions. (Compare second order
ordinary differential equations, where the general solution involves
two arbitrary constants.) In terms of x and t we have

y(x, t) = f(x − ct) + g(x + ct)

which is equation (11.7).

388  •  Mathematical Physics

11.4.  A STRING INITIALLY AT REST

In applications, the functions f and g in the general solution

(Equation 11.7) are determined by the way in which the wave motion
is set up, i.e. through given conditions at some specified time, often
t = 0.

Example 11.3 Find the solution of the wave equation for the
displacement of a string y(x, t), given that
∂y
=y ( x,0 ) sin
= ( x ) and ( x,0 ) 0
∂t
i.e. the string is initially at rest.
Starting from the general solution for the displacement, the
general solution is
y(x, t) = f(x − ct) + g(x + ct).(11.8)
To impose the second initial condition we need ∂y/∂t:
∂y
− cf ′ ( x − ct ) + cg′ ( x + ct ) .
= (11.9)
∂t 
Now impose the initial conditions on y and ∂y/∂t. We are given
y = sin(x) and ∂y/∂t = 0 initially (the string is at rest), and so
y(x, 0) = f(x) + g(x) = sin(x)(11.10)
and
∂y
( x,0 ) =
− cf ′ ( x ) + cg′ ( x ) =
0. (11.11)
∂t 
Thus we have two simultaneous equations for f and g, equations
(11.10) and (11.11). If we now integrate equation (11.11) we
obtain
−f(x) + g(x) = k, k a constant. (11.12)
 Partial Differential Equations  •  389

If we add this to equation (11.10) we find

2g(x) = k + sin(x),
and so
1
g ( x)
=
2
( sin ( x ) + k ) .
And from equation (11.12),
1
f ( x)
=
2
( sin ( x ) − k ) .
These are the solutions for f and g. Hence, from f (x) we get
f (x − ct) by substitution of x − ct for x. Similarly for g(x). Note
that k cancels, and so we arrive at
1 1
y ( x, t ) = f ( x − ct ) + g ( x + ct ) = sin ( x − ct ) + sin ( x + ct ) .
2 2

For a string released from rest (and only in this case) the solu-
tion can be interpreted graphically as follows: the initial configura-
1
tion y = h(x) (in this case sin(x)) is made up of two waves: 2 h ( x )
moving to the right and 1 h ( x ) moving to the left. These combine to
2
give the solution
1 1
y= h ( x − ct ) + h ( x + ct )
2 2
for a string initially at rest in the configuration y = h(x).

Exercise 11.9
(i) Repeat the working of Example 11.3 to find the solution of
the wave equation for the displacement of a string, y(x, t),
released from rest under the initial conditions

∂y
=y ( x,0 ) e=
− x2
, ( x,0 ) 0.
∂t
(ii) Interpret the solution as was done for Example 11.3 above
(see Figure 11.3).
390  •  Mathematical Physics

y(x, t)

FIGURE 11.3:  Motion of a string released from rest, with an initially Gaussian displacement
y(x, 0) = exp(−x2). As time increases (from back to front in the figure), the displacement splits
into two oppositely moving parts (Exercise 11.9).

Example 11.4 An infinite string is released from rest at time

t = 0 with the displacement
 x + L −L ≤ x ≤ 0

y ( x,0 ) =  L − x 0 ≤ x ≤ L
 0 otherwise

where L is a given constant. By using the graphical
interpretation described above draw a series of figures to show
the displacement of the string at times t < L/2c, L/2c < t < L/c
and t > L/c (where c is the wavespeed).
The string is released from rest so can be considered as two
equal parts moving to the left and right. The sum of these parts
gives the shape of the wave (Figure 11.4).

(a) (b) (c) (d)

FIGURE 11.4:  Displacement of string released from rest with an intially triangular displacement.
The sequence (a)–(d) shows the evolution with increasing time. (a) At t = 0 the two waves overlap
completely. (b) At 0 < t < L/(2c) the two waves are mostly overlapping. (c) At L/(2c) < t < L/c the
two waves are only partially overlapping. (d) At t > L/c the two waves have separated completely.
 Partial Differential Equations  •  391

Exercise 11.10 An infinite string is released from rest at time

t = 0 with the displacement

 a, − L < x < L
y ( x,0 ) =  ,
 0 otherwise
where L and a are given constants. By using the graphical inter-
pretation above draw a series of figures to illustrate the displace-
ment of the string at times t < L/2c, L/2c < t < L/c, t > L/c, where
c is the wavespeed. (Note that this is an idealized example: the
string is displaced to a height a between two end-points where it
is fixed (so has zero height). But a real string cannot be discon-
tinuous at the points x = ±L. The picture therefore approximates
a string that has been given a constant displacement along almost
the whole of its length 2L.)

11.5.  A STRING GIVEN AN INITIAL VELOCITY

The graphical method discussed above works only for strings

released from rest. In this section and the next (11.6) we give exam-
ples of the general method for strings with a non-zero initial velocity.

Example 11.5 An infinite undisplaced string is given an initial

transverse velocity
∂y
( x,0 ) = sin ( x ) .
∂t
Find its subsequent displacement.
For an infinite string we start from the general solution for the
displacement and transverse velocity which is
y(x, t) = f(x − ct) + g(x + ct)(11.13)
and
∂y
− cf ′ ( x − ct ) + cg′ ( x + ct ) .
=
∂t (11.14)
392  •  Mathematical Physics

Write down the initial conditions: at t = 0:

y(x, 0) = 0 (11.15)
and
∂y
( x,0 ) = sin ( x ) .
(11.16)
∂t 
Putting t = 0 in equation (11.13) and using equation (11.15)
f(x) + g(x) = 0 or f = −g(11.17)
and putting t = 0 in equation (11.14) and using equation (11.16)
gives
−cf ′(x) + cg′(x) = sin(x).

Integrating both sides (with respect to x):

− cf(x) + cg(x) = −cos(x),

and since f = −g in equation (11.17) we have

2cg(x) = −cos(x).
So
1
f ( x) =
− g ( x ) =cos ( x ) ,
2c
and hence
1 1
y ( x, t ) = f ( x − ct ) + g ( x + ct ) = cos ( x − ct ) − cos ( x + ct ) .
2c 2c
If we use the rule that
 A+ B  A−B
cos ( A ) − cos ( B) =
−2sin   sin  
 2   2 
with A = x − ct and B = x + ct then the cosine terms can be
combined in this solution to give
1
y ( x, t ) = sin ( x ) sin ( ct ) .
c
The initial conditions therefore generate a standing wave
pattern.
 Partial Differential Equations  •  393

Exercise 11.11 An infinite string is given an initial displacement

y(x, 0) = tan−1 (x) and an initial velocity yt(x, 0) = c(1+x2)−1 (with
c a constant). Find the subsequent displacement y(x, t).

11.6. A FORMULA FOR GENERAL INITIAL

CONDITIONS

In this section, we give a general formula for the propagation of

a wave on an infinite string from arbitrary starting conditions. It is
more important that you understand the method than memorize the
general formula.

Example 11.6 Suppose that an infinite string is given the

initial displacement y(x, 0) = q(x) and transverse velocity
yt(x, 0) = p(x), with q and p given functions. Find the subsequent
displacement of the string.
Since this is an infinite string, start from the general solution of
the wave equation:

y(x, t) = f (x − ct) + g(x + ct),

with
∂y
− cf ′ ( x − ct ) + cg′ ( x + ct ) .
=
∂t

Write down the initial conditions. These will determine f and g,

and so at t = 0

y(x, 0) = q(x) and yt(x, 0) = p(x).

Put t = 0 in the general solution and use the initial conditions to
give us two equations for f and g:

f(x) + g(x) = q(x) and − cf′(x) + cg′(x) = p(x).

Integrating the second of these gives
1 x
− f ( x) + g ( x) = p ( x′ ) dx′,
c ∫x0
394  •  Mathematical Physics

where x0 is arbitrary and we have omitted an arbitrary constant

of integration (since it will cancel as in Example 11.3). We
have relabeled the integration variable on the right to avoid
confusing it with the upper limit, which we now want to call x.
Solving for g using f + g = q:

1 x 1
=g ( x) ∫ p ( x′ ) dx′ + q ( x ) ,
2c x0 2
and hence
1 1 x
f ( x)
= q ( x ) − ∫ p ( x′ ) dx′.
2 2 c x0
Finally, substitute for f and g in the general solution and we
have that
1 1 x − ct 1 1 x + ct
y ( x, t =
) q ( x − ct ) − ∫ p ( x′ ) dx′ + q ( x + ct ) + ∫ p ( x′ ) dx′.
2 2 c x0 2 2 c x0

We can combine these integrals using the fact that

c b c
∫a
f dx − ∫ f dx =
a ∫ f dx,b

and so
1 1 x + ct
y ( x, =
t)  q ( x − ct ) + q ( x + ct )  + ∫ p ( x′ ) dx′. (11.8)
2 2 c x − ct 

The solution (11.18) for the motion of an infinite string under

any initial conditions is sometimes called d’Alembert’s formula. Note
that the arbitrary constant x0 has canceled out. Any particular prob-
lem can now be solved by substituting for the given p(x) and q(x).

Example 11.7 Let y(x, 0) = sin(kx), with k a constant, and

∂y
( x,0 ) = 0.
∂t
Use equation (11.18) to find y(x, t).
 Partial Differential Equations  •  395

Here p(x) = 0, so the integral term in equation (11.18) is equal

to 0. Inserting q(x) = sin(kx) we get
1 1
y (=
x, t ) sin  k ( x − ct )  + sin  k ( x + ct )  .
2 2
as in example 11.3.

Example 11.8 Let y(x, 0) =0 and

∂y
( x,0 ) = sin ( x ) .
∂t
Use equation (11.18) to find y(x, t).
Here q = 0, p = sin(x), so

1 x + ct 1
( x, t )
y= ∫ sin ( x=
′ ) dx′ cos ( x + ct ) − cos ( x − ct )  .
2c x − ct 2c 

Exercise 11.12 Solve Exercise 11.11 using the general formula

(Equation 11.18) and hence check your solution.

11.7.  SEMI-INFINITE STRING

Take an infinite string and fix the point at x = 0. To obey this

condition the general solution must satisfy

y(0, t) = f(−ct) + g(ct) = 0

for all t. This implies

g(u) = −f (−u).

Thus the general solution must be of the form

y(x, t) = f (x − ct) − f (− (x + ct)).

396  •  Mathematical Physics

Example 11.9 Sketch this solution for the case

 −1, −2 ≤ u ≤ −1
f ( u) = 
 0, otherwise.
See Figure 11.5.
(a) At t = 0, y(x) = f(x) − f(−x). The function −f(−u) is 1 for
2 ≥ u ≥ 1 and 0 otherwise (f is reflected in u = 0 with a change
of sign). Thus at t = 0, y(x, 0) = f(x) − f(−x) is the sum of two top
hat functions.
(b) At t = 1/c, y(x) = f (x − 1) − f (− (x + 1)). To the left of
the origin, f is non-zero where x−1 lies between −2 and −1, i.e.
where x lies between −1 and 0. To the right of the origin f is
reflected with a change of sign.
(c) At t = 3/(2c), y(x) = f (x − 3/2) − f (− (x + 3/2)) On the left f
is non-zero where (x − 3/2) lies between −2 and −1, hence for
x between −1/2 and +1/2. Since this is symmetrical about the
origin, it must lie on top of the right top hat and the two cancel
giving no net disturbance on the string.
(d) At t = 2/c, y(x) = f(x − 2) − f (− (x + 2)) so f = −1 where x−2
lies between −2 and −1, hence where x lies between 0 and 1.
And y is + 1 in the reflected region to the left of the origin.

(a) (b) (c) (d)

FIGURE 11.5:  See Example 11.9. At t = 0, a “top hat” displacement f(u) at x < 0 and its
counterpart −f(−u). The sequence (a)–(d) shows the evolution of the displacement at times
t = 0, 1/c, 3/(2c) and 2/c.

Now look at these pictures from the point of view of the semi-
infinite string x > 0: the left-moving wave is reflected at the origin,
with a change of sign, into a right-moving wave. Effectively, a wave
coming in from infinity (or at least a long way away) reflects off the
fixed point at the origin (e.g. a wall) and changes its phase by 180°.
 Partial Differential Equations  •  397

11.8.  SIMPLE HARMONIC WAVES

Any wave progressing in the +x direction with speed c has the

form y = f (x − ct). If the function f is such that the shape of the wave
repeats over a fixed distance (and fixed time) the wave is called a
harmonic wave. If f has the special form

y = A cos[k(x − ct) + f]
(or y = A sin[k(x − ct) + f]) the wave is called simple harmonic. Let

w = ck.(11.19)
Then y = A cos[k(x − ct) + f] = A cos[k(ct − x) − f] can be
written as

y = A cos(wt − kx − f).(11.20)

This gives us the following quantities (Figure 11.6):

• A is the amplitude of the wave,
• 2p/k is the wavelength,
• k is the wavenumber,
• w is the frequency,
• 2p/w is the period,
• c = w/k is the phase speed,
• (wt − kx − f) is the phase of the wave at t, x).

Exercise 11.13 What is the change of phase of a simple harmonic

wave on reflection at a fixed boundary? (See Example 11.9.)

11.9. BEATS

Suppose we add two simple harmonic waves on a string with

slightly different frequencies and equal amplitudes:

y(x,t) = Acos(wt − kx) + A cos(w′t − k′x)(11.21)

398  •  Mathematical Physics

where w′ = w+∆w, k′ = k+∆k with ∆w and ∆k small (i.e. ∆w/w << 1

and ∆k/k << 1) and w = ck, w′ = ck′. Let = w ( w + w ′ ) 2 be the
mean frequency and k= ( k + k′ ) 2 be the mean wavenumber. Also,
we have
∆ w w ′ −w ck′ −ck
= = = c.
∆k k′ −k k′ −k

phase

amplitude
y(x, 0)

x
wavelength

FIGURE 11.6:  Some properties of a harmonic wave pictured at a given time.

FIGURE 11.7:  Beats in the sum of two simple harmonic waves of the same amplitude with
slightly different frequencies. The top curve shows A cos(wt − kx). The second curve shows
A cos(w′t − k′x). The two curves are overlaid to highlight where they are in phase and out of
phase, and the fourth curve shows the sum of the two waves.

Using cos (a ) + cos ( b ) = a − b  a + b  , equation

2 cos   cos  
 2   2 
(11.21) can be rewritten as
 Partial Differential Equations  •  399

 ∆w ∆k 
y( x, t ) = 2 A cos 
 2
t − x cos w t −kx .
2 
( ) (11.22)

If we let
 ∆w ∆k 
A( x, t ) 2 A cos 
= t − x ,
 2 2 
then equation (11.22) is

=y( x, t ) A( x, t ) cos w t −kx .

( (11.23) )
This is therefore a wave (the carrier wave) with frequency w
and speed w k = c , with varying amplitude A(x, t). The amplitude
A(x, t) is itself a wave of large wavelength, 4p/∆k, and low frequency,
∆w/2, the beat frequency. The beat pattern moves with velocity
∆w/∆k = c. (See Figures 11.7 and 11.8.)

–20 –10 0 10 20 –20 –10 0 10 20 –20 –10 0 10 20

x x x

(a) (b) (c)

FIGURE 11.8:  The beat pattern (at t = 0) from the sum of harmonic waves with slightly
different wavenumbers: y(x, t) = Acos(wt − kx) + B cos(w′t − k′x). (a) Waves with the same
amplitudes, A = B = 1, and (k′ − k)/k = 0.1, (b) with same amplitudes, A = B = 1 and (k′ − k)/k = 0.05,
(c) with (k′ − k)/k = 0.1 and A = 0.4 and B = 1.

|A+B|

|A–B|

–|A–B|

–|A+B|
FIGURE 11.9:  Beats in the sum of two simple harmonic waves of different amplitudes.
400  •  Mathematical Physics

Exercise 11.14 Let y(x, t)= Acos(wt − kx) + B cos(w′t − k′x)

(A and B given constants) be the sum of two simple harmonic waves
of different amplitudes, B ≠ A with w/k = w′/k′ = c. By writing
A = (A + B)/2 + (A − B)/2 and B = (A + B)/2 − (A − B)/2, show
that this also gives rise to beats, but of a diminished amplitude
(compared with the case A = B). See Figure 11.9.

11.10.  GROUP VELOCITY

Waves on a string move with the same speed c independent of

frequency. Waves in other media can move at speeds that differ for
different frequencies, so c = c(w). Such media are said to be disper-
sive. In a dispersive medium, we have w = c(w)k or, equivalently,
solving for w, we have w = w(k). Thus, in a dispersive medium, the
frequency is a function of wavenumber. If the wavespeed depends on
frequency, we can ask, at what speed does the beat pattern of waves
at frequencies w and w + ∆w move? The calculation of the previous
section gives the answer: the beat pattern moves with speed ∆w/∆k.
Consider now a group of waves, with frequencies in the range
w to w + ∆w. In the limit ∆w → 0 the beat pattern moves with
speed vg = dw/dk. The speed vg is called the group velocity. The
group velocity is in general itself a function of frequency and
gives the speed with which energy is propagated in a dispersive
medium (Figure 11.10).

Exercise 11.15 Show that the group velocity vg is equal to the

wave velocity c (for all wavenumbers) if and only if c is a constant,
i.e. independent of frequency, as in the case of a stretched string.

Exercise 11.16
(i) Show that if the frequency w and wavenumber k of waves
1
in a certain medium are related by w = k2 , then the group
velocity is double the wave velocity. 2

(ii) What is the relation between w and k if, instead, waves have
group velocity equal to half the wave velocity?
 Partial Differential Equations  •  401

Exercise 11.17 For a wave of frequency w, wavenumber k(w)

propagating in a general medium with wavespeed c(w), show that

dc 1  c 
= 1 − ,

dw k  vg 

and hence that the group velocity is respectively greater than the
wave velocity, equal to it, or smaller, according as the wave veloc-
ity increases with the frequency, is constant (as in the case of a
stretched string), or decreases as the frequency increases.

Waves for which the group and phase velocities are equal are
called non-dispersive. In particular, if a wave is non-dispersive then
its phase speed is constant with frequency (and conversely). The
group speed provides a useful estimate of the modes (of different
frequencies) that transfer the most energy.

0 5 10 15 20 0 5 10 15 20
x x

(a) (b)

0 5 10 15 20 0 5 10 15 20
x x

(c) (d)

FIGURE 11.10:  The beat pattern from the sum of harmonic waves with different wavenumbers
and different phase velocities: y1(x, t) = cos(wt − kx) (top curve) and y2(x, t) = cos(w′t − k′x)
(second curve). Here v1 = w/k = 5/5 = 1, while v2 = w′/k′ = 6.6/5.5 = 1.2. (a) At time t = 0 the
two simple waves are in phase, and they are in phase with their sum. (b)–(d) Times t = 2.5, 5
and 7.5 showing the effect of the different phase speeds of the simple waves on the sum. The
solid dots represent points of constant phase on each wave (y1, y2 and the carrier wave of their
sum); these all move with their respective phase velocities. The open circle shows a point of
constant phase on the envelope; this moves with the group velocity.
402  •  Mathematical Physics

Exercise 11.18 A wave propagates along the x-axis with a fre-

quency given by
w = −k/(k2 + 1).
Determine the minimum frequency with which the wave can
propagate in
(i) the negative x-direction
(ii) the positive x-direction.

11.11.  SEPARATION OF VARIABLES

So far we have considered infinite strings (i.e. strings so long that

we need not consider what is happening at the ends) and semi-infi-
nite strings, where we need consider only one end. For a finite string
we have to impose conditions at both ends. This can be quite hard
to solve if we start from the general solution. We also want to solve
other equations, which, unlike the wave equation in one dimension,
do not have an explicit general solution. A method that works for a
limited but useful class of partial differential equations is separation
of variables, which we describe in this section.
Now, the possible motions of a stretched string fixed at its end-
points P and Q, say, are identical, in the region between P and Q,
with the possible motions of an infinite string which happen to have
nodes (points at which there is no transverse motion) at P and Q. For
1 1
example, in Example 11.3, we saw that= y sin ( x − ct ) + sin ( x + ct ) is
2 2
a possible solution of the wave equation (11.6). We can use the iden-
tity for sin(A − B) to rewrite this solution as follows:
1
=y ( x, t )
2
( sin ( x ) cos ( ct ) − cos ( x ) sin ( ct ) )
1
+ ( sin ( x ) cos ( ct ) + cos ( x ) sin ( ct ) )
2
= sin ( x ) cos ( ct ) .

This solution has the property that y(0, t) = 0 and y(2p, t) = 0 for
all t. If we restrict our attention to the region 0 ≤ x ≤ 2p, this solution
 Partial Differential Equations  •  403

also represents one possible wave motion on a string of length 2p

fixed at its ends (so the displacement, y, is constrained to be zero at
the ends). This leads to two questions:

1. Can other motions of a string fixed at its ends be represented

in a form like equation (11.11), namely with y(x, t) = (func-
tion of x only) × (function of t only)? This question will be
answered below.
2. Can all motions of the string be derived from such solutions?
This question will be answered in Chapter 12.

