100% found this document useful (1 vote)

2K views

MST209 Mathematical Methods and Models

This publication forms part of an Open University course. It introduces the topic of differential equations. The recommended study pattern is to study one section per study session.

Uploaded by

abusmoh

Available Formats

Download as PDF, TXT or read online on Scribd

100% found this document useful (1 vote)

2K views

MST209 Mathematical Methods and Models

This publication forms part of an Open University course. It introduces the topic of differential equations. The recommended study pattern is to study one section per study session.

Uploaded by

abusmoh

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 39

MST209 Mathematical methods

and models

First-order differential equations

�

��

� ��
This unit introduces the topic of differential equations. It is an important
field of study, and several subsequent units are also devoted to it. There are
many applications of differential equations throughout the course.
The subject is developed without assuming that you have come across it
before, but the unit assumes that you have previously had a basic grounding
in calculus. In particular, you will need to have a good grasp of the basic
rules for differentiation and integration. (These were revised in Unit 1 of
this course.)
From the point of view of later studies, Sections 3 and 4 contain the most
important material.
The recommended study pattern is to study one section per study session,
and to study the sections in the order in which they appear.
You will need the computer algebra package for the course for Subsection 2.3
and for all of Section 5. The computer work for Subsection 2.3 may be
postponed until later (for example until your study of Section 5) without
affecting your ability to study the subsequent sections.

This publication forms part of an Open University course. Details of this and other Open Uni-
versity courses can be obtained from the Course Information and Advice Centre, PO Box 724,
The Open University, Milton Keynes, MK7 6ZS, United Kingdom: tel. +44 (0)1908 653231, e-mail
[email protected]
Alternatively, you may visit the Open University website at http://www.open.ac.uk where you can
learn more about the wide range of courses and packs oﬀered at all levels by The Open University.
To purchase a selection of Open University course materials, visit the webshop at www.ouw.co.uk,
or contact Open University Worldwide, Michael Young Building, Walton Hall, Milton Keynes,
MK7 6AA, United Kingdom, for a brochure: tel. +44 (0)1908 858785, fax +44 (0)1908 858787,
e-mail [email protected]

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

First published 2005.
Copyright
c 2005 The Open University
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, record-
ing or otherwise, without written permission from the publisher or a licence from the Copyright
Licensing Agency Ltd. Details of such licences (for reprographic reproduction) may be obtained
from the Copyright Licensing Agency Ltd, 90 Tottenham Court Road, London W1T 4LP.
Open University course 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 course 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 course 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 re-transmit, 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 the Open University TEX System.
Printed and bound in the United Kingdom by Martins the Printers Ltd, Berwick-upon-Tweed,
TD15 1RS.
ISBN 0 7492 6698 8
1.1
��

��

An important class of the equations that arise in mathematics consists of

those that feature the rates of change of one or more variables with respect to
one or more others. These rates of change are expressed mathematically by
derivatives, and the corresponding equations are called differential equations.
Equations of this type crop up in a wide variety of situations. They are
found, for example, in models of physical, electronic, economic, demographic
and biological phenomena.
First-order differential equations, which are the particular topic of this unit,
feature derivatives of order one only; that is, if the rate of change of variable
y with respect to variable x is involved, the equations feature dy/dx but not
d2 y/dx2 , d3 y/dx3 , etc.
When a differential equation arises, it is usually an important aim to solve
the equation. For an equation that features the derivative dy/dx, this entails
expressing the dependent variable y directly in terms of the independent
variable x. The solution process requires the effect of the derivative to be
‘undone’. The reversal of differentiation is achieved by integration, so it is
to be expected that integration will feature significantly in the methods of
solution for differential equations.
Integration can be attempted either symbolically, to obtain an exact formula
for the integral, or numerically, to give approximate numerical values from
which the integral can be tabulated or graphed. The same two approaches
can therefore be tried to obtain solutions of differential equations, and both
are introduced in this unit.
Section 1 considers in detail one example of how a differential equation
arises in a mathematical model. This is followed by some basic definitions
and terminology associated with differential equations and their solutions.
We also note how errors and accuracy are defined.
Section 2 starts by looking at the direction field associated with a first-order
differential equation. This is a device for visualizing the overall behaviour of
the differential equation and of its solutions, and leads to a basic numerical
method of solution known as Euler’s method. Both direction fields and
Euler’s method are implemented in a computer subsection.
Section 3 turns to analytic (that is, symbolic) methods of solution, consid-
ering first direct integration and then separation of variables.
Section 4 describes a further analytic approach to solving differential equa-
tions, called the integrating factor method. It applies only to equations that
are linear. Linear first-order differential equations are important in their
own right, but also give valuable clues on how to solve linear second-order
differential equations, which are the subject of Unit 3.
In Section 5 you will see how each of the analytic methods from Sections 3
and 4 can be implemented on your computer.

�
��

Subsection 1.1 develops a mathematical model for a speciﬁc situation which

leads naturally to a first-order differential equation. A key step in deriving
this equation is to apply the input–output principle, which is a useful device
for building relations between variables.
Subsection 1.2 addresses what is meant by the term ‘solution’ in the context
of first-order differential equations, and brings out the distinction between
the general solution and the various possible particular solutions. The spec-
ification of a constraint, or initial condition, usually permits us to find a
unique function that is a particular solution of the differential equation and
also satisfies the constraint.
The short Subsection 1.3 provides the definition and description of numerical
errors, in anticipation of Euler’s method in Section 2.

��

In the course you will meet many examples of differential equations. Fre-
quently these arise from studying the motion of physical objects, but we
shall start with an example drawn from biology and show how this leads
naturally to a particular differential equation.
Suppose that we are interested in the size of a particular population, and in
how it varies over time. The first point to make is that any population size is
measured in integers (whole numbers), so it is not clear how differentiation
will be relevant. (Differentiable functions must be continuous, and therefore
defined on an interval of real numbers in R.) Nevertheless, if the population
is large, say in the hundreds of thousands, a change of one unit will be
relatively very small, and in these circumstances we may choose to model
the population size as a continuous function of time. We shall write this
function as P (t), and our task is to show how P (t) may be described by a
differential equation.
Let us assume a fixed starting time (which we shall label t = 0). If the
population is not constant, then there will be ‘leavers’ and ‘joiners’. For
example, in a population of humans in a particular country, the former
will be those who die or emigrate, whilst the latter represent births and
immigrants.
It is usual to express birth rates as a proportion of the current population
size. For example, the UK Office for National Statistics quotes birth rates in
various age groups as a number per 1000 women. Death rates are specified
in a similar way. To emphasize that these rates are expressed as a proportion
of the current population, we shall use the terms ‘proportionate birth rate’ Note that in our model the
and ‘proportionate death rate’. proportionate birth rate is
expressed as a proportion of
For our simple model we shall ignore immigration and emigration, and con- the whole population, not
centrate solely on births and deaths. Denote the proportionate birth rate just the number of women.
by b and the proportionate death rate by c. Then, in a short interval of
time δt, we would expect
number of births bP (t)δt, (1.1)
number of deaths cP (t)δt, (1.2)
where P (t) is the population size at time t.

�
��

At this stage, we seek some relationship between the chosen variables. In

order to ﬁnd this, we make use of the input–output principle, which can
be expressed as
accumulation = input − output .

This principle applies to any quantity whose change, over a given time in-
terval, is due solely to the specified input and output.
The accumulation δP of population over the time interval δt is the popu-
lation at the end of the interval minus the population at the start of the
interval; that is,
δP = P (t + δt) − P (t).
The input is the number of births (Equation (1.1)), and the output is the
number of deaths (Equation (1.2)). The input–output principle now enables
us to express the accumulation δP of the population over the time interval
δt as
δP bP (t)δt − cP (t)δt = (b − c)P (t)δt.
Dividing through by δt, we obtain
δP
(b − c)P (t).
δt
The approximations involved in deriving this equation become progressively
more accurate for shorter time intervals. So, finally, by letting δt tend to
zero, we obtain This is the step that requires
P to be a continuous (rather
dP
= (b − c)P (t). than discrete) function of t.
dt
(This follows because
dP P (t + δt) − P (t)
= lim
dt δt→0 δt
is the definition of the derivative of P .)
This is a differential equation because it describes dP /dt rather than the
eventual object of our interest (which is P itself). The purpose of this unit
is to enable you to solve a wide variety of such equations.
Of course, we can simplify the above equation slightly by using the pro-
P
portionate growth rate r, which is the difference between the proportionate
birth and death rates: r = b − c. Then our model becomes
dP
= rP.
dt r>0
For very simple population models, r is taken to be a constant. As we shall
see, this leads to a prediction of exponential growth (or, if r < 0, decay)
in population size with time, as illustrated in Figure 1.1. This may be a t
very good approximation for certain populations, but it cannot be sustained
indefinitely if r > 0. ��
In practice, both the proportionate birth rate and the proportionate death
rate will vary, and so therefore will the proportionate growth rate. It turns
out to be convenient to model these changes as being dependent on the
population size, so that the proportionate growth rate r becomes a function
of P . The justification for this is as follows. When the population is low, one
may assume that there is potential for it to grow (assuming a reasonable
environment). The proportionate growth rate should therefore be high.
However, as the population grows, there will be competition for resources.

�
��

Thus the proportionate growth rate will decline, and in this way unlimited
(exponential) growth does not occur.
A particularly useful model arises from taking r(P ) to be a decreasing linear
function of P . We shall write this as
� �
P You will see later why this
r(P ) = k 1 − , (1.3) particular form is chosen.
M
where k and M are positive constants. Looking at this formula, you can
see that the proportionate growth rate r decreases linearly with P , from the
value k (when P = 0) to 0 (when P = M ).
Using this expression for r, the above differential equation satisfied by P
becomes
� �
dP P
= kP 1 − . (1.4)
dt M
This is well known to biologists as the logistic equation — we shall consider
it further in Section 2, and see how to solve it in Section 3. For now, we have
achieved our objective of showing that differential equations arise naturally
in modelling the real world.

��

Suppose that a population obeys the logistic model (with the proportionate
growth rate given by Equation (1.3)), and that you are given the following
information. When P = 10 the proportionate growth rate is 1, and when
P = 10 000 the proportionate growth rate is 0. Find the corresponding
values of k and M .

��

This subsection introduces some of the fundamental concepts associated
with differential equations. First, however, you are asked to recall some
terminology and notation from your previous exposure to calculus. Some of this terminology and
notation was discussed in
The derivative of a variable y with respect to another variable x is denoted Unit 1.
in Leibniz notation by dy/dx. In this derivative expression we refer to y as
the dependent variable and to x as the independent variable.
There are other notations in use for derivatives. If the relation between
variables x and y is expressed in terms of a function f , so that y = f (x),
then the derivative may be written in function notation as f (x).
A further notation, attributed to Newton, is restricted to cases in which
the independent variable is time, denoted by t. The derivative of y = f (t)
could be written in this case as ẏ, in which the dot over the y stands for the
d/dt of Leibniz notation. Thus we may express this derivative in any of the
equivalent forms
dy
= ẏ = f (t).
dt
Further derivatives are obtained on differentiating this first derivative. The
second derivative of y = f (t) could be represented by any of the forms
d2 y
= y¨ = f (t).
dt2
These possible notations have different strengths and weaknesses, and which
is most appropriate in any situation depends on the purpose at hand. You
will see all of these notations employed at various times during the course.

