School of Mathematics

Lecture Series
Ordinary Differential Equations I
SMA 208

Adu A.M. Wasike

May 19, 2014

i
First impression May 19, 2014 School of Mathematics
University of Nairobi
P.O. Box 30197 Nairobi
email:[email protected]
Tel: +254722748492 Or +254733374383
c Adu Adenauer Wasike

All rights reserved. This work may not be translated or copied in whole or in part without
the written permission of the publisher author, except for brief excerpts in connection with
review or scholarly analysis. Use in connection with any form of information storage and
retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now or hereafter developed is forbidden.

The use of general descriptive names, trade names, trademarks, etc., in this publication,
even if the former are not especially identified, is not to be taken as a sign that such names,
as understood by trade Marks and Merchandise Act, may accordingly be used by anyone.

Photocomposed copy prepared from the authors , TE X file

Printed and bounded by Faculty of science, University of Nairobi, Nairobi Kenya.

ISBN-XXXX-XXXX-X-X

ii
 To students:
Who are the primary reason
for the existence of
our profession
and
this book

iii
Greeting
♥♣  ♠
Thank you for opening this book. Inside, you will find beautiful applications of the ideas and
techniques of calculus applied to the study of differential equations . It is not easy to convey
the beauty of the subject in the traditional first course on differential equations because
the number of equations that can be treated analytically are limited. We have in this course
introduced the concept of using differential equations to understand real life problems by
using basic calculus and graphical tools. These techniques are very valuable not only to
nonlinear problems but to the whole range of differential equation problems. Since we do
not intend to leave any void, we have also covered the traditional techniques. You are advised
to get acquitted to both techniques, qualitative and analytic.
Adu A.M. Wasike
May 19, 2014, Nairobi

♥♥♥
♥♥♥♥♥♥

♥♥♥
♥

iv
Contents

1 Lecture 1: Fundamental Ideas 3

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.2 What is a Differential Equation? . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.3 Definitions and Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . 4

1.3.1 Linearity and Degree of an ODE . . . . . . . . . . . . . . . . . . . . 6

2 Lecture 2: First-order Differential Equations:Analytic

Techniques 8
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2.2 The Separation of Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2.2.1 How do we solve separable differential equations? . . . . . . . . . . . 10

2.3 Initial Value Problem and more examples . . . . . . . . . . . . . . . . . . . . 13

3 Lecture 3: Some Applications 15

3.1 A savings model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

3.2 A Mixing Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

3.3 Orthogonal Trajectories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

4 Lecture 4: Equations with homogeneous coeffi-

cients 28
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

4.2 What is a homogeneous equation or function ? . . . . . . . . . . . . . . . . . 28

4.3 Method of Solution of Equations with homogeneous coefficients . . . . . . . 29

4.4 Equations reducible to homogeneous form . . . . . . . . . . . . . . . . . . . 31

4.4.1 x0 = f (at + bx + c). . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

at+bx+c
4.4.2 x0 = f ( αt+βx+γ ). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

v
5 Lecture 5: Exact Differential Equations 35
5.1 What is an exact differential equation? . . . . . . . . . . . . . . . . . . . . . 35

5.1.1 When is an equation exact? . . . . . . . . . . . . . . . . . . . . . . . 36

5.1.2 How do we solve an exact differential equation? . . . . . . . . . . . . 37

6 Lecture 6:Integrating Factors, Substitution Suggested

by the Equation 40
6.1 Integrating Factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

6.2 Substitution Suggested by the Equation . . . . . . . . . . . . . . . . . . . . 46

6.2.1 Change of Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

7 Lecture 7: Bernoulli’s Equations,

Ricatti Equations,
The general Linear Equation of first order 49
7.1 Bernoulli’s Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

7.2 Riccati Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

7.3 The general Linear Equation of first order . . . . . . . . . . . . . . . . . . . 53

8 Lecture 8: First Order First Degree differential Equa-

tions in three Variables 56
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

8.2 Total Differential Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

8.3 Simultaneous Total differential equations . . . . . . . . . . . . . . . . . . . . 58

8.3.1 Method of Grouping . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

8.4 Method of Multipliers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

9 Lecture 9: Qualitative Analysis of First order Dif-

ferential Equations 65
9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

vi
 9.2 Direction Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

9.3 An example of Qualitative Analysis . . . . . . . . . . . . . . . . . . . . . . . 68

9.4 Equilibrium Solutions, Stability, and Phase diagrams . . . . . . . . . . . . . 72

9.4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

9.4.2 Equilibria of Autonomous Equations . . . . . . . . . . . . . . . . . . 72

9.4.3 Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

9.4.4 Phase Line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

10 Lecture 10: Bifurcations 83

10.1 Introduction: What is bifurcation? . . . . . . . . . . . . . . . . . . . . . . . 83

10.2 The Bifurcation Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

10.3 When do bifurcations not occur? . . . . . . . . . . . . . . . . . . . . . . . . 85

10.3.1 Notation for differential equations depending on a parameter . . . . . 86

11 Lecture 11: Second order Linear Differential Equa-

tion 90
11.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

11.2 Second Order Linear Differential Equation: Basic Ideas . . . . . . . . . . . 90

11.3 Linear Homogeneous Equations: Basic Properties . . . . . . . . . . . . . . . 91

11.4 Linear Independence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

11.4.1 The Wronskian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

11.5 General solution of Homogeneous Linear equations . . . . . . . . . . . . . . 95

11.5.1 Method of solution of Second order Linear Homogeneous Equations

with constant coefficients . . . . . . . . . . . . . . . . . . . . . . . . . 96

11.6 Reduction of order Technique . . . . . . . . . . . . . . . . . . . . . . . . . . 98

11.7 Initial value problems for second order equations . . . . . . . . . . . . . . . 99

11.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

11.9 The nonhomogeneous equation of order two . . . . . . . . . . . . . . . . . . 101

vii
 11.10 Finding a particular solution of L(y) = b(x) . . . . . . . . . . . . . . . . . . 102

11.10.1 Method of Undetermined Coefficients . . . . . . . . . . . . . . . . . 102

11.10.2 Exrecise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

11.10.3 Method of Variation of constants . . . . . . . . . . . . . . . . . . . . 106

11.11 Euler-Cauchy Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

12 Lecture 12: Higher order Linear Equations 111

12.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

12.2 Linear Differential Equations: Basic Ideas . . . . . . . . . . . . . . . . . . . 111

12.3 Linear Independence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

12.3.1 The Wronskian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

12.4 General Solution of nth order Homogeneous Linear Equations With Constant
Coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

12.5 Linear Nonhomogeneous Equations of order n With Constant Coefficients . 117

12.5.1 Method of Undetermined Coefficients or

Guessing Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

12.5.2 Method of Variation of Parameters . . . . . . . . . . . . . . . . . . . 119

13 Lecture 13: Series Solution 122

13.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122

13.2 Power series: Taylor series approach . . . . . . . . . . . . . . . . . . . . . . . 122

13.3 Series Solutions: Taking Derivatives and Index Shifting . . . . . . . . . . . . 126

13.4 Series Solutions: First Examples . . . . . . . . . . . . . . . . . . . . . . . . 127

13.5 Series Solutions: Airy’s Equation . . . . . . . . . . . . . . . . . . . . . . . . 132

13.6 The Radius of Convergence of Series Solutions . . . . . . . . . . . . . . . . 135

14 Answers 138
♥♣  ♠

viii
What is in this course?
Introduction

These notes provide an introduction to both the quantitative and qualitative methods of solv-
ing ordinary differential equations. Emphasis is placed on first and second order equations
with constant coefficients. The equations studied are often derived directly from physical
considerations in applied problems.

Objectives of the course

At the end of the journey through the course you should be able to do the following:

• Identify and classify ordinary differential equations.

• State the characteristics for a function to be a solution of an ordinary differential

equation.

• Find solutions to certain simple first order ordinary differential equations by analytic
techniques; mainly by the methods of separation of variables, integrating factors, sub-
stitution.

• Determine some properties of the solution of an ordinary differential equation by qual-

itative and graphical methods; long-term behaviour of solutions.

• Derive and analyse ordinary differential equation as a mathematical model for some
physical phenomenon; population dynamics and others.

• Solve simple linear equation using the method of series.

Requirements

In this course we assume a good knowledge of SMA 103: calculus I, SMA 104: calculus II
and SMA 203: Linear Algebra I

Overview of the course

The course is divided into thirteen lectures of varying lengths. In Lecture 1 we, shall learn
basic definitions, terminologies and have a first glance at ordinary differential equations. In
Lectures 2 to 7 we learn how to find analytic solutions to various kinds of first order ordinary

1
differential equations. The understanding of these lectures is basic to the appreciation of the
usefulness and limitations of analytic methods. We also look at applications to some simple
real life problems. In Lecture 8, we look at more analytic techniques to differential equation
in three variables.

In Lecture 9 to 10, we devote a lot of time to qualitative techniques. These are the
methods used to analyse all kinds of ordinary differential equations. We develope both
graphical and use classical calculus techniques to study the nature of solutions even when
formulas for these solutions are not available. We shall mainly consider long-term behaviour
of solutions. It is a lecture that we shall find exciting! The lecture provides a window to
modern techniques of studying differential equations. We shall as a motto have a glance at
the applications, especially in population dynamics.

In Lecture 11, we get back to analytic techniques for finding solutions to second order lin-
ear differential equations. In Lecture 12, we look at higher order linear differential equations.
We finalise the course with Lecture 13 where simple series solutions are discussed.

Within each lecture, we have interludes of exercises. It is advisable to tackle them before
you proceed. Furthermore, at the end of each lecture we have several exercises that you have
to do! We of course, have solutions for self-checking.

As a culture we shall have an examination that is split into two sections: course work
tests that constitutes 30% and a final examination that will constitute 70%. The Final
examination has five questions. Question one will cover the entire course and carries 30%.
You will also be required to answer two other questions out of the remaining four questions
that will be testing an in depth understanding of some aspects of the course. Each of this will
carry 20%. The grading will be as follows: A 70% ≤ Marks ≤ 100 ; B 60% ≤ Marks ≤
69%; C 50% ≤ Marks ≤ 59%; D 40% ≤ Marks ≤ 49%; Fail E 0% ≤ Marks ≤ 39%.

There are many reference books that you will find useful to consult from time to time.
Please look at the reference list at the end of the book.

2
1 Lecture 1: Fundamental Ideas

1.1 Introduction

Our aim is to introduce the basic ideas that underlie the study of ordinary differential
equations(ODE).

Objectives:

At the end of this lecture we should be able to answer the following questions:

• What is a differential equation?

• What is meant by an order of a differential equation?
• What is a linear and nonlinear differential equation?
• What is an initial condition?
• What is meant by an explicit solution of a differential equation?

1.2 What is a Differential Equation?

A Motivating Example

Consider the problem of forecasting the future size of the population of some country. Census
data can be used to determine birth and death rates. For instance, we might find that among
every 1000 people alive today, on average there will be 25 births and 20 deaths in the next
year. We might also find that, on average, 3 people out of every 1000 emigrate every year
and a total of 100000 people immigrate. If the population today is 50 million, what will be
the population 10 years from now if present trends continue?
Notice the following aspects of the problem:

• Most of the relevant information describes the rate at which the population changes.
Namely; birth, death, immigration and emigration rates, which are expressed as people
per year. If p(t) represents the population at time t, then the stated information tells
us only about the rate of change of the population which is symbolised by dp(t) dt
. The
subject of differential equations is the study of equations that involve the derivative(s)
of an unknown function.
• To forecast the population 10 years from now, we must know the size of the popula-
tion today. (Identical demographic trends lead to different predictions of the future

3
 population if the initial population is 100 million instead of 50 million.) The present
population is an example of an initial condition, so called because we must know the
initial size of the population to forecast the future.

These information may be given symbolically thus:

dp(t)
= Birth rate − Death rate + Immigration rate − Emigration rate. (1)
dt
Observe the following from Equation(1). Births and immigration increase the population;
therefore, they contribute positively to the rate of population increase. On the other hand
deaths and emigration decrease the population; therefore, they are negative terms in the
equation.

According to the given information:

25
Birth rate = People (p(t)) per year,
1000
and
20
Death rate = People (p(t)) per year ,
1000
where p(t) is the present population. Similarly:

Emigration = 0.003p(t) per year .

The immigration rate is assumed to be constant; here,

Immigration = 100000 per year .

Thus Equation (1) with these symbols becomes:

dp(t)
= 0.025p(t) − 0.020p(t) − 0.003p(t) + 100000. (2)
dt
Equation(2) is an example of a differential equation.

1.3 Definitions and Terminology

From Equation(2), we observe the following:

• The passage of time t is independent of the population trends as described above. Thus
t is referred to as an independent variable in the differential equation(2).

• The population at any time p(t) depends on the time t (years). Thus p(t) is known as
a dependent variable.

4
 • The only derivative in Equation (2) dp(t)
dt
is with respect to the variable t and it is ordi-
nary. Thus Equation(2) is referred to as an ordinary differential equation (ODE).

dp(t)
• The highest-order derivative in Equation (2) is the first derivative dt
. Thus it is
referred to as a first order equation.

• The derivative in Equation (2) is with respect to one independent variable t.

We thus refer to Equation(2) as a first order ordinary differential equation.

Various notations are used to denote the dependent variable and the independent vari-
ables in differential equations. The derivative dp
dt
may be expressed as p0 , ṗ. Where the prime
or dot indicate differentiation with respect to time. For purpose of notational brevity, we
also simply write p instead of p(t).

Example 1. The equations

ṗ = p + 3t and ṗ = p2
are first-order ODEs. The independent variable is t the dependent variable is p.
The equation
p̈ + ṗ + p = 0
is a second-order ODE because it involves the second derivative of the function p. As above,
the independent variable is t.
The equation
∂u ∂ 2 u
=
∂t ∂x2
is a second-order partial differential Equation (PDE) because it involves partial derivatives
of u with the highest being of order two. The dependent variable is u, and the independent
variables are t and x.
The Equation
˙ = −x(t − τ ),
x(t)
where τ > 0, is an example of a delay differential equation(DDE).

Now the formal definitions are:

1.1 Definition. An equation involving independent and dependent variables and derivatives
or differentials of one or more variables with respect to one or more independent variables
is called a differential equation.

1.2 Definition. A differential equation which involves derivatives with respect to a single
independent variable is known as an Ordinary differential equation

1.3 Definition. The order of a differential equation is the order of the highest-ordered
derivative in the equation.

5
Exercise 1. In each of the following, determine the order, the independent variable and the
dependent variable:

1 x2 + y 2 − 2xy dy
dx
= 1,
d2 y
2 dx2
+ ω 2 y = 0.

1.3.1 Linearity and Degree of an ODE

The equation
F (t, y, ẏ, ÿ, . . . , y (n) ) = 0, (3)
where n ≥ 1 is a positive whole number, is called an nth-order ordinary differential equation.
It is referred to as ordinary since the only derivatives involved in the equation are ordinary
derivatives. Under suitable restrictions on the function F , (3) can be solved explicitly for
y (n), in terms of n + 1 variables to obtain
y (n) = f (t, y, ẏ, ÿ, . . . , y (n−1) ), (4)
where
dn y
y (n) = .
dtn
For example
(ẏ)2 + 4ẏ − 6t2 = 0 (5)
could be written as √
ẏ = −2 ± 4 + 6t2 .
We notice that Equation (5) has products of y and/or its derivatives with themselves. This
kind of equation is said to be nonlinear. We are thus lead to the following definition.
1.4 Definition. Linearity. An equation
F (t, y, ẏ, . . . , y (n) ) = 0
is called linear if the function F is linear in y, ẏ, ÿ, . . . , y (n).

Thus, a general linear equation of order n may be written as

Xn
ai (t)y (n−i) = r(t),
i=0

where ai (t), and r(t) are functions of t.

We also notice that, in Equation(5), the highest power to which the highest derivative,
ẏ, is raised is 1. We say that this equation is of degree 1. Equations
w2
(ẍ)2 +
(1 + ẋ2 ), (x − tẋ)2 = k 2 [1 + ẋ2 ], k 2(ẍ)2 = (1 + ẋ2)3 , (6)
H2
are all of second degree.
1.5 Definition. The degree of a differential equation is the highest order derivative present
in the equation, after the differential equation has been made free from radicals and fractions
as far as the derivatives are concerned.

6
Solution

One of the problems in ODEs is usually to determine a function φ(t) that satisfies a differ-
ential equation. Consider, for instance, the first order differential equation

ẋ = f (t, x). (7)

A function φ(t) is a solution to Equation(7) if,

˙ = f (t, φ(t)).
φ(t)

When we substitute the function φ(t) := 1 + 2 cos t in the differential equation

ẍ + x = 1, (8)

we find,
ẍ(= −2 cos t) + x(= 1 + 2 cos t) = 1.
The equation is identically satisfied. The function φ(t) = 1 + 2 cos t is, therefore, said to
be a solution to the differential equation (8).

Check that
φ(t) = e2t
is a solution to the equation
ÿ + ẏ − 6y = 0.
Let us have a quick check on wheather we have followed this brief lecture. Please do:

Exercise 2.

1. For each of the following, describe the equation completely; state whether the equation
is ordinary, partial linear or non-linear, and give its order.

∂ 2u 2
2∂ u
(a) (t2+x2 ) dt+2tx dx = 0, (b) ẍ+k 2x = 0, (c) = a , (d) xẍ = t,
∂t2 ∂x2
tan t 1
(e) y (4) = w(t), (f) t(ẍ)3 +(ẋ)4 −x = 0 (g) ẍ−tan tẋ− x = 2 x3 .
t t

2. Show that the functions φ1(t) = e−t and φ2 (t) = te−t are each a solution to the equation
ẍ + 2ẋ + x = 0.

3. Show that c1e−t + c2 te−t is a solution to ẍ + 2ẋ + x = 0, where c1, c2 are constants.

4. Show that the function x(t) = cat, with a and c as constant, is a solution to the equation
ẋ = (ln a)x.

7
2 Lecture 2: First-order Differential Equations:Analytic
Techniques

2.1 Introduction

In this lecture, we describe procedures for determining solutions by analytical means to some
type of first order equations. We consider differential equations that may be written as
dx
= f (t, x). (9)
dt
Equation(9) can be written in the form

M(t, x)dt + N (t, x)dx, (10)

where the functions M(t, x) and N (t, x), have been derived from f (t, x). For example
dx t 2 + x2
=− =: f (t, x) can be written thus (t2 + x2 )dt + (x2 − t)dx = 0. (11)
dt t − x2
We shall use either form in this sequel.

The nature of the function f (t, x), and thus the functions M(t, x), N(t, x), in Equations
(9) and (10) respectively, dictates the method one applies in getting an analytic solution to
the first order equation. We thus scan through every equation and determine an appropriate
method of solving it.

Objectives

At the end of the lecture we should be able to:

• Determine if an equation is separable

• Solve separable equations
We now take the path of exploiting the characteristics of f (t, x) to get solutions for some
equations.

2.2 The Separation of Variables

In my view, this is a very important section for solving odes. You must grasp it
by all possible means!

8
What is a separable equation?

A differential equation is called separable if the function f (t, x) can be written as the
product of two functions: one that depends on t alone and another that depends only on x.
That is, a differential equation is separable if it assumes the form:
dx
= h(x)g(t). (12)
dt
Example 2. The differential equation
dx
= xt, is clearly separable,
dt
while equation
dx
= x + t, is not.
dt

At times, we might have to do a little work to see that an equation is separable. For
instance

Example 3. The equation

dx t+1 dx  t + 1  1 
= , is separable since we can write it as = .
dt tx + t dt t x+1

Two important types of separable equations occur if either t or x is missing from the
right-hand side of the given equation.

Example 4. The differential equation

dx
= g(t), (13)
dt
is separable since we regard the right-hand side as g(t).1, where we consider 1 as a (very
simple) function of x. Similarly
dx
= h(x), (14)
dt
is also separable.

This last type of differential equation, Equation(14), is said to be autonomous. Many

of the most important first-order differential equations that arise in applications are au-
tonomous. For instance, the right-hand of the equation
dp  p
= rp 1 − , (15)
dt K
depends on the dependent variable p alone, so this is autonomous.

9
2.2.1 How do we solve separable differential equations?

We begin with a simple example. Consider the differential equation:

dx t
= 2. (16)
dt x
The first step one takes is to separate variables in such a manner that the function of x
“sits”close to and, on the left, of the differential dx while the function of t “sit ”close to
and, on the left, of the differential dt. Thus through some “informal”algebra, Equation(16)
becomes
x2 dx = tdt.
Integrating both sides we have: Z Z
2
x dx = tdt

that yields
1 3 1 2
x = t + C(constant).
3 2
Technically there are constants of integration on both sides of this equation, but we can
lump them together as a single constant C on the right!

What is really going on in the informal algebra above?

If you peruse through the above example, you probably become nervous at one point. Treat-
ing dt as a variable! It sounds criminal mathematically. However, it is a tip-off that some-
thing a little more complicated is actually going on. Here is the real story.

We began with a separable equation

dx
= g(t)h(x), (17)
dt
and then rewrote it as
1 dx
= g(t), assume h(x) 6= 0. (18)
h(x) dt
This equation actually has a function of t on both sides of the equal sign since x = x(t) is a
function of t. So we really should write as
1 dx
= g(t). (19)
h(x(t)) dt
In this form, we can integrate both sides with respect to t to get
Z Z
1 dx
dt = g(t) dt. (20)
h(x(t)) dt

Now for the important step: We make a “u−substitution”just as in calculus by replacing

the function x(t) by a new variable, say x. (In this case, the substitution is actually an x−

10
substitution). Of course, we must also replace the expression (dx/dt)dt by dx. The method
of substitution from calculus tells us that
Z Z
1 dx 1
dt = dx, (21)
h(x(t)) dt h(x)

and therefore we can combine the equations in Equations(20), (21) to obtain:

Z Z
1
dx = g(t)dt + C. (22)
h(x)

Hence, we can integrate the left-hand side with respect to x and the right-hand side with
respect to t.

Separating variables and multiplying both sides of the differential equation by dt is simply
a notational convention that helps us remember the method. It is justified by the foregoing
argument.

Equation(22) is an implicit solution of Equation(12) insofar as it expresses a function

of x as a function of t. The constant C is arbitrary, so the separation of variables yields
a general solution. Each choice of C leads to different functional relationship between x
and t; the corresponding graphs are called integral curves. The value of C can usually
be determined if one point on the integral curve is prescribed aprior. This point is usually
called an initial condition.

A little example will surely shade some light on this somewhat hazy ideas!

Example 5. Find the solution to the differential equation

dx
= ax, subject to x(0) = 2. (23)
dt

Solution 5. By separating variables: we have

dx
= adt
Z x Z
dx
= a dt + C
x
ln x = at + C(constant)
x(t) = eat eC
x(t) = Keat

This is what we have done:

In the first equation we separated variables. In the second equation we have integrated with
C as a constant of integration. In the fourth equation we have taken the exponential. In
the fifth equation we have K = eC . The solution x(t) in the fifth equation is referred to as
a general solution to the Equation(23). If we substitute the value of x at t = 0, x(0) = 2,
we find that x(0) = 2 = C. Thus the specific solution to Equation(23) is x(t) = 2eat .

At times we say that x(t) = 2eat is a solution to Equation (23) passing through (0, 2).

11
 A look at the graphs will give some insight into understanding, a general solution, and a
specific solution. We give in Figure 2.1 graphs for a > 0, x(0) = 2, x(0) = 3, x(0) = 4; while
in Figure 2.2 graphs for a < 0, x(0) = 2, x(0) = 3, x(0) = 4 are given.

Integral curves for x.=ax, a=2

250
x(0)=2
x(0)=3
x(0)=4

200

150
(x(t)

100

50

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time t

Figure 2.1 Integral curves for Equation(23) with a > 0

.
Integral curves for x =ax, a=−2
4
x(0)=2
x(0)=3
x(0)=4
3.5

2.5
(x(t)

1.5

0.5

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time t

Figure 2.2 Integral curves for Equation(23) with a < 0

We see that for different values of C, we get different integral curves. We also see that once
x(0) is specified, we get a specific integral curve for a corresponding value of a.

From the graphs in Figures 2.1 and 2.2, we can describe how each solution changes as
time changes. We shall, in simple cases give graphical solutions as they provide a vivid
picture of how a solution depends on t and other parameters as a.

12
2.3 Initial Value Problem and more examples

2.1 Definition. A differential equation ẋ = f (t, x) subject to an initial condition x(t0) = x0

is called an initial value problem(IVP).

The time t0 is any time of interest at which x is known.

Further examples:

Example 6. Solve the equation

dx
= h(x)g(t). (24)
dt

Solution 6. By straightforward integration in the following way

Z Z
dx
= g(t) + C dt. (25)
h(x)

Furthermore if (24) is subject to x(t0) = x0, then the solution is given thus
Z x Z t
ds
= g(σ) dσ. (26)
x0 h(s) t0

Example 7. Solve
2x
ẋ = ,
t
for t > 0 and x > 0.

Solution 7.
dx 2x
=
Z dt Zt
dx dt
= ,
2x t
1/2 ln x = ln t + c,
ln x = 2 ln t + 2C,
x = e2c t2,
x = Kt2 ,

where (K = e2c constant).

Example 8. Solve
(1 + x2 )dt + (1 + t2)dx = 0.

Solution 8. Dividing through by (1 + x2)(1 + t2), and integrating, we get:

Z Z
dt dx
2
=− ,
1+t 1 + x2

13
that upon solving yields
tan−1 t = − tan−1 x + C.
Simplifying further we obtain  t+x 
tan−1 =c
1 − tx
or
t+x
= K,
1 − tx
where K = tan c.

We see that it is not difficult to solve a separable differential equation ! By the way all
subsequent methods are related to it. That is why you have to check your understanding
before you move on.

Exercise 3. Separation of Variables

Solve the differential equations below

dx dx dx
1. = tx 2. = t4 x 3. = 2x + 1
dt dt dt
dx dx dx
4. = 2−x 5. = e−x 6. = (tx)2
dt dt dt
dx t dx 1 dx 1
7. = 8. = 9. = 1+ 2
dt 1 + x2 dt tx + t + x + 1 dt x
dx etx dx t2 + 1 dw w
10. = 11. = 12. =
dt 1 + x2 dt x4 + 3x dt t
In exercise 13-21 solve the given initial value problem.
dx dx dx 1
13. = 2x+1, x(0) = 3 14. = tx2 +2x2 , x(0) = 1 15. = −xt, x(0) = √
dt dt dt π

dx dx t dx t2
16. = −x2, x(0) = 0 17. = , x(0) = 4 18. = , x(0) = −2.
dt dt x − t2 x dt x + t3 x
19. θ̇ = 1−2k cos 2θ, θ(0) = 0. 20. ẋ = (x2 +1)t, x(0) = 1 21. ṙ = r(1−r2 ), r(0) = 1

14
3 Lecture 3: Some Applications

Objectives

At the end of the lecture we should be able to:

• Apply some differential equations to real life problems such as in:

◦ A savings model
◦ Mixing problem
◦ Newton’s Law of cooling
◦ Radioactive decay
◦ Geometry; Orthogonal trajectories
A simple example on something useful, money should be interesting especially when you
want to spent some savings.

3.1 A savings model

Example 9. Suppose we deposited ksh. 50000 in a savings account with interest accruing
at the rate of 5% compounded continuously.
Write down a differential equation to represent this information.
What will be the amount on the account after 10 years?

Solution 9. Let A(t) denote the amount of money in the account at time t, then the
differential equation for A is
5
Ȧ = A, A(0) = 50000.
100
Solving this equation we find that

A(t) = A(0)e0.05t = 50000e0.05t .

Assuming interest rates never change, then after 10 years we will have

A(10) = 50000e0.05×10 = 50000e0.5 ≈ 82440

shillings on the account.

Now suppose we decide to withdraw sh. 10000 per year from the account after this 10
years. Two questions arise:

15
What is the differential equation for this scenario?
After how long will all the money get finished?

Solution 9. The differential equation for A(t) must change, but only beginning in year
10. For 0 ≤ t ≤ 10, our previous model works fine. However, for t > 10, the differential
equation becomes
Ȧ = 0.05A − 10000.
Thus we really have a differential equation of the form

0.05A, for t ≤ 10;
Ȧ =
0.05A − 10000, for t > 10,

whose right-hand consists of two pieces.

