MS327

Deterministic and stochastic dynamics

Handbook
This publication forms part of an Open University module. Details of this and other Open University modules
can be obtained from Student Recruitment, The Open University, PO Box 197, Milton Keynes MK7 6BJ,
United Kingdom (tel. +44 (0)300 303 5303; email [email protected]).
Alternatively, you may visit the Open University website at www.open.ac.uk where you can learn more about
the wide range of modules and packs offered at all levels by The Open University.

The Open University, Walton Hall, Milton Keynes, MK7 6AA.

First published 2017. Second edition 2018.
c 2017, 2018 The Open University
Copyright
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, transmitted or
utilised in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without
written permission from the publisher or a licence from the Copyright Licensing Agency Ltd, Saffron House,
6–10 Kirby Street, London EC1N 8TS (website www.cla.co.uk).
Open University materials may also be made available in electronic formats for use by students of the
University. All rights, including copyright and related rights and database rights, in electronic materials and
their contents are owned by or licensed to The Open University, or otherwise used by The Open University as
permitted by applicable law.
In using electronic materials and their contents you agree that your use will be solely for the purposes of
following an Open University course of study or otherwise as licensed by The Open University or its assigns.
Except as permitted above you undertake not to copy, store in any medium (including electronic storage or use
in a website), distribute, transmit or retransmit, broadcast, modify or show in public such electronic materials in
whole or in part without the prior written consent of The Open University or in accordance with the Copyright,
Designs and Patents Act 1988.
Edited, designed and typeset by The Open University, using LATEX.
Printed in the United Kingdom by Latimer Trend and Company Ltd, Plymouth.

SUP 06129 4
2.1
 1 Notation

Introduction
This Handbook is a reference that you can take into the MS327 exam. It
will be more effective if you are already familiar with it before you sit the
exam, and we suggest that you consult it when you attempt the
assignments in the module. This Handbook is not designed as a teaching
document, and reading it is not a substitute for studying the module units.
The first few sections consist of general mathematical reference material,
largely based on the topics taught at Levels 1 and 2. It is not intended to
be a self-contained or logically complete account of basic mathematics; it is
just a set of useful definitions and results. Some of these results are used
repeatedly in the module.
The later sections of the Handbook are brief summaries of the units,
emphasising the most important results.

1 Notation
1.1 Greek alphabet
α A alpha ι I iota ρ P rho
β B beta κ K kappa σ Σ sigma
γ Γ gamma λ Λ lambda τ T tau
δ ∆ delta µ M mu υ Υ upsilon
ε E epsilon ν N nu φ Φ phi
ζ Z zeta ξ Ξ xi χ X chi
η H eta o O omicron ψ Ψ psi
θ Θ theta π Π pi ω Ω omega

1.2 Symbols
√
= is equal to positive square root
6 = is not equal to e the number 2.718 28 . . .
' is approximately equal to π the number 3.141 59 . . .
± plus or minus ∞ infinity
∓ minus or plus Z the integers
< less than R the real numbers
≤ less than or equal to C the complex numbers
> greater than
≥ greater than or equal to

1.3 Limits and sums

An ordered list of numbers x0 , x1 , x2 , . . . , xn , . . . is said to converge to the
limit x if successive terms in the ordered list are better and better
approximations to x. We write ‘xn → x as n → ∞’ or ‘ lim xn = x’.
n→∞

3
Handbook

Given numbers a1 , a2 , . . . , an , we define

n
X
ai to mean a1 + a2 + · · · + an−1 + an .
i=1

2 Numbers
2.1 Real numbers
The integers are the positive and negative whole numbers, together with
zero. Non-integer numbers that can be expressed exactly as fractions are √
called rational numbers; those that cannot be so expressed, such as 2,
e and π, are called irrational. The collection of the rational numbers
(including the integers) and the irrational numbers is called the set of real
numbers.
When a real number is expressed in decimal notation, if it is approximated
then the approximation can be given to so many decimal places, or to so
many
√ significant figures. For example, 1.4142 is the approximation to
2 to four decimal places and five significant figures, while 0.000 000 342 is
given to nine decimal places but three significant figures, and 342 000 has
no decimal places but at least three significant figures. The process of
reducing the number of decimal places or significant figures to which a
number is expressed is referred to as rounding. To round a given number
to n decimal places or n significant figures, take the number expressed to
n decimal places or n significant figures that is closest to the given
number, where it is conventional to round the digit 5 up. (Be aware that
other conventions for rounding exist, and that computer programs and
calculators do not always use this convention.)
A number given in the form ±b × 10c , where 1 ≤ b < 10 and c is an integer,
is said to be in scientific notation. For example, the number 342 000 can
be expressed as 3.42 × 105 in scientific notation.
For any two real numbers a and b with a < b, we write:
[a, b] for the set of all real numbers x such that a ≤ x ≤ b;
[a, b) for the set of all real numbers x such that a ≤ x < b;
(a, b] for the set of all real numbers x such that a < x ≤ b;
(a, b) for the set of all real numbers x such that a < x < b.
These sets of numbers are called intervals. The interval [a, b] is a closed
interval, (a, b) is an open interval, and the two others are half-open
intervals. For practical problems the distinction between open and closed
intervals is rarely significant. If we know that a real number x is 1.274 to
three decimal places, then x lies in the interval [1.2735, 1.2745).

4
 2 Numbers

2.2 Complex numbers

A complex number z is written in Cartesian form as z = a + bi, where
a and b are real numbers, and i2 = −1. We refer to a as the real part
of z, written <(z), and to b as the imaginary part of z, written =(z).
Complex numbers can be added, e.g.
(a + bi) + (c + di) = (a + c) + (b + d)i,
or multiplied, e.g.
(a + bi)(c + di) = (ac − bd) + (ad + bc)i.
These formulas make use of the ordinary rules of algebra, together with
the relation i2 = −1.
The complex conjugate of z = a + bi is z = a − bi. Note that The notation z ∗ is also used in
some texts for the complex
zz = (a + bi)(a − bi) = a2 + b2 conjugate of z.
z and z ∗ are read as ‘z bar’ and
is a positive real
√ number
√ (unless a = b = 0). The modulus of z is the ‘z star’.
number |z| = zz = a2 + b2 .
To calculate a quotient of complex numbers, multiply top and bottom by
the complex conjugate of the bottom, e.g.
(a + bi)(c − di) bc − ad
   
a + bi ac + bd
= = + i,
c + di (c + di)(c − di) c2 + d2 c2 + d2
which expresses the quotient in the form p + qi with p and q real.
The Argand diagram is a representation of complex numbers as points
in a plane, where the complex number a + bi is represented by the point
with Cartesian coordinates (a, b). A point can also be represented in polar
coordinates as (r, θ) and related to its Cartesian coordinates by Polar coordinates are discussed
in Subsection 5.2.
a = r cos θ, b = r sin θ.
The angle coordinate θ is referred to as an argument of z, written arg(z),
and the unique value of θ in the range −π < θ ≤ π is referred to as the
principal value of the argument, written Arg(z).
The multiplication of complex numbers in polar form is given by the rule
(r, θ) × (s, φ) = (rs, θ + φ). So powers of a complex number can be
expressed as (r, θ)n = (rn , nθ), for n a positive integer. The special case
when r = 1 is known as de Moivre’s theorem:
(cos θ + i sin θ)n = cos(nθ) + i sin(nθ).
Euler’s formula is
eiθ = cos θ + i sin θ.
This extends to any complex number a + bi as
ea+bi = ea ebi = ea (cos b + i sin b).

5
Handbook

The exponential form of a complex number z = r(cos θ + i sin θ) is

z = reiθ .
This form is useful for multiplying, dividing and taking powers of complex
numbers.

3 Functions and graphs

3.1 Functions
A variable is a quantity, represented by a symbol, that can vary over a set
of values. If its value does not vary, then it is a constant.
Any expression or formula that involves a variable x, and whose value is
uniquely determined by the value of x, is called a function of x.
If a variable y is a function of x (i.e. if y is equal to a function of x), then
we call x the independent variable and y the dependent variable, and
we may write y = y(x). Here y(x) stands for the function of x (i.e. for the
formula involving x).
If f and g are two functions, then their sum is a function f + g defined by
(f + g)(x) = f (x) + g(x) (for all x).
Moreover, if A and B are any two numbers, then the function Af + Bg is
defined by
(Af + Bg)(x) = A f (x) + B g(x) (for all x).
The function f (g(x)) is called the composite function or composition
of the functions f and g.
The graph of a function f (x) is the curve in the (x, y)-plane whose
equation is y = f (x).
A continuous function is one whose graph has no breaks or jumps in it,
i.e. it can be drawn without lifting your pen from the paper.
A constant function f (x) is one that assigns the same value to any
input. Its graph is a straight line parallel to the x-axis. A special case is
the zero function, which assigns the value 0 to any input.
A linear function is one having the form a1 x + a0 (with a1 6= 0), where
a1 and a0 are constants. Its graph is a straight line with slope a1 and
y-intercept a0 .
A quadratic function is one having the form a2 x2 + a1 x + a0 (with
a2 6= 0), where a2 , a1 , a0 are constants. Its graph is a parabola, similar in
shape to the one shown in Subsection 5.3 if a2 > 0, but the other way up if
a2 < 0.
A cubic function is one having the form a3 x3 + a2 x2 + a1 x + a0 (with
a3 6= 0), where a3 , a2 , a1 , a0 are constants.

6
 3 Functions and graphs

3.2 Polynomials
Linear, quadratic and cubic functions are all particular examples of
polynomial functions, or simply polynomials. An nth-order polynomial,
or polynomial of degree n, is a function of the form
an xn + an−1 xn−1 + · · · + a1 x + a0 ,
where n is a positive integer, x is a variable and a0 , a1 , . . . , an are
constants with an 6= 0. A linear polynomial has n = 1 (or n = 0), a
quadratic polynomial has n = 2, and a cubic polynomial has n = 3.
The roots of a polynomial p(x) are the solutions of the equation p(x) = 0.
Every polynomial of degree n can be written as a product of an and
n factors of the form x − ck (k = 1, 2, . . . , n), with each ck a complex
number (which may be real). Each of these factors corresponds to a root
x = ck of the polynomial. If a factor x − c occurs more than once, then the
root x = c is a repeated root; repeated roots are also sometimes referred
to as equal roots or coincident roots.
The roots of a quadratic equation ax2 + bx + c = 0, a 6= 0, are given by the
formula method as
√
−b ± b2 − 4ac
.
2a
The quantity b2 − 4ac is referred to as the discriminant of the quadratic
equation.
To factorise a polynomial is to express it as a product of two or more
polynomials of lower degree. For example, the difference of two squares
x2 − a2 factorises as
x2 − a2 = (x − a)(x + a),
and the perfect square x2 + 2ax + a2 factorises as
x2 + 2ax + a2 = (x + a)(x + a) = (x + a)2 .

3.3 Exponentials and logarithms

A function of the form y = bax , where a and b are non-zero constants (with
a > 0 and a 6= 1), and x is real, is said to exhibit exponential
behaviour. In ax , a is referred to as the base and x as the exponent (or
index or power). Properties of such a function include
a0 = 1,
a−x = 1/ax ,
ax × ay = ax+y ,
ax /ay = ax−y ,
(ax )y = axy = (ay )x .
The function ex , where e = 2.718 28 . . ., is referred to as the exponential Some texts refer to y = bax as
function. It is also written as exp x. an exponential function; this is
not to be confused with the
exponential function, exp x.

7
Handbook

The natural logarithm function ln x is defined to be the inverse

function of the exponential function exp x, i.e. each reverses the effect of
the other, so that
ln(exp x) = x for all real x,
exp(ln x) = x for all real x > 0.
In other words, if ey = x, then y = ln x, and vice versa.
The natural logarithm function ln x, for x > 0, has the properties
ln 1 = 0,
ln(1/x) = − ln x,
ln(xy) = ln x + ln y, y > 0,
ln(x/y) = ln x − ln y, y > 0,
ln xy = y ln x.
Any function y = bax can be written in the form y = bekx , where k = ln a.
Another logarithm function is log10 x, for x > 0, where y = log10 x if
10y = x (and vice versa). The properties given above for ln also hold for
log10 .
The hyperbolic functions sinh x, cosh x and tanh x are defined as
combinations of exponential functions:
sinh x = 21 (ex − e−x ),
cosh x = 21 (ex + e−x ),
sinh x ex − e−x
tanh x = = x .
cosh x e + e−x
The inverses of these functions
p are
arcsinh x = ln(x + x + 1),2
p
arccosh x = ln(x + x2 − 1), x ≥ 1,
 
1 1+x
arctanh x = 2 ln , |x| < 1.
1−x

8
 3 Functions and graphs

3.4 Graphs of some common functions

y y y √
x2 x3 x

x x x

y y y
1 ex ln x
x 1

x x 1 x

y y y
tan x

sin x 1 1
cos x

−π π x −π π x −π π x
−1 − π2 π
−1 2

y y y
cosec x sec x cot x

1 1

−π π x −π π x −π π x
−1 − π2 π
2
−1

9
Handbook

y y y

π
π π
π
2 π
arcsin x π arccos x 2 arctan x
2

−1 1 x −1 − π 1 x x
2 − π2
− π2
−π −π
−π

y y y

π π π
π
2 π arcsec x π
arccosec x 2
2 arccot x

−1 1 x −1 1 x x
− π2 − π2
− π2
−π −π −π

4 Trigonometry
4.1 Radians and degrees
In this module we usually measure angles in radians rather than degrees.
There are 2π radians in a full circle, corresponding to 360◦ , so 1 radian is
(180/π)◦ ' 57◦ . An advantage of working in radians is the simplicity of
the formula for the arc length subtended by an angle in a circle of radius r:
the length of the arc subtended by an angle of θ radians is simply rθ. The
following radian measures of standard angles are worth knowing:
π
• a right angle is 2 radians
π
• the angles of an equilateral triangle are 3 radians.
π
An angle is acute if its radian measure lies between 0 and 2, and obtuse
if its radian measure lies between π2 and π.

10
 4 Trigonometry

4.2 Trigonometric functions and their

inverses
For an acute angle θ, the values of trigonometric functions are related to
the ratios of lengths of the sides of a triangle as follows.

Function Definition for acute angles Definition in terms

in terms of triangle shown of sin and cos
a
sin θ
c
b
cos θ
c
a sin θ
tan θ c
b cos θ a
b cos θ θ
cot θ
a sin θ b
c 1
sec θ
b cos θ
c 1
cosec θ
a sin θ

Inverse function Definition

arcsin x = θ where sin θ = x
arccos x = θ where cos θ = x
arctan x = θ where tan θ = x
arccot x = θ where cot θ = x
arcsec x = θ where sec θ = x
arccosec x = θ where cosec θ = x

4.3 Two useful triangles

From the two triangles in the margin it can be seen that:
√
3
sin π6 = 12 , cos π6 = 2 , tan π6 = √1 ,
3
√
cosec π6 = 2, sec π6 = √2 ,
3
cot π6 = 3; π
6
sin π
= √1 , cos π
= √1 , tan π4 = 1, 2 √
4 2 4 2 √ 3
√ √
cosec π4 = 2, sec π
= 2, cot π4 = 1; 2
4 1
√ √ π π
3 3
sin π3 = 2 , cos π3 = 21 , tan π3 = 3, 4

π 1 1
cosec √2 sec π3 = 2, cot π3 = √1 .
3 = 3, 3

11
Handbook

Other values of the trigonometric functions worth remembering are:

sin 0 = 0, cos 0 = 1, tan 0 = 0;
π π
sin 2 = 1, cos 2 = 0;
sin π = 0, cos π = −1, tan π = 0;
3π 3π
sin 2 = −1, cos 2 = 0.

4.4 Trigonometric identities

Pythagoras’ theorem states that for any right-angled triangle, if c is the
length of the hypotenuse (the side opposite the right angle) and a and b
are the lengths of the other two sides, then
c2 = a2 + b2 .
This leads to the following trigonometric identities:
sin2 θ + cos2 θ = 1,
tan2 θ + 1 = sec2 θ,
1 + cot2 θ = cosec2 θ.

Addition formulas
sin(α + β) = sin α cos β + cos α sin β,
sin(α − β) = sin α cos β − cos α sin β,
cos(α + β) = cos α cos β − sin α sin β,
cos(α − β) = cos α cos β + sin α sin β,
tan α + tan β
tan(α + β) = ,
1 − tan α tan β
tan α − tan β
tan(α − β) = ;
1 + tan α tan β
1 1
sin α cos β = 2 sin(α + β) + 2 sin(α − β),
cos α sin β = 12 sin(α + β) − 21 sin(α − β),
cos α cos β = 12 cos(α + β) + 12 cos(α − β),
sin α sin β = 21 cos(α − β) − 12 cos(α + β).
In particular, these formulas give
sin(α + 2π) = sin α, cos(α + 2π) = cos α, tan(α + π) = tan α;
sin(−α) = − sin α, cos(−α) = cos α, tan(−α) = − tan α.

12
 4 Trigonometry

Double-angle formulas
sin 2α = 2 sin α cos α,
cos 2α = cos2 α − sin2 α = 1 − 2 sin2 α = 2 cos2 α − 1,
2 tan α
tan 2α = ,
1 − tan2 α
sin2 α = 12 (1 − cos 2α),
cos2 α = 12 (1 + cos 2α).

Cosines of related angles

cos( π2 − α) = sin α, cos( π2 + α) = − sin α,
cos(π − α) = − cos α, cos(π + α) = − cos α.

4.5 General sinusoidal functions

A sinusoidal function or sinusoid is a function x(t) of time t of the form
x = x0 + A cos(ωt + φ) = x0 + A sin(ωt + ψ),
where x0 is a constant, A is a positive constant called the amplitude, ω is
a positive constant called the angular frequency, and φ and ψ are
constants called phase constants.
A sinusoidal function oscillates between x0 − A and x0 + A, repeating the
same pattern of oscillations through each time interval of length 2π/ω,
known as the period of the function. For these reasons, sinusoidal
functions are examples of oscillatory functions and of periodic functions.
The phase constants in the two forms of the sinusoidal function are related
according to φ = 3π π
2 + ψ (or equivalently φ = ψ − 2 ).
Alternative forms of the sinusoidal functions are given by
x = x0 + A cos(ωt + φ) = x0 + B cos(ωt) + C sin(ωt), (1)
x = x0 + A sin(ωt + ψ) = x0 + D sin(ωt) + E cos(ωt). (2)
√ B = A cos φ and C = −A sin φ. Solving these
In equation (1) we have
equations gives A = B 2 + C 2 and φ = ± arccos(B/A), with + sign if
C < 0.
In equation (2) we have
√ D = A cos ψ and E = A sin ψ. Solving these
equations gives A = D2 + E 2 and ψ = ± arccos(D/A), with + sign if
E > 0.