�
��

It is common practice in applied mathematics to reduce the proliferation of

symbols as far as possible. One aspect of this practice is that we often avoid
allocating separate symbols to variables and to associated functions. Thus,
in place of the equation y = f (t) (where y and t denote variables, and f
denotes the function that relates them), we could write y = y(t), which is Strictly speaking, this is an
read as ‘y is a function of t’. (You have seen examples of this in the previous abuse of notation, since there
subsection.) is ambiguity as to exactly
what the symbol y represents:
The following definitions explain just what are meant by a differential equa- it is a variable on the
tion, by the order of such an equation, and by a solution of it. left-hand side of y = y(t) but
a function on the right-hand
side. However, it is a very
convenient abuse.
��
(a) A differential equation for y = y(x) is an equation that relates
the independent variable x, the dependent variable y, and one or
more derivatives of y.
(b) The order of such a differential equation is the order of the high-
est derivative that appears in the equation. Thus a first-order
differential equation for y = y(x) features only the first derivative,
dy/dx.
(c) A solution of such a differential equation is a function y = y(x)
that satisfies the equation.

These deﬁnitions have been framed in terms of an independent variable x

and a dependent variable y. You should be able to translate them to apply
to any other independent and dependent variables. Thus Equation (1.4) is
a differential equation in which t is the independent variable and P is the
dependent variable. It is a first-order equation, since dP /dt appears in it
but higher derivatives such as d2 P /dt2 do not. By contrast, the differential
equation
d2 y
3 + y 2 sin x = x2
dx2
is of second order, since the second derivative d2 y/dx2 appears in it but
higher derivatives do not.
The topic of this unit is first-order differential equations. Moreover, it con- Second-order differential
centrates upon first-order equations that can be expressed (possibly after equations are the subject of
some algebraic manipulation) in the form Unit 3.

dy
= f (x, y). Equation (1.4) is of� this form,
�
dx P
The right-hand side here stands for an expression involving both, either or with f (t, P ) = kP 1 − .
M
neither of the variables x and y, but no other variables and no derivatives.
According to the definition above, a function has to satisfy a differential
equation in order to be regarded as a solution of it. The differential equation
is satisfied by the function provided that when the function is substituted This substitution includes the
into the equation, the left- and right-hand sides of the equation give an requirement that the function
identical expression. should be differentiable
(i.e. that it should have a
You are asked to verify in the next exercise that several functions are solu- derivative) at all points where
tions of corresponding first-order differential equations. Later in the unit, it is claimed to be a solution.

�
��

you will see how all of these diﬀerential equations may be solved; but even

when a solution has been deduced, it is worth checking in the manner of this

exercise (i.e. by substitution) that the supposed solution is indeed correct.

��

Verify that each of the functions given below is a solution of the correspond
-
ing diﬀerential equation.

dy
*(a) y = 2ex − (x2 + 2x + 2); = y + x2 . Remember that an asterisk
dx denotes an exercise (or part
dy of one) that is considered
(b) y = 12 x2 + 32 ; = x.
dx particularly important.

2
*(c) u = 2ex /2 ; u
= xu.
�
27 − x2 √ √ dy x
(d) y = (−3 3 < x < 3 3); =− (y = 0). Note that the restriction
3 dx 3y y= 0 placed on the
*(e) y = t + e−t ; ẏ = −y + t + 1. diﬀerential equation in
part (d) is necessary to ensure
*(f) y = t + Ce−t ; ẏ = −y + t + 1. (Here C is an arbitrary constant.) that −x/3y is well deﬁned.

In the last two parts of Exercise 1.2 you were asked to verify that
y = t + e−t and y = t + Ce−t
are solutions of the differential equation ẏ = −y + t + 1, where in the second
case C is an arbitrary constant. In saying that C is arbitrary, we mean that
it can assume any real value. Whatever number is chosen for C, the corre-
sponding expression for y(t) is always a solution of the differential equation.
The particular function y = t + e−t is just one example of such a solution,
obtained by choosing C = 1.
This demonstrates that solutions of a differential equation can exist in pro-
fusion; as a result, we need terms to distinguish between the totality of all
these solutions for a given equation and the individual solutions that are
completely specified.

��
(a) The general solution of a diﬀerential equation is the collection
of all possible solutions of that equation.
(b) A particular solution of a diﬀerential equation is a single solu-
tion of the equation, and consists of a solution function whose rule
contains no arbitrary constant.

In many cases it is possible to describe the general solution of a ﬁrst-order

diﬀerential equation by a single formula involving an arbitrary constant. For
example, y = t + Ce−t is the general solution of the equation ẏ = −y + t + 1;
this means that not only is y = t + Ce−t a solution whatever the value of C,
but also every particular solution of the equation may be obtained by giving
C a suitable value.

�
��

��

(a) Verify that, for any value of the constant C, the function y = C − 13 e−3x
is a solution of the diﬀerential equation
dy
= e−3x .
dx
(b) Verify that, for any value of the constant C, the function u = Cet − t − 1
is a solution of the diﬀerential equation
u̇ = t + u.
(c) Verify that, for any value of the constant C, the function
CM ekt
P =
1 + Cekt

is a solution of Equation (1.4) on page 7.

As you have seen, there are many solutions of a differential equation. How-
ever, a particular solution of the equation, representing a definite relation-
ship between the variables involved, is often what is needed. This is achieved
by using a further piece of information in addition to the differential equa-
tion. Often the extra information takes the form of a pair of values for the
independent and dependent variables.
For example, in the case of a population model, it would be natural to
specify the starting population, P0 say, and to start measuring time from
t = 0. We could then write
P = P0 when t = 0, or, equivalently, P (0) = P0 .
A requirement of this type is called an initial condition.

��
(a) An initial condition associated with the differential equation
dy
= f (x, y)
dx
specifies that the dependent variable y takes some value y0 when
the independent variable x takes some value x0 . This is written
either as
y = y0 when x = x0
or as
y(x0 ) = y0 .
The numbers x0 and y0 are referred to as initial values.
(b) The combination of a first-order differential equation and an initial
condition is called an initial-value problem.

The word ‘initial’ in these definitions arises from those (frequent) cases in
which the independent variable represents time. In such cases, the differen-
tial equation describes how the system being modelled behaves once started,
while the initial condition specifies the configuration in which the system is
started off. In fact, if the initial condition is y(x0 ) = y0 , then we are often
interested in solving the corresponding initial-value problem for x > x0 . If x represents time, then
x > x0 is ‘the future’ after the
system has been started off.
��
��

We usually require that an initial-value problem should have a unique solu-

tion, since then the outcome is completely determined by how the system
behaves and its conﬁguration at the start. Almost all the diﬀerential equa-
tions in this course do have unique solutions.

��
Using the result given in Exercise 1.3(b), solve the initial-value problem
dy
= x + y, y(0) = 1.
dx
��
From Exercise 1.3(b), on replacing the variables t, u by x, y, respectively,
the general solution of the differential equation here is
y = Cex − x − 1.
The initial condition says that y = 1 when x = 0, and on feeding these values
into the general solution we find that
1 = Ce0 − 0 − 1 = C − 1.
Hence C = 2, and the particular solution of the differential equation that
solves the initial-value problem is
y = 2ex − x − 1.

��

The size of a population (measured in hundreds of thousands) is modelled
by the logistic equation
� �
dP P
= kP 1 − , P (0) = 1,
dt M
where k = 0.15 and M = 10.
*(a) Use your answer to Exercise 1.3(c) to ﬁnd a solution to this initial-value
problem.
(b) Can you predict the long-term behaviour of the population size from
your answer?

y
Finally in this subsection, note that one needs to keep an eye on the domain 4
of the function defining the differential equation. ‘Gaps’ in the domain 2
usually show up as some form of restriction on the nature of a solution –10 –5 5 10 x
–2
curve. For example, consider the differential equation
–4
dy 1 –6
= . (1.5)
dx x
It turns out that there are two distinct families of solutions of this equation, ��
given by y = ln x + C (if x > 0) and y = ln(−x) + C (if x < 0). These two Since |x| = −x if x < 0, you
families of solutions are illustrated in Figure 1.2. Notice that the right-hand can see that this agrees with
side of Equation (1.5) is not defined at x = 0, and that there is no solution what we know from Unit 1,
that crosses the y-axis. namely
� that
1
This unit deals with numerical and analytic (symbolic) methods of solving dx = ln |x|.
x
differential equations. However, before we can discuss numerical methods,
we need to know something about the way that errors and accuracy are
described: this is the topic of the next subsection.

��
��

��

We often find that we need to make a calculation based on numerical values
that are not exact — for example, there may be limitations on the accuracy
to which a measurement can be taken. Finding a numerical solution to a
differential equation almost always involves inexact arithmetic: any calcula-
tor or computer will have some limit on the accuracy (number of significant
figures) to which a decimal can be stored, and rounding errors will occur.
In fact, using a numerical method usually involves repeated calculations, so
inaccuracies may build up and have a significant effect on the result. When
using a numerical method, one wants (if possible) to know how accurate
the result is. With this in mind, we recall a few basic ideas relating to
approximation and accuracy.
One simple form of calculation is the evaluation of a function. Suppose that
we want the value of f (π) for some function f , and use a value of π which
is rounded to three decimal places, that is, 3.142. Since this is the value of
π to three decimal places, we know that
|3.142 − π| ≤ 0.0005 = 5 × 10−4
(since 3.1415 ≤ π < 3.1425). This gives us some idea of how accurate 3.142
is as an estimate of π. In general, we refer to the difference
approximate value − true value
as the error in the approximate value, and to the modulus of this difference,
that is,
|approximate value − true value|,
as the absolute error in the approximate value. So if ε is the absolute Notice that, by definition, the
error in using 3.142 as an estimate of π, then we know that absolute error is always
greater than or equal to zero,
ε ≤ 0.0005. whereas the error can be
positive or negative.
In this context, the quantity 0.0005 is referred to as an absolute error
bound.
If we take π to be 3.142, what is the consequent error in f (π)? This will
depend on the function f .

��

For each of the functions f given below, use your calculator to ﬁnd f (3.142). You are welcome to use your
Then estimate f (π) more accurately, using π = 3.141 592 6 (to seven decimal computer rather than a
places). Hence estimate the error in using f (3.142) as an approximation calculator if you prefer.
to f (π).
(a) f (x) = x3 (b) f (x) = e10x

In Exercise 1.5(b), you saw that an absolute error of less than 0.0005 in
the value of π results in an absolute error of the order of 1.8 × 1011 in
the calculated value of f (π), for f (x) = e10x . For this function f , errors
are severely magnified! However, the situation is not quite so bad as this
statement might suggest. The calculated value of f (π) is not completely
unreliable — it is accurate to two significant figures. The value of f (π) is
itself very large (4.4 × 1013 ), so an error of 1.8 × 1011 is not quite so serious
as it sounds. We often want a measure of ‘error’ that takes into account the
size of the error relative to the size of the number being calculated.