To find the answer to (1) we take y(x, t) to be of the proposed

form, substitute in the wave equation, and determine what condi-
tions must be satisfied if y is to be a solution. So, since we want a
function of x times a function of t, let us write y(x, t) = X(x)T(t) as the
general form of solution and attempt to determine what functions
X and T will give a solution of the wave equation. Note that X is a
function of x only and T a function of t only. A solution of this form
is said to be separable and the process of seeking a solution in this
way is referred to as the method of separation of variables. (We are
separating the time and space parts of the solution.)
It is important to realize that y = X(x)T(t) is a less general solu-
tion than y = f(x−ct)+g(x−ct). This latter class encompasses all sepa-
rable solutions, in addition to all other possible solutions.
We substitute our expression for y(x, t) in the wave equation

∂2 y 1 ∂2 y
= .
∂x2 c2 ∂t 2 (11.24)
Take the left side first. We have

∂2 y ∂2 X
= T,
∂x2 ∂x2
since T(t) is constant for partial differentiation with respect to x, and
similarly for the right hand side

∂2 y d 2T
= X .
∂t 2 ∂t 2
404  •  Mathematical Physics

Hence, substituting in equation (11.24), we require

d 2 X X d 2T
T = . (11.25)
dx2 c2 dt 2 
Divide through by y = XT:

1 d2 X 1 d 2T
= .
X dx2 c2 T dt 2 (11.26)
In equation (11.26) we have been able to write our equation
in a very special form: the left side is a function of x only; the right
side is a function of t only, not of x. Thus, a function of x on the left
equals something that is independent of x on the right. The only way
this can happen is if each side of the equation equals a constant. We
have the freedom (at the moment) to choose this constant. For the
applications we shall be making it will turn out that the best choice
is for this constant to be negative. With hindsight we therefore write
this separation constant as −a2. (Writing + a2 for the constant would
lead a to be imaginary; this is not wrong but it is not so convenient.)
Thus equation (11.4) gives

1 d2 X 1 d 2T
=
−a 2
, −a 2c2 .
= (11.27)
X dx2 T dt 2 
Thus we have separated out the x and t parts of the equation. The
dependence on x and t is sometimes said to be decoupled. Equations
(11.27) can be solved for X and T; possible solutions are
X = cos(ax) or sin(ax),
and
T = cos(act) or sin(act).
Since each of these four solutions are possible, we make linear
combinations to form a solution:

y(x, t) = [a cos(ax) + b sin(ax)] [c cos(act) + d sin(act)]

= Asin(ax + y) sin(act + f).(11.28)

For the second line we have made use of the fact that
 Partial Differential Equations  •  405

   Asin(z + f) = A [sin(z) cos(f) + cos(z) sin(f)]

= (A sin(f)) cos(z) + (A cos(f)) sin(z)
 = a cos(z) + b sin(z).
The two constants on the left side (A, f) are related to the two
constants on the right side (a and b) by a = A sin(f) and b = A cos(f).
From the second form of equation (11.28) it is clearer that y(x, t)
involves three independent constants, not four, but the first form
is often simpler in applications. Note that in a separable solution y
varies in time in the same way (e.g. like sin(ct + f)) at all points, x.
Separable solutions are also referred to as normal modes of the
string. Physically, these solutions are standing waves.
Exercise 11.19 Using the first form of the separable solutions
above, show that this is consistent with the solution to Example 11.3.

Exercise 11.20 What are all the separable solutions of the wave
equation obtained by taking +a2 in equation (11.27)? Hence
explain why the choice of −a2 is made in applications involving
the wave equation.

Exercise 11.21 For Laplace’s equation in 2D,

∂2 f ∂2 f
+ =
0,
∂x2 ∂y2
use the method of separation of variables to find all solutions of
the form f(x, y) = X(x)Y(y) which vanish on x = 0 (for all y).

Exercise 11.22 Use the method of separation of variables to find

all solutions of the form y(x, t) = X(x)T(t) of Schrödinger’s equa-
tion
∂2y ∂y
= −i
∂x 2
∂t
for which X is real.
Note that most equations do not have separable solutions; i.e. in
general this method does not work because it cannot lead to an equa-
tion like equation (11.26) in which the variables have been separated!
But those that do are among the most important in physical science.
406  •  Mathematical Physics

11.12. EIGENVALUES

So far we have found separable solutions of the wave equation

but we have not yet considered boundary conditions (e.g. at the
fixed ends of the string) or initial conditions (e.g. how the string is
set vibrating at time t = 0). The former will be considered here, the
latter in Chapter 12. We shall illustrate the effect of boundary condi-
tions by some examples. The result is always the same: the boundary
conditions determine the possible values of the separation constant.
Possible values of a separation constant, giving rise to non-
zero solutions satisfying the given boundary conditions, are called
eigenvalues.

Example 11.10 Find the separable solutions of the wave

equation on a string of length l subject to the conditions
y(0, t) = 0, y(l, t) = 0 for all t (i.e. the string is fixed at its end-
points x = 0 and x = l).
First, write down the general form of the separable solutions:

y = A sin(ax + y) sin(act + f).

Next, impose the boundary conditions. For this problem these

are given at x = 0 and x = l. At x = 0:

y(0, t) = A sin(y) sin(act + f) = 0

for all t. So Asin(y) = 0, hence y = 0 (since A ≠ 0). Similarly, at

x = l:

y(l, t) = Asin(al)sin(act + f) = 0,

for all t. So it must be that Asin(al) = 0. If

sin(al) = 0
then al is an integer multiple of p, so we can put

a = an = np/l, for n = 1, 2, 3, . . . .
 Partial Differential Equations  •  407

These are all the solutions of sin(al) = 0, and we see al must

be an integer multiple of p. We have found an infinite number
of separable solutions satisfying the given conditions. The
separable solutions are

y(x, t) = Asin(anx) sin(anct + f),

where an = np/l for n = 1, 2 . . . . (See Figure 11.11.)

y(x, t)

0.0 0.2 0.4 0.6 0.8 1.0

x

FIGURE 11.11:  The first five eigenfunctions from Example 11.10. Bottom to top: n = 1, 2,
3, 4, 5.

The numbers an2 are the eigenvalues of the problem; the cor-
responding solutions for y are called the eigenfunctions of the prob-
lem. So for the example above the eigenvalues are (np/l)2 and the
eigenfunctions are
Asin(npx/l) sin(npct/l + f),
or, expressed differently,
sin(npx/l)(a cos(npct/l) + b sin(npct/l)).
In the case of the wave equation we also refer to the eigenfunc-
tions as the normal modes of the system and the eigenfrequencies as
the normal frequencies or harmonics.
408  •  Mathematical Physics

Exercise 11.23 Find the separable solutions of the wave equa-

tion on a string subject to the conditions y = 0 at x = 0 for all t and
∂y/∂x = 0 at x = l for all t. What are the eigenvalues and eigenfunc-
tions? What are the normal modes of the string?

Exercise 11.24 Find the separable solutions of Laplace’s equa-

tion (in 2D) subject to the conditions f = 0 on x = 0 and x = l, for
all y. What are the eigenvalues and eigenfunctions?

Exercise 11.25 Show that for a string with the end at x = 0 fixed
and the end at x = l free, the wave equation has separable solu-
tions of the form
y(x, t) = sin(wx/c)(Acos(wt) + Bsin(wt))
Find the transverse velocity of the free end of the string at x = l
corresponding to this solution. If the free end at x = l is driven
with velocity v0 sin(pt) show that the displacement of the string
is given by
v sin( px / c)
y( x, t ) = − 0 cos( pt ).
p sin( pl / c)

11.13.  REFLECTIONS AT A DISCONTINUITY

We now look at two developments of the above theory. The first is

to waves on a string which consists of two halves with different prop-
erties joined at the origin. The second is to a matter wave, satisfying
the Schrödinger equation, incident on a potential step. The approach
to these problems is similar. We consider a wave incident on the join
in the strings, or the potential step, and determine what fraction of it
is reflected and what fraction transmitted in order to satisfy certain
conditions at the join. To solve these problems efficiently it is conve-
nient to represent waves in exponential form. For example, the wave
Acos(wt − kx) is represented as Re(Aeiwte−ikx). The Re(…) is usually
omitted, it being understood that the real part is intended.
Let a semi-infinite string (x < 0), having density per unit length r,
wavespeed c, be joined at x = 0 to a semi-infinite string (x > 0) with
density r′, wavespeed c′. The tensions in the two strings are equal,
 Partial Differential Equations  •  409

from which it can be shown that r′(c′)2 = rc2 = F, say. Let an incident
wave of unit amplitude
yi(x, t) = eiwte−ikx

be incident from the left. (Note that the sign in the exponent is
related to the direction of propagation in the usual way.) A wave

ytrans(x, t)= Beiw′te−ik′x

is transmitted to the right in the region x > 0, and a wave

yref(x, t) = Aeiwteikx

is reflected to the left.

Thus, the total disturbance on the string in x < 0 is

y1(x, t) = eiwt(eikx + Ae−ikx),

and for x > 0
y2(x, t) = Beiw′teik′x.

Consider now the conditions at x = 0.

1. The string does not break, so y is continuous:

  y1(0, t) = y2(0, t),

eiwt(1 + A) = Beiw′t.
y(x, t)

x
FIGURE 11.12:  Wave incident on a join. In the text we show that w = w′ so this diagram is
often drawn omitting the factors eiwt.
410  •  Mathematical Physics

For this to hold at all times we must have w = w′ and

1 + A = B.(11.29)

2. The vertical components of the force at x = 0 must balance.

Since the tension on each side of the origin is the same, this
means the slope must be the same also. Thus
∂y1 (0, t ) ∂y2 (0, t )
= ,
∂x ∂x
or
ik − ikA = ik′B.(11.30)

We now have two equations (11.29) and (11.30) for A and B.

Solving gives
k − k′ 2k
=A = , B .
k + k′ k + k′

The energy flux in a harmonic wave of amplitude a, frequency

w, on a string with density r, wavespeed c is known to be

1
rw 2 a2 c.
2

So the incident energy flux is

1 1
rw 2 c = w kF,
2 2
using F = rc2 and w = ck. Similarly the reflected flux is

1 2
A w kF
2
and the transmitted flux is

1 2
B w k ′ F.
2
 Partial Differential Equations  •  411

The reflection coefficient is defined as

reflected flux
=R = A2
incident flux

and the transmission coefficient as

transmitted flux k′ 2
=T = B.
incident flux k

Notice that the transmission coefficient is not B2 because the

energy flux depends on wavenumber, which changes between the
two strings.

Exercise 11.26 Show that R + T = 1.

Exercise 11.27 Show from first principles that the proportion of

energy of a transverse wave which is reflected at a join between
two different strings is given by
(z1 − z2)2/(z1 + z2)2
where z is the impedance of a string, z = density × wavespeed.
Ene rgy

x
FIGURE 11.13:  A matter wave incident from the left on a potential barrier V0. The common
exponential factors e−iwt are often omitted.
412  •  Mathematical Physics

11.14. MATTER WAVES INCIDENT ON A

POTENTIAL STEP

For matter waves incident on a small potential step the math-

ematics follows closely that of the previous section, although the
interpretation is different.
The incident wave of unit amplitude is yi = e−iwteikx, where
w = E  and k = p  . The fact that this is a matter wave comes from
the relation between w and k: the Schrödinger equation for a free
particle gives E = p2/2m or w = k2 2 m and hence

2 mE
k= .

(Note the opposite signs in the exponentials in yi relative to
Section 11.13. This is dictated by the interpretation as matter waves.
Solutions of the Schrödinger equation are proportional to e− iEt  and
positive momentum means e+ ipx  . For waves on strings only the
relative sign between t and kx matters.)
Similarly the reflected wave is yr = Ae−iwte−ikx.
To the right of the origin the Schrödinger equation requires
p2
=
E + V0 ,
2m
or, with p = k′,
2 m(E − V0 )
k´ = .

So the transmitted wave is yt = Be−iwteik′x. Thus in x < 0 the wave-
function is
y1 = e−iwt(eikx + Ae−ikx),
and in x > 0
y2 = Be−iwteik′x.
There are two cases to consider, depending on whether E > V0
(k′ real) or E < V0 (k′ imaginary).
 Partial Differential Equations  •  413

(i) The small potential step. Assume first that E > V0 so k′ is

real. In a steady state there can be no build up of probability
at x = 0, so y must be continuous there:
y1(0,t) = y2(0,t),

or
1 + A = B.(11.31)

Note that this holds for all time only if we can cancel the
factors e−iwt, which requires w (or E) to be the same on both
sides of the barrier at x = 0. This corresponds to conservation
of energy. Similarly, the conservation of probability means
that the probability current must be continuous across x = 0,
so ∂y/∂x must be continuous. Thus
∂y 1 (0, t ) ∂y 2 (0, t )
= ,
∂x ∂x
or
ik − ikA = −ik′B.(11.32)

Equations (11.31) and (11.32) are the same as equations

(11.29) and (11.30) and yield the solutions

k − k′ 2k
=A = , B .
k + k′ k + k′
The reflection and transmission coefficients are the same as
for the string example (above). Again, we see an extra factor
of k′/k in the transmission region to take account of the dif-
ferent speed of the wave there.
(ii) The large potential step. In this case E < V0 so k′ is imagi-
nary. Thus A and B are complex and the definition of reflec-
tion and transmission coefficients requires a little more care.
The flux of energy is proportional to
E × particle probability density × speed ∝ E|ψ|2k.
414  •  Mathematical Physics

Thus the reflection coefficient is not A2 but R = |A|2. In gen-

eral (from the probability current)

2 2 Re(k′)
= =
R A , T B
Re(k )

Exercise 11.28 Show for a large potential step the reflection

coefficient is R = 1 and find the wavefunction in the region x > 0.

Exercise 11.29 From first principles derive the reflection and

transmission coefficients at a small potential step in terms of the
energy E of the incoming particle and the height of the barrier V0.

Revision Notes

After completing this chapter you should be able to

• Write down the wave equation in one, two or three
dimensions
• Understand the language used in the study of waves:
frequency, wavenumber, wavespeed, phase, beats, phase
and group velocity, eigenfunctions and eigenvalues,
transmission and reflection coefficients
• Write down and derive the general solution of the wave
equation in one space dimension
• Obtain the solution of the wave equation for an infinite
string given its initial position and velocity
• Compute the group velocity of a wave given a relation
between frequency and wavenumber
• Obtain separable solutions of simple linear partial
differential equations and explain the terms eigenvalue,
eigenfunction, normal mode
• Match solutions of the wave equation (and Schrödinger’s
equation) across a boundary and find the corresponding
reflection and transmission coefficients
 Partial Differential Equations  •  415

11.15. EXERCISES

1. Verify that y = Asin(ax) cos(act) is a solution of the wave

equation
∂2 y 1 ∂2 y
=
∂x2 c2 ∂t 2

where A, a are constants. Hence find the solution satisfying

 px 
y( x,0) = 2sin   and y(0, t) = y(l, t) = 0 for all t.
 l 
1
2. Verify that =
y
2
[ f ( x − ct) + f (x + ct)] is the solution of
∂2 y 1 ∂2 y
the wave equation 2 = 2 2 satisfying y(x, 0) = f(x),
∂x c ∂t
∂y
( x,0) = 0.
∂t
1 x + ct
3. Verify that y( x, t ) = ∫ G(s) ds is the solution of the
2 c x − ct
∂2 y 1 ∂2 y
wave equation = satisfying the initial conditions
∂x2 c2 ∂t 2
∂y
y(x, 0) = 0, ( x,0) = G( x).
∂t
4. What is the general solution of ∂ y( x, t ) = xt ?
2

∂x∂t
∂2 y 1 ∂2 y
5. If y(x,t) = exp[i(kx − wt)] satisfies 2 = 2 2 what is the
∂x c ∂t
relation between w and k?
6. Write down the normal modes of a stretched string with
wavespeed c which is of unit length fixed at its end­-points.
Such a string is given a displacement y(x, 0) = 3 sin(2px)
and released from rest. Show that the subsequent displace-
ment of the string is y(x,t) = 3 sin(2px) cos(2pct).
7. A semi-infinite string 0 ≤ x < ∞ is fixed at x = 0. A wave
y = f(x + ct) impinges on x = 0 from the right. What is the
reflected wave?
 416  •  Mathematical Physics

11.16. PROBLEMS

1. Verify that
1 1 x + ct 
y( x,=
t)  F( x − ct) + F( x + ct) + ∫ G(s)ds
2 c x − ct

∂2 y 1 ∂2 y
is the solution of the wave equation = satisfying
∂x2 c2 ∂t 2
∂y
the initial conditions y(x, 0) = F(x), ( x,0) = G( x).
∂t
2. An infinite string is at rest along the x-axis. At time t = 0 it is
struck in such a way that its velocity is non-zero only in the
region |x| ≤ L. Show that, after a time L/c, at the point x = 0
the string is again at rest. What parts of the string are at rest
for times t ≥ L/c?
3. An infinite string is released from rest with the initial shape
 a(l + x) if − l < x < 0

=y  a(l - x) if 0 < x < l
0 otherwise

where a, l are constants. Show that after a time t = l/c the
string at the point x = 0 is again at rest. What parts of the
string are at rest for t > l/c?
4. An infinite string lies along the x-axis except for the region
−l < x < l, where it is deformed from the axis. At t = 0 the
string is released from rest. Show that, whatever the initial
deformation, the string is again at rest at the origin (i.e.
y(0, t) = 0) for t > l/c.
∂2 y 1 ∂2 y
5. Separate variables in the wave equation 2 = 2 2 and
∂x c ∂t
hence find the solution obeying the conditions
 px  ∂  px 
=y( x,0) B=
sin   and ( ,0) sin  
 l  ∂t  l 
 Partial Differential Equations  •  417

for 0 < x < p, where B, l, v are constants. Verify that the

combination
2 2
 ∂y   pc  2
 ∂t  +  l  y
   
is constant in time at each x.
6. By seeking solution of the form y(x, t) = X(x)T(t) to the
∂2 y 1 ∂2 y
wave equation = find the solution obeying the
∂x2 c2 ∂t 2
∂y 2 cp  p x 
conditions y(x, 0) = 0, ( x,0) = sin   and y(0, t) =
∂t l  l 
y(1, t) = 0 for all t.
7. By seeking solutions of the form y(x, t) = X(x)T(t) to the
∂2 y 1 ∂2 y
wave equation 2 = 2 2 find the solution obeying the
∂x c ∂t
∂y
conditions y(0, t) = y(1, t) = 0 for all t and ( x,0) = 0 and
∂t
 px   3p x 
= y( x,0) a sin   + b sin   with a and b given con-
stants.  l   l 

8. Use separation of variables to find all the separable solu-

tions of the equation
∂2 f 2 ∂2 f
−c + w02 f =0,
∂t 2 ∂x 2
where c and w0 are constants.
9. An infinite stretched string, with wavespeed c, is given an
initial displacement in the form of a Gaussian curve
y(x, 0) = exp(−x2) and an initial velocity of the form
yt(x, 0) = cx exp(−x2). Show that the subsequent displace-
ment consists of a Gaussian wave of fixed shape with
amplitude 3/4 moving to positive x and a Gaussian wave of
amplitude 1/4 moving to negative x. If the two waves can-
not be resolved while the displacement of the string has its
greatest value at x = 0, show that the two waves will become
1/ 2
1 4 
apparent (i.e. can be resolved) after a time  ln( ) .
c 3 
 CHAPTER

12
FOURIER SERIES

12.1. INTRODUCTION

Many problems in physics involve vibrations and oscillations.

Often the oscillatory motion is simple (e.g. weights on springs,
pendulums, harmonic waves, etc.) and can be represented as a
single sine or cosine function. However, in many cases (electro-
magnetism, heat conduction, quantum theory, etc.) the waveforms
are not simple and, unlike sines and cosines, can be difficult to
treat analytically.
Fourier methods give us a set of powerful tools for represent-
ing any periodic function as a sum of sines and cosines. The Fourier
series of a function, f(x), with period 2L is
∞
1  np x np x 
f ( x) = a0 + ∑  an cos + bn sin  . (12.1)
2 n =1  L L 

To see how this works, let us consider a sound wave. The vibra-
tion of a tuning fork produces a sound wave of a given frequency.
If we plot the pressure as a function of distance, x, or time, t, it
looks like a single sine wave, or a “pure tone” (Figure 12.1: left).
When a note is played on a flute, we get a more complex sound
(Figure 12.1: center). The note that we get is made up from the sum
of many pure tones: the fundamental and different harmonics with
 420  •  Mathematical Physics

frequencies 2, 3, 4, . . . times the frequency of the fundamental (Fig-

ure 12.1: right). This series is the Fourier series representation of
the complex waveform.

FIGURE 12.1:  Left: a pure sine wave, sin(wt). Center: example waveform, f(t), from a flute.
Right: the note from the flute is made up of the sum of the fundamental sine wave and a series
of harmonics. In this example, f(t) = sin(wt) + sin(2wt) + 0.2sin(3wt) + 0.4sin(4wt). This is
the Fourier series.

Fourier methods are used very heavily in signal and data analy-
sis. By Fourier analyzing a signal – essentially by expanding it in the
form of equation (12.1) – we can immediately tell which harmon-
ics are the important ones. For example, in the note from the flute
(Figure 12.1), the harmonic at frequency 2w has relatively large
amplitude, while the harmonic at 3w is small. If, for example, a
poorly designed speaker filtered out the harmonic at 2w it would
greatly change the character of the sound, while filtering out the
harmonic at 3w would have a much less discernible effect.
Fourier methods are also commonly used in mathematical phys-
ics. In this chapter, we will focus on using them to solve differen-
tial equations, and the wave equation in particular. We will examine
Fourier half range series and Fourier full range series, study some
applications of Fourier series, then finish by introducing Fourier
transforms and the convolution theorem.

12.2.  FOURIER HALF RANGE SINE SERIES

In Chapter 11, we calculated the separable solutions for a wave

on a string that is fixed at both ends, at x = 0 and at x = L. In general,
the displacement of such a string is
∞
np x  npct npct 

= y( x, t) ∑ sin
n =1 L 
 an sin
L
+ bn cos
L 
 , (12.2)

where each of the an and bn for n = 1, 2, 3 . . . is an arbitrary constant

that we can set once we know the boundary conditions for any given
 Fourier Series  •  421

problem. The range 0 to L is called the half range because it is half

the maximum wavelength or spatial period.
Consider the case when the string is initially at rest and has ini-
tial displacement y(x, 0) = f(x). Then, by substituting t = 0 into equa-
tion (12.2), we find
∞
np x
f ( x) = ∑ bn sin . (12.3)
n =1 L

Given any physically reasonable function, f(x), can we find the coef-
ficients, bn, such that equation (12.3) is satisfied? Remarkably, yes!
This is known as Fourier’s theorem.
Equation (12.3) is the Fourier half range sine series of a func-
tion, f(x). This is a very powerful result. It tells us that, within the
range 0 to L, we can write any (physically reasonable) function as a
sum of sine waves.

12.3.  FOURIER SINE SERIES COEFFICIENTS

The Fourier half range sine series coefficients, bn, are given by

2 L np x
bn = ∫ f ( x)sin dx. (12.4)
L 0 L

Derivation of the Fourier Sine Series Coefficients

The formula for the bn can be derived directly from the Fourier
series representation (Equation 12.3). First, multiply both sides of
equation (12.3) by sin(mpx/L), then integrate from 0 to L. This gives
∞
mp x np x mp x
dx = ∑ bn ∫ sin
L L
∫0
f ( x)sin
L n =1
0 L
sin
L
dx.