13
Handbook

5 Geometry
5.1 Cartesian coordinates
The Cartesian coordinates (x, y) of a point P in a plane specify the
position of that point relative to two perpendicular axes, the x-axis (or
horizontal axis) and y-axis (or vertical axis), which meet at a point O
called the origin, with Cartesian coordinates (0, 0). The directions of the
axes indicate increasing numerical values for the x- and y-coordinates.
Values of x to the right of the y-axis are positive, and those to the left are
negative; similarly, values of y above the x-axis are positive, and those
below are negative. The four parts into which a plane is divided by
Cartesian coordinate axes are known as quadrants of the plane. A plane
on which Cartesian coordinate axes have been specified is often referred to
as the (x, y)-plane.
y
y
Second quadrant First quadrant
P (x, y)
y-axis x<0 x>0
y>0 y>0

O x
origin x<0 x>0
x y<0 y<0
O
x-axis Third quadrant Fourth quadrant

5.2 Polar coordinates

The point P whose polar coordinates are (r, θ) has Cartesian coordinates
x (x, y) where
P
r y x = r cos θ, y = r sin θ.
θ The value of r is always positive (except at the origin, where it is zero).
For a given point P , the value of θ is not unique: we can add or subtract
The relationship between polar
any integer multiple of 2π and obtain another value for θ that describes
and Cartesian coordinates
the same point. The value of θ satisfying −π < θ ≤ π is called the
principal value of θ.

14
 5 Geometry

5.3 Plane figures and curves

A closed plane figure with straight sides is called a polygon. A polygon
with 3 sides is a triangle, one with 4 sides is a quadrilateral, one with
5 sides is a pentagon, one with 6 sides is a hexagon, and in general one
with n sides is called an n-gon.
A polygon is said to be regular if all its sides have equal length and all its
angles are equal. A regular triangle is referred to as an equilateral
triangle, and a regular quadrilateral is a square.
An isosceles triangle is one with two sides of equal length (or
equivalently with two equal angles). A right-angled triangle is one in
which one angle is a right angle. The angle sum of a triangle is
π radians (180◦ ).
A parallelogram is a quadrilateral with opposite sides parallel.
A rectangle is a parallelogram all of whose angles are right angles.
A square is a rectangle all of whose sides have equal length.
The angle sum of a quadrilateral is 2π radians (360◦ ).
1
The area of a triangle = 2 × base × height.
The area of a parallelogram = base × height.

height height

base base
The areas of a triangle and parallelogram

A circle is a set of points in a plane that are a constant distance from a

fixed point in the plane. The fixed point is the centre of the circle, and
the constant distance is its radius. If a straight line cuts a circle at two
points, then the segment of that straight line within the circle is known as
a chord of the circle. The length of a chord that passes through the centre
of a circle is the diameter of the circle. The terms diameter and radius
are also used to refer to a chord through the centre of a circle and to a
straight line from a point on the circle to its centre, respectively.

15
Handbook

radius arc

diameter

centre

angle subtended
chord by arc

Each continuous segment of a circle is known as an arc of the circle; the

angle made at the centre of a circle by two radii drawn from the ends of the
arc is known as the angle subtended by the arc. (In such circumstances
the arc itself is sometimes called the arc subtended by the angle.)
The distance around a circle is known as its circumference, and for a
circle of radius r this is given by 2πr. The area of a circle of radius r
is πr2 . The arc length of an arc of a circle of radius r subtended by an
angle θ radians is rθ. The area of a sector of a circle of radius r
subtended by an angle θ radians is 12 r2 θ.
The equation of a circle in the (x, y)-plane with centre (a, b) and
radius r is (x − a)2 + (y − b)2 = r2 . Other curves in the (x, y)-plane that
can be represented by quadratic formulas are the ellipse, the parabola
and the hyperbola, examples of which are shown below.

Ellipse Parabola Hyperbola

A straight line between two distinct points on a curve is known as a chord

of the curve. A straight line that just touches a curve is known as a
tangent to that curve at the point where it touches.

6 Linear algebra
6.1 Vectors
A (non-zero) vector x ∈ Rn (n integer, n ≥ 2) consists of:
• a positive real number |x|, called the magnitude or modulus or
length of the vector
• a direction in n-dimensional space (in the plane for two-dimensional
vectors).

16
 6 Linear algebra

The zero vector 0 has zero magnitude and no associated direction. A

unit vector is a vector with magnitude 1.
A vector x ∈ Rn can be written as
 
x1
 x2 
x =  .  = (x1 x2 . . . xn )T ,
 
 .. 
xn
where the real numbers x1 , x2 , . . . , xn are the elements or components
or coordinates of the vector, and T is the transpose operator. The
transpose notation is used to save space in printed documents; in
handwritten work, the column notation is preferred. The magnitude of
such a vector is given by
q
|x| = x21 + x22 + · · · + x2n ,
i.e. by the square root of the sum of the squares of its components. The
direction of such a vector is determined by its position in n-dimensional
space (Rn ) relative to the origin, and hence is determined by its
components.
Two vectors x, y ∈ Rn are equal if they have the same magnitude and
direction, i.e. if their components are equal.
The addition of two vectors x = (x1 x2 ... xn )T and
y = (y1 y2 . . . yn )T is defined by
x + y = (x1 + y1 x2 + y2 ... xn + yn )T .
The scalar multiplication of a vector x by a real number or scalar
k ∈ R is the operation defined by
kx = (kx1 kx2 ... kxn )T .
The magnitude of kx is |k||x|.
Scalar multiplication by zero gives the zero vector:
0x = (0 0 ... 0)T = 0.
Scalar multiplication of a non-zero vector x by k > 0 gives the vector kx
with the same direction as x but with magnitude k|x|.
Scalar multiplication of a non-zero vector x by −1 gives the vector of equal
magnitude but opposite direction:
−1x = (−x1 −x2 ... −xn )T = −x.
Scalar multiplication of any non-zero vector x by 1/|x| gives the unit
vector with the same direction as x.
The dot product or scalar product of two vectors x and y is the scalar
defined by
xT y = x · y = |x| |y| cos θ = y · x = yT x,

17
Handbook

where θ is the angle between the directions of x and y. If

x = (x1 x2 . . . xn )T and y = (y1 y2 . . . yn )T are two
n-dimensional vectors, then
xT y = x1 y1 + x2 y2 + · · · + xn yn ;
the formula
|x|2 = xT x = x21 + x22 + · · · + x2n
is a special case.
If xT y = 0 then either x or y is 0, or x is perpendicular to y. Two
non-zero vectors x and y are orthogonal (i.e. perpendicular) if and only if
xT y = 0.

6.2 Position vectors and the vector

equation of a line
The position vector x of a point X relative to the origin in
n-dimensional space is given by the coordinates x1 , x2 , . . . , xn of that point
and is written as x = (x1 x2 . . . xn )T .
The vector equation of a line through a point given by the position
vector x, parallel to the vector y, is
z = x + ty,
where t is a scalar, and any point on the line is given by z for a particular
value of t.
The vector equation of a line through the two points given by the position
vectors x and y is
z = x + t(y − x) = (1 − t)x + ty,
where t is a scalar, and any point on the line is given by z for a particular
value of t.

6.3 Matrices
The coefficients aij of a system of m linear equations in n variables
x1 , . . . , xn , given by
a11 x1 + a12 x2 + · · · + a1n xn = b1 ,
a21 x1 + a22 x2 + · · · + a2n xn = b2 ,
..
.
am1 x1 + am2 x2 + · · · + amn xn = bm ,

18
 6 Linear algebra

may be written as the elements or entries of a rectangular array, an

m × n matrix or coefficient matrix
 
a11 a12 . . . a1n
 a21 a22 . . . a2n 
A= . .. ..  = (aij ),
 
 .. . . 
am1 am2 . . . amn
where m and n refer to the number of rows and columns of the matrix,
respectively, and aij is the element in the ith row and jth column of the
matrix. The size or dimension or order of A is m × n.
A zero matrix is a matrix all of whose elements are zero; it is written
as 0. The scalar multiple kA of a matrix A by a real number or
scalar k has elements that are k times the corresponding elements of A,
i.e. kA = (kaij ). Two matrices are equal if they are of the same size and
their corresponding elements are equal.
If A = (aij ) and B = (bij ) are of the same size, then they may be
combined by matrix addition; the sum A + B is formed by adding the
corresponding elements of A and B, i.e. A + B = (aij + bij ).
The matrix product of an m × n matrix A = (aij ) and an n × r matrix
Pan m × r matrix AB whose element in the ith row and jth
B = (bij ) is
column is nk=1 aik bkj . In such a case we may say that the matrix A
operates on the matrix B by premultiplication to form the matrix
product AB. Matrix multiplication is not, in general, commutative,
i.e. in general BA 6= AB.
The transpose of a matrix A, written as AT , where T is the transpose
operator, is obtained by interchanging the rows and columns of A, so
that the element in the ith row and jth column of AT is aji . From this, we
can also deduce that
(AT )T = A, (A1 A2 . . . An−1 An )T = ATn ATn−1 . . . AT2 AT1 .
A square matrix is one with the same number of rows as columns; its
leading or main diagonal consists of elements of the form aii . An upper
triangular matrix is a square matrix in which all elements below the
leading diagonal are zero; a lower triangular matrix is a square matrix
in which all elements above the leading diagonal are zero; a diagonal
matrix is a square matrix in which all elements not on the leading
diagonal are zero. An identity matrix, written as I, is a square matrix
with ones on the leading diagonal and zeros everywhere else. For any
square matrix A and identity matrix I of the same size,
AI = IA = A.
A square matrix A is symmetric if AT = A.
The trace of a matrix A, written
Pn as tr A, is equal to the sum of the
diagonal entries, i.e. tr A = i=1 aii .

19
Handbook

6.4 Linear combinations, dependence and

independence
A linear combination of the variables x1 , x2 , . . . , xn is any expression of
the form
c1 x1 + c2 x2 + · · · + cn xn (ci ∈ R, i = 1, 2, . . . , n).
A linear combination of the real functions f1 (x), f2 (x), . . . , fn (x) with
the same domain is any function of the form
c1 f1 (x) + c2 f2 (x) + · · · + cn fn (x) (ci ∈ R, i = 1, 2, . . . , n).
A linear combination of the vectors x1 , x2 , . . . , xn is any vector of the
form
c1 x1 + c2 x2 + · · · + cn xn (ci ∈ R, i = 1, 2, . . . , n).
A finite set of vectors {x1 , x2 , . . . , xn } is linearly independent if
c1 x1 + c2 x2 + · · · + cn xn = 0
only when all the scalars ci (i = 1, 2, . . . , n) are zero. Otherwise the set is
said to be linearly dependent.
If {x1 , x2 , . . . , xn } is a set of linearly independent vectors in Rn , then any
vector x in Rn can be expressed as a linear combination of x1 , x2 , . . . , xn in
a unique way.

6.5 Singular matrices and matrix inverses

If, for a square matrix A, there exists a matrix B of the same size such
that AB = BA = I, then A is said to be an invertible or non-singular
matrix with inverse B. If A has no inverse, then it is said to be a
singular matrix. If A does have an inverse, then its inverse is unique and
is written as A−1 . Moreover, A−1 is non-singular and (A−1 )−1 = A. Also,
(A−1 )T = (AT )−1 .
Matrix inverses have the property that
−1 −1
(A1 A2 . . . An )−1 = A−1
n An−1 . . . A1 .

A matrix is invertible if and only if its rows are linearly independent.

Similarly, a matrix is invertible if and only if its columns are linearly
independent.
The inverse of the 2 × 2 matrix
 
a b
A=
c d
is
 
−1 1 d −b
A = ,
ad − bc −c a
provided that ad − bc 6= 0. If ad − bc = 0, then the matrix is singular.

20
 6 Linear algebra

6.6 Determinants
To every square matrix A there corresponds a number called its
determinant, written as det A or |A|. If A = (aij ) is a 2 × 2 matrix, then
 
a11 a12 
det A = 
  = a11 a22 − a12 a21 .
a21 a22 
If A = (aij ) is a 3 × 3 matrix, then
 
a11 a12 a13 
 
det A = a21 a22 a23 
a31 a32 a33 
     
a22 a23  a21 a23  a21 a22 
= a11    − a12 
  + a13 
 
a32 a33  a31 a33  a31 a32 
= a11 (a22 a33 − a23 a32 ) − a12 (a21 a33 − a23 a31 ) + a13 (a21 a32 − a22 a31 ).
The determinant of an n × n matrix is defined in terms of smaller
determinants in a similar way.
Determinants satisfy the following properties:
det I = 1,
det(kA) = k n det A (k ∈ R, for an n × n matrix A),
det(AB) = det A det B,
det(AT ) = det A,
det(A−1 ) = 1/ det A
The determinant of an upper triangular or lower triangular or diagonal
matrix is the product of the elements on its leading diagonal.
Linear combinations of rows (or columns) of a matrix can be added to any
other row (or column) without changing the value of its determinant.
Interchanging two rows of a matrix changes the sign of its determinant.
If x = x1 i + x2 j + x3 k and y = y1 i + y2 j + y3 k are vectors in R3 expressed
in a Cartesian coordinate system i, j, k then the cross product x × y can
be expressed as a 3 × 3 determinant:
 
i j k 

x × y = x1 x2 x3  ,
 y1 y2 y3 
= (x2 y3 − y2 x3 )i − (x1 y3 − y1 x3 )j + (x1 y2 − y1 x2 )k.

6.7 Eigenvalues and eigenvectors

An eigenvector of a square matrix A is a non-zero vector v such that
Av = λv
for some real number λ, where λ is called an eigenvalue of A. Any
non-zero scalar multiple of an eigenvector is also an eigenvector, so there is
an infinite number of eigenvectors of A corresponding to each eigenvalue.

21
Handbook

The eigenvalues of A may be found by solving the characteristic

equation
det(A − λI) = 0.
Eigenvectors can then be found by solving the eigenvector equations
(A − λI)v = 0 for v.
The eigenvalues of a diagonal, upper triangular or lower triangular matrix
are equal to the elements on its leading diagonal. If A is singular, then at
least one of its eigenvalues is zero.
If the characteristic equation for an n × n matrix has n different solutions
λ1 , λ2 , . . . , λn , then the eigenvalues are distinct and the matrix has n
linearly independent eigenvectors. If there are repeated eigenvalues, then
the matrix may or may not have n linearly independent eigenvectors. A
symmetric matrix always has n linearly independent eigenvectors. If v1
and v2 are eigenvectors corresponding to distinct eigenvalues λ1 and λ2 of
a symmetric matrix A, then v1 and v2 are orthogonal vectors. The
eigenvalues of a real symmetric matrix are always real.

7 Differentiation
7.1 Notation and terminology
If f (x) is a function, then its derived function or derivative f 0 (x) is
defined by
f (x + h) − f (x)
f 0 (x) = lim .
h→0 h
The process of calculating f 0 (x) from f (x) is called differentiation of
f (x) with respect to x. Differentiation with respect to x can also be
d
denoted by the symbol written to the left of the expression or variable
dx
being differentiated, so that, for example,
d df (x)
(f (x)) and both mean f 0 (x).
dx dx
dy
If y = f (x), then f 0 (x) can also be written as y 0 or , where to save space
dx
dy
we often print dy/dx in place of .
dx
When the independent variable is t (time), we often use a dot to indicate a
derivative, so that u̇ means the same thing as u0 (t) or du/dt. If x is the
position of a particle along an axis, then dx/dt or ẋ represents the
velocity along the axis, and d2 x/dt2 or ẍ represents the acceleration
along the axis.
The notation f 0 (x) is referred to as function notation, dy/dx as Leibniz
notation, and u̇ as Newton’s or Newtonian notation.

22
 7 Differentiation

The derivative of a derivative is called a second derivative. For example,

the second derivative of the function f (x), denoted by f 00 (x), is the
derivative of f 0 (x). If y = f (x), then the second derivative is also written
as y 00 or d2 y/dx2 . If u is a function of t, then its second derivative can be
written as ü.
Third and higher derivatives are defined analogously. The nth
derivative of f is denoted by f (n) or, if y = f (x), by dn y/dxn ; n is referred
to as the order of the derivative. The prime and dot notations are not
used for higher derivatives, except that f 000 is sometimes used in place
of f (3) .
A complex-valued function f (x) = g(x) + i h(x), where g and h are
real-valued functions, can be differentiated in a natural way as
f 0 (x) = g 0 (x) + i h0 (x).

7.2 Rules of differentiation

Constant multiple rule: If k is a constant and u is a function of x, then
d du
(ku)0 = ku0 , or equivalently
(ku) = k .
dx dx
Sum rule: If u and v are functions of x, then
d du dv
(u + v)0 = u0 + v 0 , (u + v) =
or equivalently + .
dx dx dx
Product and quotient rules: If u and v are functions of x, then
d du dv
(uv)0 = u0 v + uv 0 , or equivalently (uv) = v+u ,
dx dx dx
and
u0 v − uv 0
 u 0  .
d u du dv
= , or equivalently = v−u v2.
v v2 dx v dx dx
Composite rule or chain rule: If g and u are two functions, and This is sometimes called the
h(x) = g(u(x)), then ‘function of a function’ rule.

h0 (x) = g 0 (u(x)) u0 (x).

Another way to write this is as
dh dh du
= .
dx du dx
Implicit differentiation: Given an equation connecting two variables x
and y, we can use implicit differentiation to calculate dy/dx by
differentiating both sides with respect to x and then solving the resulting
equation algebraically for dy/dx (instead of solving for y before
differentiating). For example,
d 3 2 dy
(x y ) = 3x2 y 2 + 2x3 y .
dx dx

23
Handbook

7.3 Partial derivatives

The partial derivatives of a function f (x, y) with respect to x and y may
be defined as
∂f f (x + δx, y) − f (x, y)
fx (x, y) = = lim ,
∂x δx→0 δx
∂f f (x, y + δy) − f (x, y)
fy (x, y) = = lim ,
∂y δy→0 δy
where δx and δy denote (small) increments in the values of x and y,
respectively.
The rate of change ofz = f (x, y) with respect to t along a parametrised
curve (x(t), y(t)) is given by the following version of the chain rule.
dz ∂z dx ∂z dy
= + .
dt ∂x dt ∂y dt

7.4 Standard derivatives

In each case, a is a constant.

Function Derivative
a 0
xa axa−1
eax aeax
1
ln(ax)
x
sin(ax) a cos(ax)
cos(ax) −a sin(ax)
tan(ax) a sec2 (ax)
cot(ax) −a cosec2 (ax)
sec(ax) a sec(ax) tan(ax)
cosec(ax) −a cosec(ax) cot(ax)
a
arcsin(ax) √
1 − a2 x2
a
arccos(ax) −√
1 − a2 x2
a
arctan(ax)
1 + a2 x2
a
arccot(ax) −
1 + a2 x2
a
arcsec(ax) √
|ax| a2 x2 − 1
a
arccosec(ax) − √
|ax| a2 x2 − 1

24
 7 Differentiation

7.5 Stationary points

The gradient of a function f at a point x0 is the slope of the tangent to
the graph of f at that point, and is given by the derivative of f at that
point, i.e. f 0 (x0 ). A function is increasing on an interval if its gradient is
positive throughout that interval; it is decreasing if its gradient is
negative throughout that interval. A stationary point of f is a point x0
where the gradient is zero, i.e. f 0 (x0 ) = 0.
A function is smooth if it is continuous and has a continuous derivative.
Consider a stationary point x0 of a smooth function f .
• x0 is a local maximum if, for all x in the immediate vicinity of x0 ,
f 0 (x) > 0 if x < x0 and f 0 (x) < 0 if x > x0 . An alternative condition is
f 00 (x0 ) < 0.
• x0 is a local minimum if, for all x in the immediate vicinity of x0 ,
f 0 (x) < 0 if x < x0 and f 0 (x) > 0 if x > x0 . An alternative condition is
f 00 (x0 ) > 0.
• x0 is a point of inflection if f 00 (x0 ) = 0 and f 00 (x) changes sign as x
increases through x0 .
A global maximum of a function f is a point x0 such that f (x0 ) ≥ f (x)
for all x where f is defined. A global minimum of a function f is a
point x0 such that f (x0 ) ≤ f (x) for all x where f is defined. A function f
is bounded above by an upper bound A if f (x) ≤ A for all x where f
is defined. A function f is bounded below by a lower bound B if
f (x) ≥ B for all x where f is defined.