��
��

We deﬁne the relative error as

� �
� approximate value − true value �
� �.
� true value �
So the relative error in f (3.142) as an estimate of f (π) here is (roughly)
1.8 × 1011
0.4 × 10−2 .
4.4 × 1013
A relative error of this size corresponds to a value that is accurate to two
significant figures, as obtained in Exercise 1.5(b). Relative errors provide
a guide to the number of significant figures that can be relied on, while Usually, a relative error of
absolute errors relate to the number of decimal places that are accurate. 0.5 × 10−n corresponds to a
value that is accurate to n
significant figures.
��
��
(a) Verify that, for any value of the constant C, the function
y = arcsin x + C (−1 < x < 1)
is a solution of the differential equation

dy 1

=√ .
dx 1 − x2
(b) Using the result of part (a), ﬁnd the solution of the initial-value problem
dy 1
=√ , y( 12 ) = π2 .
dx 1 − x2

��

(a) Verify that, for any value of the constant C, the function

� π �
x = tan(t + C) − 2 < t + C <
π2
is a solution of the diﬀerential equation
ẋ = 1 + x2 .
(b) Using the result of part (a), ﬁnd the solution of the initial-value problem
� �
ẋ = 1 + x2 , x π4
= 1.

� ��

Subsection 2.1 shows that qualitative information about the solutions of a
first-order differential equation may be gleaned directly from the equation
itself, without undertaking any form of integration process. The main con-
cept here is the direction field, sketches of which usually give a good idea of
how the graphs of solutions behave.
Direction fields can also be regarded as the starting point for a numeri-
cal (that is, calculational rather than algebraic) method of solution called
Euler’s method, which is described and applied in Subsection 2.2.
In Subsection 2.3 you will see how both direction fields and Euler’s method
can be implemented on your computer.

��
��

��

We start this subsection by considering what can be deduced about solutions
of the diﬀerential equation
dy
= f (x, y) (2.1)
dx
from direct observation of this equation.
In Section 1 we encountered the logistic equation Here we have
� � � �
dP P P
= kP 1 − , (2.2) f (t, P ) = kP 1− .
dt M M

where k and M are positive constants. In certain circumstances this is a

useful mathematical model of population changes, in which P (t) denotes
the size of the population at time t. The right-hand side of this equation
is equal to zero if either P = 0 or P = M . Hence, since dP/dt = 0 in both
cases, each of the constant functions P = 0 and P = M is a particular
solution of Equation (2.2). Within the model, these solutions correspond to
a complete absence of the population (P = 0), and an equilibrium population
y
level (P = M ) for which the proportionate birth and death rates are equal.
Such spotting of constant functions that are particular solutions is useful f ( x0 , y0 )
y0
on occasion but of limited applicability. In general, more useful information 1
can be deduced from the observation that, for any given point (x, y) in the
plane, the equation
x0 x
dy
= f (x, y) (2.1)
dx
�� A graphical
describes the direction in which the graph of the particular solution through representation of the slope at
that point is heading (see Figure 2.1). This is because if y = y(x) is any the point (x0 , y0 )
solution of the differential equation, then dy/dx is the gradient or slope of
the graph of that function. Equation (2.1) therefore tells us that f (x, y)
represents the slope at (x, y) of the graph of the particular solution that For example, if
passes through (x, y). If the slope f (x, y) at this point is positive, then the f (x, y) = x + y, then the
corresponding solution graph is increasing (rising) from left to right through slope at the point (1, 2) is
f (1, 2) = 1 + 2 = 3, the slope
the point (x, y); if the slope is negative, then the graph is decreasing (falling); at the point (2, −7) is
and if the slope is zero, then the graph is horizontal at the point. f (2, −7) = 2 − 7 = −5, and
When looking at f (x, y) in this light, it is referred to as a direction field, the slope at the point (3, −3)
is f (3, −3) = 3 − 3 = 0.
since it describes a direction (slope) for each point (x, y) where f (x, y) is
defined.

��
A direction field associates a unique direction to each point within
a specified region of the (x, y)-plane. The direction corresponding to
the point (x, y) may be thought of as the slope of a short line segment
through the point.
In particular, the direction field for the differential equation
dy
= f (x, y)
dx
associates the direction f (x, y) with the point (x, y).

Direction ﬁelds can be visualized by constructing the short line segments

referred to above for a ﬁnite set of points in an appropriate region of the
plane, where typically the points are chosen to form a rectangular grid.

��
��

An example is shown in Figure 2.2, which corresponds to the diﬀerential

equation
dy
= x + y. (2.3) Here f (x, y) = x + y.
dx
In this case the chosen region is the set of points (x, y) such that −2 ≤ x ≤ 2
and 0 ≤ y ≤ 2, and the rectangular grid consists of the points at intervals
of 0.2 in both the x- and y-directions within this region.

0 1 2 x
–2 –1

��

From this diagram, we can gain a good qualitative impression of how the
graphs of particular solutions of Equation (2.3) behave. The aim is mentally
to sketch curves on the diagram in such a way that the tangents to the curves
are always parallel to the local slopes of the direction ﬁeld. For example,
starting from the point (−1, 0.5) (that is, taking the initial condition to be
y(−1) = 0.5), we expect the solution graph initially to fall as we move to
the right. The magnitude of the negative slope decreases, however, and
eventually reaches zero, after which the slope becomes positive and then
increases. On this basis, we could sketch the graph of the corresponding
particular solution and obtain something like the curve shown in Figure 2.3.

(– 1 , 0 . 5)

0 1 2 x
–2 –1

��
��

��

(a) Part of the direction ﬁeld for the logistic equation
� �
dP P
=P 1− This is Equation (2.2) with
dt 1000 k = 1 and M = 1000.
is sketched in Figure 2.4. Using this diagram, sketch the solution curves
that pass through the following points:
(0, 1500), (0, 1000), (0, 100), (0, 0), (0, −100).
From your results, describe the graphs of particular solutions of the
diﬀerential equation.

P
1500

1000

500

0
2 4 6 8 10 t

–500

��

(b) What does your answer to part (a) tell you about the predicted be-
haviour of a population whose size P (t) at time t is modelled by this
logistic equation?

Drawing by hand precise grids of line segments to represent direction ﬁelds

is not a good use of your time. However, it is a task that your computer
can be programmed to perform, as you will see in Subsection 2.3. Before
investigating this, you will see in Subsection 2.2 how the concept of direction
fields helps in constructing approximate numerical solutions for first-order
differential equations.

��

In the previous subsection it was suggested that the graphs of particular
solutions of a differential equation
dy
= f (x, y)
dx
could be ‘mentally sketched’ on a diagram of the direction field given by
f (x, y). This was to be done in such a way that the tangent to the solution
curve is always ‘parallel to the local slope’ of the direction field. While this
gives a good visual image of the connection between the direction field and

��
��

the graph of a solution, it is somewhat short on precision. We could not

expect, by this approach, to predict with any accuracy the actual solution to
an initial-value problem. That task is the subject of the current subsection,
which shows how an initial-value problem
dy
= f (x, y), y(x0 ) = y0 , (2.4)
dx
may be ‘solved’ by calculational means. However, the direction field diagram
is still of use in explaining how this numerical method arises.
Suppose that instead of trying to sketch a solution curve to fit the direction
field, we move in a sequence of straight-line steps whose directions are gov-
erned by the direction field. The aim is to produce a sequence of points that
provide approximate values of the solution function y(x) for the initial-value
problem at a sequence of x-values. The steps are constructed as follows.
Corresponding to the given initial condition y(x0 ) = y0 , there is a point P0
in the (x, y)-plane with coordinates (x0 , y0 ), and this is our starting point.
At P0 , the direction field f (x, y) defines a particular slope, namely f (x0 , y0 ).
We move off from P0 along a straight line that has this slope, and continue
until we have travelled a horizontal distance h to the right of P0 . The point
that has now been reached is labelled P1 , as in Figure 2.5.
y

slope = f ( x 0, y0)
P1 ( x1, Y1)

P0 Y 1 – y0
(x 0, y0)
h

0 x0 x1 x

��
The idea is that the point P1 , whose coordinates have been denoted by
(x1 , Y1 ), provides an approximate value Y1 of the solution function y(x) at The reason for using Y1 here,
x = x1 . Now, unless the solution function follows a straight line as x moves rather than y1 , will be
from x0 to x1 , Y1 is unlikely to give the exact value of y(x1 ). However, the explained shortly.
hope is that, because we headed off from x0 along the correct slope, as given
by the direction field, Y1 will be reasonably close to the exact value. Before
worrying about accuracy, let us continue with the construction of the points
in our sequence.
The next thing that we need to do, before proceeding to the second step
in the construction process, is determine formulae for x1 and Y1 in terms
of x0 , y0 , h and f (x0 , y0 ). By the construction described, as the point P1
is reached from P0 by taking a step to the right of horizontal length h, we
have
x1 = x0 + h. (2.5)
We can express Y1 in terms of other quantities by equating two expressions
for the slope of the line segment P0 P1 ,
Y1 − y0
= f (x0 , y0 ),
h
and then rearranging to give
Y1 = y0 + hf (x0 , y0 ). (2.6)
This completes the first step, and we now take a second step to the right.

��
��

The direction of the second step is along the line with slope deﬁned by the
direction ﬁeld at the point P1 , namely f (x1 , Y1 ). The second step moves us
from P1 through a further horizontal distance h to the right, to the point
labelled P2 , as illustrated in Figure 2.6. This point provides an approximate
value Y2 of the solution function y(x) at x = x2 .

y
slope = f ( x1, Y1) P2 ( x2, Y2)

Y 2 – Y1
( x1, Y1) P1
h
P0

0 x0 x1 x2 x

��
As in the first step, we need now to express the coordinates (x2 , Y2 ) of P2
in terms of x1 , Y1 , h and f (x1 , Y1 ). We have
x2 = x1 + h (2.7)
and (equating two expressions for the slope of the line segment P1 P2 )
Y2 − Y1
= f (x1 , Y1 ),
h
which can be rearranged to give
Y2 = Y1 + hf (x1 , Y1 ). (2.8)
Having carried out two steps of the process, it is possible to see that the
same procedure can be applied to construct any number of further steps,
and we next generalize to a description of what happens at the (i + 1)th
step, where i represents any non-negative integer.
Suppose that after i steps we have reached the point Pi , with coordinates
(xi , Yi ). For the (i + 1)th step, we move away from Pi along the line with
slope f (xi , Yi ), as defined by the direction field at Pi . After moving through a
horizontal distance h to the right, we reach the point Pi+1 , whose coordinates
are denoted by (xi+1 , Yi+1 ), as illustrated in Figure 2.7. The point Pi+1
provides an approximate value Yi+1 of the solution function y(x) at x = xi+1 .

y
P i+1 ( xi+1 , Y i+1 )
slope = f ( x i , Y i )