The integral on the right hand side is a standard integral (equa-

tion 12.35) with result (L/2)dnm, where dnm is the Kronecker delta
defined by dnm = 1 if m = n and dnm = 0 otherwise. Essentially,
when the two sine waves in the integral on the right have a different
wavelength they interfere destructively and cancel to zero. We get
422  •  Mathematical Physics

a non-zero result for the integral only when the wavelengths are the
same and hence when n = m.
Substituting in the result from equation (12.35), we have
∞
mp x L L
∑=
L
∫0
=
f ( x)sin
L
dx b d
2
n =1
n nm bm
2
.

Finally, rearranging this equation and replacing the symbol m with

n we find equation (12.4), the formula for the Fourier series sine
coefficients.

12.4.  USING THE FOURIER SERIES RESULTS

We can now use the results from equation (12.3) and equation
(12.4) to find the Fourier half range sine series for any function, f(x).

Example 12.1 Calculate the Fourier series representation of

the function f(x) = 1 in 0 ≤ x < L.
We wish to represent the function f(x) as a Fourier sine series,
∞
np x
f ( x )= 1= ∑b
n =1
n sin
L
.

To do this, we simply need to calculate the appropriate Fourier

coefficients bn using equation (12.4):

2 L np x 2 L np x
=bn =∫ f ( x ) sin dx ∫ sin dx
L 0 L L 0 L
L
2 L  np x 
= − cos
L np  L  0
2 L
= −
L np
( cos ( np ) − 1) .
 Fourier Series  •  423

We can simplify this equation using cos(np) = (–1)n. Thus

=
bn
2
np
(
1 − ( −1 )
n
) 

0 if n iseven
 (12.5)
= 4
if n isodd.
 np

So the Fourier sine series of f(x) = 1 for 0 ≤ x < L is

4 np x 
1= ∑ np sin
n odd L
, (12.6)

where the notation “n odd” simply means to take only the

odd integer terms in the summation. We could also write this
explicitly by defining a new integer counter, m = 0, 1, 2, . . ., and
setting n = 2m + 1 so that n is always odd:

∞
4 ( 2m + 1) p x
1= ∑ sin .
m =0 ( 2 m + 1) p L

Figure 12.2 illustrates how the representation of f(x) = 1

is built up by adding together sine waves from the series.
Writing out the first few terms in equation (12.6) explicitly,
we have

4  px 1 3p x 1 5p x 
1=  sin + sin + sin + .
p L 3 L 5 L 

As we add each successive sine term from the infinite series we

get closer and closer to an exact representation of the function
(Figure 12.2: right).
424  •  Mathematical Physics

y y

1 1

0 0
0 L x 0 L x

FIGURE 12.2:  Left: the function f(x) = 1 in 0 ≤ x < L. Right: the first three partial sums of its
Fourier sine series, dashed line f1 = 4 sin(px/L)/p, dotted line f2 = f1 + 4 sin(3px/L)/(3p), solid
line f3 = f2 + 4sin(5px/L)/(5p). As we add each successive term from the infinite series we get
closer and closer to the exact representation of the function.

Exercise 12.1 If f(x) is given by f(x) = x in 0 ≤ x < L, show that the

Fourier sine series coefficients are given by bn = 2L(–1)n+1/(np).
Hence show that the Fourier sine series for f(x) is
2L  p x 1 2p x 1 3p x 
x=  sin − sin + sin −   . (12.7)
p  L 2 L 3 L 

Figure 12.3 shows f(x) = x and the first three partial sums of the
Fourier sine series of f(x).

y y
L 1

0 0
0 L x 0 L x

FIGURE 12.3: Left: the function f(x) = x in 0 ≤ x < L. Right: the first three partial sums
of its Fourier sine series, dashed line f1 = 2L sin(px/L)/π, dotted line f2 = f1 – 2L sin(2px/L)/(2p),
solid line f3 = f2 + 2L sin(3px/L)/(3π).
 Fourier Series  •  425

Exercise 12.2 A function g(x) is defined by

x L if 0 ≤ x < L 2
g ( x) = 
1 − x L if L 2 ≤ x < L.

By expanding g(x) as a Fourier sine series show that

n −1
4 ( −1 ) 2 np x
g ( x) = ∑ 2 2
sin .
n odd np L

Hint: the integral for bn can be split into the sum of two parts
b c b
of the form ∫=
f ( x)dx ∫ f ( x)dx + ∫ f ( x)dx where a < c < b. See
a a c

Figure 12.4 for a depiction of f(x) and the first three partial sums
of its Fourier sine series.

y y
0.5 0.5

0 0
0 L x 0 L x

FIGURE 12.4:  Left: the function g(x) from exercise 12.2. Right: the first three partial sums of
its Fourier sine series, dashed line f1 = 4sin(px/L)/p2, dotted line f2 = f1 – 4sin(3px/L)/(9p2),
solid line f3 = f2 + 4sin(5px/L)/(25p2).

12.5.  APPLICATION TO THE WAVE EQUATION

The separable solutions to the wave equation for a string fixed at

x = 0 and x = L are
∞
np x  np ct np ct 
=
y ( x, t ) ∑ sin
n =1 L 
 An sin
L
+ Bn cos
L 
 . (12.8)
426  •  Mathematical Physics

Armed with results 12.3 and 12.4, we can now find the coeffi-
cients An and Bn given a set of initial conditions.
Let us examine the general case when the string is given an
initial displacement, y(x, 0) = p(x), and an initial velocity, yt(x, 0)
= q(x). By substituting t = 0 into equation (12.8) we immediately
find that
∞
np x
y( =
x,0 ) p=
( x) ∑B
n= 1
n sin
L
.

This looks like the Fourier sine series of p(x) with Fourier series
coefficients Bn. So, to find the coefficients Bn we simply need to
apply the formula in equation (12.4):

2 L np x
p ( x ) sin
L ∫0
Bn = dx . (12.9)
L
We can follow a similar process to find the An. First, find the
transverse velocity of the string:

∂y ∞
np x  np c np ct np c np ct 
yt ( x, =
t) =
∂t
∑ sin
n =1 L 
 An
L
cos
L
− Bn
L
sin
L 
.

So at t = 0
∞
 An np c  np x
yt ( x=
,0 ) q=
( x) ∑ 
n =1 L 
 sin
L
.

Again, this looks like the Fourier sine series of q(x) with Fourier
series coefficients Annpc/L. So to find Annpc/L we simply need to
apply the formula in equation (12.4):

 An np c  2 L np x
  = ∫0 q ( x ) sin dx,
 L  L L
giving

2 L np x
An =
n pc ∫ q ( x ) sin L dx . (12.10)
0
 Fourier Series  •  427

Example 12.2 A string fixed at x = 0 and at x = L is given

constant initial velocity, yt(x, 0) = v, and zero initial displacement,
y(x, 0) = 0. Find y(x, t).
In general,
∞
np x  np ct np ct 
=y ( x, t ) ∑ sin
n =1 L 
 An sin
L
+ Bn cos
L 
.

To find y(x, t) given a set of initial conditions, substitute the initial

conditions into the general solution, find equations involving the
unknown coefficients An and Bn, then calculate An and Bn using the
formula for the Fourier sine series coefficients, equation (12.4).
At t = 0 the displacement of the string is zero, so
∞
np x
0 = ∑ Bn sin
n =1 L
giving Bn = 0.
At t = 0 the initial velocity is v, so
∞
 A np c  np x
v = ∑ n  sin .
n =1  L  L
This is just the Fourier series representation of a constant, v. So
 An np c  2 L np x
  = ∫0 v sin dx.
 L  L L
Then using the result from equation (12.5),
0 if n iseven
 An np c  
 =
  4v
 L   if n isodd
 np
from which
0 if n iseven
An =  2 2
4 vL n p c if n isodd.
Once we have calculated the An and Bn we can write down
the full solution y(x, t) that describes the displacement of
the string as a function of x and t. In the case when the initial
displacement is 0 and the initial velocity is v,
4 vL np x np ct
y ( x, t ) = ∑ n p c sin
n odd
2 2
L
sin
L
. (12.11)
428  •  Mathematical Physics

Exercise 12.3 The initial displacement of a string of length L

fixed at its end points x = 0 and x = L is given by y(x, 0) = ax,
where a is a constant. The initial velocity is zero. Find the solution
for y(x, t) as an infinite series.

Exercise 12.4 A string is fixed at its end-points at x = 0 and x = L.

If the initial displacement is y(x, 0) = sin(px/L) and the initial
velocity is yt(x, 0) = x, find the solution for y(x, t) as an infinite
series.

Exercise 12.5 A string fixed at its end-points x = 0 and x = L

is released from rest with initial displacement y(x, 0) = exp(–a2
(x – L/2)2) where a >> 1/L. Find the displacement y(x, t) at time t.
You may assume that if aL is large then
L −a 2 ( x − L 2 )
2
np x p − n2 p 2 np
∫
4 L2a 2
e sin dx ≈ e sin .
0 L a 2
Figure 12.5 shows the displacement of the string, y(x, t), at four
different times.

y y y y
1 1 1 1

0 0 0 0
L x L x L x L x

—1 —1 —1 —1

FIGURE 12.5:  Wave from exercise 12.5. From left to right the panels show the displacement
(solid line) and transverse velocity (dashed line) of the string at t = 0, t = L/4c, t = 3L/4c and
t = L/c.

12.6. APPLICATION TO OTHER DIFFERENTIAL

EQUATIONS

We can use the techniques from Section 12.5 to find the solu-
tions to other differential equations.
As a brief example, let us consider the solution to the Laplace
equation for the electrostatic potential f(x, y) on a metal plate,
 Fourier Series  •  429

fxx + fyy = 0. Suppose the potential f(x, y) → 0 as y → ∞ and is set

to zero at x = 0 and x = L; then
∞
np x − np y L
f ( x, y ) = ∑ Bn sin e .
n =1 L
If the potential at y = 0 has the form p(x), then
∞
np x
f ( x=
,0 ) p=
( x) ∑B
n =1
n sin
L
,

and we can calculate the coefficients simply by applying the formula

for the Fourier sine series coefficients,
2 L np x
Bn = ∫ p ( x ) sin dx.
L 0 L
Figure 12.6 shows a solution to the Laplace equation, f(x, y),
satisfying the boundary equations in Exercise 12.6.

Exercise 12.6 The electric potential f(x, y) has boundary con-

ditions f(x, 0) = x2, f(0, y) = 0, f(L, y) = 0 and f(x, y) → 0 as
y → ∞. Show that

 2 L2  np x − np y L (12.12)
4 L2
( ( −1) )
∞
f (=
x, y ) ∑ ( ) n +1 n
 −1 + − 1  sin e .
n =1  np n3 p 3  L

L 1 L

0.5

0 0 0
0 L 0 L

FIGURE 12.6:  φ(x, y) from equation (12.12). Left: map of the electric potential. Right:
equipotential lines.
430  •  Mathematical Physics

12.7.  FOURIER HALF RANGE COSINE SERIES

Another representation of a function defined in the range 0 ≤ x < L

uses cosine instead of sine functions. This is equally valid: within the
range 0 ≤ x < L both sines and cosines form mathematically complete
sets. This means we can expand any function within this range in
terms of either sines or cosines.
The cosine representation of a function f(x) is
∞
1 np x
( x)
f= a0 + ∑ an cos . (12.13)
2 n =1 L
1
The factor of in front of a0 is simply for convenience in that it
2
allows us to give a single formula for the an for all n.
The coefficients an, n = 0, 1, 2, . . . are given by
2 L np x
f ( x ) cos
L ∫0
an = dx. (12.14)
L
It is often easier to perform the calculation for a0 separately from all
the other an. Then a0 is simply
2 L
f ( x ) dx. (12.15)
L ∫0
a0 =
1
Note that a0 is simply the average of the function f(x) in the range
0 ≤ x < L. 2

Derivation of Fourier Cosine Series Coefficients

The derivation for an is along similar lines to that of the Fourier
sine series coefficients. First multiply both sides of equation (12.13)
by cos(mpx/L) and integrate from 0 to L:
L mp x 1 L mp x
∫ f ( x ) cos L dx = 2 a ∫
0
0
0
cos
L
dx 
∞
(12.16)
np x mp x
+ ∑ an ∫ cos
L
cos dx.
n =1
0 L L
 Fourier Series  •  431

From the standard integral (Equation 12.36) all the terms in the
sum on the right of equation (12.16) are zero except for the one with
m = n. Hence, if m = n,
L mp x L
2 mp x L
∫ f (=
0
x ) cos
L
dx m ∫ cos
a=
0 L 2
am ,

and by rearranging and replacing m by n we have the result in

equation (12.4).
If we choose m = 0 in equation (12.16), then cos(mpx/L) = 1 and
all the terms in the sum go to zero. We are left with
L 1 L L
f ( x ) dx
∫= =a0 ∫ dx a0 ,
0 2 0 2
which gives the result for a0 in equation (12.15).

Exercise 12.7 The function f(x) = x in 0 ≤ x < p is expanded as

a Fourier cosine series. Calculate the coefficients a0 and an, and
show that
p 4 1
x= − ∑ 2 cos nx. (12.17)
2 p n odd n 

The function f(x) = x and the first three partial sums in its Fourier
cosine series are shown in Figure 12.7.

y y
L L

0 0
0 L x 0 L x

FIGURE 12.7:  Left: the function f(x) = x in 0 ≤ x < p. Right: the first three partial sums of its
Fourier cosine series, dashed line f1 = p/2 – 4cos(x)/p, dotted line f2 = f1 – 4 cos(3x)/(9p), solid
line f3 = f2 – 4 cos(5x)/(25p). As we add each successive term from the infinite series we get
closer and closer to the exact representation of the function f(x) = x.
432  •  Mathematical Physics

12.8.  NUMERICAL SERIES

Both sine and cosine Fourier representations can be used to

derive interesting numerical series results for constants. Take, for
example, the result from equation (12.17). Writing out the first few
terms in the cosine series for x defined between 0 and π we have

p 4 1 1 
x=−  cos x + cos 3 x + cos 5 x + . . .  .
2 p 9 25 
This equation is true for any x where 0 ≤ x < p. We can substi-
tute a particular value for x into both the left and right sides and the
equality will still hold. Let us choose, for example, x = 0. Then all the
values cos nx = 1 and so

p 4 1 1 
0= − 1 + + + . . .
2 π 9 25 
giving
 1 1 
p 2= 8  1 + + + . . .,
 9 25 
which is a series expansion for p2.

Exercise 12.8 A function f(x) is defined by

1 0 ≤ x < L 2
f ( x) = 
0 L 2 < x < L.
Expand f(x) as a Fourier cosine series and show that an = 0 if n is
even and that, if n is odd,

2
=
an ( −1)( n−1) / 2 . (12.18)
np
Write down the cosine series for f(x) in 0 ≤ x < L and deduce that

p 1 1 1
=1 − + − .
4 3 5 7
 Fourier Series  •  433

12.9.  PERIODIC EXTENSION OF FOURIER SERIES

So far we have examined the sine and cosine Fourier series rep-
resentations of functions within a limited range, 0 ≤ x < L. However,
both sine and cosines repeat periodically. So, if we plot the Fourier
series representations outside of this range we will get functions that
repeat periodically with wavelength 2L.
Outside the given finite range, the Fourier series of f(x) repre-
sents a periodic extension of the function with f(x + 2L) = f(x).
To understand the periodic extension of Fourier series it is impor-
tant to first understand the symmetry of even and odd functions. There
is a short section on even and odd functions in Chapter 3.

12.10. EVEN AND ODD SYMMETRY OF PERIODIC

FUNCTIONS

Sine waves are odd, so any Fourier sine series representation of

a periodic function must have odd symmetry. Similarly, cosine waves
are even, so any Fourier cosine series representation of a periodic
function must have even symmetry.

y y y

–L 0 L 2L x —L 0 L 2L x —L 0 L 2L x

FIGURE 12.8:  Left: Fourier sine or cosine representation of f(x) = x within 0 ≤ x < L. Center:
periodic extension of sine series representation of f(x). Right: periodic extension of cosine
series representation of f(x).

Figure 12.8 shows sine and cosine representations of f(x). Earlier

we saw that within 0 ≤ x < L we could expand the function f(x) = x
either as a sine series, or as a cosine series (Figure 12.8: left). However,
because sines and cosines have different symmetry, when we expand the
range we obtain different shape waveforms for each of the series: the
“sawtooth” wave for the sine series (Figure 12.8: center) has odd symme-
try, while the “triangle” wave for the cosine series (Figure 12.8: left) has
even symmetry. For both periodic extensions, f(x) = f(x + 2L). For the
sine series we also have f(x) = –f(–x) and, for the cosine series, f(x) = f(–x).
434  •  Mathematical Physics

Example 12.3 Sketch the Fourier cosine representation of

f(x) = x3 in the range —L ≤ x < 3L.

f(x)

First sketch the function f(x) = x3

between 0 and L.
0 L

Next, we know that cosines are even

f(x)
functions, so the Fourier cosine series
must be even. This allows us to sketch
the Fourier cosine series between –L and
0.
-L 0 L

f(x)
Cosines and hence the Fourier cosine
series must repeat every 2L. So the
shape of the waveform between –L and
L simply repeats, and we can sketch the
-L 0 L 2L 3L cosine series of f(x) = x3 between –L
and 3L.

Exercise 12.9 A function is defined by f(x) = x2, for 0 ≤ x < L.

Sketch the Fourier sine and cosine series representations of f(x)
in –L < x < 3L.

Exercise 12.10 A function g(x) is defined by.

x L if 0 ≤ x < L 2
g ( x) = 
1 − x L if L 2 ≤ x < L.
Sketch the Fourier sine and Fourier cosine representations of
g(x) in –L ≤ x < 3L.

Exercise 12.11 Within 0 ≤ x < L, f(x) = x can be expanded as

a sine series equation (12.7) or a cosine series equation (12.17).
What functions do each of these series represent in the range
–L ≤ x < 0?
 Fourier Series  •  435

Figure 12.9 shows three examples of waveforms that are com-

mon in signal analysis. We can obtain a square wave by expanding
the range of the Fourier half range sine series representation of
f(x) = 1. We can similarly obtain a triangle wave from the Fourier half
range cosine series representation of f(x) = x from Exercise 12.7. But
what happens if we consider a function, f(x), defined in the full range
from x = –L to x = L? If f(x) is neither even nor odd then we cannot
expand it as only a sum of sine waves or only a sum of cosine waves.
Instead we must use a Fourier full range series (Section 12.11).
The rectified half wave (Figure 12.9: right) is one such function.
This is obtained from the full series representation of a function
defined by f(x) = 0 where –L ≤ x < 0, f(x) = sin(px/L) where 0 < x < L.

FIGURE 12.9:  Some common waveforms: square wave (left), triangle wave
(center), rectified half wave (right).

12.11.  FOURIER FULL RANGE SERIES

The range 0 to 2L (or, alternatively, –L to L) is called the full

range because it contains a full wavelength of the periodic function.
In this section we will mostly use the range –L to L; however, the
ranges 0 to 2L and –L to L are exactly equivalent.
In the range –L to L neither sine nor cosine waves form a com-
plete set. If a function defined between –L and L has odd symmetry
it can be represented as a sine series. If a function has even sym-
metry it can be represented as a cosine series. However, in the gen-
eral case, to represent a function of arbitary symmetry, we need to
include both sine and cosine terms in the representation.
The full range Fourier series for f(x) in the range –L < x < L is
∞
1  np x np x 
f ( x) = a0 + ∑  an cos + bn sin  . (12.19)
2 n =1  L L 
436  •  Mathematical Physics

The formulae for the Fourier series full range coefficients can
be derived in a similar way to the formulae for the sine and cosine
half range coefficients. We have
1 L
f ( x ) dx
L ∫− L
a0 = 
1 L np x
an = ∫ f ( x ) cos dx (12.20)
L − L L
1 L np x
bn = ∫ f ( x ) sin dx.
L − L L

Note that if the full range is defined to be between 0 to 2L the

formulae remain the same except that the limits of the integration
are from 0 to 2L.

Exercise 12.12 A function f(x) = 1 + x within –p ≤ x < p. Calcu-

late the Fourier full range series of f(x).

Exercise 12.13 Calculate the full range Fourier series for a

“sawtooth” wave, f(x) = x, –p < x < p. Explain why the series is the
same as the half range sine representation in Exercise 12.1. By writ-
ing out the result for an appropriately chosen value of x, show that
∞
( −1)m p
∑ 2m + 1 = 4 .
m =0

12.12.  COMPLEX FORM OF FOURIER SERIES

Instead of equation (12.19) we could equally well write the com-

plex form
∞

f ( x) = ∑ce
n =−∞
n
inp x L
, (12.21)

where

1 L
cn = ∫ f ( x ) e− inp x L dx. (12.22)
2L − L
 Fourier Series  •  437

Sometimes this form is more convenient than the sine and cosine
forms.

Derivation of Complex Series Results

Recalling that sines and cosines can be written in terms of com-
plex exponentials we can obtain equation (12.21) directly from equa-
tion (12.19):
px px 2p x 2p x
f ( x) =
a0 + a1 cos + b1 sin + a2 cos + b2 sin +...
L L L L
e ip x L + e− ip x L e ip x L − e− ip x L
=
a0 + a1 + b1
2 2i
e i 2 p x L + e− i 2 p x L e i 2 p x L − e− i 2 p x L
     + a2 + b2 + .
2 2i

Collecting together exponentials with the same powers we have

 ( a2 2 − b2 ( 2 i ) ) e− i 2 p x L + ( a1 2 − b1 ( 2 i ) ) e− ip x L + a0
f ( x ) =+

+ ( a1 2 + b 1 ( 2 i ) ) e i px L + ( a 2 2 + b 2 ( 2 i ) ) e i 2 px L +

=  + c−2 e− i 2 p x L + c−1 e− ip x L + c0 + c1 e ip x L + c2 e i 2 p x L + ,

which is identical to equation (12.21).

To find the formula for the complex coefficients, cn, multiply
both sides of equation (12.21) by e-ipmx/L and integrate:
∞

∑c ∫
L L
∫ f ( x ) e− imp x L dx = n e i( n− m )p x L dx.
−L −L
n = −1

The integral on the right is a standard integral, equation (12.40).