To solve this equation, we solve the first case, for 0 < t ≤ 10, and determine A(10),
already done. Then we solve the second case, t > 10 with A(10) = 82440 as the initial
condition. This equation is also separable, and we have:
Z Z
dA
= dt.
0.05A − 10000
Solving it we find that
A(t) = 200000 − 117560e0.05t . (27)
At t = 10, we have
dA
= 0.05 × 82440 − 10000 − 5878 < 0. (28)
dt
Clearly we can see from Equation(28) that our account is being depleted. After how many
years t will it tend to zero? Depletion occurs when A(t) = 0. So let us persue this line of
thought.

0 = 200000 − 117560e0.05t
117560e0.05t = 200000
200000
e0.05t =
117560
1 200000 
t = ln
0.05 117560
t = 10.627
1
t = 10 years and 7 months
2
So under this scheme of saving and expenditure, one should be broke after 20 years and 7 12
months.

3.2 A Mixing Problem

The name mixing problem refers to a large class of different problems where two or more
substances are mixed together. This can range from pollutants in a lake to chemicals in a
vat to cigar smoke in the air in a room to spices in a serving of curry.

16
Mixing in a vat. Consider a large vat containing sugar-water that is to be made into soft
drinks( See figure below).
Suppose:

• The vat contains 100 litres of liquid. Moreover, the amount flowing in equals the
amount flowing out, so there are always 100 litres in the vat.
• The vat is kept well mixed, so the concentration of sugar is uniform throughout the
vat.
• Sugar-water containing 5 tablespoons of sugar per liter enters the vat through Pipe A
at a rate of 2 litres per minute.
• Sugar-water containing 10 tablespoons of sugar per liter enters the vat through Pipe
B at a rate of 1 litres per minute.
• Sugar-water leaves the vat through Pipe C at a rate of 3 litres per minute.

Figure 2.1 Mixing Vat

To make the model, we let t be time measured in minutes (the independent variable).
For the dependent variable, we have two choices. We could choose either the total amount
of sugar, S(t), in the vat at time t measured in tablespoons, or C(t), the concentration of
sugar in the vat at time t measured in tablespoons per litre. We will develop the model for
S, leaving the model for C as an exercise.

Using the total sugar S(t) in the vat as the dependent variable, the rate of change of S
is the difference between the amount of sugar being added and the amount of sugar being
removed. The sugar entering the vat comes from pipes A and B and can be easily computed
by multiplying the number of litres per minute of sugar mixture entering the vat by the
amount of sugar per litre. The amount of sugar leaving the vat through pipe C at any given
moment depends on the concentration of sugar in the vat at that moment. The concentration
is given by S/100, so the sugar leaving the vat is the product of the number of litres leaving
per minute(3 litres per minute) and the concentration (S/100). The model is
dS S
= 2| {z
× 5} + |1 ×
{z10} − 3× . (29)
dt 100
| {z }
sugar in from Pipe A sugar in from Pipe B
sugar out from Pipe C

That is,
dS 3S 2000 − 3S
= 20 − = . (30)
dt 100 100

17
By the method of separation of variables, we have:
2000  2000 − 3S(0)  −0.03t
S(t) = − e ,
3 3
where S(0) is the total amount of sugar in the vat at time t = 0.

Exercise 4.

Question 1. A 100-litre bucket is full of pure water. Suppose we begin dumping salt into
the bucket at a rate of 14 kilogram per minute. Also, we open the spigot so that 10 litres per
minute leaves the bucket, and add pure water to keep the bucket full. IF water is always
well mixed, what is the amount of salt in the bucket after

(a)1 minute (b)10 minutes (c) 60 minutes

(d)1000 minutes (e) a very long time?

Question 2. High levels of cholesterol in the blood are known to be a risk factor for heart
disease. Cholesterol is manufactured by the body for use in the construction of cell walls
and is absorbed from foods containing cholesterol. The following is a very simple model of
blood cholesterol levels. Let C(t) be the amount of cholesterol in the blood of a particular
person at time t(in milligrams per deciliter). Then
dC
= k1 (C0 − C) + k2 E, (31)
dt
where
C0 = the person’s “natural”cholesterol level,
k1 = “production”parameter,
E = amount of cholesterol eaten (per day), and
k2 = “absorption”parameter.

(a) Suppose C0 = 200, k1 = 0.1, k2 = 0.1, E = 400, and C(0) = 150. What will the person’s
cholesterol level be after 2 days on this diet?
(b) With the initial conditions as above, what will the person’s cholesterol level be after 5
days on this diet?

(c) What will the person’s cholesterol level be after a long time on this diet?
(d) Suppose that, after a long time on the high cholesterol diet described above, the person
goes on very low cholesterol diet, so E changes to E = 100. (The initial cholesterol
level at the starting time of this diet is the result of part (c).) What will the person’s
cholesterol level be after a 1 day on the new diet, after 5 days on the new diet, after a
very long time on the new diet?

18
 (e) Suppose the person stays on the high cholesterol diet but takes drugs that block some
of the uptake of cholesterol from food, so k2 changes to k2 = 0.075. With the cholesterol
level from part (c), what will the person’s cholesterol level be after a 1 day on the new
diet, after 5 days, after a very long time?

Question 3. Many radioactive materials disintegrate at a rate proportional to the amount

present. For example, if the radioactive material present now, at time t, is Q(t), then the
rate of change of Q(t) with respect to time t is given by
dQ
= −rQ,
dt
where r is a positive constant. The time taken for the material to disintegrate to half its
present amount is called its half-life.

If the half life of some radioactive material is 16 days and, you wish to have 30g at the
end of 30 days, how much radioactive material would you start with?

Question 4. From experimental observation it is known that, upto a “satisfactory ”ap-

proximation, that the surface temperature of an object changes at a rate proportional to its
relative temperature. That is, the difference between its temperature and the temperature
of the surrounding environment. This is what is known as Newton’s law of cooling. Thus
if, θ(t) is the temperature of the object at time t, then we have

θ̇ = −k(θ − S), (32)

where S is the temperature of the surrounding environment, k > 0.

Suppose a corpse was discovered in a motel room at midnight and its temperature was
27 C. The temperature of the room was kept constant at 16◦ C. Two hours later the
◦

temperature of the corpse dropped to 24◦ C. Find the time of death.

3.3 Orthogonal Trajectories

We have seen before (see separable equations for example) that the solutions of a differential
equation may be given by an implicit equation with a parameter something like

F (y, t) = C.

This is an equation describing a family of curves. Whenever we fix the parameter C we

get one curve and vice-versa. For example, consider the families of curves

y = mt or y 2 + t2 = C 2

where m and C are parameters. Clearly, we may change the names of the variables and still
have the same geometric curves. For example, the above families define the same geometric
object as
y = mx or y 2 + x2 = C 2

19
Note that the first family describes all the lines passing through the origin (0, 0) while the
second the family describes all the circles centered at the origin (including the limit case
when the radius 0 which reduces to the single point (0, 0)) see the Figure 2.3.

Figure 2.3

and

Figure 2.4

20
In this page, we will only use the variables x and y. Any family of curves will be writ-
ten as
F (x, y, C) = 0
One may ask whether any family of curves may be generated from a differential equation?
In general, the answer is no. Let us see how to proceed if the answer were to be yes. First
differentiate with respect to x, and get a new equation involving, in general x, y, y 0, and
C. Using the original equation, we may able to eliminate the parameter C from the new
equation.

Example 10. Find the differential equation satisfied by the family

x2 + y 2 = Cx

Solution 10. We differentiate with respect to x, to get

2x + 2yy 0 = C

Since we have
x2 + y 2
C=
x
then we get
x2 + y 2
2x + 2yy 0 = (33)
x
You may want to do some algebra to make the new equation easy to read. The next step is
to rewrite this equation in the explicit form

y 0 = f (x, y)

this is the desired differential equation.

Example 11. Find the differential equation (in the explicit form) satisfied by the family

x2 + y 2 = Cx

Solution 11 . We have already found the differential equation in the implicit form,
Equation(33). Algebraic manipulations give

0 1  x2 + y 2  y 2 − x2
y = − 2x = .
2y x 2xy
Let us reconsider the example of the two families

y = mt or y 2 + t2 = C 2

If we draw the two families together on the same graph we get

21
 Figure 2.5

As we see here something amazing happened. Indeed, it is clear that whenever one line
intersects one circle, the tangent line to the circle (at the point of intersection) and the
line are perpendicular or orthogonal. We say the two curves are orthogonal at the point of
intersection.
3.1 Definition. Consider two families of curves Γ1 and Γ2 . We say that Γ1 and Γ2 are
orthogonal whenever any curve from Γ1 intersects any curve from Γ2 , the two curves are
orthogonal at the point of intersection.

For example, we have seen that the families y = mx and x2 + y 2 = C 2 are orthogonal.
One may then ask the following natural question:
Given a family of curves Γ, is it possible to find a family of curves which is
orthogonal to Γ?
The answer to this question has many implications in many areas such as physics, fluid-
dynamics, etc.. In general this question is very difficult. But in some cases, we may be able
to carry on the calculations and find the orthogonal family. Let us show how.

Consider the family of curves Γ. We assume that an associated differential equation may
be found, say
y 0 = f (x, y).
We know that for any curve from the family passing by the point (x, y), the slope of the
tangent at this point is f (x, y). Hence the slope of the line perpendicular (or orthogonal) to
1
this tangent is − f (x,y) which happens to be the slope of the tangent line to the orthogonal
curve passing by the point (x, y). In other words, the family of orthogonal curves are solutions
to the differential equation
1
y0 = − .
f (x, y)

22
 From this we see what we have to do. Indeed consider a family of curves Γ. In order to find
the orthogonal family, we use the following practical steps:

Step 1. Find the associated differential equation.

Step 2. Rewrite this differential equation in the explicit form

y 0 = f (x, y).

Step 3. Write down the differential equation associated to the orthogonal family
1
y0 = − .
f (x, y)

Step 4. Solve the new equation. The solutions are exactly the family of orthogonal curves.

Step 5. You may be asked to give a geometric view of the two families. Also you may be asked
to find a specific curve from the orthogonal family (something like an IVP).

Example 12. Find the orthogonal family to the family of circles

x2 + y 2 = C 2 .

Solution 12. First, we look for the differential equation satisfied by the circles. We
differentiate with respect to the variable x to get

2yy 0 + 2x = 0

We rewrite this equation in the explicit form

x
y0 = −
y
Next we write down the equation for the orthogonal family
1 y
y0 = − x =
−( y ) x

This is a linear as well as a separable equation. If we use the technique separation of variables,
we get the integrating factor
y = mx
We recognize the family of lines and we confirm our earlier observation (that the two families
are indeed orthogonal). This example is somehow easy and was given here to illustrate the
technique.

Example 13. Find the orthogonal family to the family of circles

x2 + y 2 = 2cx

23
 Figure 2.6

Solution 13. We have seen before that the explicit differential equation associated to the
family of circles is
y 2 − x2
y0 = .
2xy
Hence the differential equation for the orthogonal family is
2xy
y0 = .
x2 − y2
We recognize an homogeneous equation. Let us use the technique developed to solve this
kind of equations. Consider the new variable v = yx (or equivalently y = xv). Then we have

y 0 = xv 0 + v

and
2xy 2v
=
x2 −y 2 1 − v2
Hence we have
2v
4xv 0 + v = .
1 − v2
Algebraic manipulations imply
1 v + v3
v0 = .
x 1 − v2
This is a separable equation. The constant solutions are given by

v + v3 = 0

24
which gives v = 0. The non-constant solutions are found once we separate the variables

v − v2 1
3
dv = dx.
v+v x
and then we integrate Z Z
v − v2 1
dv = dx
v + v3 x
Before we perform the integration for the left-hand side, we need to use partial decomposition
technique. We have
v − v2 1 − v2 A Bv + C
3 3
= + .
v+v v+v v 1 + v2
We will leave the details to you to show that A = 1, B = −2, and C = 0. Hence we have
Z Z 
v − v2 1 2 
dv = − dv = ln |v| − ln(v 2 + 1).
v + v3 v 1 + v2
Hence
ln |v| − ln(v 2 + 1) = ln |x| + C
which is equivalent to
v
= Cx
v2
+1
where C =
6 0. Putting all the solutions together we get

v=0
v
v 2 +1
= Cx

Going back to the variable y, we get


y=0
y
y 2 +x2
=C

which is equivalent to 
y=0
y 2 + x2 = my
We recognize a family of circles centered on the y-axis and the line y = 0 (the x-axis which
was easy to guess, isn’t it?) See Figure 2.7 for the curves y 2 + x2 = my

25
 Figure 2.7

If we put both families together, we appreciate better the orthogonality of the curves (see
Figure 2.8).

Figure 2.8 Orthogonality for y = 0, y 2 + x2 = my

26
Exercise 5. Orthogonal Trajectories

(a) Find an equation of the family orthogonal to the family y = cx.

(b)Find the family orthogonal to the family y = ce−x of exponential curves. Determine the
member of each family passing through (0, 4).
(c)Find the orthogonal trajectories of x2 + y 2 = cx.
(d)Find the orthogonal trajectories of the family y = x + ce−x and determine that particular
member of each family that passes through (0, 3).
(e)Find the orthogonal trajectories of the family of rectangular hyperbolas y = c/x.
(f)Find an equation of orthogonal trajectory of the family of circles having a polar equation
r = f (θ) = 2a cos θ.

27
4 Lecture 4: Equations with homogeneous coefficients

Objectives

At the end of the lecture we should be able to:

• Solve first order Odes with homogeneous coefficients;

• Solve Equations which are reducible to ones with homogeneous coefficients.

4.1 Introduction

In this section, we consider the so called homogeneous first order differential equations and
the method(s) for solving such equations. We shall also consider non-homogeneous equations
but which can be reduced (or transformed) to homogeneous equations.

4.2 What is a homogeneous equation or function ?

4.1 Definition. A function f (t, x) is said to be homogeneous of degree k in t and x if, and
only if
f (λt, λx) = λk f (t, x) (34)
for some constants λ and k ≥ 0.

Example 14. f (t, x) = t4 − t3x is homogeneous of degree 4 since

f (λt, λx) = (λt)4 − (λt)3 (λx),

= λ4 (t4 − t3x),
= λ4 f (t, x).

Check That
x x
f (t, x) = e t + tan( )
t
is homogeneous of degree 0.

Check That
f (t, x) = t2 + sin t cos x
is not homogeneous.

28
Exercise 6. Determine whether each of the following functions is homogeneous or not.
Indicate the degree of homogeneity
3x
1. 4t2 − 3tx + x2 2. et 3. tan( )
y

x2 + 3xy
4. tan x 5. (x2 +y 2) exp(2x/y)+4xy 6. .
x − 2y
1 t 2t
7. x5/(x2 +2y 2 ) 8. t ln t−x ln x 9. +tan( )+ln( +1).
(x/t)2 x x
Theorem 4.1. If M(t, x) and N (t, x) are both homogeneous and of the same degree, then
the function
M(t, x)
N (t, x)
is homogeneous of degree zero.

Exercise 7. Proof Theorem 4.1.

x
Theorem 4.2. If f (t, x) is homogeneous of degree zero then f (t, x) is a function of t
alone.

Proof. Let u(t) = xt . We show that if f (t, x) is homogeneous of degree zero then f (t, x) is a
function of u alone.
f (t, x) = f (t, ut) = t0f (1, u) = f (1, u), (35)
where t is playing the role of parameter λ in definition 4.1. Clearly f (1, u) is a function of
u alone.

4.3 Method of Solution of Equations with homogeneous coeffi-

cients

Suppose the coefficients M and N in an equation

M(t, x) dt + N (t, x) dx = 0 (36)

are homogeneous and of the same degree, then (36) could be written in the form

dx x
= f ( ). (37)
dt t

This suggests the introduction of a new variable x(t)

t
= u(t) to facilitate the determination
of the solution of (36). With this transformation, (37) reduces to

du dt
= , (38)
f (u) − u t

that is readily solved by a method of separation of variables.

29
Summary:

Let us summarize the steps to follow:

1. Check if the vector field f (t, x) is homogeneous;
2. Write out the substitution
x(t)
= u(t);
t

3. Through easy differentiation, find the new equation satisfied by the new function u. You
may want to remember the form of the new equation as the one in Equation(38);
4. Solve the new equation, which is always separable, to find u;
5. Go back to the old function x through the substitution x = ut;
6. If you have an IVP, use the initial condition to find the particular solution.

Since you have to solve a separable equation, you must be particularly careful about the
constant solutions.

Example 15. Solve the initial value problem

x t2
x0 = − 2 , x(1) = 1.
t x

Solution 15. Follow the steps:

1. It is easy to check that

x t2
f (t, x) = − 2
t x
is homogeneous.

2. Because of homogeneity, make the substitution:

x(t)
u(t) = .
t

3. We thus have:
x t2 x
x0 = − 2 =: f ( ) (39)
t x t
f (u) − u
u0 = (40)
t
u − 1/u2 − u
= (41)
t
−1
= . (42)
u2 t
The last equation is separable. If you don’t get a separable equation at this point, then your
equation is not homogeneous, or something went wrong along the way!

30
4. Upon integration of the last equation, we get:
1 3
u = − ln t + C.
3

5. Back to the function x, we get

1 x
c − ( )3 = ln t
3 t

6. Using the initial condition, we get: c = 13 . Hence the solution to the IVP is

1 1 x 3
− ( ) = ln t
3 3 t
Now try to see if you can do the little exercise below.

Exercise 8. Exact

solve the following equations:

1.
 dx 
2 2 2
(a) (t −x ) dt+2tx dx = 0 (b) (t−x) dt+2tx dx = 0 (c) t = x(log x−log t+1)
dt
dy x y
(d) v 2 dx + x(x + v) dv = 0 (e) 3y dx + (7x − y) dy = 0 (f ) = + .
dx y x
−2x + 5y
(g) x0 = .
2x + y

2. Solve the following IVPs

(a) (t2 +x2) dt+2tx dx = 0, x(1) = −1 (b) (t−x) dt+(3t+x) dx = 0, x(2) = −1

(c) tx dt + 2(t2 + 2x2 ) dx = 0, x(0) = 1. (d) v 2 dx + x(x + v) dv = 0, x(1) = 1

4.4 Equations reducible to homogeneous form

The method of homogeneous equations may be extended to some non-homogeneous equa-

tions. We consider in this subsection cases which are reducible to homogeneous equations.

31
4.4.1 x0 = f (at + bx + c).

Here a, b, and c are constants. We consider the case where b 6= 0. We use the substitution
u(t) = at + bx(t) + c and obtain

u0 (t) = a + bx0(t) = a + bf(u).

This transformation enables us to achieve an equation of the form

du
= a + bf(u),
dt
that can be solved by a method of separation of variables.

Example 16. Solve x0 = (t + x)2 .

Solution 16. Let u = t + x, u0 = 1 + x0 = 1 + f (u)

Z Z
du
= dt (43)
1 + f (u)
Z Z
du
= dt (44)
1 + u2
tan−1 u = t + c. (45)

Therefore
u = tan(t + c)
⇔ x = tan(t + c) − t

at+bx+c
4.4.2 x0 = f ( αt+βx+γ ).

Here a, b, c, α, β, and γ are constants.

Case 1.
a b
= 0, ⇒ a = λα, b = λβ.
α β
A solution of this kind of equation has already been done.

Case 2.
a b
6= 0,
α β
then the linear system,

at + bx + c = 0 (46)
αt + βx + γ = 0 (47)

has exactly one solution, (ξ, η).

32
 Let t̄ := t − ξ, x̄ := x − η. With these transformation, we have

x̄(t̄) := y(t̄ + ξ) − η.

dx̄  a(t̄ + ξ) + b(x̄(t̄) + η) + c 

0
= x̄ (t̄ + ξ) = f (48)
dt̄ α(t̄ + ξ) + β(x̄(t̄) + η) + γ
 at̄ + bx̄(t̄) 
= f , (49)
αt̄ + β x̄(t̄)
which is of the form amenable to methods of separation of variables.

Example 17. Solve the differential equation

dx t + 2x − 4
= . (50)
dt 2t + x − 5

Solution 17. We notice that the functions t + 2x − 4 and 2t + x − 5 are linear in t and x,
and that
1 2
6= 0.
2 1
This suggests the use of the method described above in the determination of the solution to
(50). (ξ, η) = (2, 1). Substituting

t̄ := t − 2, x̄ := x − 1,

where (2, 1) is the unique solution of the system

t + 2x − 4 = 0, (51)
2t + x − 5 = 0, (52)

in (50) we get
dx̄  t̄ + 2x̄ 
= .
dt̄ 2t̄ + x̄
This is the same as:
dx̄ 1 + 2x̄/t̄ x̄
= =: f ( ). (53)
dt̄ 2 + x̄/t̄ t̄
Substituting
x̄(t̄)
u(t̄) =
t̄
in (53) we get:
du(t̄) f (u) − u
= .
dt̄ t̄
Integration by separation of variables yields:
Z Z
du dt̄
=
Z f (u) − u Z t̄
2+u dt̄
du = = ln t̄ + c
1 − u2 t̄

33
 Z  
3 1
+ du = ln t̄ + c (54)
2(1 − u) 2(1 + u)
1  1+u 
ln = ln t̄ + c (55)
2 (1 − u)3
1 + u = C(1 − u)3t̄2
x̄
Using u = t̄
in (55), we obtain
t̄ + x̄ (t̄ − x̄)3 2
=K t̄
t̄ t̄3
this is the same as
t̄ + x̄ = K(t̄ − x̄)3
t + x − 3 = K(t − x − 1)3

Exercise 9. Equations reducible to homogeneous form

Solve the differential equations below

dy x + 2y − 3
(1) = (2) (x + y) dx + (3x + 3y − 4) dy
dx 2x + y − 3
dy 4(3x + y − 2)
(3) (x + 2y − 1) dx + 3(x + 2y)dy = 0 (4) =
dx 3x + y
(5) (x+2y−1) dx+(2x+4y−3) dy = 0 (6) (x+2y+1) dx−(2x+4y+3) dy = 0
dy x+y+4
(7) = (8) (x + 2y − 4) dx − (2x + y − 5) dy
dx x−y−6
dy y
(9) (2y − 3x − 7) dx + (2x + 3y + 9) dy = 0 (10) =
dx x−y+1
(11) (x−y+1) dy−(2x+y−4) dx = 0, (12) (2x−y+1) dx+(2y−x−1)dy = 0

34
5 Lecture 5: Exact Differential Equations

Objectives

At the end of the lecture we should be able to:

• Determine if a differential equation is exact

• Solve exact differential equations

We consider in this section a method for solving first order differential equations for which
the method of separation of variables may not be applied directly.

5.1 What is an exact differential equation?

We begin with the following definition

5.1 Definition. The total differential of a function F (t, x) is given by

∂F ∂F
dF = dt + dx = 0,
∂t ∂x
∂F ∂F
provided that the partial derivatives ∂t
, and ∂x
exist.

This definition leads to the following definition:

5.2 Definition. A differential equation of the form

M(t, x) dt + N (t, x) dx = 0, (56)

is called exact if there is a function F (t, x) whose total differential is equal to the left-hand
side of Equation(56); that is,

dF = M(t, x) dt + N (t, x) dx.

In this case F (t, x) is called an integral of the exact differential equation .

Suppose we can find a function F (t, x) such that

dF = M dt + N dx = 0. (57)

35
This implies that F (t, x) = c, c =constant and it implicitly defines a set of solutions of (56).

Two questions arise:

(1) First, under what conditions on M and N does a function F exist such that dF =
M dt + N dx = 0?
(2) Second, if these conditions are satisfied, how do we determine this F ?

5.1.1 When is an equation exact?

By definition 5.2, a differential equation is exact if, there exists F such that
dF = M dt + N dx.
From calculus
∂F ∂F
dF = dt + dx;
∂t ∂x
∂F ∂F
⇒ = M and = N.
∂t ∂x
These two, again from calculus, lead to
∂M ∂ 2F ∂ 2F ∂N
= = = ,
∂x ∂x∂t ∂t∂x ∂t
provided these partial derivatives are continuous. Therefore for (56) to be exact, we see that
∂M ∂N
= (58)
∂x ∂t
We now show that if Equation(58) is satisfied, then (56) is an exact equation.

Let φ(t, x) be a function for which ∂φ∂t

= M. The function φ is the result of integrating
M dt w.r.t. t while holding y constant. Now
∂ 2φ ∂M
= ;
∂t∂x ∂x
hence, if (58) holds, then also
∂ 2φ ∂N
= . (59)
∂t∂x ∂t
Integrating both sides of (59) w.r.t t, while holding x fixed, we obtain
∂φ
= N + B 0(x) (60)
∂x
where B 0(x) is an arbitrary function of x. Now a function F can be exhibited, namely,
Z
F = φ(t, x) − B(x), B(x) = B 0(x)dx.

To this end we have that

∂φ ∂φ
dF = dt + dx − B 0 (x)dx (61)
∂t ∂x
= Mdt + [N + B 0(x)]dx − B 0(x)dx (62)
= Mdt + Ndx. (63)
Hence (56) is exact.

36
Summary:

If (56) is exact then (58) holds. If (58) holds then (56) is an exact differential
equation

Theorem 5.1. If, M(t, x), N(t, x) ∈ C 1 (D), then a necessary and sufficient condition that

M(t, x) dt + N (t, x) dx = 0

be an exact equation is that

∂M ∂N
= .
∂x ∂t

5.1.2 How do we solve an exact differential equation?

The proof of Theorem 5.1 contains the germ of a method for obtaining a set of solutions.
This are the steps to follow:

(1) Check that the equation is indeed exact;

(2) Write down the system  ∂φ

∂t
= M(t, x),
∂φ
∂x
= N (t, x).

(3) Integrate either the first equation with respect to the variable t or the second with
respect to the variable x. The choice of the equation to be integrated will depend on
how easy the calculations are. Let us assume that the first equation was chosen, then
we get Z
φ(t, x) = M(t, x) dt + h(x).

The function h(x) should be there, since in our integration, we assumed that the
variable x is constant.

(4) Use the second equation of the system to find the derivative of h(x). Indeed, we have
Z
∂φ ∂  
= M(t, x) dt + h0 (x) = N (t, x),
∂x ∂x
which implies Z
0 ∂  
h (x) = N (t, x) − M(t, x) dt .
∂x
Note that h is a function of x only. Therefore, in the expression giving h0 (x) the
variable, t, should disappear. Otherwise something went wrong!

(5) Integrate to find h(x);

(6) Write down the function φ(t, x);

37
 (7) All the solutions are given by the implicit equation

φ(t, x) = C.

(8) If you are given an IVP, plug in the initial condition to find the constant C.

You may ask, what do we do if the equation is not exact? In this case, one can try to find
an integrating factor which makes the given differential equation exact. That is discussed in
the next section. But before we proceed, let us fix an example.
Example 18. Solve the differential equation

(cos(t + x2) + 3x)dt + (2x cos(t + x2) + 3t)dx = 0.

Solution 18. We solve this problem by religiously following the steps outlined above.

1. Clearly
∂ ∂
(cos(t + x2 ) + 3x) = −2x(sin(t + x2) + 3) = (2x cos(t + x2) + 3t) .
∂x | {z } ∂t | {z }
M N

Hence the given equation is exact.

2. Let φ(t, x) = c be the solution to this equation. Then,

 ∂φ
∂t
= cos(t + x2) + 3x
∂φ
∂x
= (2x cos(t + x2 ) + 3t)

3. Integrating the first equation, we get:

φ(t, x) = sin(t + x2) + 3tx + h(x), (64)

where h(x) is an arbitrary function of x or a constant.

4. We use the second equation:

If (60) is a solution, then
∂φ dh
= 2x cos(t + x2) + 3t + = 2x cos(t + x2) + 3t
∂x dx
dh
⇒ = 0. (65)
dx

5. We integrate Equation(65) to find that h(x) = a constant.

6. We write down φ(t, x); that is,

φ(t, x) = sin(t + x2) + 3tx = a constant

is the solution of the differential equation .

38
Exercise 10. Exact differential Equations

In each of the following, determine whether the equation is exact. If so, solve it.

(1) 3x(xy−2) dx+(x3+2y) dy = 0. (2) (2t3 −tx2−2x+3) dt−(t2x+2t) dx = 0

1 sin t 9
(3) [(cos t ln(2x−8)+ ] dt+ dx = 0, x(1) = (4) (t+2x) dt+(2t+x) dx = 0.
x x−4 2
(5) v(2uv 2−3) du+(3u2 v 2−3u+4v) dv = 0 (6) (r+sin θ−cos θ) dr+r(sin θ+cos θ) dθ = 0
(7) (yexy − 2y 3 ) dx + (xexy − 6xy 2 − 2y) dy, y(0) = 2.

