Lecture Notes on Differential Equations

Fei Ye

Department of Mathematics and Computer Science

Queensborough Community College - CUNY

Version 2021.12.16
Dr. Fei Ye
Department of Mathematics and Computer Science
Queensborough Community College of CUNY
222-05 56th Street, Bayside, NY, 11364
email: [email protected]

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0

International License.
 These lecture notes are based on the open source textbook Elementary Differential Equa-
tions with Boundary Value Problems by William F. Trench, the open source textbook Differen-
tial Equations for Engineers by Daniel An, lecture notes on Ordinary Differential Equations by
Zhixian Zhu and other online resources.
 Contents

1 Introduction 1
1.1 Motivating Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Basic Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2 Separable and Linear First Order Equations 15

2.1 Separable Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2 Linear First Order Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.3 Change of Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

3 Substitution Methods and Exact Equations 33

3.1 Substitution Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.2 Exact Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

4 Integrating Factors and Applications 44

4.1 Integrating Factors for Non-exact Equations . . . . . . . . . . . . . . . . . . . . 44
4.2 Existence and Uniqueness* . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.3 Applications of First Order Differential Equations . . . . . . . . . . . . . . . . . 53
4.A Integrating factor methods for linear first order equations revisited . . . . . . . 72

5 Linear Second Order Equations 75

5.1 Homogeneous Linear Second Order Equations . . . . . . . . . . . . . . . . . . . 75
5.2 Constant Coefficient Homogeneous Equations . . . . . . . . . . . . . . . . . . . 84
5.3 Non-Homogeneous Linear Equations . . . . . . . . . . . . . . . . . . . . . . . . 91

6 Linear Second Order Equations II 96

6.1 The Method of Undetermined Coefficients . . . . . . . . . . . . . . . . . . . . . 96
6.2 Variation of Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110

7 Applications of Linear Second Order Equations 120

7.1 Vibrating Springs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
7.2 Free Vibrations with Damping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
7.3 Forced vibrations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130

8 Series Solutions 135

8.1 Review of Power Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
8.2 Series Solutions Near an Ordinary Point . . . . . . . . . . . . . . . . . . . . . . . 143
8.3 Method of Frobenius and Euler Equations . . . . . . . . . . . . . . . . . . . . . . 151

9 Introduction to Laplace Transforms 159

9.1 Laplace Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
 CONTENTS

9.2 Existence and Additional Properties* . . . . . . . . . . . . . . . . . . . . . . . . . 164

9.3 Inverse Laplace Transforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171

10 Solving Differential Equations using Laplace Transforms 176

10.1 Solving IVP using Laplace Transforms . . . . . . . . . . . . . . . . . . . . . . . . 176
10.2 Convolutions* . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
10.3 Dirac Delta Function* . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180

11 Linear System of Equations 184

11.1 Basics of Linear Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
11.2 Linear Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
11.3 Constant Coefficient Homogeneous Linear Systems . . . . . . . . . . . . . . . . 201
11.4 Variation of Parameters for Nonhomogeneous Linear Systems* . . . . . . . . . 205
11.5 Solving Linear System by Laplace Transform . . . . . . . . . . . . . . . . . . . . 206

12 Bounded Value Problems and Fourier Series 208

12.1 Introduction to Bounded Value Problems . . . . . . . . . . . . . . . . . . . . . . 208
12.2 Eigenvalues and Eigenfunctions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
12.3 Introduction to Fourier Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213

ii Lecture Notes for MA451

 Week 1: Introduction

8/26–9/1

1.1 Motivating Examples

1.1.1 What is a Differential Equation

Definition 1.1
A differential equation is an equation that contains one or more derivatives of an unknown
function.
A differential equation is an ordinary differential equation (ODE) if it involves an unknown
function of only one variable.
♣

In this course, we will consider only ordinary differential equations and simply call them
differential equations.

Example 1.1 Which one is a differential equation?

1. 𝑦 ′ = 𝑦 +𝑥 is a differential equation. The unknown function is 𝑦 depending on the variable

𝑥 and the derivative of 𝑦 is involved in the equation.
2. sin(𝑥) = 1 is not a differential equation. There is no unknown function.
3. 𝑦 = 5 + 𝑥 is not a differential equation. There is a function 𝑦 in the equation but the
derivative is not involved.

 Exercise 1.1 Determine whether the followings are differential equations.

1. 𝑦 ′ = 𝑦.
2. (sin 𝑥)′ = cos 𝑥.
3. 𝑔 ′′ (𝑥) − 2𝑔 ′ (𝑥) + 𝑔(𝑥) = 𝑥 2

Solution

1. 𝑦 ′ = 𝑦 is a differential equation. The unknown function is 𝑦 depending on the variable 𝑥

and the derivative of 𝑦 is involved in the equation.
2. (sin(𝑥))′ = cos 𝑥 is not a differential equation. There is no unknown function.
3. 𝑔 ′′ (𝑥)−2𝑔 ′ (𝑥)+𝑔(𝑥) = 𝑥 2 is a differential equation. The unknown function is 𝑔(𝑥) depending
on the variable 𝑥 and the derivatives of 𝑔(𝑥) are involved in the equation.

■
From the above equations and exercise, you see that ordinary differential equations are
 1.1 Motivating Examples

usually in the form

𝑦 (𝑛) = 𝑓 (𝑥, 𝑦, 𝑦 ′ , … , 𝑦 (𝑛−1) ).

1.1.2 Where to Find Them

In physical and many real life problems, we want to study relations between changing
quantities. To apply mathematical methods to such a problem, we need to formulate the
problem using mathematical concepts and construct a mathematical model to describe it. The
process of developing a mathematical model is known as mathematical modeling.

Comparing rates of change often helps to model relations by sufficiently simple mathe-
matical equations. Those equations often involve functions and their derivatives. They are
called differential equations. The focus of this course is to study how to solve differential
equations.

A good the mathematical model should have, but not limited to, the following qualities:

1. It’s sufficiently simple so that the mathematical problem can be solved.

2. It should fit the actual situation sufficiently well so that it can be used to make predictions
that are verifiable by experimental data.

Now let’s see some examples of mathematical models involving differential equations.
You will learn how to solve the various types of differential equations in those examples later.
d𝑦
In this course, we will use the prime notation 𝑦 ′ and the Leibniz notation interchange-
d𝑡
ably.

Example 1.2 Population Growth and Decay The number 𝑃 of members of a population (peo-
ple in a given country, bacteria in a laboratory culture, etc.) at any given time 𝑡 can be modeled
using differential equations. In most models, it is assumed that the differential equation takes
the form
𝑃 ′ (𝑡) = 𝑎(𝑃)𝑃(𝑡), (1.1)

where 𝑎 is a continuous function of the population 𝑃(𝑡) that represents the relative rate of
change per unit time, known as the growth rate.

In the Malthusian model, the growth rate 𝑎 is assumed to be a constant 𝑟, and the equation
1.1 becomes
𝑃 ′ (𝑡) = 𝑟𝑃(𝑡). (1.2)

From Calculus, we know that the equation 1.2 has a solution 𝑃(𝑡) = 𝑃(0)𝑒 𝑟𝑡 .

The Malthusian model has the limitation. Suppose that we are modeling the population
of a country. Starting from a time 𝑡 = 0, as time goes, the population might either be 0 if 𝑎 < 0
or infinity if 𝑎 > 0 which is not reasonable. Indeed, the population breaks the country’s limit of

2 Lecture Notes for MA451

 1.1 Motivating Examples

resources, the model will no longer be valid. Because of the limitation of space and resources,
the relative population growth rate should decrease as the population increase.

Another model that reflects the above mentioned phenomenon is the Verhulst Model:

𝑃 ′ (𝑡) = 𝑟𝑃(𝑡)(1 − 𝛼𝑃(𝑡)), (1.3)

where 𝑟 is the growth rate and 1/𝛼 is the carrying capacity. As long as 𝑃 is relatively small com-
pared to 1/𝛼, in other words, 𝛼𝑃 is approximately 0, the growth is approximately exponential
′ (𝑡)
because the ratio 𝑃𝑃(𝑡) is approximately 𝑟. However as 𝑃 increases, the growth rate decrease.

Equation 1.3 is known as the logistic equation. It can be re-written as

d d
( ln(𝑃(𝑡)) − ( ln(1 − 𝛼𝑃(𝑡)) = 𝑟.
d𝑡 d𝑡
Integrating both sides implies that the logistic equations has the solution

𝑃(0)
𝑃(𝑡) = .
𝛼𝑃(0) + (1 − 𝛼𝑃(0))𝑒 −𝑟𝑡

1 1
Note that lim 𝑃(𝑡) = 𝛼
and 𝛼
is independent of 𝑃(0).
𝑡→∞

The following figure shows the solutions of the logistic equation for various 𝑃0 .

𝑃(𝑡)

1/𝛼

𝑡
Figure 1.1: Solutions to a logistic equation

Example 1.3 Newton’s Law of Cooling According to Newton’s law of cooling, the temper-
ature of a body changes at a rate directly proportional to the difference in the temperatures
between the temperature of the body and the temperature of its surroundings. If 𝑇𝑚 is the
temperature of the surrounding and 𝑇 = 𝑇 (𝑡) is the temperature of the body at time 𝑡, then

𝑇 ′ (𝑡) = −𝑘(𝑇 (𝑡) − 𝑇𝑚 (𝑡)), (1.4)

3 Lecture Notes for MA451

 1.1 Motivating Examples

where 𝑘 is a positive constant and the minus sign indicates that heat will transfer from hot to
cold objects.

When the surrounding temperature 𝑇𝑚 (𝑡) = 𝑇𝑚 is constant, the equation 1.4 has a solution

𝑇 (𝑡) = 𝑇𝑚 + (𝑇0 − 𝑇𝑚 )𝑒 −𝑘𝑡 ,

where 𝑇0 = 𝑇 (0) is the initial temperature of the body. As you can imagine, the temperature
will be approximately balanced as time goes. A mathematical explanation is that lim 𝑇 (𝑡) = 𝑇𝑚 .
𝑡→∞

The following figure shows typical graphs of the function 𝑇 (𝑡) with various values of 𝑇0
and a fixe 𝑇𝑚 .

𝑇 (𝑡)

𝑇𝑚

𝑡
Figure 1.2: Functions from Newton’s Law of Cooling

Assuming that the surrounding remains at constant temperature seems reasonable in

some situations such as cooling a cup of hot coffee in a room. In this case, the heat transferred
to the room won’t increase its temperature much. However, if we cool down the cup of hot
coffee in a small pot of water, you will find the water temperature increases as time goes.
In this situation, to model the temperature changes more accurate, we must consider the
heat exchanges. Suppose energy is conserved, that is, the total heat in the object and its
surrounding remain constant.

In physics, it is known that the heat transfer is directly proportional the change of tem-
perature. Then we have an extra equation in addition to Equation 1.4:

𝑎(𝑇 (𝑡) − 𝑇0 ) = 𝑎𝑚 (𝑇𝑚 (𝑡) − 𝑇𝑚0 ),

where 𝑇𝑚0 = 𝑇𝑚 (0) is the initial temperature of the surrounding. Solving 𝑇𝑚 (𝑡) and plug in

4 Lecture Notes for MA451

 1.1 Motivating Examples

Equation 1.4, we get

𝑇 ′ (𝑡) = − 𝑘(𝑇 (𝑡) − 𝑇𝑚 (𝑡))

𝑎 𝑎
=−𝑘 1+𝑇 (𝑡) − 𝑇 + 𝑇𝑚0
(( 𝑎𝑚 ) ( 𝑎𝑚 0 ))
𝑎 𝑎𝑇0 + 𝑎𝑚 𝑇𝑚0
=−𝑘 1+ 𝑇 (𝑡) −
( 𝑎𝑚 ) ( 𝑎 + 𝑎𝑚 )

Again, the equation can be rewritten in the form ln(𝐹 (𝑡)) = 𝐶 form (Can you find the 𝐹 (𝑡) and
𝐶?). Using Calculus, you will find that it has a solution

𝑎𝑇0 + 𝑎𝑚 𝑇𝑚0 𝑎𝑚 (𝑇0 − 𝑇𝑚0 ) −𝑘(1+ 𝑎𝑎 )𝑡

𝑇 (𝑡) = + 𝑒 𝑚 .
𝑎 + 𝑎𝑚 𝑎 + 𝑎𝑚

Remark In the above examples, the differential equations can be re-written in the form

𝐹 ′ (𝑡)
= 𝐶, (1.5)
𝐹 (𝑡)

where 𝐹 (𝑡) is a function and 𝐶 is a constant. Note that

𝐹 ′ (𝑡) ′
= ( ln(𝐹 (𝑡))) .
𝐹 (𝑡)

Integrate both sides of Equation 1.5, you will find that 𝐹 (𝑡) = 𝐹 (0)𝑒 𝑐𝑇 .