7.6 Curve sketching

The following is a possible procedure for sketching the graph of y = f (x),
where f (x) is some given function.
(a) Check whether f (x) is a standard function whose graph you already
know, or is a simple modification of such a function. If not, proceed
to Step 2.
(b) Determine how y behaves when x is very large and positive, and when
x is very large and negative.
(c) Look for any obvious symmetries or repetitions in the behaviour of f .
(d) Find where the curve crosses the x- and y-axes, if at all.
(e) Look for any values of x at which f (x) is undefined, and examine the
behaviour of f (x) near these values of x.
(f) Find the locations of any local maxima, local minima or points of
inflection.
(g) Try to determine whether there are any intervals over which the
function is increasing or decreasing.

25
Handbook

(h) Transfer the information found in Steps 4 and 6 to a sketch graph,

then use this information together with any information found in
Steps 2, 3, 5 and 7 to try to sketch a smooth curve. If you are still
unsure about any parts of the curve, choose suitable values of x and
plot the corresponding points (x, f (x)) before completing the curve.
Lines y = c where f (x) → c as x → ±∞, and lines x = c where
f (x) → ±∞ as x → c, where c is a constant, are known as asymptotes of
the graph of the function f (x).

7.7 Taylor polynomials and series

Factorials
For any positive integer n, we define n factorial, written n!, by
n! = 1 × 2 × 3 × · · · × (n − 1) × n.
The first few factorials are 1! = 1, 2! = 2, 3! = 6, 4! = 24. We also define
0! = 1.

Taylor polynomials
For a function f (x) with n continuous derivatives near x = a, the Taylor
polynomial of degree n about x = a or the nth-order Taylor
polynomial about x = a is
1 1
pn (x) = f (a) + (x − a)f 0 (a) +
(x − a)2 f 00 (a) + · · · + (x − a)n f (n) (a).
2! n!
When used to approximate f (x) near x = a, we refer to this polynomial as
the nth-order Taylor approximation to f (x) near x = a, and write
1 1
f (x) ' f (a) + (x − a)f 0 (a) + (x − a)2 f 00 (a) + · · · + (x − a)n f (n) (a).
2! n!
In particular, n = 1 gives the tangent approximation
f (x) ' f (a) + (x − a)f 0 (a),
and n = 2 gives the quadratic approximation
f (x) ' f (a) + (x − a)f 0 (a) + 12 (x − a)2 f 00 (a).
These approximations are good when x is close to a.

Taylor series
The Taylor series about x = a for a function f (x) with infinitely many
continuous derivatives near x = a is
1
f (x) = f (a) + (x − a)f 0 (a) + (x − a)2 f 00 (a) + · · ·
2!
1 n (n)
+ (x − a) f (a) + · · ·
n!
∞
X 1
= (x − a)n f (n) (a).
n!
n=0

26
 7 Differentiation

Some standard Taylor series about x = 0

1 3 1 1
sin x = x − x + x5 − · · · + (−1)n−1 x2n−1 + · · · ,
3! 5! (2n − 1)!
1 1 1
cos x = 1 − x2 + x4 − · · · + (−1)n x2n + · · · ,
2! 4! (2n)!
1 1
ex = 1 + x + x2 + · · · + xn + · · · ,
2! n!
1 1 1
ln(1 + x) = x − x2 + x3 − · · · + (−1)n−1 xn + · · · (−1 < x < 1).
2 3 n

Small-angle approximations
If the angle θ is small (sometimes written θ  1) and is measured in
radians, then we can obtain useful approximations by truncating the above
Taylor series:
sin θ ' θ,
cos θ ' 1 − 21 θ2 .

7.8 Surfaces
A stationary point of a function f (x, y) is a point (a, b) at which
fx (a, b) = fy (a, b) = 0.
A function f (x, y) has a local minimum at (a, b) if for all (x, y)
sufficiently close to (a, b), we have f (x, y) ≥ f (a, b).
A function f (x, y) has a local maximum at (a, b) if for all (x, y)
sufficiently close to (a, b), we have f (x, y) ≤ f (a, b).
A point that is either a local maximum or a local minimum is an
extremum.
A stationary point that is not an extremum is a saddle point.
A stationary point can be classified using the AC − B 2 test as follows.
Suppose that the function f (x, y) has a stationary point at (a, b). Let
A = fxx (a, b), B = fxy (a, b), C = fyy (a, b).
(a) If AC − B2 > 0, then there is:
• a local minimum at (a, b) if A > 0
• a local maximum at (a, b) if A < 0.
(b) If AC − B 2 < 0, then there is a saddle point at (a, b).
(c) If AC − B 2 = 0, then the test is unable to classify the stationary point.

27
Handbook

8 Integration
8.1 Notation and terminology
The indefinite integral of a continuous function f (x) is
Z
f (x) dx = F (x) + C,

where F is a function such that F 0 (x) = f (x), known as an integral or

antiderivative of f , and C is a constant, often referred to as an
arbitrary constant or constant of integration.
The definite integral of a continuous function f (x) from a to b is
Z b
 b
f (x) dx = F (x) a = F (b) − F (a),
a
where F is any integral of f . The numbers a and b are called the lower
limit of integration and upper limit of integration, respectively. If
the areas bounded by the graph of f (x) above and below the x-axis
between a and b are A1 and A2 , respectively, then
Z b
f (x) dx = A1 − A2 .
a
The process of finding an indefinite or definite integral is known as
integration, and the function f being integrated is known as the
integrand. If F1 and F2 are two integrals of f , then they differ by a
constant, i.e. F1 (x) = F2 (x) + C, where C is a constant.

8.2 Rules of integration

Constant multiple rule:
Z Z
k f (x) dx = k f (x) dx (where k is a constant).

Sum rule:
Z Z Z


f (x) + g(x) dx = f (x) dx + g(x) dx.

Integration by substitution:
Z Z

0
f g(x) g (x) dx = f (u) du (where u = g(x)),

or in Leibniz notation
du
Z Z
f (u) dx = f (u) du.
dx
The following formula, which can be derived by integration by
substitution, is also useful:
Z 0
g (x)
dx = ln |g(x)| + C (where g(x) 6= 0).
g(x)

28
 8 Integration

Integration by parts:
Z Z
f (x) g 0 (x) dx = f (x) g(x) − f 0 (x) g(x) dx.

For definite integrals,

Z b Z b
0
b
f 0 (x) g(x) dx.

f (x) g (x) dx = f (x) g(x) a −
a a

A function f is odd if f (−x) = −f (x). If f is odd and a is a positive

constant, then
Z a
f (x) dx = 0.
−a

8.3 Standard integrals

In each case, a is a non-zero constant, b is any constant, and n is any
integer. When using the table below to obtain indefinite integrals, add an
arbitrary constant.

29
Handbook

Function Integral
a ax
xa+1
xa (a 6= −1)
a+1
1 1
ln |ax + b|
ax + b a
1 ax
eax e
a
ln(ax) x(ln(ax) − 1)
1
sin(ax) − cos(ax)
a
1
cos(ax) sin(ax)
a
1
tan(ax) − ln |cos(ax)|
a
1
cot(ax) ln |sin(ax)|
a
1
sec(ax) ln |sec(ax) + tan(ax)|
a
1
cosec(ax) ln |cosec(ax) − cot(ax)|
a
1
sec2 (ax) tan(ax)
a
1
cosec2 (ax) − cot(ax)
a
1

x sin(ax) 2
sin(ax) − ax cos(ax)
a
1

x cos(ax) 2
cos(ax) + ax sin(ax)
a
1 1 x
arctan
x2 + a2 a a
 
1 1 a − x
ln  
(x − a)(x − b) a − b x − b
1 √
√ ln(x + x2 + a2 ) = arcsinh (x/a)
x + a2
2

1 √
√ ln |x + x2 − a2 | = arccosh (x/a)
x − a2
2

1 x
√ arcsin
a2 − x2 a

30
 9 Coordinate systems

9 Coordinate systems
9.1 Cartesian coordinates
In Cartesian coordinates the gradient operator is given by
∂ ∂ ∂
∇=i +j +k ,
∂x ∂y ∂z
where i, j, k are unit vectors in the x-, y-, z-directions.
The divergence of a vector field quantifies outflow (or inflow if negative)
of an infinitesimal neighbourhood of a point. The divergence of a vector
field F = F1 i + F2 j + F3 k is the scalar field
∂F1 ∂F2 ∂F3
div F = ∇ · F = + + ,
∂x ∂y ∂z
where x, y and z are Cartesian coordinates.
The Laplacian operator. is defined to be the divergence of the gradient
∇2 = ∇ · ∇.
In Cartesian coordinates the Laplacian operator can be expressed as
∂2 ∂2 ∂2
∇2 = + +
∂x2 ∂y 2 ∂z 2

9.2 Polar coordinates

Polar coordinates (r, θ) are related to Cartesian coordinates by the
equations x = r cos θ, y = r sin θ.
In polar coordinates the gradient operator is given by
∂ 1 ∂
∇ = er + eθ ,
∂r r ∂θ
where er , eθ are unit vectors in the r-, θ-directions.
The divergence of a vector field F(r, θ, z) = Fr er + Fθ eθ in polar
coordinates is
 
1 ∂ 1 ∂Fθ
div F = ∇ · F = rFr + .
r ∂r r ∂θ
In polar coordinates (r, θ), the Laplacian operator is
1 ∂2
 
2 1 ∂ ∂
∇ = r + 2 .
r ∂r ∂r r ∂θ2

31
Handbook

9.3 Cylindrical coordinates

Any point P can be represented in cylindrical coordinates by the triple
(r, θ, z), where z is the distance of P from the (x, y)-plane, and (r, θ) are
the polar coordinates of the projection N of P onto the (x, y)-plane (see
the figure).
Cylindrical coordinates are related to Cartesian coordinates (x, y, z) by
x = r cos θ, y = r sin θ, z = z,
r = (x2 + y 2 )1/2 , sin θ = y/r, cos θ = x/r.
We require that
r ≥ 0, −π < θ ≤ π, z ∈ R.
The value of θ for points on the z-axis (r = 0) is undefined. By convention,
we put θ = 0 for such points.
The gradient operator is given by
∂ 1 ∂ ∂
∇ = er + eθ + ez ,
∂r r ∂θ ∂z
where er , eθ , ez are unit vectors in the r-, θ-, z-directions.
The divergence of a vector field F(r, θ, z) = Fr er + Fθ eθ + Fz ez in
cylindrical coordinates is
∂Fr 1 1 ∂Fθ ∂Fz
div F = ∇ · F = + Fr + + .
∂r r r ∂θ ∂z

9.4 Spherical coordinates

Any point P can be represented in spherical coordinates by the triple
(r, θ, φ), where r is the distance of P from the origin, and θ and φ are the
polar and azimuthal angles, respectively. In the figure, N is the projection
of P onto the (x, y)-plane.
Spherical coordinates are related to Cartesian coordinates (x, y, z) by
x = ON cos φ = r sin θ cos φ, y = ON sin φ = r sin θ sin φ,
p
z = r cos θ, r = x2 + y 2 + z 2 .
We require that
r ≥ 0, −π < φ ≤ π, 0 ≤ θ ≤ π.
The gradient operator is given by
∂ 1 ∂ 1 ∂
∇ = er + eθ + eφ ,
∂r r ∂θ r sin θ ∂φ
where er , eθ , eφ are unit vectors in the r-, θ-, φ-directions.
The divergence of a vector field F(r, θ, φ) = Fr er + Fθ eθ + Fφ eφ in
spherical coordinates is
   
∂Fr 1 ∂Fθ 1 ∂Fφ
div F = ∇ · F = + + 2Fr + + Fθ cos θ .
∂r r ∂θ r sin θ ∂φ

32
 10 Area and volume integrals

10 Area and volume integrals

The area integral in Cartesian coordinates of a function f (x, y) over
a region S contained between the lines x = a and x = b together with the
curves y = α(x) and y = β(x) is obtained as two successive integrals as
follows
Z Z x=b Z y=β(x) !
f (x, y) dA = f (x, y) dy dx.
S x=a y=α(x)

In polar coordinates (r, θ) there is an extra factor of r in the integrand,

which is called the Jacobian factor. So in polar coordinates the integral
of a function f (r, θ) over a disc D of radius a is
Z Z θ=π Z r=a 
f (r, θ) dA = f (r, θ) r dr dθ.
D θ=−π r=0

In cylindrical coordinates (r, θ, z) the Jacobian factor is also r. So the

volume integral in cylindrical coordinates of a function f (r, θ, z) over
a cylinder B of radius a lying between the planes z = 0 and z = h is given
by
Z Z r=a Z θ=π Z z=h  
f (r, θ, z) dV = f (r, θ, z) r dz dθ dr.
B r=0 θ=−π z=0

In spherical coordinates (r, θ, φ) the Jacobian factor is r2 sin θ. So the

volume integral in spherical coordinates of a function f (r, θ, φ) over
a thick spherical shell with inner radius R1 and outer radius R2 is given by
Z Z r=R2 Z θ=π Z φ=π  
2
f (r, θ, φ) dV = f (r, θ, φ) r sin θ dφ dθ dr.
B r=R1 θ=0 φ=−π

33
Handbook

Unit summaries
Unit 1 Introduction and two-dimensional plane the phase space or phase
plane for the system, and the point is called the
differential equations phase point.
1 The logistic equation is the first–order We can think of x and y as the coordinates of the
differential equation position vector x = (x y)T of the phase point.
The evolution of the position and velocity can then
dx R0
= (x0 − x)x. be represented by a path, called a phase path or
dt x0
phase trajectory in the phase space.
2 An equilibrium point or fixed point of a The dynamical system equations can be written in
differential equation a very simple form using vector notation:
ẋ = F (x)
ẋ = u(x),
is a point xe such that x(t) = xe is a constant
where u(x) = (u(x, y) v(x, y))T . This is an
solution of the system of differential equations,
equation of motion for the phase point: the
that is, xe is a point at which ẋ(t) = 0. To find the
velocity ẋ of the phase point is equal to u(x), that
equilibrium points xe , solve the equation
is, to the value of a vector field u evaluated at the
F (xe ) = 0.
position of the phase point, x = (x, y). It can be
3 Suppose that the differential equation shown that the phase path at (x, y) is tangential to
ẋ = F (x) the vector field u(x, y).
6 A constant of motion is a function K(x, y)
has an equilibrium point at x = xe . The
that remains constant as we follow any given phase
equilibrium point is said to be:
path.
• stable if all points arbitrarily close to the
7 To test whether the differentiable function
equilibrium point remain in the neighbourhood
K(x, y) is a constant of motion of the dynamical
of the equilibrium point as time increases
system ẋ = u(x, y), ẏ = v(x, y), calculate
• unstable otherwise.
dK ∂K ∂K
4 To check the stability of fixed points of
dt
=
∂x
u+
∂y
v.
one-dimensional motion of ẋ = F (x), do the
following. K(x, y) is a constant of motion if and only if
dK/dt = 0 for all points (x, y) in the phase space
(a) Find the equilibrium points xe by solving
of the system.
F (xe ) = 0.
(b) Each fixed point xe is classified as 8 To solve the differential equation
• stable if F 0 (xe ) < 0 dy
= g(x) h(y), where h(y) 6= 0,
• unstable if F 0 (xe ) > 0, dx
where F 0 (x) = dF/dx. using separation of variables, do the following.
5 Consider the two-variable dynamical system (a) Divide both sides of the equation by h(y), and
given by first-order differential equations integrate with respect to x, to obtain
1
Z Z
ẋ = u(x, y), ẏ = v(x, y), dy = g(x) dx.
h(y)
where u(x, y) and v(x, y) are two functions. (b) If possible, perform the two integrations,
At any given time, we represent this system by a obtaining an implicit form of the general solution,
point in the (x, y)-plane. We call this which should include one arbitrary constant.

34
 Unit 1 summary

(c) If possible, rearrange the formula found in is given by

Step (b) to give y in terms of x; this is the explicit
y = yc + yp ,
general solution of the differential equation.
9 To solve the linear second-order where:
constant-coefficient homogeneous differential • yc , the complementary function, is the
equation general solution of the associated
homogeneous equation
d2 y dy
a 2
+b + cy = 0, d2 y dy
dx dx a +b + cy = 0
dx 2 dx
first solve the auxiliary equation • yp , a particular integral, is any particular
aλ2 + bλ + c = 0 solution of the original inhomogeneous equation.
This leads to the following procedure for finding
to determine λ. Because this is a quadratic
the general solution of an inhomogeneous equation.
equation, there are two solutions. There are three
cases to consider depending on these solutions. (In (a) Find the complementary function, by solving
each case A and B are arbitrary constants.) the auxiliary equation of the associated
homogeneous equation.
• If the auxiliary equation has two distinct real
(b) Find a particular integral, using the method
roots λ1 and λ2 , then the general solution of the
of undetermined coefficients.
differential equation is
The method involves using a trial solution y(x)
y(x) = Aeλ1 x + Beλ2 x . of a form similar to that of f (x) (see the table
• If the auxiliary equation has complex roots, then below), with coefficients initially undetermined.
they will be complex conjugates of each other: Substituting the trial solution into the differential
λ1 = α + iβ and λ2 = α − iβ, where α and β are equation gives equations that determine the
real numbers. The general solution is coefficients.
y(x) = exp(αx) [A cos(βx) + B sin(βx)]. Target function f (x) Trial solution y(x)
• If the auxiliary equation has equal real roots a p
λ1 = λ2 , then the general solution is a0 + a1 x p0 + p1 x
y(x) = (A + Bx)eλ1 x . aekx pekx
aeikx peikx
10 The principle of superposition states that a cos(kx) + b sin(kx) p cos(kx) + q sin(kx)
if y1 (x) is a solution of the inhomogeneous
equation ay 00 + by 0 + cy = f1 (x), and y2 (x) is a (c) Add the particular integral to the
solution of ay 00 + by 0 + cy = f2 (x), then for any complementary function.
constants k1 and k2 , 12 To solve the equation
y(x) = k1 y1 (x) + k2 y2 (x) dy
+ g(x) y = h(x)
dx
is a solution of the inhomogeneous equation
by the integrating factor method, do the
d2 y dy following.
a 2
+b + cy = k1 f1 (x) + k2 f2 (x).
dx dx (a) Determine the integrating factor
11 The general solution of the second-order
Z 
p(x) = exp g(x) dx .
inhomogeneous linear constant-coefficient
equation (b) Rewrite the equation as
d2 y dy d

a +b + cy = f (x) p(x) y = p(x) h(x).
dx2 dx dx

35
Handbook

(c) Integrate to obtain 3 A normal mode of oscillation is a solution

in which all of the coordinates of the system
Z
p(x) y = p(x) h(x) dx + C, oscillate sinusoidally with the same angular
where C is a constant of integration. frequency.
(d) Divide through by p(x), to obtain the general 4 For a coupled oscillator attached to some fixed
solution in explicit form. point, there are two normal modes, where both
masses oscillate sinusoidally with the same angular
frequency. These are given by
Unit 2 Oscillations of linear
x(t) = Ai cos(ωi t + φi )vi , for i = 1, 2.
systems
The displacements of the in-phase normal mode
1 The motion of a simple harmonic are always in the same direction, while the
oscillator is described by a single function of time displacements of the phase-opposed normal
x(t) satisfying the equation of motion mode are always in opposing directions.
d2 x The frequency of oscillation of the in-phase mode
+ ω 2 x = 0, is always smaller than that of the phase-opposed
dt2
for some real constant ω. The motion x(t), called mode.
simple harmonic motion, is obtained by solving 5 The Lissajous curve for a coupled oscillator
the differential equation, giving with two particles is obtained by plotting the path
x(t) = (x(t) y(t))T in the plane.
x(t) = A cos(ωt + φ),
If the ratio of the normal mode frequencies ω1 /ω2
or alternatively is rational, then the curve is closed and the motion
x(t) = C cos(ωt) + D sin(ωt), is periodic. Otherwise, if ω1 /ω2 is irrational, then
the curve never traces back on itself and the
where A, C, D and φ are arbitrary constants. motion is not periodic.
The parameter ω is called the angular frequency, 6 The differential equations describing a
A is called the amplitude of the oscillation (and is conservative system of n coupled linear oscillators
normally positive), and φ is called the phase can always be written in the form ẍ = Ax, where
constant. C and D have no special names. A is an n × n matrix. To solve a general linear
2 The equations of motion of a coupled oscillatory system, do the following.
oscillator with two masses can be written in the (a) Find the n eigenvalues λ1 , λ2 , . . . , λn of A, and
matrix form their corresponding eigenvectors v1 , v2 , . . . , vn .
ẍ = Ax, (b) Each negative eigenvalue λ and its
corresponding eigenvector v give rise to a solution
where A is a 2 × 2 matrix. √
x(t) = Av cos( −λt + φ),
If the system is attached to some fixed point, then
where A and φ are arbitrary constants.
the most general motion of the system is described
by (Or equivalently
√ √



x(t) = v C cos( −λt) + D sin( −λt) ,
x(t) = A1 cos(ω1 t + φ1 ) v1 + A2 cos(ω2 t + φ2 ) v2 ,
where C and D are arbitrary constants.)
where These are the normal modes of oscillation.
p
ωi = −λi , i = 1, 2. Each zero eigenvalue and its corresponding
eigenvector v give rise to a solution
Here λ1 and λ2 are the eigenvalues of A, and v1
and v2 are the corresponding eigenvectors; A1 , A2 , x(t) = v(A + Bt),
φ1 and φ2 are arbitrary constants. where A and B are arbitrary constants.