39
6 Lecture 6:Integrating Factors, Substitution Suggested
by the Equation

Objectives

At the end of the lecture we should be able to:

• Determine an integrating factor for some differential equation;

• Make some equations exact by use of the integrating factor;
• Use a substitution suggested by the equation in order to solve it;
• Change of Variables.

6.1 Integrating Factors

We have studied first-order differential equations that are exact. However some differential
equations may not be exact and thus require other techniques to solve them. If an equation
is not exact, it is natural to attempt to make it exact by the introduction of an appropriate
factor, which is then, called an integrating factor.

In general, very little can be said about the theory of integrating factors (IF) for first-
order DEs. Nonetheless, there is one important class of equations where the existence of an
IF can be demonstrated. This is the class of linear first-order differential equations of the
form,
M dt + N dx = 0. (66)
Suppose that u := u(t, x) is an IF of (66). Then, the equation
uM dt + uN dx = 0, (67)
must be exact; that is,
∂(uM) ∂(uN )
= .
∂x ∂t
Hence
∂M ∂u ∂N ∂u
u +M =u +N ,
∂x ∂x ∂t ∂t
or
∂M ∂N ∂u ∂u
u( − )=N −M . (68)
∂x ∂t ∂t ∂x
We see that if u satisfies (68) above, then it is an IF to (66). However, this is not an ordinary
differential equation since it involves more than one variable. This is what is called a partial

40
 differential equation. These types of equations are very difficult to solve, which explains why
the determination of the integrating factor is extremely difficult except for two special cases:

Case 1: There exist an IF u(t) which is a function of t alone if,

1  ∂M ∂N 
−
N ∂x ∂t
were a function of t alone. Indeed, assume that u = u(t), then (68) becomes
 ∂M ∂N  du
u − =N ,
∂x ∂t dt
1  ∂M ∂N  du
⇔ − dt = N (69)
N ∂x ∂t u
If the L.H.S of (69) were a function of t alone, then we can determine u(t), by an
appropriate method. Indeed, if
1  ∂M ∂N 
− = g(t), (70)
N ∂x ∂t
then the IF of (66) is Z
u(t) = exp( g(t) dt).

Case 2: Similarly, there exist an IF u(x) which is a function of x alone if,

1  ∂N ∂M 
−
M ∂t ∂x
were a function of x alone. Assume u = u(x) then
 ∂M ∂N  du
u − = −M ,
∂x ∂t dx
1  ∂N ∂M  du
⇔ − dt = (71)
M ∂t ∂x u
should the LHS of (71) be a function of x alone; that is,

1  ∂N ∂M 
− =: h(x),
M ∂t ∂x
then the IF Z
u = exp( h(x) dx).

Once the IF is found, multiply the old equation by u to get a new one which is exact.
Then you are left to use the previous technique to solve the new equation. Advice: if you
are not pressured by time, check that the new equation is in fact exact!

41
 Summary of IF technique

Consider the equation

M(t, x) dt + N (t, x) dx = 0.
If the equation is not given in this form, then we have to first, transform it to look like the
above.

Step 1: Check for exactness; that is, compute

∂M ∂N
and ,
∂x ∂t
then compare them.

Step 2: Assume that the equation is not exact (if it is exact use the previous technique to find
the solution). Then evaluate
1  ∂M ∂N 
− .
N ∂x ∂t
If this expression is a function of t only, then go to step 3.1. Otherwise, evaluate
1  ∂N ∂M 
− .
M ∂t ∂x
If this expression is a function of x alone, then go to step 3.2. Otherwise, we cannot
solve the equation using the technique developed above!

Step 3: Find the IF.

Step 3.1: If the expression

1  ∂M ∂N 
−
N ∂x ∂t
is a function of t only. Then the IF is given by
Z 1  ∂M ∂N  
u(t) = exp − dt ;
N ∂x ∂t

Step 3.2: If the expression

1  ∂N ∂M 
−
M ∂t ∂x
is a function of x only, then an integrating factor is given by
Z 1  ∂N ∂M  
u(x) = exp − dx .
M ∂t ∂x

Step 4: Multiply the old equation by u, and, if you can, check that you have a new equation
which is exact.

Step 5: Solve the new equation using the steps described in the previous section.

The following example illustrates the use of the integrating factor technique:

42
Example 19. Find all the solutions to
dx 3tx + x2
=− 2 .
dt t + tx

Solution 19 Note that this equation is in fact homogeneous. But let us use the technique
of exact and nonexact to solve it. Let us follow these steps:

(1) We rewrite the equation to get

(3tx + x2 ) dt + (t2 + tx) dx = 0.
Hence, M(t, x) = (3tx + x2 ) and N (t, x) = (t2 + tx).
(2) We have
∂M ∂N
= 3t + 2x and = 2t + x,
∂x ∂t
which clearly implies that the equation is not exact.
(3) Let us find an integrating factor. We have
1  ∂M ∂N  1
− = .
N ∂x ∂t t
Therefore, an integrating factor u(t) exists and is given by
1 
u(t) = exp dt = eln(t) = t.
t
(4) The new equation is
(3t2 x + tx2) dt + (t3 + t2x) dx = 0,
which is exact. (Check it!)
(5) Let us find φ(t, x). Consider the system:
 ∂φ
∂t
= (3t2 x + tx2 )
∂φ
∂x
= (t3 + t2 x)

(6) Let us integrate the first equation. We get

t2 2
φ(t, x) = t3x + x + h(x).
2
(7) Differentiate with respect to x and use the second equation of the system to get
∂φ
= t3 + t2 x + h0(x) = t3 + t2x,
∂x
which implies h0(x) = 0; that is, h(x) = C is constant. Therefore, the function φ(t, x)
is given by
t2
φ(t, x) = t3 x + x2.
2
We don’t have to keep the constant C due to the nature of the solutions (see next
step).

43
 (8) All the solutions are given by the implicit equation
t2 2
t3 x + x = C.
2

Remark: Note that if you consider the function

1
u(t, x) = ,
tx(2t + x)
then we get another integrating factor for the same equation. That is, the new equation
1 1
(3tx + x2) dt + (t2 + tx) dx = 0
tx(2t + x) tx(2t + x)
is exact. So, from this example, we see that we may not have uniqueness of the integrating
factor. Also, we learn that if the integrating factor is given, the only thing we have to do is
to multiply it with the given equation and check that the new one is exact.

Another example which assumes some steps is as follows:

Example 20. Solve the equation

x(t + x + 1) dt + t(t + 3x + 2) dx = 0 (72)

Solution 20.
∂M
M = x(t + x + 1), = t + 2x + 1
∂x
∂N
N = t(t + 3x + 2), = 2t + 3x + 2
∂t
∂N ∂M
− =t+x+1
∂t ∂x
1 ∂N ∂M 1
( − )=
M ∂t ∂x x
We see that the integrating factor is
Z
1
exp ( dt) = x
x
Multiplying Equation(72) by x we obtain

(tx2 + x3 + t2) dt + (t2 x + 3tx2 + 2tx) dx = 0,

an exact equation whose solution is tx2(t + 2x + 2) = c.

Note that if neither of the preceding criteria is satisfied, we can say only that the equation
does not have an integrating factor; that is, a function of t and x alone.

On the other hand, if Equation(66) is homogeneous and Mt +Nx 6= 0, then the integrating
factor takes the form
1
µ(t, x) = .
Mt + Nx

44
Sometimes, the equation (66) can be written in the form

xf (tx) dt + tg(tx) dx = 0,

where f (tx) 6= g(tx); then, in this case, the required integrating factor takes the form
1 1
µ(t, x) = = , (73)
tx{f (tx) − g(tx)} Mt − Nx

where M = xf (tx) and N = tg(tx).

Example 21. Solve the equation

(t4 + x4 ) dt − tx3 dx = 0 (74)

Solution 21.First we note that Equation(74) is a homogeneous equation of degree four and
is also not an exact equation. However

Mt + N x = t5 + tx4 − tx4 = t5 6= 0;

and so the required integrating factor is

1 1
µ= = 5.
Mt + Nx t
Multiplying through Equation(74) by µ we obtain
1 x4  x3
+ dt − dx = 0
t t5 t4
which is exact. Upon integration, we get
x4
ln |t| − = constant.
4t4

Exercise 11. Integrating factor

Solve the following equations:

(1) (x2+y 2+x) dx+xy dy = 0 (2) (xy+y 2+y)dx+(x2+3xy+2x)dy = 0

(3) (x4+y 4)dx−xy 3 dy = 0 (4) (4xy+3y 2−x) dx+x(x+2y) dy = 0

(5) (x2y 2 +2y) dx+(2x−2x3y 2) dy = 0, (6) y dx−x dy = 0
2 2
(7) (x +y +1) dx+x(x−2y) dy = 0, (8) (xy+1) dx+x(x+4y−2) dy = 0
(9) (2y 2+3xy−2y+6x) dx+x(x+2y−1) dy = 0, (10) xy dx−(x2+2y 2 ) dy = 0

45
6.2 Substitution Suggested by the Equation

Introduction

Sometimes, first order equations by the nature of f (t, x) or M(t, x) N (t, x) suggest cer-
tain substitutions which make them amenable to certain solution techniques. There are no
benchmark techniques, you just have to be hawk eyed. We discuss a few of the familiar
problems.

6.2.1 Change of Variables

Motivating example- u-substitution .

The idea behind changing of variables is not new. We have seen it before in calculus when
we did the “u-substitution”method for computing antiderivatives. For instance given
Z
t sin t2 dt,

we define a new variable u by u = t2 and rewrite the integral in terms of u instead of t. Since
du = 2t dt, we obtain upon substitution
Z Z
2 1
t sin t dt = sin udu.
2
This is the same integral as before but written using the new variable u. The variable u was
chosen so that the integral is easy to evaluate. We have
Z
1 cos u
sin u du = − + C.
2 2
We can now do the back- substitution to obtain
cos u cos t2
− +C =− +C
2 2
We see that changing variables has aided in making the job of integration a little lighter.

Given a differential equation, it is likely that it will not be in a form appropriate for the
particular technique we would like to use. It would be nice if we could transform it into a
new equation that has a proper structure. This is the idea behind changing variables. By
rewriting the equation in terms of a new variable, we may be able to put the equation into
a form that is appropriate for a particular technique. In this section we give a method for
transforming differential equations by changing variables into (hopefully) simpler forms.

An example . Consider the differential equation

dx x t−1
=− + .
dt t 2x

46
This equation looks pretty complicated. It is neither separable nor linear. It is nonlinear.
We attempt to simplify by x multiplying through the equation. This yields:

dx x2 t − 1
x =− + .
dt t 2
However, the situation is still unclear. Try u = x2. This substitution leads to
du dx 1 du dx
= 2x , so =x .
dt dt 2 dt dt
Substituting this into the L.H.S and replacing the x2 with u on the R.H.S, we obtain the
new equation
1 du u t−1
=− + .
2 dt t 2
Multiplying through by 2 gives
du 2
= − u + (t − 1).
dt t
This is a linear equation whose solution is

t2 t c
u(t) = − + 2.
4 3 t
We can easily then compute r
t2 t c
x(t) = − + 2.
4 3 t
Exercise 12. (a) Solve
a2 (t dx − x dt)
t dt + x dx = .
t 2 + x2
(b) Solve
 dx 
2
sec x + 2t tan x = t3.
dt

Exercise 13. Substitution suggested by equation

Solve the following equations

1. (x + 2y − 1) dx + (3x + 6y) dy = 0

2. (1 + 3x sin y) dx − x2 cos y dy = 0

3. (3x − 2y + 1) dx + (3x − 2y + 3) dy = 0

47
4. sin y(x + sin y) dx + 2x2 cos y dy = 0
dy
5. dx
= (9x + 4y + 1)2
dy
6. dx
= sin(x + y)

7. (3 tan x − 2 cos y) sec2 dx + tan x sin y dy = 0

8. (x + 2y − 1) dx + (2x + 4y − 3) = 0

48
7 Lecture 7: Bernoulli’s Equations,
Ricatti Equations,
The general Linear Equation of first order

Objectives

At the end of the lecture we should be able to:

• Solve a Bernoulli’s Equation

• Solve Ricatti Equations;
• Get the general solution to a general Linear Equation of first order differential equation.

7.1 Bernoulli’s Equations

A Bernoulli equation is of the form

x0 + g(t)x + h(t)xn = 0, (75)
where n 6= 1 is a real number. The above equation named after Jacob Bernoulli (1654-1705)
can be transformed to a linear differential equation which can be solved by methods already
discussed. Equation(75) can be rewritten as
x−n x0 + g(t)x−n+1 + h(t) = 0, n 6= 1 (76)
Let
x1−n = z,
then
dx dz
(1 − n)x−n = .
dt dt
this implies
dx 1 dz
x−n = .
dt 1 − n dt
So (76) in the variables z and t is
dz
+ (1 − n)g(t) + (1 − n)h(t) = 0.
dt
Thus (76) in the variables z and t is
z 0 + (1 − n)g(t)z + (1 − n)h(t) = 0, (77)

49
a linear equation in the standard form that may be solved by an appropriate technique.
Once it is solved, we obtain the function
1
x = z 1−n .

Note that if n > 1, then we have to add the solution x = 0 to the solution found via the
technique described above.

Let us summarize the steps to follow in the solution of a Bernoulli Equation.

Summary:

1. Recognize that the differential equation is a Bernoulli equation. Then find

the parameter n from the equation;

2. Write out the substitution z = x1−n ;

3. Through easy differentiation, find the new equation satisfied by the new
variable z. You may want to remember the form of the new equation:

z 0 + (1 − n)g(t)z + (1 − n)h(t) = 0

4. Solve the new linear equation to find z;

5. Go back to the old function x through the substitution

1
x = z 1−n ;

6. If n > 1, add the solution x = 0 to the ones you obtained in (4).

7. If you have an IVP, use the initial condition to find the particular solution.

Example 22. Solve the equation

x
x0 + + (1 + t)x4 = 0 (78)
1+t

Solution 22.
(1) We have a Bernoulli equation with n = 4;
(2) Consider the new function
z := x1−4 = x−3 ;
(3) Then we have
z 0 = −3x−4 x0.
This transformation used in (76) yields
3z
z0 − − 3(1 + t) = 0.
1+t

50
(4) The solution of this system is

z(t; c) = c(1 + t)3 − 3(1 + t)2.

(5) Back to the function x we have:

sgn(ct + c − 3)
x(t; c) = p
3
(1 + t)2|ct + c − 3|
(6) All solutions are of the form
 sgn(ct+c−3)
x(t; c) = √
3 2
(1+t) |ct+c−3|
x=0

Exercise 14. Bernoulli’s Equations

Solve the differential equations

dy
1. dx
− y = xy 5,

2. 6x2 dt − t(2t3 + x) dx = 0,

dx
3. dt
− 12 (1 + 1t )x = − 3t x3.

dx 1 1
4. dt
+ t−1
x = tx 3 .

dy
5. 2 dx − y sec x = y 3 tan x.

6. 2xy dy
dx
= y 2 − 2x3 , y(1) = 2.
dy
7. dx
= y + y 3.

7.2 Riccati Equations

Before we give the formal definition of Riccati equations, a little introduction may be
helpful. Indeed, consider the first order differential equation

y 0 = f (x, y).

51
If we approximate f (x, y), while x is kept constant, we will get

f (x, y) = P (x) + Q(x)y + R(x)y 2 + · · ·

If we stop at y, we will get a linear equation. Riccati looked at the approximation to the
second degree: He considered equations of the type

y 0 = P (x) + Q(x)y + R(x)y 2. (79)

These equations bear his name, Riccati equations. They are nonlinear and do not fall
under the category of any of the classical equations. In order to solve a Riccati equation,
one will need a particular solution. Without knowing at least one solution, there is absolutely
no chance to find any solutions to such an equation. Indeed, let y1 be a particular solution
of Equation(79). Consider the new function z defined by
1
z= .
y − y1
Then easy calculations give
 
z 0 = − Q(x) + 2y1 R(x) z − R(x)

which is a linear equation satisfied by the new function z. Once it is solved, we go back to
y via the relation
1
y = y1 + .
z
Keep in mind that it may be harder to remember the above equation satisfied by z. Instead,
try to do the calculations whenever you can.
Example 23. Solve the equation

y 0 = −2 − y + y 2

knowing that y1 = 2 is a particular solution.

Solution 23. We recognize a Riccati equation. First of all we need to make sure that y1
is indeed a solution. Otherwise, our calculations will be fruitless. In this particular case, it
is quite easy to check that y1 = 2 is a solution. Set
1
z= .
y−2
Then we have
1
y =2+
z
which implies
z0
y0 = − .
z2
Hence, from the equation satisfied by y, we get
z0  1  1 2
− 2 = −2 − 2 + + 2+ .
z z z

52
Easy algebraic manipulations give
z0 3 1
= + .
z2 z z2
Hence
z 0 = −3z − 1.
This is a linear equation. The general solution is given by
− 13 e3x + C 1
z= 3x
= − + Ce−3x.
e 3
Therefore, we have
1
y =2+ .
− 13 + Ce−3x

Note: If one remembers the equation satisfied by z, then the solutions may be found a
bit faster. Indeed in this example, we have P (x) = −2, Q(x) = −1, and R(x) = 1. Hence
the linear equation satisfied by the new function z, is
 
0
z = − Q(x) + 2y1 R(x) z − R(x) = −(−1 + 4)z − 1 = −3z − 1.

Exercise 15. (i) Check that y1 = sin x is a solution to

2 cos2 x − sin2 x + y 2
y0 = , y(0) = −1.
2 cos x
Then solve the IVP
dy 2 cos2 x − sin2 x + y 2
= , y(0) = −1
dx 2 cos x

Solution Exercise 15
1
y = sin x + −1 .
2
sin x − cos x
Exercise 16. Check that y1 = x is a solution to
y
y0 = + x3 y 2 − x5 ,
x
then find the general solution.

7.3 The general Linear Equation of first order

The first-order differential equations we have looked at so far can nearly be written in the
form
y 0 = f (x, y), (80)
where f (x, y) is a fairly simple function. If we make the requirement that f (x, y) = −a(x)y +
b(x), then we get what is called a linear differential equation in normal form. Let us
now write a general first order linear differential equation

y 0 + a(x)y = b(x). (81)

53
When b(x) = 0, the equation is referred to as a first oder homogeneous . If b(x) 6= 0, it is
referred to as a first order linear nonhomogeneous differential equation .

We wish to find a solution to Equation(81).

Method of solution . The methods of solution of the homogeneous differential equa-

tion have already been discussed. Indeed the solution to

y 0 + a(x)y = 0

is given by the method of separation of variables as

R
φ(x) := Ce− a(x)dx
6= 0,

where C is a constant. To solve (81), y 0 = b(x) − a(x)y, we assume a solution of the form

ψ(x) = c(x)φ(x),

where c(x) is a function to be determined and φ(x), satisfies the homogeneous differential
equation ; that is,
φ0 + a(x)φ = 0.
Then
ψ 0(x) = c0 φ + cφ0 = b(x) − a(x)c(x)φ(x) (82)
But φ0 = −a(x)φ(x). This used in (82) gives

c0 (x)φ(x) − c(x)(a(x)φ(x)) = b(x) − a(x)c(x)φ(x)

c0 (x)φ(x) = Z
b(x)
b(x)
c = dx
φ(x)
Z
b(s)
⇒ ψ(x) = φ(x) ds. (83)
φ(s)

Since (81) is linear, Z

b(s)
Φ(x) = cφ(x) + φ(x) ds
φ(s)
is also a solution, where φ(x) is a solution of L(y) = 0.

Exercise 17. The general Linear equation of first order

1. Find all solutions of the following Equations

(a) y 0 + 2xy = x, (b)xy 0 + y = 3x3 − 1, for x > 0

54
 π
(c) y 0 + exy = 3ex, (d) y 0 − y tan x = esin x , for 0 < x <
2

2. Consider the equation y 0 + y cos x = e− sin x .

(a) Find the solution φ, which satisfies φ(π) = π.
(b) Show that any solution φ has the property that

φ(kπ) − φ(0) = kπ,

where k ∈ Z.

3. Find the solution to

π
(a) 2tẏ−y = t+1, y(2) = 4. (b) cos2 t sin ty 0 = −y cos3 t+1, y( ) = 0.
4

55
8 Lecture 8: First Order First Degree differential Equa-
tions in three Variables

8.1 Introduction

We have thus far handled differential equations that involve only two variables. We now
consider ordinary differential equations of the first order and first degree involving three
variables.

Objectives

At the end of this lecture we are expected to be able to solve Total Differential equations in
three variables by one of the methods below:

• The Method of grouping

• The Method of multipliers

8.2 Total Differential Equations

An ordinary differential equation of the first order and first degree involving three variables
is of the form
dy dz
P +Q +R = 0, (84)
dx dx
where P, Q, R are functions of x, y, and z, and x is the independent variable.

In terms of differentials, Equation(84) has the form

P dx + Q dy + R dz = 0. (85)

Equation(85) is integrable only when

 ∂Q ∂R   ∂R ∂P   ∂P ∂Q 
P − +Q − +R − =0 (86)
∂z ∂y ∂x ∂z ∂y ∂x

Observe the relation between Equation(86) with

P Q R
∂P ∂Q ∂R = 0,
∂x ∂y ∂z

56
 ∂P
where ∂P ∂y here denotes ∂y
,e.t.c. Alternatively if

X = P~i + Q~j + R~k,

then Equation(85) is integrable iff

X.curlX = 0.
We shall not provide a prove for this, you may consult, for instance [5].

To get a solution of Equation(85) we have the following rule.

Rule

Once the condition in Equation(86) is satisfied, take one of the variables, say z,
as constant so that dz = 0. Then integrate the equation P dx + Q dy = 0. Replace
the arbitrary constant appearing in its integral by φ(z). Now differentiate the
integral just obtained with respect to x, y, z. Finally compare this result with
the given differential equation to determine φ(z).

Let us look at an example to shade more light on this procedure.

Example 24. Solve

(y 2 + yz) dx + (z 2 + zx) dy + (y 2 − xy) dz = 0. (87)

Solution 24. We see that P = y 2 + yz, Q = z 2 + zx, R = y 2 − xy. It is easily seen

that Equation (86) is satisfied. We then seek a solution to Equation(87) by the above rule.
Treating z as a constant, Equation(87) becomes
dx dy
(y 2 + yz) dx + (z 2 + zx) dy = 0, or + = 0.
z(z + x) y(y + z)
Integrating, we obtain:

ln(z + x) + ln y − ln(y + z) = (constant),

or
y(z + x)
= φ(z) = constant (88)
y+z
or
y(z + x) − φ(z)(y + z) = 0.
Differentiating w.r.t. x, y, z, we obtain:

y dx + [z + x − φ(z)] dy + [y − (y + z)φ0 (z) − φ(z)] dz = 0, (89)

where 0 denotes differentiation with respect to z. Comparing Equations(89) and (87), we get

y 2 + yz z 2 + zx y 2 − xy
= = .
y z + x − φ(z) y − (y + z)φ0 (z) − φ(z)

57
The relation
y 2 + yz z 2 + zx
=
y z + x − φ(z)
reduces to Equation(88) and thus gives no new information as far as the determination of
φ(z) goes. Therefore we try
y 2 + yz y 2 − xy
= ,
y y − (y + z)φ0(z) − φ(z)
that upon simplification with the help of Equation(88) yields
y 2 − xy = y 2 − xy − (y + z)2 φ0(z) or (y + z)2 φ0(z) = 0,
which gives φ0(z) = 0 and thus φ(z) = C1 , a constant. Hence the required solution to
Equation(87) is
y(z + x) = (y + z)C1.
Remark. Sometimes the integral is obtained simply by regrouping the terms in the given
equation as illustrated in the next example.
Example 25.
Solve x dx + z dy + (y + 2z) dz = 0.

Solution 25. After regrouping the terms, the given equation can be written thus:
x dx + (y dz + z dy) + 2z dz = 0.
Integrating this equation, we obtain
x2
+ yz + z 2 = C1 .
2

8.3 Simultaneous Total differential equations

The equations
P1 dx + Q1 dy + R1 dz = 0, P2 dx + Q2 dy + R2 dz = 0, (90)
where P1 , Q1 , R1 and P2 , Q2, R2 are any functions of x, y,and z, are called simultaneous
total differential equations.
(a) If each of the above equations is integrable and has solutions φ(x, y, z) = C1 and
ψ(x, y, z) = C2 , respectively, then these equations taken together form the solution to
Equation(90)
(b) If one or both of Equation(90) is not integrable, then solving Equations(90), we get
dx dy dz
= =
Q1R2 − R1 Q2 R1P2 − P1 R2 P1 Q2 − Q1 P2
which is of the form
dx dy dz
= =
P Q R
and solve these by the methods given below.

58
8.3.1 Method of Grouping

Note that if it is possible to take two fractions dx

P
= dz
R
, from which y can be canceled or
absent, leaving the equation in x and z only. Then integrate it giving
φ(x, z) = C1. (91)
Again, note that if one variable, say x, is absent or can be canceled( may be with the help
of Equation(91)) from equation dyQ
= dz
R
, then integrate it to get
ψ(y, z) = C2 . (92)
The two independent solutions Equations(90) and (92) taken together form the complete
solution of the given equation.
Example 26.
dx dy dz
Solve 2
= 2 = 2 .
z y z x y x

Solution 26.Taking
dx dy
2
= 2 .
z y z x
2
Dividing through by z and rearranging, we get
x dx − y dy = 0,
which on integration gives
x2 − y 2 = C1 . (93)
Now take
dy dz
2
= 2 .
z x y x
Dividing through by x and rearranging, we get
y 2 dy − z 2 dz = 0
which upon integration gives
y 3 − z 3 = C2. (94)
Equations(93) and (94) taken together constitute a solution to the given equation.

8.4 Method of Multipliers

Before we get down to business, let us remind ourselves of a basic property of ratios. Take
any two equal ratios like ab = 2a
2b
. We always have
a 2a αa + 2aβ a
= = = ,
b 2b αb + 2bβ b
where α and β are nonzero. Try anything crazy like
1 2 4×1+2×5 14 1
= = = = .
3 6 4×3+6×5 42 3

59
Method 1

Let us now apply this little property of ratios. By a proper choice of multipliers l, m, n which
are not necessarily constants, we write
dx dy dz l dx + m dy + n dz
= = = (95)
P Q R lP + mQ + nR
such that lP + mQ + nR = 0. Then l dx + m dy + n dz = 0 can be solved giving the solution
φ(x, y, z) = C1 . (96)
Again look for another set of multipliers λ, µ, ν such that λP + µQ + νR = 0, giving λ dx +
µ dy + ν dz = 0, which on integration gives the solution as
ψ(x, y, z) = C2. (97)
Equations(96) and (97) taken together constitute a solution to the question under consider-
ation. They are also referred to as integral curves of the set of differential equations
dx dy dz
= = .
P Q R
They form a two-parameter family of curves in three-dimensional space.
Example 27.
dx dy dz
Solve = = .
x(y 2 2
−z ) 2 2
−y(z + x ) z(x + y 2)
2

Solution 27. Using the multipliers x, y, z in each fraction and using Equation(95) we get:
x dx + y dy + z dz x dx + y dy + z dz
= .
x2 (y 2 − z 2) 2 2 2 2 2 2
− y (z + x ) + z (x + y ) 0
Thus
x dx + y dy + z dz = 0,
which on integration gives
x2 + y 2 + z 2 = C1.
Again, using the multipliers x1 , − 1y , − 1z in each fraction and using Equation(95) we get:
1 1
x
dx − y
dy − 1z dz 1
x
dx − 1
y
dy − 1
z
dz
= .
(y 2 − z2) + (z 2 + x2) − (x2 + y 2 ) 0
Thus
1 1 1
dx − dy + dz = 0,
x y z
which on integration gives
ln x − ln y − ln z = Constant
or
yz = C2 x.
Hence, the solution of the given equation is
x2 + y 2 + z 2 = C1 , yz = C2x.