Y i+1 – Yi

Pi
(xi , Yi )
h

0 xi xi+1 x

��

��
��

Arguing as before, we have

xi+1 = xi + h (2.9)
and (equating two expressions for the slope of the line segment Pi Pi+1 )
Yi+1 − Yi
= f (xi , Yi ),
h
which can be rearranged to give
Yi+1 = Yi + hf (xi , Yi ). (2.10)
Note that Equations (2.5) and (2.7) are the special cases of Equation (2.9)
for i = 0 and i = 1, respectively, and that Equation (2.8) is the special case
of Equation (2.10) for i = 1. If we also define Y0 to be equal to the initial
value y0 , then Equation (2.6) is the special case of Equation (2.10) for i = 0.
To sum up, for the initial-value problem (2.4), we have a procedure for
constructing a sequence of points
Pi with coordinates (xi , Yi ) (i = 1, 2, . . .),
where the values of xi and Yi for each value of i are determined by the re-
spective formulae (2.9) and (2.10). The starting point for the sequence is the
point P0 with coordinates (x0 , Y0 ), where Y0 = y0 . Because the procedure
is based on the direction field, each Yi provides an approximation at x = xi
to the value of the solution function y(x) for the initial-value problem. The
horizontal distance h by which we move to the right at each stage of the
procedure is called the step size or step length.
Figure 2.8 shows the constructed sequence of points, and for comparison in-
cludes a curve representing the graph of the exact solution of the initial-value
problem (2.4). This makes clear that the successive points P1 , P2 , P3 , . . . are
only approximations to points on the solution curve. In fact, the situation
shown in Figure 2.8 is typical of the behaviour of the constructed approx-
imations, in that they gradually move further and further from the exact
solution curve. This is because, at each step, the direction of movement is
along the slope of the direction field at Pi = (xi , Yi ) and not along the slope
of the direction field at (xi , yi ), where yi = y(xi ) denotes the value of the The common use of
exact solution function at x = xi ; that is, for each xi , the slope for the next yi = y(xi ) to represent the
step is defined by a point close to the solution curve rather than by the point exact solution at x = xi
explains why we use a
exactly on that curve. different notation, namely Yi ,
for the numerical
approximation to y(xi ).
y

y = y(x) P6

P4
P3
P2
P1
P0

0 x0 x1 x2 x3 x4 x5 x6 x

��
��

Nevertheless, the formulae (2.9) and (2.10) provide a method for ﬁnding ap-
proximate solutions to the initial-value problem (2.4), in terms of numerical The accuracy of such
estimates Y1 , Y2 , Y3 , . . . at the respective domain values x1 , x2 , x3 , . . .. This approximate solutions, and
is called Euler’s method, and is summarized below. ways of improving accuracy,
will be considered shortly.

��

��
To apply Euler’s method to the initial-value problem Leonhard Euler (1707–1783)
was one of the most proliﬁc
dy mathematicians of all time.
= f (x, y), y(x0 ) = y0 , (2.4)
dx (His surname is pronounced
proceed as follows. ‘oiler’.) He ﬁrst devised this
method in order to compute
(a) Take x0 and Y0 = y0 as starting values, choose a step size h, and the orbit of the moon.
set i = 0.
(b) Calculate the x-coordinate xi+1 , using the formula
xi+1 = xi + h. (2.9)
(c) Calculate a corresponding approximation Yi+1 to y(xi+1 ), using
the formula
Yi+1 = Yi + hf (xi , Yi ). (2.10)
(d) If further approximate values are required, increase i by 1 and
return to Step (b).

��
Consider the initial-value problem
dy
= x + y, y(0) = 1.
dx
Use Euler’s method, with step size h = 0.2, to obtain an approximation to
y(1).

��
We have x0 = 0, Y0 = y0 = 1, and f (xi , Yi ) = xi + Yi . The step size is given
as h = 0.2. Equation (2.9) with i = 0 gives
x1 = x0 + h = 0 + 0.2 = 0.2,
and Equation (2.10) with i = 0 gives
Y1 = Y0 + hf (x0 , Y0 ) = 1 + 0.2 × (0 + 1) = 1.2.
For the second step, we have (from Equation (2.9) with i = 1)
x2 = x1 + h = 0.2 + 0.2 = 0.4,
and (from Equation (2.10) with i = 1)
Y2 = Y1 + hf (x1 , Y1 ) = 1.2 + 0.2 × (0.2 + 1.2) = 1.48.
If more than a couple of steps of such a calculation have to be computed by
hand, then it is a good idea to lay out the calculation as a table. In this
case, by continuing as above and putting i in turn equal to 2, 3 and 4, we
obtain Table 2.1.

��
��

i xi Yi f (xi , Yi ) = xi + Yi Yi+1 = Yi + hf (xi , Yi )

0 0 1 1 1.2 After each value of Yi+1 has
1 0.2 1.2 1.4 1.48 been calculated from the
2 0.4 1.48 1.88 1.856 formula and entered in the
3 0.6 1.856 2.456 2.347 2 last column, it is transferred
to the Yi column in the next
4 0.8 2.347 2 3.147 2 2.976 64
row.
5 1.0 2.976 64

So, at x = 1, Euler’s method with step size h = 0.2 gives the approximation
y(1) 2.976 64.

��

Use Euler’s method, with step size h = 0.2, to obtain an approximation to
y(1) for the initial-value problem
dy
= y, y(0) = 1.
dx

The solution of the initial-value problem given in Example 2.1 is in fact

known exactly, and is y = 2ex − x − 1. Putting x = 1, this gives This exact solution was found
in Example 1.1 (page 11).
y(1) = 2e − 1 − 1 = 3.436 56,
correct to five decimal places. This value may be compared with the ap-
proximation 2.976 64 for y(1) obtained by Euler’s method in Example 2.1,
and the comparison indicates that the approximation is not at all accurate.
Indeed, the absolute error in this case is
|2.976 64 − 3.436 56| = 0.459 92,
which is about 13% of the exact value, and indeed not even one decimal
place accuracy is achieved.
Similarly, the other values Yi (i = 1, 2, 3, 4) found in Example 2.1 contain
significant absolute errors when considered as approximations to the cor-
responding exact values yi = y(xi ). This is illustrated in general terms in
Figure 2.8, where the absolute error in approximation Yi is the vertical dis-
tance from the point Pi to the point directly above it on the exact solution
curve. As shown there, and for reasons given earlier, the absolute error tends
to increase as more and more steps are taken.
The realization that Euler’s method can produce values that are poor ap-
proximations to the exact solution of an initial-value problem invites us to
ask whether the accuracy of the approximations can be improved, using this
method. In fact, it is not hard to see that improvements in accuracy ought
to be achieved by reducing the step size h. Our earlier prescription for con-
structing the sequence of points P1 , P2 , P3 , . . . from the starting point P0
and the given direction field amounts loosely to saying ‘match the direction
of the solution curve at the current point, take a step, then adjust direction
so as to try not to move further away from the curve’. It seems natural,
therefore, that the approximations will improve if we reduce the size of the
steps taken and correspondingly ‘adjust direction’ more frequently. This is
illustrated for a hypothetical case in Figure 2.9.

��
��

y
estimate with
h = 0.4

with h = 0.2
y = y (x)
with h = 0.1

exact solution
at x = 0.4

0 0.1 0.2 0.3 0.4 x

��

In fact, it can be shown that the accuracy of Euler’s method does indeed
usually improve when we take a smaller step size.
To demonstrate this, consider the initial-value problem from Exercise 2.2.
This has the exact solution y = ex (as you can verify), and its value at
x = 1 is y(1) = e = 2.718 282, to six decimal places. In Exercise 2.2 you
showed that with a step size h = 0.2, Euler’s method gives the approximation
2.488 32 for y(1). Table 2.2 shows the corresponding results (to six decimal
places) obtained when we apply Euler’s method to this same initial-value
problem but with successively smaller step sizes h. In Exercise 2.2, where
h = 0.2, the value of y(1) was
approximated by Y5 . From
�� the column for ‘number of
steps’ in Table 2.2, you can
h Approximation Absolute Number of see that y(1) is approximated
to y(1) error steps by:
0.1 2.593 742 0.124 539 10 Y10 when h = 0.1;

Y100 when h = 0.01;

0.01 2.704 814 0.013 468 100

Y1000 when h = 0.001;

0.001 2.716 924 0.001 358 1000

Y10000 when h = 0.0001.

0.0001 2.718 146 0.000 136 10000

As expected, the absolute errors in the third column of the table become
progressively smaller as h is reduced.
Looking more carefully at these absolute errors, we notice that they seem
to tend towards a sequence in which each number is a tenth of the previous
one. Since each value of h in the table is a tenth of the previous one, this
suggests that:
absolute error is approximately proportional to step size h.
This turns out to be a general property of Euler’s method, for suﬃciently You will see this property
small values of the step size. So, not only do we know that accuracy can be stated formally in Unit 26,
improved by decreasing the step size h, but this general property also tells us where you will also see how
the property can be used to
that, by making h small enough, the absolute error in an approximation can estimate the size of absolute
errors.

��
��

be made as small as desired. In other words, the absolute error approaches

the limit zero as h approaches zero (as you might have expected from the
intuitive argument preceding Figure 2.9).

��

Suppose that when Euler’s method is applied to the problem in Exercise 2.2,
the absolute error in approximating y(1) is proportional to the step size h,
for suﬃciently small h.
Use the last row of Table 2.2 to estimate the constant of proportionality, k
say, and hence to estimate the step size required to compute y(1) correct to
ﬁve decimal places (that is, so that the absolute error is less than 5 × 10−6 ).

A few words of caution are necessary at this point. Although the absolute
error can be made as small as we please by making the step size h sufficiently
small, this is valid only if the arithmetic is performed using sufficient decimal
places. Where a calculator or computer is involved, the number of decimal
places that can be used is limited, and as a result rounding errors may
be introduced into the calculations. After a certain point, any increase in
accuracy brought about by reducing the size of h may be swamped by these
rounding errors.
Moreover, rounding errors are not the only problem. Before concluding that
h should always be chosen to be very small, we must also consider the cost
of this additional accuracy. Now, by cost is meant the effort involved, which
can be measured in a variety of ways; commonly for iterative methods (such
as Euler’s method) it is measured by the number of steps taken. In general
for numerical methods, the greater the accuracy required, the greater the
cost. To illustrate this, look back at Table 2.2. The last column of the
table shows how the number of steps required for the calculation goes up
in inverse proportion to the step size: to move from x = 0 to x = 1, it In general, to move from a
takes 10 steps with step size h = 0.1, 100 steps with step size h = 0.01, to b (where b > a) with step
and so on. Since, for sufficiently small h, the error in Euler’s method is size h takes (b − a)/h steps.
approximately proportional to the step size, it follows that for this method
a ten-fold improvement in accuracy is paid for by a ten-fold increase in the
number of steps required.
So, for Euler’s method and similar methods, the choice of step size has to be
based on a compromise between the two opposing requirements of accuracy
and cost. Methods for choosing the step size are discussed in Unit 26, which
also introduces other numerical methods for solving initial-value problems
that are considerably more efficient than Euler’s method. In fact, Euler’s Greater efficiency means that
method is not suitable for high-accuracy work. Its virtue lies rather in its the same or better numerical
simplicity and its clear illustration of the basic principles of how differential accuracy is achieved with
fewer numerical
equations may be solved numerically. computations.
In any practical case, calculations of the type described in this subsection
are ideally suited to being performed on a computer, as you will see in the
next subsection.

��
��

��

In this subsection you will see how the computer can be used to construct
direction ﬁelds and to implement Euler’s method.

Use your computer to complete the following activities. PC

��
Plot the direction field for the differential equation
dy
=x+y This is Equation (2.3)
dx (page 15).
in the region
−2 ≤ x ≤ 2, 0 ≤ y ≤ 2.
On the basis of the plot of the direction field, what can you say about the
graphs of solutions of the differential equation?