Example 1.4 Newton’s Second Law of Motion For an object with a constant mass 𝑚, New-
ton’s second law of motion states that the force 𝐹 acting on the object and the instantaneous
acceleration 𝑎 of an object are related by the equation 𝐹 = 𝑚𝑎.

In many applications, there are multiple forces that may act on the object.

Assume that the motion of an object with the mass 𝑚 = 1 is moving along a vertical line
above the surface of the Earth. Let 𝑦 be the displacement of the object from some reference
point above the surface. The following type of forces normally act:

The gravity 𝑔(𝑦) that depends only on the position 𝑦, where 𝑔(𝑦) < 0.
The atmospheric resistance −𝑟(𝑦, 𝑦 ′ )𝑦 ′ that depends on the position and velocity of the
object, where 𝑟 is a nonnegative function. The −𝑦 ′ “outside” of the function is used to
indicate that the resistive force is always in the direction opposite to the velocity 𝑦 ′ .
The force 𝑓 = 𝑓 (𝑡) from other external sources (such as a towline from a helicopter) which
depends only on 𝑡.

5 Lecture Notes for MA451

 1.1 Motivating Examples

In this case, Newton’s second law implies that

𝑦 ′′ = −𝑟(𝑦, 𝑦 ′ )𝑦 ′ − 𝑔(𝑦) + 𝑓 (𝑡),

which is usually rewritten as

𝑦 ′′ + 𝑟(𝑦, 𝑦 ′ )𝑦 ′ + 𝑔(𝑦) = 𝑓 (𝑡).

Since the second order derivative of 𝑦 occurs in this equation and no more higher order
derivative, we say that this equations is a second order differential equation.

Example 1.5 Interacting Species: Competition Let 𝑃 = 𝑃(𝑡) and 𝑄 = 𝑄(𝑡) be the populations of
two species at time 𝑡. Assume that each population would grow exponentially by the Malthu-
sian model if the other did not exist; that is, in the absence of competition we would have

𝑃 ′ = 𝑎𝑃 and 𝑄 ′ = 𝑏𝑄 (1.6)

where 𝑎 and 𝑏 are positive constants.

To model the effect of competition, one way is to assume that the growth rate per indi-
vidual of each population is reduced by an amount proportional to the other population, so
Equation 1.6 is replaced by

𝑃 ′ = 𝑎𝑃 − 𝛼𝑄
𝑄 ′ = −𝛽𝑃 + 𝑏𝑄,

where 𝛼 and 𝛽 are positive constants. The relation between the populations of the competing
species can be described by the following figure. The arrows indicate direction of rates of

Figure 1.3: A model for populations of competing species

6 Lecture Notes for MA451

 1.2 Basic Concepts

change of populations with increasing 𝑡. The dashed line 𝐿 through the origin depends only
on 𝑎, 𝑏, 𝛼 and 𝛽. If (𝑃0 , 𝑄0 ) is above 𝐿, then the species with population 𝑃 will extinct, but if
(𝑃0 , 𝑄0 ) is below 𝐿, the species with population 𝑄 will extinct.

In the example, we are dealing with a homogeneous system of differential equations with
constant coefficients. The slope of the dashed line is related to a eigenvector of the coefficient
matrix of the system. For more information, you may read Section 10.4 of Trench’s book.

1.2 Basic Concepts

1.2.1 What is the Order

Definition 1.2
The order of a differential equation is the order of the highest derivative that it contains.
♣

Example 1.6 What’s the order of the differential equation?

1. 𝑦 ′ − 𝑥 2 = 0 is a first order differential equation.

2. 𝑦 ′ + 𝑥𝑦 2 = 𝑦 3 is also a first order differential equation.
3. 𝑦 ′′ − 𝑥𝑦 ′ + 𝑦 = 𝑥𝑦 2 is a second order differential equation.
4. 𝑥𝑦 (4) + 𝑦 2 = sin 𝑥 is a fourth order differential equation.

 Exercise 1.2 Determine the order of each differential equation.

1. 𝑦 ′ = 𝑥.
2. 𝑦 ′′ ⋅ 𝑦 ′ + 𝑦 = cos 𝑥.

Solution

1. 𝑦 ′ = 𝑥 is a first order differential equation.

2. 𝑦 ′′ ⋅ 𝑦 ′ + 𝑦 = cos 𝑥 is a second order differential equation.

1.2.2 What is a Solution

Definition 1.3
A solution of a differential equation is a function that satisfies the differential equation on
some open interval.
♣

Example 1.7 Is the function a solution?

1. 𝑦 = 𝑒 𝑥 is a solution of 𝑦 ′ − 𝑦 = 0 on the interval (−∞, ∞).

7 Lecture Notes for MA451

 1.2 Basic Concepts

If 𝑦 = 𝑒 𝑥 , then 𝑦 ′ = 𝑒 𝑥 and
𝑦 ′ − 𝑦 = 𝑒 𝑥 − 𝑒 𝑥 = 0.

In fact, for any number 𝑐, 𝑦 = 𝑐𝑒 𝑥 is a solution.

2. 𝑦 = sin(𝑥) is a solution of 𝑦 ′′ + 𝑦 = 0.
If 𝑦 = sin(𝑥), then 𝑦 ′ = cos(𝑥) and 𝑦 ′′ = − sin(𝑥) and

𝑦 ′′ + 𝑦 = − sin(𝑥) + sin(𝑥) = 0.

So 𝑦 is a solution of 𝑦 ′ − 𝑦 = 0 on the interval (−∞, ∞)

3. 𝑦 = sin(𝑥) is not a solution of 𝑦 ′ − 𝑦 = 0.
If 𝑦 = sin(𝑥), then 𝑦 ′ = cos(𝑥) and

𝑦 ′ − 𝑦 = sin(𝑥) − cos(𝑥) ≠ 0.

Example 1.8 Show that for any constants 𝑐1 and 𝑐2 the function

𝑦 = 𝑐1 sin 𝑥 + 𝑐2 cos 𝑥

is a solution of
𝑦 ′′ + 𝑦 = 0

on (−∞, ∞).

Solution

Differentiating function twice yields

𝑦 ′ = 𝑐1 cos 𝑥 − 𝑐2 sin 𝑥,

𝑦 ′′ = −𝑐1 sin 𝑥 − 𝑐2 cos 𝑥.

Therefore 𝑦 is a solution of the differential equation on (−∞, ∞). ■

 Exercise 1.3 Determine whether the following functions are solutions of 𝑦 ′ = 𝑦 2

1. 𝑦 = − 𝑥1 .
2. 𝑦 = 𝑥.
2
Solution If 𝑦 = − 𝑥1 , then 𝑦 ′ = 1
𝑥2
and 𝑦 2 = (− 𝑥1 ) = 1
𝑥2
. So, 𝑦 = − 𝑥1 is a solution.

If 𝑦 = 𝑥, then 𝑦 ′ = 1. But 𝑦 2 = 𝑥 2 ≠ 1 = 𝑦 ′ . Thus, 𝑦 = 𝑥 is not a solution. ■

8 Lecture Notes for MA451

 1.2 Basic Concepts

1.2.3 What is an Initial Value Problem

We’ve seen that a differential equation may have an infinite family of solutions. To speci-
fies one solution, extra constants will be needed. Those constants are initial conditions.
Definition 1.4
An initial value problem (IVP for short) for an 𝑛-th order differential equation requires 𝑦 and
its first 𝑛 − 1 derivatives to have specific values at a point. Such a specific value is called an
initial condition.
A solution of an initial value problem is a function which satisfies both the differential
equation and initial conditions.
♣

Example 1.9 The following is an initial value problem for a first order differential equation.
⎧
⎪
⎪ 𝑦 ′ = 𝑦.
⎨
⎪
⎪ 𝑦(0) = 3.
⎩

The initial condition is 𝑦(0) = 3.

To solve this initial value problem, we first find the general solution and then determine
the parameter.

Note that the equation is equivalent to (ln 𝑦)′ = 1. Integrating both sides yields the general
solution 𝑦 = 𝑐𝑒 𝑥 . The initial condition to determine the value of the constant 𝑐 = 𝑦(0) = 3 Hence
𝑦 = 3𝑒 𝑥 is the solution of the initial value problem.

 Exercise 1.4 Consider the ordinary differential equation 𝑦 ′ − 2𝑦 + 𝑒 𝑥 = 0.

1. Show that 𝑦 = 𝑐𝑒 2𝑥 + 𝑒 𝑥 is the general solution.

2. Solve the initial value problem
⎧
⎪
⎪ 𝑦 ′ − 2𝑦 + 𝑒 𝑥 = 0
⎨
⎪
⎪ 𝑦(0) = 2.
⎩
Solution

1. If 𝑦 = 𝑐𝑒 2𝑥 + 𝑒 𝑥 , then 𝑦 ′ = 2𝑐𝑒 2𝑥 + 𝑒 𝑥 and

𝑦 ′ − 2𝑦 + 𝑒 𝑥 = 2𝑐𝑒 2𝑥 + 𝑒 𝑥 − 2(𝑐𝑒 2𝑥 + 𝑒 𝑥 ) + 𝑒 𝑥 = 0.

Hence, 𝑦 = 𝑐𝑒 2𝑥 + 𝑒 𝑥 is the general solution.

2. We use the initial condition to determine the choice of 𝑐. For 𝑦 = 𝑐𝑒 2𝑥 + 𝑒 𝑥 we have
𝑦(0) = 𝑐 + 1. We match it with the initial condition and we have

𝑐+1=2

9 Lecture Notes for MA451

 1.2 Basic Concepts

or simply 𝑐 = 1. Hence 𝑦 = 𝑒 2𝑥 + 𝑒 𝑥 is the solution of the initial value problem.

1.2.4 What is a General Solution

In the above example, the solution is a particular solution of the differential equation.

Like antiderivatives, without any constraint, a differential equation may have a family of
solutions. Such a family parametrizes all solutions of the differential equation.
Definition 1.5
A general solution consists of all solutions of a differential equation.
♣

Example 1.10 Show that 𝑦 = 𝑐1 𝑒 𝑥 + 𝑐2 𝑒 −𝑥 is the general solution of

𝑦 ′′ − 𝑦 = 0

on (−∞, ∞), where 𝑐1 and 𝑐2 can be any numbers.

Solution Differentiating the function 𝑦 yields

𝑦 ′ = 𝑐1 𝑒 𝑥 − 𝑐2 𝑒 −𝑥 ,

𝑦 ′′ = 𝑐1 𝑒 𝑥 + 𝑐2 𝑒 −𝑥 .

Then
𝑦 ′′ − 𝑦 = 𝑐1 𝑒 𝑥 + 𝑐2 𝑒 −𝑥 − (𝑐1 𝑒 𝑥 + 𝑐2 𝑒 −𝑥 ) = 0.

So the function 𝑦 is a solution. The fact that any solution can be written in this form is equivalent to
the uniqueness of a solution to the initial value problem with the general initial conditions 𝑦(𝑥0 ) = 𝑦0
and 𝑦 ′ (𝑥0 ) = 𝑦0′ . The proof of the uniqueness is above the level of this course. However, one can also
argue using Calculus. The differential equation can be re-written as follows

(𝑦 ′ − 𝑦)′ + (𝑦 ′ − 𝑦) =0
(𝑦 ′ − 𝑦)′
=−1
𝑦′ − 𝑦
′
′
( ln(𝑦 − 𝑦)) = − 1
′
Integrating the equation ( ln(𝑦 ′ − 𝑦)) = −1 yields the general solution 𝑦 ′ − 𝑦 = 𝑐𝑒 −𝑥 . Similarly, the
equation may be re-written as
′
−𝑥
( ln(𝑦 − 𝑐2 𝑒 )) = 1,
where 𝑐2 = − 2𝑐 . Integrating the equation yields that 𝑦 = 𝑐1 𝑒 𝑥 + 𝑐2 𝑒 −𝑥 . ■

10 Lecture Notes for MA451

 1.2 Basic Concepts

 Exercise 1.5 Find the general solution to the differential equation

𝑦 ′′′ = sin 𝑥.