60
Example 28. Find the integral curves of the equations
dx dy dz
= = . (98)
y(x + y) + az x(x + y) − az z(x + y)

Solution 28. Clearly

P = y(x + y) + az, Q = x(x + y) − az, R = z(x + y). (99)

Let the multipliers be l = 1, m = 1 now for

lP + mQ + nR = 0,

we need  
y(x + y) + az + x(x + y) − az + n z(x + y) = 0. (100)

Hence on solving (100) we get

x+y
n=− .
z
Using it in l dx + m dy + n dz = 0 we get
x+y
dx + dy − dz = 0. (101)
z
This yields
dx + dy dz
=
x+y z
which on integration gives
x+y
= C1.
z
Again, using the multipliers λ = x, µ = −y in each fraction and using Equation(95) we get:
     
x y(x + y) + az − y x(x + y) − az + ν z(x + y) = 0. (102)
 
az(x + y) + ν z(x + y) = 0.
This implies that az + νz = 0 and thus ν = −a. Hence upon the use of these multipliers,
we get:
2x dx − 2y dy − 2ax dz = 0
Which on integration gives
x2 − y 2 − 2az = C2.
Therefore the required integral curves are the intersection of the surfaces:
 x+y
 z = C1

x2 − y 2 − 2az = C2

61
Method 2

We have derived the solutions in this manner, Method 1, to illustrate the general argument.
Written down in this manner, the derivation of the solution of these equations seems to
require a good deal of intuition in determining the forms of the multipliers l, m, n, λ, µ, ν.
In any actual example, it is simpler to try to cast the differential equations into a form
which suggests their solution. We usually use that little property of ratios already discussed.

Consider Equation(98) and look at each term like a “fraction ”.

dx dy dz
= = .
y(x + y) + az x(x + y) − az z(x + y)
| {z } | {z } | {z }
1 2 3

A new ratio can be found by adding the tops of 1 and 2 and then dividing by the sum of
their respective bottoms to yield:
dx dy dz dx + dy
= = = . (103)
y(x + y) + az x(x + y) − az z(x + y) (x + y)2
| {z } | {z } | {z } | {z }
1 2 3 4

We could also take 1 times x and then subtract 2 times y to give:

dx dy dz x dx − y dy x dx − y dy
= = = = .
y(x + y) + az x(x + y) − az z(x + y) xy(x + y) + azx − xy(x + y) + azy az(x + y)
| {z } | {z } | {z } | {z } | {z }
1 2 3 4 5
(104)
Using 3 and 4 in Equation(103) we get

dz dx + dy
=
z(x + y) (x + y)2
| {z } | {z }
3 4

Which on solving quickly gives

x + y = zC1 .
Also the from Equation(104) we have upon using 3 and 5

dz x dx − y dy
= ;
z(x + y) az(x + y)
| {z } | {z }
3 5

that is,
dz x dx − y dy
= .
1 a
Easy computation yields
x dx − y dy − az dz = 0.
This yields
x2 − y 2 − 2az = C2.

62
Method 3

In some instances it is comparatively easy to derive one of the solution sets by Method
2 but difficult to obtain a second solution by the same method. If this happens, then
use the determined solution to get the second one. Suppose, for instance, we have solved
Equation(98) and found
x+y
= z.
C1
We use it to eliminate z from the first equation and get:
dx dy
x+y = . (105)
y(x + y) + a C1 x(x + y) − a x+y
C1

This crumbles to
dx dy
a = . (106)
y + C1 x − Ca1
We are experts at solving such kind of things! Yes! the method of separation of variables
yields:
a a 2a
(x − ) dx = (y + ) dy, x2 − y 2 − (x + y) = C2 .
C1 C1 C1
Upon the use of x+y
z
= C1 , we get
x2 − y 2 − 2az = C2.

Method 4

When one of the variables is absent from one of the equations, then the derivation of one of
the solutions is a simple matter. Suppose, for the sake of definiteness, that the equation
dy dz
=
Q R
may be written in the form
dy
= f (y, z).
dz
Then by the theory of ordinary differential equations this equation has a solution of the
form
φ(y, z, C1) = 0.
Solving this equation for z and substituting the value of z so obtained in the equation
dx dy
= ,
P Q
we obtain an ODE of the type
dy
= g(x, y, C1)
dx
whose solution
ψ(x, y, C1, C2) = 0
can readily be obtained.

63
Example 29. Find the integral curves of the equation
dx dy dz
= = (107)
x+z y z + y2

Solution 29. We notice that x is absent from

dy dz
= .
y z + y2
This equation is equivalent to
dz z
− =y
dy y
that on solving gives
z = C1 y + y 2 .
From the first equations we have
dx dy
=
x + y 2 + C1y y
that is linear in x and can be solved to give
x = C1 y ln y + C2 y + y 2 .

For further reading on differential equations in three variable see for instance [5]

Exercise 18. Solve the following equations

1. (y+z) dx+(z+x) dy+(x+y) dz = 0. 2. yz dx−2xz dy+(xy−zy 3) dz = 0.
dx dy dz
3. (x + z)2 dy + y 2 (dx + dz) = 0. 4. 2
= 2 = .
x y nxy
dx dy dz dx dy dz
5. = = . 6. = = 2 .
mz − ny nx − lz ly − mx y − zx yz + x x + y2
dx dy dz dx dy dz
7. 2 2
= 2 2
= 2 2
. 8. 2 2 2
= = .
x(y − z ) y(z − x ) z(x − y ) x −y −z 2xy 2xz
dx dy dz a dx b dy c dz
9. = = . 10. = = .
x(y − z) y(z − x) z(x − y) (b − c)yz (c − a)zx (a − b)xy
dx dy dz dx dy dz
11. = = 2
. 12. 2 3 3
= 2 3 3
= 2 3 .
xz − y yz − x 1−z x (y − x ) y (z − x ) z (x − y 3)

64
9 Lecture 9: Qualitative Analysis of First order Dif-
ferential Equations

9.1 Introduction

It is not easy to convey the beauty of ordinary differential equations in the traditional ap-
proach. The traditional approach dwells on a few and specialized tricks and techniques for
solving differential equations. However, most of the most important differential equations
are nonlinear systems that cannot be easily solved by these hodgepodge of tricks. A more
accommodating approach that makes use of the current technology, computers and calcu-
lators, involves both qualitative and numerical techniques. The techniques that we shall
discuss in these lecture are usable in contemporary mathematical research where ODEs are
the tools of modeling.

The techniques make use of basic calculus. Just as a graph of a function can be said
to be a complete geometric representation of a function, so the direction field of first order
differential equation, described below, represents the differential equation. This fundamental
idea can be used to get information about solutions without going to the trouble of finding
analytic forms (formulas) for solutions.

Even if we can find an explicit expression for a solution, we shall often work with equation
both numerically and qualitatively to understand the geometry and the long-term behaviour
of solutions.

Objectives

At the end of this lecture, we should be able to:

• Obtain graphical representation of some differential equations.

• Obtain direction fields of some differential equations.
• Describe a solution without actually obtaining its analytic form.
• Obtain equilibrium solutions to some differential equations.
• Deduce the long-term behaviour of some solutions by studying their stability.

We now take a journey through these objectives. We shall depart from direction fields to
an example of qualitative analysis of solution and then to long-term behaviour of solutions.

65
9.2 Direction Fields

Since ẋ can be interpreted as a slope, a differential equation written in the normal form

ẋ = f (t, x)

assigns slope f (t, x) to the point (t, x) in the tx-plane. Such an assignment is called direction
field or slope field. Note that if x(t) is a solution to the differential equation, then

ẋ(t) = f (t, x(t)).

But since ẋ(t) is the slope of the tangent to the graph of the solution at the point with
coordinates (t, x(t)), the number f (t, x(t)) is necessarily also equal to that slope. The slope
then specifies a direction along the solution curve x = x(t). From one point of view, we can
think of starting with the solution x = x(t) and locating a short segment of tangent line at
each point of the graph x = x(t). Consider the differential equation

Example 30.
ẋ = −2tx =: f (t, x). (108)

The slope, ẋ = −2tx, of the solutions x(t), is determined once we know the values for t
and x. For instance if, at t = 1, x = −1 the slope is ẋ(1) = (−2).1.(−1) = 2; that is, x(t)
passing through or at (1, −1) and has a slope of 2. We indicate this graphically by inserting
a small line segment at the point (1, −1) of slope 2 as shown in Figure 1.

Figure 1: A Slope field for Equation(108).

Thus, the solution of the differential equation with initial condition x(1) = −1 will look
similar to this line segment as long as we stay close to t = −1.

Of course, doing this at just one point does not give much information about the solu-
tions. We have to do this simultaneously for as many points as possible. For this example

66
we get the shape in Figure 2.

Figure 2: Complete Slope field for Equation(108).

We can get an idea as to the form of the differential equation’s solutions by “connecting
the dots.”So far, we have graphed little pieces of the tangent lines of our solutions. If the
sketch is skillfully made, it is often possible to get a good idea of what the graph of solutions
x = x(t) should look like; the principal is that a solution graph should be tangent to the
segment located at each point that the graph passes through. The “true”solutions should
not differ very much from those tangent line pieces! For Equation (108), it has the shape
shown in Figure 3.

Figure 3: Complete Slope field with solution curves for Equation(108).

Exercise 19. Draw the direction fields for ẋ = t − x

67
9.3 An example of Qualitative Analysis

Example 31. Give a qualitative analysis of the equation

ẋ = e−x . (109)

Solution 31. Note that the vector field is a function of the dependent variable x alone.
Such equations are referred to as autonomous. Autonomous differential equations have
a very special property; their slope fields are horizontal-shift-invariant; that is, along a
horizontal line the slope does not vary. Figure 4 illustrates this phenomena.

Figure 4: Complete Slope field for Equation(109).

Exercise 20. What is special about the solutions to an autonomous differential equation?

We now give a qualitative analysis for a simple differential equation which describes the
growth with a natural ceiling.
Example 32. Give a qualitative analysis of the equation
x
ẋ = 0.3x(1 − ). (110)
15

Solution 32. The the direction field and the solution curves are as given in the figures 5
and 6.

68
 Figure 5: Slope field for Equation(110).

The solutions of the logistic Equation(110) have the form:

e0.3t
x(t) = 1 0.3t ,
15
e +C
where C is a constant of integration.

Figure 6: Solution curves and Slope field for Equation(110).

Note the following facts from Figure 6.

• If the initial condition is 0 < x(t0) < 15, the solution increases with time to x(t) = 15.
Here, t0 can be any arbitrary starting time.
• If the initial condition is x(t0) > 15, the solution always decrease to x(t) = 15

69
 • The solution cannot change once x(t) = 15, it will remain so for all t.

• The zero function is a solution corresponding to initial conditions of the form x(t0 ) = 0.

The foregoing discussion is an example of a qualitative analysis of Equation(110).

Example 33. Give a qualitative analysis of the equation

ṗ = 2 − p. (111)

Solution 33.The solution curves and the direction field are as given in Figure 7.

2+2 exp(−t)
4

1
p

−1

−2
−2 0 2 4 6 8 10
t

Figure 7: A graphs of five solutions superimposed on the slope field for Equation (111)

Note the following facts from Figure 7:

• The solution increases with time if the initial condition is positive; that is if 0 < p(t0 ) <
2. Here, t0 can be any arbitrary starting time.

• The solution curve increases if p < 2 and decreases if p > 2.

• The solution curves level off as they approach the line p = 2.

• The function p(t) = 2 is a constant solution to Equation(111)

• No solution curve can cross the line p = 2.

• Solutions of Equation(111) cannot oscillate. The slope of an oscillating function

changes sign, because the function alternately increase and decreases. However, in
the case of Equation(111), the slope of a solution curve cannot change sign unless the
solution curve crosses the line p = 2, which is impossible!

The foregoing discussion is an example of a qualitative analysis of Equation(111).

70
Exercise 21. 1. Consider the differential equation Ṡ = S 3 − 2S 2 + S.
(a) By hand, give a rough sketch of the slope field.
(b) Using this sketch, sketch graphs of solutions S(t) with initial conditions S(0) = 0, S(0) =
1
2
, S(1) = 12 , S(0) = 32 , S(0) = −1
2
.

2. Consider the differential equation ẋ = f (x), where the graph of f (x) is given by Figure 8

Figure 8: f (x) versus x

(a) Sketch the slope fields of this differential equation.

(b) Sketch the graph of the solution of the differential equation with x(0) = 12 , find the
limt→∞ x(t).
(c) Sketch the graph of the solution of the differential equation with x(0) = −1
2
, find the
limt→∞ x(t).

3. Suppose that a population can accurately be modeled by the logistic equation

p
ṗ = 0.4(1 − )p.
30
Suppose that, at time t = 5, a disease is introduced in the population that kills 25% of the
population per year. To adjust the model, we change the differential equation to
 p
0.4(1 − 30 )p, for t < 5;
ṗ = p
0.4(1 − 30 )p − 0.25p, for t > 5.

(a) Sketch the slope field for this equation.

(b)Using the slope field, sketch some representative solution curves for this equation.
(c)Find the formulas for the solutions of this equation for initial conditions p(0) = 30 and

71
p(0) = 20.
(d) In a few sentences, describe the behaviour of the solutions with initial conditions p(0) =
30 and p(0) = 20.( You can use either the sketches from the slope field or the formulas, but
give a qualitative description of solutions.)

9.4 Equilibrium Solutions, Stability, and Phase diagrams

9.4.1 Introduction

Solving a differential equation can be done in three main ways: analytic, qualitative, and
numerical. We have seen some examples where the above methods are applicable.

When a differential equation is autonomous and contains parameter(s) and we need to

make conclusions relative to this parameter(s), the above techniques may not be appropriate.
For example, consider the logistic equation with a constant harvesting rate E,

ṗ = p(1 − p) − E, (112)

where p := p(t) is the population of the given species at time t. In order to find the optimal
rate E(which makes hunters happy as well as ecologist worried about preserving the species),
numerical techniques may not be the best tool to use. Here we will see how ideas based on
graphical approach will help us to say something about the long-term solutions. Be aware
that these ideas are only valid for autonomous equations.

9.4.2 Equilibria of Autonomous Equations

We frequently speak of equilibrium processes in nature. In the simplest case, an equilibrium

is a state in which the system does not change.

A formal definition is

9.1 Definition. Let ẋ = f (x) be an autonomous first-order differential equation. An equi-

librium solution is a constant function x(t) = x̄ such that f (x̄) = 0 for all time t.

Remark. The concept of an equilibrium solution applies only to autonomous differential

equations.

Example 34. Consider the cooling or warming rate of a class of water in a room at a
constant temperature R.

Suppose that the rate at which a glass of water warms or cools is proportional to the
current temperature difference between the water and the room. That is, if T := T (t) is the
temperature of the water at time t and the room temperature R is held constant, the rate
of temperature change is given by

Ṫ = k(T − R).

72
If the water is initially at room temperature, then no heat flows between the water and
the room, and the water temperature remains constant. Thus, T (t) = R is an equilibrium
solution.
Example 35. Consider the differential equation

ẋ = (1 − x)x.

The equilibrium solutions are found by solving

ẋ = (1 − x)x = 0.

Here we find the solutions to be either x = 1, or x = 0 as the points of equilibria.

9.4.3 Stability

Equilibrium solutions are often of interest, because they are easy to analyse qualitatively
and they also provide important information about the behaviour of a system. For example,
suppose that a population of fish that is initially at equilibrium, constant for along time, is
overhavested to almost a point of extinction. If the harvesting is stopped, will the population
return to its original equilibrium levels?

This kind of question concerns the stability of the equilibrium state. Suppose a system
is initially at equilibrium and is perturbed in some way; a perturbation is this case is some
“small ”disturbance. Will the system return to the equilibrium? If the answer is yes, then
we say that the equilibrium is stable or attracting or a sink. If not, then we say that the
equilibrium is unstable or repelling or asource. If it is neither a source nor a sink it is
called a node. We give an informal definition of stability as follows:
9.2 Definition. Let x = x̄ be an equilibrium point of the differential equation ẋ = f (x).
The equilibrium is stable if solution curves starting at initial values near x̄ return to x̄ at
t → ∞. Otherwise, the equilibrium is unstable.

Note: The concept of stability refers only to small perturbations, however, as the next
example shows.
Example 36. The population of some species of plants and animals does not recover if the
number of individuals in a given area drops below a certain threshold. For instance, sexual
reproduction is inhibited if individuals are too widely scattered. If female pawpaw and male
pawpaw are far apart, the pollen transfer is much less and fertilisation is greatly reduced.
For animals less encounter between mates results in lower reproduction.

Suppose that a population can recover provided that the number of individuals is not
reduced below half of the carrying capacity, where carrying capacity of an environment
is the maximum number of species that environment can carry. A simple model of this
situation is given by the equation
p p
ṗ = rp(1 − )( − 1) =: f (p), (113)
N M

73
where p := p(t) is the population at time t,
r = 2 is growth-rate coefficient (parameter),
N = 2 is the carrying capacity(parameter), and
M = 1 is the “sparsity”constant(parameter).
Let us look at the solution curves the in Figure 9.

Figure 9: Solution curves for Equation(113).

We observe the following from Figure 9:

• The Equilibrium points, f (p̄) = 0, to Equation(113) are p̄ = 0, 1, 2.
• For initial condition in 1 < p0 < 2, ṗ > 0. This indicates that the solution increases to
p̄ = 2 with time. By uniqueness it remains there for all t.
• Values of p large than 2, the carrying capacity, are not sustainable, because if p > 0, then
ṗ < 0, reflecting the influence of population pressures, as the graph shows. The solution
curve decreases to p̄ = 2 if the initial condition 2 < p0 < ∞.
• The solution curves starting near p̄ = 2 level to it while any with initial condition p0 = 2
stays with p(t) = 2 for all t. Thus p = 2 is a stable equilibrium. Implying the population
remains at p(t) = 2 for all time.
• The solution curve decreases to p̄ = 0 if the initial condition 0 < p0 < 1. Note that the
population can survive indefinitely at p = 1 as long as there is no disturbance; something
unlikely in real life. Any small reduction below this threshold, p = 1, results in eventual
extinction. While any small increase to a value slightly above p = 1 leads to p = 2. Thus
p = 1 is an unstable equilibrium.
• The solution curve starting near p̄ = 1 move away from it to either p̄ = 2 if p > 1 or to
p̄ = 0 if p0 < 1.
• The solution curves starting near p̄ = 0 level off to it as t → ∞.

Thus p̄ = 2, 0 are stable equilibrium states whilst p̄ = 1 is unstable.

This example also serves to illustrate why “small”perturbations are emphasised in the

74
 definition of equilibrium. For instance p = 2 is a stable equilibrium, but we can only revert
to it if the perturbation does not go below p = 1. The perturbation δ has to be such that
1 < 2 − δ.

Is there a way of testing for stability without such a sketch? The answer is provided in
the theorem below:

Theorem 9.1. Let ẋ = f (x) for some continuously differentiable function f , and suppose
x(t) = x̄ is an equilibrium solution. If f 0 (x̄) > 0 the equilibrium point is unstable. If
f 0 (x̄) < 0 then the equilibrium point is stable.

Remark. If f 0 (x̄) = 0 or if f 0 (x̄) does not exist, then we need additional information to
determine the nature of stability of x̄. It could be stable or unstable

Exercise 22. Check the stability of the fixed points in Equation(113) using Theorem 9.1.

Exercise 23. Consider the differential equation ẋ = f (x) given graphically in Figure 10.

Figure 10: f (x) versus x

Analyse this equation in a manner similar to 113 and check the stability of the fixed points
using Theorem 9.1.

Summary Let us write down some general steps to follow when dealing with the au-
tonomous equation
ẋ = f (x).

Step 1. Find the constant or equilibria solutions through solving:

f (x) = 0.

Draw the constant solutions. Note that here we are drawing x versus t.

Step 2. Find the sign of the function f (x) through its graph (versus x).

Step 3. In the region between any two equilibria, the solutions will be all increasing (if f (x) is
positive) or decreasing (if f (x) is negative).

75
Step 4. If a solution is increasing and bounded above by a constant solution x = L, then the
limit of the solution when t → +∞ is the number L. Otherwise the limit is +∞.
If this solution is bounded below by a constant solution x = `, then the limit of the
solution when t → −∞ is the number `. Otherwise the limit is −∞.

Step 5. If a solution is decreasing and bounded below by a constant solution x = L, then the
limit of the solution when t → +∞ is the number L. Otherwise the limit is −∞.
If this solution is bounded above by a constant solution x = `, then the limit of the
solution when t → −∞ is the number `. Otherwise the limit is +∞.

Step 6. Draw some solutions.

Remark. You must be very careful here, since two graphs will be drawn: one (x, f (x)) and
another one for the solutions, where this time we are plotting (t, x) for various appropriate
initial conditions. This second graph is the most important as it contains what we are looking
for : the solutions of the differential equation.
Example 37. Draw some solution for the equation

ẋ = 0.2x(5 − x)(x − 2) =: f (x) (114)

Solution 37. The equilibrium points are given by x̄ = 0, 5, 2. The graph of (x, f(x)) is
given by Figure 11.

Figure 11: Graph of (x, f (x)) for Equation(114).

From the graph we observe the following:

Clearly f 0 (0) < 0 and f 0 (5) < 0. These points are thus stable. f 0 (2) > 0, hence x̄ = 2 is
unstable.
• The solution x(t) with initial condition −∞ < x(0) < 0 will tend to 0 as t → ∞.
• The solution x(t) with initial condition 0 < x(0) < 2 will tend to 0 as t → ∞.

76
• The solution x(t) with initial condition 2 < x(0) < 5 will tend to 5 as t → ∞.
• The solution x(t) with initial condition 5 < x(0) < ∞ will tent to 5 as t → ∞.
Figure 12 shows some solution curves.

Figure 12: Solution curves for Equation(114).

9.4.4 Phase Line

Direction fields are often time consuming to draw, and information about the stability of
equilibrium solutions sometimes is difficult to discern. In the case of autonomous equations
of the form
ẋ = f (x),
all of the information that is needed can be extracted from a plot of (x, f(x)). Such a plot
is called phase diagram.

From the phase diagram, we can determine the following:

• The equilibrium points;

• The intervals of x-values for which f (x) > 0 or f (x) < 0;

This information is useful in determining the so called Phase line. It gives information on
a fixed points, and how x varies on some intervals. We now describe how to draw a phase
line.

• Along the x-axis, locate the equilibrium points.

77
 • Find the intervals of x-values for which f (x) > 0, and draw arrows pointing to the
right in this interval.
• Find the intervals of x-values for which f (x) < 0, and draw arrows pointing to the left
in this interval.

We look at an example where we give a complete picture of a phase diagram, phase line,
solution curves, and direction field.
Example 38. Consider the equation
ẋ = f (x) := 0.2x(5 − x)(x − 2), (115)
ẋ = f (x) := x2. (116)
Show their respective phase lines.

Solution 38

Figure 13:Phaseline for Equation(115).

78
 Figure 14: Phase line for Equation(116).

By considering the phase diagram, phase line, and the direction field of an autonomous
system, we are able to say when its solution is in equilibrium. We are also able to know
when it is increasing or decreasing and, how it depends on initial conditions. All these is
gained without having to get its analytic form. That is the power of graphical and qualitative
analysis.

Exercise 24.

Question 1. Consider the differential equation ẇ = (2 − w) sin w.

(a)Draw the phase line for the differential equation and classify the equilibrium points.
(b) State the initial conditions which lead to a stable or unstable equilibrium.
(c) Sketch graph of solution curves.

Question 2. Consider the differential equation ẋ = f (x) with f (x) given in Figure 15.

Figure 15: Graph of (x, f (x))

(a) Draw the phase line for the differential equation and classify the equilibrium points
as sinks, sources,or nodes.
(b)Give a rough sketch of the slope field for this differential equation, and draw a few solu-
tions into the slope field.
(c) Consider the solution to the differential equation which satisfies the initial condition

79
(i)x = 2. Find
lim x(t)
t→∞

(ii)x = 1. Find
lim x(t).
t→∞

Question 3. Consider the modified logistic differential equation

S S
Ṡ = rS(1 − )( − 1) =: f (S),
N M
where S := S(t) is the population at time t,
r is growth-rate coefficient (parameter),
N is the carrying capacity(parameter), and
M is the “sparsity”constant(parameter) with 0 < M < N and r > 0.
(a)Find the equilibrium points. Classify them as: source, sink or node. Justify your answers.
(b) Draw the phase line for the differential equation.
(c) State the initial conditions which lead to a stable or unstable equilibrium.
(d) Sketch graph of solution curves for all viable initial conditions.

Question 4. Suppose you wish to model a population with a differential equation of the
form ṗ = f (p), where p(t) is the population at time t. Experiments have been performed on
the population that give the following information:
• The only equilibrium points are p = 0, 10, and p = 50.
• If the population is 100, the population decreases.
• If the population is 25, the population increases.
(a)Sketch the possible phase lines for this system for p > 0
(b)Give a rough sketch of the corresponding functions f (p) for each of your phase lines.
(c)Give a formula for functions f (p) whose graph agree(qualitatively) with the rough sketches
in part (b) for each of your phase lines.

Question 5 Consider the population model

dP  P 
= 0.4 1 − P, (117)
dt 230
where P (t) is the population at time t.
(a) For what values of P is the population increasing?
(b) For what values of P is the population decreasing?
(c)For what values of P is the population in equilibrium?

Question 6 Consider the differential equation

dy
= y 3 − y 2 − 12y (118)
dt
where y = y(t)
(a) For what values of y is y(t) increasing?
(b) For what values of y is y(t) decreasing?

80
(c)For what values of y is y(t) in equilibrium?

Question 7 The table below provides the land area in Australia colonized by the American
marine toad( Bufo marinis) every five years from 1939-1974. Model the migration of this
toad using an exponential growth model
dP
= kP, (119)
dt
where P (t)=land area occupied at time t. Make predictions about the land area occupied
in the years 2010, 2050, and 2100. You should do this by
(a) solving the initial-value problem,
(b) determining the constant k,
(c)computing the predicted areas, and?
(d)comparing your solutions to the actual data. Do you believe your prediction?

Year Cumulative area occupiedin km2

1939 32800
1944 55800
1949 73 600
1954 138000
1959 202000
1964 257000
1969 301000
1974 584000
The area of Australia is 7619000km2 .

Question 8 Consider an elementary model of the learning process of certain list of SMA
208 ODE I formulae. If we let L(t) be the fraction of the list learned at time t, where L = 0
corresponds to knowing nothing and L = 1 corresponds to knowing the entire list, then we
can form a simple model of this type of learning based on the assumption:
• The rate of learning is proportional to the amount left to be learned. Since L = 1
corresponds to knowing the entire list, the model is
dL
= k(1 − L), (120)
dt
where k is the constant of proportionality.
(a) For what value of L, 0 ≤ L ≤ 1, does learning occur most rapidly?
(b) Suppose two students memorize lists according to the same model:

dL
= 2(1 − L), (121)
dt
(i) If one of the students knows one-half of the list at time t = 0 and the other knows none
of the list, which student is learning most rapidly at this instant?
(ii) Will the student who starts out knowing none of the list ever catch up to the student
who starts out knowing one-half of the list?

81
(c)Consider two equations showing the rate of memorizing by two students’ Juma and Brenda
respectively,
dLJ
= 2(1 − LJ ), (122)
dt
and
dLB
= 3(1 − LB )2. (123)
dt
(i)Which student has a faster rate of learning at t = 0 if they both start memorizing together
having never seen the list of formulae before?
(ii)Which student has a faster rate of learning at t = 0 if they both start memorizing together
having already learned half of the formulae list?
(iii)Which student has a faster rate of learning at t = 0 if they both start memorizing
together having already learned one-third of the formulae list?

82
 10 Lecture 10: Bifurcations

Objectives

At the end of this lecture, we should be able to:

• Define Bifurcation.
• Define and determine a bifurcation point for a given differential equation.
• Sketch a bifurcation diagram for a given differential equation.
• Determine the conditions under which bifurcation is not possible for a given differential equation.
• Describe how a solution qualitatively depends on parameters around the bifurcation point.

10.1 Introduction: What is bifurcation?

In many of our models, a common feature is the presence of parameters along with the other
variables involved. Parameters are quantities that do not depend on time( the independent
variable) but assume different values depending on the specifics of the application at hand.
For instance, consider the logistic equation describing a certain fish population

ṗ = p(1 − p) − E, (124)

where p := p(t) is the population of fish at time t, and E > 0 is a constant harvest rate.
Here is a simple example of real-world problem modeled by a differential equation involving
a parameter (constant rate E). Clearly, the fisherman will be pleased if E is big, whilst
ecologist will argue for a smaller E (in order to protect the fish population). What then is the
“optimal”constant E (if such exists) which allows maximal harvesting without endangering
the survival of the fish population? We see that different values of E may lead to different
values of p. This is the problem of how solutions depend on the parameter E.

We wish to study how the solutions behave as one varies a parameter in a differential
equation. Occassionaly changes in parameters may lead to a drastic change in the long-term
behaviour of solutions. Such a change is called a bifurcation. We say that a differential
equation that depends on parameters “bifurcates”if there is a qualitative change in the
behaviour of solutions as the parameter changes.