��
Use Euler’s method to obtain approximations to y(1) for the initial-value
problem
dy
= x + y, y(0) = 0,
dx
using step sizes h = 1, 0.5, 0.2, 0.1, 0.01, 0.001, 0.0001, in turn. In each case,
plot the graph of the solution on an appropriate direction ﬁeld diagram, and
observe how each graph compares with the previous one.

��
Euler’s method is to be used to estimate the value of the function y(x) at
x = 0.1, 0.2, . . . , 1 for the initial-value problem
dy
= x + y, y(0) = 0.
dx
(a) Use the step sizes h = 0.1, 0.01, 0.001, 0.0001, in turn. Compare the
results in each case with the exact solution y = ex − x − 1, and comment
on how the size of the absolute error varies with h.
(b) Compare your estimates for the step sizes h = 0.1 and h = 0.01. Then
compare your estimates for all four step sizes. What can you conclude?

��

(a) Without plotting the direction field, say what you can about the slopes
defined by the differential equation
dy
= f (x, y) = y + x2 .
dx
(b) Verify that your conclusions are consistent with the direction field dia-
gram in Figure 2.10.

��
��

–2 –1 0 1 2 x

–1

–2

��
(c) On the basis of the direction ﬁeld, what can be said about the graphs
of solutions of the diﬀerential equation?
(d) Write down the formulae required in order to apply Euler’s method to
the initial-value problem
dy
= y + x2 , y(−1) = −0.2,
dx

using a step size h = 0.1.

� ��

This section and the next look at methods for finding analytic solutions
of first-order differential equations — that is, solutions expressed in terms
of exact formulae. It is not always possible to find analytic solutions, and
in these cases numerical methods of approximate solution, such as Euler’s
method, are applied. Even when a formula for the solution is obtainable,
it may be so complicated that a numerical solution is preferred. However,
where a simple formula can be found, this is likely to be more informative
than use of a numerical method.
This section specializes from the form of differential equation
dy
= f (x, y)
dx
considered earlier. Subsection 3.1 looks at cases in which f (x, y) is taken to
be a function of x alone, f (x). You will see that the differential equa-
tion can then be solved by direct integration, assuming that the neces-
sary integration can be performed. Subsection 3.2 considers cases in which
f (x, y) = g(x)h(y) (the product of a function of x and a function of y).
These can be solved in principle by the method of separation of variables.

��
��

��

An example of a differential equation that can be solved is
dy
= x2 . (3.1)
dx
In order to do this, we need to find functions y(x) whose derivatives are x2 ;
one such function is y =
13 x3 . There are other functions with this same
derivative, for example y = 13 x3 + 1 and y = 13 x3 − 2. In fact, any function
of the form
y = 13 x3 + C, The values C = 0, C = 1 and
C = −2 give the three
where C is an arbitrary constant, satisfies the differential equation (3.1). particular solutions
This is an expression for the general solution. mentioned above.
The expression 13 x3 + C is also the indefinite integral of x2 : that is, This is hardly surprising,
� since integration ‘undoes’ or
x2 dx = 13 x3 + C. reverses the effect of
differentiation.
In this case, therefore, the indefinite integral of x2 is the general solution of
Equation (3.1), and a similar connection applies more generally.
Consider the differential equation
dy
= f (x), (3.2) The function f (x) is assumed
dx to be continuous (i.e. its
where the right-hand side, f (x), is a function of x alone. Suppose that we graph has no breaks).
have a particular solution y = F (x) of this differential equation; in other
words, F (x) is an integral of f (x). In such circumstances, the general so-
lution of Equation (3.2) is given by y = F (x) + C, where C is an arbitrary
constant; and the indefinite integral of f (x) is given by the same expression,
�
f (x) dx = F (x) + C.

This means that the general solution of Equation (3.2) can be written down
directly as an indeﬁnite integral; and, if the integration can be performed,
then the equation is solved.

��

The general solution of the diﬀerential equation
dy
= f (x) (3.2)
dx
is �
y = f (x) dx = F (x) + C,

where F (x) is an integral of f (x) and C is an arbitrary constant.

Once the general solution has been found, it is possible to single out a
particular solution by specifying a value for the constant C. As before, this
value may be found by applying an initial condition.

��
(a) Find the general solution of the diﬀerential equation
dy
= e−3x .
dx

��
��

(b) Find the particular solution of this diﬀerential equation that satisﬁes
the initial condition y(0) = 53 .

��
(a) On applying direct integration, we obtain the general solution
�
y = e−3x dx = − 13 e
−3x + C,

where C is an arbitrary constant.

(b) In order to satisfy the initial condition y(0) =

53 (that is, y
=
53 when

x = 0), we must have

5
3 = − 13 e
0 + C,
so C = 2. The required particular solution is therefore

y = − 13 e
−3x + 2.

Procedure 3.1 uses x for the independent variable and y for the dependent

variable. As usual, you should be prepared to translate this into situations

where other symbols are used for the variables. But remember that the

method of direct integration applies solely to ﬁrst-order diﬀerential equations

for which the derivative is equal to a function of the independent variable

alone. Thus direct integration can be applied, for example, to the diﬀerential

equation

dx
= cos t,
dt
to give the general solution
�
x = cos t dt = sin t + C,

where C is an arbitrary constant. (Here t is the independent variable and x

is the dependent variable.) On the other hand, the diﬀerential equation
dx
= x2
dt
cannot be solved by direct integration, since the right-hand side here is a
function of the dependent variable, x.

��

Solve each of the following initial-value problems.
dy
(a) = 6x, y(1) = 5.
dx
dv
(b) = e4u , v(0) = 2.
du
Remember that ẏ stands for
*(c) ẏ = 5 sin 2t, y(0) = 0. dy/dt, where t denotes time.

The method of direct integration succeeds in solving a diﬀerential equation

of the speciﬁed type whenever it is possible to carry out the integration that
arises, and this task may require you to apply any of the standard techniques
of integration, such as integration by parts and integration by substitution. Integration by parts and
For more diﬃcult integrals, a computer algebra package can be used. integration by substitution
are revised in Unit 1. Many
integrations can be performed
by reference to the table of
standard integrals in the
course Handbook.

��
��

��

Find the general solution of each of the following diﬀerential equations.
dy
(a) = xe−2x
dx
t
*(b) ṗ = (Hint : For the integral, try the substitution u = 1 + t2 .)
1 + t2

The answer to Exercise 3.2(b) can be generalized to any diﬀerential equation This is a simple extension of
of the form the�result from Unit 1 that
f (x)
dy f (x) dx = ln |f (x)| + C,
=k (f (x) = 0), f (x)
dx f (x) for f (x) = 0.
where k is a constant, to give the general solution
y = k ln |f (x)| + C,
where C is an arbitrary constant.

��

Direct integration applies, in an immediate sense, only to the very simplest
type of differential equation, as described by Equation (3.2). However, all
other analytic methods of solution for first-order equations eventually also
boil down to performing integrations. In this subsection, we consider how
to solve first-order differential equations of the form
dy
= f (x, y)
dx
where the right-hand side f (x, y) is the product of a function of x and a
function of y; that is, equations of the form
dy
= g(x)h(y). (3.3)
dx
One example of this type of differential equation is
dy Here we have g(x) = x and
= x(1 + y 2 ). (3.4) h(y) = 1 + y 2 .
dx
We divide both sides of this equation by 1 + y 2 , to obtain Note that 1 + y 2 is never zero,
so it is safe to divide by it.
1 dy

= x,

1 + y 2 dx
and then integrate both sides with respect to x, which gives
� �
1 dy
dx = x dx. (3.5)
1 + y 2 dx
Applying the rule for integration by substitution (in Leibniz notation) to See Section 6 of Unit 1.
the left-hand side, we obtain
� �
1 dy 1

2
dx = dy,

1 + y dx 1 + y2
so Equation (3.5) becomes
� �
1
dy = x dx.
1 + y2
On performing the two integrations, we obtain See the table of standard
1 2 integrals in the Handbook.
arctan y = 2x + C, (3.6)

��
��

where C is an arbitrary constant. Making y the subject of the equation, we Note that one arbitrary
obtain the solution expression constant suffices.
y = tan(
12 x
2 + C).
The approach just demonstrated applies more widely. In principle, it works
for any differential equation of the form
dy
= g(x)h(y). (3.3)
dx
On dividing this equation through by h(y) (for all values of y other than
those where h(y) = 0), we obtain
1 dy
= g(x).
h(y) dx
Integration with respect to x on both sides gives
� �
1 dy
dx = g(x) dx,
h(y) dx
and, on applying the rule for integration by substitution to the left-hand
side, this becomes
� �
1
dy = g(x) dx. (3.7) This is the form that you
h(y) need to remember! Note that
If the two integrals can be evaluated at this stage, then we reach an equation you can obtain it ‘informally’
that relates x and y and features an arbitrary constant. This equation is by dividing Equation (3.3) by
the general solution of the differential equation (for values of y other than h(y), ‘multiplying through
by dx’, and then adding the
those where h(y) = 0); but usually y will not be the subject of this equation. two integral signs.
It is a form of the general solution called an implicit (general) solution of
the differential equation. (An example of an implicit solution is provided
by Equation (3.6).) Usually, the final aim is to make y the subject of the
equation, if possible — that is, to manipulate the equation into the form
y = function of x.
This is called the explicit (general) solution of the differential equation.
In either case (implicit or explicit), a particular solution may be obtained
from the general solution as before, by applying an initial condition.
The method just described for solving differential equations of the form (3.3)
is called the method of separation of variables since, in Equation (3.7), we
have separated the variables to either side of the equation, with only the
dependent variable appearing on the left and only the independent variable
on the right. The method is summarized below.

��

This method applies to separable diﬀerential equations, which are of

the form

dy
= g(x)h(y). (3.3)
dx
(a) Divide both sides by h(y) (where h(y) = 0), and integrate both

sides with respect to x, to obtain

� �
1
dy = g(x) dx. (3.7)
h(y)
(b) If possible, perform the integrations, to obtain an implicit form of

the general solution.

It is a good idea to check, by
substitution into the original
(c) If possible, rearrange the formula found in Step (b) to give y in
diﬀerential equation, that the
terms of x. This is the explicit (general) solution.
function obtained is indeed a
solution.

��
��

The separation of variables method is useful, but there are some difficulties
with it. First, it may not be possible to perform the necessary integrations.
Second, the general solution obtained is restricted to those values of y such
that h(y) = 0. Third, it may not be possible to perform the necessary
manipulations to obtain an explicit solution.
Of these difficulties, the first can be overcome by use of a numerical method,
such as Euler’s method. The second will be discussed shortly. The third
will usually also need numerical techniques.
It is necessary to be careful about the domain or image set of the solution
obtained, as the following example illustrates.

��
(a) Find the general solution of the diﬀerential equation
dy x
=− (y > 0).
dx 3y
(b) Find the particular solution that satisﬁes the initial condition y(0) = 3.

��
(a) The equation is of the form
dy
= g(x)h(y),
dx

where the obvious choices for g and h are

g(x) = −x and h(y) = 1/(3y). Notice that since y > 0,

h(y) is never zero.