Using this result we find
∞

∑ c 2 Ld
L
∫ f ( x ) e− imp x L dx = n nm . (12.23)
−L
n =−∞

Then rearranging equation (12.23) for cn we obtain the result in

equation (12.22).
438  •  Mathematical Physics

Exercise 12.14 If f(x) = 1 + x, –L < x < L, show that the complex

Fourier series coefficients are given by

1 n=0
cn = 
( −1 ) L ( inp ) n ≠ 0.
n +1

Exercise 12.15 The function f(x) = exp(px) in –p ≤ p < p. Expand

f(x) as a sum of complex exponentials to show that
∞
( −1)n
f ( x) = ∑ p ( p − in ) sinh ( pp ) e
n = −∞
inx
.

Example 12.4 Verify that the result for the complex

Fourier series in Exercise 12.14 is equal to that obtained in
Exercise 12.12.
The complex Fourier series for f(x) = 1 + x in –L < x < L is

−1
( −1)n+1 L ∞
( −1)n+1 L
1+ ∑
f ( x) = e inp x L
+∑ e inp x L .
n =−∞ inp n =1 inp

Then replacing n by –n in the first summation, and factorizing

out –L/π, we have

− L ∞  ( −1 ) i( − n)p x L ( −1 ) inp x L  
−n n

f ( x) =
1+ ∑
p n=1  i ( − n )
e +
in
e 

− L ∞ ( −1 )
n

=
1+ ∑
p n=1 in 
 e inp x L − e− inp x L 

2 ( −1 ) L
∞ n +1
np x (12.24)
= 1+∑ sin ,
n =1 np L
where, in the last step, we have used the fact that
2 i sin ( np x L ) .
e inp x L − e− inp x L =
 Fourier Series  •  439

12.13.  PROPERTIES OF FOURIER SERIES

Here we quote without proof some facts about Fourier series

that you should know.

General Properties
a) Except for some pathological functions which do not occur in
physical problems, we can always expand a function, f(x), defined
in a finite interval as a Fourier series which will converge with
sum f(x) at all points at which f(x) is continuous.
b) If f(x) has a discontinuity at x = x0 its Fourier series will con-
verge to the average of the limit from the left and the limit
y from the right. The Fourier series of
a discontinuous function sums to a
value midway along the discontinuity.
f (x0)
1
x0 x f ( x0 ) lim
=
Œ →0 2
( f ( x0 + Œ ) + f ( x0 − Œ ) ) .
c) A Fourier series can be integrated term by term: the resulting
series always converges to ∫ f ( x ) dx .
d) Term by term differentiation of a Fourier series may produce a
divergent series. If the series produced by differentiation does
converge, then it is the Fourier series for f ¢(x).

Convergence of Series
The properties above apply to the full Fourier series containing
an infinite number of terms. In practice we often calculate the sum
of only a finite number of the terms in the series which we use as an
approximation. It is important to know when this is likely to give a
good approximation.
a) If f(x) or its periodic extension has discontinuities we expect an,
bn, cn to be of order 1/n and the convergence is slow.
b) If f(x) or its periodic extension is continuous we expect an, bn, cn
to be of order 1/n2 and the convergence is rapid.
c) The Gibbs phenomenon. At a discontinuity (at x0, say) conver-
gence of a Fourier series is slow and a finite sum of N terms will
persistently under- or over-estimate f(x) near x0. The size of the
440  •  Mathematical Physics

overshoot (or undershoot) does not tend to zero as N → ∞, but

the region of the overshoot (or undershoot) does become nar-
rower as the series converges.
y
The figure on the left
shows the partial Fourier
sine series of f(x) = 1, 0 ≤
x x < p with N = 3 (dashed),
N = 5 (dotted) and N = 25
(solid line) terms.

Exercise 12.16 By using the results above briefly explain why

the convergence of the half range cosine series representation of
x (Equation 12.17) is much faster than that of the sine series rep-
resentation (Equation 12.7).

Exercise 12.17 State whether each of the following functions,

defined in the range −p < x < p, can be expanded as (a) a Fourier
sine series, (b) a cosine series or (c) a Fourier series containing
both sine and cosine terms: (i) x3; (ii) x2 + x; (iii) exp(x); (iv) |x|.

12.14. FOURIER TRANSFORMS AND FOURIER

INTEGRALS

So far, we have seen that we can express an arbitrary periodic

function as a Fourier series, a sum of harmonic (sine and cosine)
waves.
The Fourier transform gives us an analogous way to represent a
general function that is not periodic. Fourier transforms are used in
an enormous range of pure and applied science, including informa-
tion processing, electronics and communications.
The Fourier series of a function f(x) with period λ = 2L can be
written as
∞ ∞

=f ( x) ∑
= ce
n= −∞
∑c e
n
inp x L

n= −∞
n
2 inp x l
. (12.25)
 Fourier Series  •  441

So what happens if the function is non-periodic? A non-periodic

function is equivalent to a periodic function in the limit that λ → ∞.
So, let us consider what happens to 12.25 when λ becomes large.
First we define a new variable k = 2pn/λ. Then we can think of
the Fourier coefficients cn as a function f ( k ) = ckl 2 p defined on a
line of points in k-space. The distance between the points in k-space
is ∆k = 2p/λ. Then we can rewrite the sum in equation (12.25) as
∞ ∞
1 ∞
f ( x)
= ∑
= ce f ( k ) e
∑= n
ikx ikx
∑ F ( k ) e ikx ∆ k
n = −∞ n = −∞ 2 π n= −∞
where F (k ) = l f (k ) . In the limit λ → ∞ the spacing between the
points, ∆k, tends to zero and the sum becomes an integral:
1 ∞
f ( x) = F ( k ) e ikx dk. (12.26)
2 π ∫−∞

This is a Fourier integral. It is a representation of an arbitrary (non-

periodic) function, f(x), in terms of simple harmonics.

12.15.  FOURIER TRANSFORMS

To obtain a result for F(k) in equation (12.26) we simply apply

the formula (Equation 12.22) for the Fourier series coefficients,
cn = ckλ/2π, in the limit as λ → ∞,
∞
F ( k ) lim
= = l ckL 2 π ∫ f ( x ) e− ikx dx. (12.27)
λ→∞ −∞

F(k) is called the Fourier transform of f(x). The two functions, F(k)
and f(x), are called a Fourier transform pair; f(x) is the inverse Fou-
rier transform of F(k).
Unfortunately there is no standard definition in the literature
of what constitutes the transform and what constitutes the inverse
transform. The only requirement is that one of equations (12.26)
and (12.27) contains e-ikx and one contains eikx. Similarly there is no
set convention for the constant factors in front of these integrals.
We have used a factor 1/2p on the inverse transform and 1 on the
442  •  Mathematical Physics

transform, but you will often see the opposite of this, or sometimes
1 2p is used in front of both. This means that, when reading the
literature, you should be careful to identify the conventions used.
The argument of the Fourier transform, k, has units that are the
reciprocal of the dimensions of the variable x. We will often use x to
denote position, in which case k is the wavenumber. Similarly, if the
variable is a time, t, the transform variable will be a frequency, often
denoted by w. Physicists often talk about using Fourier transforms
to transform from real space to reciprocal or k-space, or from the
time-domain to the frequency-domain.

12.16.  CALCULATING FOURIER TRANSFORMS

Fourier transforms and integrals are ordinary integrals that can

be evaluated in the usual way. We will consider a specific example:
the Fourier transform of a Gaussian.

Example 12.5 Find the Fourier transform of

( x ) exp ( − x2 s 2 ) . From equation (12.27) we have
f=
∞ ∞ (
− x2 s 2 + ikx )
=F ( k) ∫=
e e dx ∫
− x2 s 2 − ikx
e dx.
−∞ −∞

This integral is performed with a standard trick. First, we

complete the square in the argument of the exponential,

( x s + iks 2 )
2
x2 s 2 + ikx = + k2s 2 4.

Then
∞ − ( x s + iks 2 )
2

F ( k ) = e− k s ∫
2 2
4
e dx,
−∞ (12.28)
∞
∫
− k2s 2 4 − x′2
=s e e dx′,
−∞

where we have changed variable from x to x¢ = x/s − iks/2, so

dx¢ = dx/s. In this particular case, the change of variable leaves
the limits on the integral unchanged at −∞ and ∞. Then, using
 Fourier Series  •  443

exp ( − x2 ) dx =
∞
the fact that ∫−∞
p, we find that the Fourier
transform of the Gaussian is

F ( k ) = s p e− k s
2 2
4
. (12.29)

So the Fourier transform of a Gaussian of half width s is a

Gaussian of half width 2/s; thus a wide Gaussian function in
real space transforms to a narrow Gaussian function in k-space
and vice-versa.

Exercise 12.18 Show that the Fourier transform of the function

defined by f(x) = a for |x| ≤ L, f(x) = 0 for |x| > L is
2 a sin kL
F ( k) = .
k

Exercise 12.19 Find the inverse Fourier transform of F(k) =

exp(−k2/4).

12.17. CONVOLUTIONS

Here we will briefly introduce a useful result: the convolution

theorem.
A convolution integral has the general form
∞
C( x)
= ∫ f ( x − x′ ) g ( x′ ) dx′. (12.30)
−∞

This integral relates an output function, C(x), to the input g(x)

and the response f(x). Figure 12.10 shows an example from imag-
ing processing where we degrade an initially sharp image with
a Gaussian “blur.” The process is illustrated in one dimension
in Figure 12.11: we convolve the input, g(x), with a Gaussian
response function, f(x) = exp(−x2), to give the new blurred output
444  •  Mathematical Physics

∞
( x)
C= exp  − ( x − x′ )  g ( x′ ) dx′ . You will often see a similar
∫ −∞
2

effect in experimental physics where some initially sharp “input” sig-

nal will suffer from Gaussian broadening as a result of an imprecise
experimental response.

FIGURE 12.10:  Image processing with convolutions. A sharp image of a circle (left) is con-
volved with a Gaussian response function to give the blurred image (right).

1 1 1

0.5 0.5 0.5

0 0 0
—10 —5 0 5 10 —10 —5 0 5 10 —10 —5 0 5 10

FIGURE 12.11:  Example convolution. Left: sharp “input” signal, g(x) = 1, |x| < 3. Center:
Gaussian response function, f(x) = exp(−x2). Right: blurred output signal,
∞
C( x)
= ∫ f ( x − x′ ) g ( x′ ) dx′ .
−∞

12.18.  CONVOLUTION THEOREM

The convolution theorem is a useful relation between the Fou-

rier components of the input and output functions. According to the
convolution theorem,

C ( k ) = F ( k ) G ( k ) . (12.31)

To prove this relation we simply calculate the Fourier transform of

the output function C(x),
∞ ∞ ∞
C ( k)
= ∫ C (=
x ) e− ikx dx ∫ ∫ f ( x − x′ ) g ( x′ ) e− ikx dx′dx.
−∞ −∞ −∞
 Fourier Series  •  445

Putting u = x − x¢ and using u as the integration variable instead of

x we have
∞ ∞
C ( k) = ∫ ∫ f ( u ) g ( x′ ) e− ik( x + u) dx′du
′
−∞ −∞
∞ ∞
=∫ f ( u ) e− iku du∫ g ( x′ ) e− ikx′ dx′
−∞ −∞

= F ( k) G ( k).

12.19.  FOURIER DECONVOLUTION

The convolution theorem tells us that convolutions in real space

are simply multiplications in reciprocal space. This is very useful if
we want to deconvolve a signal.
For example, let us imagine that, in an experiment, we want
to measure some input g(x). However, because of our experi-
mental response, f(x), we actually measure a degraded output
signal, C(x). How do we get back to the actual input that we
would like to measure? We could try to solve the integral equa-
tion (12.30) for g(x) but it is much easier to make use of the
convolution theorem.
In reciprocal space we can find G(k) by dividing C(k) by F(k).
Then we can easily obtain g(x) with an inverse transform,

1 ∞ C ( k ) ikx
g ( x) =
2 π ∫−∞ F ( k )
e dk. (12.32)

Figure 12.12 shows an illustration of this process. An initially blurred

image is sharpened by deconvolving the image with a Gaussian
response function.

FIGURE 12.12: Sharpening of an image by deconvolution. From left to right:

original blurred image C(x, y); Fourier transform of blurred image, C(kx, ky); response function
in reciprocal space, F(kx, ky); input function in reciprocal space, G(kx, ky) = C(kx, ky)/F(kx, ky);
deconvolved image, g(x, y).
 446  •  Mathematical Physics

Exercise 12.20 Write down the analogous results for the con-
volution integral and convolution theorem, in the time and fre-
quency domain.

∞
Exercise 12.21 If= c ( t ) ∫ f ( t − t ′ ) g ( t ′ ) dt ′ show that C(w) =
−∞
F(w)G(w) where C(w), F(w) and G(w) are the Fourier transforms
of c(t), f(t) and g(t), respectively.

Exercise 12.22 In reciprocal space an experimen-

tal signal that has been broadened with a Gaussian
F ( k)
response function, = p exp ( − ak2 4 ) , has the form
=C (k ) (2 )
p k exp ( −ak 2 4 ) sin k . Use the convolution theo-
rem to write down the input signal, G(k), in reciprocal space, then
transform this into real space to find the original form of the input,
g(x). (Hint: you can often find a Fourier transform simply by recog-
nizing that it is related to a known inverse transform.)

12.20.  EXTENSION: LAPLACE TRANSFORM

We define the Laplace transform of the function f(t) by

∞
F ( s ) ≠ ∫ e − st f (t )dt .
0

Note how this differs from the Fourier transform: we replace the
complex exponential by a real exponential and integrate only from
0 to ∞. The Laplace transform is useful for finding the solutions of
linear differential equations with constant coefficients, particularly
equations with time as the independent variable, because it readily
allows us to incorporate the initial conditions. It has the disadvan-
tage that the inverse transform is not straightfoward; in particular
∞
f ( t ) ≠ ∫ F (s )e − st ds in general. In practice, one can look up tables
0
of transform pairs.
 Fourier Series  •  447

To see how this works, consider the equation of a driven har-

monic oscillator,

x + x =f ( t ) . (12.33)

We take the Laplace transform of both sides by multiplying by e−st

and integrating. We have
∞ ∞ ∞
∫0 e xdt  e x ( t ) 0 + s∫0 e x ( t ) dt
− st − st − st
=
∞ ∞
− x ( 0 ) + s  e− st x ( t )  0 + s2 ∫ e− st x ( t ) dt
=
0

− x ( 0 ) − sx ( 0 ) + s2 X ( s ) .
=

Thus the transform of equation (12.33) gives

F ( s ) + sx ( 0 ) + x ( 0 )
X ( s) = . (12.34)
s2 + 1
We can now find the solution of the differential equation by
looking up the inverse Laplace transforms of s/(s2 + 1) and 1/(s2 + 1)
(which are sin(t) and cos(t), respectively) and, if it is known,
F(s)/(s2 + 1) for the particular form of F(s), the Laplace transform
of f(t). More complicated linear equations can be treated similarly,
although explicit solutions will be obtained only where the inverse
Laplace tranforms can be found.
You may be wondering what is special about the exponential
here and in the Fourier transform. In fact, transforms can be and are
defined for various classes of functions (besides the exponential) for
which an inverse transform can be found. (A transform from which
the original function cannot be retrieved is obviously losing informa-
tion and is unlikely to be useful.)
448  •  Mathematical Physics

Revision notes

After completing this chapter you should be able to

• Appreciate how an arbitrary periodic function can be
represented by an infinite summation of sines and
cosines, the Fourier series of a function
• Use the Fourier series representation to find the
solution of partial differential equations, including the
wave equation, satisfying given boundary conditions
• Distinguish between full range (both complex and
real form) and half range sine and cosine Fourier
representations of a given function
• Write down the formulae for the Fourier series
coefficients in each case
• Find the Fourier half range and full range series
representations of a given function in a finite interval
• Understand the periodic extensions of Fourier series
• Appreciate how an arbitrary non-periodic function can
be represented by a Fourier transform
• Calculate the Fourier transform and inverse transform
of a given function
• Prove the convolution theorem
 Fourier Series  •  449

12.A.  APPENDIX: STANDARD INTEGRALS

In this appendix, we gather together some standard integrals

that are useful for Fourier series.

0 if m= n= 0
L np x mp x  (12.35)
∫ sin sin dx =  L
L L  2 d mn otherwise
0


L if m= n= 0
L np x mp x 
∫0 cos cos dx = L (12.36)
L L  2 d mn otherwise

if m= n= 0 
L np x mp x 0
∫ −L
sin
L
sin
L
dx = 
 Ld mn otherwise
(12.37)

L np x mp x 2 L if m= n= 0
∫−L
cos
L
cos
L
dx = 
 Ld mn otherwise  (12.38)

L np x mp x

∫−L
sin
L
cos
L
dx = 0 (12.39)

L
∫ e i( n− m ) πx L dx = 2 Ld nm . (12.40)
−L

Derivation of Standard Integrals

The results for the standard integrals 12.35, 12.36, 12.37, 12.38,
12.39 can be found using the product formulae for trigonometric
functions.
For example, to obtain equation (12.35) we write

L mp x np x 1 L ( m − n) p x ( n + m)px 
∫=
0
sin
L
sin
L
dx ∫
2 
0
 cos
L
− cos
L
dx 

 ( m − n ) πx 
L
1 L
= sin 
2 ( m − n) p  L 0
450  •  Mathematical Physics

 ( m + n) p x 
L
1 L
− sin
2 ( m + n ) p  L 
0
= 0 if m ≠ n, ( all the sine terms are zero ) .

If m = n ≠ 0 then the integral becomes

L mp x np x L np x
∫0
sin
L
sin
L
dx = ∫ sin 2
0 L
1 L 2 np x 
=
2 ∫0  1 − cos
L
 dx

L
= .
2

To obtain equation (12.40) we have, if n ≠ m,

 i (n − m )p x  i (n − m )p x
L
L   L 
∫− L exp   dx = 
L  i ( n − m ) p
exp 
L
 
     − L
L
= e i ( n − m )p − e − i ( n − m )p 
i (n − m )
= 0,

because exp(i(n − m)π) = exp(−i(n − m)π) = cos((n − m)π).

However, if n = m then

L  i( n − m)px  L
∫−L
exp 
 L


dx = ∫ dx
−L

           = 2 L.
 Fourier Series  •  451

12.21. EXERCISES

1. A function f(q), −p ≤ q ≤ p is known to have a Fourier series

of the form
∞
1
f (=
q) a0 + ∑ an cos ( nq ) . (12.41)
2 n=0

Show that f(q) = f(−q).

2. If the function f(q) is given in the range 0 ≤ q ≤ π to be
f(q) = q sketch the extension of this function to the range
−p ≤ q ≤ p if it is to have a Fourier series of the form
(12.41). Sketch also the function represented by the
Fourier series outside this range.
3. The function f(q) = q3, 0 ≤ q ≤ p is expanded in (a) a Fou-
rier sine series, and (b) a Fourier cosine series. Sketch the
forms of the functions represented by these series in the
range −3p ≤ q ≤ 3p.
4. Starting from the assumption that any function defined
between −L and L can be expanded as a sum of sine and
cosine functions,
1 ∞
 np x np x 
f ( x) = a0 + ∑  an cos + bn sin ,
2 n =1  L L 
derive the formulae for the Fourier coefficients, a0, an and bn.
5. The following functions are all defined in the range
−L < x < L. For each of the functions, give the symmetry of
the function and state whether the function can be expand-
ed as a half range sine series, a half range cosine series or
whether it must be expanded as a full range Fourier series.
    (i)  1/(|x| + d),
  (ii)  x3 sin(2x),
(iii) exp(−4x),
 (iv)  x exp(ax2),
   (v)  ax4 + bx2 + c.
452  •  Mathematical Physics

12.22. PROBLEMS

1. Find the Fourier series for f(q) = q2, −p ≤ q < p, and by set-
ting q = 0, show that
p2 ∞
( −1) n +1

=∑ .
12 n=1 n2
2. Expand the function f(x) = x as a half range Fourier cosine
series in the range 0 ≤ x < L.
By considering the series for a suitable value of x, show that
∞
1 p2
∑ (2 n + 1 )
n =0
2
=
8
.

What function does the series represent in the range

−L < x < L?
3. Expand the function f(x) = sin(px) as a half range Fourier
cosine series in the range 0 ≤ x ≤ 1.
By considering the series for a suitable value of x show that
∞
( −1) p 1 p
∑ 4p
p =1
2
= − .
−1 2 4

1 1
( x)
4. If f = ( ( x) ( h ( x) − h ( − x)) ,
h ( x ) + h ( − x ) ) and g=
2 2
show that f(x) is even and g(x) is odd.
Derive the following results:

a
0 if h ( x ) isodd
∫− a h ( x ) dx =  a
2 ∫0 h ( x ) dx if h ( x ) iseven.

Hint: first split the integral into two regions, a part from
−a to 0 and a part from 0 to a, then change variable from x
to x¢ = −x in the region where x < 0.
 Fourier Series  •  453

5. Find the Fourier series for the function

a , −p ≤ q < 0
f (q ) = 
b , 0 < q < p ,
where a and b are constants. (Hint: the neatest approach
is to write f(q) as the sum of a symmetric function and an
antisymmetric one.)
By setting q = π/2 show that
∞
(−1)n p
∑
n =0 2 n + 1
=
4
6. Find the Fourier series for
− sin t, −p < t < 0
f (t) = 
sin t, 0 < t < p.
Hence show that
∞
1 1
∑
n = 2,4,6, n − 1
2
= .
2
7. Calculate the full range complex Fourier series representa-
tion of f(x) = x + x2.
8. Show that the Fourier transform of the function defined by
f(x) = exp(−a2x) for x ≥ 0, f(x) = 0 otherwise, is 1/(ik + a2).
9. Calculate the inverse Fourier transform of F(k) = exp(−k2).
 CHAPTER

13
INTRODUCTION TO
VECTOR CALCULUS

In physics we are often concerned with physical properties that

vary over some region of space, e.g. air temperature in a room,
charge density through a solid body, velocity within a fluid. For
these, we need to use the idea of a field. In this chapter we will
discuss fields, and the most important differential operators that
physicists use to work with fields. These are called grad, div and
curl. Here we will study the mathematical definitions and some
rules for manipulating fields and operations on fields, and we will
begin to look into their geometrical interpretations. These will be
crucial to your studies of physics, especially electromagnetic fields,
but also fluid mechanics.
In this chapter, we will consider only three-dimensional space
and use only Cartesian coordinate systems, i.e. using coordinates
(x, y, z) and basis vectors i, j, k.