36
 Unit 2 summary

(c) The general solution is a linear combination of 10 If a continuous periodic function f (t) is
the n solutions found in Step (b), involving sufficiently smooth and has fundamental period τ ,
2n arbitrary constants. then its derivative f 0 (t) has the same fundamental
period τ , and its Fourier series is obtained by
7 Consider a system of n oscillators with
differentiating the Fourier series of f term by term.
equation of motion ẍ = Ax that has identical
masses and no zero eigenvalues. To find the 11 Consider a linear differential equation
solution subject to a set of initial conditions,
d2
 
do the following. d
a0 + a1 + a2 2 + · · · x(t) = f (t),
dt dt
(a) Calculate the n eigenvectors vi and the
corresponding eigenvalues λi (i = 1, 2, . . . , n) of the where f (t) is a periodic function with period τ .
coefficient matrix A.
To find a particular solution using Fourier
(b) Put the initial conditions into vector form series, do the following.
x(0) = x0 and ẋ(0) = u0 , (a) Calculate the Fourier coefficients fn and the
where x0 and u0 are n-dimensional vectors. resulting Fourier series of f :
(c) Calculate the n numbers x0 · vi , the n numbers ∞
X 2π
u0 · vi and the n numbers vi · vi (i = 1, 2, . . . , n). f (t) = fn einωt , where ω = .
n=−∞
τ
(d) The solution satisfying the initial conditions is
then given by (b) Assume that x(t) is also a periodic function
Xn 
x 0 · vi
 with period τ , and that its Fourier series is
x(t) = cos(ωi t)+ ∞
vi · vi X
i=1
   x(t) = xn einωt ,
u0 · v i n=−∞
sin(ωi t) vi ,
ωi vi · vi where the xn are to be determined.
√
where ωi = −λi . (c) Calculate the Fourier series of the derivatives
8 The complex exponential functions exp(int) of x(t):
for integer n satisfy the orthogonality relations ∞
X
Z π ẋ(t) = (inω)xn einωt ,
eint e−imt dt = 2πδnm for any integers n and m, n=−∞
−π ∞
X
where the Kronecker delta is defined by ẍ(t) = (inω)2 xn einωt ,
( n=−∞
1 if n = m, and so on.
δnm =
0 if n 6= m. (d) Substitute the Fourier series for f , and for
x, ẋ, ẍ, . . ., into the differential equation, giving
9 For a periodic function f (t) with fundamental ∞
period τ , the exponential Fourier series is
X
a0 + a1 (inω) + a2 (inω)2 + · · · xn einωt


∞
X n=−∞
f (t) = cn ei2nπt/τ . ∞
X
n=−∞ = fn einωt .
n=−∞
The Fourier coefficients cn are calculated from
(e) Equate the coefficients of einωt on both sides,
1 τ /2
Z
and solve the resulting algebraic equation for xn ,
cn = f (t) e−i2nπt/τ dt (n = 0, ±1, ±2, . . .).
τ −τ /2 giving
The Fourier series defines a real function if fn
xn = .
cn = c−n . a0 + a1 (inω) + a2 (inω)2 + · · ·

37
Handbook

12 The equation of motion of a forced damped the fixed point is stable; whereas if all orbits
harmonic oscillator is given by (apart from the one starting at xe ) flow away
from xe , then the fixed point is unstable. In cases
d2 x dx
2
+ 2Γ + ω02 x = f (t). where f 0 (xe ) 6= 0, the following rule can be applied:
dt dt
the fixed point is stable if f 0 (xe ) < 0 and unstable
It describes a single oscillator (of mass m) subject if f 0 (xe ) > 0.
to a damping force (−2mΓẋ) and an applied
external driving force (m f (t)).
2 The evolution of two interacting populations
X and Y can be modelled by the Lotka–Volterra
Here Γ is a constant called the damping equations
parameter, and ω0 is the angular frequency of the
oscillator in the absence of damping or external Ẋ = kX − AXY, Ẏ = −hY + BXY (X ≥ 0, Y ≥ 0),
forces, called the natural angular frequency. where X is the population of the prey, Y is the
The solution of this equation is the sum of a population of the predator, and k, h, A and B are
complementary function xc (t) (sometimes known positive constants.
as the transient) and a particular integral xp (t). The Lotka–Volterra equations are nonlinear
If Γ 6= 0, then for large t the transient dies away because their right-hand sides contain a term
and the solution is given by xp . proportional to XY .
If the driving term f (t) is periodic in t, with The rescaled Lotka–Volterra equations are
period τ , then xp can be found using the method ẋ = kx(1 − y), ẏ = −hy(1 − x) (x ≥ 0, y ≥ 0),
described in item 11 above. In particular, if f has
Fourier series f (t) = ∞ inωt (with where the rescaled variables x and y are related to
P
n=−∞ fn e
ω = 2π/τ ), then xp (t) has Fourier series X and Y via
representation given by X = hx/B, Y = ky/A.
∞
xp (t) =
X fn
einωt .
3 An equilibrium point (also called a fixed
n=−∞
(ω02 − n ω2)
2 + 2Γinω point) of a system of differential equations
ẋ = u(x, y) is a point (xe , ye ) where u(xe , ye ) = 0,
For small values of the damping parameter, xp can that is, a point where the phase velocity vanishes.
be very large if ω0 ' nω, that is, whenever the
natural angular frequency is some integer multiple 4 To find the equilibrium points of the
of the angular frequency of the driving force. system of differential equations ẋ = u(x, y), for
some vector field u, solve the equation
u(x, y) = 0
Unit 3 Nonlinear dynamics
for x and y.
1 The phase flow of the autonomous
one-dimensional system 5 Suppose that the system of differential
equations
ẋ = f (x), x(0) = x0 ,  
u(x, y)
ẋ = u(x) =
can be determined by examining the function f (x) v(x, y)
without obtaining an explicit expression for the has an equilibrium point at (xe , ye ). To linearise
solution x(t). this system of differential equations, do the
The trajectory x(t) for t > 0, from the initial following.
value x0 , is called the orbit of x0 . Orbits must (a) Determine the Jacobian matrix
flow towards positive x in regions where f (x) > 0, 
∂u ∂u

and towards negative x in regions where f (x) < 0.  ∂x (x, y) ∂y (x, y)
If in the vicinity of a fixed point xe (i.e. a point J(x, y) = 
 ∂v
.
∂v 
where f (xe ) = 0) all orbits flow towards xe , then (x, y) (x, y)
∂x ∂y

38
 Unit 3 summary

(b) Near the equilibrium point (xe , ye ), the (b) Use item 5 to determine the linear system
differential equations can be approximated by the ṗ = Ap
linear system
    that approximates the original nonlinear system
ṗ p when close to each equilibrium point.
= J(xe , ye ) ,
q̇ q (c) For each equilibrium point, use item 7 to
where x(t) = xe + p(t) and y(t) = ye + q(t). classify the linear system.
6 Consider the system ẋ = u(x) with an The behaviour of the original nonlinear system
equilibrium point at xe . near an equilibrium point is the same as that of
the linear approximation, except when the linear
The equilibrium point is stable if x(t) stays
system has a centre. If the linear system has a
arbitrarily close to xe for all t > 0 whenever x(0) is
centre, then the equilibrium point of the original
sufficiently close to xe . (A more mathematical way
nonlinear system may be a centre (neutrally
of expressing this is to state that xe is stable if for
stable), a stable spiral (asymptotically stable) or a
all ε > 0 there exists a δ > 0 such that
unstable spiral.
|x(t) − xe | < ε for all t > 0 whenever
|x(0) − xe | < δ.) 9 Consider the system of differential equations

If, in addition, x(t) → xe as t → ∞, then the ẋ = u(x, y), ẏ = v(x, y).

equilibrium point is said to be asymptotically To find centres using constants of motion, do
stable; otherwise it is non-asymptotically the following.
stable (or neutrally stable).
(a) Look for a constant of motion of the system
If the equilibrium point is not stable, then we say by attempting to find the general solution of the
that it is unstable. differential equation
7 Consider the linear system ṗ = Ap, for a dy
=
v(x, y)
,
2 × 2 matrix A. The eigenvalues of A are dx u(x, y)
determined from det A and tr A via the formula and writing the arbitrary constant K as a function
 p  of x and y. This function K(x, y) is a constant of
λ = 12 tr A ± (tr A)2 − 4 det A .
motion.
The nature of the equilibrium point at p = 0 is (b) Either use item 4 to find the fixed points of
determined by these eigenvalues according to the the system, or solve
table below. ∂K ∂K
= =0
∂x ∂y
Eigenvalues real λ1 > λ2 > 0 Unstable node
to find the stationary points of K(x, y).
λ1 , λ2 ∈ R λ1 < λ2 < 0 Stable node
λ1 > 0 > λ2 Saddle point (c) Check whether any of the points found in
Step (b) are extrema of K(x, y). If so, then these
Eigenvalues ν>0 Unstable extrema are centres of the system of differential
complex spiral equations.
λ1,2 = ν ± iω ν<0 Stable spiral 10 A limit cycle is an isolated closed orbit, that
ν, ω ∈ R, ω 6= 0 ν=0 Centre is, an orbit closed in phase space and therefore
8 To classify the equilibrium points of the periodic in time. It is isolated in the sense that
nonlinear system of differential equations neighbouring orbits (i.e. all orbits in some
neighbourhood of the limit cycle) are not closed
ẋ = u(x, y), but either approach the limit cycle asymptotically,
as in the case of a stable limit cycle, or move
do the following.
further away from it, as for an unstable limit
(a) Find the equilibrium points by using item 4. cycle. One may also find a half-stable limit

39
Handbook

cycle, where neighbouring orbits on one side Set n = 1

approach the limit cycle but neighbouring orbits
on the other side move further away. Limit cycles
occur only in nonlinear systems. Carry out commands
cmd 1, cmd 2, cmd 3, etc.
11 The Poincaré–Bendixson theorem can be
stated as follows. Consider the two-dimensional
autonomous system ẋ = u(x), with x(0) = x0 and Increase n by 1
u sufficiently well-behaved (i.e. continuously
differentiable). If a solution of the system x(t)
remains in a closed, bounded region of R2 for all
t ≥ 0, then this orbit must: (a) be a closed path
(i.e. a periodic orbit); or (b) approach a closed yes
Is n ≤ N ?
path (i.e. a stable limit cycle) as t → ∞; or
(c) approach a fixed point as t → ∞.
12 The following properties serve as a working no
definition of a chaotic orbit.
Proceed to the next
(a) It has extreme sensitivity to initial step in the program
conditions, which means that two orbits that are
initially very close together rapidly move apart in
such a way that the distance between the orbits
15 Consider the iterated map
grows exponentially fast in time until they are
sufficiently far apart. xn = f (xn−1 , n), n = 1, 2, 3, . . . .
(b) It remains bounded for all time.
The following pseudocode uses numerical
(c) It is aperiodic (i.e. not periodic) and does not iteration to print the value of xN given the value
tend to fixed points or settle into periodic motion of x0 .
as time t → ∞.
(a) Set a value for N .
13 Consider an orbit x(t) and an initially nearby (b) Set x = x0 .
orbit x(t) + δ(t), where |δ(0)| is very small. In the
(c) For n from 1 to N do (Replace x by f (x, n)).
case of a chaotic orbit,
(d) Print x.
|δ(t)| ' |δ(0)| eht , h > 0,

as t increases from zero until |δ(t)| gets sufficiently Unit 4 Random motion and
large. Thus the orbit has an exponential sensitivity
to its initial condition. The constant h is called the probability
Lyapunov exponent. 1 The simple random walk models the
14 A loop repeats a set of commands (cmd 1, position of a particle on a line. Given the position
cmd 2, cmd 3, etc.) N times by using a counter n, xn at time n, the position at time n + 1 is
which starts at n = 1 and increases by 1 each time determined by tossing a coin. If the coin falls
the loop is repeated until n = N . heads up, then xn+1 = xn + 1. If the coin falls
heads down, then xn+1 = xn − 1. This can be
It is expressed in pseudocode as summarised in the form of a recurrence relation as
For n from 1 to N do (cmd 1 ; cmd 2 ; etc.) xn+1 = xn + rn ,

and illustrated by the flowchart below. where rn is a random variable obtained from

40
 Unit 4 summary

tossing a coin: 9 The probability density for the uniform

( distribution on the interval [X1 , X2 ], where
+1, nth coin toss lands heads,
rn = X1 < X2 , is
−1, nth coin toss lands tails. 
1

 , X1 ≤ x ≤ X2 ,
2 Each outcome of a stochastic process (under ρ(x) = X2 − X1
specified conditions) is referred to as a 0, otherwise.

realisation; the act of obtaining a realisation is
called a trial, and a succession of trials is referred 10 The probability density for the normal
to as an experiment. distribution is
(x − x0 )2
 
3 If an event labelled by an index i occurs 1
ρ(x) = √ exp − .
Ni times in N trials, then the probability of this 2πσ0 2σ02
event is The distribution depends on two parameters: x0 is
Ni the mean value, and σ0 is the standard
Pi = lim .
N →∞ N deviation.
4 If there are n events labelled by an index i A function of the same form as the right-hand side
that describe all possible outcomes, then their of the equation above is called a Gaussian
probabilities Pi satisfy function.
n
X 11 The probability density function of the normal
Pi = 1. distribution is normalised because of the following
i=1
result on the integral of a Gaussian function:
5 If A and B are mutually exclusive events, Z ∞
√
with probabilities PA and PB , respectively, then exp(−x2 ) dx = π.
−∞
the probability of observing either event is
12 The probability density for the exponential
PA or B = PA + PB . distribution is
6 If a and b are independent events at
 1 exp(−x/x0 ), x ≥ 0,


successive trials, with probabilities Pa and Pb , ρ(x) = x0
then the probability of observing both events in 0, x < 0,

successive trials is
where x0 is a positive constant, called the mean
Pa,b = Pa Pb ,
value.
where Pa,b denotes the probability of the events 13 In the limit as N → ∞, the average xav of a
occurring in succession. set of N realisations of a random variable x
7 If x is a continuously varying random variable, approaches hxi, the mean value or expectation
let P (x1 , x2 ) be the probability that x1 < x < x2 . value (or sometimes just the mean). In the case
The probability density function, or just where the random variable takes one of a finite
probability density, ρ(x) is defined by writing discrete set of values Xj with probability Pj , the
Z x2 mean value is given by
P (x1 , x2 ) = ρ(x) dx. n
x1
X
hxi = Pj Xj .
8 A valid probability density function ρ(x) must j=1
satisfy the following conditions:
Z ∞ 14 A discrete random variable takes one of a
ρ(x) dx = 1 (normalisation), discrete set of values Xj , each occurring with
−∞ probability Pj . There may be a finite or an infinite
ρ(x) ≥ 0 for all x (non-negativity). number of possible values.