Let us have a quick analysis of Equation(124) in order to have a feel of Bifurcation.

The fixed points of Equation(124) are given thus:

p(1 − p) − E = 0,

83
which gives √
1± 1 − 4E
p= . (125)
2
Let us draw this on a diagram with two axes. See Figure 16.

Figure 16: Graph of P versus E; Equation(125).

We see that from Figure 16 that:

• if E < 14 , then we have two equilibria solutions;
• if E = 14 , then we have one equilibrium solution;
• if E > 14 , then we do not have any equilibrium solution.

This is an example of what is meant by “bifurcation”. As we can see the number of

equilibria solutions changes from two, for E < 14 to one, for E = 14 , and no equilibrium for
E > 14 . Hence the phase line will change at E = 14 . This is called the bifurcation value.
Note that this is just one form of bifurcations; there are other forms or changes, which are
also called bifurcations.

10.2 The Bifurcation Diagram

An extremely helpful way to view bifurcations is through a Bifurcation Diagram. This

is a picture of the phase lines near a bifurcation value, highlighting the changes that the
phase lines undergo as the parameter passes this value. To plot the bifurcation diagram for
the differential equation ẋ = f (x, µ), where µ is a parameter, we plot the parameter values
(µ) along the horizontal axis. For each µ, we draw the phase line corresponding to it on the
verticle line through (µ, 0). We think of the bifurcation diagram as a movie: As our eyes
scans the picture from left to right, we see the phase lines evolve through the bifurcation.
Figure 17 is the bifurcation diagram for Equation(124).

84
 Figure 17: Bifurcation diagram for Equation(125).

Let us use this diagram to discuss the fate of the fish population as the parameter E in-
creases.
When E = 0, no fishing, the fish population tends to the carrying capacity p = 1 which is a
sink.
If E increases but stays smaller than 14 , the fish population still tends to a new smaller
number given in Equation (125) which is also a sink.
When E is increased to exceed 14 , then the differential equation has no equilibrium points.
The fish population is decreasing and crosses the t−axis at finite time. This means the fish
population will vanish completely in finite time. Hence, in order to avoid such a catastrophic
outcome, E needs to slightly lower than 14 , which is called the optimal harvesting rate. For
E = 14 the only equilibrium point is P = 0.5, a node, and as soon as P falls under 0.5, we
will again witness extinction in finite time.

10.3 When do bifurcations not occur?

We wish to know whether bifurcation will or will not occur for a given differential equation
with a parameter. Consider a differential equation

ẋ = f (x, µ),

where µ is a parameter. Let (x0 , µ0 ) be a fixed point that is either a sink or a source. A
small change in a parameter around this fixed will create a fixed point (x1 , µ1), say, that is
either a sink or a source respectively. See the figure below:

85
 Figure 18 Graphs of f (x, µ0) and f (x, µ1)

The graph indicates the curves for f (x, µ0 ), f (x, µ1 ) for µ1 close to µ0 . Note that f (x, µ1 )
decreases across the x-axis at x = x1 near x0 , so ẋ = f (x, µ1 ) has a sink at x = x1 . There
shall be no bifurcation.

If
f (x0, µ0 ) = 0
and
∂f
|(x ,µ ) = 0,
∂x 0 0
then there may be a bifurcation at (x0, µ0 ). The general conditions on f (x, µ) to undergo a
bifurcation at a particular fixed point (x0 , µ0 ) shall be studied in a higher level course.

Determining Bifurcation Values for ẋ = f (x, µ).

Just solve for µ in 
f (x0 , µ0 ) = 0,
∂f
|
∂x (x0 ,µ0 )
= 0. .

10.3.1 Notation for differential equations depending on a parameter

An example of a differential equation depending on a parameter is

ẋ = x2 − 2x + µ. (126)

The parameter is µ. The independent variable is t and the dependent variable is is x, as

usual. Note that for different values of µ, we get a different differential equations from
Equation(126); that is, a family os equations. We call an equation such as Equation(126)
as a one-parameter family of differential equations. To emphasise the importance of

86
the dependence of the differential equation on the parameter µ, we write Equation(126) as
follows:
ẋ = x2 − 2x + µ =: f (x, µ); (127)
that is,
ẋ = f (x, µ) (128)
where
f (x, µ) = x2 − 2x + µ.
Exercise 25. Sketch the phase line for Equation(127) for µ ∈ {−4, −2, 0, 2, 4} and classify
the equilibrium points if they exist.
At what value of µ do we have a qualitative change in the number of equlibrium points and
their nature of stability?

Exercise 26.

Question 1. Consider the constant harvest rate model of fish given below.
p
ṗ = rp(1 − ) − E, (129)
N
where r is the growth rate parameter, p := p(t) is the population of fish at time t, N is the
carrying capacity of the area, and E > 0 is a constant harvest rate. How does the population
of fish vary as E is increased? We answer this question via the following little questions.

(a) What will be the long-term population when harvesting is banned; that is, E = 0?

(b) Find an expression for fixed points for positive values of E and classfiy each.
(c) Sketch, on the same grid, the graph of f (p, E) := rp(1 − Np ) − E against P for several
positive values of E. What do you notice about the separation of the equilibrium
points?
(d) Using E ≥ 0 as a parameter, determine the bifurcation value and sketch the bifurcation
diagram. On the bifurcation diagram indicate all possible phase lines.
rN
(e) What is likely to happen to the fish population when E > 4
.

(f) Consider the population model

p2
ṗ = 2p − , (130)
50
for a species of fish in a lake. Time is measured in years. Suppose it is decided that
fishing will be allowed in the lake, but it is unclear how many fishing licenses should be
issued. Suppose the average catch of a fisherman with a license is 3 fish per year(these
are hard fish to catch). [Hint; To answer the following questions use Equations (129)
and (130) and the results in (a).]

87
 (i) What is the largest number of licenses that can be issued if the fish are to have a
chance to survive?
(ii) Suppose the number of fishing licenses in part (b)(i) is issued. What will happen
to the fish population; that is, how does the behaviour of the population depend
on the initial population?

Question 2. Consider the population of some squirrel on the slopes of mount Elgon that
is modeled thus:
S S
Ṡ = rS(1 − )( − 1) =: f (S), (131)
N M
where S := S(t) is the population of the squirrel at time t,
r is growth-rate coefficient (parameter),
N is the carrying capacity(parameter), of the area,
M is the “sparsity”constant(parameter) with 0 < M < N and r > 0. The parameters M,
and r remain relatively constant for a long time, but as more people settle into the area, the
carrying capacity N (parameter) decreases.

(a) Sketch the graph of the function f (S) for fixed values of r and M and several values
of N .
(b) At what value of N does a bifurcation occur?
(c) How does the population of squirrels behave if the parameter N is slowly and contin-
uously decreased through the bifurcation value?

Question 3. Suppose we observe that an increase in the human population of an area

makes the area less desirable to squirrels, so they emigrate from the region. We can adjust
the model in Equation(131) to take account of this emigration by subtracting a fixed rate E
of emigration. We have:
S S
Ṡ = rS(1 − )( − 1) − E =: f (S, E), (132)
N M
where the symbols are as in question 2.

(a) Sketch the graph of the function f (S, E) for fixed values of r, N and M and several
values of E.
(b) At what value of E does a bifurcation occur? (The answer should be in terms of
r, M, N).
(c) How does the population of squirrels behave if the parameter E is slowly and contin-
uously increased through the bifurcation value?

Question 4. Consider the differential equation

ẋ = µ − x2 , x ∈ R1, µ ∈ R1 (133)

where µ is a parameter.

88
 (a) Draw the bifurcation diagram for this differential equation.

(b) Find the bifurcation values, and describe how the behaviour of the solutions changes
close to each bifurcation value.

Question 5. Repeat Question 4 for

ẋ = µx − x2 , x ∈ R1, µ ∈ R1 (134)

Question 6. Repeat Question 4 for

ẋ = µx − x3 , x ∈ R1, µ ∈ R1 (135)

Question 7. Consider the equation

ẋ = µx − x3 , x ∈ R1, µ ∈ R1 (136)

Is there a bifurcation at (x, µ) = (0, 0)? Explain why.

89
11 Lecture 11: Second order Linear Differential Equa-
tion

11.1 Introduction

This lecture is devoted to a study of second order linear differential equations. We shall look
at basic definitions and properties associated with such equations. We begin by considering
equations with constant coefficients and then look at those with variable coefficients. In each
case we shall discuss methods of finding the solutions to these equations. The Wronskian
determinant is defined and its properties and applications given.

Objectives

At the end of this lecture we should be able:

• To solve homogeneous linear equations.

• To determine Linear Independence of Solutions by use of the Wronskian.
• To use the method of reduction of order.
• To solve homogeneous Equations with Constant Coefficients,
• To solve nonhomogeneous linear Equations by:
◦ The method of Undetermined Coefficients,
◦ The method of Variation of Parameters.
• To solve Euler-Cauchy Equations.

11.2 Second Order Linear Differential Equation: Basic Ideas

A second order linear differential equation is often written as

p0 (x)y 00 + p1 (x)y 0 + p2 (x)y = d(x), (137)

where pi (x), d(x), i = 0, 1, 2 are continuous functions on some interval I. The interval I is
called the interval of definition. Points where p0 (x) = 0 are called singular points, and
often the equation requires special consideration at such points. In this Lecture, we assume
that p0 (x) 6= 0. By dividing through Equation(137) by p0 (x) 6= 0 we get:

y 00 + a1(x)y 0 + a2(x)y = b(x), (138)

90
where
p1 (x) p2 (x) d(x)
a1 (x) = , a2(x) = , b(x) = .
p0 (x) p0 (x) p0 (x)
For purpose of notational brevity, we designate the left side of Equation (138) by L(y). Thus
(138) becomes
L(y) = b(x). (139)

11.1 Definition. (i) If b(x) = 0 for all x ∈ I, then L(y) = 0 is a homogeneous linear
differential equation, whereas if b(x) 6= 0 for some x ∈ I, then L(y) = b(x) is called a
nonhomogeneous linear differential equation. To a nonhomogeneous differential equa-
tion, Equation(138), we can associate the so called associated homogeneous equation.
(ii) If ai (x) = ai , i = 1, 2 are constants then we refer to (138) as a linear differential equa-
tion with constant coefficients otherwise it is a linear differential equation with variable
coefficients.

Example 39. The equation

y 00 + 2y 0 + 3y = cos x (140)
is an example of a nonhomogeneous linear differential equation with constant coefficients
while
y 00 + (2x − 1)y 0 + sin(ex )y = 0 (141)
is an example of a homogeneous linear differential equation with variable coefficients.

Remark.

1. The term linear refers to the fact that each expression in the differential equation is of
degree one in variables y, y 0 and, y 00

2. L is a differential operator
L is a differential operator which takes each function φ, which has 2 derivatives on I, into
the function L(φ) on I whose value at x is given by

L(φ(x)) = φ00(x) + a1(x)φ0(x) + a2(x)φ(x) = b(x). (142)

Thus a solution of (138) on I is a function φ on I having 2 derivatives there, and is such

that Equation(142) is satisfied.

3. The operator L is linear

That is, for any functions φ1 (x), and φ2 (x) defined on I and any constants c1 , c2 , we have

L(c1φ1 (x) + c2φ2 (x)) = L(c1φ1 (x)) + L(c2 φ2(x))

= c1L(φ1 (x)) + c2L(φ2 (x)).

11.3 Linear Homogeneous Equations: Basic Properties

Our main objective here is to find all solutions to L(y) = 0. For this purpose we shall always
use the following:

91
Basic Property: It is not difficult to see that if, φ1(x), and φ2(x) are any two solutions
of L(y) = 0, then
φ(x) = c1φ1 (x) + c2φ2 (x), (143)
where c1, c2 are arbitrary constants is also a solution.
Remark. The function φ(x) in Equation(143) is said to be a linear combination of the
functions φ1 (x), φ2(x).

This fact is correct as the little proof below indicates:

We use the linearity of L(y) and the fact that L(φ1(x)) = 0 and L(φ2 (x)) = 0; as they are
solutions of L(y) = 0. Hence

L(φ(x)) = L(c1 φ1(x) + c2 φ2(x))

= c1 L(φ1(x)) + c2 L(φ2(x)) = 0.

This implies that φ(x) = c1φ1 (x) + c2φ2 (x) is a solution of L(y) = 0.

Exercise 27. Check that sin x and cos x are solutions to y 00 + y = 0. Further, check that for

φ(x) = c1 sin x + c2 cos x,

where ci , i = 1, 2 are constants, is also a solution.

This result holds in general and we have:

Theorem 11.1. Any linear combination of solutions of a homogeneous linear differential

equation is also a solution.

The main result of this section will be addressed if we could answer the following question:

Question: Is any, general, solution φ of a second order homogeneous linear

differential equation of the form (143)?
The answer to this question uses the notion of linear independence of solutions. It is provided
at the end of the next two subsections.

11.4 Linear Independence

Given two nonzero functions φ1 (x), φ2(x), if constants c1 , c2 , not all zero can be found such
that
c1 φ1 (x) + c2 φ2 (x) = 0 (144)
for all x ∈ I, then φ1 (x), φ2(x) are said to be linearly dependent. If no such ci ’s exist;
that is, ci = 0, i = 1, 2 for all i, then the functions φi (x), i = 1, 2 are said to be linearly
independent.

92
11.4.1 The Wronskian

We shall give sufficient conditions under which two differentiable functions, φ1 (x), φ2(x)
defined on I are linearly independent.

Let
c1φ1 (x) + c2φ2 (x) = 0. (145)
Equation(145) with its first derivative give us the system
c1 φ1(x) + c2 φ2(x) = 0 (146)
c1 φ01(x) + c2 φ02(x) = 0. (147)
This is equivalent to
Φc = 0, (148)
(i−1) 2
where Φ = [φj ]i,j=1 , and c = [c1, c2 ]T .

The function W (φ1 , φ2)(x), defined as

φ1 (x) φ2(x)
W (φ1, φ2 ) ≡ det Φ = = φ1 (x)φ02(x) − φ01(x)φ2(x)
φ01 (x) φ02(x)
is called the Wronskian of the two functions φ1 (x), φ2 (x) defined on I. As a consequence of
the properties of determinants, it clearly follows that W (φ1, φ2)(x) 6= 0 implies φ1 (x), φ2(x)
are linearly independent on I. Furthermore, we have the following important properties:

1. If φ1 (x), φ2(x) are two solutions of L(y) = 0, then

 Z x 
W (φ1, φ2)(x) = W (φ1, φ2)(x0 ) exp − a1 (s)ds , (149)
x0

where x0, x ∈ I.

2. If φ1(x), φ2 (x) are two solutions of L(y) = 0, then

W (φ1, φ2)(x) 6= 0 for every x ∈ I ⇐⇒ ∃x0 ∈ I such that W (φ1, φ2)(x0 ) 6= 0.
In this case, we say that the functions φ1 (x), φ2(x) are linearly independent.

3. If φ1(x), φ2(x) are two linearly independent solutions of L(y) = 0, then any solution φ
is given by
φ(x) = c1 φ1 (x) + c2φ2 (x)
for some constants c1 , c2. In this case, the set {φ1(x), φ2(x)} is called the fundamental set
of solutions.

4. If
W (φ1, φ2 ) = 0, (150)
then, the functions φ1 , φ2 are linearly dependent on I.

A few deductions from the above results are in order:

93
 • Statement 2 tells us that once W (φ1, φ2 )(x0) 6= 0 at some point x0 ∈ I, then it is
nonzero for all x ∈ I.

•  φ 0
1 W (φ1 , φ2)
= . (151)
φ2 φ21
Example 40. The functions cos x, sin x are independent as

cos x sin x
W (cos x, sin x) = = cos2 x + sinx = 1, ∀ x.
− sin x cos x

We can also see that φ(x) := c1 cosx + c2 sin x with c1 and c2 as constants, is the general
solution of y 00 + y = 0.
Example 41. Write an expression of the Wronskian to Equation(141) without finding the
solutions.

Solution 41. Let W (x) = W (φ1, φ2 )(x) where φ1, φ2 are the solutions to Equation(141).
The formula of the Wronskian in (149) applied to our problem gives:
 Z x  2
W (x) = W (0) exp − (2s − 1)ds = W (0)e−x +x . (152)
0

Exercise 28.

1. Prove that if, φ1 φ2 are two solutions of L(y) = y 00 + a1y 0 + a2y = 0 on an interval
I 3 x0 , then
W (φ1 , φ2)(x) = W (φ1 , φ2)(x0 )e−a1 (x−x0 )

2. Use the Wronskian formula, Equation(149), to check that the Wronskian of

2 +x
y 00 + (2x − 1)y 0 + sin(exy) = 0, is W (x) = 3e−x .

Exercise 29.

1. Show that the given functions form a linearly independent set of solutions for the corre-
sponding differential equation. Write down the general solution of the differential equation
in each case.
(a) sin 3x, cos 3x; y 00 + 9y = 0.
(b) e2x, e−3x ; y 00 − 4y = 0. √
(c) eiat, e−iat ; y 00 + a2 y = 0, i = −1.
(d)e3x, ex , y 00 + 4y 0 + 3y = 0.
(e) e−2x , xe−2x; y 00 + 4y 0 + 4y = 0.
(f)ex cos 2x, ex sin 2x; y 00 − 2y 0 + 5y = 0.

94
(g) cos x + sin x, cos x + sin x, cos x − sin x; y 00 + y = 0.

2. In the following problems, write an expression for the Wronskian of two linearly
independent solutions of the given equation, without actually finding the solutions.
(a) 2y 00 − 3y 0 + y = 0
(b) xy 00 − 3y 0 − 5y = 0
(c) y 00 + cos xy 0 + ex y = 0.

11.5 General solution of Homogeneous Linear equations

In this section, we look at techniques for finding general solutions for homogeneous lnear
differential equations.
Theorem 11.2. Let φ1, φ2 defined on I, be linearly independent solutions of L(y) = 0. If
φ(x) is a solution of L(y) = 0 on I, then it is necessarily of the form
X
2
φ(x) = ci φi (x),
i=1

where ci ’s are constants.

The theorem says that span {φ1 , φ2} is the solution space of L(y) = 0.

Proof. Theorem 11.2. Consider

L(y) = y 00 + a1(x)y 0 + a2(x)y = 0.
Let φ1 (x) and φ2 (x) be the linearly independent solutions of L(y) = 0 on I, then for any
ξ ∈ I, W (φ1, φ2 )(ξ) 6= 0 implies
c1 φ1(ξ) + c2 φ2(ξ) = φ(ξ)
c1 φ01(ξ) + c2 φ02(ξ) = φ0(ξ) (153)
has a unique solution c1 = α1 , c2 = α2 , say; that is,
α1 φ1(ξ) + α2 φ2(ξ) = φ(ξ)
α1 φ01(ξ) + α2 φ02(ξ) = φ0 (ξ). (154)
Now consider ψ(ξ) = α1 φ1 (ξ) + α2 φ2(ξ), a linear combination of the solutions of L(y) = 0
on I. It is also a solution of L(y) = 0. Besides,
ψ(ξ) = α1 φ1(ξ) + α2 φ2(ξ)
ψ 0 (ξ) = α1 φ01(ξ) + α2 φ02(ξ).
Hence ψ(ξ) = φ(ξ), ψ 0 (ξ) = φ0(ξ). By the uniqueness theorem ψ(ξ) = φ(ξ) for all ξ ∈ I.

Now that we know what a solution to a second order homogeneous differential equa-
tion is, let us see how we can find some of these solutions for some equations. We begin
with linear differential equations with constant coefficients and later look at equations with
variable coefficients.

95
11.5.1 Method of solution of Second order Linear Homogeneous Equations
with constant coefficients

Assume a solution of the form

φ(x) = erx 6= 0. (155)
Substituting this in L(y) = 0, we obtain

L(erx ) = p(r)erx , (156)

where
p(r) = r2 + a1r + a2,
is called a characteristic polynomial of L. Since erx 6= 0, L(erx) = 0 implies that
p(r) = r2 + a1 r + a2 = 0. This is called a characteristic equation. We see that r must be
a root of p(r). Hence once all the zeroes of p(r) are determined, the solution of L(y) = 0 is
simply given by erx .

Note : The roots of p(r) can be obtained from L(y) by simply replacing y (k) with rk , where
y (k) is the k − th derivative of y.

How do these roots look like? Yes, we know, from algebra, that p(r) always has at most
two roots r1 and r2 . The roots may be distinct; that is, r1 6= r2 in which case they can be
real or complex, or be repeated; that is, r1 = r2 . Let us take on each case in turn.

Case 1. r1 6= r2
If r1 6= r2 then the functions er1 x and er2x are the solutions of L(y) = 0, and the general
solution is given by
φ(x) = c1 er2 x + c2er2 x ,
where c1 and c2 are arbitrary constants.

Case 2. r1 = r2
If r1 is double root of p(r) = 0, then p(r) = 0, and p0 (r) = 0.

This suggests differentiating

L(erx) = p(r)erx
with respect to r. In doing this we observe that, since L involves only differentiation with
respect to x,
∂ ∂
L(erx ) = L( erx ) = L(xerx ),
∂r ∂r
and therefore
L(xerx) = [p0 (r) + xp(r)]erx.
Setting r = r1 in this equation we see that L(xerx) = 0, thus showing that apart from
erx, xerx is a another solution. The general solution is given by

φ(x) = c1 er1 x + c2 xer1x ,

96
where c1 and c2 are arbitrary constants.

Case 3. r1 = α + iβ, r2 = r¯1 .

The roots occur in complex conjugates. In this case the solutions are eαx cos(βx) eαx sin(βx)
and, the general solution is then given by

φ(x) = c1 eαx cos(βx) + c2 eαx sin(βx),

where c1 and c2 are arbitrary constants.

Example 42. Solve the equation y 00 + y 0 − 2y = 0.

Solution 42. The characteristic polynomial is p(r) = r2 + r − 2, and its roots are −2, 1.
Therefore the general solution is

φ(x) = c1 e−2x + c2 ex ,

where c1 , c2 are any constants.

Example 43. Solve the equation y 00 + ω 2 y = 0, ω > 0.

Solution 43. The characteristic polynomial is

p(r) = r2 + ω 2 ,

and its roots are iω, −iω. Consequently all solutions are of the form

φ(x) = c1 eiωx + c2e−iωx

= (c1 + c2 ) cos ωx + i(c1 − c2) sin ωx. (157)

Let c1 − c2 = α1 and i(c1 − c2 ) = α2, be any constants. Then the general solution becomes

φ(x) = α1 cos ωx + α2 sin ωx

Exercise 30.

1. Find the general solution to each of the following differential equations

(a) y 00 + 2y 0 + 2y = 0.
(b) y 00 − 2y 0 + 5y = 0
(c) y 00 − 4y 0 + 7y = 0
(d) y 00 + y 0 − 6y = 0
(e) y 00 + a2y = 0, a is a constant,
(f) y 00 − 4y 0 + 4y = 0.
(g) Consider equation ay 00 + by 0 + cy = 0, where a, b, c are positive constants. Describe the
conditions on a, b, c which guarantee that the characteristic equation will have:
Two distinct real roots, repeated roots, and complex conjugates.

97
11.6 Reduction of order Technique

This technique is very important since it helps us find a second solution independent from
a known one. Let φ1 be a non-zero solution to the homogeneous linear equation

y 00 + a1(x)y 0 + a2(x)y = 0.

Then, a second solution φ2 independent of φ1 can be found as

φ2 (x) = φ1 (x)v(x).

Please, as an exercise, confirm that

Z  
1 − R a1 (x) dx
v(x) = C e dx,
φ21 (x)

where C is a non-zero arbitrary constant. Since we are looking for a second solution we take
C = 1, to get Z 
1 − R a1 (x) dx 
φ2(x) = φ1 (x) e dx.
φ21(x)
Example 44. Find the general solution to the Legendre equation

(1 − x2)y 00 − 2xy 0 + 2y = 0, − 1 < x < 1,

using the fact that φ1 = x is a solution.

Solution 44. First we need to rewrite the equation in the form

2x 0 2
y 00 − 2
y + y = 0, − 1 < x < 1,
1−x 1 − x2
We may try φ2(x) = v(x)φ1(x) which works or, use the formula thus
Z  
1 − R − (1−x
2x
2 ) dx
φ2 (x) = x e dx.
x2
Techniques of integration of rational functions give
Z  Z 
1 R 2x 2 dx  1  1 1 1 + x
e 1−x dx = dx = − + ln ,
x2 x2(1 − x2) x 2 1−x

which gives
x 1 + x
φ2(x) = −1 + ln .
2 1−x
The general solution is then given by
 x  1 + x 
φ = c1 x + c2 −1 + ln .
2 1−x

98
Exercise 31. Reduction of order

Use the reduction of order method to find the solution of the following equation; one solution
of the homogeneous equation is given:

1. y 00 − x2 y 0 + 2
x2
y = 0, φ1 = x.

2. (2x2 + 1)y 00 − 4xy 0 + 4y = 0, φ1 = x.

3. y 00 − x2 y 0 + 2
x2
y = x ln x, φ1 = x.

4. x2 y 00 + xy 0 − y = x2 e−x , φ1 = x.

5. Consider the equation

L(y) := y 00 + a1(x)y 0 + a2(x)y = b(x), (158)

where a1, a2 and b are continuous functions on some interval I.

(a) Suppose φ(x) =6 0 ∀x ∈ I is a solution of

L(y) = 0.

Show that there is a particular solution ψp (x) of

L(y) = b(x),

of the form ψp = uφ, where v = u0 is a particular solution of the first order equation

φ(x)v 0 + [2φ0 (x) + a1 (x)φ(x)]v = b(x). (159)

(b) Use the idea in part (a) to find all solutions of

x2 y 00 − xy 0 + y = x2, for x > 0. (160)

Hint: φ(x) = x in this case.

11.7 Initial value problems for second order equations

We consider the initial value problem

L(y) = 0, y(x0) = α, y 0 (x0) = β (161)

Theorem 11.3 (Existence Theorem). For any real number x0 , and constants α, β, there
exists a solution φ of (161) on R.

99
Proof. Let φ1 (x) and φ2(x) be solutions of L(y) = 0. We show that there are unique
constants c1, c2 such that φ(x) = c1 φ1 (x) + c2φ2 (x) satisfies (161). If φ(x) is a solution then

φ(x0) = c1φ1 (x0) + c2 φ2 (x0) = α,

φ0(x0 ) = c1φ01 (x0) + c2 φ02 (x0) = β. (162)

The linear system (162) will have a unique solution c1 , c2 if W (φ1 , φ2) 6= 0. Indeed

φ1 (x0) φ2(x0 )
= φ1 (x0)φ02 (x0) − φ01 (x0)φ2 (x0).
φ01 (x0) φ02(x0 )

If r1, r2 satisfies the characteristic polynomial, the solutions are given by

φ1 (x0) = er1x0 , φ2(x0 ) = er2 x0 ,

and
φ01 (x0) = r1 er1x0 , φ02(x0 ) = r2 er2 x0 .
If r1 6= r2 , then
W (φ1, φ2)(x0 ) = (r2 − r1 )e(r1+r2 )x0 6= 0.
If r1 = r2 , then
φ1 (x0) = er1x0 , φ2(x0 ) = xer1 x0 ;
φ01 (x0) = r1er1 x0 , φ02(x0 ) = (1 + r1x)er1 x0 ,
and
W (φ1, φ2)(x0 ) = e2r1 x0 6= 0
This implies that φ1(x), and φ2 (x) are independent solutions, and

φ(x) = c1 φ1 (x) + c2φ2 (x)

is a unique solution of (161).

Theorem 11.4 (Uniqueness Theorem). Let α, β be any constants, and let x0 be any real
number. On any interval I 3 x0 , there exists a unique solution for the IVP

L(y) = 0, y(x0) = α, y 0(x0 ) = β.

The proof to this not the main concern of this course. It can be found for instance in [?]

Example 45. solve

y 00 − 5y 0 + 6y = 0, y(0) = 1, y 0 (0) = −1.

Solution 45 The characteristic equation is

p(r) = r2 − 5 + 6

whose roots are r = 2, 3. Thus the general solution is

y(x) = c1 e2x + c2 e3x,

100
where c1 and c2 are arbitrary constants. To determine c1 and c2, we use the conditions given
y(0) = c1 e2.0 + c2 e3.0 = c1 + c2 = 1.