13.1.  VECTOR FIELDS

A scalar field associates a scalar (a real or complex number) to

each point in space. Let f(P) be a function defined on the points P
in some region of three-dimensional space. For example f might be
the temperature at each point in a room. Then f is called a scalar
456  •  Mathematical Physics

field. In terms of the usual Cartesian coordinates a general point P is

labeled by (x, y, z) and the scalar field f is the function f(x, y, z). So
f takes as input a point in three-dimensional space, or equivalently
three numbers for the coordinate values, and returns as output a
single number.
4
3
2

z
y
1
0

x
–1

–1 0 1 2 3 4
x

(a) (b)
FIGURE 13.1:  Two different ways to visualize a scalar field f(x, y). (a) A heat-map or intensity
image, where the shade represents the value of f. (b) A surface plot where z = f(x, y) is the
height of the surface.

You may wonder why, if f expressed in this way is just a familiar

function, we go to the trouble of calling it a scalar field. The reason
is to emphasize that a scalar field takes on values at the points in a
region and these values do not depend on the coordinates used to
label the points. For example, the temperature obviously depends
on where you measure, but not on what system of coordinates you
use to specify the point where you make the measurement. See
Figure 13.1 for a visualization of a scalar field. Other examples of
scalar fields are the air pressure in a room and the gravitational
potential around a planet.
A vector field associates a vector to each point in space. Let V(P)
be a vector defined on the points P in some region of space (see
Figure 13.2); i.e. V associates a vector with each point. Then V is
called a vector field. For example V might be the velocity at each
point of a fluid. In terms of the usual Cartesian coordinates (x, y, z)
the vector field has components (f(x, y, z), g(x, y, z), h(x, y, z)) for
some functions f, g and h (each of which is a function of position).
Instead of (f, g, h) we usually write
 Introduction to Vector Calculus  •  457

4
3
2
y
1
0
−1

−1 0 1 2 3 4
x
FIGURE 13.2:  A vector field V(x, y): a vector is defined at each point (x, y). Here, vectors are
drawn as arrows at grid points.

V = V(x, y, z) = (Vx (x, y, z), Vy (x, y, z), VZ (x, y, z)),(13.1)

the subscripts indicating that the components are along the x-, y- and
z-axes (not to be confused with shorthand notation for the partial
derivatives of a scalar V(x, y, z). If there is any danger of confusion, the
derivatives would be denoted by V,x etc. with the additional comma.)
A vector field is simply a vector-valued function, i.e. an assign-
ment of a vector (three real numbers) to each point of a region of
space. Examples of vector fields are the electric field in space and
the velocity of fluid in a pipe.
Examples:
•• A = (x, y, z) is a vector field with Ax = x, Ay = y, Az = z at the
point (x, y, z).
•• B = (x3, x2y, y2(z + 1)) is a vector field.
•• V = (0, 0, 2) is a vector field with V equal to this constant
value throughout space.

A large part of this chapter is devoted to discussing differential

operations on scalar fields and vector fields–in other words, different
458  •  Mathematical Physics

kinds of fields that can be constructed out of derivatives of other

fields. Many of these involve the del operator or nabla symbol which
is defined (in Cartesian coordinates) by
∂ ∂ ∂
∇≡i + j +k
∂x ∂y ∂z
∂ ∂ ∂ 
≡  , , . (13.2)
 ∂x ∂y ∂z  

This is not a function; it is an operator, which means it is “waiting

for a function” on which to operate. In the next few sections, we will
see various ways in which it works.

13.2.  THE GRADIENT OF A FUNCTION

Let f(x, y, z) be a scalar field. We now define the gradient of f,

which is a new vector field, by
 ∂f ∂f ∂f 
∇f = , ,  , (13.3)

 ∂x ∂y ∂z 
or equivalently
∂f ∂f ∂f
∇=
f i + j + k .  (13.4)
∂x ∂y ∂z
We take a scalar field f and make a vector field ∇f out of the par-
tial derivatives of f. To emphasize that ∇f is a vector ∇ is often printed
in bold or written as ∇. We read this as “grad f,” and sometimes it is
also written as grad f. To get equation (13.3) we inserted the scalar
function f into the “slots” in the del operator, equation (13.2).

Example 13.1 Let f(x, y, z) = xz; find ∇f and its value at the
point P with coordinates (3, 2, 17).
According to the definition we have to work out the three
partial derivatives. Recall that z is a constant for differentiation
with respect to x etc.
 Introduction to Vector Calculus  •  459

Thus:
∂f ∂f ∂f
= z=, 0,= x, (13.5)
∂x ∂y ∂z 
and so we have
∇f = (z, 0, x).  (13.6)
Substituting the values of x, y and z at P = (3, 2, 17) gives

∇f = (17, 0, 3).  (13.7)

Example 13.2 Let f(x, y, z) = (x2 + y2 + z2)1/2; find ∇f.

Once more, according to the definition, we have to work out
the three partial derivatives of f. Using the chain rule (i.e “the
function of a function” rule):
∂f 1 1 x
= = × 2x . (13.8)
∂x 2 ( x + y + z2 )1 2
2 2
(x 2
+ y + z2 )
2 12


Note that there is no need to differentiate to get fy and fz in this

problem. Since f is symmetrical in x, y and z, fy is obtained from
fx by interchanging x and y in the numerator. (And similarly for
fz.) Hence
∂f y ∂f z
= , and . (13.9)
∂y ( x + y + z )
2 2 2 12
∂z ( x + y + z2 )1 2 
2 2

Now, ∇f is a vector with three components:

 x y z 
∇f = , , .
 ( x2 + y2 + z2 ) ( x2 + y2 + z2 ) ( x2 + y2 + z2 )1 2
1 2 1 2
 (13.10)
 
We can extract a common factor from each component:

∇f = (x2 + y2 + z2)−1/2(x, y, z).  (13.11)

460  •  Mathematical Physics

Exercise 13.1 Find the gradients of the following functions:

(i) f(x, y, z) = (x2 + y2 + z2)−1/2,
(ii) f(x, y, z) = exp(lx + my + nz) where l, m, n are constants.

13.3.  MAXIMUM RATE OF CHANGE

The gradient of a scalar field, ∇f, has a simple geometrical inter-

pretation. Its magnitude |∇f | gives the maximum rate of change of
the function f and the vector ∇f points in the direction of this maxi-
mum rate of change.
Consider two neighboring points in three-dimensional space,
with position vectors r = (x, y, z) and r + dr = (x + dx, y + dy, z + dz),
and let f(x, y, z) be a function of position (a scalar field). We are
going to calculate the change in the value of f as we pass from r to
r + dr:
df = f(r + dr) − f(r).  (13.12)
To do this we add up the change in f as a result of first increasing
x (to x + dx), then increasing y, and finally increasing z. (See Figure
13.3 and compare the chain rule in Section 10.15.) The increase in
f is
d f = f ( x + d x, y + d y, z + d z ) − f ( x, y, z )
∂f ∂f ∂f
≈ d x + dy + dz.  (13.13)
∂x ∂y ∂z
The approximation in the last line is sometimes known as the
incremental approximation. This consists of the first order terms
of the Taylor series for a function of three variables (x, y, z). (See
Section 10.11.)
Let dr = uds where u is a unit vector and ds = |dr|. We can write
equation (13.13), using the scalar product of vectors, as
 ∂f ∂f ∂f 
d f = , ,  ⋅ ( d x, d y, d z ) =∇f ⋅ d r = ( ∇f ⋅ u ) d s.
 ∂x ∂y ∂z 

(13.14)
 Introduction to Vector Calculus  •  461

But the change in f is

df = (rate of change of f in direction u) ×
(distance between the two points). (13.15)

∂f ∂f ∂f
f+ δ x + δy+ δz
∂x ∂y ∂z

f ( x , y, z)
δr
δz

δx
r

δy
∂f
f + δx
∂x ∂f ∂f
f+ δx+ δy
∂x ∂y
FIGURE 13.3:  Changes in f(x, y, z). We start at point r and end up at point r + dr by moving
first along the x-axis, then the y-axis, and then the z-axis.

Comparing equations (13.14) and (13.15), and noting that ds is

the distance between the two points, we see that
(rate of change of f in direction u) = (∇f) · u.(13.16)
If a is the angle between the vectors ∇f and u, then equation
(13.16) becomes (by the definition of the scalar product)

(rate of change of f in direction u) = |∇f | cos(a),(13.17)

remembering that |u| = 1 (as it is a unit vector). Since 0 ≤ |cos(a)| ≤ 1,

from equation (13.17) we have
462  •  Mathematical Physics

(maximum rate of change of f) = |∇f|,  (13.18)

and this will occur when a = 0, i.e. when ∇f and u are parallel. Hence
∇f gives the magnitude and direction
of the maximum rate of increase of f.  (13.19)
You have seen this before – take another look at section 10.3.

Exercise 13.2 Find the direction of the maximum rate of

increase of each of the functions in Exercise 13.1 at a general
point (x, y, z).

Exercise 13.3 Given the scalar field f(x, y, z) = 3x + y2z2

(i) Find the magnitude of the maximum rate of increase of f,
(ii) Find the gradient of the field in the direction (2, 2, 2).

13.4.  LEVEL SURFACES

Consider first a function of two variables f(x, y). The level curves
are defined to be the curves f(x, y) = k for the various possible values
of the constant k. They are the curves along which f(x, y) has a level,
i.e. constant, value.
This idea is familiar from meteorology. If f = pressure, level
curves are isobars; or if f = temperature, level curves are iso-
therms. And in geography, if f = height of land, level curves are
contour lines.
The surface f(x, y, z) = k, with k a constant, is said to be a level
surface of the function f(x, y, z). A set of level surfaces is generated
by taking a range of values for k. The notion of level surface captured
by this definition is not the same as the non-technical use of “level” to
mean a “flat” surface – it is the function that is “flat,” not the surface.
 Introduction to Vector Calculus  •  463

4
3
2

y
1
0
–1

–1 0 1 2 3 4
x
FIGURE 13.4:  Level curves of of the function from Figure 13.1.

Example 13.3 Find level surfaces of the function f(x, y, z)

= z and show that ∇f is normal (i.e. perpendicular) to these
surfaces at any point.
The level surfaces are the planes obtained as the constant runs
through a range of possible values. On a level surface, f = constant.
As f(x, y, z) = z, the level surfaces z = constant are planes parallel
to the (x, y) plane. We now find ∇f. Remember that
∂z ∂z
≡ =
0. (13.20)
∂x ∂x y, z= const 

Here we have that f(x, y, z) = z, and so

∇f = (fx, fy, fz) = (0, 0, 1).  (13.21)

The vector (0, 0, 1) points along the z-axis, and hence is normal
to the (x, y) plane.
464  •  Mathematical Physics

Exercise 13.4 Describe geometrically the level surfaces of the

functions
(i) f = (x2 + y2 + z2)1/2,
(ii) f = x2 + y2,
(iii) f = lx + my + nz (l, m, n constants).

Exercise 13.5 For each of the functions in Exercise 13.4, show

explicitly that ∇f is normal to the level surface at any point.

13.5.  NORMALS TO A SURFACE

The examples in the previous section are special cases of a gen-

eral relationship between level surfaces and the vector gradient:
∇f is always in a direction normal to the surface f = constant.
(13.22)
To see this, consider a scalar function f and a level surface
f(x, y, z) = k. Now let r and r+dr be the position vectors of neigh-
boring points in the level surface. As they are both in the level
surface, f(r) = k and f(r + dr) = k, the difference in f at the two
points is zero, i.e. df = 0. Hence the directional derivative of f
along the vector joining the points is df = ∇f · dr = 0. This means
∇f is normal (i.e. perpendicular) to dr, hence is in the direction of
the normal to the surface.
We have therefore shown that a unit normal to the surface
f(x, y, z) = k is
∇f
n= . (13.23)
∇f 

∇f
Note that − is also a unit normal to the surface (but it points
∇f
in the opposite direction).
 Introduction to Vector Calculus  •  465

4
3
2
1
0
–1

–1 0 1 2 3 4

FIGURE 13.5:  Level curves and gradient vectors of of the function from Figure 13.1. Notice
how the gradient vectors are perpendicular to the level curves (contours). Also notice how the
gradients are larger (longer arrows) when the contours are closer together.

FIGURE 13.6:  Surface defined by z = x2 + y2 – 1; see example 13.4.

Example 13.4 Find the outward pointing unit normal to the

surface z = x2 + y2 − 1 at the point (1, 1, 1).
466  •  Mathematical Physics

Begin by writing the surface in the form f = constant:

f = z − x2 − y2 = constant = −1. (13.24)

See Figure 13.6. (We could equally choose to write f = z−x2−

y2+1 = constant = 0. The result would be unaffected.) Next, we
find ∇f = (fx, fy, fz ):
∇f = (−2x, −2y, 1).  (13.25)
The magnitude of ∇f is |∇f | = [(−2x)2 + (−2y)2 + (1)2]1/2, i.e.

|∇f | = (4x2 + 4y2 + 1)1/2.  (13.26)

We have shown that ∇f/|∇f | and −∇f/|∇f | are unit normals, and
so
 1 
n= ±  ( −2 x, −2 y,1 ) . (13.27)
 ( 4 x2 + 4 y2 + 1 )1 2  
 
Next, we substitute the values of x, y and z at the given point
(1, 1, 1):
n = (−2/3, −2/3, 1/3) or (2/3, 2/3, −1/3). (13.28)
Since the surface expands outwards from the z-axis
above z = −1, the outward normal points “down,” so is
(2/3, 2/3, −1/3).

Exercise 13.6 Find the outward pointing unit normal to the sur-
face x2 + y2 = 8 at the point (2, 2, 1).

Exercise 13.7 Find the outward pointing unit normal to the sur-
face f(x, y, z) = x2 − y2 at the point (−20, −20, 0).
 Introduction to Vector Calculus  •  467

Exercise 13.8 Show that the unit normal to the surface

f = z − g(x, y) = 0 is
12
 ∂g ∂g    ∂g 2  ∂g 2 
±  , , −1   1 +   +    . (13.29)
 ∂x ∂y    ∂x   ∂y  
 

13.6.  THE DIVERGENCE OF A VECTOR FIELD

The divergence is another vector operator. The divergence oper-

ates on a vector field and produces a scalar field. (Compare grad
which operates on a scalar field and produces a vector field.) The
divergence of A(x, y, z), which is read and sometimes also written
as “div A,” is
∂Ax ∂Ay ∂Az
∇ =
⋅ A + + , (13.30)
∂x ∂y ∂z 
where Ax, Ay and Az are the components of A, which are functions
of position (x, y, z).
Look at this expression carefully. The x component of A, Ax, is
differentiated with respect to x, the y component with respect to y
etc. and the results added together to give one function (i.e. a scalar).
There is a reason why we write the divergence using the nabla
and dot product symbols. Let’s write ∇ and A as vectors and multiply
as if taking a dot product.

 ∂ ∂ ∂ 
∇= (
⋅ A  i + j + k  ⋅ iAx + jAy + kAz )
 ∂x ∂y ∂z 
∂A ∂Ay ∂Az
= x+ + . (13.31)
∂x ∂y ∂z

We have used the results for the dot products of the basis vec-
tors: i · i = 1, i · j = 0, and so on.
468  •  Mathematical Physics

Example 13.5 Let r = (x, y, z). Find ∇ · r.

First, identify the three components of the vector field,

rx = x, ry = y, rz = z. (13.32)

Now work out the required partial derivatives:

∂rx ∂ry ∂rz

= 1,= 1,= 1.  (13.33)
∂x ∂y ∂z

Finally, substitute in the definition of ∇ · A:

∇· r = 1 + 1 + 1 = 3. (13.34)

The answer is always a scalar function, in this case a constant.

The vector field r = (x, y, z) is simply the radius vector at each
point (Figure 13.7).
4
2
0
y
–2
–4

–4 –2 0 2 4
x
FIGURE 13.7:  The vector field r = (x, y, z) shown in the plane z = 0. See Example 13.5.
 Introduction to Vector Calculus  •  469

Example 13.6 Let A = (x3, x2y, y2(z + 1)). Find ∇ · A at the

point (1, 0, 0).
First we identify the three components of the vector field

Ax = x3, Ay = x2y, Az = y2(z + 1). (13.35)

Next, work out the required partial derivatives

∂Ax ∂Ay ∂Az

= 3 x=
2
, x=
2
, y2 .  (13.36)
∂x ∂y ∂z
Now we can determine the divergence of A:

∇ · A = 3x2 + x2 + y2 = 4x2 + y2(13.37)

and the answer is a single function. We now substitute for (x, y, z)

at the given point; i.e. we let (x, y, z) = (1, 0, 0) in ∇· A and so

∇ · A = 4 × 12 + 02 = 4. (13.38)

Exercise 13.9 Find the divergence of the following vector fields:

(i) A = (yz, xz, xy),
(ii) B = (x2 + y2 + z2)n/2(x, y, z) (i.e. B = rnr),
(iii) F = ∇ φ for some function φ(x, y, z).

We can think of divergence as representing the “spreading

out” of a vector field. Negative divergence at a point means the
vector field converges at that point. If V(x, y, z) is a vector field
representing the velocity of a fluid, then its divergence at any
point ∇· V is the outward flow of fluid (per unit volume) at that
point (evaluated over an infinitesimal volume around the point).
Negative divergence means there is a net inflow (convergence
at the point), and zero divergence means the inflow and outflow
balance.
470  •  Mathematical Physics

13.7.  CURL OF A VECTOR FIELD

Another vector operator is curl. The curl of a vector field is

another vector field. The curl of a vector field A, which is read and
sometimes written “curl A,” is

 ∂A ∂Ay ∂Ax ∂Az ∂Ay ∂Ax 

∇=
×A  z − , − , −  (13.39)
 ∂y ∂z ∂z ∂x ∂x ∂y  

or equivalently

 ∂A ∂Ay   ∂A ∂A   ∂Ay ∂Ax 

∇=
× A i z − + j x − z  + k  − .
 ∂y ∂z   ∂z ∂x   ∂x ∂y 

(13.40)

Look at the definition carefully. The x component of the curl

of A involves derivatives of the z and y components of A with
respect to y and z, respectively, and so on cyclically. The result
is a vector field (three components, each in general a function of
x, y and z).
There is a reason we write curl using the nabla and vector (cross)
product symbols. Let’s write ∇ and A as vectors and work out the
components of their cross product

 ∂ ∂ ∂ 
∇= (
× A  i + j + k  × iAx + jAy + kAz )
 ∂x ∂y ∂z 
 ∂A ∂Ay   ∂Ax ∂Az   ∂Ay ∂Ax 
= i z −  + j −  + k − .
 ∂y ∂z   ∂z ∂x   ∂x ∂y 

(13.41)
using the usual rules for vector (cross) products (see Section 4.4).
This is the same as equation (13.40).
 Introduction to Vector Calculus  •  471

Example 13.7 Let A = (3xz, 2yz, −z2); find ∇ × A.

First, identify the components of A:
Ax = 3xz, Ay = 2yz, Az = −z2.(13.42)

We now work out each of the components of ∇ × A. For the x

component:
∂A ∂A
( ∇ × A ) x = z − y = 0 − 2 y = −2 y; (13.43)
∂y ∂z 
while for the y component:

∂Ax ∂Az
( ∇ × A )=y − = 3 x − 0= 3 x; (13.44)
∂z ∂x 
and finally for the z component:

∂Ay ∂Ax
(∇ × A )z = − = 0 − 0= 0. (13.45)
∂x ∂y 
Thus we have that ∇ × A = (−2y, 3x, 0). It is a vector field
so has three components each of which is a function of x, y
and z.

Exercise 13.10 Find the curls of the following vector fields. In

each case begin by identifying explicitly the components of the
vector for which the curl is required.
(i) (1+ x2, xy, y(z + 1)),
(ii) (a · r)a where r = (x, y, z) and a = (ax, ay, az) is a constant
vector,
(iii) w × r where r = (x, y, z) and w = (wx, wy, wz) is a constant
vector.
472  •  Mathematical Physics

4.0

4
2
3.0
y

y
0
2.0

–2
1.0

–4
1.0 2.0 3.0 4.0 –4 –2 0 2 4
x x

(a) (b)
FIGURE 13.8:  Two examples of vector fields plotted with the z-axis suppressed. (a) The vec-
tors always lie along the x-axis with y-and z-components zero. The curl is therefore zero every-
where. If we placed a small paddle wheel in the field it would not be made to rotate. (b) The
vectors change direction as we move around the space; if we placed a paddle wheel anywhere
in the field it would begin to rotate clockwise. For example, at the point (2, 0) there is a stron-
ger “current” on the right than on the left side. This means there is a non-zero curl; at any
point the curl vector here points into the page in agreement with the “right hand screw rule”
(see last paragraph in Section 13.7).

Example 13.8 Let A = (3xz, 2yz, −z2) as in example 13.7.

Verify that ∇ · ∇ × A) = 0.
We worked out ∇ × A in example 13.7 so we just quote it here.
We will call the resulting vector field B:

B = ∇ × A = (−2y, 3x, 0). (13.46)

As B is a vector field we can take its divergence,

∂Bx ∂By ∂Bz

∇ ⋅ (∇ × A ) = ∇ ⋅ B = + + = 0 + 0 + 0 = 0. (13.47)
∂x ∂y ∂z 

Exercise 13.11 Show that ∇ · (∇ × A) = 0 for any vector field A.

 Introduction to Vector Calculus  •  473

In fact, although we shall not show it here, the converse also

holds: if B is a vector field such that ∇ · B = 0, then it is always pos-
sible to find a vector field A such that B = ∇ × A. This is an impor-
tant result for theoretical physics.

Exercise 13.12 Show that ∇ × ∇f = 0 for any scalar field f.