41
Handbook

A continuous random variable takes a value are N random variables, then the mean value of
within a given range (which may be infinite). It is the sum
described by giving a probability density ρ(x). N
X
Formulas for statistics such as the mean value are X= Xi
expressed in a slightly different form depending on i=1
whether we are considering a discrete or a is the sum of the mean values:
continuous random variable. *N N
+
15 The mean value of a continuous random
X X
hXi = Xi = hXi i.
variable is given by i=1 i=1
Z ∞
hxi = ρ(x) x dx. This result applies in all cases (even when the
−∞ random variables are not independent).
16 For a continuous random variable x with If X and Y are two independent random variables,
probability density ρ(x), the moments Mk are then the mean value of their product is the
Z ∞ product of their mean values:
k
Mk = hx i = xk ρ(x) dx,
−∞ hXY i = hXihY i.
where k = 0, 1, 2, . . .. Note that M1 is the mean,
These results apply for both discrete and
and (recalling the normalisation condition (item 8)
continuous distributions.
M0 = 1. We call M1 the first moment, M2 the
second moment, and so on. 21 The recurrence relation for simple
random walk probability is
In the case where x has n discrete values Xj , with
probabilities Pj , j = 1, . . . , n, we have P (x, n) = 1


P (x − 1, n − 1) + P (x + 1, n − 1) .
2
n
k
hx i =
X
Xjk Pj . 22 To calculate the statistics of a random
j=1 walk, consider the random walk determined by the
recurrence relation
17 The variance of x, written as Var(x), is the
second moment of the difference between x and its xn+1 = xn + rn+1 , x0 = 0,
mean value:
where the rn are independent random variables
2
Var(x) = h(∆x) i, where ∆x = x − hxi. each with the same probability distribution as a
random variable r.
18 The standard deviation σ of x is the square
root of its variance: The mean and variance of the displacement xn
p after n steps are equal to n times the mean and
σ = Var(x). variance of the displacements of a single step:
19 Calculating the moments M1 = hxi and hxn i = n hri,
M2 = hx2 i of the probability density ρ(x) using Var(xn ) = n Var(r).
item 16, the variance is obtained from In the case where the single-step displacements rn
Var(x) = hx2 i − hxi2 = M2 − M12 , have a continuous distribution, the probability
distribution of xn approaches a normal distribution
These formulas are valid for discrete distributions as n → ∞, so the probability density is
as well as continuous distributions. approximated by

20 (x − nhri)2
 
We can calculate the statistics of sums and 1
ρn (x) = p exp − .
products of random variables as follows. If Xi 2πn Var(r) 2n Var(r)

42
 Unit 5 summary

Unit 5 Introduction to the where g is the magnitude of the acceleration due

to gravity. The stationary curve y(x) that solves
calculus of variations the problem is called the brachistochrone.
• Finding the surface of minimal area that is
1 A functional S[f ] is a map from functions f
generated by revolving a curve y(x) about the
to the real numbers.
x-axis. The functional for the surface area is
2 The distance along a curve y = y(x), in the
Z b p
(x, y)-plane, that passes between the points (a, A) S[y] = 2π y 1 + (y 0 )2 dx,
a
and (b, B) is given by the functional y(a) = A, y(b) = B,
Z bp and the stationary curve y(x) solves the problem.
S[y] = 1 + y 0 (x)2 dx, y(a) = A, y(b) = B. This minimal surface of revolution is a good
a
approximation to the shape formed by a soap
3 Let S[y] be a functional that maps functions y film suspended between two circular hoops.
that satisfy y(a) = A and y(b) = B to the real
numbers. Any such function y(x) for which 6 Each of the partial derivatives of a function
 of several variables is obtained by differentiating
d  with respect to one variable while holding the
S[y + εg] = 0,
dε ε=0 others constant. The total derivative is obtained
by differentiating with respect to some variable
for all functions g(x) that satisfy g(a) = g(b) = 0, without holding the other variables constant.
is said to be a stationary path of S, or
alternatively a stationary curve or a stationary For example, if F (x, y, y 0 ) = xy 2 y 0 is a function of
function of S. an independent variable x and functions y(x) and
y 0 (x) = dy/dx, then the partial derivatives of F
Also, if S and y(x) satisfy the above equation, then with respect to each variable are ∂F/∂x = y 2 y 0 ,
we say that S is stationary at y(x), and ∂F/∂y = 2xyy 0 , ∂F/∂y 0 = xy 2 .
sometimes we abbreviate this by merely saying
that S[y] is stationary. The total derivative of F with respect to x is given
by the chain rule:
4 Leibniz’s integral rule states that for any d ∂F ∂F dy ∂F dy 0
sufficiently well-behaved function f (x, ε) and F (x, y, y 0 ) = + + 0
dx ∂x ∂y dx ∂y dx
independent variables x, ε, we have
= y y + 2xyy 02 + xy 2 y 00 .
2 0
Z b  Z b
d d
f (x, ε) dx = f (x, ε) dx, 7 Generally, if F (x, y, y 0 ) is a function of the
dε a a dε
independent variable x and the dependent variables
where a and b are fixed constants (that do not y(x) and y 0 (x) = dy/dx, then the functional
depend on ε). Z b
5 Examples of well-known problems that can be S[y] = F (x, y, y 0 ) dx, y(a) = A, y(b) = B,
a
solved by finding the stationary paths of
functionals include the following. has stationary paths that are given by solving the
Euler–Lagrange equation
• Finding the shape of a frictionless wire between  
two points, so that a bead will slide down the d ∂F ∂F
0
− = 0, y(a) = A, y(b) = B.
wire in the shortest time. The functional for the dx ∂y ∂y
time to slide down the wire is
Z bs 8 A special case of the Euler–Lagrange
1 1 + (y 0 )2 equation is when a functional has the form
T [y] = √ dx,
2g 0 y Z b
y(0) = 0, y(b) = B, S[y] = F (x, y 0 ) dx, y(a) = A, y(b) = B,
a

43
Handbook

in which the integrand does not depend explicitly Unit 6 Lagrangian mechanics
on y. Then ∂F/∂y = 0, so the Euler–Lagrange
equation can be integrated to give 1 Let x be a Cartesian coordinate, and consider
a particle with position x(t) at time t.
∂F
= constant. • The kinetic energy of the particle, if it has
∂y 0
mass m and velocity v = dx/dt, is given by
9 If a functional has the form T = 12 mv 2 .
Z b
• A particle is acted on by a conservative force
S[y] = F (y, y 0 ) dx, y(a) = A, y(b) = B,
a F (x) if the force can be written as
dV
Z
where the integrand does not depend explicitly on F =− , or equivalently V = − F dx.
dx
the independent variable x, then the
Euler-Lagrange equation reduces to We say that V (x) is the potential energy or
potential for the force. At a given point x, we
∂F say that the particle has potential energy V (x).
y0 − F = c, y(a) = A, y(b) = B,
∂y 0 The constant of integration dictates where the
where c is a constant determined by the boundary potential energy is zero (the datum). This
conditions. This equation is called the choice is arbitrary, and does not affect the
first integral of the Euler–Lagrange equation. equation of motion.
The stationary path of the functional S can be • If several conservative forces Fi (x) act on the
determined by solving this first-order differential particle, then it has total potential energy
equation. X dVi
V = Vi , where Fi = − .
dx
10 Fermat’s principle says that the path taken i
between two points (a, A) and (b, B) by a ray of • If all the forces are conservative, then the
light is a stationary path of the time functional particle has total energy
1 b E = T + V.
Z p
T [y] = n(x, y) 1 + (y 0 )2 dx,
c a 2 In general, we may wish to use other
y(a) = A, y(b) = B, (non-Cartesian) coordinates to describe the motion
where n(x, y) is the refractive index of the medium of a particle.
through which the ray passes and For a single particle moving in one dimension
c ' 2.9 × 108 m s−1 is the speed of light in a (which may be on a curve), its position at time t
vacuum. This equation is sometimes written in the can always be specified by a single function q(t).
form
In Lagrangian mechanics it is conventional to label
Z b
coordinates of particles by the letter q when we
T [y] = L(x, y, y 0 ) dx, y(a) = A, y(b) = B,
a
have not specified what they are. If we specify that
they are Cartesian coordinates, then we normally
where use the letters x, y and z.
1 3 Hamilton’s principle for a particle
L(x, y, y 0 ) =
p
n(x, y) 1 + (y 0 )2
c moving in one dimension can be stated as
is called the optical Lagrangian and T is called follows. The trajectory q(t) of a particle of mass m,
the optical action. Fermat’s principle tells us starting from a position q(ta ) = a at time ta , and
that light rays passing between (a, A) and (b, B) ending at a position q(tb ) = b at time tb , is given
travel along paths that make the optical action by the stationary path of the functional
stationary. Z tb
S[q] = L(q, q̇, t) dt, q(ta ) = a, q(tb ) = b.
ta

44
 Unit 6 summary

Here the integrand L is called the Lagrangian of variables

the system, defined as the difference between the 
d ∂F

∂F
kinetic energy T and potential energy V of the 0 − = 0, yk (a) = Ak , yk (b) = Bk ,
dx ∂yk ∂yk
particle:
for k = 1, 2, . . . , n.
L = T − V.
7 The number of degrees of freedom of a
The functional S is called the action of the system is the smallest number of coordinates
system. needed to describe its configuration at any instant
4 The stationary path of the action S[q] (and in time. In any coordinate system,
hence the trajectory of the particle) is given by the number of coordinates − number of constraints
Euler–Lagrange equation for a particle = number of degrees of freedom.
moving in one dimension If a dynamical system has n degrees of freedom
and its configuration is specified by n coordinates,
 
d ∂L ∂L
− = 0, q(ta ) = a, q(tb ) = b, then those coordinates are said to be generalised
dt ∂ q̇ ∂q
coordinates for the system.
where L = L(q, q̇, t).
We often denote the n generalised coordinates for a
In mechanics, ‘Euler–Lagrange equation’ is often system by q = (q1 , q2 , . . . , qn ).
abbreviated to Lagrange’s equation, and the
two terms are used interchangeably. 8 If the generalised coordinates are
q = (q1 , q2 , . . . , qn ), and their first derivatives with
5 Carry out the following steps to find the
respect to time (sometimes called the generalised
equation of motion of a particle that is
velocities) are q̇ = (q̇1 , q̇2 , . . . , q̇n ), then the
constrained to move on a fixed path with
kinetic energy T in generalised coordinates
coordinate q = q(t) at time t.
is a function of q, q̇ and t, so we can write
(a) Express the Cartesian coordinates x, y and (if T = T (q, q̇, t).
necessary) z in terms of q.
9 For a system of N particles with Cartesian
(b) By taking the derivatives of these functions, positions ri (t) = (xi (t) yi (t) zi (t))T , the total
calculate ẋ, ẏ and (if necessary) ż in terms of q kinetic energy is
and q̇.
N
(c) Determine the kinetic energy X
T = 1
mi (ẋ2i + ẏi2 + żi2 ).
T = 12 mv 2 = 12 m(ẋ2 + ẏ 2 + ż 2 ) in terms of q and q̇. 2
i=1
(d) If the potential energy V (x, y, z) is known in
Cartesian coordinates, express it in terms of q. Carry out the following steps to express the
total kinetic energy in terms of generalised
(e) Construct the Lagrangian L = T − V in terms
coordinates q = (q1 , q2 , . . . , qn ).
of q and q̇.
(f ) Use Lagrange’s equation to determine the (a) Express the Cartesian coordinates xi , yi and
equation of motion. zi in terms of q, and if necessary t: xi = xi (q, t),
yi = yi (q, t), zi = zi (q, t).
6 The functional
Z b (b) Differentiate these with respect to time to
S[y] = F (x, y, y0 ) dx, calculate ẋ, ẏ and ż in terms of q and q̇.
a
yk (a) = Ak , yk (b) = Bk , (c) Substitute these expressions into the formula
for the total kinetic energy to determine T in
of n dependent variables
terms of q, q̇ and possibly t: T = T (q, q̇, t).
y(x) = (y1 (x), y2 (x), . . . , yn (x)) has a set of
n stationary paths that are given by solving the 10 Let r = (x, y, z) represent the
Euler–Lagrange equations for n dependent three-dimensional Cartesian coordinates. If a force

45
Handbook

F(r) can be written as 12 The general formulation of Hamilton’s

  principle is as follows. The trajectories q(t)
∂V ∂V ∂V
F=− i+ j+ k , between two points q(ta ) = a and q(tb ) = b,
∂x ∂y ∂z starting and ending at times ta and tb , are given by
then it is said to be conservative, and the the stationary paths of the functional
function V (r) is said to be the potential energy Z tb
function for the force. S[q] = L(q, q̇, t) dt, q(ta ) = a, q(tb ) = b.
ta
If there are several conservative forces Fi in the Here the integrand L is called the Lagrangian of
system, then the total potential energy is the system, defined as the difference between the
X 
∂Vi ∂Vi ∂Vi
 kinetic energy (T ) and potential energy (V ) for the
V = Vi , where Fi = − i+ j+ k . system:
∂x ∂y ∂z
i
L = T − V.
11 Consider a system of particles acted on by
forces of gravity and/or springs. The functional S is called the action of the
system.
• If a spring with natural length l0 and stiffness k
has length d, then this gives rise to a potential
13 The stationary paths of the action
Z tb
V = 12 k(d − l0 )2 . S[q] = L(q, q̇, t) dt, q(ta ) = a, q(tb ) = b,
ta
• If a particle has mass m, and lies at a position z
are given by the Euler–Lagrange equations for
vertically above some given point, then this gives
a mechanical system
rise to a potential  
d ∂L ∂L
V = mgz. − = 0, k = 1, 2, . . . , n.
dt ∂ q̇k ∂qk
• If two bodies with masses m1 and m2 , a distance
In mechanics, these equations are often called
r apart, move under the influence of each other’s
Lagrange’s equations.
gravitational pull, then they give rise to a
potential 14 To determine the equations of motion
for a system of N particles, carry out the
Gm1 m2
V =− . following steps.
r
• The total potential energy of a system is (a) Identify the number of degrees of freedom, n,
obtained by summing all the individual in the system.
potentials (b) Choose suitable generalised coordinates
P due to gravity and/or springs: q = (q1 , q2 , . . . , qn ), and determine the total kinetic
V = i Vi .
energy T and potential energy V in terms of these,
If the positions of N particles are expressed in as outlined here in items 9 and 11.
Cartesian coordinates ri , then the total potential
(c) Construct the Lagrangian L = T − V , and use
energy has the form V = V (r1 , r2 , . . . , rN ).
Lagrange’s equations
Carry out the following steps to express the total
 
d ∂L ∂L
potential energy in generalised coordinates − = 0, k = 1, 2, . . . , n,
dt ∂ q̇k ∂qk
q = (q1 , q2 , . . . , qn ). to determine the n equations of motion.
(a) Express the Cartesian coordinates xi , yi and 15 If a function f (q, q̇) of the positions and
zi in terms of q and if necessary t: xi = xi (q, t), generalised velocities remains constant during the
yi = yi (q, t), zi = zi (q, t). motion of a system, then it is called an integral of
(b) Substitute these expressions into the formula the motion or constant of the motion.
for the total potential energy to determine V in When energy or momentum is a constant of the
terms of q and possibly t: V = V (q, t). motion, then it is said to be conserved.

46
 Unit 7 summary

16 If the Lagrangian L = L(q, q̇) does not • If a system has a Lagrangian that is invariant
depend explicitly on the time t, then the quantity under all translations in space r → r + δr (where
N
δr is any constant vector), then linear
X ∂L momentum is conserved.
q̇k − L = E,
∂ q̇k • If a system has a Lagrangian that is invariant
k=1
under all rotations about the origin, then
where E is a constant, is an integral of the motion,
angular momentum is conserved.
and its value is E = T + V , the total energy of the
system; that is, the total energy is conserved.
17 If the Lagrangian L of a system does not
Unit 7 Exploring dynamics
depend on a particular generalised coordinate, using maps
ql say, then ql is called an ignorable coordinate,
and the corresponding generalised momentum 1 Consider the system of differential equations
ẋ = u(x, t). Sample the solution at regular time
∂L intervals T to obtain a sequence of points
pl = = constant,
∂ q̇l xn = x(nT ). The vector-valued map F that relates
is an integral of the motion. successive points via xn+1 = F(xn ) is called the
time-T map.
We distinguish two important types of momentum.
For a particle of mass m, with position vector 2 Imagine a surface Σ, embedded in
r(t) = (x(t) y(t) z(t))T , its linear momentum three-dimensional space as illustrated in the figure.
is given by p = mṙ, and its angular z
momentum l, about the origin, is given by the
cross product l = r × p = m(r × ṙ).
Σ
• If an ignorable coordinate ql is a Cartesian xn xn+1 y
coordinate, then the integral of motion is the
component of linear momentum in the direction
of ql . For example, for a particle with position
(x(t), y(t), z(t)), if the Lagrangian is independent
of x, then px = mẋ is conserved.
x
• If an ignorable coordinate ql is rotation about a
given axis, then the integral of motion is the An orbit pierces the surface Σ in a downward
component of angular momentum in the direction at points xn , n = 0, 1, 2, . . ., with xn
direction of the axis of rotation. For example, being the nth such intersection. The surface Σ is
for a particle with position (r(t), θ(t), z(t)) in called the surface of section or the Poincaré
cylindrical coordinates, if the Lagrangian is section. Now, each time the orbit pierces the
independent of θ, then lz = m(xẏ − y ẋ) is surface Σ in the same sense (the downward
conserved. direction in the case illustrated), the point of
intersection is recorded, the nth such intersection
18 A Lagrangian is said to be invariant with
being denoted by xn . The map P defined by
respect to time t or a coordinate qi if it is
independent of t or qi . Equivalently, if we xn+1 = P(xn )
transform time or the coordinate in some way,
is called the Poincaré return map.
t → t + δt or qi → qi + δqi , then the Lagrangian
does not change. 3 We use the symbol ‘◦’ to denote the
composition of two functions, f (x) and g(x), say,
• If a system has a Lagrangian that is invariant
as follows:
under all translations in time t → t + δt, then
energy is conserved. (f ◦ g)(x) = f (g(x)),

47
Handbook

that is, the function, f , of another function, g. In y

y=x
addition, another notation is often used to denote
the composition of a function with itself, as in
x3 y = f (x)
x2
f 2 (x) = (f ◦ f )(x) = f (f (x)),
x1
or more generally for k ≥ 1,
f k (x) = (f ◦ f ◦ · · · ◦ f )(x)
| {z }
k times x0 x1 x2 x
x3 x∗
= f (f (. . . f (f (x)) . . .)),
and we use the convention that f 0 (x) = x. (c) Thus we have been able to give graphical
expression to the map x0 7−→ x1 = f (x0 ), by
4 Consider the map xn+1 = f (xn ). The starting drawing the line joining the points (x0 , 0) and
value x0 is called the initial value of the map. (x0 , x1 ) together with the line joining (x0 , x1 ) and
(x1 , x1 ).
The orbit of x0 under the map xn+1 = f (xn ) is the
These steps can be repeated on the point (x1 , x1 )
sequence of points generated by the map, starting
so that a vertical line is drawn to the point
at the initial point x0 . It can be expressed as the
(x1 , f (x1 )) = (x1 , x2 ) (as in Step (a)), followed by
sequence
a horizontal line to the point (x2 , x2 ) (as in
Step (b)). These latter two lines represent the map
{x0 , x1 , x2 , . . .} = {x0 , f (x0 ), f 2 (x0 ), . . .}.
x1 7−→ x2 = f (x1 ). The steps can be carried out
repeatedly so that a picture of the orbit is built up
This sequence, also denoted
as shown by the arrowed lines in the figure.
{xn }n≥0 = {f n (x0 )}n≥0 , is referred to as the orbit
of x0 under f . 8 Consider a map f (x) with a fixed point x∗ .
We say that x∗ is an attracting fixed point, also
5 A linear map is any map whose mapping called a sink, if all orbits starting sufficiently close
function f (x) has the form f (x) = a + bx, where a to x∗ (but not at x∗ itself) converge to x∗ under f .
and b are constants. Conversely, we say that x∗ is a repelling fixed
point, also called a source, if all orbits starting
6 For the map xn+1 = f (xn ), a fixed point x∗
sufficiently close to x∗ (but not at x∗ itself) move
solves the equation
increasingly away from x∗ under f .
f (x∗ ) = x∗ . Attracting fixed points are stable; repelling fixed
points are unstable.
7 The cobweb plot for the orbit {xn }n≥0 of 9 To classify a fixed point x∗ of a
the map xn+1 = f (xn ) is constructed as follows. one-dimensional map f (x), check the following.

(a) Let x0 be some point on the x-axis. Draw the • If |f 0 (x∗ )| < 1, then x∗ is a stable fixed point,
vertical line from (x0 , 0) until it meets the curve i.e. a sink.
y = f (x). The y-coordinate of this point of • If |f 0 (x∗ )| > 1, then x∗ is an unstable fixed
intersection is x1 since y = f (x0 ) = x1 using the point, i.e. a source.
equation xn+1 = f (xn ). 10 A sink is called superstable if f 0 (x∗ ) = 0. In
(b) Now draw the horizontal line y = x1 until it this case the orbit converges to x∗ faster than any
intersects the line y = x. The x-coordinate of this exponential.
point of intersection is clearly x1 , as can be seen in 11 The set of all initial values whose orbits
the figure by tracing along the vertical dashed line converge to the sink x∗ is called the basin of
between the point and the x-axis. attraction of x∗ .