Solution Integrating the differential equation repeatedly yields

𝑦 ′′ = − cos 𝑥 + 𝑐1 ,

𝑦 ′ = − sin 𝑥 + 𝑐1 𝑥 + 𝑐2 ,

𝑦 = cos 𝑥 + 𝑐1 𝑥 + 𝑐2 𝑥 + 𝑐3 ,

where 𝑐1 , 𝑐2 and 𝑐3 are arbitrary constants. Since 𝑦 ′′ and 𝑦 ′ are both general solutions by
Rolle’s theorem, so is 𝑦. ■

1.2.5 Solving Differential Equations by Direct Integration

To solve a general differential equation is usually difficult. In this course, we will only focus
on some special types of differential equations. For example, some differential equations can
be solved by simply integrating both sides using techniques from Calculus. We call them direct
integration problem.

Example 1.11 Solve the initial value problem.

1
𝑦′ = , 𝑦(1) = 0.
𝑥2 +𝑥

Solution Using partial fraction expansion, the differential equation can be written as

1 1
𝑦′ = − .
𝑥 𝑥 +1
Integrating both sides yields
3
𝑦 =∫ 3 d𝑥
𝑥 +𝑥
1 1
=∫ ( − d𝑥
𝑥 𝑥 + 1)
= ln |𝑥| − ln(|𝑥 + 1|) + 𝑐

Now using the initial condition 𝑦(1) = 0 to determine 𝐶:

0 = ln 1 − ln 2 + 𝑐

Since ln 1 = 0, we get 𝑐 = ln 2. The solution of this initial value problem is

𝑦 = ln |𝑥| − ln(|𝑥 + 1|) + ln 2.

11 Lecture Notes for MA451

 1.2 Basic Concepts

Note that the domain of the solution function is (−∞, −1) ∪ (−1, 0) ∪ (0, ∞). ■
In order to solve direct integral problems, you will need to review integrations of some
elementary functions and various methods of integrations such as integration by parts, inte-
gration by substitution, partial fraction expansion.

 Exercise 1.6 Solve the initial value problem.

𝑦 ′ = 𝑥𝑒 𝑥 , 𝑦(0) = 1.

Solution Using integration by parts, one will find that

𝑦 = 𝑥𝑒 𝑥 − 𝑒 𝑥 + 𝑐.

The initial condition 𝑦(0) = 1 implies that 𝑐 = 2. So the solution of this initial value problem is
𝑦 = 𝑥𝑒 𝑥 − 𝑒 𝑥 + 2. ■

1.2.6 Integral Curve*

Definition 1.6
The graph of a solution of a differential equation is called a solution curve.
A curve 𝐶 is called an integral curve of a differential equation if every solution curve is a part
of it.
♣

From the definition, we see that any solution curve of a differential equation is an integral
curve, but an integral curve need not be a solution curve.

Example 1.12 For any positive constant 𝑎, consider the circle defined by 𝑥 2 + 𝑦 2 = 𝑎2 . Verify
the circle is an integral curve of 𝑦 ′ = − 𝑦𝑥 but not a solution curve.
√ √
Solution The circe is made of two solution curves defined by 𝑦 = 𝑎2 − 𝑥 2 and 𝑦 = − 𝑎2 − 𝑥 2 .
But the circle is not a function. Note that the differential equation can be re-written as 𝑦𝑦 ′ = 𝑥.
Integrate both sides, you will see that all solutions of 𝑦 ′ = − 𝑦𝑥 satisfy the equation 𝑥 2 + 𝑦 2 = 𝑎2 for
some 𝑎. Thus, the circle is an integral curve. ■
 Exercise 1.7 Verify that the cuspidal curve defined by 𝑦 2 = 𝑥 3 + 𝑐 is an integral curve for
2
𝑦 ′ = − 3𝑥2𝑦 but not a solution curve.
2 √
Solution The equation 𝑦 ′ = − 3𝑥2𝑦 has two solution curves defined by 𝑦 = 𝑥 3 + 𝑐 and 𝑦 =
√
− 𝑥 3 + 𝑐. Apply the same method as in the above example, one can verify that the equation
𝑦 2 = 𝑥 3 + 𝑐 defines an integral curve. ■

12 Lecture Notes for MA451

 1.2 Basic Concepts

1.2.7 Direction Fields for First Order Equations

When finding an explicit formula for the solution of a differential equation is impossible
or the formula is too complicated, we may use graphical or numerical methods to investigate
how the solution behaves.

In this section, we consider a graphical method for first order differential equations 𝑦 ′ =
𝑓 (𝑥, 𝑦).

The idea is to sketch the integral curve using the first derivative 𝑦 ′ . To be specific, the
slope 𝑦0′ of an integral curve of the equation 𝑦 ′ = 𝑓 (𝑥, 𝑦). through a given point (𝑥0 , 𝑦0 ) is given
by the number 𝑓 (𝑥0 , 𝑦0 ). Through the point (𝑥0 , 𝑦0 ), we may draw a small line segment with the
slope 𝑦0′ .
Definition 1.7
Let 𝑓 be a function defined on a set 𝑅 ∈ ℝ2 . The graph consists of line segments through every
point (𝑥, 𝑦) in 𝑅 with the slope 𝑓 (𝑥, 𝑦) is called the direction filed (also called the slope field)
for 𝑦 ′ = 𝑓 (𝑥, 𝑦) in 𝑅.
♣

In practice, we can’t actually draw line segments through very points in 𝑅 if 𝑅 is an infinite
set. Instead, we select a finite subset of 𝑅 and draw line segments through points in it. For
example, we may create a rectangular grid and draw line segments through grid points.

Example 1.13 Consider the differential equation 𝑦 ′ = − 𝑦𝑥 .

Given a grid, the direction field (Figure 1.4 can be constructed by calculating 𝑦 ′ at grid
points and drawing a short line segment with the slope 𝑦 ′ and trough (𝑥, 𝑦).

From the direction field, you may guess what does an integral curve look like (see Figure
1.5). It’s a circe (see Example 1.12)!

𝑦 𝑦

𝑥 𝑥

Figure 1.4: The direction field for 𝑦 ′ = − 𝑦𝑥 . Figure 1.5: The direction field with an in-
tegral curve for 𝑦 ′ = − 𝑦𝑥 .

13 Lecture Notes for MA451

 1.2 Basic Concepts

 Exercise 1.8 Consider the differential equation 𝑦 ′ = 𝑦𝑥 . Sketch the direction field and guess
a solution.

Solution The direction field for 𝑦 ′ = 𝑦𝑥 suggests a solution curve is a piece of hyperbola. The
general solutions of the differential equations satisfy the equation 𝑦 2 − 𝑥 2 = 𝑐 (see Figure 1.5).
If 𝑐 > 0, then an integral curve will look like the red curve in Figure 1.5. If 𝑐 < 0, then an
integral curve will look like the green curve in Figure 1.5.

Figure 1.6: The direction field with an integral curve for 𝑦 ′ = 𝑦𝑥 .

14 Lecture Notes for MA451

 Week 2: Separable and Linear First Order
Equations

9/2–9/19

A first order differential equation can always be written in the standard form

𝑦 ′ = 𝑓 (𝑥, 𝑦).

Sometimes, it might be easier to consider the differential form of the equation.

𝑀(𝑥, 𝑦) d 𝑥 + 𝑁 (𝑥, 𝑦) d 𝑦 = 0.

Like direction integration, different type of differential equations will be solved using dif-
ferent techniques.

In the coming weeks, you will learn how to solve following types of first order differential
equations.

1. Separable equation
𝐴(𝑥) d 𝑥 + 𝐵(𝑦) d 𝑦 = 0.

or
𝐹 (𝑥)
𝑦′ = .
𝐺(𝑦)
2. Linear first order equations
𝑦 ′ + 𝑝(𝑥)𝑦 = 𝑞(𝑥).

3. Bernoulli equations
𝑦 ′ + 𝑝(𝑥)𝑦 = 𝑞(𝑥)𝑦 𝑛 .

4. Homogeneous equations 𝑦 ′ = 𝑓 (𝑥, 𝑦) if 𝑓 (𝑡𝑥, 𝑡𝑦) = 𝑓 (𝑥, 𝑦).

5. Exact equations
𝑀(𝑥, 𝑦) d 𝑥 + 𝑁 (𝑥, 𝑦) d 𝑦 = 0

and
𝜕𝑀 𝜕𝑁
= .
𝜕𝑦 𝜕𝑥
6. Other nonlinear first order equations that can be solved by substitution.
 2.1 Separable Equations

2.1 Separable Equations

Definition 2.1
A first order differential equation is said to be separable if it can be written as

ℎ(𝑦)𝑦 ′ = 𝑔(𝑥) (2.1)

Rewriting a separable differential equation in this form is called separation of variables.

♣

2.1.1 How to solve separable equations

′
d
Suppose that 𝐻 (𝑦) is an antiderivative of ℎ(𝑦), i.e d𝑦
𝐻 (𝑦) = ℎ(𝑦). Then (𝐻 (𝑦(𝑥))) =
ℎ(𝑦(𝑥))𝑦 ′ (𝑥) and Equation 2.1 can be written as
′
(𝐻 (𝑦(𝑥))) = 𝑔(𝑥)

Suppose 𝐺(𝑥) is an antiderviative of 𝑔(𝑥). Integrating both sides of the above equation yields

𝐻 (𝑦(𝑥)) = 𝐺(𝑥) + 𝑐. (2.2)

It can be checked that a solution of Equation 2.1 must satisfy Equation 2.2.

Equation 2.2 is called an implicit solution of ℎ(𝑦)𝑦 ′ = 𝑔(𝑥).

In conclusion, to solve a separable equation, it suffices to find antiderivatives using direct

integrations.

From Calculus, we know that continuous functions have an antiderivatives. Therefore, as

long as ℎ(𝑦) and 𝑔(𝑥) are continuous, the differential equation 2.1 will have an implicit solution.
The existence of a solution function follows from a result from advance Calculus called the
implicit function theorem. Moreover, the solution of an initial value problem for Equation 2.1
is unique.

If the constant 𝑐 in Equation 2.2 satisfies the initial condition, then we say the implicit
solution is an implicit solution of the initial value problem for Equation 2.1

Example 2.1 Solve the equation

𝑦 ′ = 𝑥(1 + 𝑦 2 ).

Solution Separating variables yields

𝑦′
= 𝑥.
1 + 𝑦2

16 Lecture Notes for MA451

 2.1 Separable Equations

Integrating leads to
−1 𝑥2
tan 𝑦 = + 𝑐.
2
Therefore, the general solution of the equation is

𝑥2
𝑦 = tan +𝑐 .
(2 )

■
 Exercise 2.1 Solve the equation
sin 𝑥
𝑦′ =
𝑦

Solution Rewrite the equation as

𝑦𝑦 ′ = sin 𝑥.

Integrating gives
1 2
𝑦 = − cos 𝑥 + 𝑐.
2
Hence, the general solutions are
√
𝑦 = ± −2 cos 𝑥 + 2𝑐.

2.1.2 Constant solutions

In the above example, dividing 1 + 𝑦 2 is an equivalent transformation because it is always
𝑦′
non-zero. In general, re-writing the function 𝑦 ′ = 𝑔(𝑥)𝑝(𝑦) into 𝑝(𝑦) = 𝑔(𝑥) may lose some
constant solutions.

Example 2.2 Solve the differential equation

𝑦 ′ = 2𝑥𝑦 2

Solution It can be check that the function 𝑦 = 0 is a solution. Suppose that 𝑦 is a solution that is
not identically zero. Then there must be intervals on which 𝑦 is never zero. Over this interval, we
can separate variables, which yields
𝑦′
= 2𝑥.
𝑦2
Integrating both sides leads to
1
− = 𝑥2 + 𝑐
𝑦
Therefore, 𝑦 = − 𝑥 21+𝑐 is the general solution of the equation that is not identically zero. ■

17 Lecture Notes for MA451

 2.1 Separable Equations

Remark If a first order separable is in the form

𝑦 ′ = 𝑔(𝑥)𝑝(𝑦).

Before rewrite the differential equation in the form

𝑦′
= 𝑔(𝑥),
𝑝(𝑦)

we need to check for values of 𝑦 that make 𝑔(𝑦) = 0. The equation 𝑔(𝑦) = 0 often leads to
constant solutions.

 Exercise 2.2 Find the general solution of

𝑦 ′ = 𝑒𝑥 𝑦 2.

Solution Setting 𝑦 2 = 0 gives a constant solution 𝑦 = 0.

Now suppose that 𝑦 is not a constant solution.

Rewrite the equation as

𝑦′
2
= 𝑒𝑥 .
𝑦
Integrating yields
1
− = 𝑒𝑥 + 𝑐
𝑦
Therefore, the general solution is
1
𝑦= .
−𝑐 − 𝑒 𝑥
■

2.1.3 Initial value problems

Example 2.3 Solve the initial value problem
cos 𝑥
𝑦′ = , 𝑦(0) = 3.
𝑦

Solution We first find the general solution. The equation is the same as

𝑦𝑦 ′ = cos 𝑥.