Also
y 0(x) = 2c1 e2x + 3c2 e3x
and
y 0(0) = 2c1 + 3c2 = −1.
Solving the linear algebraic system

c1 + c2 = 1
2c1 + 3c2 = −1

gives us c1 = 4, c2 = −3. Thus

y(x) = 4e2x − 3e3x .
We can see that this is not a difficult task. Let us try the following:

11.8

Exercise 32.

1. y 00 − 4y = 0, y(0) = 1, y 0 (0) = 2.

2. y 00 + 2y 0 + y = 0, y(0) = 1, y 0(0) = 0.

3. y 00 + 2y 0 + y = 0, y(0) = 1, y 0(0) = 0.

4. Given that y = x is a solution of

(1 − x2 )y 00 − 2xy 0 = 2y, |x| < 1.

Find a second solution. Obtain the solution satisfying the initial condition y(0) = y 0(0) = 1.

11.9 The nonhomogeneous equation of order two

We seek a particular solution to differential equation L(y) = y 00 + a1y 0 + a2y = b(x), where
b is continuous on I.

Suppose ψp is a particular solution to L(y) = b(x); that is, L(ψp) = b(x). Let ψ be any
other such solution, then

L(ψ − ψp) = L(ψ) − L(ψp ) = b − b = 0

101
on I. This shows that ψ − ψp is a solution of the homogeneous equation L(y) = 0. Let ψc
be the solution of L(y) = 0, then by uniqueness ψc = ψ − ψp; that is,

ψ = ψc + ψp . (163)

To mean that the general solution ψ, of L(y) = b(x) is given by Equation (163). Here ψc is
the unique solution to the associated homogeneous equation L(y) = 0, to L(y) = b(x). The
function ψc is called a complementary solution while ψp is called a particular solution.

Main result: If φ1, and φ2 are the two linearly independent solutions of L(y) = 0, an
associated homogeneous equation to L(y) = b(x), then, ψc = c1 φ1 + c2 φ2 , where c1 and c2
are any constants. The solution to the nonhomogeneous equation is

ψ = c1 φ1 + c2 φ2 + ψp.

We see that the problem of solving L(y) = b(x) reduces to finding a particular solution
ψp , and the linearly independent solutions φ1, φ2 of L(y) = 0. We have dispensed with the
determination of ψc , we now wish to delve into finding particular solutions.

11.10 Finding a particular solution of L(y) = b(x)

There are many methods available for solving nonhomogeneous equations. We shall be
concerned with two methods, namely: The method of Undetermined coefficients and,
The variation-of-constants formula.

11.10.1 Method of Undetermined Coefficients

This method is based on a guessing technique. That is, we will guess the form of ψp and then
plug it in the equation to find it. However, it works only under the following two conditions:
• Condition 1: The associated homogeneous equations has constant coefficients;
• Condition 2: The non homogeneous term b(x) is of a special form; that is

b(x) = P (x)eαx cos βx, or b(x) = Q(x)eαx sin βx, (164)

where P (x) and Q(x) are polynomial functions. Note we may assume that b(x) is the sum
of such functions.

Assume that the two conditions are satisfied. Consider the equation

a0y 00 + a1y 0 + a2y = b(x),

where a0, a1 , a2 are constants and,

b(x) = Pn (x)eαx cos βx, or b(x) = Pn (x)eαx sin βx, (165)

102
where Pn (x) is a polynomial function of degree n. Then a particular solution ψp is given by
 
ψp = xs Tn (x)eαx cos βx + Rn (x)eαx sin βx , (166)

where

Tn (x) = A0 + A1 x + · · · + An xn , and Rn (x) = B0 + B1 x + · · · + Bn xn , (167)

where Ak , Bk , k = 0, 1, . . . n have to be determined. The power s is equal to 0 if α + iβ is

not a root of the characteristic equation. If α + iβ is a simple root, then s = 1 and s = 2 if
it is a double root.
Remark. : If the nonhomogeneous term b(x) satisfies the following

X
N
b(x) = bk (x),
k=1

where bk (x) are of the forms cited above, then we split the original equation into N equations

a0y 00 + a1 y 0 + a2y = bk (x), k = 1, 2, . . . , N

then find a particular solution ψpk , for each bk (x). A particular solution to the original
equation is given by
XN
ψp = ψpk (x).
k=1

Summary

: Let us summarise the steps to follow in applying this method

(1) First, check that the two conditions are satisfied;

(2) If the equation is given as
X
N
a0y 00 + a1y 0 + a2y = bk (x),
k=1

where

bk (x) = Pn (x)eαk x cos βk x, or bk (x) = Pn (x)eαk x sin βk x,

where Pn (x) is polynomial function of degree n, then split this equation into
N equations
a0 y 00 + a1y 0 + a2 y = bk (x), k = 1, 2, . . . , N ;

(3) Write down the characteristic equation

a0 r2 + a1r + a2 = 0,

and find its roots;

103
 (4) Write down the number αk + iβk . Compare this number to the roots of the
characteristic equation found in the previous step.

(4.1) If αk + iβk is not one of the previous roots, then set s = 0;

(4.2) If αk + iβk is one of the previous roots, then set s = 1;
(4.3) If αk +iβk is equal to both (which means that the characteristic equation
has a double root), set s = 2.
In other words, s measures how many times αk + iβk is a root of the
characteristic equation;

(5) Write down the form of the particular solution

 
ψpk = xs Tn (x)eαx cos βx + Rn (x)eαx sin βx, ,

where

Tn (x) = A0 + A1x + · · · + An xn , and Rn (x) = B0 + B1 x + · · · + Bn xn .

(6) Find the constants Ak and Bk by plugging ψpk into the equation

a0y 00 + a1 y 0 + a2y = bk (x).

(7) Once all the particular solutions ψpk are found, then the particular solution
of the original equation is
XN
ψp = φ pk .
k=1

Example 46. Find a particular solution to the equation

y 00 − 3y 0 − 4y = 3e2x + 2 sin x − 8e−x .

Solution 46. Let us follow these steps:

(1) First, we notice that the conditions are satisfied to invoke the method of undetermined
coefficients.
(2) We split the equation into the following three equations:

y 00 − 3y 0 − 4y = 3e2x
y 00 − 3y 0 − 4y = 2 sin x (168)
y 00 − 3y 0 − 4y = −8e−x

(3) The roots of the characteristic equation r2 − 3r0 − 4 = 0 are r = −1 and r = 4.

(4.1) Particular solution to the first equation in Equation(168):

Since α = 2, and β = 0, then α + iβ = 2, which is not one of the roots. Then s = 0.

The particular solution is given as

ψ1 = Ae2x.

104
 If we plug it into the first equation in Equation(168), we get

4Ae2x − 6Ae2x − 4Ae2x = 3e2x,

which implies A = − 12 ; that is,

1
ψ1 = − e2x.
2

(4.2) Particular solution to the second equation in Equation(168):

Since α = 0, and β = 1, then α + iβ = i, which is not one of the roots. Then s = 0.

The particular solution is given as

ψ2 = A cos x + B sin x.

If we plug it into the second equation in Equation(168), we get

(−A cos x − B sin x) − 3(−A sin x + B cos x) − 4(A cos x + B sin x),

which implies n
−5A − 3B = 0
3A − 5B = 2
3 5
Easy calculations give A = 17
, and B = − 17 ; that is

3 5
ψ2 = cos x − sin x.
17 17

(4.3) Particular solution to the third equation in Equation(168):

Since α = −1, and β = 0, then α + iβ = −1, which is one of the roots. Then s = 1.

The particular solution is given as

ψ3 = x1 Ae−x.

If we plug it into the third equation in Equation(168), we get

A(x − 2)e−x − 3A(−x + 1)e−x − 4Axe−x = −8ex ,

which implies A = 85 ; that is,

8
ψ3 = xe−x .
5

(5) A particular solution to the original equation is the sum of ψ1, ψ2 , ψ3;
1 3 5 8
ψp = − e2x + cos x − sin x + xe−x .
2 17 17 5

105
Exercise 33. 11.10.2 Exrecise

Find the general solution of the differential equations below by the method of undetermined
coefficients:
(a) y 00 + 4y 0 + 4y = 4x2 + 6ex
(b) y 00 + 2y 0 + 5y = 12ex − 34 sin 2x
(c) y 00 − 3y 0 + 2y = 2x2 + 3e2x
(d) y 00 − 3y 0 + 2y = xe2x + sin x
(e) y 00 + 4y 0 + 4y = 3xe−2x
(f ) y 00 + y = 2 sin x sin 2x
(g) 4y 00 − y = ex
(h) 6y 00 + 5y 0 − 6y = x
(i) y 00 + 2y 0 + y = e−x , y(0) = y 0(0) = 1.

11.10.3 Method of Variation of constants

This method has no prior conditions to be satisfied. Therefore, it may sound more general
than the previous method. We will see that this method depends on integration while the
previous one is purely algebraic which, for some at least, is an advantage!

Every solution ψc , of L(y) = 0, is in span{φ1 , φ2 }. Such a function cannot satisfy

L(y) = b(x) 6= 0. However, assume that c1 and c2 are functions u1(x), u2 (x) (not necessarily
constants) on I, and ask, whether there is a function

ψ p = u1 φ 1 + u2 φ 2 (169)

such that L(ψp ) = b(x). If there are such u1, u2 , then what should they satisfy? The
procedure of determining such u1 and u2 is what is referred to as the variation-of-constants
method.

We argue as follows: Suppose there is a ψp as given in Equation (169), then the problem
is to find the variables u1 (x) and u2(x).

Procedure:

L(u1φ1 + u2φ2 ) = L(u1φ1 ) + L(u2 φ2)

= u1L(φ1 ) + u2 L(φ2) + (u001 φ1 + u002 φ2) + 2(u01φ01 + u02φ02 )
+a1(u01φ1 + u02 φ2)
= 0 + (u001 φ1 + u002 φ2) + 2(u01 φ01 + u02φ02 ) + a1(u01 φ1 + u02 φ2 )
= (u01φ1 + u02φ2 )0 + (u01φ01 + u02φ02 ) + a1(u01 φ1 + u02 φ2) = b(x)

We notice that if
u01φ1 + u02φ2 = 0, (170)

106
then we must have
u01φ01 + u02φ02 = b(x). (171)
From the above arguments in reverse we see that if we can find two functions u1 , u2 satisfying
(170) and (171), then indeed u1φ1 + u2φ2 will satisfy L(y) = b(x).

Equations(170) and (171) are linear equations in u01, u02 . Since W (φ1, φ2 ) 6= 0 on I, we
can solve for u01 , u02 in (170) and (171), obtaining

(−1)i φk b(x)
u0i = , i, k = 1, 2, i 6= k.
W (φ1, φ2 )
Thus Z x
φ2(s)b(s)
u1 (x) = − ds,
x0 W (φ1, φ2 )(s)
Z x
φ1 (s)b(s)
u2 (x) = ds.
x0 W (φ1, φ2 ))(s)
Then Z x Z x
φ1 (s)b(s) φ2(s)b(s)
ψp = φ2(x) ds − φ1 (x) ds (172)
x0 W (φ1 , φ2)(s) x0 W (φ1, φ2 )(s)
Remark. In case
L(y) = y 00 + a1 y 0 + a2y = b(x),
we may take
φi (x) = eri x , i = 1, 2
where r1 6= r2 are the zeroes of p(r) = r2 + a1r + a2 , and then

W (φ1, φ2 )(x) = (r2 − r1 )e(r1 +r2 )x

Also
φ1 (s)φ2(x) − φ1(x)φ2(s) = e(r1 s+r2 x) − e(r1 x+r2 s) .
Thus every solution ψ of L(y) = b(x) in this case has the form
Z x
r1 x r2 x 1
ψ(x) = c1 e + c2e + [er1(x−s) − er2 (x−s) ]b(s) ds, (173)
r1 − r2 x0

where x0 ∈ R, and c1, c2 are arbitrary constants.

Exercise 34. Find the equivalent of Equation(173) when r1 = r2 .

Summary

Let us summarize the steps to follow in applying this method:

(1) Find the fundamental solution set {φ1(x), φ2(x)} of the associated homoge-
neous equation
L(y) = y 00 + a1(x)y 0 + a2y = 0.

107
 (2) Write down the form of the particular solution

ψp = u1(x)φ1(x) + u2 (x)φ2(x).

(3) Write down the system

    
φ1 φ2 u01 0
= .
φ01 φ02 u02 b(x)

(4) Solve it; that is, find u1 and u2.

(5) Plug u1 and u2 into the equation giving the particular solution.
Example 47. Solve the differential equation y 00 − y 0 − 2y = e−x .

Solution 47. Let L(y) = y 00 − y 0 − 2y = e−x .

Let us follow the steps:
(1) A set of fundamental solutions of L(y) = 0:
The characteristic equation of L(y) = 0 is p(r) = r2 − r − 2, with −1 and 2 as its roots.
Thus φ1 = e−x , φ2 = e2x are the fundamental solutions.
(2) To determine a particular solution ψp , assume that

ψp = u1e−x + u2 e2x.

(3) We have the system     

e−x e2x u01 0
=
−e−x 2xe2x u02 e−x

(4) Solving for u01 and u02, we find that

1 1
u01 = − , u02 = e−3x,
3 3
from which we get
x 1
u1 = − , u2 = − e−3x.
3 9
x 1
ψp (x) = − e−x − e−x
3 9
We note that − 19 e−x is a solution of L(y) = 0, so we take
x
ψp(x) = − e−x
3
as a simpler particular solution of the inhomogeneous equation.
(5) Thus
x
ψ = − e−x + c1e−x + c2 e2x.
3
Try this exercise
Exercise 35. Do problems in the previous exercise.

108
11.11 Euler-Cauchy Equations

An Euler-Cauchy equation is
x2 y 00 + bxy 0 + cy = 0, (174)
where b and c are constant numbers. Let us consider the change of variable x = et. Then we
have
dy dy d2 y  d2 y dy  −2t
= e−t , = − e .
dx dt dx2 dt2 dt
Thus with this substitution (174) reduces to the new equation
d2 y dy
2
− (b − 1) + cy = 0.
dt dt
We recognize a second order differential equation with constant coefficients. Therefore, we
use the previous section to solve it. We summarise below all the cases:

(1) Write down the characteristic equation

r2 + (b − 1)r + c = 0.

(2) If the roots r1 and r2 are distinct real numbers, then the general solution of Equation(174)
is given by
y(x) = c1|x|r1 + c2 |x|r2 .
(3) If the roots r1 and r2 are equal real numbers; r1 = r2, then the general solution of
Equation(174) is given by
 
y(x) = c1 + c2 ln |x| |x|r1 .

(4) If the roots r1 and r2 are complex conjugates, then the general solution of Equation(174)
is given by  
y(x) = c1 cos(β ln |x|) + c2 sin(ln β|x|) |x|α,
where p
(b − 1) 4c − (b − 1)2
α=− and β= .
2 2
Example 48. Find the general solution to
xy 00 − xy 0 + y = 0.

Solution 48. First we recognize that the equation is an Euler-Cauchy equation, with
b = −1 and c = 1.
The characteristic equation is r2 − 2r + 1 = 0, with 1 as a double root.
Thus the solution is of the form
 
y(x) = c1 + c2 ln |x| |x|.
Exercise 36. Solve the Euler-Cauchy equations below:
(a)x2 y 00 + xy 0 − 4y = x2
(b)x2y 00 − xy 0 − 3y = x2 ln x
(c)x2y 00 + xy 0 − y = x2e−x
(d)x2y 00 − xy 0 + y = 2 ln x

109
Exercise 37. Mixed Grill

Question 1. Use the method of undetermined coefficients‘to solve the following:

(a) y 00 − 5y 0 + 6y = 2ex
(b) y 00 + y 0 − 2y = 2x − 40 cos 2x
(c) y 00 − y = −2x2 + 5 + 2ex
(d) y 00 + 2y 0 − 3y = 3x2 ex + e2x + x sin x + 2 + 3x
(e) y 00 − y 0 − 2y = e3x cos 2x.

Question 2. Solve the following equations:

(a) y 00 + y = 10e2x ; y(0) = y 0 (0) = 0
(b) y 00 − 4y = 2 − 8x; y(0) = 0, y 0(0) = 5
(c) y 00 + 3y 0 = −18x, y(0) = 0, y 0 (0) = 5.
(d) y 00 + 4y 0 + 5y = 10e−3x , y(0) = 4, y 0(0) = 0

Question 3. Use the method of variation of constants to solve:

(a) y 00 − y = ex .
(b) y 00 + y = csc x
(c) y 00 − 3y 0 + 2y = 1+e1−x
(d)y 00 − 2y 0 + y = x1 ex , x > 0
(e) y 00 + y = tan x + 3x − 1.

Question 4. In the following equations, a non-trivial solution φ1 is given. Find the second
linearly independent solution φ2.
(a) x2y 00 + 6xy 0 + 6y = 0, x > 0, φ1 = x−2 .
(b) x2 y 00 + (2x2 − x)y 0 − 2xy = 0, x > 0, φ1 = e−2x.
(c) (x − 1)y 00 − xy 0 + y = 0, φ1 = ex
(d) xy 00 + (1 − 2x)y 0 + (x − 1)y = 0, x > 0, φ1 = ex .
(e) Verify that φ1 = ex is a solution of

(x − 1)y 00 − xy 0 + y = 0.

Use this fact to find the general solution of

(x − 1)y 00 − xy 0 + y = 1.

110
12 Lecture 12: Higher order Linear Equations

12.1 Introduction

What we have done for second order equations can easily be carried over to higher equations.
We shall thus gloss over most facts in this chapter with minimal details where necessary.

Objectives

At the end of this lecture we should be able:

• To solve homogeneous linear equations,

• Determine Linear Independence of Solutions by use of the Wronskian,
• To use the method of reduction of order in solving some differential equations ,
• To solve homogeneous Equations with Constant Coefficients,
• To solve non-Homogeneous Linear Equations by:
◦ The method of Undetermined Coefficients,
◦ The method of Variation of Parameters,
• To solve Euler-Cauchy Equations.

12.2 Linear Differential Equations: Basic Ideas

The general linear differential equation of order n is an equation that can be written as
X
n
pi (x)y (n−i) = d(x), (175)
i=0

where pi (x), d(x), i = 0, 1 . . . , n are continuous functions on some interval I and, y (n−i) is
the (n − 1) − th derivative of y with respect to x. Points p0 (x) = 0 are called singular points,
and often the equation requires special consideration at such points. In this Lecture, we
assume that p0 (x) 6= 0. By dividing through Equation(175) by p0 (x) 6= 0 we get:

X
n
ai (x)y (n−i) = b(x), (176)
i=0

where
pi (x) d(x)
ai (x) = , b(x) = , i = 0, 1, 2, . . . , n.
p0 (x) p0 (x)

111
For purpose of notational brevity, we designate the left side of Equation (176) by L(y). Thus
(176) becomes
L(y) = b(x). (177)

12.1 Definition. (i) If b(x) = 0 for all x ∈ I, then L(y) = 0 is a homogeneous linear
differential equation, whereas if b(x) 6= 0 for some x ∈ I, then L(y) = b(x) is a nonho-
mogeneous linear differential equation. To a nonhomogeneous Equation(176), we can
associate the so called associated homogeneous equation.
(ii)If ai (x) = ai , i = 1, 2, . . . , n are constants then we refer to (176) as a linear differen-
tial equation with constant coefficients otherwise, it is a linear differential equation with
variable coefficients.

It is instructive to understand L as a differential operator which takes each function φ,

which has n derivatives on I, into the function L(φ) on I whose value at x is given by

X
n
ai(x)φ(n−i) (x) = b(x).
i=0

Thus a solution of (176) on I is a function φ on I having n derivatives there, and is such

that L(φ) = b. If φ1 (x), and φ2 (x) are two solutions of L(y) = 0, then it is not difficult to
see that for any constants c1 and c2

φ(x) = c1 φ1 (x) + c2φ2 (x)

is also a solution. Since φ1 (x) and φ2 (x) are solutions of L(y) = 0, it follows that L(φ1 (x)) = 0
and L(φ2 (x)) = 0. The operator L is linear. Thus

L(c1φ1 (x) + c2φ2 (x)) = L(c1φ1 (x)) + L(c2 φ2(x))

= c1L(φ1 (x)) + c2L(φ2 (x)) = 0.

This implies that c1φ1 (x) + c2φ2 (x) is a solution of L(y) = 0.

In general if φi (x), i = 1, . . . , n are solutions of L(y) = 0, so is

X
n
ci φi (x), (178)
i=1

where ci are constants. This can readily be shown as follows. Since L is linear
X
n X
n
L( ci φi (x)) = ci L(φi (x)) = 0.
i=1 i=1

From the preceeding results, we have

Theorem 12.1. Any linear combination of solutions of a homogeneous linear differential

equation is also a solution.

112
12.3 Linear Independence

Given the functions φi (x), i = 1, . . . , n, if constants ci , i = 1, . . . , n, not all zero exist such
that
X n
ci φi (x) = 0 (179)
i=1

for x ∈ I, then φi (x)’s are said to be linearly dependent. If no such ci ’s exist; that is, ci = 0
for all i, then the φi (x)’s are said to be linearly independent.

12.3.1 The Wronskian

We shall give sufficient conditions under which n functions φ ∈ C n−1 (I, C) are linearly
independent.

Let
X
n
ci φi (x) = 0. (180)
i=1

Differentiate (180) successively (n − 1) times, to obtain

X
n
ci φi (x) = 0,
i=1
X
n
ci φ0i (x) = 0,
i=1
..
.
X
n
(n−1)
ci φi (x) = 0. (181)
i=1

This is equivalent to
Φc = 0, (182)
(i−1) n
where Φ = [φj ]i,j=1 , c = [c1, c2 , . . . , cn ]T .

The function W (φ1 , . . . , φn ) defined by W (φ1 , . . . , φn ) ≡ detΦ is called the Wronskian

of the n functions involved. Clearly W (φ1, . . . , φn ) 6= 0 implies that φ1 , . . . , φn are linearly
independent. This is a consequence of the properties of determinants. If W (φ1 , . . . , φn ) = 0,
then, φ1, . . . , φn are linearly dependent.

Note : The nonvanishing of the Wronskian is a sufficient condition for the n functions to be
linearly independent on the interval under consideration. However, this is not a necessary
condition for linear independence. W may be zero and yet the functions be linearly
independent. This will be shown shortly by a specific example.

Example Consider 
1 + x3, if x ≤ 0,
φ1(x) =
1 if x ≥ 0.

113
 
1, if x ≤ 0,
φ2(x) =
1 = x3 if x ≥ 0.
φ3(x) = 3 + x3 ∀x
For x ≤ 0.
1 + x3 1 3 + x3
W (φ1, φ2 , φ3)(x) = 3x2 0 3 + 3x2 = 0
6x 0 6x
While upon solving
c1 φ1 + c1φ2 + c2 φ2 = 0 (183)
for c1, c2 , c3 yields c1 + c2 + 3c3 = 0, c1 + c3 = 0.
For x ≥ 0.
1 1 + x3 3 + x3
W (φ1, φ2 , φ3)(x) = 0 3x2 3x2 = 0
0 6x 6x
While upon solving for c1 , c2 , c3 in (183) yields c1 = c2 = c3 = 0.

If the n function are solutions of a linear equation L(y) = 0, the situation is simplified
by
Theorem 12.2. If the functions φ1 , . . . , φn defined on I are solutions of L(y) = 0, then
W (φ1, . . . , φn ) 6= 0 is a necessary and sufficient condition for their linear independence.
Theorem 12.3. Let φ1, . . . , φn be n solutions of L(y) = 0 on an interval I 3 x0. Then
W (φ1, . . . , φn )(x) = W (φ1 , . . . , φn )(x0 )e−a1 (x−x0 ) (184)
Exercise 38. Prove the Theorem 12.3.

12.4 General Solution of nth order Homogeneous Linear Equa-

tions With Constant Coefficients

Consider the nth-order linear equation with constant coefficients

X
n
L(y) = ak y (n−k) = 0, (185)
k=0

where an = 1.

In order to generate n linearly independent solutions, we need to perform the following:

(1) As in the case of n = 2 assume a solution of the form erx 6= 0. Substituting in L(y) = 0,
we see that
L(erx ) = p(r)erx ,
where
X
n
p(r) = ak rn−k = 0 (186)
k=0

114
 with ak ’s as defined in Equation (185), is the characteristic equation of L(y) = 0.
Then, look for the roots. These roots will either be simple or multiple. Let us show
how they generate independent solutions of (185).

(2) First Case: simple root

Let r be a simple root of the characteristic equation.
(2.1) If r is a real number, then it generates the solution erx ;
(2.2) If r = α + iβ is a complex root, then since the coefficients of the characteristic
equation are real, r = α − iβ is also a root. The two roots generate two solutions:

eαx cos βx, eαx sin βx;

(3) Second case: Multiple root

Let r be a root of the characteristic equation with multiplicity m. If r is a real number,
then generate the m independent solutions of the form:

xk erx, k = 0, 1, . . . , m.

If r = α + iβ occurs with multiplicity m. We shall have 2m independent solutions

given thus:
xk eαx cos βx, xk eαx sin βx, k = 0, 1 . . . m.

Using properties of roots of polynomial equations, we will generate n linearly independent

solutions {φ1, . . . , φn }. Hence the general solution to (185) is given by

φ = c1 φ1 + · · · + φn ,

where ck , k = 1, . . . cn are arbitrary constants.

Therefore, the real problem in solving (185) has to do with finding roots of the associated
characteristic polynomial. So we have to be experts at finding roots of some simple
polynomials.

Example 49. Solve L(y) = y 000 − 3y 0 + 2y = 0.

Solution 49. The characteristic equation for L(y) = 0 is given by

p(r) = r3 − 3r + 2.

Its roots are r = 1, 1, −2. Thus the general solution is

φ(x) = (c1 + c2 x)ex + c3 e−2x ,

where c1 , c2 and c3 are arbitrary constants.

Exercise 39. Find the general solution of

y (4) + y = 0.

115
 An initial value problem(IVP) of L(y) = 0 is given by

L(y) = 0, y (k−1)(x0 ) = αk , k = 1, 2, . . . , n, (187)

for some x0 ∈ I. The IVP in Equation (187) has only one solution.
Example 50. Solve the equation L(y) = y 000 − 3ay 00 + 3a2 y 0 − a3y = 0, where a is a constant.

Solution 50. The characteristic equation is given by

p(r) = (r − a)3.

We take
φ1 (x) = eax, φ2 (x) = xeax, φ3 (x) = x2 eax.
y(x) = c1 φ1 + c2 φ2 + c3φ3 , with c1, c2, c3 are constants.

Exercise 40. 1. Find the general solution of the following:

(a) y (4) + 8y 00 + 16y = 0
(b) y 000 + y = 0

dk
(c) (D3 +3D2 +3D +1)y = 0, where Dk :=
dxk
(d) (D5 − D3 )y = 0
(e) (D4 +5D2 +3D −9)y = 0

2. Consider the equation

y 000 − 4y 0 = 0.
(a) Compute three linearly independent solutions.
(b) Compute the Wronskian of the solutions found in (a).
(c)Find that solution φ satisfying

φ(0) = 0, φ0(0) = 1, φ00(0) = 0.

3. Consider the equation

y (5) − y (4) − y 0 + y = 0;

(a) Compute all its solutions

(b) Find that solution φ that satisfies

φ(0) = 1, φ0 (0) = φ00(0) = φ000(0) = φ(4)(0) = 0.

116
12.5 Linear Nonhomogeneous Equations of order n With Con-
stant Coefficients

We consider an equation of the form

L(y) = y (n) + an−1) y (n−1) + · · · + · · · + a1 y 0 + a0y = b(x),

where ak , k = 0, . . . (n − 1) are constants. Our main interest is to determine solutions of

this equation. The solution is of the form ψ = ψc + ψp , where L(ψc ) = 0 and L(ψp) = b(x).
Finding ψc has already been handled. We now see how ψp may be determined.

This will be handled in a manner similar to that for n = 2. That is, by the methods of
Undetermined Coefficients and Variation-of-constants formula. These two methods
are still valid in the general case, but the second one is a little involved.