48
 Unit 7 summary

12 The logistic map is the nonlinear map given 17 A two-dimensional map describes the
by xn+1 = fa (xn ), where fa (x) = ax(1 − x) and a discrete-time evolution of points x = (x y)T on
is a constant parameter. the (x, y)-plane. Such a map is expressed in vector
13 An invariant set under a map is a set of notation as
points S that maps into itself under the map; that xn+1 = f (xn ),
is, each point in S is mapped to a point also in S.
Mathematically, we say that the set S is invariant where xn = (xn yn )T and the vector-valued map
under the map f if x ∈ S implies f (x) ∈ S. function is
   
f (x, y) x
14 A period-2 orbit under a map f (x) is an f (x) = for x = .
g(x, y) y
orbit whose points alternate between two distinct
values after successive iterations of the map. The So an equivalent and more explicit way to express
period-2 orbit is denoted by {x∗1 , x∗2 }, x∗1 6= x∗2 , and this map is
must satisfy    
xn+1 f (xn , yn )
= .
f (x∗1 ) = x∗2 , f (x∗2 ) = x∗1 . yn+1 g(xn , yn )

15 A period-k orbit under a map is an orbit A point with coordinate (xn , yn ) is mapped under
that cycles through k distinct values, repeating the f to a point with coordinates (xn+1 , yn+1 ). With
pattern at the kth iterate, that is, returning to the initial point x0 , the map will generate the orbit
first point after k iterations. More mathematically, given by the sequence of points
we say that the orbit of x∗j , j = 1, . . . , k, under the {xn }n≥0 = {x0 , x1 , x2 , . . .}, where xn = f n (x0 ).
map f (x) is a period-k orbit of f , denoted 18 The Hénon map is the two-dimensional map
{x∗1 , . . . , x∗k }, if it has the property xn+1 = f (xn ), where
(
x∗j+1 , 1 ≤ j ≤ k − 1, a − x2 + by
 
∗ f (x) = ,
f (xj ) = x
x∗1 , j = k,
and a and b are constant parameters.
and all the x∗j (j = 1, . . . , k) must be distinct in
order for the orbit to be defined as period-k. This 19 A linear two-dimensional map is given by
also means that xn+1 = Axn , where A is a constant 2 × 2 matrix.
This is also referred to as a linear
f k (x∗j ) = x∗j for 1 ≤ j ≤ k, transformation.
and that k is the smallest positive integer for 20 The elementary linear transformations
which this is true, so that points of a period-k are as follows.
orbit are fixed points of f k (x). This can also be
 
cos θ sin θ
regarded as a defining property of a period-k orbit. Rotation: Arot (θ) = .
− sin θ cos θ
16 To classify a period-k orbit {x∗1 , . . . , x∗k } of
 
λ1 0
the map f (x), perform the following stability test. Stretch: Astr (λ1 , λ2 ) = .
0 λ2
The orbit is:  
1 τ
• a period-k sink if Shear: Ash (τ ) = .
0 1
k
 
Y  Note that a stretch is expanding in the
f 0 (x∗j ) < 1
 

  x-direction (respectively, y-direction) if |λ1 | > 1
j=1
(respectively, |λ2 | > 1) and contracting in the
• a period-k source if x-direction (respectively, y-direction) if |λ1 | < 1
k (respectively, |λ2 | < 1). Rotations and shears are
 
Y 
f 0 (x∗j ) > 1.
 

  area-preserving linear transformations. A
j=1 stretch is area-preserving only if λ1 λ2 = ±1.

49
Handbook

21 A general linear transformation A can be 25 Consider a map with Jacobian matrix J(x).
constructed by successive application of the three The map is area-contracting if |det J(x)| < 1 for
elementary linear transformations: all x, whereas it is area-preserving if
det J(x) = ±1 for all x.
A = Arot (θ) Astr (λ1 , λ2 ) Ash (τ ).
26 Let x∗ be a fixed point of an area-preserving
22 For the map xn+1 = f (xn ), a fixed point x∗
map with stability matrix J(x∗ ) satisfying
solves the vector equation
det J(x∗ ) = 1.
f (x∗ ) = x∗ ,
• If |tr J(x∗ )| > 2, then x∗ is a saddle, i.e. a
or, in terms of the individual components of hyperbolic fixed point.
x∗ = (x∗ y ∗ )T , we have for the map • If |tr J(x∗ )| < 2, then x∗ is a centre, i.e. an
f (x) = (f (x, y) g(x, y))T that elliptic fixed point.
f (x∗ , y ∗ ) = x∗ , g(x∗ , y ∗ ) = y ∗ .
23 Let x∗ be a fixed point of a two-dimensional Unit 8 Quantifying chaotic
map, and let {xn }n≥0 be an orbit of the map. The
fixed point is stable if xn stays arbitrarily close to dynamics
x∗ for all n > 0 provided that x0 (6= x∗ ) is 1 If x(t) is a solution of a system of differential
sufficiently close to x∗ . A fixed point that is not equations, and x(t) + δx(t) is an initially nearby
stable is said to be unstable. solution with |δx(0)| very small, then the
24 Consider the two-dimensional map Lyapunov exponent h is defined by
xn+1 = f (xn ), where f (x) = (f (x, y) g(x, y))T for
x = (x y)T . To find and classify its fixed |δx(t)| ' |δx(0)| eht ,
points x∗ , do the following. which holds for large t > 0 provided that |δx(0)| is
(a) Find the fixed points by solving the vector sufficiently small.
equation f (x∗ ) = x∗ . 2 The Lyapunov exponent h(x0 ) of the orbit
(b) For each x∗ , determine the stability matrix of x0 under f (x) is given by
J(x∗ ), where J(x), the Jacobian matrix for the
n−1
map f (x), is given by 1X
h(x0 ) = lim ln |f 0 (xj )|,
∂f ∂f n→∞ n
 
 ∂x (x, y) ∂y (x, y)
j=0
J(x) =   ∂g
.
provided that this limit exists.
∂g 
(x, y) (x, y)
∂x ∂y 3 An asymptotically periodic orbit is an
(c) For each x , find the eigenvalues of J = J(x∗ )
∗
orbit that converges to a periodic orbit as the
using the formula iterate index n increases.
 p 
λ = 12 tr J ± (tr J)2 − 4 det J . 4 The orbit {xn }n≥0 of x0 under the map f is
The nature of the fixed point is determined by chaotic if the following three conditions hold.
these eigenvalues according to the following table. (a) {xn }n≥0 is bounded.
(b) {xn }n≥0 is not asymptotically periodic.
Eigenvalues |λ1 | ≥ |λ2 | > 1 Repeller (unstable)
real and |λ1 | ≤ |λ2 | < 1 Attractor (stable) (c) h(x0 ) > 0 for the Lyapunov exponent h(x0 ).
distinct |λ1 | > 1 > |λ2 | Saddle (unstable) 5 The tent map is defined by
Eigenvalues ρ>1 Repelling spiral (
complex (unstable) cx, x ≤ 21 ,
Tc (x) =
λ1,2 = ρe±iθ , ρ<1 Attracting spiral c(1 − x), x > 21 ,
ρ > 0, (stable)
0<θ<π ρ=1 Centre (stable) where c > 0.

50
 Unit 9 summary

6 Let δx0 be the initial displacement of an orbit these cubes we determine whether there is an
from a reference orbit of x0 under a element of S contained within the cube, in which
two-dimensional map. Then the displacement δxn case we say that the cube belongs to the
after n iterations of the map becomes ε-covering set of S.
δxn ' Jn (x0 ) δx0 , 10 The box-counting dimension of a set S is

where the 2 × 2 matrix Jn is the product of ln N (ε)

D = − lim ,
Jacobian matrices J at each application of the ε→0 ln ε
map: where N (ε) is the number of boxes in the
Jn (x0 ) = J(xn−1 ) J(xn−2 ) · · · J(x1 ) J(x0 ). ε-covering set of S.

This is the linearised equation for small

11 The baker’s map is a map of the unit square
to itself. It is defined by the mapping of
separations.
coordinates
7 For the two-dimensional map 
x(n + 1) = f (x(n)), there are two Lyapunov ( 1 x, 2y), 0 ≤ y ≤ 12 ,
0 0 2
(x, y) 7−→ (x , y ) =
exponents, denoted as h1 and h2 . The first ( 1 x + 1 , 2y − 1), 1 < y ≤ 1.
2 2 2
Lyapunov exponent, h1 , is defined as
1 12 The skinny baker’s map is defined by
h1 = lim ln(|Jn e1 |),
n→∞ n

( 1 x, 2y), 0 ≤ y ≤ 12 ,
0 0 3
and the second Lyapunov exponent, h2 , is (x, y) 7−→ (x , y ) =
( 1 x + 2 , 2y − 1), 1 < y ≤ 1.
defined via 3 3 2
1
h1 + h2 = lim ln |(Jn e1 ) × (Jn e2 )|
n→∞ n
13 The pre-image (or inverse image) of the
set S under the map f , denoted f −1 (S), is the set
or
of all points that map to S under f .
1
h1 + h2 = lim ln |det(Jn )|, 14 If a two-dimensional map has Lyapunov
n→∞ n
exponents h1 and h2 , then the Kaplan–Yorke
where Jn is the product of Jacobian matrices
dimension or Lyapunov dimension DL is given
Jn = J(x(n − 1)) J(x(n − 2)) · · · J(x(1)) J(x(0)), by

and e1 and e2 are perpendicular unit vectors (here 
 0, h1 < 0,
e1 = a1 i + a2 j can be chosen to be almost any unit

 h1
vector). DL = 1 + , h1 > 0, h1 + h2 < 0,

 |h2 |
8

Suppose that an orbit of a two-dimensional
 2, h1 + h2 ≥ 0.
map is bounded and not asymptotically periodic
with Lyapunov exponents h1 and h2 . Then we
have the following.
Unit 9 Random walks and
(a) The orbit is chaotic if h1 > 0. diffusion
(b) The orbit is an attractor if h1 + h2 < 0. 1 The mean and variance of the random
(c) The orbit is a strange attractor if walk with displacement xn+1 = xn + rn after
h1 + h2 < 0 but h1 > 0. Note that strange n steps, starting at x0 = 0, are
attractors are (because of point (a) above) chaotic.
hxn i = nhri, Var(xn ) = n Var(r),
9 Let S be a subset of a d-dimensional space.
We pick a value for a positive number, denoted respectively, where hri and Var(r) are the mean
by ε, and divide the d-dimensional space into and variance of the random variable rn (note that
d-dimensional cubes of side length ε. In each of these statistical quantities are independent of n).

51
Handbook

2 If the concentration of molecules is c(r, t), diffusion equation gives the concentration at
then the number of molecules inside a time t > t0 as
three-dimensional region V at time t is given by a
(x − x0 )2
 
N
volume integral c(x, t) = p exp − .
Z 4πD(t − t0 ) 4D(t − t0 )
N (t) = c(r, t) dV. The mean squared displacement satisfies
V

3 If ∆N particles cross a surface in a small h(x − x0 )2 i = 2D(t − t0 ).

time ∆t, then the flux of particles is defined as
10 The one-dimensional advection–diffusion
∆N equation is
Φ= .
∆t
∂c ∂2c ∂

That is, flux is the number of particles crossing the =D 2 − v(x, t) c ,
surface per unit time. ∂t ∂x ∂x
where D is the diffusion coefficient and v(x, t) is a
We often write this as
function that represents the velocity of the fluid in
dN
Φ= . which the diffusion is taking place.
dt
4 The scalar flux density (or just flux
11 If xn is the displacement at the nth time step
of a random walk, and the steps are separated in
density) J is the number of particles per unit area
time by ∆t, then the probability density P (x, t)
per unit time passing through an imaginary surface
(which gives the probability that the random walk
element. In general, J depends on position, time,
will reach position x at time t) obeys an
and the orientation of the surface.
advection–diffusion equation with diffusion
5 If some material that is neither created nor coefficient
destroyed has concentration c and flux density J,
Var(rn )
then these are related by the continuity D=
2 ∆t
equation, which in one dimension is
and velocity
∂c ∂J
+ = 0. hrn i
∂t ∂x v= ,
6 Diffusion is a process in which materials are ∆t
mixed by random microscopic motion of atoms, which is called the drift velocity.
without any large-scale (macroscopic) motion 12 A unit normal vector nb is a vector of
(such as occurs when a liquid is stirred). length 1 that is perpendicular to a surface element.
7 Fick’s law states that
The oriented area element with area δA and
∂c unit normal vector n
b is defined as δA = n
b δA.
J = −D ,
∂x
13 Given any vector field F and a planar element
where D is a positive constant called the diffusion
with oriented area δA, the flux of the vector field
coefficient. It describes the phenomenon that
over the element is defined as
molecules tend to move from regions of high
concentration to regions of low concentration by flux = F · δA = (F · n
b ) δA = F δA cos θ,
the process of diffusion.
where the field F is evaluated at the position of the
8 The one-dimensional diffusion equation is element, and θ is the angle between the directions
∂c ∂2c of F and the unit normal n b to the planar element.
= D 2.
∂t ∂x 14 The flux of a vector field over a planar element
9 If N particles are released from position x0 at is the normal component of the field multiplied by
time t0 , then the point source solution of the the area of the element.

52
 Unit 9 summary

Flux is a scalar quantity, which can be positive, where ∇ · J denotes the divergence of J.
negative or zero depending on the relative
18 For a steady flow or steady state,
orientations of F and the unit normal nb.
∂c/∂t = 0.
15 An open surface has at least one boundary
19 The following identity holds for any closed
curve marking the furthest extent of the surface.
surface S containing a volume V in which a
In this case there are two choices for the unit
differentiable vector field F is defined:
normal vectors, and we must choose one of these Z Z
consistently across the surface. F · dA = ∇ · F dV.
S V
A closed surface has no boundary curves, and
divides three-dimensional space into two parts: the This is known as the divergence theorem or
space inside the surface and the space outside the Gauss’s theorem, a fundamental result in the
surface. For any closed surface, all the unit theory of vector calculus (which extends the
normals are conventionally chosen to point concepts of differentiation and integration to
outwards into the exterior space, rather than vectors). It enables integrals over surfaces to be
inwards towards the enclosed volume (see figure). expressed as integrals over volumes, and vice versa.
20 Fick’s law in three dimensions states that

J = −D ∇c,

where the components are

 
∂c ∂c ∂c
(Jx , Jy , Jz ) = −D , , .
∂x ∂y ∂z

A closed surface 21 The diffusion equation in three

dimensions can be written as
16 The flux Φ of an arbitrary vector field F
∂c
 2
∂ c ∂2c ∂2c

across a surface S is given by the surface integral =D + + .
Z ∂t ∂x2 ∂y 2 ∂z 2
Φ= F · dA. This is often written in the compact notation
S
If we are considering the motion of particles, then ∂c
= D ∇2 c,
the flux density is denoted by J. The rate at ∂t
which particles cross a surface S is determined by where
a surface integral of the flux density: ∂2 ∂2 ∂2
∇2 = + +
dN
Z
∂x2 ∂y 2 ∂z 2
= J · dA.
dt S is known as the Laplacian operator.
17 If some material that is neither created nor
Note that ∇2 = ∇ · ∇.
destroyed has concentration c(r, t) and flux
density J(r, t), then in three dimensions these are 22 When heat energy flows through a material by
related by a three-dimensional continuity thermal conduction, the temperature Θ(r, t) obeys
equation that takes the form the heat equation
∂c ∂Jx ∂Jy ∂Jz ∂Θ
+ + + = 0, = D ∇2 Θ,
∂t ∂x ∂y ∂z ∂t
or, using the notation of vector calculus, where D is a material-dependent constant called
the thermal diffusivity. The heat equation has
∂c
+ ∇ · J = 0, an identical form to the diffusion equation.
∂t

53
Handbook

Unit 10 Fourier transforms Name f (x) fe(k)

1 Given a function f , its Fourier transform fe

r
2 1
is the function defined by Exponential–Lorentzian e−|x|
π 1 + k2
Z ∞ 2 /2 2 /2
1 Gaussian–Gaussian e−x e−k
f (k) = √
e f (x) e−ikx dx
2π −∞ r
2 sin(k)
Top-hat–sinc χ(x)
for all real k. The original function f can be π k
recovered from its Fourier transform by the
inverse Fourier transform as 7 A monomial is a simple type of polynomial
Z ∞ that consists of only one term, with coefficient 1.
1 So x, x2 , x3 are all monomials of the independent
f (x) = √ fe(k) eikx dk.
2π −∞ variable x.
We also use the notation F(f (x)) for the Fourier 8 The Fourier transform rules are as follows.
transform of f (x).
Rule Function Fourier transform
The variables x and k in these definitions are
called conjugate variables. Another pair of Linearity α f (x) + β g(x) α fe(k) + β ge(k)
commonly used conjugate variables is t and ω. Scaling rule (α 6= 0) f (x/α) |α| fe(αk)
2 The Lorentzian function is given by the Translation along x rule f (x − a) e−ika fe(k)
formula
Translation along k rule eiax f (x) fe(k − a)
1
L(k) = . Inversion rule fe(k) f (−x)
1 + k2
Even symmetry rule f even fe even
3 The top-hat function is denoted by χ(x)
and defined by Odd symmetry rule f odd fe odd
dn
(
1, |x| ≤ 1, Derivative rule f (x) (ik)n fe(k)
χ(x) = dxn
0, |x| > 1. dn e
Monomial rule xn f (x) in f (k)
4 The characteristic function of a set S is dk n
denoted by χS (x) and defined by 9 To find the Fourier transform of f (x) using
( the scaling rule, do the following.
1, x ∈ S,
χS (x) = (a) Start by writing down the nearest known
0, otherwise. Fourier transform.
So the top-hat function is equal to the (b) Apply the scaling rule with scaling α to this
characteristic function of the interval [−1, 1]. equation.
(c) Compare the resulting Fourier transform with
5 The sinc function is defined as f (x) to obtain an equation involving α.

 sin(x) , x =

6 0, (d) Solve the equation for α (equating coefficients
sinc(x) = x as necessary, using the fact that the equation must

1, x = 0. hold for all x).
(e) Substitute the α obtained in Step (d) into the
The sinc function is continuous for all x ∈ R. equation derived in Step (b) to give the required
6 Three standard Fourier transform pairs are Fourier transform.
given in the following table. 10 To solve a differential equation for an

54
 Unit 11 summary

unknown function y(x) using Fourier (a) Take the Fourier transform of the partial
transforms, do the following. differential equation with respect to x (using k as
the conjugate variable).
(a) Take the Fourier transform of the differential
equation. (b) Solve the resulting ordinary differential
equation to find ye(k, t). Express any constants of
(b) Derive an algebraic equation for the Fourier
integration in terms of the Fourier transforms of
transform ye.
the initial state (such as ye(k, 0)).
(c) If possible, solve this equation to obtain an (c) Write the resulting particular solution in the
expression for ye. form
(d) Take the inverse Fourier transform to obtain y. √
ye(k, t) = 2π K(k,
e t) ye(k, 0),
11 The convolution of two functions f and g is where K(k,
e t) is an explicit expression obtained in
the function f ⊗ g defined by Step (b).
Z ∞
(d) Use the inverse Fourier transform to calculate
(f ⊗ g)(x) = f (x − y) g(y) dy.
−∞ K(x, t).