Integrating both sides yields

1 2
𝑦 = sin 𝑥 + 𝑐
2

18 Lecture Notes for MA451

 2.1 Separable Equations

or
√
𝑦 = ± 2𝑐 + 2 sin 𝑥.

The initial condition leads to

√
𝑦(0) = 2𝑐 = 3

Hence, 2𝑐 = 9 and the solution of the initial value problem is

√
𝑦= 9 + 2 sin 𝑥.

■
 Exercise 2.3 Solve the initial value problem

𝑥3 + 2
𝑦′ = , 𝑦(0) = 1.
𝑦3 + 2

Solution Rewrite the equation as

(𝑦 3 + 2)𝑦 ′ = 𝑥 3 + 2

Integrating both sides yields

1 4 1
𝑦 + 2𝑦 = 𝑥 4 + 2𝑥 + 𝑐.
4 4
The initial condition implies that
1 9
𝑐 = 14 + 2 ⋅ 1 = .
4 4
Therefore, the solution of the initial value problem is

1 4 1 9
𝑦 + 2𝑦 = 𝑥 4 + 2𝑥 + .
4 4 4
■

2.1.4 Autonomous Equations*

Definition 2.2
A differential equation is called autonomous if it can be written as

d𝑥
= 𝑓 (𝑥).
d𝑡 ♣

Autonomous differential equations are separable. The general solution satisfies

1
∫ 𝑓 (𝑥) d 𝑥 = 𝑡 + 𝑐.

19 Lecture Notes for MA451

 2.1 Separable Equations

In other words, an autonomous differential equation always has a general implicit solution.

Those equations have many applications.

Example 2.4 The population 𝑃 = 𝑃(𝑡) of a species satisfies the logistic equation

𝑃 ′ = 𝑎𝑃(1 − 𝛼𝑃)

and 𝑃(0) = 𝑃0 > 0, where 𝑎 and 𝛼 are constants. Find 𝑃 for 𝑡 > 0.

Solution The equation can be re-written as

𝑃′
= 𝑎.
𝑃(1 − 𝛼𝑃)

Applying the partial fraction decomposition to left hand side yields

𝑃′ 𝛼𝑃 ′
+ = 𝑎,
( 𝑃 1 − 𝛼𝑃 )

or equivalently
′
( ln(𝑃) − ln(1 − 𝛼𝑃)) = 𝑎.
Therefore,
𝑃
= 𝑐𝑒 𝑎𝑡 . (2.3)
1 − 𝛼𝑃
By the initial condition 𝑃(0) = 𝑃0 ,
𝑃0
𝑐= .
1 − 𝛼𝑃0
Solving for 𝑃 from Equation 2.3 leads to

𝑃0 𝑒 𝑎𝑡
𝑃=
1 − 𝛼𝑃0 + 𝛼𝑃0 𝑒 𝑟𝑡
𝑃0
=
𝛼𝑃0 + (1 − 𝛼𝑃0 )𝑒 −𝑟𝑡

■
 Exercise 2.4 A frozen pizza, initially at 32◦ 𝐹 is put into an oven that is pre-heated to 400◦ 𝐹 .
The pizza warmed up to 50◦ 𝐹 in 2 minutes. Find how long would it take to reach 200◦ 𝐹 by using
Newton’s cooling law 𝑇 ′ (𝑡) = −𝑘(𝑇 (𝑡) − 𝑇𝑚 (𝑡)), where 𝑇 (𝑡) is the temperature of the pizza after
𝑡 minutes and 𝑇𝑚 (𝑡) is the ambient temperature,

Solution Since the oven temperature is constantly 400◦ 𝐹 , the Newton cooling law implies

𝑑𝑇
= 𝑘(400 − 𝑇 )
𝑑𝑡

20 Lecture Notes for MA451

 2.1 Separable Equations

Dividing by (400 − 𝑇 ) and integrating yields

− ln(400 − 𝑇 ) = 𝑘𝑡 + 𝐶

Solving for 𝑇 gives the general solution

𝑇 = 400 − 𝑒 −𝑘𝑡−𝐶 ,

or equivalent
𝑇 = 400 − 𝑐𝑒 −𝑘𝑡 .

Since the initial temperature of the pizza was 32, that is 𝑇 (0) = 32, plugging it into the function
𝑇 (𝑡) and solving for 𝑐 gives 𝑐 = 368. Since it takes 2 minutes for the temperature to reach
50◦ 𝐹 , that is 𝑇 (2) = 50, the value 𝑘 must satisfy the equation

50 = 400 − 368𝑒 −2𝑘 .

Solving for 𝑘 yields

1 350
𝑘 = − ln = 0.02507
2 368
Thus the temperature of the pizza at time 𝑡 is:

𝑇 (𝑡) = 400 − 368𝑒 −0.02507𝑡 .

To reach 200◦ 𝐹 , the time that it takes should satisfies

200 = 400 − 368𝑒 −0.02507𝑡 .

Solving this for 𝑡 implies that it take 24.3 minutes for the pizza to reach 200◦ 𝐹 ■

2.1.5 More examples

Example 2.5 Solve the differential equation

𝑦 ′ = −2𝑥𝑦 + 3𝑥

Solution Factoring out 𝑥 from the right-hand side separates variables

𝑦 ′ = −2𝑥(𝑦 − 3).

Setting 𝑦 − 3 = 0 leads to a constant solution 𝑦 = 3.

21 Lecture Notes for MA451

 2.2 Linear First Order Equations

Suppose that 𝑦 is not identically 3. Then the equation may be re-written as

𝑦′
= −2𝑥.
𝑦−3

Integrating both sides implies

ln(|𝑦 − 3|) = 𝑥 2 + 𝑐1 .

Solve for 𝑦 leads to the general solutions

2
𝑦 = ±𝑐𝑒 𝑥 + 3,

where 𝑐 > 0.
2
Note that the constant solution may includes in the general solutions 𝑦 = ±𝑐𝑒 𝑥 + 3 by allowing
𝑐 = 0. ■
Example 2.6 Solve the initial value problem

𝑦 ′ = 𝑒 𝑥+2𝑦 , 𝑦(0) = 1.

Solution The equation can be re-written as

𝑦 ′ = 𝑒 𝑥 𝑒 2𝑦

Since 𝑒 2𝑦 > 0 for any 𝑦, dividing by 𝑒 2𝑦 leads to the following equivalent differential equation

𝑒 −2𝑦 𝑦 ′ = 𝑒 𝑥

Direct integration yields

1
− 𝑒 −2𝑦 = 𝑒 𝑥 + 𝑐.
2
Solving for 𝑦 gives the general solution

1
𝑦 = − ln (−2𝑒 𝑥 − 2𝑐).
2
The initial condition 𝑦(0) = 1 implies that

1
1 = − ln (−2 − 2𝑐).
2

Therefore, 𝑐 = − 21 𝑒 −2 − 1. Then the solution of the initial value problem is

1
𝑦 = − ln (−2𝑒 𝑥 + 𝑒 −2 + 2).
2
■

22 Lecture Notes for MA451

 2.2 Linear First Order Equations

2.2 Linear First Order Equations

Definition 2.3
A first order differential equation is called linear if it can be written as

𝑦 ′ + 𝑝(𝑥)𝑦 = 𝑓 (𝑥).

A first order differential equation that cannot be written like this is nonlinear.
A linear first order differential equation is said to be homogeneous if 𝑓 (𝑥) is identically 0.
♣

Example 2.7 Determine whether the equation is linear, homogeneous linear or not.

1. 𝑦 ′ − 2𝑦 = −𝑒 𝑥 .
2. 𝑥 2 𝑦 ′ + 𝑒 𝑥 𝑦 = 0.
3. 𝑥𝑦 ′ + 𝑦 2 = 0.

Solution

1. 𝑦 ′ − 2𝑦 = −𝑒 𝑥 is a linear first order differential equation because the highest order and expo-
nents of 𝑦 are both 1.
2. 𝑥 2 𝑦 ′ + 𝑒 𝑥 𝑦 = 0 is also a linear first order differential equation. Moreover, it is homogeneous
because, it has no constant term, i.e. it is in the form 𝑦 ′ + 𝑝(𝑥)𝑦 = 0.
3. 𝑥𝑦 ′ + 𝑦 2 = 0 is not linear because the highest exponent of 𝑦 is 2.

■
Note that a homogeneous linear first order differential equation 𝑦 ′ + 𝑝(𝑥)𝑦 = 0 is as special
separable differential equation. It always has 𝑦 = 0 as a solution. We call it the trivial solution.

If 𝑝(𝑥) = 0, the the equation becomes 𝑦 ′ = 𝑓 (𝑥) which can be solve by direct integration if
𝑓 (𝑥) is continuous.

How can we solve a general linear non-homogeneous differential equation? There are
two methods both uses the product rule in some ways.

2.2.1 The integrating factor method

Let’s first see an example.

Example 2.8 Solve the equation

2 𝑒𝑥
𝑦′ + 𝑦 = 2 .
𝑥 𝑥

Solution Let’s first clear the denominator by multiplying 𝑥 2 to both sides which yields

2𝑥𝑦 + 𝑥 2 𝑦 ′ = 𝑒 𝑥 .

23 Lecture Notes for MA451

 2.2 Linear First Order Equations
′
Now, notice that the left side is same as (𝑥 2 𝑦 ) by the product rule. Therefore, the equation can be
re-written as
2 ′ 𝑥
(𝑥 𝑦 ) = 𝑒
Integrating both sides leads to
𝑥 2 𝑦 = 𝑒 𝑥 + 𝑐.

Solving for 𝑦 gives the general solution

1 3𝑥 𝐶
𝑦= 𝑒 + 2.
𝑥2 𝑥
■
It seems that we are lucky that the left hand side becomes the derivative of a product.
But it also suggests that we may look for a multiplier so that the left hand side will be the
derivative of a product. The existence a multiplier should be clear once we find it.

Let’s suppose that there is a function 𝑟(𝑥) such that

′
𝑟(𝑥)𝑦 ′ + 𝑟(𝑥)𝑝(𝑥)𝑦 = (𝑟(𝑥)𝑦 ) .

Apply the product rule to 𝑟(𝑥)𝑦 and compare both sides of the above equation, we see that
𝑟(𝑥) must satisfy the separable differential equation

𝑟 ′ (𝑥) = 𝑟(𝑥)𝑝(𝑥).

Solving this separable differential equation, we get 𝑟(𝑥) = 𝑒 ∫ 𝑝(𝑥) d 𝑥 which is called the integrating
factor for 𝑦 ′ + 𝑝(𝑥)𝑦 = 𝑓 (𝑥). After finding 𝑟(𝑥), then the linear first order differential equation
becomes
′
( 𝑟(𝑥)𝑦 ) = 𝑟(𝑥)𝑓 (𝑥)
which has a solution
𝑦 = 𝑒− ∫ 𝑝(𝑥) d 𝑥
𝑒∫ 𝑝(𝑥) d 𝑥
𝑓 (𝑥) d 𝑥 + 𝑐 .
(∫ )
This method is called the integrating factor method.

Remark Note that it doesn’t matter which antiderivative we take when computing the inte-
grating factor. Because, it will eventually alter the constant 𝑐 by a factor.

In the following, we will abuse the notation ∫ 𝑝(𝑥) d 𝑥 and set it equals a specific antideriva-
tive.

Example 2.9 Find the general solution of

𝑦 ′ + 2𝑦 = 𝑥 3 𝑒 −2𝑥 .

24 Lecture Notes for MA451

 2.2 Linear First Order Equations

Solution The integrating factor 𝑟(𝑥) can be taken as

2d𝑥
𝑟(𝑥) = 𝑒 ∫ = 𝑒 2𝑥 .

Multiplying both sides of the equation by 𝑒 2𝑥 transforms it to

𝑒 2𝑥 𝑦 ′ + 2𝑒 2𝑥 𝑦 = 𝑥 3 ,

or equivalently
2𝑥 3′
(𝑒 𝑦 ) = 𝑥 .
Integrating both sides yields
1
𝑦𝑒 2𝑥 = 𝑥 4 + 𝑐.
4
So 𝑦 = 41 𝑒 −2𝑥 (𝑥 4 + 𝑐) is the general solution of the equation. ■
 Exercise 2.5 Find the general solution of

𝑦 ′ + 𝑦 = 1.