12.5.1 Method of Undetermined Coefficients or

Guessing Method

As for the second order case, we have to satisfy the two conditions. Thanks God, the first
one is already fixed as we have assumed that our equation has to have constant coefficients.
The second condition has to do with the nonhomogeneous term b(x). Indeed, in order to
use the undetermined coefficients method, b(x) should be one of the elementary forms
b(x) = P (x)eαx cos βx, or b(x) = Q(x)eαx sin βx, (188)

where P (x) and Q(x) are polynomial functions. Note we may assume that b(x) is the sum
of such functions. In order to guess the form of particular solutions we follow these steps:

(1) Write down the characteristic equation

an rn + an−1 r(n−1) + · · · + a1 r + a0 = 0

and find its roots and (especially) their multiplicity. Note that it will help
strongly if we factorise this equation. This way we get the roots and their
multiplicity;
(2) If the equation is given as
X
N
(n) (n−1) 0
an y + an−1 y + · · · + a1y + a0y = bk (x),
k=1

where
bk (x) = Pn (x)eαk x cos βk x, or bk (x) = Pn (x)eαk x sin βk x,

where Pn (x) is polynomial function of degree n, then split this equation into
N equations

an y (n) + an−1 y (n−1) + · · · + a1 y 0 + a0y = bk (x), k = 1, 2, . . . , N ;

117
 (3) Write down the number αk + iβk . Compare this number to the roots of the
characteristic equation found in the previous step 1.

(4.1) If αk + iβk is not one of the previous roots, then set s = 0;

(4.2) If αk + iβk is one of the previous roots, then s is it multiplicity;

(5) Write down the guessed form for the particular solution
 
φp = xs Tn (x)eαx cos βx + Rn (x)eαx sin βx, ,

where

Tn (x) = A0 + A1x + · · · + An xn , and Rn (x) = B0 + B1 x + · · · + Bn xn .

Note the degree of (T )= degree of (R)= degree of (P ).

(6) Find the constants Ak and Bk by plugging φp into the equation

an y n + an−1 y (n−1) + · · · + a1y 0 + a0 y = bk (x), k = 1, 2, . . . , N ;

(7) Once all the particular solutions ψk are found, then the particular solution
of the original equation is
X
N
ψp = ψk .
k=1

Example 51. Find a particular solution to the equation

y 000 − 4y 0 = x + 3 cos x.

Solution 51. Let us follow two steps:

(1) First, we notice that the conditions are satisfied to invoke the method of undetermined
coefficients.
(2) We split the equation into the following three equations:

y 000 − 4y 0 = x
y 000 − 4y 0 = 3 cos x (189)

(3) The roots of the characteristic equation r3 − 4r = r(r2 − 4) = 0 are r = 0, r = 2, r = −2.

(4.1) Particular solution to the first equation in Equation(189):

Since α + iβ = 0, which is a simple the root. Then s = 1.

The particular solution is given as

ψ1 = x1 (Ax + B).

118
 If we plug it into the first equation in Equation(189), we get
1
A = − , B = 0;
8
that is,
1
ψ 1 = − x2 ;
8

(4.2) Particular solution to the second equation in Equation(189):

We have α + iβ = i, which is not one of the roots. Then s = 0.

The particular solution is given as

ψ2 = x0 (A cos x + B sin x).

If we plug it into the second equation in Equation(189), we get

3
A = 0, B=− .
5
Therefore, we have
3
ψ2 = − sin x.
5

(5) A particular solution to the original equation is

1 3
ψp = − x2 − sin x.
8 5

12.5.2 Method of Variation of Parameters

This method is interesting whenever the previous method does not apply; when b(x) is not
of the desired form. The idea is analogous to what we did for second order linear equations
except that, in that case, we were dealing with a small system and here we may be dealing
with a bigger one. Let us describe the general case( constant coefficients or not).

Consider the equation

y (n) + an−1 (x)y (n−1) + · · · + a1 (x)y 0 + a0(x)y = b(x). (190)

Suppose that the set of independent solutions {φ1 , . . . , φn } of the associated homogeneous
equation is known. Then we try to find n functions u1, . . . , un so that the function

ψp = u1φ1 +, . . . , un φn

is a solution. By a procedure similar to that for n = 2, we take

(i)
u01φ1 + . . . , u0n φ(i)
n = 0, i = 0, . . . , (n − 2)

119
 (n−1)
u01 φ1 + · · · + u0n φ(n−1)
n = b(x),
and then solve for u01, . . . , u0n . This can be done readily by Cramers rule, and we obtain
Wk (x)b(x)
u0k = , k = 1, . . . , n
W (φ1, . . . , φn )(x)
where Wk is the determinant obtained from W (φ1, . . . , φn )(x) by replacing the k−th column;
(n−1)
that is, φk , φ1k , . . . , φk by en = [0, 0, . . . , 1], an n-dimensional vector with zero’s all over
except at the n-th entry. If x0 ∈ I, then
Z x
Wk (s)b(s)
uk = ds, k = 1, . . . , n
x0 W (φ1 , . . . , φn )(s)b(s)

and Z
X
n x
Wk (s)b(s)
ψp = φk (x) ds. (191)
x0 W (φ1 , . . . , φn )(s)b(s)
k=1

Example 52. Solve the IVP

L(y) = y 000 + y 00 + y 0 + y = 1, ψ(0) = 0, ψ 0(0) = 1, ψ 00(0) = 0.

Solution 52. The solutions of L(y) = 0, are

φ1 = cos x, φ2 = sin x, φ3 = e−x .
To obtain a particular solution ψp, we assume a solution of the form
ψp = u1 cos x + u2 sin x + u3e−x
where the function u1(x), u2 (x), u3 (x) are to be determined.
W (φ1 , φ2, φ3 )(x) = e−(x−0)W (φ1 , φ2, φ3 )(0) = 2e−x .
1 1
u01(x) = − (cos x + sin x); u1(x) = − (sin x − cos x).
2 2
Similarly
1 1
u02(x) = − (cos x − sin x); u1 (x) = − (sin x + cos x).
2 2
1 1 x
u03(x) = ex; u3(x) = e .
2 2
Therefore
1 1 1
ψp = (cos x − sin x) cos x + (sin x + cos x) sin x + = 1
2 2 2
−x
⇒ ψ(x) = c1 cos x + c2 sin x + c3 e + 1.
Using the initial conditions, we obtain
1
ψ(x) = 1 − (cos x − sin x + e−x ).
2
Exercise 41. Find a particular solution of
π π
y 000 + y 0 = tan x, − <x< .
2 2

120
Exercise 42. Nonhomogeneous

Question 1
(a) D4 y = 6
(b) (D4 − 2D3 + 2D2 − 2D + 1)y = x2 + x + 1
(c) (D5 − 3D4 + 3D3 − D2 )y = x2 + 2x + 3ex (d) (D3 + 4D)y = 4 cot 2x

Question 2
(a) (D5 − D4 − D + 1)y = x2 − 1, y(0) = 1, y 0(0) = y 00(0) = y 000(0) = y (4) = 0.

121
13 Lecture 13: Series Solution

13.1 Introduction

Through out this lecture, we will assume that you are familiar with power series as learned
in Advanced calculus and the concept of radius of convergence of power series.

Objectives:

At the end of this lecture we should be able to do the following:

• Get a power series solution to second order differential equation.

• Shift index of a power series.
• Solve Airy’s Equation.
• Find radius of convergence of a power series

13.2 Power series: Taylor series approach

Consider the example of the second order differential equation

y 00 + 4y = 0, y(0) = 1, y 0(0) = 0. (192)

We already know that the solution is given by y(t) = cos 2t. This function has its Taylor
series
22 t2 24 t4 25 t5
cos 2t = 1 − + − + ···,
2! 4! 6!
and it is valid for all t ∈ R.

Recall the taylor series expansion of y(t) is given by

y(t) = a0 + a1 t + a2 t2 + a3t3 + · · · +,

where
y (n)(0)
an = .
n!
Here y (n)(0) denotes the nth derivative of y at the point t = 0. What does this mean?
To compute the Taylor series for the solution to our differential equation, we just have to
compute its derivatives. Note that the initial condition gives us a head start: y(0) = 1, so
a0 = 1. y 0(0) = 0, so a1 = 0.

122
 We can rewrite the differential equation as
y 00 = −4y,
so in particular
y 00(0) = −4y(0) = −4.1 = −4.
Consequently a2 = − 2!4 = −2. By now, we have figured out that the solution to our differ-
ential equation has as its second degree Taylor polynomial the function
1 − 2t2 .

Next we differentiate y 00 = −4y on both sides with respect to t, to obtain

y 000 = −4y 0 ,
so in particular
y 000(0) = −4y 0 (0) = 0,
yielding a3 = 0. We can continue in this fashion as long as we like: Differentiating
y 000 = −4y 0 ,
yields
y (4) = −4y 00,
in particular
y (4)(0) = −4y 00(0) = −4.(−4) = 16,
so a4 = 16
4!
= 23 . At this point we have figured out that the solution to our differential
equation has as its fourth degree Taylor polynomial the function
2
1 − 2t2 + t4.
3
We expect that this Taylor polynomial is reasonably close to the solution y(t) of the differ-
ential equation, at least close to t = 0. In Figure 7.1, the solution y(t) = cos 2t and, the
power series approximation f (t) = 1 − 2t2 + 23 t4 are shown.

123
 Figure 7.1: Curves for y(t) and f (t).

The solutions do not agree for the entire time shown because we have approximated y(t) by
a finite limit of only three terms! However should the number of terms in f (t) be many, they
should differ by a very small error margin.

The method outlined works also in theory for non-linear differential equations, even
though the computational effort usually becomes prohibitive after the first few steps. Let’s
consider the example
y 00 + sin y = 0, y(0) = 0, y 0(0) = 1.
We try to find the taylor polynomials for the solution, of the form
y(t) = a0 + a1 t + a2 t2 + a3t3 + · · · +,
where
y (n)(0)
an = .
n!
The initial condition yields a0 = 0 and a1 = 1. To find a2 , we rewrite the differential equation
as
y 00 = − sin y,
and plug in t = 0:
y 00(0) = − sin(y(0)) = − sin(0) = 0;
consequently a2 = 0. Next we differentiate y 00 = − sin y, on both sides with respect to t,
to obtain
y 000 = − cos y.y 0,
so in particular
y 000(0) = − cos(y(0)).y 0(0) = − cos(0).1 = −1,
yielding a3 = − 3!1 = − 16 . Note that since we were differentiating with respect to t, we had
to use the chain rule to find the derivative of sin(y(t)).

Let’s continue a little bit longer: We differentiate

y 000 = − cos y.y 0,
to obtain
y (4) = sin(y).(y 0)2 − cos(y).y 00. (193)
(4)
Consequently y (0) = 0, and thus a4 = 0.
Differentiating Equation(193) yields
y (5) = cos(y).(y 0)3 + sin(y).2(y 0)0 y 00 + sin(y).(y 0).y 00 − cos(y).y 00.
2
We see that y (5) = y 00)3 − y 000(0) = 1 − (−1) = 2 and thus a5 = 5!
.

The fifth degree Taylor polynomial approximation to the solution of our differential equa-
tion has the form
f (t) = a0 + a1t + a2t2 + a3t3 + a4t4 + a5 t5
1 2
= t − t3 + t5 .
6 5!

124
We again expect that this Taylor polynomial is reasonably close to the solution y(t) of the
differential equation, at least close to t = 0. In Figure 7.2, we indicate both y(t) and f (t).
We see that they agree close to t = 0.

Figure 7.2: Curves for y(t) and f (t).

See the minor difference between y(t) and f (t). By Maclaurins formula we have in gen-
eral
X
∞
tk
y(t) = y(0) + y (k)(0) .
k!
k=1

The next sections will develop an organized method to find power series solutions for second
order linear differential equations. Here are a couple of examples to practice what we have
learned so far:

Exercise 43.

1. Find the fifth degree Taylor polynomial of the solution to the differential equation

y 00 − 3y = 0, y(0) = 1, y 0(0) = −1.

2. Find the fourth degree Taylor polynomial of the solution to the differential equation

y 00 + 2y 2 = 0, y(0) = 1, y 0(0) = 0.

125
13.3 Series Solutions: Taking Derivatives and Index Shifting

We have assumed familiarity with power series and the concept of the radius of convergence
of a power series. Given a power series
X
∞
y(t) = an tn = a0 + a1 t + a2 t2 + a3t3 + · · · , (194)
n=0

we can find its derivative by differentiating term by term:

X
∞
0
y (t) = nan tn−1 = a1 + 2a2t + 3a3 t2 + 4a4 t3 + · · · , (195)
n=1

Here we used that the derivative of the term an tn equals an ntn−1 . Note that the start of the
summation changed from n = 0 to n = 1, since the constant term a0 has 0 as its derivative.
The second derivative is computed similarly:
X
∞
00
y (t) = n(n − 1)an tn−2 = 2a2 + 6a3 t + 12a4 t2 + · · · , (196)
n=1

Taking the derivative of a power series does not change its radius of convergence, so y(t), y 0(t), y 00(t), . . .
will all have the same radius of convergence.

The rest of this section is devoted to “index shifting ”. Consider the example
Z 5
(x + 1)5 dx. (197)
2

Using a simple substitution u = x + 1, we can rewrite this integral as

Z 6
u5 du. (198)
3

on changing the dummy variable u back to x, we get:

Z 5 Z 5
5
(x + 1) dx = x5 dx. (199)
2 3

The expression (x + 1) is “shifted down”by one unit to x, while the limits of integration
are “shifted up”by one unit from 2 to 3, and 5 to 6. Summation is just a special case of
integration, so an analogous “index shifting”will work:

X
5 X
6
5
(n + 1) = n5 . (200)
2 3

You should convince yourself that both of these expressions are indeed the same, by writing
out explicitly the four terms of each of the two formulas!

126
 Let’s try this for our derivative formulas:
X
∞
0
y (t) = nan tn−1
n=1
X
∞
= (n + 1)an+1 t((n+1)−1)
n=1
X∞
= (n + 1)an+1 tn
n=0

We shifted each occurrence of n in the expression up by one unit, while the limits of
summation were shifted down by one unit, from 1 to 0, and from ∞ to ∞ − 1 = ∞.

You should once again convince yourself that the first and the last formula are indeed
the same, by writing out explicitly the first few terms of each of the two formulas! As a last
example, let’s shift the formula for the second derivative by 2 units:
X
∞
y 00(t) = n(n − 1)an tn−2
n=1
X
∞
= (n + 1)((n + 2) − 1)an+2 t((n+2)−2)
n=1
X∞
= (n + 2)(n + 1)an+2 tn
n=1

13.4 Series Solutions: First Examples

Let us look (again) at the example (192)

y 00 + 4y = 0.

Using other techniques it is not hard to see that the solutions are of the form

y(t) = A sin 2t + B cos 2t,

where A and B are arbitrary constants.

We want to illustrate how to find power series solutions for a second-order linear differ-
ential equation. The generic form of a power series is
X
∞
y(t) = an tn . (201)
n=0

We have to determine the right choice for the coefficients an in order for Equation(201)
to be a solution of Equation(192). As in other techniques for solving differential equations,

127
once we have a “guess ”for the solutions, we plug it into the differential equation. Recall
from the previous section that
X
∞
00
y (t) = n(n − 1)an tn−2 .
n=2

Plugging this information into the differential equation we obtain:

X
∞ X
∞
n−2
n(n − 1)an t +4 an tn = 0.
n=2 n=0

P
Our next goal is to simplify this expression such that only one summation sign ” ”
n−2
remains. The obstacle we encounter is that the powers of both sums are different, t for
n
the first sum and t for the second sum. We make them the same by shifting the index of
the first sum by 2 units to obtain
X
∞ X
∞
n
(n + 2)(n + 1)an+2 t + 4an tn = 0. (202)
n=0 n=0

Now we can combine the two sums in Equation(202) as follows:

∞ 
X 
n n
(n + 2)(n + 1)an+2 t + 4an t = 0; (203)
n=0

and factor out tn to obtain:

∞ 
X 
(n + 2)(n + 1)an+2 tn + 4an tn = 0; (204)
n=0

Next we need a result you probably already know in the case of polynomials: A polynomial
is identically equal to zero if and only if all of its coefficients are equal to zero. This
results also holds true for power series:
Theorem 13.1. A power series is identically equal to zero if and only if all of its coeffi-
cients are equal to zero.

This theorem applies directly to our example: The power series on the left is identically
equal to zero, consequently all of its coefficients are equal to 0:
(n + 2)(n + 1)an+2 + 4an = 0 for all n = 0, 1, 2, . . . (205)
Solving these equations for the “highest index”n + 2, we can rewrite as
4
an+2 = − an for all n = 0, 1, 2, . . . (206)
(n + 2)(n + 1)
These equations are known as the “recurrence relations” of the differential equations.
The recurrence relations contain all the information about the coefficients we need. Recall
that
X∞
y(t) = an tn = a0 + a1 t + a2 t2 + a3t3 + · · · , (207)
n=0

128
in particular y(0) = a0 , and
X
∞
0
y (t) = nan tn−1 = a1 + a1t + 1a2t + 3a3 t2 + · · · , (208)
n=1

in particular y 0 (0) = a1. This means, we can think of the first two coefficients a0 and a1 as
the initial conditions of the differential equation.

How can we evaluate the next coefficient a2? Let us read our recurrence relations for the
case n = 0:
4
a2 = − a0.
1.2
Reading off the recurrence relation for n = 1 yields
4
a3 = − a1.
2.3
Continue ad nauseam:
4 4 4 24
a4 = − a2 = − .(− a0) = a0.
3.4 3.4 1.2 4!
4 4 4 25
a5 = − a3 = − (− a1) = a1.
4.5 4.5 2.3 2.5!
What do we know about the solutions to our differential equation at this point? They look
like this:
X
∞
y(t) = an tn
n=0
= a0 + a1t + a2t2 + a3t3 + a4t4 + a5 t5 + · · ·
4 4
= a0 + a1t − a0 t2 − a1t3 + · · ·
 1.2 2.3 
4 2 24 4
= a0 t − t + t + ···
 1.24 4!
25 5 
3
+a1 t − t + t + ···
 3! 2.5! 
22 2 24 4
= a0 1 − t + t + · · ·
2! 4!
a1  23 3 25 5 
+ 2t − t + t + · · ·
2 1
3! 5!
1 
= a0 1 − (2t) + (2t)4 + · · ·
2
2! 4!
a1  1 1 
+ (2t) − (2t)3 + (2t)5 + · · ·
2 3! 5!
Of course the power series inside the parentheses are the familiar functions cos(2t) and
sin(2t):
a1
y(t) = a0 cos(2t) + sin(2t),
2
so we have found the general solution of the differential equation (with a0 instead of B, and
a1
2
instead of A).

129
 The series solutions method is mainly used to find power series solutions of differential
equations whose solutions can not be written in terms of familiar functions such as polyno-
mials, exponential or trigonometric functions. This means that in general you will not be
able to perform the last few steps of what we just did (less worries!), all we can try to do is
to come up with a general expression for the coefficients of the power series solutions.

As another introductory example, let’s find the solution to the initial value problem
y 00 − 4y 0 + 3y = 0, y(0) = 1, y 0 (0) = 2.
The generic form of a power series is
X
∞
y(t) = an tn .
n=0

We have to determine the right choice for the coefficients (an ). Start by plugging this
“guess”into the differential equation. Recall from the previous section that
X
∞
y 0(t) = nan tn−1 ,
n=1

and
X
∞
00
y (t) = n(n − 1)an tn−2 .
n=2
Plugging this information into the differential equation we obtain:
X
∞ X
∞ X
∞
n(n − 1)an tn−2 − 4 nan tn−1 + 3 an tn = 0.
n=2 n=1 n=0
P
Our next goal is to simplify this expression such that only one summation sign “ ”remains.
The obstacle we encounter is that the powers of the three sums are different: tn−2 for the
first sum, tn−1 for the second sum, and tn for the third. We make them the same by shifting
the index of the first two sums to obtain
X
∞ X
∞ X
∞
(n + 2)(n + 1)an+2 tn − 4 (n + 1)an+1 tn + 3 an tn = 0.
n=0 n=0 n=0

Now we can combine the sums as follows:

X∞  
(n + 2)(n + 1)an+2 tn − 4(n + 1)an+1 tn + 3an tn = 0.
n=0

and factor out tn :

∞ 
X 
(n + 2)(n + 1)an+2 − 4(n + 1)an+1 + 3an tn = 0.
n=0

Since the power series on the left is identically equal to zero, all of its coefficients are equal
to zero:
 
(n + 2)(n + 1)an+2 − 4(n + 1)an+1 + 3an = 0 for all n = 0, 1, 2, 3, 4 . . . .

130
Solving these equations for the “highest index”n + 2, we can rewrite as
4 3
an+2 = an+1 − an for all n = 0, 1, 2, 3, 4 . . . .
(n + 2) (n + 1)(n + 2)

These recurrence relations contain all the information about the coefficients we need. Recall
that y(0) = a0 and y 0(0) = a1, so our initial conditions imply that a0 = 1 and a1 = 2.
Reading off the recurrence relations we can compute the next coefficients:
5 5
a2 = =
2 2!
14 14
a3 = =
6 3!
41 41
a4 = =
24 4!
122 122
a5 = =
120 5!
365 365
a6 = = e.tc.
720 6!
Can you see a pattern evolving for the numerators? Going from an to an+1 we multiply by 3
and subtract 1. With mostly patience, experience and some magic, we can see that the nth
numerator is of the form
3n + 1
.
2
This implies that
3n + 1
an = for all n = 0, 1, 2, 3, . . .
2.n!
This implies that our solution has the power series:
X
∞
3n + 1
y(t) = tn .
n=0
2.n!

We can rewrite this to retrieve the solution in more familiar form:

X
∞
3n + 1
y(t) = tn
n=0
2.n!
X∞
3n n X 1 n
∞
= t + t
n=0
2.n! n=0
2.n!
X∞
(3t)n 1 X tn
∞
= +
n=0
n! 2 n=0 n!
1 3t 1 t
= e + e
2 2
Isn’t Math fun?

131
Exercise 44. 1. Solve the following equations:
(a) y 00 + 3xy 0 + 3y = 0
(b) (1 + x2)y 00 − 4xy 0 + 6y = 0
(c) y 00 + (x − 1)2 y 0 − 4(x − 1)y = 0 Hint change variables to v = x − 1
(d) (1 − x2)y 00 − 6xy 0 − 4y = 0

13.5 Series Solutions: Airy’s Equation

The general form of a homogeneous second order linear differential equation looks as follows:
y 00 + p(t)y 0 + q(t)y = 0.
The series solutions method is used primarily, when the coefficients p(t) or q(t) are non-
constant. One of the easiest examples of such a case is Airy’s Equation
y 00 − ty = 0,
which is used in physics to model the defraction of light.

We want to find power series solutions for this second-order linear differential equation.
The generic form of a power series is
X
∞
y(t) = an tn . (209)
n=0

We have to determine the right choice for the coefficients an .

As in other techniques for solving differential equations, once we have a “guess”for the
solutions, we plug it into the differential equation. Recall that
X
∞
00
y (t) = n(n − 1)an tn−2 .
n=2

Plugging this information into the differential equation we obtain:

X
∞ X
∞
n−2
n(n − 1)an t −t an tn = 0,
n=2 n=0

or equivalently
X
∞ X
∞
n−2
n(n − 1)an t − an tn+1 = 0.
n=2 n=0

Our
P next goal is to simplify this expression such that (basically) only one summation sign
“ ”remains. The obstacle we encounter is that the powers of both sums are different, tn−2
for the first sum and tn+1 for the second sum. We make them the same by shifting the index
of the first sum up by 2 units and the index of the second sum down by one unit to obtain
X
∞ X
∞
(n + 2)(n + 1)an+2 tn − an−1 tn = 0.
n=0 n=1

132
Now we run into the next problem: the second sum starts at n = 1, while the first sum has
one more term and starts at n = 0. We split off the 0th term of the first sum:
X
∞ X
∞
n
(n + 2)(n + 1)an+2 t = 2.1.a2 + (n + 2)(n + 1)an+2 tn = 0.
n=0 n=1

Now we can combine the two sums as follows:

∞ 
X 
2a2 + (n + 2)(n + 1)an+2 tn − an−1 tn = 0,
n=1

and factor out tn : ∞ 

X 
2a2 + (n + 2)(n + 1)an+2 − an−1 tn = 0.
n=1

The power series on the left is identically equal to zero, consequently all of its coefficients
are equal to 0:

2a2 = 0
(n + 2)(n + 1)an+2 − an−1 = 0 for all n = 1, 2, 3, . . .

We can slightly rewrite as


a2 = 0
an−1
an+2 = (n+2)(n+1)
for all n = 1, 2, 3, . . .

These equations are known as the “recurrence relations” of the differential equations.
The recurrence relations permit us to compute all coefficients in terms of a0 and a1.

We already know from the 0th recurrence relation that a2 = 0. Let’s compute a3 by
reading off the recurrence relation for n = 1:
a0
a3 =
2.3
Let us fire on:
a1
a4 =
3.4
a2
a5 = =0
4.5
a3 a0
a6 = =
5.6 (2.3)(5.6)
a4 a1
a7 = =
6.7 (3.4)(6.7)
a5
a8 = =0
7.8
a6 a0
a9 = =
8.9 (2.3)(5.6)(8.9)

The hardest part, as usual, is to recognize the patterns evolving; in this case we have to
consider three cases:

133
1. All the terms a2, a5, a8 , . . . are equal to zero. We can write this in compact form as

a3k+2 = 0 for all k = 1, 2, 3, . . .

2. All the terms a3, a6, a9 , . . . are multiples of a0. We can be more precise:
1
a3k = a0 for all k = 1, 2, 3, . . .
(2.3)(5.6) · · · ((3k − 1).(3k))

(Plug in k = 1, 2, 3, 4 to check that this works!)

3. All the terms a4, a7, a10, . . . are multiples of a1. We can be more precise:
1
a3k+1 = a1 for all k = 1, 2, 3, . . .
(3.4)(6.7) · · · ((3k).(3k + 1))

(Plug in k = 1, 2, 3, 4 to check that this works!)

Thus the general form of the solutions to Airy’s Equation is given by
 X
∞
t3k 
y(t) = a0 1 +
(2.3)(5.6) · · · ((3k − 1).(3k))
k=1
 X∞
t3k+1 
+a1 t + .
k=1
(3.4)(6.7) · · · ((3k).(3k + 1))

Note that, as always, y(0) = a0 and y 0(0) = a1. Thus it is trivial to determine a0 and a1
when you want to solve an initial value problem.

In particular
X
∞
t3k
y1 (t) = 1 +
k=1
(2.3)(5.6) · · · ((3k − 1).(3k))
and
X
∞
t3k+1
y2 (t)t + .
k=1
(3.4)(6.7) · · · ((3k).(3k + 1))
form a fundamental system of solutions for Airy’s Differential Equation.

Below you see a picture of these two solutions. Note that for negative t, the solutions
behave somewhat like the oscillating solutions of y 00 + y = 0, while for positive t, they behave
somewhat like the exponential solutions of the differential equation y 00 − y = 0.

134
 Figure 7.3: Curves for y(t) and f (t).

Note: The solutions of

y 00 + ay = 0
has infinitely many values if a.) and at most one such value if a < 0. Analogously it can be
shown that a solution of the Airy equation has infinitely many positive zeroes. (a > 0), but
at most one negative zero.

In the next section we will investigate what one can say about the radius of convergence
of power series solutions.
Exercise 45. Solve:
y 00 + ty = 0

13.6 The Radius of Convergence of Series Solutions

In the last section we looked at one of the easiest examples of a second-order linear homo-
geneous equation with non-constant coefficients:Airy’s Equation

y 00 − ty = 0,

which is used in physics to model the defraction of light. We found out that

X
∞
t3k
y1 (t) = 1 +
(2.3)(5.6) · · · ((3k − 1).(3k))
k=1

and
X
∞
t3k+1
y2 (t)t + .
k=1
(3.4)(6.7) · · · ((3k).(3k + 1))

135
form a fundamental system of solutions for Airy’s Differential Equation. The natural ques-
tions arise, for which values of t these series converge, and for which values of t these series
solve the differential equation. The first question could be answered by finding the radius
of convergence of the power series, but it turns out that there is an elegant Theorem, due
to Lazarus Fuchs (1833-1902), which solves both of these questions simultaneously.
Fuchs’s Theorem. Consider the differential equation
y 00 + p(t)y 0 + q(t)y = 0
with initial conditions of the form y(0) = y0 and y 0(0) = y00 . Let r >0. If both p(t) and q(t)
have Taylor series, which converge on the interval (−r, r), then the differential equation has
a unique power series solution y(t), which also converges on the interval (−r, r). In other
words, the radius of convergence of the series solution is at least as big as the minimum of
the radii of convergence of p(t) and q(t).

In particular, if both p(t) and q(t) are polynomials, then y(t) solves the differential
equation for all t ∈ R.