12 The convolution theorem states that given

4 To apply the separation of variables
procedure to find the solution of a homogeneous
two functions f and g with Fourier transforms fe
linear partial differential equation with dependent
and ge, we have

√ variable u and independent variables x and t,
F (f ⊗ g)(x) = 2π fe(k) ge(k). subject to boundary and initial conditions, do the
following.
(a) Separate the variables by substituting the trial
Unit 11 Solving the diffusion solution
equation u(x, t) = X(x) T (t)
1 The non-transform derivative rule states into the partial differential equation, and rearrange
that if c(x, t) is sufficiently smooth and we are so that each side of the equation involves only one
considering Fourier transforms in the variable x, of the independent variables. Both sides must then
then be equal to a separation constant µ. Rearranging
  then gives two separate ordinary differential
∂c ∂ ∂
F = F(c) = e c(k, t), equations for X and T . The boundary conditions
∂t ∂t ∂t for u will give boundary conditions for X.
where e
c(k, t) is the Fourier transform of c(x, t). (b) Find the general solution of the ordinary
differential equations for X and T found in
2 The solution c(x, t) of the one-dimensional
Step (a). Use the boundary conditions for X to
diffusion equation without boundary conditions
find the normal mode solutions un (x, t).
can be expressed as
Z ∞ (c) Write down the general solution as a linear
c(x, t) = K(x − y, t) c(y, 0) dy. combination of the normal mode solutions:
∞
−∞ X
u(x, t) = an un (x, t).
The propagator K(x, t) can be interpreted as the n=0
probability that a particle will diffuse a distance x The initial conditions (and results about Fourier
in time t, and it is given by series) can be used to determine the constants an
1 appearing in this solution.
K(x, t) = √ exp(−x2 /4Dt).
4πDt 5 The general solution of the growth/decay
equation T 0 (t) − µ T (t) = 0 is
3 To calculate the propagator for a partial
differential equation for y(x, t), do the following. T (t) = Ceµt ,

55
Handbook

where C is a constant. interval [0, L]:

∞
6 The general solution of the simple harmonic X
motion (SHM) equation X 00 (x) − µ X(x) = 0 is g(x) = an fn (x),
n=0

rx −rx
Ae + Be , µ > 0, where {fn } is an orthonormal set of eigenfunctions.


X(x) = Ax + B, µ = 0, The coefficients an are given by

 Z L
A cos(kx) + B sin(kx), µ < 0,
an = fn (x) g(x) dx.
√ 0
where A and B are constants, and r = µ and
√
k = −µ. 13 If c(x, y, t) satisfies the diffusion equation on
the rectangular region 0 < x < a, 0 < y < b
7 A Dirichlet boundary condition is a together with Neumann boundary conditions, then
boundary condition where the function value is X∞ X ∞
zero at a boundary, such as u(0, t) = 0, u(L, t) = 0 c(x, y, t) = amn cos(km x) cos(Kn y)×
or X(L) = 0. m=0 n=0
2 2
A Neumann boundary condition is a boundary e−(km +Kn )Dt ,
condition where a derivative of the function is zero for some constants amn , where km = mπ/a and
at a boundary, such as ∂u/∂x = 0 at x = 0, Kn = nπ/b.
∂u/∂x = 0 at x = L, or X 0 (L) = 0. 14 The second-order linear ordinary differential
8 An eigenproblem is a differential equation of equation
the form L(X) = µX, together with Dirichlet or d2 R dR
Neumann boundary conditions, where L(X) = 0 is z2 2
+z + (z 2 − n2 )R = 0
dz dz
a linear homogeneous differential equation. A
is known as Bessel’s equation. The solutions of
non-zero solution is called an eigenfunction, and
this equation can be expressed in terms of two
the corresponding value of µ is called an
families of functions, Jn (z) and Yn (z),
eigenvalue. The set of all eigenvalues is called the
n = 0, 1, 2, . . ., which are known as Bessel
spectrum.
functions of the first and second kind. Bessel
9 Two eigenfunctions fn and fm defined on an functions of the first kind, Jn (z), are finite for all
interval [0, L] are orthogonal if real z. Bessel functions of the second kind, Yn (z),
Z L tend to infinity as z tends to zero.
fm (x) fn (x) dx = 0, m 6= n. 15 The R(r) eigenproblem
0
1 0 n2
10 An eigenfunction fn defined on an interval R00 (r) + R (r) − 2 R(r) = µ R(r), R0 (a) = 0,
[0, L] is normalised if r r
Z L with R(r) finite for 0 ≤ r ≤ a, has eigenfunctions
fn (x)2 dx = 1. Rmn (r) = Jn (kmn r), m = 1, 2, 3, . . . , n = 1, 2, 3, . . .
0
2 , where
with corresponding eigenvalues µ = −kmn
11 A set {fn } of functions defined on an interval
[0, L] is said to be an orthonormal set if kmn = zmn /a and zmn is the mth zero of Jn0 (z).
Z L 16 For the solution of an equation, a guess with
fm (x) fn (x) dx = δmn , undetermined constants or functions is called an
0 ansatz.
where δmn is the Kronecker delta. 17 To solve a differential equation using an
12 A generalised Fourier series is a ansatz, do the following.
representation of a function g(x) defined on an (a) Guess an ansatz.

56
 Unit 12 summary

(b) Substitute the ansatz into the differential is defined as

equation and boundary conditions to derive Z ∞
relations between the unknown constants and τ= F (t) t dt.
0
functions.
(c) Solve the resulting relationships to find the It can also be related to the survival probability as
Z ∞
unknown functions and constants, and hence find
the solution. τ= S(t) dt.
0

6 The mean exit time of a random walk with

Unit 12 Case studies in initial position x0 in a finite domain 0 ≤ x0 ≤ L is
given by
stochastic processes
L  x0 
1 An absorbing boundary condition is when τ (x0 ) = x0 1 −
2D L
a particle ceases to exist when a boundary x = xc
is reached. Mathematically, this condition is if both boundary conditions are absorbing, or
imposed by L  x0 
τ (x0 ) = x0 1 −
P (x, t) = 0 at x = xc . D 2L
if there is a mix of absorbing and reflecting
2 A reflecting boundary condition is when a boundary conditions at x = 0 and x = L,
particle bounces back into the domain once a respectively.
boundary x = xc is reached. Mathematically, this
condition is imposed by 7 A generalised random walk is defined as a
random walk with a drift velocity v(x, t) and
∂P (x, t) diffusion coefficient D(x, t) that depend on position
=0 at x = xc .
∂x and/or time.
3 Given a one-dimensional random walk moving 8 The probability density function P (x, t) for a
in a domain D with a probability density function particle that follows a generalised random walk to
P (x, t), the survival probability S(t) is the reach position x at time t is given by the
probability that the random walk has not left the Fokker–Planck equation
domain by time t and is defined as
∂P (x, t) ∂
Z =− [v(x, t) P (x, t)] +
S(t) = P (x, t) dx. ∂t ∂x
D ∂2
[D(x, t) P (x, t)] ,
∂x2
4 The first-passage probability density F (t) where v(x, t) and D(x, t) are the drift velocity and
is the probability that a random walk leaves the
diffusion coefficient, respectively.
domain at time t:
∂S(t) This equation can also be written as a continuity
F (t) = − , equation,
∂t
where S(t) is the survival probability. The integral ∂P ∂J
+ = 0,
version of this relation is ∂t ∂x
Z t where
S(t) = 1 − F (t0 ) dt0 ,
0 ∂
J(x, t) = v(x, t) P (x, t) − [D(x, t) P (x, t)]
which is useful for finding the survival probability ∂x
from the first-passage density function. is called the probability flux.
5 The mean exit time τ is the average time 9 If the drift and diffusion coefficient do not
taken by a random walk to leave the domain, and depend on time, then the system may evolve into a

57
Handbook

stationary regime in which the statistical random walk model describes the evolution of
properties are independent of time, that is, the fluctuations of S, and it is defined as
∂Ps δS
= 0, = µ ∆t + σ δX,
∂t S
where Ps (x) is the stationary probability where σ is called the volatility of the stock price,
density function and is given by which measures how uncertain we are about future
price movements, and µ is called the drift rate
d
[D(x) Ps (x)] = v(x) Ps (x). and corresponds to the expected rate of return per
dx
unit of time from the stock, and δX is a random
10 The Ornstein–Uhlenbeck process is variable.
described by a drift term of the form v(x) = −αx,
If S is a random variable satisfying a log-normal
where α is a constant, and a constant diffusion
random walk, then the transformed variable
coefficient D(x) = D.
Y = ln S satisfies the probability density
11 An option is a derivative contract that gives !
the owner the right, but not the obligation, to buy 1 (Y − Y0 − µT + 12 σ 2 T )2
P (Y, T ; Y0 ) = √ exp − ,
or sell an underlying asset for an agreed price at or 2πT σ 2σ 2 T
before a specific time. In these transactions, the where Y0 = ln S0 and S0 is the value of S at t = 0.
agreed price K is called the strike price, and T is
the expiration date. 14 The prices of a European call option
c(S0 , K, T ) and put option p(S0 , K, T ) are given by
There are two types of options: the Black–Scholes formulas
• a call option gives the owner of the option the c(S0 , K, T ) = S0 N (d1 ) − Ke−rT N (d2 ),
right to buy an underlying asset at a strike price p(S0 , K, T ) = Ke−rT N (−d2 ) − S0 N (−d1 ),
• a put option gives the owner of the option the where r is the interest rate, σ is the stock price
right to sell an underlying asset at a strike price. volatility, T is the expiration date, and
In addition, put or call options are further
ln(S0 /K) + r + 21 σ 2 T √


classified into other categories: d1 = √ , d2 = d1 − σ T .
σ T
• a European option gives the owner of the
option the right to buy or sell an underlying The function N (x) is given by the cumulative
asset at a strike price only at the expiration date distribution function of the normal probability
density function defined below.
• an American option gives the owner of the
option the right to buy or sell an underlying 15 There are two ways to find the numerical
asset at a strike price at any time until the value of the cumulative normal distribution
expiration date. function
Z x
1 2
12 The pay-off of an European option with N (x) = √ e−y /2 dy,
strike price K, expiration date T and stock price 2π −∞
S(T ) is at a given point x.
Ψ(S, K) = max{S(T ) − K, 0} (1) Numerical approach by using a computer.
Most current software has a built-in function
if it is a call option, and that computes N (x) or the error function erf(x).
Ψ(S, K) = max{K − S(T ), 0} For example, in Maxima you can either use the
command erf to evaluate the error function and
if it is a put option. then find N (x) as
13 Consider the change in the price S occurring 1
 
x

in a short time ∆t to be δS. The log-normal N (x) = 1 + erf √ ,
2 2

58
 Unit 12 summary

or first load the distrib package using For negative values of x, we use the symmetry of
load(distrib); and then cdf normal(x,0,1) the normal distribution, which gives
will give N (x). N (−x) = 1 − N (x). For example,
In Excel and MATLAB you can use NORMDIST N (−0.12) = 1 − N (0.12) = 1 − 0.5478 = 0.4522.
and normcdf, respectively, directly to find N (x). Values can be interpolated in the same way as
(2) Tables. Below there is a table that gives the for positive x, so
numerical values of N (x) for x ≥ 0. These values N (−0.1234) = N (−0.12) − 0.34[N (−0.12) − N (−0.13)]
are tabulated at intervals of 0.01 in x. For = 0.4522 − 0.34 × (0.4522 − 0.4483)
intermediate values, the table should be used = 0.4509.
with interpolation. For example:
N (0.1234) = N (0.12) + 0.34[N (0.13) − N (0.12)]
= 0.5478 + 0.34 × (0.5517 − 0.5478)
= 0.5491.

59
Handbook

Probabilities for the standard normal distribution N (z) = P (Z ≤ z)

z 0 1 2 3 4 5 6 7 8 9
0.0 0.5000 0.5040 0.5080 0.5120 0.5160 0.5199 0.5239 0.5279 0.5319 0.5359
0.1 0.5398 0.5438 0.5478 0.5517 0.5557 0.5596 0.5636 0.5675 0.5714 0.5753
0.2 0.5793 0.5832 0.5871 0.5910 0.5948 0.5987 0.6026 0.6064 0.6103 0.6141
0.3 0.6179 0.6217 0.6255 0.6293 0.6331 0.6368 0.6406 0.6443 0.6480 0.6517
0.4 0.6554 0.6591 0.6628 0.6664 0.6700 0.6736 0.6772 0.6808 0.6844 0.6879
0.5 0.6915 0.6950 0.6985 0.7019 0.7054 0.7088 0.7123 0.7157 0.7190 0.7224
0.6 0.7257 0.7291 0.7324 0.7357 0.7389 0.7422 0.7454 0.7486 0.7517 0.7549
0.7 0.7580 0.7611 0.7642 0.7673 0.7704 0.7734 0.7764 0.7794 0.7823 0.7852
0.8 0.7881 0.7910 0.7939 0.7967 0.7995 0.8023 0.8051 0.8078 0.8106 0.8133
0.9 0.8159 0.8186 0.8212 0.8238 0.8264 0.8289 0.8315 0.8340 0.8365 0.8389
1.0 0.8413 0.8438 0.8461 0.8485 0.8508 0.8531 0.8554 0.8577 0.8599 0.8621
1.1 0.8643 0.8665 0.8686 0.8708 0.8729 0.8749 0.8770 0.8790 0.8810 0.8830
1.2 0.8849 0.8869 0.8888 0.8907 0.8925 0.8944 0.8962 0.8980 0.8997 0.9015
1.3 0.9032 0.9049 0.9066 0.9082 0.9099 0.9115 0.9131 0.9147 0.9162 0.9177
1.4 0.9192 0.9207 0.9222 0.9236 0.9251 0.9265 0.9279 0.9292 0.9306 0.9319
1.5 0.9332 0.9345 0.9357 0.9370 0.9382 0.9394 0.9406 0.9418 0.9429 0.9441
1.6 0.9452 0.9463 0.9474 0.9484 0.9495 0.9505 0.9515 0.9525 0.9535 0.9545
1.7 0.9554 0.9564 0.9573 0.9582 0.9591 0.9599 0.9608 0.9616 0.9625 0.9633
1.8 0.9641 0.9649 0.9656 0.9664 0.9671 0.9678 0.9686 0.9693 0.9699 0.9706
1.9 0.9713 0.9719 0.9726 0.9732 0.9738 0.9744 0.9750 0.9756 0.9761 0.9767
2.0 0.9772 0.9778 0.9783 0.9788 0.9793 0.9798 0.9803 0.9808 0.9812 0.9817
2.1 0.9821 0.9826 0.9830 0.9834 0.9838 0.9842 0.9846 0.9850 0.9854 0.9857
2.2 0.9861 0.9864 0.9868 0.9871 0.9875 0.9878 0.9881 0.9884 0.9887 0.9890
2.3 0.9893 0.9896 0.9898 0.9901 0.9904 0.9906 0.9909 0.9911 0.9913 0.9916
2.4 0.9918 0.9920 0.9922 0.9925 0.9927 0.9929 0.9931 0.9932 0.9934 0.9936
2.5 0.9938 0.9940 0.9941 0.9943 0.9945 0.9946 0.9948 0.9949 0.9951 0.9952
2.6 0.9953 0.9955 0.9956 0.9957 0.9959 0.9960 0.9961 0.9962 0.9963 0.9964
2.7 0.9965 0.9966 0.9967 0.9968 0.9969 0.9970 0.9971 0.9972 0.9973 0.9974
2.8 0.9974 0.9975 0.9976 0.9977 0.9977 0.9978 0.9979 0.9979 0.9980 0.9981
2.9 0.9981 0.9982 0.9982 0.9983 0.9984 0.9984 0.9985 0.9985 0.9986 0.9986
3.0 0.9987 0.9987 0.9987 0.9988 0.9988 0.9989 0.9989 0.9989 0.9990 0.9990
3.1 0.9990 0.9991 0.9991 0.9991 0.9992 0.9992 0.9992 0.9992 0.9993 0.9993
3.2 0.9993 0.9993 0.9994 0.9994 0.9994 0.9994 0.9994 0.9995 0.9995 0.9995
3.3 0.9995 0.9995 0.9995 0.9996 0.9996 0.9996 0.9996 0.9996 0.9996 0.9997
3.4 0.9997 0.9997 0.9997 0.9997 0.9997 0.9997 0.9997 0.9997 0.9997 0.9998
3.5 0.9998 0.9998 0.9998 0.9998 0.9998 0.9998 0.9998 0.9998 0.9998 0.9998
3.6 0.9998 0.9998 0.9999 0.9999 0.9999 0.9999 0.9999 0.9999 0.9999 0.9999
3.7 0.9999 0.9999 0.9999 0.9999 0.9999 0.9999 0.9999 0.9999 0.9999 0.9999
3.8 0.9999 0.9999 0.9999 0.9999 0.9999 0.9999 0.9999 0.9999 0.9999 0.9999
3.9 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
4.0 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000

60
 Index

Index
area-preserving map, 50(25)
Index entries referring to unit summaries are in ,
Argand diagram, 5
the form ‘page number(item number)’.
argument of a complex number, 5
associated homogeneous equation, 35(11)
absorbing boundary condition, 57(1) asymptote, 26
AC − B 2 test, 27 asymptotic stability, 39(6)
acceleration, 22 asymptotically periodic orbit, 50(3)
action, 45(3), 46(12) attracting fixed point, 48(8)
acute angle, 10 attractor, 51(8)
addition formulas, trigonometric, 12 auxiliary equation, 35(9)
addition of vectors, 17
advection–diffusion equation, 52(10) baker’s map, 51(11)
American option, 58(11) base of an exponential function, 7
amplitude basin of attraction, 48(11)
of a sinusoidal function, 13 Bessel functions, 56(14)
of an oscillation, 36(1) Bessel’s equation, 56(14)
angle subtended by an arc of a circle, 16 Black–Scholes formulas, 58(14)
angle sum boundary condition
of a quadrilateral, 15 absorbing, 57(1)
of a triangle, 15 reflecting, 57(2)
angular frequency, 13, 36(1) boundary curve, 53(15)
angular momentum, 47(17) bounded above, 25
ansatz, 56(16), 56(17) bounded below, 25
antiderivative, 28 box-counting dimension, 51(10)
arbitrary constant of integration, 28 brachistochrone, 43(5)
arc length, 16 call option, 58(11)
arc of a circle, 16 Cartesian coordinates for two-dimensional space,
arccos, 11 14
arccosec, 11 Cartesian form of a complex number, 5
arccosh, 8 centre, 39(7), 39(9)
arccot, 11 centre of a circle, 15
arcsec, 11 chain rule
arcsin, 11 for partial derivatives, 24
arcsinh, 8 chain rule of differentiation, 23
arctan, 11 chaotic orbit, 40(12)
arctanh, 8 one-dimensional, 50(4)
area two-dimensional, 51(8)
of a circle, 16 characteristic equation, 22
of a parallelogram, 15 characteristic function, 54(4)
of a sector, 16 chord
of a triangle, 15 of a circle, 15
under a graph, 28 of a curve, 16
area integral circle, 15
in Cartesian coordinates, 33 circumference of a circle, 16
area-contracting map, 50(25) classification of an equilibrium point, 39(7),
area-preserving linear transformation, 49(20) 39(8)