We now apply Procedure 3.2. On dividing through by h(y) = 1/(3y)

(that is, multiplying through by 3y) and integrating with respect to x,

the diﬀerential equation becomes

� �
3y dy = −x dx. With practice, you will be
able to move directly to this
Evaluating the integrals gives stage, as shown in
Procedure 3.2.
3 2
2y = −
12 x
2 + B,
where B is an arbitrary constant. This is an implicit form of the general
solution.
On solving for y (and noting the condition y > 0 given above, which
determines the sign of the square root), we obtain the explicit general
solution
�
y =
13 (2B − x2 ).
This can be simpliﬁed slightly by writing C in place of 2B, where C

is another arbitrary constant. However, we need to recognize that the

formula for y represents a real quantity greater than zero only when the

argument of the square root is positive, so we must have C − x

2 > 0.
This in turn means that C cannot be completely arbitrary, since it must
at least be positive. The general solution in this case is therefore Since x2 ≥ 0 for all x,
� √ √ C − x2 > 0 implies that
y =
3 (C − x ) (− C < x < C),
1 2 C > x2 ≥ 0, so C must be
positive.
where C is a positive but otherwise arbitrary constant.

��
��

(b) The initial condition is y(0) = 3, so we substitute

� x = 0 and y = 3 into
1
the general solution above. This gives 3 = 3 C, so C = 27, and the
required particular solution is
� √ √ You veriﬁed in Exercise 1.2(d)
y = 13 (27 − x2 ) (−3 3 < x < 3 3). that this function is a solution
of the diﬀerential equation.

��

A mass m(t) of a uranium isotope, which is present in an object at time t,

declines over time due to radioactive decay. Its behaviour is modelled by

the diﬀerential equation

dm
= −λm (m > 0), This model can be applied to
dt other radioactive substances
where the decay constant λ is a positive constant characteristic of the ura- by selecting the appropriate
nium isotope. value of the parameter λ.
(a) Find the general solution of this diﬀerential equation.
(b) Find the particular solution for which the initial amount of uranium

present (at time t = 0) is m0 .

The condition m > 0 in Exercise 3.3 arose from the modelling context. This
condition enabled us to ﬁnd the general solution without needing to worry
about dividing by zero at Step (a) of the separation of variables method
(and hence without needing to restrict the image set further). Suppose we
were to forget the modelling context — that is, suppose we were to remove
the restriction m > 0. How does this aﬀect the solution process? And how
do we cope with the case where m = 0? These questions are answered in the
following example where, to emphasize the absence of the previous modelling
context, the variables used are x and y.

��
Find the general solution of the diﬀerential equation
dy
= −λy,
dx
where λ is a non-zero constant.
��
To apply the separation of variables method, we need to exclude the cases
where y = 0. So, for y = 0, on dividing through by y, integrating with
respect to x, and using the rule for integration by substitution on the left-
hand side, we obtain
� �
1
dy = (−λ) dx. (3.8)
y
Integrating, we obtain You saw in Unit 1 that
�
1
ln |y| = −λx + B, dy = ln |y| (y =
0).
y
where B is an arbitrary constant. Taking exponentials gives
|y| = e−λx+B
or, removing the modulus sign,
y = ±e−λx+B = ±eB e−λx = Ce−λx ,
where C = ±eB is a non-zero but otherwise arbitrary constant.

��
��

This is not quite the general solution, as we have to consider what happens
when y = 0. Now, looking at the above solution, it is natural to ask what
happens when C = 0. This gives the zero function, y = 0 for all x, and
inspection of the diﬀerential equation shows that this is a particular solution.
So we now have the general solution
y = Ce−λx ,
where C is an arbitrary constant. (Positive C corresponds to y > 0, negative
C to y < 0, and C = 0 to the particular solution y = 0.)

The above example illustrates that:

• the separation of variables method requires that h(y) = 0 and gives a
family of solutions containing an arbitrary constant;
• the case when h(y) = 0 is exceptional and can give extra solutions that
may or may not have the same form as the family of general solutions.
The following exercises provide you with some practice at applying the sepa-
ration of variables method and at completing the general solution for values
of y such that h(y) = 0.
��
Find the general solution of each of the following diﬀerential equations.
dy y − 1 dy 2y
*(a) = (x > 0) (b) = 2
dx x dx x +1
��

Solve the initial-value problem

dv
= eu+v , v(0) = 0.
du

��

Find the general solution of each of the following differential equations,
where a is a non-zero constant.
dy 1
(a) = a)
(u =
du u−a
dy 1
(b) = 0, x = 1/a)
(x =
dx x(1 − ax)
��
Find the general solution of each of the following differential equations.
(a) u = xu (b) ẋ = 1 + x2
��
(a) Solve the initial-value problem The differential equation here
� � is the logistic equation
dP P
= kP 1 − , P (0) = P0 (where P0 > 0), (Equation (2.2)) which, as
dt M was pointed out earlier, may
where k and M are positive constants. be used as a model for the
size P (t) of a population at
(Hint : For the integral involving P , the solution for Exercise 3.6(b) time t. The direction field of
should be of use.) this equation in a specific case
(b) Describe what happens to the solution P (t) as t becomes large. was examined in Exercise 2.1
(page 16).

��
��

This section presents one ﬁnal method of analytic solution for ﬁrst-order

diﬀerential equations. The details of this integrating factor method appear

in Subsection 4.2. It applies only to a particular form of equation known

as a linear diﬀerential equation. The deﬁnition and some properties of this

type of equation are introduced in Subsection 4.1.

��

This subsection introduces the concept of linearity as applied to diﬀeren
-
tial equations. Here the concept is introduced in the context of ﬁrst-order

diﬀerential equations, but you should be aware that the idea generalizes to

higher-order diﬀerential equations and is important from a theoretical point Linear second-order

of view. diﬀerential equations are

considered in Unit 3.

��
(a) A ﬁrst-order diﬀerential equation for y = y(x) is linear if it can be

expressed in the form

dy
+ g(x)y = h(x), (4.1) This differential equation can
dx be written in the general form
where g(x) and h(x) are given functions. dy
= f (x, y)
(b) A linear first-order differential equation is said to be homoge- dx
that we have been using by
neous if h(x) = 0 for all x, and inhomogeneous or non- putting
homogeneous otherwise. f (x, y) = −g(x)y + h(x).

For example, the diﬀerential equation

dy
− x2 y = x3
dx

is linear, with g(x) = −x2 and h(x) = x3 , whereas the equation

dy
= xy 2
dx
is not, due to the presence of the non-linear term y 2 .

��

Decide whether or not each of the following ﬁrst-order diﬀerential equations

is linear.

dy dy dz
(a) + x3 y = x5 (b) = x sin x (c) = −3z 1/2
dx dx dt
dy dy
(d) ẏ + y 2 = t (e) x + y = y2 (f) (1 + x2 ) + 2xy = 3x2
dx dx

An important theorem guarantees that an initial-value problem based on a

linear ﬁrst-order diﬀerential equation has a unique solution.

��
��

��
If the functions g(x) and h(x) are continuous throughout an interval
(a, b) and x0 belongs to this interval, then the initial-value problem This includes the possibility
that either a = −∞ or b = ∞,
dy so the interval might be all of
+ g(x)y = h(x), y(x0 ) = y0 ,
dx the real line.

has a unique solution throughout the interval.

This is a very powerful result, since it means that once you have found a
solution in a particular interval, that solution will be the only one.
There is a particularly useful technique for solving linear diﬀerential equa-
tions, to which we turn next.

��

As you have seen, the method of separation of variables relies upon an
application of the rule for integration by substitution, which is equivalent to
the Composite Rule (or Chain Rule) for derivatives. It is natural to enquire
whether there might similarly be a method for solving first-order differential
equations that derives from the rule for integration by parts or, equivalently,
from the Product Rule for derivatives. There is indeed such a method, and
it is the subject of this subsection.
To introduce the topic, consider the differential equation
dy
(1 + x2 ) + 2xy = 3x2 . (4.2) As you saw in Exercise 4.1(f),
dx this differential equation is
Note first that 2x (the coefficient of y) is the derivative of 1 + x2 (the co- linear; but it is not soluble by
efficient of dy/dx). It follows from the Product Rule that direct integration or by
separation of variables.
d � � dy
(1 + x2 )y = (1 + x2 ) + 2xy.
dx dx
The right-hand side of this equation is the same as the left-hand side of
Equation (4.2), so we can rewrite the latter as
d � �
(1 + x2 )y = 3x2 . (4.3)
dx
Now the left-hand side here is just the derivative of (1 + x2 )y, so we can
apply direct integration to Equation (4.3) to obtain
�
(1 + x )y = 3x2 dx = x3 + C,
2

where C is an arbitrary constant. Division by 1 + x2 then gives the general

solution of Equation (4.2) explicitly, as
x3 + C
y= .
1 + x2

��
��

This solution was arrived at by noting that the left-hand side of Equa-
tion (4.2) is of the form
dy dp
p + y, (4.4)
dx dx
where p = 1 + x2 , and that this form can be re-expressed, using the Product
Rule, as
d
(py).
dx
Linear differential equations need not come in this convenient form. For
example, the left-hand side of the equation
� �
dy 2x 3x2
+ y = (4.5)
dx 1 + x2 1 + x2
is not of the form (4.4). However, Equation (4.2) can be obtained from
Equation (4.5) on multiplying through by p = 1 + x2 . For this reason,
p = 1 + x2 may be called an integrating factor for Equation (4.5): it is
the factor by which Equation (4.5) needs to be multiplied in order that the
resulting differential equation has a left-hand side of the form (4.4), enabling
direct integration to be performed.
This leaves the question of how such an integrating factor can be found,
starting from Equation (4.5). The answer comes from writing down the two
properties that such a function p = p(x) must satisfy, as follows.
• Multiplying Equation (4.5) by p gives, on the left-hand side,
� �
dy 2x
p +p y.
dx 1 + x2
• The left-hand side must be of the form
dy dp
p + y. (4.4)
dx dx
Comparison of these two expressions shows that p must itself be a particular
solution of the differential equation
� �
dp 2x
= p. (4.6)
dx 1 + x2
This is a homogeneous linear first-order differential equation, and we can
solve it by separation of variables. Indeed, following Procedure 3.2, the
0)
equation becomes (for p =
� �
dp 2x
= dx.
p 1 + x2
Performing the left-hand integral gives
�
2x
ln |p| = dx,
1 + x2
so
��
2x
|p| = exp dx . (4.7)
1 + x2

��
��

Now, performing the integral on the right,

|p| = exp(ln(|1 + x2 | + A)) Note that 1 + x2 > 0, so
= exp(A)|1 + x2 | |1 + x2 | = 1 + x2 .