Since in the case of Airy’s Equation p(t) = 0 and q(t) = −t are both polynomials, the
fundamental set of solutions y1 (t) and y2 (t) converge and solve Airy’s Equation for all t ∈ R.

Let us look at some other examples:

Hermite’s Equation of order n has the form

y 00 − 2ty 0 + 2ny = 0,
where n is usually a non-negative integer. As in the case of Airy’s Equation, both p(t) = −2t
and q(t) = 2n are polynomials, thus Hermite’s Equation has power series solutions which
converge and solve the differential equation for all t ∈ R .

Legendre’s Equation of order n has the form

(1 − t2)y 00 − 2ty 0 + n(n + 1)y = 0,
where n is a real number.

Be careful! We have to rewrite this equation to be able to apply Fuchs’s Theorem. Let’s
divide by 1 − t2:
2ty 0 n(n + 1)
y 00 − 2
+ y = 0.
1−t 1 − t2
Now the coefficient in front of y 00 is 1 as required.

What is the radius of convergence of the power series representations of

2t n(n + 1)
p(t) = and q(t) = ?
1 − t2 1 − t2
The center as in all our examples will be t = 0. We really have to investigate this question
only for the function
1
f (t) = ,
1 − t2

136
since multiplication by a polynomial( −2t, and n(n + 1), respectively) does not change the
radius of convergence.

The geometric series

1 X ∞
= xn
1 − x n=0
converges when −1 < x < 1. If we substitute x = t2, we obtain the power series representa-
tion we seek:
1 X∞
f (t) = 2
= t2n,
1−t n=0

which will be convergent when −1 < x = t2 < 1, i.e., when −1 < t < 1. Thus both
2t n(n + 1)
p(t) = and q(t) = ?
1 − t2 1 − t2
will converge on the interval (−1, 1). Consequently, by Fuchs’s result, series solutions to
Legendre’s Equation will converge and solve the equation on the interval (−1, 1).

Bessel’s Equation of order n has the form

t2y 00 + ty 0 + (t2 − n2 )y = 0,

where n is a non-negative real number. Once again we have to be careful! Let’s divide by
t2 :
1 (t2 − n2 )
y 00 + y 0 + y = 0.
t t2

Now the coefficient in front of y 00 is 1 as required by Fuchs’s Theorem. The function has
a singularity at t = 0, thus p(t) fails to have a Taylor series with center t = 0. Consequently,
Fuchs’s result does not even guarantee the existence of power series solutions to Bessel’s
equation. As it turns out, Bessel’s Equation does indeed not always have solutions, which
can be written as power series. Nevertheless, there is a method similar to the one presented
here to find the solutions to Bessel’s Equation. If you are interested in Bessel’s Equation,
look up the section on “The Method of Frobenius”in a differential equations or advanced
engineering mathematics textbook. See Coddington[5]

Exercise 46. Question 1.

(a) Find the Hermite polynomials of order 1 and 3.
(b) Find the Hermite polynomials of order 2,4 and 6.
(c) Consider the Hermite Equation of order 5:

y 00 − 2ty 0 + 10y = 0.

Find the solution satisfying the initial condition a0 = 1, a0 = 0.

137
14 Answers

Exercise 1

1 Order 1, inependent variable is x and dependent variable is y.

2 Order 2, inependent variable is x and dependent variable is y.

Exercise 2

1. (a) Nonlinear first oder odinary differential equation.

(b) Linear Second oder ordinary differential equation.
(c) Second oder partial differential equation.
(d) A nonlinear Second oder ordinary differential equation.
(e) Fourth order linear ordinary differential equation.
(f) Second oder third degree ordinary differential equation.
(e) Second oder nonlinear ordinary differential equation.

Exercise 3

t2
1. x = ce 2 , where C is any real constant.
1 5
2. x = Ce 5 t , where C is any real constant.
1
3. x = Ce2t− , where C is any real constant.
2
−t
4. x = 2−Ce , where C is any real constant.
5. x(t) = ln(t+C), where C is any real constant.
1
6. x = − 1 3 where C is any real constant.
3
t +C
t2
7. tan x = +C, where C is any real constant.
2
x2
8. ln(1+t) = x+ +C, where C is any real constant.
2
9. x−tan x = t+C, where C is any real constant.
x2
10. +ln |x| = et+C, where C is any real constant.
2
x5 x 2 t3
11. +3 = +C, where C is any real constant.
5 2 3

138
12. w = Ct, where C is any real constant.
7 1
13. x = e2t −
2 2
h  t2 i−1
14. x = 1 − + 2t .
2
1 t2
15. x = √ = −
π 2
16. x(t) = 0.
1 + t
17. ln = x2 −16
1−t
r
2
18.x(t) = − 4 + ln(t3 + 1).
3
r 1 − 2k √ 
19.θ(t) = `+arctan tan 1 − 4k 2 t , where ` ∈ Z( set of integers)
1 + 2k
t2 π
20.x(t) = tan( + ).
2 4
 − 12
21. r(t) = r0 r02 + (1 − r02)e−2t , r0 = r(0)

Exercise 4

Question 1.: Mixing problem The differential equation to be solved is

dS 1 S 5−S
= − = . (210)
dt 4 2×5 20
By the method of separation of variables, we have:
5
S(t) = (1 − e−0.01t).
2
Therefore;
(a) Amount of salt ≈ 0.238 kg
(b) Amount of salt ≈ 1.580 kg
(c) Amount of salt ≈ 2.494 kg
(d) Amount of salt ≈ 2.50 kg
(e) Amount of salt ≈ 2.50 kg

Question 2.: Cholestrol problem By the method of separation of variables, the solution
to Equation(31) is:
 k2   k2 
C(t) = C0 + E − C0 − C(0) + E e−k1 t . (211)
k1 k1
Therefore;
(a) 231.571 milligrams per deciliter

139
(b) 327.061 milligrams per deciliter
(c) 600 milligrams per deciliter
(d) Using C(0) = 600, E = 100 in Equation(211) we have:
After 1 day 571.45 milligrams per deciliter
After 5 day 481.959 milligrams per deciliter
After a long time 300 milligrams per deciliter
(e) Using k2 = 0.075, E = 400 in Equation(211) we have:
After 1 day 590.484 milligrams per deciliter
After 5 day 560.653 milligrams per deciliter.

Question 3.: Radioactive Decay Clearly, in order to determine Q(t) we need to find
the constant r. This can be done using the half-life T of the material Q. So, we have
Q(T ) = 12 Q0. An easy calculation gives rT =∈ (2). Therefore, if we know T , we can get r
and vice-versa.

Since the half-life is given in days we will measure time in days. Let Q(t) be the amount
present at time t and the amount we are looking for (the initial amount). We know that
Q(t) = Q0 e−rt ,
where r is a constant. We use the half-life T to determine r. Indeed, we have
1 1
r = ln(2) = ln(2).
T 16
Hence, since
Q(30) = 30 = Q0e−r30 ,
we get
30
Q0 = 30er30 = 30e 16 ln(2) = 110.04g.

Question 4: Newton’s Law of Cooling Equation(32) is a first order linear differential

equation. The solution, under the initial condition θ(0) = θ0, is given by
θ(t) = S + (θ0 − S)ekt.
Hence,
θ(t1) − S
= e−k(t1 −t2 ) ,
θ(t2) − S
which implies
 θ(t ) − S 
1
k(t1 − t2) = − ln .
θ(t2) − S
This equation makes it possible to find k if the interval of time t1 −t2 is known and vice-versa.

First we use the observed temperatures of the corpse to find the constant k. We have
1  24 − 16 
k = − ln = 0.159.
2 27 − 16
In order to find the time of death we need to remember that the temperature of a corpse at
time of death is 37◦ (assuming the dead person was not sick!). Then we have
1  37 − 16 
td = − ln = −4.061 hours
k 27 − 16
which means that the death happened around 7:57 P.M.

140
Exercise 5

(a) x2 + y 2 = k 2 , a one-parameter family of circles centered at the origin.

(b) y 2 = 2(x + k), a one-parameter family of parabolas.
Putting y(0) = 4, we get c = 4 and k = 8. The family members through (0, 4) have equations
y = 4e−x and y 2 = 2(x + 8). The slopes of these curves at (0, 4) are −4 and − 14 .
(c) x2 + y 2 = cy with c1 as parameter is the orthogonal family.
(d) xey − ey (y − 2) = c1 , is orthogonal to y = x + ce−x .
Putting y(0) = 3, we have equations y = x + 3e−x and x − y + 2 + e3−y = 0 as the orthogonal
curves through (0, 3).
(e) y 2 − x2 = c2 with c2 as parameter is the orthogonal family.
(f) r = 2c sin θ. The required family is a family of circles having centers on the line θ =
π
2
y(−axis).

Exercise 8

h 3x − t) √ i
1. (a) (x2 + t2) = ct, (b) ln(3x2 t − 2xt2 − t3) + 2 arctan 2 = 3C.
t
(c) x = tect , (d) xv 2 = C(x + 2v)
3 7
(e) ln |10x−y|+ ln |y| = ln c., (f ) y 2 = 2x2 ln x+cx2 .
10 10
(g) (y−x) = C1(y−x)4.
3

r
4 − t3
2. (a) x(t) = − (b) 2(t+2x)+(t+x) ln(t+x) = 0.
3t
1
(c) x4 (3t2 +4x2 ) = 4, (d) xv 2 = (x+2v)
3

Exercise 9 Equations reducible to homogeneous form

1. (x+y−2) = c2 (x−y)3 2. x+3y+2 ln |2−x−y| = c.

8
3. x+3y = 3 ln |x+2y+2|+c 4. 2x+y+ ln |21x+7y−8| = c.
49
5. (x+2y)2−2x−6y = c 6. 4x+8y+5 = c1e4(x−2y)
x − y  1
7. arctan +ln(x2+y 2 −2x+10y+26) = c 8. (x−y−1)3 = c(x+y−3) 2
y+5
1 2(x − 1)
9. (x+3)(y+1) 2 10. ln y− =c
y
2  y−2  (y − 2)2
11. +arctan √ −ln 2+ 2
= c 12. 3x2 +3y 2−3xy−3y+3x+1 = c
x−1 2(x1 ) (x − 1)

141
Exercise 10: Exact differential Equations

(1) x3 y −3x2 +y 2 = c. (2) −t2x2 −4tx = t4 +6t = c.

1 ln t
(3) x = 4 + e− sin t . (4) t2 + 4tx + x2 = C.
2
2 2
(5) v(u v −3u+2v) = c. (6) r2 +2r(sin θ −cos θ) = c.
(7) exy = 2xy 3 +y 2 −3.

Exercise 11: Integrating factor

x2 y 2
(1) 3x4 +4x3 +6x2 y 2 = c. (2) +xy 3 +xy 2 = C.
2
1
(3) y 4 = 4x4 ln |x| + cx4. (4) x4 y + x3y 2 − x4 = C.
4
1 1 2
(5) ln |x| − 2 2 − ln |y| = C. (6) y = cx.
3 3x y 3
(7) x2 −y 2 +xy −1 = cx. (8) xy +ln |x+2y 2 −2y = C.
(9) x2 y 2 + x3 y − x2 y + 2x3 = c. (10) 2x4 y − 3x3 y 2 = c.

Exercise 13: Substitution Suggested by the Equation

1. x+3y−3 ln |x+2y+2| = c 2. 4x sin y = cx4 −1.

3. 5(x+y+c) = 2 ln(15x−10y+1) 4. x3 sin2 y = c(3x+sin y)2
5. 3 tan(6x+C) = 2(9x+4y+1) 6. x+c = tan(x+y)−sec(x+y)
7. cos y tan2 x = tan3 x+C 8. (x+2y−1)2 = 2y+c

Exercise 14: Bernoulli’s Equations


y −4 = −x + 14 e−4x
1. 2. (2t3 − x)2 = cxt6
y=0

t = x2 (6 + ce−t ) 2 1 −2
3. 4. y 3 = (5x2 −2x−3)+c(x−1) 3
x=0 20

c+x
5. y −2 = sec x+tan x
−1 6. y 3 = x(5 − x2)
x=0
(   −1
2
−2x
7. y = ± ce −1
y=0

142
Exercise 17 : The general Linear Equation of first order

1 2 c 3 3
1.(a) y = (1 − e−x C) (b) y= + x −1
2 x 4
x
(c) y = ce−e + 3 (d) y = sec x(C + esin x
2. (a) y = xe− sin x =: φ(x) (b) Follows by direct substitution in φ(x)
3
3.(a) y=√ (b) y = sec t − csc t
t(t − 1)

Exercise 18 Simultaneous Total differential Equations

1. xy+yz+zx = c2 2. 2xz = y 2z 3+cy 2

1 1 nxy x
3. y(x+z) = c(x+y+z) 4. = +c1, z = c2 − log .
x y y−x y
5. lx+my+nz = c1 , x2+y 2 +z 2 = c2. 6. x2 −y 2−2xy = c1, x2−y 2−z 2 = c2 .
7. xyz = c1, x2 +y 2+z 2 = c2 . 8. y = c1 z, x2+y 2+z 2 = c2 z.
9. x+y+z = c1 , xyz = c2 . 10. a(c−a)x2−by 2(b−c) = c1 , (a−b)by 2−(c−a)cz 2 = c2 .
1 1 1
11. (x−y)(1−z) = c1 , (x+y)(1+z) = c2 . 12. + + = c1 , x2+y 2+z 2 = c2 .
x y z

Exercise 21

1.

143
2.

(a) Since we do not know the function f (x), we will only be able to sketch the slope fields.
This will give us an idea about the behavior of the solutions. Therefore, we should be looking
for the critical solutions (given by the roots of f (x) = 0), and the sign of f (x) which will
give the variation of the solutions. Note that we should be careful not to mix between the
graph of f (x) and the graphs of the solutions x(t). So, according to the graph of f(x), the
critical solutions are x = −1, x = 0, and x = 1. Using the sign of f (x), we conclude that
• the solutions located in the region x < −1 are decreasing,
• the solutions located in the region −1 < x < 0 are increasing,
• the solutions located in the region 0 < x < 1 are decreasing,
• the solutions located in the region 1 < x are increasing.
The sketch of the slope fields is given below.

144
(b) Using the slope fields, we sketch the graph of the solution satisfying the initial condition
x(0) = 0.5.

Clearly, we have
lim x(t) = 0
t→∞

(c) Using the slope fields, we sketch the graph of the solution satisfying the initial condition
x(0) = −0.5.

145
Clearly, we have
lim x(t) = 0
t→∞

3. (a)

Five solution curves superimposed on the slope field for dp/dt=2−p

30

25

20

15
p

10

0
0 5 10 15 20 25 30
t

(b)

146
 120/79+2 exp(−79/300 t)
4

−1

−2
−2 0 2 4 6 8 10
t

30
(c) p(t) = 2 , p(0) = 20, p(t) = 30, p(0) = 30 for t < 5
(1 + 12 e− 5 t )
45 45
p(t) = 3 , p(0) = 20, p(t) = 3 , p(0) = 30 for t > 5
(4 − 74 e − 20 t
) (4 − 52 e− 20 t)
(d) For t < 5, we notice the following:
• All solutions with initial condition p(0) = 30 remain there for for all t. This is an equilib-
rium solution.
• Those solutions with p(0) = 20 increase with t until p(t) = 30.
See Figure below:

30/(1+60 exp(−2/5 t))

40

35

30

25

20
p

15

10

0 5 10 15 20
t

For t > 5, we notice the following:

45
• All solutions with initial condition p(0) = 20 approach 4
and remain so for all future
times.
45
• All solutions with initial condition p(0) = 30 approach 4
and remain so for all future

147
times.

Exercise 24

1. (a)

Equlibrium points are: ω = 2π, 2 − π, all are stable.

and ω = π, 0, are unstable.
(b) Initial conditions which lead to stable equlibrium are: ω ∈ (0, π).
Initial conditions which lead to unstable equlibrium are: ω = π, ω = 0
(c)

148
 8
ω

2π
6

ω
2

0 0

−2

−π

−4
−10 0 10 20 30 40 50 60
t

2.
(a) Here is a picture of the phase line (with the slope field):

The equilibrium points are x = 0, x = 1 and x = 3.

The equilibrium point x = 0 is a source, x = 1 is a node, and x = 3 is a sink.

(b) Below you can see the same picture with some solutions:

149
(c)
(i) Since x(1) = 2, the solution will increase over time and eventually approach the sink at
x = 3. Thus
lim x(t) = 3.
t→∞

(ii) Since x = 1 is an equilibrium point, the solution will be constant, in particular

lim x(t) = 1.
t→∞

3. (a) Equilibrium points are: S = 0, M, N .

The equilibrium points 0 and N are sinks. This is so bcause f 0 (S) < 0 at S = 0 and S = N .
The point S = M is a source as f 0 (M) > 0.

(b)

(c) Initial conditions leading to stable equilibrium are S ∈ (M, ∞), S ∈ (0, M)
Initial conditions leading to unstable equilibrium are S = M
(d)

150
4. (a)

4
x 10
1.5

0.5

(b)
f(p)

−0.5

−1

−1.5
−10 0 10 20 30 40 50 60
p

Graph of f (p) = p(p − 10)(50 − p)

151
 5
x 10
4

f(p)
−1

−2

−3

−4

−5

−6
−10 0 10 20 30 40 50 60
p

Graph of f (p) = p(p − 10)2(50 − p)

(c) f (p) = p(p − 10)(50 − p) and f (p) = p(p − 10)2 (50 − p).

5. (a) Increasing in P ∈ (0, 230)

(b) Decreasing in P ∈ (230, ∞)
(c) In equilibrium at P = 0 and p = 230

6. (a) Increasing in y ∈ (−3, 0) and y ∈ (4, ∞)

(b) Decreasing in y ∈ (−∞, 0) and y ∈ (0, 4)
(c) In equilibrium at y = −3, 0, 4

7. (a) L = 0.
(b) (i)The one who knows nothing.
(ii) Yes. Since for any initial condtion the

lim = 1.
t→∞

Indeed L = 1 is a stable equilibrium point.

(c) (i)Brenda. (ii) Juma (iii) They have the same rate.

Exercise 26

1. (a) Long-term population is N if the initial population is nonzero.

(b)Fixed points are given by solving the quadratic equation rp(N − p) − NE = 0. This gives
r
∗ N N 2 EN
p = ± − .
2 4 r

152
 r
∗ N N 2 EN
p = + − is satble.
2 4 r
r
N N 2 EN
p∗ = − − is unstable.
2 4 r

(c)

p
Graph of ṗ = rp(1 − N) − E = f (p, E), E ≥ 0

The equilibrium points move closer until they coalesce and then finally vanish.
(d)

Bifurcation value E0 = N4r .

(e) The population will vanish.

2. Using the graph of (s, f (s))

153
 Graph of (s, f (s)) for Equation(131).

(b) Bifurcation occurs at N = M.

(c) It decreases and tends to zero as N tends to zero.

3. The graph of the function f (S, E) can be obtained by shifting the graph of
S S
f (S) := rS ( 1 − N
)( M −1)

down E-unit along the vertical axis. Clearly we have three cases according to the value of
E and the value of f at the local maximum:
√
101 + 9901
h= .
3
• If 0 < E < f(h), then we have similar behavior as for E = 0

154
 Graph for (s, f (s)) for 0 < E < f(h) for Equation(132).

If E = f (h), we have two critical points:

Graph for (S, f (S)) for E = f (h) for Equation(132).

• If E > f (h), we have only one critical point:

155
 Graph for (S, f (S)) for E > f (h) for Equation(132).

Note that
h
f (h) = h(1 − h) ( 100 −1).
(b) Clearly, the bifurcation is happening when E = f (h).
(c) The population continuously becomes smaller.

Exercise 29

1.

(a) y(x) = c1 sin 3x + c2 cos 3x (b) y(x) = c1 e2x + c2 e−3x

(c) y(t) = c1 cos at + c2 sin at (d) y(x) = c1 e3x + c2 ex
(e) y(x) = c1 e−2x+c2 xe−2x (f ) y(x) = ex (c1 cos 2x+c2 sin 2x
(g) y(x) = c1 cos x + c2 sin x

2.
3 3
(a) W (x) = W (0)e 2 x (b) W (x) = W (0)x 2 (c) W (x) = W (0)e− sin x

Exercise 30

(a) y(x) = e−x (c1 sin x+c2 cos x) (b) y(x) = ex (c1 sin 2x+c2 cos 2x)

156
 √ √
(c) y(x) = e2x(C1 sin( 3x)+C2 cos( 3x)) (d) y(x) = C1 e2x+C2 e−3x
(e) y(x) = c1 sin ax+c2 cos ax (f ) y(x) = C1 e2x +C2 xe2x
(g) The characteristic equation is given by

p(r) = ar2 + br + c = 0.

This is a quadratic equation.

Let r1 and r2 be its roots we have
√
−b ± b2 − 4ac
r1, 2 = ;
2a
We know from algebra the following:
• r1 and r2 are distinct real numbers if b2 − 4ac > 0.
• r1 = r2 are equal real numbers if b2 − 4ac = 0.
• r1 and r2 are complex conjugates if b2 − 4ac < 0

Exercise 31: Reduction of order

√ √
(−1 + i 2x)2(−2C1 + 4iC1 2x + 4C1 x2 + C2 x)
1. y(x) = (−C1+C2 x)x 2. y(x) = √
(i 2 + 2x)2
1  
4 4 3 −x −x 2 2
3. y(x) = (4x ln x−5x +16C1 +16C2 x )/x 4. y(x) = xe +e +C1x +C1+C2 x −C2 /x
16
5.(a)

L(ψp) = L(uφ)
= u(φ00 + a1 φ0 + a2 φ) + u00φ + [2φ0 + a1φ]u0 = b(x)
= u00φ + [2φ0 + a1φ]u0 = b(x)
= v 0φ + [2φ0 + a1φ]v = b(x)

Thus v = u0 is a particular solution of

v 0φ + [2φ0 + a1 φ]v = b(x).

5.(b) x2 + C1 x + C2 x ln x

Exercise 32: IVP

1. y(x) = sinh 2x + cosh 2x 2. y(x) = (1 + x)e−x

1 x + 1  1 x + 1
3. y(x) = C1 x+C2 x ln −1 , y(x) = x− x ln +1
2 1−x 2 1−x

157
Exercise 33: Undertemined Coefficients

1
(a) y(x) = (6x2 − 12x + 9 + 4ex ) + (C1 + C2 x)e−2x
6
  3
(b) e C1 cos 2x + C2 sin 2x + ex + 8 cos 2x − 2 sin 2x
x
2
7
(c) C1 ex +C2 e2x+x2 +3x+ −3e2x +3xe2x
2
 1  1 
(d) C1 ex + C2e2x + e2x 1 − x + x2 + 3 cos x + sin x
2 10
1 
(e) e−2x x3+C1 +C2x
2
1 1 1
(f ) C1 sin x+C2 cos x+ x sin x+ cos x cos 2x− cos 3x
2 2 8
1 1 1
(g) C1 e− 4 x + C2 e 4 x + ex
3
2 3 5 1
(h) C1e 3 x +C2 e− 2 x − − x
36 6
1 
−x 2
(i) y(x) = e x +x+2
2

Exercise 36: Euler-Cauchy

1
(a) y(x) = C1 x2 + C2 x−2 + x2 ln x
4
1 3
(b) y(x) = C1 x−1 + C2 x3 − x2( + ln x)
3 2
(c) y(x) = C1 x + C2 x + e (1 + x−1 )
−1 −x

(d) y(x) = (C1 + C2 ln x)x + 2 ln x + 4.

Exercise 37: Mixed Grill

Question 1.

(a) y(x) = C1 e2x + C2e3x + ex

1
(b) y(x) = C1 e2x +C2e−2x − −x+6 cos 2x−2 sin 2x
2
(c) y(x) = C1e + (C2 + x)ex + 2x2 − 1
−x

4 1 1 1 1 1
(d) y(x) = C1ex − −(60+18x+3x2 )ex + e2x+( − x) cos x+( − x) sin x
5−x 6 5 16 5 5

158
 1
(e) y(x) = C1 ex + C2e2x + e3x sin 2x
6

Question 2.

(a) y(x) = 2(e2x − cos x − 2 sin 2x)

1 1
(b) y(x) = e2x − e−2x + 2x −
2 2
2 −3x
(c) y(x) = 1 + 2x − 3x − e
(d) y(x) = 5e−3x + 13e−2x sin x − e−2x cos 2x

Question 3.
1
(a) y(x) = C1 e−x + C2ex + xex
2
(b) y(x) = (C1 −x) cos x+(C2 +ln | sin x|) sin x
(c) y(x) = C1ex +C2 e2x +(ex +e2x ) ln(1+e−x )
(d) y(x) = (C1 + C2x)ex + x ln xex
(e) y(x) = (C1 +3x) cos x+C2 sin x−ln | sec x+tan x|

Question 4.

(a) φ2(x) = x−3 , y(x) = C1x−2 + C2 x−3

(b) φ2(x) = x−1, y(x) = C1(x−1)+C2e−2x

(c) φ2(x) = x, y(x) = C1x + C2 ex
(d) φ2 (x) = ex. ln x, y(x) = C1 ex. ln x+C2ex
(e) y(x) = 1 + C1 x + C2 ex

Exercise 40

Question 1.

(a) y(x) = (C1 +C2x) cos 2x+(C3 +C4x) sin 2x

(b) y(x) = (C1 + C2x)e−x + C2 ex

(c) y(x) = (C1 + C2 x + C3x2 )e−x
(d) y(x) = C1 + C2 x + C3 x2 + C4 ex + C5e−x
√ √
(e) y(x) = (C1 +C2 x)e−x +C3 e(1+ 3)x
+C4 e(1− 3)x

159
Question 2.

(a) 1, e2x, e−2x,

(b) 16
1
(c) y(x) = sinh 2x
2

Question 3.

(a) ex, xex, e−x , cos x, sin x

5 1 1 1
(b) y(x) = ex − xex + e−x + (sin x−cos x)
8 4 8 4

Exercise 42: Nonhomogeneous

Question 1

(a) y(x) = C1 + C2x + C3x2 + x3

(b) y(x) = (C1 +C2x)e−x +C3 cos x+C4 sin x+7+5x+x2

1 1 4
(c) y(x) = C1 +C2x+(C3 +C4x+C5x2 )ex+ x3ex − x4 − x3−9x2
2 12 3
1 1
(d) y(x) = C1+C2 cos 2x+C3 sin 2x+ ln sin 2x− cos 2x ln(csc 2x−cot 2x)
2 2

Question 2
5 1 1 1
(a) y(x) = ex − xex+ e−x + (sin x−cos x)+1+2x+x3
8 4 8 4

Exercise 44: Series

1.

(a)
h X
∞
(−3)k x2k i h X
∞
(−3)k x2k+1 i
y = a0 1 + + a1 x + valid for all finite x.
2k k! 3.5.7. · · · (2k + 1)
k=1 k=1

(b)
1
y + a0(1 − 3x2 ) + a1 (x − x3); valid for all finite x.
3

160
(c)
X
∞
4(−1)k (x − 1)3k  1 
4
y = a0 + a1 (x − 1) + (x − 1) .
3k k!(3k − 1)(3k − 4) 4
k=0

(d)
X
∞
1
y = a0 [1 + (k + 1)x2k ] + a1 [x − 1 (x − 1)4 .
4
k=1

Exercise 46: Airy Equations

X
∞
(−1)k t3k X
∞
(−1)k t3k+1
y1 (t) = 1 + , y2(t) = t + .
k=1
2.3.5.6 · · · (3k − 1)3k k=1
3.4.6.7 · · · 3k.(3k + 1)

Hermite

2
(a) H1 = t, H3 = t− t2
3
4 8
(b) H2 (t) = 1−2t2 , H4 (t) = 1−4t2 + t4, H6 (t) = 1−6t2 +4t4 − t6
3 15
X∞
2n (−5)(−3)(−1) · · · (2n − 7)
(c) y(t) = t62n
n=0
(2n)!

161
References
[1] Ahsan,Z., Differential Equations and their Applications, Prentice-Hall of India, 2001.

[2] Blanchard,P., Devaney,R.L., Hall, G.R.; Differential Equations, Brooks/ Cole Publish-
ing Company, (1996).

[3] Braun, M., Differential Equations and their applications, Springer-Verlag, Berlin (1993)

[4] Coddington, E.A.,An introduction to Ordinary Differential equations, Prentice-Hall of

India, New Delhi, (1995)
[5] Snedon, I.N., Elements of Partiall Differential Equations, McGraw-Hill Books Company,
Inc. (1975)

162