A vector field with zero curl is said to be irrotational or curl

free. The result of Exercise 13.12 shows that any vector field that
is the gradient of a scalar field is irrotational. In fact, although we
shall not show it here, the converse of this also holds: if E is a vec-
tor field such that ∇ × E = 0, then it is always possible to find a
scalar field f such that E = ∇f. This is another important result for
theoretical physics.
We will return to the geometrical interpretation of curl a little
later. But for now we can think of the curl of a vector field at a
point as representing the amount of “circulation” of the field about
that point. If V(x, y, z) is a vector field representing the velocity of
a fluid, then its curl ∇ × V at any point is a measure of the angular
velocity of the fluid about that point. Imagine holding a tiny paddle
wheel at a point in the fluid (Figure 13.8). If the curl is zero, the
paddle wheel would not rotate at that point. If the paddle wheel
rotates, the curl there is non-zero. The direction of the curl follows
the “right hand screw rule”: if you curl the fingers of your right
hand in the direction of the circulation, your thumb will point in
the direction of positive curl.

13.8. FLUX OF A VECTOR FIELD THROUGH A

SURFACE

In this and the next section, we will introduce two important

quantities associated with vector fields: flux and circulation. Here
we will try to emphasize what these mean, and also show how they
can be calculated for simple plane geometries.
474  •  Mathematical Physics

y
x

FIGURE 13.9:  The flux of a field through a plane surface.

Flux1 is the amount of “something” passing through a surface.

If the vector field is the velocity of a fluid, then flux is the flow rate
through the surface. We can also consider heat flux (the rate of heat
energy transfer through a surface), and electric flux, which we can
think of as the number of electric field lines passing through a sur-
face (although notice that nothing is actually flowing in this case).
The flux depends on the vector field and the size, shape and
location of the surface. If the vector field is stronger (faster moving
fluid, stronger electric field etc.) then the flux will be higher; if the
surface is made larger or oriented better to “catch” more of the field,
then the flux will be higher. Imagine catching fish with a fishing net:
a larger area fishing net will catch more fish, and moving the fishing
net to a region with a higher density of fish will catch more fish.
First, a simple example. Let B = (b(x), 0, 0) be a vector field in
the x-direction, with b(x) depending only on x. Let S be a rectangular
surface of area A at x = x0, aligned normal to the positive x-axis, as in
Figure 13.9. The flux of the vector field B through the surface S is

(flux of B through X) = Area × field strength = Ab(x0).

 (13.48)
1
The word flux is from the Latin fluere meaning “to flow” and was intro-
duced into calculus by Newton’s use of the word fluxion to mean derivative.
 Introduction to Vector Calculus  •  475

More generally, if

B(x, y, z) = (Bx, By, Bz)  (13.49)

where each of Bx, By and Bz depends on position, then the flux of B

through X is defined as an integral over the surface X,

( flux of B through X ) = ∫∫ Bx ( x0 , y, z ) dydz. (13.50)

X

The flux of a vector field through a surface is defined as the inte-

gral of the normal component of the field over the surface. Later we
shall find an alternative way to calculate the flux through a closed
surface.

Exercise 13.13 Write down the integral analogous to equation

(13.50) defining the flux of the vector field B(x, y, z) through the
rectangle Y normal to the y-axis, cutting the y-axis at y = y0.

Note that to calculate the flux through a surface we need to

be given the direction of the positive normal to the surface. In the
examples below, we take this along the positive coordinate axes.

y 1
x

FIGURE 13.10:  The surface X.

476  •  Mathematical Physics

Example 13.9 Calculate the flux of the vector field

B = ((1 + x)−1, (y + x), 0) through the surface X, in the (y, z)
plane bounded by the lines y = 0, y = 1, z = 0, z = 1.
When tackling problems like this, it is often helpful to draw the
region of integration. This is done in Figure 13.10. We start
from the definition. The component of B normal to the surface
X is the component Bx. The element of area in X is dydz, so

flux = ∫∫ Bx dydz. (13.51)

X 
For a general value of z in X, y runs between 0 and 1. Within X, z
runs between 0 and 1. So these are the limits of integration:
1 1
flux = ∫ ∫ Bx dydz. (13.52)
0 0 
Next substitute for Bx. On the surface X we have x = 0, so
B = (1, y, 0) on X. Thus Bx = 1. This leaves a simple double integral:
1 1 1 1
[=
z ]0
1
∫∫ ∫ ∫
1
=
flux =
1 dydz y  =
dz 1=
dz 1. (13.53)
0 0 0  0 0 
The field has 1 unit of flux through X.

Example 13.10 Using the same vector field as in Example

13.9, calculate the flux through the surface Y in the (x, z) plane
bounded by the lines x = 1, x = 2, z = 0, z = 1.
The component of B normal to Y is By. The element of area in
Y is dxdz. Thus we have that
flux = ∫∫ By dxdz. (13.54)
Y 
To cover Y, x runs between 1 and 2 and z between 0 and 1, so
1 2
flux = ∫ ∫ By dxdz. (13.55)
0 1 
 Introduction to Vector Calculus  •  477

On the surface Y (in the (x, z) plane) y = 0, so B = (1/(1 + x), x, 0)

and By = x. This leaves a standard double integral:
2
1  x2 
2 1 1 1
flux ∫ ∫ x=
=
0 1
dxdz ∫  =
0  dz
 2 1
∫ 2 ( 4 − 1) dz
0

13 3
= ∫= dz . (13.56)
0 2 2


The field has 3/2 units of flux through Y.

Example 13.11 Calculate the flux of the vector field

B = (ye−x, 1, 0) through the surface X in the (y, z) plane bounded
by the lines y = 0, y = 1, z = 0, z = 1.
The component of B normal to the surface X is Bx, and so

flux = ∫∫ Bx dydz. (13.57)

Y 

The surface of integration is the square 0 ≤ y ≤ 1, 0 ≤ z ≤ 1 so

these are the limits of integration. The surface lies at x = 0,
in the (y, z) plane, and so Bx = ye0 = y. This gives a standard
double integral:
1
1 1 1  y2  1 1 1
=
flux ∫∫
0 0
y=
dydz ∫0  2=  dz
0
∫=
02
dz . (13.58)
2



Exercise 13.14 Calculate the flux of the vector field (x3 + 1, yz, 2)
through
(i) the surface X1, in the y-z plane bounded by the lines y = 0,
y = 1, z = 0, z = 1,
(ii) the surface X2, in the plane x = 1 bounded by the lines y = 0,
y = 1, z = 0, z = 1.
(iii) Hence obtain the difference in the fluxes through X2 and X1.
478  •  Mathematical Physics

13.9.  CIRCULATION OF A VECTOR FIELD

The circulation of a vector field around a loop (closed curve)

is defined as the integral of the tangential component of the vec-
tor field along the loop. If the vector field represents the force on
a body, then the circulation represents the work done by the force
in moving the body around the loop. There are many other physical
applications in electro-magnetism and fluid mechanics.
Let’s see an example. We have a vector field V = (u(y), v(x), 0)
where, for the moment, u depends on y only and v depends on x
only. Let PQRS be a rectangle in the (x, y) plane; sides PS and QR
have length (x2 − x1) and PQ and SR have length (y2 − y1) (Figure
13.11). We are going to define the circulation of V around the rect-
angle PQRS in the counterclockwise direction.2

v V
y
u(y2) u
y2
Q R

P S
y1
u(y1) x
x1 x2

FIGURE 13.11:  The rectangle PQRS. Above is shown the vector V at a point, and its compo-
nents in the x-and y-directions.

We define the contribution from the side RQ to the circulation

of the vector field V around the rectangle to be

(length of RQ) × (tangential component of V along RQ )

= (x2 − x1) × (−u(y2)),(13.59)

2
The direction is chosen to follow the right hand screw convention with
respect to the (x, y, z) coordinate system.
 Introduction to Vector Calculus  •  479

where the minus sign for u arises because the tangential component
of V is positive in the direction QR, opposite to RQ. Similarly, for the
contribution from PS,

(length of PS) × (tangential component of V along PS )
= (x2 − x1) × u(y1).

The net contribution of the pair of sides QR and SP is

(x2 − x1)(u(y1) − u(y2)).(13.60)

Similarly, from the pair of sides QP and SR we get a net

contribution
(y2 − y1)(−v(x1) + v(x2)).

The circulation of V around PQRS (i.e. around the sides in the

order RQ, QP, PS, SR) is the sum of these net contributions:

circulation = (x2 − x1)(u(y1) − u(y2)) + (y2 − y1)(v(x2) − v(x1))

= (x2 − x1)u(y1) + (y2 − y1)v(x2) − (x2 − x1)u(y2) − (y2 − y1)v(x1).
(13.61)

To extend this to more general vector fields with tangential com-

ponents that vary along an arc we replace the products (length) ×
(tangential component) by an integral.
For a vector field V = (Vx,Vy, VZ) we define the contribution to
the circulation from RQ to be

Vx ( x, y2 , 0 ) dx = − ∫ Vx ( x, y2 , 0 ) dx
x1 x2

∫x2 x1

(13.62)

(replacing Equation 13.59). We have put z = 0 in Vx because we have

assumed above that PQRS lies in the z = 0 plane. The contribution
to the circulation from PS is

Vx ( x, y2 ,0 ) dx.
x2

∫ x1

(13.63)
480  •  Mathematical Physics

The net contribution from PQ and RS is

∫ ( − V ( x, y , 0 ) + V ( x, y , 0 ) ) dx 
x2
x 2 x 1 (13.64)
x1

(replacing Equation 13.60). The total circulation of V around PQRS

is therefore

∫ ( − Vx ( x, y2 ,0 ) + Vx ( x, y1 ,0 ) ) dx
x2
circulation =
x1

( V ( x , y,0 ) − V ( x , y,0 ) ) dx.

y2
+∫ y 2 y 1
y1

(13.65)

Example 13.12 Calculate the circulation of the vector field

V = (1+x2, xy, y(z + 1)) around the unit square in the first
quadrant of the (x, y) plane having one vertex at the origin
(Figure 13.12).
The circuit PQRS is in the (x, y) plane, meaning z = 0. The simplest
way to get the signs right is to let the limits take care of them.
On RQ the tangential component of V is Vx = (1 + x2), and x
runs between 1 (lower limit) and 0 (upper limit). So, the
contribution to the circulation from RQ is

∫ (1 + x ) dx =
− ∫ (1 + x2 ) dx.
0 1
circulation from RQ = 2
1 0

1
Q R

P S x
1
FIGURE 13.12:  The square PQRS (see Example 13.12).
 Introduction to Vector Calculus  •  481

On QP the tangential component of V is Vy = xy = 0 since x = 0, so

0
=
circulation from QP ∫=
0.y dy
1
0.

On PS the tangential component of V is Vx = (1 + x2), so

∫ (1 + x ) dx. 
1
circulation from=
PS 2
(13.66)
0

On SR the tangential component of V is Vy = xy = y since x = 1

on SR; y runs between 0 and 1 and so

1
circulation from SR = ∫ y dy.
0

The net circulation from PQRS is the sum of these contributions

around the loop,

− ∫ (1 + x2 ) dx + ∫ y dy + ∫ (1 + x2 ) dx.
1 1 1
circulation PQRS =
0 0 0

The contributions from QR and SP cancel since Vx is not a

function of y, and so

1
1  y2  1
=
circulation PQRS ∫0
= y dy = 
 2 0 2
.

Exercise 13.15 Find the (counterclockwise) circulation of the

vector field
V(x, y, z) = (1 + x2, xy, y(z + 1))

around the unit square in the (x, y) plane, centered on the origin
with sides parallel to the axes.
482  •  Mathematical Physics

13.10. DIVERGENCE THEOREM
(GAUSS’S THEOREM)

We have defined two physically important quantities associated

with vector fields–the divergence and the flux through a surface.
For a special type of surface, namely a closed surface, these two are
related. The relation between them is called the divergence theo-
rem, also known as Gauss’s theorem. The aim of this section is to
derive this theorem for some special cases.
We start by considering a vector field B and a closed surface S
enclosing a volume V. The divergence theorem says the flux of B
(outwards) through the closed surface is equal to the integral of the
net divergence inside the volume

∆flux= ∫ ∇ ⋅ B dV. (13.67)

V 

Example 13.13 Verify the divergence theorem for a simplified

case in which B = (b(x), 0, 0) is a vector field in the x-direction,
with b(x) depending only on x, and V is a cuboid with faces
X1 and X2 of areas A perpendicular to B.
The normal component of B on X2 is Bx = b(x2). The flux of B
through X2 is

b ( x ) dydz b=
∫∫= 2 ( x ) ∫∫ dydz
2 Ab ( x2 ) . (13.68)
X2
X2


since Bx is constant on X2. Similarly the flux of B through X1 into

V is (area)×(normal component) since the normal component
Bx is constant on X1, thus

flux through X1 = Ab(x1). (13.69)

The net flux out of the volume is the sum of the net flux out of
each face. As the vector field is along the x-direction, crossing
only the surfaces X1 and X2, these are the only faces with
 Introduction to Vector Calculus  •  483

non-zero flux. Therefore the net flux is the difference of the

flux out of X2 and the flux into X1, i.e.
Δflux = A(b(x2) − b(x1)).(13.70)
We want to relate the expression for Δflux to ∇· B, so we bring
in derivatives of b(x):
x2 ∂b
∆flux= A ∫ dx. (13.71)
x1 ∂x 
However, since
∂Bx ∂By ∂Bz
∆ ⋅ B= + + (13.72)
∂x ∂y ∂z 

where Bx = b(x), By = 0 and Bz = 0, so

∂b
∆ ⋅ B= . (13.73)
∂x 
We can write
z2 y2
A=∫ ∫ dydz. (13.74)
z1 y1

Then

=∆flux (∫ z2

z1 ∫
y1
y2
dydz )(∫
x1
x2
∇ ⋅ B dx , ) 
(13.75)

or
z2 y2 x2

∆flux= ∫ ∫ ∫
z1 y1 x1
∇ ⋅ B dxdydz= ∫
V
∇ ⋅ B dV.

(13.76)

This example shows us the physical meaning of the divergence

of a vector field. The divergence of a field gives the rate per unit
volume at which flux is being gained or lost from a point. If we
think of a vector field as a fluid flowing through space (the arrows
representing the velocity at each point) then we gain or lose flux
where there are sources or sinks of the fluid. At a point where
∇ · B = 0 there are no sources or sinks. And if there are no sources
or sinks inside a volume, then ∇ · B = 0 everywhere inside, and the
484  •  Mathematical Physics

divergence theorem tells us that the net flux through the surround-
ing surface is also zero.
More generally, if B = (b(x, y, z), 0, 0) the definition of the flux
of B through the rectangle X2 gives

∫ b ( x , y, z ) dydz
z2 y2
flux of B through the plane surface X 2 = ∫ 2
z1 y1

(13.77)
and similarly for X1. Now the difference in flux between X2 and X1 is

 b ( x2 , y, z ) − b ( x1 , y, z )  dydz.
z2 y2

=∆flux ∫ ∫
z1 y1

(13.78)

FIGURE 13.13:  Flux through a closed surface.

Example 13.14 Verify the divergence theorem for the case in

which B = (Bx, By, Bz) is a general vector field (with components
in all three directions each of which is a function of x, y and z),
and V is a cuboid with faces X1, X2, Y1, Y2, Z1, Z2 perpendicular
to the x-, y- and z-axes.
We calculate the flux through each of the six faces of the
rectangular parallelepiped (cuboid) in Figure 13.13. For each
pair of faces we write down an equation like equation (13.78).
So, for faces X1 and X2,
 Introduction to Vector Calculus  •  485

 Bx ( x2 , y, z ) − Bx ( x1 , y, z )  dydz
z2 y2
( ∆flux ) x
= ∫ ∫
z1 y1
(13.79)
z2 y2  x2 ∂Bx 
=∫ ∫  ∫x1 dx  dydz.
   
z1 y1
 ∂x  

Similarly
z2 x2 y2 ∂By
( ∆flux ) y =
∫z ∫x ∫y dydxdz. (13.80)
1 1 1 ∂y
and

y2 x2 z2 ∂Bz
( ∆flux ) z =
∫y ∫x ∫z dzdxdy. (13.81)
1 1 1 ∂z

The flux out of the volume V is the sum of the net flux through
each pair of faces:

( flux out of V ) = ( ∆flux ) x + ( ∆flux )y + ( ∆flux )z ,

(13.82)
z2 x2 y2  ∂Bx ∂By ∂Bz 
= ∫ ∫ ∫
z1 x1 y1

 ∂x
+
∂y
+  dxdydz
∂z 

= ∫ V
∇⋅ B dV ,

(13.83)

where we change the order of integration in equations (13.80)

and (13.81).

We have therefore obtained the divergence theorem for a gen-

eral vector field, but only for a cuboidal volume. In fact, although
we shall not prove it, the theorem holds for general vector fields and
general volumes. It is usual to write the flux through a surface S in
a notation that indicates its origin as an integral of the vector field
over an area.
486  •  Mathematical Physics

In this notation the divergence theorem becomes

∫ B ⋅ dS= ∫
S V
∇ ⋅ B dV , (18.84)

where S is a surface enclosing the volume V.

The divergence theorem therefore tells us that we can calculate
the flux of a vector field through a closed surface if we know what it
is throughout the whole interior volume. It relates a surface (two-
dimensional) integral to a volume (three-dimensional) integral. It says
that if we know the flux of a field through a closed surface, which can be
obtained from the value of the field only on the surface, we can deduce
the net sources and sinks of the field within the enclosed volume.
Exercise 13.16 Evaluate the flux of the vector field F = (x 3, x 2y,
y 2(z + 1)) through the closed surface of the unit cube bounded by
the six planes x = 0, x = 1, y = 0, y = 1, z = 0 and z= 1
(i) directly by evaluating the six surface integrals,
(ii) by integrating (∇∙F) over the volume of the cube. Hence
verify the divergence theorem in this case.

Exercise 13.17 Use the divergence theorem to calculate the flux

of the vector field
F(x, y, z) = (xz, y3, xy)

through the closed surface of the unit cube bounded by the six
planes x = −1, x = 0, y = −1, y = 0, z = −1 and z = 0.

13.11.  STOKES’S THEOREM

We have defined–two other physically important quantities asso-

ciated with vector fields the curl of the field and its circulation around
a closed loop. These two are also related to each other. The relation
between them is called Stokes’s theorem. The aim of this section is to
derive this theorem for some special cases.
 Introduction to Vector Calculus  •  487

Example 13.15 Let S be a closed rectangle in the (x, y) plane

with sides parallel to the axes and perimeter C. Show that the
circulation of a vector field V around C is given by the flux of
∇ × V through S.
The circulation around C ≡ PSRQ is found from the sum of the
contributions from the four sides,
Q P S R
PQRS = ∫ R
V x dx + ∫ V y dy + ∫ V x dx + ∫ V y dy
Q P S

∫ (V ( x , y , 0 ) − V x ( x , y 1 , 0 ) ) dx
x2
= x 2
x1

(V ( x )
, y , 0 ) − V y ( x 1 , y , 0 ) dy .
y2
+∫ y 2
y1

Now
y1 ∂V
Vx ( x, y2 ,0 ) − Vx ( x, y1 ,0 ) =
∫y2 ∂y ( x, y,0 ) dy
x

and
x2 ∂Vy
Vy ( x2 , y,0 ) − Vy ( x1 , y,0 ) =
∫x1 ∂x ( x, y,0 ) dx
and so
x2 y2 ∂V y2 x2 ∂Vy
−∫ ∫
circulation = x
dydx + ∫ ∫ dxdy
x1 y1 ∂y y1 x1 ∂x

x2 y2  ∂Vy ∂Vx 
= ∫ ∫
x1 y1

 ∂x
−
∂y
 dxdy.


Making use of the fact that the integrand is the same as the
z−component of the curl ∇ × V (see Equation 13.39) we have
that

Circulation of V round C= ∫ (∇ × V )
S z
dA. (13.85)

The integration in the final expression is over the area of the

plane rectangle S.
488  •  Mathematical Physics

The equality (13.85) is known as Stokes’s theorem (for a plane

area). Stokes’s theorem enables us to express the circulation of a vec-
tor field around a closed curve as an integral of the curl of the vector
field over the area enclosed by the curve.

∫
C
V ⋅ d=
l ∫ ( ∇ × V ) ⋅ dS. 
S
(13.86)

The physical meaning of the curl of a vector field is therefore

that it gives the rate per unit area at which the circulation changes
(roughly speaking, the ‘‘vorticity’’ of the vector field). We have not
here proved the result for a general vector field, only for a plane
rectangle. But in fact, the theorem holds for most open surfaces3 one
encounters in physics, even surfaces with holes. It is usual to write
the theorem in a way that emphasizes this:

circulation= ∫ (∇ × V )
S z
dS.

(13.87)

On the right side we have the flux (of ∇ × V) through a gen-

eral surface, which involves the integral of the normal component of
∇ × V through a general surface S that has unit normal n. On the left
side the circulation is written as the integral of the tangential com-
ponent (the scalar product of V with a element of length dl tangent
to the curve C) around the closed curve C that is the boundary of S.
The circle on the integral symbol is the conventional way of indicat-
ing that the curve C must be closed (otherwise the equation is false).

Example 13.16 By using Stokes’s theorem, compute the

circulation of the vector field V = (1 + x2, xy, y(z + 1)) around
the unit square S in the first quadrant of the (x, y) plane having
one vertex at the origin.
By Stokes’s theorem we can work out the circulation as an
integral of the curl of A over S:

3
Provided they are orientable–roughly speaking, that they have two sides
and so one can define a consistent choice of normal vectors to the surface
at every point, unlike a Möbius strip.
 Introduction to Vector Calculus  •  489

circulation= ∫ (∇ × V )
S z
dS.

We have already worked out ∇ × V in Exercise 13.10:

∇ × V = (z + 1, 0, y).

The integrand is the z–component of ∇ × V on the surface S:

(∇ × V)z = y.
So we have
1 1 1 1
circulation=∫
0 ∫ (∇ × V )
0 z
dxdy = ∫
0 ∫
0
y dxdy,

and so
1 1 1 1
circulation= ∫ dx ∫ y dy =1 × = .
0 0 2 2

This agrees with the direct calculation of the circulation in

Exercise 13.12.