Solution The integrating factor is

d𝑥
𝑟(𝑥) = 𝑒 ∫ = 𝑒𝑥 .

Multiplying both sides with the integrating factor 𝑒 𝑥 leads to

𝑒𝑥 𝑦 ′ + 𝑒𝑥 𝑦 = 𝑒𝑥

which is the same as

(𝑒 𝑥 𝑦)′ = 𝑒 𝑥 .

Integrating both side gives the equation

𝑒 𝑥 𝑦 == 𝑒 𝑥 + 𝑐.

Hence, the general solution of the equation is

𝑦 = 1 + 𝑐𝑒 −𝑥 .

■
Example 2.10 Solve the initial value problem.

𝑦 ′ + 4𝑦 = 𝑒 −4𝑥 𝑦(1) = 3.

25 Lecture Notes for MA451

 2.2 Linear First Order Equations

Solution Since 𝑝(𝑥) = 4, the integral factor is ∫ 𝑝(𝑥) d 𝑥 = 4𝑥. Multiplying the equation with 𝑒 4𝑥
leads to
𝑒 4𝑥 𝑦 ′ + 4𝑒 4𝑥 𝑦 = 1

which is the same as

(𝑒 4𝑥 𝑦)′ = 1

Integrating both sides yields

𝑒 4𝑥 𝑦 = 𝑥 + 𝑐.

So the general solution is

𝑦 = 𝑥𝑒 −4𝑥 + 𝑐𝑒 −4𝑥 .

The initial condition means when 𝑥=1, 𝑦 = 4. Plugging the point into the function 𝑦 produces
an equation of 𝑐
𝑒 4 ⋅ 3 = 1 + 𝑐,

which implies that

𝑐 = 3𝑒 4 − 1.

Hence, the solution to the initial value problem is

𝑦 = 𝑥𝑒 −4𝑥 + (3𝑒 4 − 1)𝑒 −4𝑥 .

■
 Exercise 2.6 Solve the initial value problem

𝑦 ′ − 3𝑦 = 𝑒 𝑥 , 𝑦(0) = 0.

Solution Since 𝑝(𝑥) = −3, ∫ 𝑝(𝑥) d 𝑥 = −3𝑥 and the integrating factor is 𝑒 −3𝑥 . Multiplying both
sides with 𝑒 2𝑥 yields
𝑒 −3𝑥 𝑦 ′ − 3𝑒 −3𝑥 𝑦 = 𝑒 −2𝑥

The product rule and chain rule together implies that

𝑒 −3𝑥 𝑦 ′ + 2𝑒 −3𝑥 𝑦 = (𝑒 −3𝑥 𝑦)′ .

Hence, the original equation can be transformed into

(𝑒 −3𝑥 𝑦)′ = 𝑒 −2𝑥 .

Integrating both sides yields

1
𝑒 −3𝑥 𝑦 = − 𝑒 −2𝑥 + 𝑐.
2

26 Lecture Notes for MA451

 2.2 Linear First Order Equations

Solving for 𝑦 gives the general solution

1
𝑦 = − 𝑒 𝑥 + 𝑐𝑒 3𝑥 .
2
Since 𝑦(0) = 0, 𝑐 satisfies
1
0 = − + 𝑐,
2
or equivalently 𝑐 = 12 .

Therefore, the solution to the initial value problem is

1 1
𝑦 = − 𝑒 𝑥 + 𝑒 3𝑥 .
2 2
■

2.2.2 The method of variation of parameters

If you read Trench’s book, you will find there is another method called variation of pa-
rameters. The idea is to solve the complimentary linear homogeneous equation 𝑦 ′ + 𝑝(𝑥)𝑦 = 0
first to get a solution 𝑦𝑜 . Then find a "parameter" 𝑢(𝑥) such that 𝑦 = 𝑢𝑦𝑜 is a solution of the
original linear non-homogeneous equation. By plugging 𝑢𝑦𝑜 into the non-homogeneous linear
equation and comparing both sides, you will find that 𝑢 ′ = 𝑓 𝑦(𝑥)
𝑜
which can be solved by direct
integration.

A possible motivation of the idea is as follows. Let 𝑦𝑝 be a particular solution for 𝑦 ′ +𝑝(𝑥)𝑦 =
𝑦
𝑓 (𝑥). Then 𝑦 = 𝑦𝑜 + 𝑦𝑝 is also a solution. Whatever the 𝑦𝑝 is, 𝑦 = 𝑦𝑜 (1 + 𝑦𝑝𝑜 ) is a solution. Here
𝑦
𝑢 = 1 + 𝑦𝑝𝑜 is the parameter that varies.

Both the integrating factor method and then variation of parameters work equally well for
first order equations. For higher order differential equations, they have there own advantages
and limitations. The integrating factor method can also be used for exact equations which
includes all linear first order equations. The method of variation of parameters can be used
for some nonlinear first order or linear higher order equations.

Example 2.11 Find the general solution of

𝑦 ′ + 2𝑦 = 4𝑥.

Solution First, we solve the complimentary linear homogeneous equation

𝑦 ′ + 2𝑦 = 0.

Note that the constant solution 𝑦 = 0 won’t be a solution to the original equation. So we assume

27 Lecture Notes for MA451

 2.2 Linear First Order Equations

that 𝑦 is not identically zero and hence dividing 𝑦 is legitimate. The linear homogeneous equation
can be re-written as
𝑦′
= −2𝑥.
𝑦
Integrating both sides implies 𝑦 = 𝑒 −2𝑥 . Again, you can check that adding a constant won’t change
the general solution.

Now, we want to find a function 𝑢 such that 𝑢𝑒 −2𝑥 is a solution of the original equation.

Suppose that 𝑢𝑒 −2𝑥 is a solution of the original equation. Then the function 𝑢 must satisfy the
following equation
𝑢 ′ 𝑒 −2𝑥 − 2𝑢𝑒 −2𝑥 + 2𝑢 −2𝑥 = 4𝑥,

or equivalently
𝑢 ′ = 4𝑥𝑒 2𝑥 .

Integrating using integration by parts yields

𝑢 = 2𝑥𝑒 2𝑥 + 2𝑒 2𝑥 + 𝑐.

Therefore, the general solution of the original equation is

𝑦 = 𝑢𝑒 −2𝑥 = 2𝑥 + 2 + 𝑐𝑒 −2𝑥 .

■
 Exercise 2.7 Find the general solution of

𝑦 ′ + 2𝑦 = 𝑥 3 𝑒 −2𝑥 .

using the method of variation of parameters.

Solution Solving the homogeneous linear equation 𝑦 ′ + 2𝑦 = 0 produces a solution 𝑦 = 𝑒 −2𝑥 .

Suppose that 𝑦 = 𝑢𝑒 −2𝑥 is a solution of the equation 𝑦 ′ + 2𝑦 = 𝑥 3 𝑒 −2𝑥 . Then the function 𝑢
satisfy the following differential equation.

𝑢 ′ 𝑒 −2𝑥 − 2𝑢𝑒 −2𝑥 + 2𝑢𝑒 −2𝑥 = 𝑥 3 𝑒 −2𝑥 ,

or equivalently
𝑢′ = 𝑥 3.

Therefore,
𝑥4
𝑢= + 𝑐,
4

28 Lecture Notes for MA451

 2.3 Change of Variables

and
1
𝑦 = 𝑢𝑒 −2𝑥 = 𝑒 −2𝑥 (𝑥 4 + 𝑐 )
4
is the general solution. ■

2.3 Change of Variables

Some non-linear first order differential equations may be solve by changing variable. In
this section, we will look into two of such types.

2.3.1 Bernoulli Equations

Definition 2.4
A Bernoulli first order differential equation is a differential equation of the form

𝑦 ′ + 𝑝(𝑥)𝑦 = 𝑞(𝑥)𝑦 𝑛 .
♣

On way to solve this type of equation is to use the method of variation of parameters.

Example 2.12 Solve the Bernoulli equation

𝑦 ′ − 𝑦 = 𝑥𝑦 2 .

Solution Solving the complimentary homogeneous equation 𝑦 ′ −𝑦 = 0 yields a solution 𝑦𝑜 = 𝑒 𝑥 . We

look for solutions of the original equation in the form 𝑦 = 𝑢𝑒 𝑥 . Plugging 𝑦 = 𝑢𝑒 𝑥 into the Bernoulli
equation yields
𝑢 ′ 𝑒 𝑥 = 𝑥𝑢 2 𝑒 2𝑥 ,

or equivalently
𝑢 ′ = 𝑥𝑢 2 𝑒 𝑥

which is a separable equation. Separating variables and integrating both sides gives

𝑢′
=𝑥𝑒 𝑥
𝑢2
1 ′
− ( ) =𝑥𝑒 𝑥
𝑢
1
− =𝑥𝑒 𝑥 − 𝑒 𝑥 + 𝑐.
𝑢

Hence,
1
𝑢=−
(𝑥 − 1)𝑒 𝑥 + 𝑐

29 Lecture Notes for MA451

 2.3 Change of Variables

and
1
𝑦=− .
𝑥 − 1 + 𝑐𝑒 −𝑥
■
Another way is to use a substitution to reduce the equation to a linear equation. Indeed,
suppose 𝑦 is not identically zero. Then dividing 𝑦 𝑛 from both sides yields

𝑦 ′ 𝑦 −𝑛 + 𝑝(𝑥)𝑦 −(𝑛−1) = 𝑞(𝑥).

′
Notice that 𝑦 ′ 𝑦 −𝑛 = (𝑦 −(𝑛−1) ) . Let 𝑣 = 𝑦 −(𝑛−1) , then the equation becomes

𝑣 ′ + 𝑝(𝑥)𝑧 = 𝑞(𝑥)

which can be solved by multiple methods.

Example 2.13 Find the solution of the initial value problem

𝑦 ′ + 2𝑦 = 𝑦 3 𝑦(0) = 2.

Solution Clearly 𝑦 = 0 is a trivial solution of the differential equation but not a solution of the
initial value problem. Assume that 𝑦 is not identically zero. Since 𝑛 = 3, let 𝑣 = 𝑦 −2 . Dividing 𝑦 3
from both sides of the equations implies that 𝑧 satisfies the equation.

𝑣 ′ − 4𝑣 = −2.

According to the method of integrating factor, the integrating factor is

𝑟(𝑥) = 𝑒 ∫ (−4) d 𝑥 = 𝑒 −4𝑥 .

Multiplying with 𝑟(𝑥) and integrating both sides yields

𝑣 ′ 𝑒 −4𝑥 − 4𝑣𝑒 −4𝑥 = − 2𝑒 −4𝑥

−4𝑥 ′ −4𝑥
(𝑣𝑒 ) = − 2𝑒
1
𝑣𝑒 −4𝑥 = 𝑒 −4𝑥 + 𝑐
2
1
𝑣 = + 𝑐𝑒 4𝑥 .
2

As 𝑣 = 𝑦 −2 , then the general solutions are

±1 ±1
𝑦 = √ = √1 .
𝑣 2
+ 𝑐𝑒 4𝑥

30 Lecture Notes for MA451

 2.3 Change of Variables

Since 𝑦(0) = 2, the constant 𝑐 satisfies

±1
√1 = 2.
2
+ 𝑐𝑒 4𝑥

Because square root is nonnegative, only the equation

1
√1 =2
2
+ 𝑐𝑒 4𝑥

may have a solution. Solving for 𝑐 gives

1
𝑐=− .
4

Therefore, the solution of the initial value problem is

1
𝑦 = √1 .
2
− 41 𝑒 4𝑥

■
Remark Note that in this exercise if we change the initial condition to 𝑦(0) = −2 the solution
will be 𝑦 = − √ 1 1 1 4𝑥 . This suggests that the solution of an initial value problem may depend on
2−4𝑒
the initial condition.

 Exercise 2.8 Solve the initial value problem

1
𝑦′ − 𝑦 = , 𝑦(0) = 1.
𝑦

Solution Multiplying both sides with 𝑦 and applying the substitution 𝑣 = 𝑦 2 implies that 𝑣
satisfies
𝑣 ′ − 2𝑣 = 2

The integrating factor is

𝑟(𝑥) = 𝑒 −2𝑥

and the equation for 𝑣 is equivalent to

(𝑣𝑒 −2𝑥 )′ = 2𝑒 −2𝑥

whose general solution is

𝑣 = 𝑒 2𝑥 ∫ 2𝑒 −2𝑥 d 𝑥 = 𝑒 2𝑥 (𝑐 − 𝑒 −2𝑥 ) = 𝑐𝑒 2𝑥 − 1.

Hence
√ √
𝑦 = ± 𝑣 = ± 𝑐𝑒 2𝑥 − 1.