61
Index

closed interval, 4 cot, 11

closed surface, 53(15) coupled oscillator, 36(2)
cobweb plot, 48(7) cross product of vectors, 21
coefficient in a system of linear equations, 18 cubic function, 6
coefficient matrix, 19 cumulative normal distribution function, 58(15)
coincident roots of a polynomial, 7 curve sketching, 25
column of a matrix, 19 cylindrical coordinates, 32
combined events, 41(5)
commutativity, 19 damping parameter, 38(12)
complementary function, 35(11) datum, 44(1)
de Moivre’s theorem, 5
complex conjugate, 5
decimal places, 4
complex number, 5
decreasing function, 25
complex-valued function, 23
definite integral, 28
component of a vector, 17
degree of a polynomial, 7
composite function, 6
degrees as a measure of angle, 10
composite rule for differentiation, 23
degrees of freedom, 45(7)
composition of functions, 6, 47(3)
dependent variable of a function, 6
concentration of molecules, 52(2)
derivative, 22
conjugate variables, 54(1)
of a complex-valued function, 23
conservation of energy, 47(16)
derivative rule for Fourier transforms, 54(8)
conservative force, 44(1), 46(10)
derived function, 22
conserved quantities, 46(15)
determinant, 21
constant, 6 diagonal matrix, 19
of integration, 28 diameter of a circle, 15
of motion, 34(6), 34(7) difference of two squares, 7
constant function, 6 differential equation, Fourier transform, 55(10)
constant multiple rule differentiation, 22
for differentiation, 23 of Fourier series, 37(10)
for integration, 28 rules, 23
constant of the motion, 39(9), 46(15) diffusion, 52(6)
continuity equation, 52(5) in a rectangular region, 56(13)
in three dimensions, 53(17) diffusion coefficient, 52(11), 52(7)
continuous function, 6 diffusion equation, 52(8)
continuous random variable, 42(14) in three dimensions, 53(21)
probability density function, 41(7) point source solution, 52(9)
contracting stretch, 49(20) solution in one dimension, 55(2)
convergence of a sequence of numbers, 3 dimension of a matrix, 19
convolution, 55(11) Dirichlet boundary condition, 56(7)
convolution theorem, 55(12) discrete random variable, 41(14)
coordinates discriminant of a quadratic equation, 7
cylindrical, 32 distance along a curve, 43(2)
of a vector, 17 distinct eigenvalues, 22
polar, 14 divergence, 53(17)
spherical, 32 in Cartesian coordinates, 31
cos, 11 in cylindrical coordinates, 32
cosec, 11 in polar coordinates, 31
cosh, 8 in spherical coordinates, 32

62
 Index

divergence theorem, 53(19) expectation value, 41(13)

dot product, 17 experiment, 41(2)
double-angle formulas, 13 expiration date, 58(11)
drift rate, 58(13) exponent of an exponential function, 7
drift velocity, 52(11) exponential behaviour, 7
exponential distribution, 41(12)
eigenfunction, 56(8) exponential form of a complex number, 6
eigenproblem, 56(8) exponential Fourier series, 37(9)
solution, 56(15) exponential function, 7
eigenvalue, 21, 56(8) extreme sensitivity to initial conditions, 40(12)
eigenvector, 21
extremum of a function of two variables, 27
eigenvector equations, 22
element
factorial, 26
of a matrix, 19
factorisation of polynomials, 7
of a vector, 17
Fermat’s principle, 44(10)
elementary linear transformations, 49(20)
Fick’s law, 52(7)
ellipse, 16
in three dimensions, 53(20)
entry of a matrix, 19
first integral, 44(9)
ε-covering set, 51(9)
first Lyapunov exponent, 51(7)
equal roots of a polynomial, 7
first moment, 42(16)
equality
first-order Taylor polynomial, 26
of matrices, 19
of vectors, 17 first-passage probability density, 57(4)
equation of a circle, 16 fixed point, 34(2), 38(3)
equation of motion of a one-dimensional map, 48(6), 48(9)
of a coupled oscillator, 36(2) of a two-dimensional map, 50(22), 50(24)
of a simple harmonic oscillator, 36(1) of an area-preserving map, 50(26)
equilateral triangle, 10, 15 flux, 52(14), 52(3)
equilibrium point, 34(2), 38(3) across an arbitrary surface, 53(16)
classification of, 39(7), 39(8) over a planar element, 52(13)
finding, 38(4) flux density, 52(4)
linearising a differential equation near, 38(5) vector, 53(16)
stability of, 34(3) Fokker–Planck equation, 57(8)
Euler–Lagrange equations, 43(7) formula method for roots of a quadratic
first integral, 44(9) equation, 7
for a mechanical system, 46(13) Fourier coefficients, 37(9)
for a particle moving in one dimension, 45(4) Fourier series
for n dependent variables, 45(6) differentiation of, 37(10)
special case, 43(8) exponential, 37(9)
Euler’s formula, 5 generalised, 56(12)
European option, 58(11) Fourier transform, 54(1)
pay-off, 58(12) of a differential equation, 55(10)
even symmetry rule for Fourier transforms, 54(8) pairs, 54(6)
events rules, 54(8)
combined, 41(5) function, 6
independent, 41(6) function notation (for derivatives), 22
mutually exclusive, 41(5) function of a function rule for differentiation, 23
expanding stretch, 49(20) functional, 43(1)

63
Index

Gaussian function, 41(10), 41(11) integral, 28

Gauss’s theorem, 53(19) of a Gaussian function, 41(11)
general solution of a differential equation, 35(11) integral of the motion, 46(15)
generalised coordinates, 45(7) integrand, 28
generalised Fourier series, 56(12) integrating factor method, 35(12)
generalised momentum, 47(17) integration, 28
generalised random walk, 57(7) by parts, 29
generalised velocities, 45(8) by substitution, 28
global maximum of a function, 25 rules, 28
global minimum of a function, 25 interval, 4
gradient (slope), 25 invariance of the Lagrangian, 47(18)
gradient operator invariant set, 49(13)
in Cartesian coordinates, 31 inverse Fourier transform, 54(1)
in cylindrical coordinates, 32 inverse function, 8
in polar coordinates, 31 inverse hyperbolic functions, 8
in spherical coordinates, 32 inverse of a matrix, 20
graph of a function, 6 inverse trigonometric functions, 11
graphs of common functions, 9–10 inversion rule for Fourier transforms, 54(8)
Greek alphabet, 3 invertible matrix, 20
growth/decay equation, 55(5) irrational number, 4
isosceles triangle, 15
half-open interval, 4
Hamilton’s principle Jacobian factor, 33
for a particle moving in one dimension, 44(3) Jacobian matrix, 38(5)
general formulation, 46(12)
heat equation, 53(22) Kaplan–Yorke dimension, 51(14)
Hénon map, 49(18) kinetic energy, 44(1)
hexagon, 15 in generalised coordinates, 45(8)
higher derivatives, 23 Kronecker delta, 37(8)
horizontal axis of a Cartesian coordinate system,
14 Lagrange’s equations, 45(4)
hyperbola, 16 for a mechanical system, 46(13)
hyperbolic functions, 8 for a particle moving in one dimension, 45(4)
hypotenuse, 12 Lagrangian, 45(3), 46(12)
Laplacian operator, 31, 53(21)
identity matrix, 19 leading diagonal, 19
ignorable coordinate, 47(17) Leibniz notation (for derivatives), 22
imaginary part of a complex number, 5 Leibniz’s integral rule, 43(4)
implicit differentiation, 23 length of a vector, 16
increasing function, 25 limit cycle, 39(10)
indefinite integral, 28 half-stable, 40(10)
independent events, 41(6) stable, 39(10)
independent variable of a function, 6 unstable, 39(10)
index of an exponential function, 7 limit of a sequence of numbers, 3
initial condition, 37(7) linear combination, 20
initial value of a map, 48(4) linear dependence of vectors, 20
in-phase normal mode, 36(4) linear function, 6
integer, 4 linear independence of vectors, 20

64
 Index

linear map minimum of a function, 25, 27

one-dimensional, 48(5) modulus
two-dimensional, 49(19) of a complex number, 5
linear momentum, 47(17) of a vector, 16
linear transformation, 49(19) molecules, 52(2)
linearised equation for small separations, 51(6) moments of a random variable, 42(16)
linearising, 38(5) monomial, 54(7)
linearity rule for Fourier transforms, 54(8) monomial rule for Fourier transforms, 54(8)
Lissajous curve, 36(5) mutually exclusive events, 41(5)
local maximum, 25
of a function, 27 natural angular frequency, 38(12)
local minimum, 25 natural logarithm function, 8
of a function, 27 Neumann boundary condition, 56(7)
log-normal random walk, 58(13) neutral stability, 39(6)
logarithm function, 8 Newtonian notation (for time derivatives), 22
logistic equation, 34(1) Newton’s notation (for time derivatives), 22
logistic map, 49(12) n-gon, 15
loop, 40(14) node, 39(7)
Lorentzian function, 54(2) non-negativity of a probability density function,
Lotka–Volterra equations, 38(2) 41(8)
rescaled, 38(2) non-singular matrix, 20
lower bound of a function, 25 non-transform derivative rule, 55(1)
lower limit of integration, 28 non-asymptotic stability, 39(6)
lower triangular matrix, 19 normal distribution, 41(10)
Lyapunov dimension, 51(14) table of probabilities, 60
Lyapunov exponent, 40(13), 50(1) normal mode, 36(3), 36(6)
first, 51(7) in-phase, 36(4)
in one dimension, 50(2) phase-opposed, 36(4)
in two dimensions, 51(7) normalisation, 56(10)
second, 51(7) of a probability density function, 41(8)
normalised eigenfunction, 56(10)
magnitude of a vector, 16 nth derivative, 23
main diagonal, 19 nth-order polynomial, 7
matrix, 19 nth-order Taylor approximation, 26
matrix addition, 19 nth-order Taylor polynomial, 26
matrix multiplication, 19 number of molecules, 52(2)
matrix product, 19 numerical iteration, 40(15)
maximum of a function, 25, 27
mean exit time, 57(5) obtuse angle, 10
in a finite domain, 57(6) odd function, 29
mean of a random walk, 51(1) odd symmetry rule for Fourier transforms, 54(8)
mean squared displacement, 52(9) open interval, 4
mean value, 41(13) open surface, 53(15)
of a continuous random variable, 42(15) operation of matrices, 19
of the exponential distribution, 41(12) optical action, 44(10)
of the normal distribution, 41(10) optical Lagrangian, 44(10)
method of undetermined coefficients, 35(11) option (contract), 58(11)
minimal surface of revolution, 43(5) orbit, 38(1)

65
Index

chaotic, 40(12) of degree n, 7

of a one-dimensional map, 48(4) position vector, 18
order potential, 44(1)
of a derivative, 23 potential energy, 44(1), 46(11)
of a matrix, 19 of a particle in three dimensions, 46(10)
of a polynomial, 7 potential energy function, 46(10)
of a Taylor approximation, 26 power of an exponential function, 7
of a Taylor polynomial, 26 pre-image of a set, 51(13)
oriented area element, 52(12) premultiplication, 19
origin of a Cartesian coordinate system, 14 principal value
Ornstein–Uhlenbeck process, 58(10) of a polar coordinate angle, 14
orthogonal functions, 56(9) of the argument of a complex number, 5
orthogonal vectors, 18 principle of superposition, 35(10)
orthogonality relations for exponentials, 37(8) probability, 41(3)
orthonormal set, 56(11) in successive trials, 41(6)
oscillatory function, 13 of combined events, 41(5)
of mutually exclusive events, 41(5)
parabola, 6, 16
parallelogram, 15 probability density, 41(7)
partial derivative, 43(6) probability density function, 41(7)
partial derivatives, 24 probability flux, 57(8)
particular integral of an inhomogeneous procedure
differential equation, 35(11) applying the scaling rule, 54(9)
particular solution, 37(11) calculating a propagator, 55(3)
pay-off, 58(12) calculating statistics of a random walk,
pentagon, 15 42(22)
perfect square, 7 calculating variance and standard deviation,
period of a sinusoidal function, 13 42(19)
periodic function, 13 checking the stability of fixed points of
period-2 orbit, 49(14) one-dimensional motion, 34(4)
period-k orbit, 49(15) classifying a fixed point of a one-dimensional
phase constant, 13, 36(1) map, 48(9)
phase flow of an autonomous one-dimensional classifying a fixed point of an
system, 38(1) area-preserving map, 50(26)
phase path, 34(5) classifying an equilibrium point of a linear
phase plane, 34(5) system, 39(7)
phase point, 34(5) classifying an equilibrium point of a
phase space, 34(5) nonlinear system, 39(8)
phase trajectory, 34(5) computing the cumulative normal
phase-opposed normal mode, 36(4) distribution function, 58(15)
Poincaré return map, 47(2) constructing a cobweb plot, 48(7)
Poincaré section, 47(2) determining the equations of motion for a
Poincaré–Bendixson theorem, 40(11) system of particles, 46(14)
point of inflection, 25 expressing total kinetic energy in terms of
polar coordinates, 14 generalised coordinates, 45(9)
polygon, 15 finding a particular solution using Fourier
polynomial, 7 series, 37(11)
function, 7 finding and classifying a fixed point of a

66
 Index

two-dimensional map, 50(24) random walk, 40(1)

finding and using constants of motion to generalised, 57(7)
locate centres, 39(9) mean, 51(1)
finding equilibrium points, 38(4) variance, 51(1)
finding potential energy, 46(11) rational number, 4
finding the equation of motion of a particle real number, 4
constrained to move on a fixed path, real part of a complex number, 5
45(5) realisation, 41(2)
finding the general solution of a rectangle, 15
second-order inhomogeneous linear recurrence relation for simple random walk
constant-coefficient equation, 35(11) probability, 42(21)
finding the solution subject to initial reflecting boundary condition, 57(2)
conditions, 37(7) regular polygon, 15
integrating factor method, 35(12) repeated eigenvalues, 22
linearising a system of differential equations repeated root of a polynomial, 7
near an equilibrium point, 38(5) repelling fixed point, 48(8)
numerical iteration, 40(15) right angle, 10
separation of variables, 55(4) right-angled triangle, 15
solving a first-order differential equation by root of a polynomial, 7
separation of variables, 34(8) rotation, 49(20)
solving a linear oscillatory system, 36(6) rounding, 4
solving a linear second-order row of a matrix, 19
constant-coefficient homogeneous
differential equation, 35(9) saddle point, 39(7)
solving differential equations using Fourier of a function, 27
transforms, 55(10) scalar, 17, 19
stability test for a period-k orbit, 49(16) scalar flux density, 52(4)
testing for a constant of motion, 34(7) scalar multiplication, 17, 19
using an ansatz, 56(17) scalar product, 17
product of random variables, 42(20) scaling rule for Fourier transforms, 54(8), 54(9)
product rule for differentiation, 23 scientific notation, 4
propagator sec, 11
calculating, 55(3) second derivative, 23
in solution of the diffusion equation, 55(2) second Lyapunov exponent, 51(7)
put option, 58(11) second moment, 42(16)
Pythagoras’ theorem, 12 second-order Taylor polynomial, 26
separation of variables, 34(8), 55(4)
quadrants of the plane, 14 shear, 49(20)
quadratic approximation, 26 SHM equation, 56(6)
quadratic function, 6 significant figures, 4
quadrilateral, 15 simple harmonic motion, 36(1)
quotient rule for differentiation, 23 simple harmonic motion equation, 56(6)
simple harmonic oscillator, 36(1)
radian, 10 simple random walk, 40(1)
radius of a circle, 15 sin, 11
random variable sinc function, 54(5)
continuous, 42(14) singular matrix, 20
discrete, 41(14) sinh, 8

67
Index

sink, 48(8) steady state, 53(18)

sinusoid, 13 strange attractor, 51(8)
sinusoidal function, 13 stretch, 49(20)
size of a matrix, 19 strike price, 58(11)
skinny baker’s map, 51(12) successive trials, 41(6)
slope of a straight line, 6 sum
small-angle approximations, 27 of functions, 6
smooth function, 25 of probabilities, 41(4)
solution of random variables, 42(20)
of a homogeneous differential equation, 35(9) sum rule
of an inhomogeneous differential equation, for differentiation, 23
35(11) for integration, 28
source, 48(8) summation notation, 3
spectrum, 56(8) superstable fixed point, 48(10)
spherical coordinates, 32 surface of section, 47(2)
spiral, 39(7) survival probability, 57(3), 57(4), 57(5)
square, 15 symmetric matrix, 19
square matrix, 19 symmetry and conservation, 47(18)
stability in a two-dimensional map, 50(23) symmetry rules for Fourier transforms, 54(8)
stability of a fixed point, 34(4), 38(1), 39(6)
tan, 11
asymptotic, 39(6)
tangent approximation, 26
neutral, 39(6)
tangent to a curve, 16
non-asymptotic, 39(6) tanh, 8
stability test for a period-k orbit, 49(16) Taylor approximation, 26
stable equilibrium point, 34(3) Taylor polynomial of a function of one variable,
stable fixed point, 48(8) 26
of a two-dimensional map, 50(23) Taylor series of a function of one variable, 26
stable node, 39(7) tent map, 50(5)
stable spiral, 39(7) thermal diffusivity, 53(22)
standard derivatives, 24 third derivative, 23
standard deviation, 41(10), 42(18) time-T map, 47(1)
standard integrals, 29 top-hat function, 54(3)
standard normal distribution, table of total derivative, 43(6)
probabilities, 60 total energy, 44(1)
stationary curve, 43(3) trace of a matrix, 19
stationary function, 43(3) transient solution, 38(12)
stationary functional, 43(3) translation along k rule for Fourier transforms,
stationary path, 43(3) 54(8)
stationary point translation along x rule for Fourier transforms,
of a function of two variables, 27 54(8)
stationary point of a function of one variable, 25 transpose, 19
stationary probability density function, 58(9) transpose operator, 17, 19
statistics trial, 41(2)
of a random walk, 42(22) trial solution for a linear differential equation,
of sums and products of random variables, 35(11)
42(20) triangle, 15
steady flow, 53(18) trigonometric addition formulas, 12

68
 Index

trigonometric double-angle formulas, 13 variance, 42(17), 42(19)

trigonometric functions, 11 of a random walk, 51(1)
trigonometric identities, 12 vector, 16
vector equation of a line, 18
uniform distribution, 41(9)
vector flux density, 53(16)
unit normal convention for a closed surface,
velocity, 22
53(15)
vertical axis of a Cartesian coordinate system, 14
unit normal vector, 52(12)
volatility, 58(13)
unit vector, 17
unstable equilibrium point, 34(3), 39(6)
x-axis of a Cartesian coordinate system, 14
unstable fixed point, 48(8)
(x, y)-plane, 14
of a two-dimensional map, 50(23)
unstable node, 39(7)
y-axis of a Cartesian coordinate system, 14
unstable spiral, 39(7)
y-intercept of a straight line, 6
upper bound of a function, 25
upper limit of integration, 28
zero function, 6
upper triangular matrix, 19
zero matrix, 19
variable, 6 zero vector, 17

69