= D(1 + x2 ),
where D (= exp(A)) is a positive but otherwise arbitrary constant. Hence The case D = 0 corresponds
to the solution p = 0 of
p = ±D(1 + x ), 2
Equation (4.6), but this
solution is not of interest.
which, by redefining D, can be written as
p = D(1 + x2 ),
where D is now a non-zero but otherwise arbitrary constant.
Thus an integrating factor for Equation (4.5) is p(x) = D(1 + x2 ). Multi-
plying through the equation by this factor yields
dy
D(1 + x2 ) + 2Dxy = 3Dx2 ,
dx
and now you can see that (since D = 0) the arbitrary constant D can be
chosen without affecting the applicability of the form (4.4). Therefore we
choose the integrating factor to have the simplest possible form — in this
case we obtain p(x) = 1 + x2 .
As you have seen, this leads to the solution of Equation (4.5) by direct inte-
gration, and the formula for this integrating factor is given by Equation (4.7)
as
��
2x
p = exp dx . (4.8)
1 + x2
This approach generalizes to any linear first-order differential equation, pro-
vided that the integrals involved can be evaluated. For an equation written
in the form
dy
+ g(x)y = h(x), (4.1)
dx
the function g(x) takes the place of 2x/(1 + x2 ) in Equation (4.5). To find
an integrating factor p = p(x) for Equation (4.1), the argument proceeds as
above, with 2x/(1 + x2 ) replaced by g(x) at each step. This leads to the
generalized form of Equation (4.8), namely
��
p = exp g(x) dx , (4.9) Remember that calculation of
the integrating factor does
which defines the integrating factor for Equation (4.1). not require the inclusion of a
constant of integration.
When Equation (4.1) is multiplied through by the integrating factor, the
resulting differential equation is
dy
p(x) + p(x)g(x)y = p(x)h(x), (4.10)
dx
the left-hand side of which, by the definition of p, is of the form (4.4); so The definition of p ensures
Equation (4.10) can be re-expressed, using the Product Rule, as that the left-hand side of
Equation (4.10) is of the
d form (4.4) since
(p(x)y) = p(x)h(x). (4.11) � ��
dx dp d
= exp g(x) dx
Direct integration can then be used on Equation (4.11) to try to find the dx dx �
� �
general solution. = exp g(x) dx g(x)
This integrating factor method is summarized below. = p(x)g(x).

��
��

��

This method applies to diﬀerential equations of the form
dy
+ g(x)y = h(x). (4.1)
dx
(a) Determine the integrating factor
��
p = exp g(x) dx . (4.9) The constant of integration is
not needed here.
(b) Multiply Equation (4.1) by p(x) to recast the diﬀerential equation

as
You can, if you wish, check
dy that you have found p
p(x) + p(x)g(x)y = p(x)h(x). correctly by checking that
dx dy
(c) Rewrite the diﬀerential equation as p(x) + p(x)g(x)y
dx
d
d = (p(x)y) ,
(p(x)y) = p(x)h(x). dx
dx i.e. by checking that
(d) Integrate this last equation, to obtain dp/dx = p(x)g(x).
�
p(x)y = p(x)h(x) dx.
It is a good idea to check, by
(e) Divide through by p(x), to obtain the general solution in explicit substitution into the original
form. equation, that the function
obtained is indeed a solution.

As with the separation of variables method, it may not be possible to perform

the necessary ﬁnal integration. However, in the remainder of this subsection
we give examples and exercises for which this method can be used.

��
Use the integrating factor method to find the general solution of each of the The first example cannot be
following differential equations. solved by separation of
variables. The latter two can,
dy 2xy dy y−1 dy 2y
(a) =x− 2 (b) = (x > 0) (c) = as you saw in Exercise 3.4.
dx x +1 dx x dx 1 + x2 You can compare these
answers with those obtained
�� earlier.
(a) On rearranging the differential equation as
dy 2xy
+ = x,
dx x2 + 1
we see that it is in the form of Equation (4.1) with

g(x) = and h(x) = x.

x2 + 1

The integrating factor (from Equation (4.9)) is therefore

��
2x
p = exp dx
x2 + 1

= exp(ln |x2 + 1|)

= exp(ln(x2 + 1)) (since 1 + x2 > 0)

= x2 + 1. Checking, we see that
dp
= 2x = g(x)p(x).
dx

��
��

Multiplying both sides of the diﬀerential equation by this factor yields

dy
(x2 + 1) + 2xy = x(x2 + 1),
dx
and the diﬀerential equation thus becomes
d � 2 �

(x + 1)y = x(x2 + 1).

dx
Integrating both sides gives
�
(x + 1)y = x(x2 + 1) dx
2

�
= (x3 + x) dx

= 14 x4 + 21 x2 + C,
where C is an arbitrary constant. Finally, to obtain an explicit solution

we divide by x2 + 1 to obtain

x4 + 2x2 + 4C
y= .
4(x2 + 1)
(b) On rearranging the diﬀerential equation as
dy 1 1
− y=− ,
dx x x

we see that it is in the form of Equation (4.1) with g(x) = h(x) = −1/x.

The integrating factor (from Equation (4.9)) is therefore

��
1
p = exp − dx
x
= exp(− ln x) (since x > 0) Recall that a ln x = ln(xa )
� � �� and hence, in particular,
1
= exp ln − ln x = ln(x−1 ) = ln(1/x).
x
1
= . Checking, we see that
x
dp 1
Multiplying through the equation by p(x) = 1/x gives = − 2 = g(x)p(x).
dx x
1 dy 1 1
− 2 y = − 2 ,

x dx x x
and the diﬀerential equation becomes

� �
d 1 1
y = − 2.
dx x x
Integration then gives
� � �

y 1
= − 2 dx

x x
1
= + C,
x

where C is an arbitrary constant. The general solution is therefore

y = 1 + Cx,
where C is an arbitrary constant.

��
��

(c) In order to put the given diﬀerential equation into the form (4.1), we
need to bring the term in y to the left-hand side to obtain
dy 2
− y = 0. (4.12)
dx 1 + x2
Hence, in this case, we have g(x) = −2/(1 + x2 ) and h(x) = 0. The The equation is homogeneous.
integrating factor is
��
2
p = exp − dx = exp(−2 arctan x) = e−2 arctan x . Checking, we see that
1 + x2
dp −2e
−2 arctan x
Multiplying through by the integrating factor gives =
dx 1 + x2
� �
dy
−2 arctan x 2y −2 arctan x −2 arctan x 2
e − e = 0. =e −
dx 1 + x
2 1 + x2
Thus the diﬀerential equation can be rewritten as
= p(x)g(x).

d � −2 arctan x �
e y = 0.
dx
It follows, on integrating, that
e−2 arctan x y = C, or, equivalently, y = Ce2 arctan x ,
where C is an arbitrary constant. This is the general solution.

��

Find the general solution of each of the following differential equations. For part (b), see
dy dy Examples 1.1 and 2.1.
(a) − y = ex sin x (b) =y+x A direction field diagram is
dx dx shown in Figure 2.2.
��
Use the integrating factor method to solve each of the following initial-value You saw the differential
problems. equation in part (a) in
Exercise 3.7(a), where you
(a) u = xu, u(0) = 2. solved it using separation of
(b) tẏ + 2y = t2 , y(1) = 1. variables.

��

��
Which method would you use to try to solve each of the following linear
ﬁrst-order diﬀerential equations?
dy dy
(a) + x3 y = x5 (b) = x sin x
dx dx

dv dy
(c) + 5v = 0 (d) (1 + x2 ) + 2xy = 1 + x2

du dx
��
Solve each of the following initial-value problems. The diﬀerential equation in
part (a) is equivalent to that
(a) ẏ + y = t + 1, y(1) = 0. considered in parts (e) and (f)
(b) e3t ẏ = 1 − e3t y, y(0) = 3. of Exercise 1.2.

��
��

��

Find the general solution of each of the following diﬀerential equations.
dy
(a) x − 3y = x (x > 0)
dx

(b) + 4v = 3 cos 2t
dt

(Hint : If a and b are non-zero constants, then

�
eat
eat cos bt dt = 2 (a cos bt + b sin bt) + C,
a
+ b2
where C is an arbitrary constant.)