31 Lecture Notes for MA451

 2.3 Change of Variables
√
Since 𝑦(0) = 1, then 𝑐 = 2 and 𝑦 = 2𝑒 2𝑥 − 1. ■

32 Lecture Notes for MA451

 Week 3: Substitution Methods and Exact
Equations

9/20–9/26

3.1 Substitution Methods

3.1.1 Homogeneous Equations

Definition 3.1
A function 𝑓 is said to be homogeneous of degree 𝑚 if 𝑓 (𝑡𝑥, 𝑡𝑦) = 𝑡 𝑚 𝑓 (𝑥, 𝑦) for any nonzero
constant 𝑡. ♣
𝑦 𝑥
Example 3.1 The functions 𝑧 = and 𝑧 = are homogeneous of degree 0.
𝑥 𝑦
𝑥𝑦
The function 𝑧 = is also homogeneous of degree 0.
𝑥2 + 𝑦2
But the function 𝑧 = 𝑥 2 + 𝑥𝑦 − 𝑦 2 is homogeneous of degree 2.

From the example, you may conjecture that a homogeneous functions of degree 0 is a
function of the single variable 𝑣 = 𝑥𝑦. Indeed, the following theorem confirms that.
Theorem 3.1
𝑦
A function 𝑓 (𝑥, 𝑦) is homogeneous of degree zero if and only if it depends on 𝑥
only.
♥

Proof If 𝑓 (𝑥, 𝑦) is homogeneous of degree zero, then

1 1 𝑦
𝑓 (𝑥, 𝑦) = 𝑓 ( ⋅ 𝑥, ⋅ 𝑦 ) = 𝑓 (1, )
𝑥 𝑥 𝑥

where the right hand side is a function depends only on 𝑦𝑥 .

Conversely, if 𝑓 (𝑥, 𝑦) = 𝑔 ( 𝑦𝑥 ), where 𝑔 is a single variable function, then

𝑡𝑦 𝑦
𝑓 (𝑡𝑥, 𝑡𝑦) = 𝑔 ( ) = 𝑔 ( ) = 𝑓 (𝑥, 𝑦).
𝑡𝑥 𝑥
■
Definition 3.2
The differential equation dd 𝑦𝑥 = 𝑓 (𝑥, 𝑦) is called a homogeneous first order differential equa-
tion if 𝑓 is homogeneous of degree 0.
♣
 3.1 Substitution Methods

To solve a homogeneous first order differential equation, we substitute 𝑦 by 𝑥𝑣, where

𝑦
𝑣 = . Then
𝑥
d𝑦 d𝑣
=𝑣+𝑥 .
d𝑥 d𝑥
By Theorem 3.1, the homogeneous first order differential equation dd 𝑦𝑥 = 𝑓 (𝑥, 𝑦) can be reduced
to the following separable equation

d𝑣
𝑥 = 𝑞(𝑣) − 𝑣,
d𝑥

𝑥
where 𝑞 is a single variable function such that 𝑓 (𝑥, 𝑦) = 𝑞 .
(𝑦 )
Example 3.2 Solve
𝑦 𝑦
𝑦 ′ = 𝑒− 𝑥 + .
𝑥

Solution
𝑦
Step 1: Since 𝑡𝑦
𝑡𝑥
= 𝑦𝑥 for any nonzero number 𝑡, the function 𝑧 = 𝑒 − 𝑥 + 𝑦𝑥 is homogeneous of degree
0 and the differential equation is a homogeneous first order equation.
Step 2: Consider the new unknown function 𝑣 = 𝑦𝑥 . The function 𝑧 can be re-written as

𝑦 𝑦
𝑧 = 𝑒− 𝑥 + = 𝑒 −𝑣 + 𝑣.
𝑥
Step 3: Differentiating the equation 𝑦 = 𝑥𝑣 with respect to 𝑥 using the product rule implies that

𝑦 ′ = 𝑣 + 𝑥𝑣 ′ .

Step 4: Then the original equation can be transformed into

𝑣 + 𝑥𝑣 ′ =𝑒 −𝑣 + 𝑣
𝑥𝑣 ′ =𝑒 −𝑣

Step 5: Note that the equation 𝑥𝑣 ′ = 𝑒 −𝑣 is a separable equation which can be re-written as

1
𝑒𝑣 𝑣′ =
𝑥
Integrating both side yields
𝑒 𝑣 = ln |𝑥| + 𝑐.

Therefore,
𝑣 = ln(ln |𝑥| + 𝑐),

34 Lecture Notes for MA451

 3.1 Substitution Methods
𝑥
Step 6: Substituting 𝑣 by 𝑦
and solving for 𝑦 yields the general solution

𝑦 = 𝑥 ln(ln |𝑥| + 𝑐).

■
 Exercise 3.1 Find the general solution of
𝑦+𝑥
𝑦′ =
𝑥

Solution

Step 1: The equation is homogeneous of first order. Because

(𝑡𝑦) + (𝑡𝑥) 𝑡(𝑦 + 𝑥) 𝑦 + 𝑥

= = .
𝑡𝑥 𝑡𝑥 𝑥
Step 2: Set 𝑦 = 𝑥𝑣. The right-hand side function can be rewritten as the follows
𝑦 + 𝑥 𝑥𝑣 + 𝑥
= = 𝑣 + 1.
𝑥 𝑥
Step 3: Applying the product rule to 𝑦 = 𝑥𝑣 with respect to 𝑥 yields

𝑦 ′ = 𝑣 + 𝑥𝑣 ′ .

Step 4: The original equation is transformed into the following separable equation

𝑣 + 𝑥𝑣 ′ =𝑣 + 1
𝑥𝑣 ′ =1.

Step 5: Solving the resulting equation yields

1
𝑣′ =
𝑥
𝑣 = ln |𝑥| + 𝑐

Step 6: Solving for 𝑦 gives the general solution

𝑦 = 𝑥(ln |𝑥| + 𝑐).

■
 Exercise 3.2 Solve
𝑥𝑦 − 𝑦 2
𝑦′ = .
𝑥2

35 Lecture Notes for MA451

 3.1 Substitution Methods

Solution

Step 1: The equation is homogeneous of first order because

(𝑡𝑥)(𝑡𝑦) − (𝑡𝑦)2 𝑡 2 (𝑥𝑦 − 𝑦 2 ) 𝑥𝑦 − 𝑦 2

= = .
(𝑡𝑥)2 𝑡 2𝑥 2 𝑥2

Step 2: Set 𝑦 = 𝑥𝑣. Rewriting the function by substitution implies

𝑥𝑦 − 𝑦 2 𝑥(𝑥𝑣) − (𝑥𝑣)2 𝑥 2 𝑣 − 𝑥 2 𝑣 2
= = = 𝑣 − 𝑣2.
𝑥2 𝑥2 𝑥2
Step 3: Differentiating 𝑦 = 𝑥𝑣 with respect to 𝑥 gives

𝑦 ′ = 𝑣 + 𝑥𝑣 ′ .

Step 4: The original equation can be transformed into

𝑣 + 𝑥𝑣 ′ = 𝑣 − 𝑣 2 .

Step 5: Solving the equation for 𝑣 yields

𝑣 + 𝑥𝑣 ′ =𝑣 − 𝑣 2
𝑥𝑣 ′ = − 𝑣 2
𝑣′ 1
− =
𝑣2 𝑥
1 ′ 1
(𝑣 ) =𝑥
1
= ln |𝑥| + 𝑐
𝑣
1
𝑣= .
ln |𝑥| + 𝑐
𝑥
Step 6: Substituting 𝑣 = 𝑦
and solving for 𝑦 gives the general solution.

𝑥
𝑦= .
ln |𝑥| + 𝑐

3.1.2 Linear substitution*

For a first order differential equation 𝐹 (𝑥, 𝑦, 𝑦 ′ ) = 0, sometimes, a substitution 𝑦 = 𝑢(𝑥, 𝑣),
where 𝑣 = 𝑣(𝑥, 𝑦) is a function that is linear in 𝑦, may reduce the equation into a new equation
that is much easier to solve.

Here are a few classes of equations that can be reduced to separable equations using a

36 Lecture Notes for MA451

 3.1 Substitution Methods

linear substitution.

If the equation is in the form 𝑦 ′ = 𝑓 (𝑎𝑥 + 𝑏𝑦 + 𝑐), then the substitution 𝑣 = 𝑎𝑥 + 𝑏𝑦 + 𝑐

reduces the equation into a separable equation 𝑣 ′ = 𝑓 (𝑣) − 𝑎.

Example 3.3 Solve the equation

𝑦 ′ = (2𝑥 + 𝑦 − 3)2 .

Solution Let 𝑣 = 2𝑥 + 𝑦 − 3. Differentiating it with respect to 𝑥 yields

𝑣′ = 2 + 𝑦 ′,

or equivalently
𝑦 ′ = 𝑣 ′ − 2.

Then substituting 2𝑥 + 𝑦 − 3 by the function 𝑣 yields

𝑣′ − 2 = 𝑣2

or equivalently,
𝑣′
= 1.
𝑣2 + 2
Integrating both sides yields

1 𝑣
√ arctan( √ ) =𝑥 + 𝐶
2 2
𝑣 √ √
arctan( √ =𝑥 2 + 𝐶 2
2
𝑣 √ √
√ = tan(𝑥 2 + 𝐶 (2))
2
√ √ √
𝑣 = 2 tan(𝑥 2 + 𝐶 2).
√
Replacing 𝐶 2 by 𝑐 and 𝑣 by 𝑥 + 𝑦 − 3, and solving for 𝑦 gives the general solution
√ √
𝑦= 2 tan(𝑥 2 + 𝑐) − 2𝑥 + 3.

■
If the equation is in the form 𝑥𝑦 ′ = 𝑦𝐹 (𝑥𝑦), then the substitution 𝑣 = 𝑥𝑦 reduces the
equation into a separable equation 𝑣 ′ = 𝑣𝑥 (𝐹 (𝑣) + 1).

Example 3.4 Solve the equation

𝑥𝑦 ′ = 𝑥𝑦 2 − 𝑦.

37 Lecture Notes for MA451

 3.2 Exact Equations
𝑣
Solution Let 𝑣 = 𝑥𝑦. Then 𝑥𝑦 ′ = 𝑣 ′ − 𝑦. Substituting 𝑦 = 𝑥
yields

𝑣 ′ − 𝑦 =𝑣𝑦 − 𝑦
𝑣 ′ =𝑣𝑦
𝑣2
𝑣′ =
𝑥
𝑣′ 1
= .
𝑣2 𝑥

Integrating both sides gives a solution

1
− = ln(|𝑥|) + 𝑐.
𝑣

Replacing 𝑣 by 𝑥𝑦 implies
1
𝑦=− .
𝑥(ln |𝑥| + 𝑐)
■

3.2 Exact Equations

Using implicit differentiation, one can show that 𝐹 (𝑥, 𝑦) = 𝑐 (with c as a constant) is an
implicit solution of the differential equation

𝜕 𝜕
𝐹 (𝑥, 𝑦) d 𝑥 + 𝐹 (𝑥, 𝑦) d 𝑦 = 0.
𝜕𝑥 𝜕𝑦

This observation suggests an approach to solve so-called exact equations.

Definition 3.3
A first order differential equation

𝑀(𝑥, 𝑦) d 𝑥 + 𝑁 (𝑥, 𝑦) d 𝑦 = 0

is said to be exact if there is a function 𝐹 (𝑥, 𝑦) such that

𝜕 𝜕
𝐹 (𝑥, 𝑦) = 𝐹 (𝑥, 𝑦) = 𝑀(𝑥, 𝑦)
𝜕𝑥 𝜕𝑥
𝜕 𝜕
𝐹 (𝑥, 𝑦) = 𝐹 (𝑥, 𝑦) = 𝑁 (𝑥, 𝑦).
𝜕𝑦 𝜕𝑦
♣

The above definition not only defines exact functions but also gives solutions. However,
it is not a very practical criterion. Given an equation of the form

𝑀(𝑥, 𝑦) d 𝑥 + 𝑁 (𝑥, 𝑦) d 𝑦 = 0,

38 Lecture Notes for MA451

 3.2 Exact Equations

how do we know it is exact? Knowing the equation is exact, how do we find 𝐹 ?