Exercise 13.18 Use Stokes’s theorem to show that the circula-

tion of the vector field

V(x, y, z) = (1 + x2, xy, y(z + 1))

around the unit square in the (x, y) plane, centered on the origin,
is zero.
Exercise 13.19
(i) The steady drift of a straight section of river can be described
by the velocity field
v = νi.  (13.88)
A lobster pot consisting of a cube of side 1m (designed by a
physicist!) is immersed in the water with all its sides parallel
or perpendicular to the flow. Write down the flux of v into
490  •  Mathematical Physics

each of the sixsides and hence the net flux into the cube.
Verify this result using Gauss’s divergence theorem.
(ii) The pot is now moved to a section of the river approaching
a weir where the water is speeding up with a constant accel-
eration a, so the velocity field becomes

v = (ν0 + 2ax)1/2i,  (13.89)

where υ0 is the velocity at x = 0. If the pot is placed in the

same orientation as before with its center at x = 10.5m, write
down the flux into each face and the net flux into the cube.
Again verify this result using Gauss’s theorem.
(iii) What is the circulation of v around the top face of the pot in
the two locations? Verify these results using Stokes’s theorem.
(iv) The pot is placed near an outcrop where the flow is dis-
turbed and the velocity field is given by

v = (ν + t 1 y)i + t 2 yj,  (13.90)

where y = 0 corresponds to the riverbank. If the pot is placed

in the same orientation as before with its center 1.5m away
from the riverbank, work out the circulation of v around the
top face. Verify this result using Stokes’s theorem.

13.12.  EXTENSIONS: INDEX NOTATION

We can write compact expressions for the vector operators using

the index notation introduced in Section 4.15. Thus

∂f
( ∇f ) i =
∂xi

gives the components of the gradient of the function ø(x 1, x 2, x 3).

The divergence and curl of a vector field A(x 1, x 2, x 3) are given by
 Introduction to Vector Calculus  •  491

∂Ai
∇⋅A =
∂xi
∂Ak
( ∇ × A ) i =Œijk
∂x j

using the notation of Section 4.15. Once one gets the hang of this nota-
tion the derivation of vector identities becomes a mechanical process.

13.13.  EXTENSIONS: OTHER COORDINATES

In Cartesian coordinates, the distances between neighboring

points along the coordinate axes are dx, dy and dz along the x-, y-
and z-axes, respectively. In other coordinate systems the distances
between points are more complicated. For example, in spherical polar
coordinates the physical distance on the sphere of radius r between
points at the same latitude q separated by a coordinate increment dø
is not dø (which is not a distance) but r sin(q)dø. This leads to more
complicated expressions for the gradient, divergence and curl of a
vector field in non−Cartesian coordinates.
In a general orthogonal coordinate system (ξ1, ξ2, ξ3) we let the
distances between neighboring points be ds 1 = h 1dξ1, ds 2 = h 2dξ2
and ds 3 = h 3dξ3. This defines (h 1, h 2, h 3). (So, for example, in spher-
ical polars h 3 = r sin(q).) Then

 1 ∂y 1 ∂y 1 ∂y 
∇y = , , ,
 h1 ∂x1 h 2 ∂x2 h 3 ∂x3 
1 ∂ 1 ∂ 1 ∂
=∇⋅A ( h2 h 3 A1 ) + ( h1 h 3 A 2 ) + ( h1 h 2 A 3 ) ,
h1 ∂x1 h2 ∂x2 h 3 ∂x3
 1  ∂ ( h 3 A 3 ) ∂ ( h 2 A2 )  1  ∂ ( h1 A1 ) ∂ ( h 3 A 3 ) 
=∇×A   − ,  − ,

 h 2 h 3  ∂x2 ∂x3  h1 h 3  ∂x3 ∂x1 
1  ∂ ( h 2 A2 ) ∂ ( h1 A1 )  
 − .
h1 h 2  ∂x1 ∂x2  
492  •  Mathematical Physics

In particular, in cylindrical coordinates (r, q, z) we have (h1, h2, h3)

= (1, r, 1) and in spherical polars (r, q, ø), (h1, h2, h3) = (1, r, r sin(q)).

Revision Notes

After completing this chapter you should be able to

• Understand the difference between scalar fields
f(x, y, z) and vector fields A(x, y, z) = (A x, A y, A z)
• Calculate the gradient

∇f = (∂f/∂x, ∂f/∂y, ∂f/∂z)

of a scalar field f(x, y, z) and use it to find the rate of

change, u · ∇f, of a field f in a given direction u, and the
unit normal, n = ±∇f/ |∇f |, to the surface f = constant
• Calculate the divergence of A, a scalar, given by

∂Ax ∂Ay ∂Az

∇=
⋅A + +
∂x ∂y ∂z

• Calculate the curl of A, a vector field,

 ∂A ∂Ay ∂Ax ∂Az ∂Ay ∂Ax 

∇× A z − , − , − 
 ∂z ∂y ∂x ∂z ∂y ∂x 

• Calculate, for a closed rectangular box S, the flux of A out

of S
∫ A ⋅ dS = sum of fluxes of A out of each face
• State the divergence theorem (Gauss’s theorem):

∫ A ⋅ dS= ∫
S V
∇ ⋅ A dV
 Introduction to Vector Calculus  •  493

where V is the volume interior to S, and explain

the meaning of the terms with reference to simple
examples
• Compute, for a closed rectangular curve C, the circulation
of A around C:

∫C
A ⋅ d1 = sum of integrals along each edge of C.

• State Stokes’s theorem:

∫
C
A ⋅ d=
l ∫ ( ∇ × A ) ⋅ dS
S

= flux of ∇ × A through interior of the closed curve C

and explain the meaning of the terms with reference to simple

examples.

13.14. EXERCISES

1. F
 ind the gradient of the function f (x, y, z) = x2 + 2xyz + z2
in the direction parallel to the vector (1, −1, 2) at the point
(0, 2, 3).
2. F
 ind expressions for the divergence and the curl of the vec-
tor function

A(x, y, x) = zexyi + xexyj + yexyk.

3. C
 alculate the flux of the vector field E = (y3, x2y, y2z)
through a rectangular surface with corners at (0, 0, 0),
(0, 0, −2), (0, 1, 0) and (0, 1, −2).
494  •  Mathematical Physics

4. Find the divergence of the vector field

z2 y
A=x2 i + 2 xyj − k
x

at the point (1, 2, 3).

5. A curve C in the form of a closed unit square spans a plane
area S which has a unit normal n̂ (so nˆ dS = nˆ dS ). Evalu-
ate (a) ∫
C
dr , (b) ∫∫ dS , (c) ∫∫ dS and (d) ∫
S S
C
dr .

6. Show that curl(xa) = i × a where a is a constant vector.

7. Calculate the divergence of the vector field 3r2r at the
point r = (0, 2,1).
8. Find the divergence of the vector field (yz, xy2z, xyz2).
9. Find the curl of the vector field (x, xz, xy).
10. Show that ∇(| r |2) n = 2n(| r |2)n−1r where r = xi + yj + zk.
11. If A = (x2 − y2, y2 − z2, z2 − x2) and Q = x + y + z, find
∇ · (QA) (the divergence of QA) at (1, 0, 1).
12. S
 how that the curl of k sin(k · r) is identically zero if k is a
constant vector.

13.15. PROBLEMS

1. State Stokes’s theorem and use it to show that

∫ ( A × r ) ⋅ dr =
C
2p A,

where C is a unit circle spanned by a surface normal to the

constant vector A.
 Introduction to Vector Calculus  •  495

Use the above result to interpret the geometrical meaning

of the integral
1
2 ∫ C
r × dr.

2. Calculate the integral over the surface of a cube whose

corners are located at (0, 0, 0), (0, 2, 0), (0, 0, 2), (2, 0, 0),
(0, 2, 2), (2, 0, 2),(2, 2, 0) and (2, 2, 2) of the vector field

B = (x2 + y2, y2 + z2, z2 + x2).

Calculate the divergence of B and verify that its integral

over the volume enclosed by the cube above has the same
value as the flux through the surface.
3. State Gauss’s theorem and use it to evaluate

∫ div ( r 2 r ) dV ,
V

where r = (x, y, z) and the integral is taken over a unit cube

with its center at the origin and sides parallel to the axes.
4. State Gauss’s theorem and verify it for the vector field
(xn, yn, zn) for a unit cube whose surface normals are paral-
lel to the x-, y- or z- axes and with its center at the origin.
(Assume n is an odd positive integer.)
5. Prove that the volume of a solid is given by

1
3∫
r ⋅ dS,

where r is the position vector of any point on its surface and

dS is the vector element of area of the surface. Show that
the expression is correct for a cube of side a, with its center
at the origin and sides parallel to the axes.
 APPENDIX

A
PREREQUISITES

This appendix contains some basic facts and methods that will be
review for most students, as well as some useful reference material.
Consult it as necessary.

A.1.  NUMBERS, FUNCTIONS, AND PROOFS

Real Numbers
Real numbers are values that represent a quantity along a line.
Imagine each real number as a point on a line with the negatives to
the left of zero and the positives to the right. There are two types of
points: rational numbers that can be represented as fractions or as
either finite or recurring decimals and irrational numbers that are
not fractions and are represented by an unending and non-repeating
decimal expansion.
There is one point about the decimal expansion of real numbers
that you may need reminding of: the recurring expansion 0.999999
(the bar shows that the 9 repeats forever) is the same as 1.00000,
and similarly for other numbers with recurring 9s. (You can see why,
once you try adding something to 0.999999, such as 0.00001 or
0.0000001. There is nothing you can add that does not give a num-
ber that exceeds 1.)
498  •  Mathematical Physics

Functions
A function of a real variable is a rule that assigns a real num-
ber to each permitted value of the variable. For example the square
function x2 assigns the square of x for any real number x. We often
write y = f (x) for a function of x. Here x is the independent variable
and y the dependent variable.

Proofs
You need to be quite clear about the notion of proof in mathemat-
ics. A finite number of verifiable instances of a general result does not
constitute a proof. Take the following example, the expansion of
(1 + x)n = 1 + nx + ... ,
where we want to prove that the coefficient of x in the expansion is
n. It is obviously true for n = 1. For n = 2 we can get the result by
explicitly multiplying out (1 + x)(1 + x) = 1 + 2x + x2. You might see
how to go on to verify the result for n = 3. But we cannot prove that
the result is true in general by enumerating instances, because we
might eventually come across a value of n for which the result does
not hold. For example, take the (incorrect) statement that every
even number, 2n, is followed by a prime number. This is true for
n = 1, 2 and 3 where, 2n = 2,4 and 6, which are followed by 3, 5, and
7, but it is clearly not true for n = 4 since 2n = 8 is followed by 9,
which is not prime.
There are various valid methods of proof in mathematics. One
example is proof by mathematical induction.

Example A.1 Show that (1 + x)n = 1 + nx + ... .

We want to show that
(1 + x)n = 1 + nx + ...,
for n = 1,2,3,4, . . . This result is clearly true for n = 1. We shall
assume that it holds for n = 2, 3, . . . up to some general n, then
consider (1 + x)n+1.
(1+ x)n+1 = (1+ x)(1 + x)n,
= (1 + x)(1 + nx + ...).
 Prerequisites  •  499

Then, multiplying out the parentheticals and collecting like

terms we find
(1 + x)n+1 = (1 + nx + ... + x + nx2 + ...),
= 1 + (n + 1)x + ....
So if (1 + x)n = 1 + nx + ... is true for n we have shown that it is
true for n + 1. But we know that it is true for n = 1. So it must
be true also for n = 2, 3, . . . and so on for all n.
This method (assuming the result for n, deduce it for n + 1;
check it for n = 1) is called proof by induction.

A.2.  NECESSARY AND SUFFICIENT CONDITIONS

The distinction between what is necessary for a result to be true

and what is sufficient is sometimes important. It is best approached
through some examples:
• For x2 > 1 to be true it is sufficient that x > 1 but not neces-
sary (x could also be < −1).
• For x2 < 1 to be true it is necessary that x < 1, but it is not suf-
ficient (x must also be > −1).
• For x1/2 > 1 it is both necessary and sufficient for x > 1.

Suppose that x > 1; then it follows from the first of the above
statements that x2 > 1.
Suppose x < 1; then it does not follow that x2 < 1 (because according
to the second statement x < 1 is necessary but not sufficient). Another
way of thinking about this is in terms of the direction of implication:
• x > 1 implies x2 > 1 so x > 1 is a sufficient condition (‘‘suffi-
cient’’ to imply the result).
• x < 1 is implied by x2 < 1 so the result x < 1 is a necessary
consequence.
500  •  Mathematical Physics

A.3. LOGARITHMS

We define the function y = loga x (‘‘ the logarithm to the base a of x’’)
for x > 0 and a > 0, as the number, y, satisfying the relation x = ay..i.e. the
log of a positive number is the power to which the base must be raised
to get that number.

For x > 0, y = loga(x) if x = ay.

Example A.2 For example,

(i) log10 (10) = 1 (put a = 10, x = 10, then 10 = 101 so y = 1)
and similarly log a a = 1,
(ii) log10 (10n)= n,
(iii) log10 (0.1) = −1 (because 0.1 = 1/10 = 10−1).

To find values of log a(x) for more general values of x you will

need to use a calculator. Usually calculators have ‘‘log’’ meaning
log10, and ‘‘ln’’ meaning log e (see Chapter 2) but you can get from
these to any base (see below). In this section, for formulae that are
true in any base we will omit explicit reference to the base and just
write log x.

Example A.3 Show that log(x1x2) = log(x1)+log(x2), and deduce

that log(1/x) = − log(x).
Let log(x1) = y1, log(x2) = y2 and log(x1x2) = y. Then, using the
definition of log, we can write

x1 = a y1 , x2 = a y2 and x1 x2 = a y .
Then

x1 x2 = a y1 a y2 = a y1 + y2 = a y .

So, by examining the exponents we have that

y = y1 + y2.
 Prerequisites  •  501

Finally, substituting for y, y1 and y2 we obtain the required

result that

log(x1x2) = log(x1) + log(x2).

Now for the second part. We know a0 = 1 and taking logs of
both sides we get 0 = log(1). Therefore
log(x × (1/x)) = log(1) = 0.
Using the result that log(x1x2) = log(x1) + log(x2) we can then
therefore write that
log(x) + log(1/x) = 0
and hence,
log(1/x) = − log(x).
This is a useful result that is worth remembering.

The formula for the logarithm of a product is true in any base.

Obviously it follows from Example A.3 that log(x2) = 2 log(x) and
similarly
log(xa) = a log(x)
for any a.

Example A.4 Change of Base

log b x
Show that log a(x) = (log a b)(log b x) and that log a ( x ) = .
log b a
First, we let y = log b x and w = log a b, then
x = by,
b = aw.
Next, substituting for b = aw, we have
x = (aw)y,
= awy.
502  •  Mathematical Physics

Taking logs of both sides gives

log a x = wy.

Then, substituting for y and w we find the required result that

log a x = log a b × log b x.
For the second part, we use the fact that b = aw, so
a = b1/w.
Taking logb of both sides gives
logb a = 1/w,
= 1/ loga b.
Then, substituting this into the result for the first part we have
log b x
= =
log a x log a b log b x .
log b a

Exercise A.1 By using logs on your calculator find

3
(i) 5
(ii) 1.254 × 0.378
(iii) 1.254/0.378.
Exercise A.2
(i) If a plot of log(y) against log(x) is a straight line of slope k
show that y = Axk for some constant A.
(ii) Express log(xx) in terms of x and log(x).
(iii) Find log3(5.2).
A.4.  TRIGONOMETRIC FUNCTIONS

h o FIGURE A.1:  Right angle triangle; h is the length of the hypotenuse, a

the adjacent side and o the opposite side with respect to the angle q.
q
a
 Prerequisites  •  503

The elementary definition of the functions sin, cos, and tan gives
them in terms of the ratios of sides of a right-angled triangle,

o a o
sin( q )
= = , cos( q ) =
and tan( q ) .
h h a

From this definition various relations can be deduced. For example,

sin( q )
tan( q )
= and sin 2 ( =
q ) + cos2 ( q ) 1.
cos( q )

However, the elementary definition in terms of a simple right

angle triangle has limitations – and more convenient and technical
definitions exist. We will not go through these here, but simply sup-
ply some relevant and general results.
First, the extension to all angles (0° ≤ q ≤ 360°) is most easily
represented by showing the signs of the sides of the triangle in each
of the quadrants (see Figure A.2). These follow the obvious pattern
of coordinate axes. So the overall signs of each of the functions are
as in Figure A.3.

+ + FIGURE A.2:  The signs of the sides of the tri-

-
- + angles for construction of the trigonometric
- + functions follow the pattern of the coordinate
- axes.

For example, sin(135°) = sin(45°), sin(192°) = −sin(12°) and

sin(307°) = −sin(53°). Also, cos(135°) = −cos(45°), cos(192°) = −cos(12°)
and cos(307°) = cos(53°).
In general,
sin(−q) = −sin(q),
cos(−q) = cos(q).
 504  •  Mathematical Physics

Radians and Degrees

Note that in general mathematical relations are given with the
angles measured in radians, not in degrees. The relation is that 2p
radians is equivalent to 360°. Thus to convert an angle A in degrees
to the same

+ + - +
q q
- - - +
sin cos

- +
q
+ -
tan

FIGURE A.3:  The signs of the trigonometric functions in the various quadrants.

angle a in radians we have

2π 360
a= A × and conversely A = α× .
360 2π
Thus p/4 corresponds to 45°, p/6 to 30° and p/3 to 60°.
Some very useful exact values you are expected to remember are

π 1 π 1 π 3 π

=sin ( 0 ) =
0, sin   = , sin   = , sin   = , sin   1,
6 2 4 2 3 2 2
π 3 π 1 π 1 π
cos ( 0 ) =
= 1, cos   = , cos   = , cos   = , cos   0,
6 2 4 2 3 2 2
 Prerequisites  •  505

and, for any angle q,

π
=
sin ( q ) cos  −q  .
2 
A graphical representation of these functions is given in
Figures A.4 (a) and (b).

tan q
1.0

6
sin q cos q

4
0.5

2
q q
0.0

0
π/2 π 3π/2 2π π/2 π 3π/2 2π

−6 −4 −2
−0.5
−1.0

(a) (b)
FIGURE A.4:  (a) The sin and cos functions. (b) The tan function.

Combining Angles
We can combine angles:
sin(a + b) = sin(a) cos(b) + cos(a) sin(b),
cos(a + b) = cos(a) cos(b) − sin(a) sin(b),
and
a + b  a −b 
sin (a ) + sin ( b ) =
2sin   cos  ,
 2   2 
a + b  a −b 
cos (a ) + cos ( b ) =2 cos   cos  .
 2   2 
And also split them up:
sin(2a) = 2 sin(a) cos(a),
cos(2a) = 2cos2(a) −1 = 1 − 2sin2(a).

Exercise A.3 Show that

(i) sin(p + a) = −sin(a),
(ii) sin(p − a) = sin(a).
 506  •  Mathematical Physics

It is not overly important to remember these formulae

exactly, although it is often convenient to have done so. However,
it is absolutely vital that you know these formulae exist and know
their general form.

Related Functions
Here are some related trigonometrical functions and the rela-
tions between them that are important to know:

1 1 1
=sec (a ) = , cosec (a ) = , cot (a ) ,
cos (a ) sin (a ) tan (a )

where

1 + tan2 (a) = sec2 (a), 1 + cot2 (a) = cosec2(a).

Finally, another useful result (see Exercise A5) to remember is

a sin(a) + b cos(a) = A sin(a + φ),

where A = (a2 + b2)1/2 and φ = tan−1(b/a).

Exercise A.4 Use your calculator to find

(i) sin(135°)
(ii) tan(15p/4)
(iii) cosec(11p/8).

Exercise A.5 By using the identity for sin(a + φ) find A and φ

such that a sin(a) + b cos(a) = A sin(a + φ).

Exercise A.6 Using the identities given, prove that

2 tan( a )
(i) tan( 2a ) = ,
1 −tan 2 ( a )
1 
(ii) sin( a ) + sin( b ) ≤ 2sin  ( a + b )  for ( 0 ≤ a + b ≤ π ) .
2 
 Prerequisites  •  507

A.5.  DIVISION OF POLYNOMIALS

There are several ways to divide a polynomial of higher degree

by one of lesser degree. Here we cover one simple example but you
will find many other methods and examples in different textbooks.

Example A.5 Simplify 2 x + 9 x + 4 x − 20 .

3 2

2x + 5
One basic way to divide a polynomial is to guess the form of the
answer and compare coefficients. In this question the answer
must start with an x2 and hence will be of the form ax2 + bx + c +
R(x) where the remainder, R(x), will be of the form d/(2x + 5).
We can therefore write
2 x3 + 9 x2 + 4 x − 20 d
2x + 5
= ( ax 2
+ bx + c ) +
2x + 5
so
2x3 + 9x2 + 4x − 20 = (2x + 5)(ax2 + bx + c) + d
= 2ax3 + (5a + 2b)x2 + (5b + 2c)x + 5c + d.
Now we can equate the coefficients, starting with the highest
power of x. This allows the unknowns to be read off directly:
equating the coefficients of x3 gives a = 1; equating the
coefficients of x2 gives b = (9 − 5 × 1)/2 = 2; then, similarly,
c = (4 − 5 × 2)/2 = −3 and d = (−20 − 5 × (−3)) = −5. Therefore
2 x3 + 9 x2 + 4 x − 20 2 5
= x + 2x − 3 − .
2x + 5 2x + 5

Exercise A.7 Using the method above, or your own preferred

method, obtain the following results:
3 x 4 + x 3 −14 x 2 + 9 x + 20 1
(i) = x 3 −2 x 2 + 3 − ,
3x + 7 3x + 7
x2 + 3 x + 5 3x + 1
(ii) = 1+ 2 ,
x +4
2
x +4
x 3 + 4 x 2 + x −1 1
(iii) = x+ 1+ 2 .
x + 3 x −2
2
x + 3 x −2
508  •  Mathematical Physics

4 3 2
Exercise A.8 Find x + 5 x + 11 x + 13 x + 7 .
2
x + 2x + 3

A.6.  PARTIAL FRACTIONS

A rational function is a polynomial divided by another polyno-

mial, i.e. a fraction with polynomials as numerator and denomina-
tor. To find the sum (or difference) of two rational functions we put
them over a common denominator. For example:
1 1 ( x + 3) − ( x − 2 ) 5
−= = .
( x − 2 ) ( 3)
x + ( x − 2 )( x + 3 ) ( x − 2 )( x + 3 )
The reverse process is called expressing a rational function in
partial fractions.

Example A.6 Express 1 in partial fractions.