��

Arabic by Radio Book 1
93% (41)
Arabic by Radio Book 1
253 pages
STEP MAT TMUA Sample Pages
No ratings yet
STEP MAT TMUA Sample Pages
17 pages
Exam Paper MST125 2016F1
100% (1)
Exam Paper MST125 2016F1
13 pages
A Level Further Mathematics For OCR A Statistics Student Book 1 (AS/A Level) Scheme of Work
No ratings yet
A Level Further Mathematics For OCR A Statistics Student Book 1 (AS/A Level) Scheme of Work
21 pages
M208 Handbook
100% (3)
M208 Handbook
112 pages
Unit 3 MST 124
100% (1)
Unit 3 MST 124
121 pages
MST121 Chapter A3 Functions
No ratings yet
MST121 Chapter A3 Functions
64 pages
M208 Assignment Book 1
No ratings yet
M208 Assignment Book 1
12 pages
Kinetic of Fluid
No ratings yet
Kinetic of Fluid
63 pages
Book C: Essential Mathematics 1
No ratings yet
Book C: Essential Mathematics 1
288 pages
BONAHON - Low-Dimensional Geometry
No ratings yet
BONAHON - Low-Dimensional Geometry
393 pages
4th Class Examination & Certification Guide, 2014 Rev5
50% (2)
4th Class Examination & Certification Guide, 2014 Rev5
13 pages
Book A
No ratings yet
Book A
264 pages
MST121 Course Guide
No ratings yet
MST121 Course Guide
12 pages
MS221 Chapter C3 Taylor Polynomials
No ratings yet
MS221 Chapter C3 Taylor Polynomials
64 pages
Ebook mst326 Block3 E1i1 n9780749223151 l1
No ratings yet
Ebook mst326 Block3 E1i1 n9780749223151 l1
184 pages
Mst121 Chapter A2
No ratings yet
Mst121 Chapter A2
60 pages
Pure Mathematics (M208) Content Listing: Mathematical Language and Proof
No ratings yet
Pure Mathematics (M208) Content Listing: Mathematical Language and Proof
1 page
MST121 Chapter B1 Modelling With Sequences
No ratings yet
MST121 Chapter B1 Modelling With Sequences
64 pages
M4 - Review Excerise 1
No ratings yet
M4 - Review Excerise 1
132 pages
MST209 Bridging Material
No ratings yet
MST209 Bridging Material
71 pages
MST121 Handbook
No ratings yet
MST121 Handbook
48 pages
m248 Block C
No ratings yet
m248 Block C
123 pages
MU123 Week 4 Unit 4 11
No ratings yet
MU123 Week 4 Unit 4 11
35 pages
Mst121 Chapter A1
No ratings yet
Mst121 Chapter A1
52 pages
Functions I: Supporting Australian Mathematics Project
0% (1)
Functions I: Supporting Australian Mathematics Project
31 pages
Just The Maths - A.J.hobson (Complex Numbers)
No ratings yet
Just The Maths - A.J.hobson (Complex Numbers)
47 pages
Fermat's Last Theorem
100% (5)
Fermat's Last Theorem
18 pages
MST121 Handbook PDF
No ratings yet
MST121 Handbook PDF
48 pages
Group Theory m303 - 2
No ratings yet
Group Theory m303 - 2
53 pages
m208 3 Section 5
100% (1)
m208 3 Section 5
11 pages
Calculus 7th Edition - Ch03
No ratings yet
Calculus 7th Edition - Ch03
72 pages
Before Calculus: September 30, 2011 18:19 Ftoc Sheet Number 1 Page Number Xiii
No ratings yet
Before Calculus: September 30, 2011 18:19 Ftoc Sheet Number 1 Page Number Xiii
5 pages
Solomon C4 Vectors
No ratings yet
Solomon C4 Vectors
11 pages
Further Pure Mathematics
100% (2)
Further Pure Mathematics
170 pages
MST125 201509
No ratings yet
MST125 201509
12 pages
M140 Unit1
No ratings yet
M140 Unit1
81 pages
Mechanics 5 (M5)
No ratings yet
Mechanics 5 (M5)
133 pages
Gabriel Klambauer Problems and Propositions in Analysis Lecture Notes in Pure and Applied Mathematics 1979 PDF
100% (1)
Gabriel Klambauer Problems and Propositions in Analysis Lecture Notes in Pure and Applied Mathematics 1979 PDF
475 pages
APM2616 Assignment 03
No ratings yet
APM2616 Assignment 03
8 pages
M2 Solutionbank
100% (1)
M2 Solutionbank
604 pages
Partial Differential Equations: MST209 Mathematical Methods and Models
No ratings yet
Partial Differential Equations: MST209 Mathematical Methods and Models
32 pages
A Functional Determinant For The Riemann Xi Function
No ratings yet
A Functional Determinant For The Riemann Xi Function
12 pages
ComplexAnalysis Notes
100% (2)
ComplexAnalysis Notes
119 pages
Sequences and Series
100% (1)
Sequences and Series
34 pages
MST22U
No ratings yet
MST22U
56 pages
Mathematical Proofs by Examples
100% (1)
Mathematical Proofs by Examples
6 pages
Coordinate Geometry and Vector
No ratings yet
Coordinate Geometry and Vector
23 pages
Mathematics, Probability and Statistics, Applied Mathematics (6222 Titles)
0% (1)
Mathematics, Probability and Statistics, Applied Mathematics (6222 Titles)
260 pages
Thomas Little Heath Archimedes
No ratings yet
Thomas Little Heath Archimedes
101 pages
FP3 All
100% (1)
FP3 All
782 pages
Computers and Processors: Reference Manual
No ratings yet
Computers and Processors: Reference Manual
40 pages
AS Level Further Maths Specification - OCR
No ratings yet
AS Level Further Maths Specification - OCR
84 pages
Institute of Mathematics and Its Applications
No ratings yet
Institute of Mathematics and Its Applications
8 pages
Zeta Function
No ratings yet
Zeta Function
85 pages
Dama50 Vector Calculus-5
No ratings yet
Dama50 Vector Calculus-5
1 page
Geometric Progression, Sequences & Series Revision Notes From A-Level Maths Tutor
No ratings yet
Geometric Progression, Sequences & Series Revision Notes From A-Level Maths Tutor
6 pages
Unit 02.5 Geometry
No ratings yet
Unit 02.5 Geometry
64 pages
Inverse Trigonometric Functions (Trigonometry) Mathematics Question Bank
From Everand
Inverse Trigonometric Functions (Trigonometry) Mathematics Question Bank
Mohmmad Khaja Shareef
No ratings yet
Applications of Model Theory to Functional Analysis
From Everand
Applications of Model Theory to Functional Analysis
José Iovino
No ratings yet
Lebesgue Measure and Integration: An Introduction
From Everand
Lebesgue Measure and Integration: An Introduction
Frank Burk
No ratings yet
Introductions to Set and Functions
From Everand
Introductions to Set and Functions
Simone Malacrida
No ratings yet
SBRS Supplementary2015 - 2016
No ratings yet
SBRS Supplementary2015 - 2016
3 pages
Price List (Fully Built Houses) - As at June 2015: Phase I
No ratings yet
Price List (Fully Built Houses) - As at June 2015: Phase I
1 page
UWestern Ontario Winter 2015 Fee Schedule Grad INTL
No ratings yet
UWestern Ontario Winter 2015 Fee Schedule Grad INTL
1 page
PTDF Exam Tip
No ratings yet
PTDF Exam Tip
9 pages
Monthly Sumary Report August 2015
No ratings yet
Monthly Sumary Report August 2015
2 pages
U of Toronto Fees
No ratings yet
U of Toronto Fees
25 pages
International Student Application: Confirmation Number: 550070029 Date: 2014-Jan-20
No ratings yet
International Student Application: Confirmation Number: 550070029 Date: 2014-Jan-20
3 pages
Fre 122 Course Material
No ratings yet
Fre 122 Course Material
181 pages
UAlberta CVform
No ratings yet
UAlberta CVform
1 page
BC Power Engineering College
No ratings yet
BC Power Engineering College
7 pages
ARA011 Arabic Grammar 1
100% (2)
ARA011 Arabic Grammar 1
79 pages
Aspen Plus
No ratings yet
Aspen Plus
140 pages
Languages in Listings Format
No ratings yet
Languages in Listings Format
3 pages
Chapter 5. Numerical Integration: Answer Is A Number and Does Not Involve X. Notation Is Convenient
No ratings yet
Chapter 5. Numerical Integration: Answer Is A Number and Does Not Involve X. Notation Is Convenient
16 pages
Formulario Calculo Modificado Integral 2021
No ratings yet
Formulario Calculo Modificado Integral 2021
2 pages
Transmission Lines Parameters
No ratings yet
Transmission Lines Parameters
11 pages
Measure Theoretic Probability With Applications to Statistics Finance
No ratings yet
Measure Theoretic Probability With Applications to Statistics Finance
262 pages
Solution To HW-5: IEDA 3250
No ratings yet
Solution To HW-5: IEDA 3250
2 pages
Maths 2a
50% (2)
Maths 2a
6 pages
David_Mumford
No ratings yet
David_Mumford
7 pages
Exploring Mathematical Systems
No ratings yet
Exploring Mathematical Systems
8 pages
SSC Stenographer Question Paper 10th December 2024 Shift 2 English
No ratings yet
SSC Stenographer Question Paper 10th December 2024 Shift 2 English
41 pages
Class: 4 (Syllabus & Sample Questions)
No ratings yet
Class: 4 (Syllabus & Sample Questions)
2 pages
Markowitz-S Portfolio Theory (For INCHARGE)
No ratings yet
Markowitz-S Portfolio Theory (For INCHARGE)
3 pages
RMO 2013 Solutions 4
No ratings yet
RMO 2013 Solutions 4
2 pages
Chapter - Two Population and Water Demand
No ratings yet
Chapter - Two Population and Water Demand
97 pages
Data Mining For Humanity: An Overview
No ratings yet
Data Mining For Humanity: An Overview
4 pages
W. F. Chen, Plasticity For Structural Engineers, 1988-156
No ratings yet
W. F. Chen, Plasticity For Structural Engineers, 1988-156
1 page
Differential Geometry of Curves and Surfaces 2nd Edition Manfredo P. Do Carmo - Read the ebook now or download it for a full experience
100% (1)
Differential Geometry of Curves and Surfaces 2nd Edition Manfredo P. Do Carmo - Read the ebook now or download it for a full experience
80 pages
2 Unit 1 Impact of Jet
No ratings yet
2 Unit 1 Impact of Jet
16 pages
Ge-Mmw Beed 1
No ratings yet
Ge-Mmw Beed 1
12 pages
Optimization PDF
No ratings yet
Optimization PDF
9 pages
Vectors
No ratings yet
Vectors
8 pages
Math F112
No ratings yet
Math F112
3 pages
[FREE PDF sample] Structural Dynamics in Engineering Design 1st Edition Nuno M. M. Maia ebooks
100% (3)
[FREE PDF sample] Structural Dynamics in Engineering Design 1st Edition Nuno M. M. Maia ebooks
71 pages
Network Analysis
No ratings yet
Network Analysis
22 pages
Math Lv1 Tillett Factions Decimal Cycle 4
No ratings yet
Math Lv1 Tillett Factions Decimal Cycle 4
6 pages
Strength of Materials - Principal Stresses - Hani Aziz Ameen
100% (1)
Strength of Materials - Principal Stresses - Hani Aziz Ameen
32 pages
GTU DOM Paper 1
No ratings yet
GTU DOM Paper 1
2 pages
PSLP Lab File 1
No ratings yet
PSLP Lab File 1
50 pages
PV Curve For Voltage Stability
No ratings yet
PV Curve For Voltage Stability
1 page
Methods of Analysis: 5.1 First-Order Elastic Analysis
No ratings yet
Methods of Analysis: 5.1 First-Order Elastic Analysis
9 pages
Act3.2 1202-1ce-2a-Group1
No ratings yet
Act3.2 1202-1ce-2a-Group1
15 pages

MST209 Mathematical Methods and Models

Uploaded by

MST209 Mathematical Methods and Models

Uploaded by

MST209 Mathematical methods

First-order differential equations

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

An important class of the equations that arise in mathematics consists of

Subsection 1.1 develops a mathematical model for a speciﬁc situation which

��� ��� ��� ������ � ���

At this stage, we seek some relationship between the chosen variables. In

���� ��� ���

��� ��������� � �� �� � �

It is common practice in applied mathematics to reduce the proliferation of

These deﬁnitions have been framed in terms of an independent variable x

exercise (i.e. by substitution) that the supposed solution is indeed correct.

���� ��� ���

In many cases it is possible to describe the general solution of a ﬁrst-order

����� ��� ���

is a solution of Equation (1.4) on page 7.

We usually require that an initial-value problem should have a unique solu-

���� ��� ���

��� �������������� �� �� ��������

���� ��� ���

We deﬁne the relative error as

���� ��� ���

� ���� ���� ����� ��� ������� ������

��� ���� ���� �����

where k and M are positive constants. In certain circumstances this is a

Direction ﬁelds can be visualized by constructing the short line segments

An example is shown in Figure 2.2, which corresponds to the diﬀerential

���� ��� ���

Drawing by hand precise grids of line segments to represent direction ﬁelds

��� ������� ������

the graph of a solution, it is somewhat short on precision. We could not

Arguing as before, we have

��� ����� ���

i xi Yi f (xi , Yi ) = xi + Yi Yi+1 = Yi + hf (xi , Yi )

����� ��� ���

The solution of the initial-value problem given in Example 2.1 is in fact

0 0.1 0.2 0.3 0.4 x

Y100 when h = 0.01;

0.01 2.704 814 0.013 468 100

0.001 2.716 924 0.001 358 1000

0.0001 2.718 146 0.000 136 10000

be made as small as desired. In other words, the absolute error approaches

����� ��� ���

��� ������� ������ �� ��������� �� ��� �������

Use your computer to complete the following activities. PC

��������� ���� ���� ���

���� ��� ���

using a step size h = 0.1.

� ������� ������� ���������

��� ���� � �����������

��� ����� ��� ���� � �����������

where F (x) is an integral of f (x) and C is an arbitrary constant.

where C is an arbitrary constant.

(b) In order to satisfy the initial condition y(0) =

x = 0), we must have

variable. As usual, you should be prepared to translate this into situations

method of direct integration applies solely to ﬁrst-order diﬀerential equations

for which the derivative is equal to a function of the independent variable

where C is an arbitrary constant. (Here t is the independent variable and x

���� ��� ���

The method of direct integration succeeds in solving a diﬀerential equation

���� ��� ���

��� ����� � �������

��� ����� ��� ���������� �� ���������

sides with respect to x, to obtain

the general solution.

where the obvious choices for g and h are

g(x) = −x and h(y) = 1/(3y). Notice that since y > 0,

We now apply Procedure 3.2. On dividing through by h(y) = 1/(3y)

the diﬀerential equation becomes

is another arbitrary constant. However, we need to recognize that the

argument of the square root is positive, so we must have C − x

(b) The initial condition is y(0) = 3, so we substitute

����� ��� ���

declines over time due to radioactive decay. Its behaviour is modelled by

the diﬀerential equation

��

��

��

��

��

��

��

��

��

� ��

��

��

��

��

��

��

��

��

��

� ��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��