The first question is answered by the following theorem. The proof of the theorem will
also answer the second question.
Theorem 3.2 (Exactness Condition)
𝜕𝑀 𝜕𝑁
Suppose that 𝑀 and 𝑁 are continuous and have partial derivatives 𝜕𝑦
and 𝜕𝑥
on an open
rectangle 𝑅. Then the equation

𝑀(𝑥, 𝑦) d 𝑥 + 𝑁 (𝑥, 𝑦) d 𝑦 = 0

is exact if and only if

𝜕 𝜕
𝑁 (𝑥, 𝑦) = 𝑀(𝑥, 𝑦).
𝜕𝑥 𝜕𝑦
on 𝑅. ♥
𝜕𝐹 𝜕𝐹
Proof If the equation is exact, then there exists a function 𝐹 (𝑥, 𝑦) such that 𝜕𝑥
= 𝑀 and 𝜕𝑦
= 𝑁.
Therefore, by the fact that
𝜕 𝜕 𝜕 𝜕
𝐹= 𝐹,
𝜕𝑦 𝜕𝑥 𝜕𝑦 𝜕𝑥
we know that
𝜕 𝜕
𝑁 (𝑥, 𝑦) = 𝑀(𝑥, 𝑦).
𝜕𝑥 𝜕𝑦

Conversely, suppose that 𝜕𝑥𝜕 𝑁 (𝑥, 𝑦) = 𝜕𝑦𝜕 𝑀(𝑥, 𝑦). We can find a 𝐹 such that 𝜕𝐹
𝜕𝑥
= 𝑀 and
𝜕𝐹
𝜕𝑦
= 𝑁 . Integrating both sides of the equation 𝜕𝐹𝜕𝑥
= 𝑀 yields that

𝐹 (𝑥, 𝑦) = ∫ 𝑀 d 𝑥 + 𝑔(𝑦),

where 𝑔 is a single variable function. Because 𝜕𝐹

𝜕𝑦
= 𝑁 . The function 𝑔 must satisfies the
equation
𝜕𝐹 𝜕
= 𝑀 d 𝑥 + 𝑔(𝑦)
𝜕𝑦 𝜕𝑦 (∫ )
𝜕
𝑁 = ∫ 𝑀 d 𝑥 + 𝑔 ′ (𝑦).
𝜕𝑦
This yields
𝜕
𝑔(𝑦) = 𝑁 − 𝑀 d𝑥 .
𝜕𝑦 (∫ )
Because
𝜕 𝜕𝑀
∫ 𝑀 d𝑥 = ∫ d 𝑥.
𝜕𝑦 ( ) 𝜕𝑦
Then
𝜕𝐹 𝜕𝑀 𝜕𝑁
=∫ d𝑥 = ∫ d 𝑥 = 𝑁 (𝑥, 𝑦).
𝜕𝑦 𝜕𝑦 𝜕𝑥

39 Lecture Notes for MA451

 3.2 Exact Equations

This shows that

𝜕𝐹 𝜕𝐹
𝑀 d𝑥 + 𝑁 d𝑥 = d𝐹 = d𝑥 + d 𝑦.
𝜕𝑥 𝜕𝑦
Hence, 𝑀 d 𝑥 + 𝑁 d 𝑥 = 0 is exact. ■
The definition of exact equations suggests the following approach to find 𝐹 such that
𝐹 (𝑥, 𝑦) = 𝑐 is the general solution.
𝜕𝐹
Step 1: Integrating both sides of the equation 𝜕𝑥
= 𝑀 yields

𝐹 (𝑥, 𝑦) = ∫ 𝑀 d 𝑥 + 𝑔(𝑦),

where 𝑔 is a singular variable function.

Step 2: The equation 𝜕𝐹𝜕𝑦
= 𝑁 together with 𝐹 (𝑥, 𝑦) = ∫ 𝑀 d 𝑥 + 𝑔(𝑦) implies that 𝑔 satisfies the
following equation to obtain a function

𝜕
𝑀 d 𝑥 + 𝑔 ′ (𝑦) = 𝑁 .
𝜕𝑦 (∫ )

Solve for 𝑔.
Step 3: The equation 𝐹 (𝑥, 𝑦) = 𝑐 gives the general solution of 𝑀 d 𝑥 + 𝑁 d 𝑦 = 0.

Example 3.5 Check whether the equation

(sin 𝑥 + 𝑦) d 𝑥 + (𝑒 𝑦 + 𝑥) d 𝑦 = 0

is exact.

Solution The coefficient function of d 𝑥 is 𝑀(𝑥, 𝑦) = sin 𝑥 + 𝑦. Taking its partial derivative with
respect to 𝑦 yields
𝜕
𝑀(𝑥, 𝑦) = 1.
𝜕𝑦
The coefficient function of d 𝑥 is 𝑁 (𝑥, 𝑦) = 𝑒 𝑦 + 𝑥. Taking its partial derivative with respect to 𝑥
yields
𝜕
𝑁 (𝑥, 𝑦) = 1.
𝜕𝑥
Because 𝜕
𝜕𝑦
𝑀(𝑥, 𝑦) = 𝜕
𝜕𝑥
𝑁 (𝑥, 𝑦). By Theorem 3.2, the equation is exact. ■
 Exercise 3.3 Check whether the equation

(sin 𝑥 + 𝑥𝑦) d 𝑥 + (𝑒 𝑦 + 𝑥𝑦) d 𝑦 = 0

is exact.

40 Lecture Notes for MA451

 3.2 Exact Equations

Solution Because
𝜕 𝜕
𝑀(𝑥, 𝑦) = (sin 𝑥 + 𝑥𝑦) = 𝑥,
𝜕𝑦 𝜕𝑦
𝜕 𝜕
𝑁 (𝑥, 𝑦) = (𝑒 𝑦 + 𝑥𝑦) = 𝑦,
𝜕𝑥 𝜕𝑥
and they are not equal. The equation is NOT exact. ■
The inverse direction will become clear after we answer the second question: How to
find 𝐹 (𝑥, 𝑦)?

The definition of exactness suggests that 𝐹 is a solution for both 𝜕𝐹

𝜕𝑥
= 𝑀 and 𝜕𝐹
𝜕𝑦
= 𝑁 . So
the idea to solve is to integrate on the those two equations for 𝐹 and plug in the other to
determined the undetermined single variable function.

Example 3.6 Solve

(4𝑥 3 𝑦 3 + 3𝑥 2 ) d 𝑥 + (3𝑥 4 𝑦 2 + 6𝑦 2 ) d 𝑦 = 0.

Solution

Step 1: Check exactness.

Because
𝑀(𝑥, 𝑦) = 4𝑥 3 𝑦 3 + 3𝑥 2 , 𝑁 (𝑥, 𝑦) = 3𝑥 4 𝑦 2 + 6𝑦 2 .

Then
𝜕 𝜕
𝑀(𝑥, 𝑦) = 𝑁 (𝑥, 𝑦) = 12𝑥 3 𝑦 2
𝜕𝑦 𝜕𝑥
for all (𝑥, 𝑦). Therefore, by the exactness condition theorem, there’s a function 𝐹 such that

𝜕
𝐹 (𝑥, 𝑦) = 𝑀(𝑥, 𝑦) = 4𝑥 3 𝑦 3 + 3𝑥 2
𝜕𝑥
and
𝜕
𝐹 (𝑥, 𝑦) = 𝑁 (𝑥, 𝑦) = 3𝑥 4 𝑦 2 + 6𝑦 2
𝜕𝑦
for all (𝑥, 𝑦).
Step 2: Integrate 𝐹𝑥 or 𝐹𝑦 .
To find 𝐹 , we integrate Equation 𝐹𝑥 (𝑥, 𝑦) = 4𝑥 3 𝑦 3 + 3𝑥 2 with respect to 𝑥 to obtain

𝐹 (𝑥, 𝑦) = 𝑥 4 𝑦 3 + 𝑥 3 + 𝑔(𝑦),

where 𝑔(𝑦) is the “constant term" of integration with respect to 𝑥.

Step 3: Differentiate 𝐹 .
To determine 𝑔 so that 𝐹 also satisfies the equation 𝜕𝑦𝜕 𝐹 (𝑥, 𝑦) = 3𝑥 4 𝑦 2 + 6𝑦 2 , assume that 𝑔
is differentiable and differentiate 𝐹 (𝑥, 𝑦) = 𝑥 4 𝑦 3 + 𝑥 3 + 𝑔(𝑦) with respect to 𝑦. That gives

𝜕
𝐹 (𝑥, 𝑦) = 3𝑥 4 𝑦 2 + 𝑔 ′ (𝑦).
𝜕𝑦

41 Lecture Notes for MA451

 3.2 Exact Equations

Step 4: Determine the equation for 𝑔 ′ .

𝜕
Comparing this equation with the equation 𝜕𝑦
𝐹 (𝑥, 𝑦) = 3𝑥 4 𝑦 2 + 6𝑦 2 shows that

𝑔 ′ (𝑦) = 6𝑦 2 .

Step 5: Integrate 𝑔 ′ .
Integrating this equation with respect to 𝑦 yields

𝑔(𝑦) = 2𝑦 3 .

Note that here we take the constant of integration to be zero because adding a con-
stant won’t change the general implicit solution 𝐹 (𝑥, 𝑦) = 𝑐.
Step 6: Find the general implicit solution.
Substituting 𝑔 in 𝐹 (𝑥, 𝑦) using this equation yields

𝐹 (𝑥, 𝑦) = 𝑥 4 𝑦 3 + 𝑥 3 + 2𝑦 3 + 𝐶.

Now Theorem 3.2 implies that

𝑥 4 𝑦 3 + 𝑥 3 + 2𝑦 3 = 𝑐

is an implicit solution of Equation.

Step 7: Find the explicit solution(s).
Solving this for 𝑦 yields the explicit solution
1/3
𝑐 − 𝑥3
𝑦= + 𝑥4 .
( 2 )

■
 Exercise 3.4 Consider the equation

(2𝑥 + 𝑦) d 𝑥 + (2𝑦 + 𝑥) d 𝑦 = 0.

1. Verify the equation is exact,

2. Find the general solution.

Solution Here 𝑀(𝑥, 𝑦) = 2𝑥 + 𝑦 and 𝑁 (𝑥) = 2𝑦 + 𝑥. Taking partial derivatives yields that

𝜕𝑀 𝜕𝑁
=1= .
𝜕𝑦 𝜕𝑥

Hence, by the exactness condition theorem, the equation is exact.

42 Lecture Notes for MA451

 3.2 Exact Equations

To find 𝐹 (𝑥, 𝑦), integrating 𝑀 = 2𝑥 + 𝑦 with respect to 𝑥 to get

2 2
∫ 𝑀(𝑥, 𝑦) d 𝑥 = 𝑥 + 𝑦𝑥 = ∫ (2𝑥 + 𝑦) d 𝑥 = 𝑥 + 𝑦𝑥.

Note that 𝑦 is treated as a number in the above integral since the left hand side is partial
derivative of 𝑥.

Hence
𝐹 (𝑥, 𝑦) = 𝑥 2 + 𝑥𝑦 + 𝑔(𝑦),

where 𝑔(𝑦) is the constant with respect to 𝑥.

Differentiating 𝐹 with respect to 𝑦 and equating with 𝑁 (𝑥, 𝑦) yields

𝜕𝐹
= 𝑥 + 𝑔 ′ (𝑦) = 2𝑦 + 𝑥.
𝜕𝑦

Hence
𝑔 ′ (𝑦) = 2𝑦,

and a particular solution

𝑔(𝑦) = 𝑦 2 .

Therefore, 𝐹 can be taken to be 𝐹 (𝑥, 𝑦) = 𝑥 2 + 𝑥𝑦 + 𝑦 2 and the general solution is

𝑥 2 + 𝑥𝑦 + 𝑦 2 = 𝑐.

43 Lecture Notes for MA451

 Week 4: Integrating Factors and Applications

9/27–9/30

4.1 Integrating Factors for Non-exact Equations

Suppose
𝑀(𝑥, 𝑦) d 𝑥 + 𝑁 (𝑥, 𝑦) d 𝑦 = 0 (4.1)

is not exact. How to solve the equation? If it is a separable or linear first order, we already
know how to solve. One of the method is to use integrating factors. It seems reasonable to
multiplying a non-zero factor to both sides of Equation (4.1) to get an exact equation.
Definition 4.1
A function 𝜇(𝑥, 𝑦) is called an integrating factor of Equation (4.1) if the following equation
is exact after multiplying it to both sides of Equation (4.1)

𝜇(𝑥, 𝑦)𝑀(𝑥, 𝑦) d 𝑥 + 𝜇(𝑥, 𝑦)𝑁 (𝑥, 𝑦) d 𝑦 = 0.

♣

Remark It is possible to lose or gain solutions when multiplying by an integrating factor. In

general, when using integrating factors, you should check whether any solution to 𝜇(𝑥, 𝑦) = 0
is in fact a solution to the original differential equation (4.1).