( x − 2 )( x + 3 )
We begin with
1 a b
= + .
( x − 2 )( x + 3 ) ( x − 2 ) ( x + 3)
To find a multiply through by the parenthetical (x − 2),
1 ( x − 2) b
= a+ .
( x + 3) ( x + 3)
The two sides should be equal for all x, including x = 2 which
will get rid of the messy term in b. By inserting x = 2 we find
a = 1/5. Similarly, to find b we multiply by (x + 3) instead:
1 ( x + 3) a
= + b.
( x − 2) ( x − 2)
We choose x = −3 to get rid of the term in a. This gives b = −1/5.
Hence we have the solution
1 1 1
= − .
( x − 2 )( x + 3 ) 5 ( x − 2 ) 5 ( x + 3 )
 Prerequisites  •  509

This is the easiest way to obtain the solution. Alternatively, if we

had chosen different values of x to substitute in when finding
a and b we would have obtained a pair of simultaneous equations
for a and b, which we could then solve to find a and b.

3
Exercise A.9 Express in partial fractions.
( x + 4 )( x − 1)

Exactly the same method works if the numerator is any polyno-

mial of degree less than that of the denominator:

Exercise A.10 Use the method of Example A.6 to express

x in partial fractions.
( x + 4 )( x + 1)

If the numerator is a polynomial of degree greater than or equal

to that of the denominator, then a rearrangement equivalent to long
division with remainder must first be carried out.

Example A.7 Express x2 + 6 x + 4 in partial fractions.

( x + 4 )( x + 1)
We have that
x2 + 6 x + 4 x2 + 6 x + 4
= 2 ,
( x + 4 )( x + 1) x + 5 x + 4

=
(x 2
+ 5x + 4) + x
,
x2 + 5 x + 4
x
= 1+ ,
( x + 4 )( x + 1)
4 1
=
1+ − .
3 ( x + 4 ) 3 ( x + 1)
510  •  Mathematical Physics

A.7. SERIES

Arithmetic Series
An arithmetic sequence (or arithmetic progression) is a succes-
sion of terms, each of which differs from the previous one by adding
a constant.
For example
1, 3, 5, 7, 9,.. .  (A.1)
The sequence may be finite or infinite. If the sequence starts at
a and each term differs from the previous one by a constant d then
the nth term is
an = a + (n − 1)d,
giving a1 = a, a2 = a + d, a3 = a + 2d, and so on. The unevaluated sum
of the terms of a sequence is called a series. For example, the series
obtained from the sequence (A.1) is
1 + 3 + 5 + 7 + 9 + ... .  (A.2)
The sum of the first N terms of an arithmetic sequence is
N N
SN=
=
n
n 1=
∑ a= ∑ ( a + ( n − 1) d ).
n 1

We can evaluate the sum of the first N terms as follows. Write

out the sum in two different ways, first with the terms in ascending
order, then in descending order.
SN = a + (a + d) + (a + 2d) + ... + (a + (N − 2)d) + (a + (N − 1)d),
SN = (a + (N − 1)d) + (a + (N − 2)d) + ... + (a + 2d) + (a + d) + a,
where aN = a + (N − 1)d is the final term in the sequence. If we add
these two together we get

2Sn = (2a + (N − 1)d) + (2a + (N − 1)d) + ... + (2a + (N − 1)d)

= N (a + aN).
So
N N
SN = ( a + aN ) = (first term + last term).
2 2
 Prerequisites  •  511

For example, in the case of the series (A.2) we have a = 1 and

d = 2. Therefore a10 = 19 and S10 = 100.
If there are an infinite number of terms the series will diverge.
If d is positive then SN increases without limit as N increases; if d is
negative SN will decrease without limit.

Geometric Series
A geometric sequence (or geometric progression) is a succes-
sion of terms in which each term differs from the previous one by a
multiplicative constant. For example
1, (1/2), (1/4), (1/8),(1/16),... . (A.3)
The sequence may be finite or infinite. If the sequence starts at
a and each term differs from the previous one by a constant r then
the nth term is
an = ar n−1
giving a1 = a, a2 = ar, a3 = ar2, and so on. The sum of the terms
is called a geometric series. The series obtained from the sequence
(A.3) is
1 + (1/2) + (1/4) + (1/8) + (1/16) + ... .  (A.4)
The sum of the first N terms of a geometric sequence is
N N
=
SN
=n 1=
∑=
an
n 1
∑ ar n −1
.

We can evaluate the sum of the first N terms as follows. Write

out the sum SN and also rSN,
SN = a + ar + ar2 + ar3 ⋯ + arN-1,
rSN = ar + ar2 + ar3 + ar4 ⋯ + arN,
and then subtract the second from the first. The only terms that do
not cancel are the first term of the first equation and the last term of
the second equation
(1 − r)SN = a − arN.
The sum of the first N terms can therefore be written as
a (1 − r N )
SN = .
1−r
512  •  Mathematical Physics

For an infinite sequence, if |r| < 1 we have limN→∞ rN = 0, and so

the sum tends to
∞
a
=S ∑=
ar
n =1
n

1−r
. (A.5)

For example, in the case of the series (A.4) we have r = 1/2 and
a = 1. Therefore S = 2. For | r | ≥ 1 the sum Sn either diverges with
N or it oscillates with increasing amplitude.
 APPENDIX

B
THE GREEK ALPHABET

alpha a A nu n N
beta b B xi x Ξ
gamma γ Γ omicron o O
delta d ∆ pi p Π
epsilon e E rho r R
zeta z Z sigma s Â
eta h H tau t T
theta q Θ upsilon t ϒ
iota i I phi φ Φ
kappa k K chi c C
lambda l Λ psi y ψ
mu m M omega w Ω
INDEX
Active transformation, 186, 191 Binomial coefficients, 52
Adjoint matrix, 180 Binomial expansion, 55
Algebraic methods, 84 binomial series, 53–54
Algebraic operations, 192 definition, 50–51
Algebraic rules, 111 factorials, 51–53
Analytic fucntion, 261 Boundary conditions, 267–277,
Angular momentum, 139, 141 289–290
Anti-derivatives, 18
Antisymmetric matrices, 160–161 Carrier waves, 399
Arbitrary function, 49, 387 Cartesian coordinate systems,
Arbitrary point, 357–358 320, 324, 455, 491
Arcsin, 11, 70 Cauchy-Riemann relations, 261
Argand diagram, 247–252, Cauchy’s integral formula, 261
255, 262 Chain rule, 65, 228, 350,
Associative law, 158 367–369,
Asymptotes, 79 372, 380, 383, 386, 459.
Asymptotically equivalent, See also “Function of a
94–96 function”
Asymptotic approximations, 103 consistency with, 369–371
Auxiliary equations, 213, 215 Closed form expression, 99
with complex roots, 277–281 Coefficients, 207
with repeated roots, Column vector, 149, 150
215–217 Commute, 157–158
Axially symmetric, 321 Complementary functions, 223,
224, 233, 271, 284
Back-substitution, 154 Complex algebra, basic rule
“Bac minus cab” rule, 134 of, 239
Basis vectors, 115–118 Complex conjugate, 240–241,
Beats, 397–400 283
Bessel functions, 231–232 Complex exponential, 252–255
516  •  Mathematical Physics

Complex Fourier series, Cylindrical polar coordinates,

436–438 320–321, 322–323, 335
Complex impedance, 292–293
Complex inhomogeneous term, D’Alembert’s formula, 394
282–289 Deconvolution theorem,
Complex numbers, 235–238 445–446
Argand diagram, 247–249 Definite integrals, 28, 29, 32,
argument of, 249–252 319–320
complex exponential, 252–255 Del operator, 458
De Moivre’s theorem, 256–257 De Moivre’s theorem, 256–257,
division, 242–243 262
inverse of, 243 Dependent variable, 348
modulus, 249–252 Derivatives
operations, 239–243 definition, 1–4
polar representation, 251 exponential, 64–66
product, 243 of “function of a function,”
quadratic equation, 244–247 5–7
real polynomial, roots of, higher order, 14–15
258–259 implicit differentiation, 11–12
square root, 244, 245, 246 of integrals, 33–35
unity, roots of, 258 piecewise differentiable
variables, 260–262 function, 12–14
Complex polynomials, roots of, product and quotient rule,
259–260 8–10
Conservation of energy, 413 standard, 5
Conservation of probability, 413 Determinants, 167–169
Constant coefficients, 79, 211, properties, 169–174
214, 215, 218, 276 Differential equations
Constants, 305 application to, 428–429
Continuous functions, 17, 30–32 arbitrary constant in, 204
Contour integration, 262 auxiliary equations with
Contour lines, 342 complex roots, 277–281
Convergence, 93, 101–102 auxiliary equation with
Convolutions theorem, 443–445 repeated roots, 215–217
Cosh functions, 71 boundary and initial
Cosine formula, 136 conditions, 267–277,
Cosine series coefficients, 289–290
430–431 with complex coefficients,
Cramer’s rule, 174, 178, 179 281–282
Cross product, 121–122 complex impedance, 292–293
Cubic polynomials, 79, 80 complex inhomogeneous
Curl of vector field, 470–473 term, 282–289
Cylindrically symmetric, 321 definition, 199
 Index  •  517

examples, 200–201 Electric circuit theory, 280

finding solutions, 209–211 Electromagnetic field, 379
first order, 203–205, 226 Electromagnetic radiation, 310
first order equations, Electro-magnetism, 478
291–292 Elementary functions, 30
Green functions, 296–297 binomial expansion. See
homogeneous linear Binomial expansion
equations, 207–209, definition, 49
211–215 equilibrium point, 59–62
inhomogeneous, 207–209, exponential function, 63
217, 222–226 integrals of, 18–20
integrating factor method, inverse function, 63–64
226–229 Maclaurin series, 55–57
linearity vs. non-linearity, Taylor series, 58–59
205–207 Equation of a line, 125–127, 141
order of, 200–201 Equation of a plane, 127–129, 141
particular integrals, 217–222 Equilibrium point, 59–62
solving, 201–202 Equivalent system, 154
special functions, 230–233 Estimates of integrals, 32–33
types, 201 Euler’s equation, 252, 278, 280,
Differentiation 285
implicit, 352 Even functions, 85–87
standard rules, 343 Exact differential, 366–367
Diffusion equation, 381 Exponential functions, 49, 63,
Directional derivatives, 346–347 67, 72, 252
Discontinuous functions,
30–32, 439 Factorials, 51–53
Discrete symmetries, 192 Finite string, 402
Dispersive medium, 400 First order equations, 291–292
Distributive law, 133 First order separable equations,
Divergence of vector fields, 203–205
467–469 Fixed vectors, 112
Divergence theorem, 93, 101-102, Fluid mechanics, 478
482–486, 492 Fourier integrals, 440–441
Dot product, 118–121 Fourier methods, 419, 420
Double integrals, 304, 306, 308 Fourier series, 101
complex form of, 436–438
Eigenfunctions, 407, 414 convolutions theorem,
Eigenvalues, 147, 181–186, 192, 443–445
406–408, 414 cosine series coefficients,
Eigenvectors, 147, 181–186, 192 430–431
Einstein summation deconvolution, 445–446
convention, 140 definition, 419
518  •  Mathematical Physics

differential equations, Gamma, 39

application to, 428–429 gradient, 458–460
even and odd functions, green, 296–297
433–435 hyperbolic, 67–69, 254
Fourier transforms. See integrals of combinations,
Fourier transforms 20–21
full range series, 435–436 integration, 21
half range sine series, inverse, 63–64, 69–71
420–421, 422 inverse hyperbolic, 71
Laplace transform, 446 of many variables, 348
numerical series, 432 maxima/minima, 15
periodic extension of, 433 mod, 13–14
properties, 439–440 piecewise differentiable,
results, 422–425 12–14
sine series coefficients, piecewise linear, 13
421–422, 424, 429 polynomial, 59, 79–81, 103
standard integrals, 449–450 quadratic, 77–78
wave equation, application rational, 27, 81–82, 103
to, 425–428 sinh, 71
Fourier’s theorem, 421 stationary point, 15–17
Fourier transform, 440–442, 448 symmetry of, 85–87
calculating, 442–443 Taylor series, 58–59
Fraktur typeface, 238 trigonometric, 4, 49, 68, 252
Free vectors, 112 of two variables, 341–344
Fubini’s theorem, 311 Fundamental theorem of
Full range series, 435–436 algebra, 259
“Function of a function” rule,
5–7, 35, 39, 350–351, 367 Gamma function, 39, 94
Functions Gaussian function, 442–444
approximation, 94–96 Gaussian integral, 67
area under a graph, 29 Gauss’s divergence theorem,
asymptotic approximations 482–486
of, 103 Geometric progression, 98
complementary, 271 Gibbs phenomenon, 439
continuous, 17, 30–32 Graphical method, 84
cosh, 71, 94 Graphical solution of
definite integrals, 28, 29 inequalities, 83–85
derivatives of, 1–4 Green functions, 296–297
discontinuous, 30–32 Group velocity, 400–402
elementary, 18–20, 49
even and odd, 85–87, Half range cosine series,
433–435 430–431
exponential, 49, 63, 67, 252 Half range sine series, 420–421
 Index  •  519

Harmonic waves, 397, 401 methods, 40

Hermite polynomials, 232 over cylindrical surfaces,
Higher derivatives, 349–350 323–324
Higher order derivatives, 14–15 over rectangles in plane,
Homogeneous linear equations, 307–310
179, 207–209, 211–215, over regions in plane,
277, 381 316–317
Hyperbolic functions, 67–69, over spherical surfaces,
72, 254 327–329
by parts, 25–26, 40
Identity matrices, 159–160 of rational function, 27
Implicit differentiation, 11–12, by substitution, 21–25
39, 352 Inverse functions, 63–64, 69–71
Incident energy flux, 410 Inverse hyperbolic functions, 71
Incremental approximation, 460 Inverse of 2 × 2 matrices,
Indefinite integral, 18, 28, 67, 165–167
211 Inverse sine function, 11
Independent variable, 69, 201, Inverses matrices, 161–165
269, 275, 348 formula for, 179–181
Indeterminate forms, 89–91 Inverse square law, 351
Index notation, 140–141, Irrational numbers, 235–236
490–491 Iterated integral, 304
Inequalities
graphical solution of, 83–85 Kronecker delta, 140, 421
triangle, 120
Inexact differentials, 367 Laguerre polynomials, 232
Infinite limits, 92 Laplace’s equation, 381, 405,
Infinite series, 97–103 408
Infinite strings, 393–395, 402 Laplace transforms, 446
Inhomogeneous linear Laurent series, 261
differential equations, Legendre polynomials, 232–233
207–209, 217, 222–226 Leibniz rule, 8
Initial conditions, 267–277, Leonard-Jones potential, 61
289–290 Level curves, 342, 462–464
Initial transverse velocity, L’Hôpital’s rule, 89, 91–92, 103
391–393 Lie groups, 192
Integrals, limits of, 92–94 Limits, 87–88
Integrating factor method, indeterminate, 89
226–229 of integrals, 92–94
Integration, 18–20 Linear algebra, 147
of exponential functions, 67 Linear combinations, 113, 141
inequality, 32 Linear dependence vectors,
logarithmic, 21 114–115
520  •  Mathematical Physics

Linearity differential equations, Mod function, 13–14

150, 381, 205–207 Multiple integrals, 335
homogeneous, 179, 211–215 cylindrical polar
inhomogeneous, 217, coordinates, 320–321
222–226 definite integral, 319–320
solution, 153–156 order of integration,
3 × 3 linear system, 177–178 313–316
2 × 2 linear system, 174–177 over irregular regions,
Linearly independent vectors, 115 310–313
Line segments, 112–113 over rectangles in plane,
Logarithmic integration, 21 307–310
Lorentz group, 191 over regions in plane,
Lower triangular matrix, 154 316–317
over spherical surfaces,
Maclaurin series, 55–57, 58, 64, 327–329
65, 72, 356 polar coordinates, 316–317
Matrices, 147–149 repeated integrals, 303–307
adjoint, 180 sketching surfaces, 333–335
determinants, 167–169 solid angle, 329–333, 335
properties, 169–174 spherical polar coordinates,
eigenvalues and 324, 325–326
eigenvectors, 181–186 volume integrals. See
identity, 159–160 Volume integrals
inverse of 2 x 2, 165–167 Multiple products, 123–125
inverses, 161–165
linear equations systems, Natural logarithm, 4
150, 153–156 Negative divergence, 469
orthonormal/unitary, 188 Non-coplanar vectors, 115, 124
products, 156–159 Non-dispersive, 401
representation, 149–152 Non-linearity differential
square, 181, 188 equations, 205–207
symmetric and Non-negative square root, 70
antisymmetric, 160–161 Non-singular, 173
3 × 3 linear system, 177–178 Non-zero initial velocity, 391
transformations, 186–190 Normal modes, 405
groups, 191–192 Notation, 151, 152
transpose of, 160 Numerical series, 432
2 × 2 linear systems, 174–177
Matrix notation, 151 Odd functions, 85–87
Matter waves incident, 412–414 Order of integration, changing,
Maximum rate of change, 313–316
460–462 Orthogonal coordinate system,
Minors, 168 491
 Index  •  521

Orthonormal basis, 116 Passive transformation, 186, 191

Orthonormal matrices, 188, 192 PDEs. See Partial differential
Oscillatory motion, 419 equations
Perfect differential. See Exact
Parallelogram law, 113 differential
Partial derivatives, 344–346 Permutation symbol, 140
application of, 353 Phase velocity, 386
arbitrary point, 357–358 Piecewise differentiable
chain rule, 367–369 function, 12–14, 31, 32
functions of two variables, Piecewise linear functions, 13
341–344 Polar coordinates, 316–317
higher derivatives, 349–350 cylindrical, 320–323
notion of, 348 Polynomial function, 59, 79–81,
with respect to parameter, 103
353–355 Power series expansions, 49
stationary points, 359–365 Precise rule, 94
total differential, 365–366 Probability theory, 67
Partial differential equations Product rule, 8–10, 39
(PDEs), 379 Pure imaginary system, 238
beats, 397–400
general initial conditions, Quadratic equation, 236,
formula for, 393–395 244–247, 364
general solution derivation, Quadratic function, 77–78
386–387 Quotient rule, 8–10
group velocity, 400–402
infinite string, 393–395 Radioactive decay law, 211
initial transverse velocity, Rational functions, 27, 81–82, 103
391–393 Rational numbers, 235
matter waves incident, Real number system, 236, 238
412–414 Real polynomial, roots of,
in physics, 380–382 258–259
reflections at discontinuity, Reciprocal vector, 124
408–411 Rectified half wave, 435
semi-infinite string, 395–396 Recurrence relation, 38. See
separation of variables, Reduction formula
402–405 Reduction formula, 35–39, 39,
simple harmonic waves, 397 40, 94
string initially at rest, Reflection coefficient, 411, 413
388–391 Reflection matrix, 190
wave equation. See Wave Related functions, 49
equation Repeated integrals, 303–307
Partial sums, 98, 100 Repeated roots, auxiliary
Particular integrals, 217–222, 223 equation with, 215–217
522  •  Mathematical Physics

Resultant vector, 112 Stirling’s formula, 95

“Right hand screw rule,” 473 Stokes’s theorem, 486–490, 493
Roots of unity, 258 Symmetric matrices, 160–161, 186
Rotation matrix, 186–190, 191 Symmetry of functions, 85–87
Row vector, 149
Taylor series, 58–59, 66, 72, 230,
Sarrus’s rule, 169 341, 355–357, 460
Scalar, multiplication by, 151 Taylor’s theorem, 359, 363
Scalar field, 455, 456, 460 Tensor product, 118
Scalar (dot) product, 113, Three-dimensional space,
118–121 115, 116
component form, 131–132 3 × 3 linear equations systems,
of vectors, 460 177–178
Scalar triple product, 123–124, Torque, 139, 141
133–134, 135, 141 Total differential, 365–366
Schrödinger equation, 381, 405, Transformation matrices,
408, 412 186–190
Semi-infinite string, 395–396, 408 groups, 191–192
Separation of variables method, Transmission coefficient, 411, 413
402–405 Transverse acceleration, 380
Sequence, 96–97 Transverse velocity, 386
Series, 96–97, 100 Triangle inequality, 120
Set of coefficients, 149 Trigonometric functions, 4, 40,
“Sign” function, 13 49, 68, 230, 252
Simple harmonic motion, 268 Triple product
Simple harmonic waves, 397 scalar, 123–124, 133–134
Simpler integral, 21 vector, 125, 134–135
Simultaneous equations Triplet of numbers, 130
method, 156 Trivial solution, 179
Sine series coefficients, 421–422 2 × 2 linear equations systems,
Sinh functions, 71 174–177
Sketching surfaces, 333–335
Solid angle, 329–333, 335 Unitary matrices, 188
Spherical polar coordinates, 324, Unit matrix, 159
325–326, 328, 335 Unit vectors, 116, 118, 120
Spherical triangle, 136 Upper-triangular matrix, 154
Square matrix, 152, 181, 188
Standard derivatives, 5 Variables
Standard form, 238 dependent, 348
Standard integral, 21 independent, 348
Standard rules, 343 Vector
Stationary points, 15–17, 40, basic properties, 112–114
359–365, 372 components, 129–130
 Index  •  523

definition, 111 index notation, 490–491

equation of a line, 125–127 level surfaces, 462–464
equation of a plane, 127–129 maximum rate of change,
identities, 135–138 460–462
index notation, 140–141 normal to surface, 464–468
linearly independent, 115 scalar field, 455, 456
magnitude, 114 Stokes’s theorem, 486–490
multiple products, 123–125 with zero curl, 473
multiplication, distributive Vector (cross) product, 121–122
law for, 122 component form, 132–133
non-parallel, 115 Vector space, 115
parallelogram law of, 111 Vector triple product, 125,
physical science examples, 134–135, 141
138–140 Velocities, 111
product, 118 Volume integrals
reciprocal, 124, 125 cylindrical polar
row and column, 149 coordinates, 322–323
scalar (dot) product, spherical polar coordinates,
118–121 326–327
unit, 116, 118
zero, 114 Wave equation, 379, 380, 381,
Vector fields, 455–458 407
circulation, 478–481 application, 425–428
curl of, 470–473 general solution of, 383–386
divergence, 467–469
flux, 473–477 Zero vector, 114