Do integrating factors always exist?

This question is hard to answer in general. However, if we assume Equation (4.1) has a
general solution 𝐹 , then the integrating factor exists and must be the ratio function

𝜕𝐹 𝜕𝐹
𝜕𝑥 𝜕𝑦
𝜇(𝑥, 𝑦) = = .
𝑀 𝑁

How to construct integrating factors?

By Theorem 3.2, a nonzero function 𝜇(𝑥, 𝑦) is an integrating factor if and only if it satisfies
the equation
𝜕 𝜕
( 𝜇(𝑥, 𝑦)𝑀(𝑥, 𝑦)) = 𝜇(𝑥, 𝑦)𝑁 (𝑥, 𝑦)),
𝜕𝑦 𝜕𝑥 (
or equivalently
𝜕𝜇 𝜕𝜇 𝜕𝑁 𝜕𝑀
𝑀 −𝑁 = − 𝜇. (4.2)
𝜕𝑦 𝜕𝑥 ( 𝜕𝑥 𝜕𝑦 )

Equation 4.2 in general is even harder than the original equation to solve. Nevertheless,
there are a few situations where Equation 4.2 can be solved relatively easily.
 4.1 Integrating Factors for Non-exact Equations
𝜕𝜇 𝜕𝜇
Clearly, if either 𝜕𝑥 = 0 or 𝜕𝑦 = 0, or equivalently, 𝜇(𝑥, 𝑦) = 𝑞(𝑦) or 𝜇(𝑥, 𝑦) = 𝑝(𝑥), then the
above Equation 4.2 can be solved easily. That leads to the following proposition.
Proposition 4.1 (Constructing Integrating Factor)
𝜕𝑀 𝜕𝑁
Given a differential equation 𝑀(𝑥, 𝑦) d 𝑥 + 𝑁 (𝑥, 𝑦) d 𝑦 = 0, suppose that 𝑀, 𝑁 , 𝜕𝑥
and 𝜕𝑦
are
continuous. Then
Type I: If
1 𝜕𝑀 𝜕𝑁
− = 𝑝(𝑥)
𝑁 (𝑥, 𝑦) ( 𝜕𝑥 𝜕𝑦 )
is a function only in 𝑥, then we take integrating factor as a function of 𝑥,

𝑝(𝑥) d 𝑥
𝜇(𝑥, 𝑦) = 𝑒 ∫ ;

Type II: If
1 𝜕𝑀 𝜕𝑁
− = 𝑞(𝑦)
(
𝑀(𝑥, 𝑦) 𝜕𝑥 𝜕𝑦 )
is a function only in 𝑦, then we take integrating factor as a function of 𝑦,

−𝑞(𝑦) d 𝑦
𝜇(𝑥, 𝑦) = 𝑒 ∫ .
♠

Remark There are also some not so-obvious-cases that integrating factors can be constructed.

If 𝑢(𝑥, 𝑦) = 𝑓 (𝑤(𝑥, 𝑦)), where 𝑓 is a single variable function, then an integrating factor is

𝜒 (𝑤) d 𝑤
𝜇(𝑥, 𝑦) = 𝑒 ∫ ,

where
𝜕𝑀 𝜕𝑁
𝜕𝑦
− 𝜕𝑥
𝜒 (𝑤) = 𝜕𝑤 .
𝑁 𝜕𝑥
− 𝑀 𝜕𝑤
𝜕𝑦

If 𝑀 and 𝑁 are both homogeneous functions of the same degree, then an integrating
factor is
1
𝜇(𝑥, 𝑦) = .
𝑥𝑀(𝑥, 𝑦) + 𝑦𝑁 (𝑥, 𝑦)

If 𝑀 = 𝑦𝑝(𝑥𝑦), 𝑁 = 𝑥𝑞(𝑥𝑦), and 𝑝(𝑥𝑦) ≠ 𝑞(𝑥𝑦), then an integrating factor is

1
𝜇(𝑥, 𝑦) = .
𝑥𝑀 − 𝑦𝑁

For more information, see the section on Integrating Factors by Vladimir Dobrushkin.

Example 4.1 Consider the equation

(2𝑥 2 + 𝑦𝑥) d 𝑥 + (2𝑥𝑦 + 𝑥 2 ) d 𝑦 = 0.

45 Lecture Notes for MA451

 4.1 Integrating Factors for Non-exact Equations

1. Show that the equation is not exact;

𝜕𝑀
𝜕𝑦
− 𝜕𝑁
𝜕𝑥
2. Show that depends only on 𝑥;
𝑁
3. Reduce the equation to an exact equation.

Solution

1. Because 𝑀(𝑥, 𝑦) = 2𝑥 2 + 𝑥𝑦 and 𝑁 (𝑥, 𝑦) = 2𝑥𝑦 + 𝑥 2 . The partial derivatives are

𝜕𝑀
=𝑥,
𝜕𝑦
𝜕𝑁
=2𝑦 + 2𝑥.
𝜕𝑥

Hence 𝜕𝑀
𝜕𝑦
≠ 𝜕𝑁
𝜕𝑥
, and the equation is not exact.
2. We compute
𝜕𝑀
𝜕𝑦
− 𝜕𝑁
𝜕𝑥 𝑥 − (2𝑦 + 2𝑥) −𝑥 − 2𝑦 1
= 2
= =− ,
𝑁 2𝑥𝑦 + 𝑥 𝑥(2𝑦 + 𝑥) 𝑥
which depends only on 𝑥.
3. By Proposition 4.1, We find an integrating factor of type I,
𝜕𝑀 − 𝜕𝑁
𝜕𝑦 𝜕𝑥
d𝑥 − 𝑥1 d 𝑥 1 1
𝜇=𝑒 ∫ 𝑁 = 𝑒∫ = 𝑒 − ln 𝑥 = = .
𝑒 ln 𝑥 𝑥
Multiplying the equation by the integrating factor 𝜇 yields

1
((2𝑥 2 + 𝑦𝑥) d 𝑥 + (2𝑥𝑦 + 𝑥 2 ) d 𝑦) =0
𝑥
(2𝑥 + 𝑦) d 𝑥 + (2𝑦 + 𝑥) d 𝑦 =0.

This equation is exact because

𝜕 𝜕
(2𝑥 + 𝑦) = 1 = (2𝑦 + 𝑥).
𝜕𝑦 𝜕𝑥

■
Example 4.2 Consider the equation

𝑦 d 𝑥 + (𝑦 − 𝑥) d 𝑦 = 0.
𝜕𝑀 𝜕𝑁
𝜕𝑦
− 𝜕𝑥
1. Show that the equation is not exact, and depends only on 𝑦.
𝑀
2. Reduce the equation to an exact equation.
3. Solve the equation.

Solution

46 Lecture Notes for MA451

 4.1 Integrating Factors for Non-exact Equations

1. Because 𝑀(𝑥, 𝑦) = 𝑦, 𝑁 (𝑥, 𝑦) = 𝑦 − 𝑥 and

𝜕𝑀
=1
𝜕𝑦
𝜕𝑁
=−1
𝜕𝑥

The equation is not exact but

𝜕𝑀 𝜕𝑁
𝜕𝑦
− 𝜕𝑥 2
=
𝑀 𝑦
depends only on 𝑦.
2. By Proposition 4.1, we look for an integrating factor of type II,
𝜕𝑀 − 𝜕𝑁 −2
𝜕𝑦 𝜕𝑥 2 1
𝐼 = 𝑒∫ − 𝑀 d𝑦
= 𝑒 − ∫ 𝑦 d 𝑥 = 𝑒 −2 ln 𝑦 = 𝑒 ln(𝑥) = 𝑦 −2 = 2 .
( ) 𝑦

Multiplying 𝜇 to the non-exact equation yields

1
(𝑦 d 𝑥 + (𝑦 − 𝑥) d 𝑦) =0
𝑦2
1 𝑦−𝑥
d 𝑥 + 2 d 𝑦 =0
𝑦 𝑦

This is an exact equation because

𝜕 1 1
=−
𝜕𝑦 𝑦 )
( 𝑦2
𝜕 𝑦−𝑥 1
=−
𝜕𝑥 ( 𝑦 2 ) 𝑦2

3. We look for 𝐹 such that

𝜕𝐹 𝜕𝐹 1 𝑦−𝑥
d𝑥 + d 𝑦 = d 𝑥 + 2 d 𝑦.
𝜕𝑥 𝜕𝑦 𝑦 𝑦

Equivalently,
⎧
⎪
⎪ 𝜕𝐹
𝜕𝑥
= 𝑦1
⎨
⎪
⎪ 𝜕𝐹
= 𝑦−𝑥
𝑦2
⎩ 𝜕𝑦
Integrating the first equation with respect to 𝑥 implies

1 𝑥
𝐹 (𝑥, 𝑦) = ∫ d 𝑥 = + ℎ(𝑦).
𝑦 𝑦

Since 𝐹 has to be a solution solves the second equation and

𝜕𝐹 𝜕 𝑥 𝑥
= + ℎ(𝑦) = − 2 + ℎ′ (𝑦),
𝜕𝑦 𝜕𝑦 ( 𝑦 ) 𝑦

47 Lecture Notes for MA451

 4.1 Integrating Factors for Non-exact Equations

the function ℎ(𝑦) must satisfy the equation

𝑥 ′ 𝑦−𝑥
− + ℎ (𝑦) = ,
𝑦2 𝑦2

or equivalently,
1
ℎ′ (𝑦) = .
𝑦
Therefore, a solution is ℎ = ln 𝑦. Hence,
𝑥
𝐹 (𝑥, 𝑦) = + ln 𝑦
𝑦

and the general solution is

𝑥
𝐹 (𝑥, 𝑦) = + ln 𝑦 = 𝑐.
𝑦
■
 Exercise 4.1 Consider the equation

−𝑦 d 𝑥 + 𝑥 d 𝑦 = 0.
𝜕𝑀 𝜕𝑁
𝜕𝑦
− 𝜕𝑥
1. Show that the equation is not exact, and depends only on 𝑥.
𝑁
2. Reduce the equation to an exact equation.
3. Solve the equation.

Solution

1. Because 𝑀(𝑥, 𝑦) = −𝑦, 𝑁 (𝑥, 𝑦) = 𝑥 and

𝜕𝑀
=−1
𝜕𝑦
𝜕𝑁
=1
𝜕𝑥
The equation is not exact but
𝜕𝑀 𝜕𝑁
𝜕𝑦
− 𝜕𝑥 −2
=
𝑁 𝑥
depends only on 𝑥.
2. By Proposition 4.1, we look for an integrating factor of type I,
𝜕𝑀 − 𝜕𝑁
𝜕𝑦 𝜕𝑥
d𝑥 −2 1
𝐼 = 𝑒∫ 𝑁 = 𝑒∫ 𝑥 d 𝑥 = 𝑒 −2 ln 𝑥 = (𝑒 ln(𝑥) )−2 = 𝑥 −2 = .
𝑥2
Multiplying the equation by 𝜇 implies

1
(−𝑦 d 𝑥 + 𝑥 d 𝑦) = 0
𝑥2

48 Lecture Notes for MA451

 4.1 Integrating Factors for Non-exact Equations

which is
𝑦 1
− 2
d𝑥 + d𝑦 = 0
𝑥 𝑥
This is an exact equation because

𝜕 𝑦 1
(− 2) = −
𝜕𝑦 𝑥 𝑥2
𝜕 1 1
=−
𝜕𝑥 ( 𝑥 ) 𝑥2
3. We look for 𝐹 such that
𝜕𝐹 𝜕𝐹 𝑦 1
d𝑥 + d 𝑦 = − 2 d 𝑥 + d 𝑦,
𝜕𝑥 𝜕𝑦 𝑥 𝑥

Equivalently,
⎧
⎪
⎪ 𝜕𝐹
𝜕𝑥
= − 𝑥𝑦2
⎨
⎪ 𝜕𝐹
⎪ = 𝑥1
⎩ 𝜕𝑦
Integrating the first equation with respect to 𝑥 implies

𝑦 𝑦
𝐹 (𝑥, 𝑦) = ∫ − 2
d 𝑥 = + ℎ(𝑦)
𝑥 𝑥

Since 𝐹 has to satisfy the second equation and

𝜕𝐹 𝜕 𝑦 1
= ( + ℎ(𝑦)) = + ℎ′ (𝑦)
𝜕𝑦 𝜕𝑦 𝑥 𝑥

the function ℎ(𝑦) satisfies the equation

1 1
+ ℎ′ (𝑦) =
𝑥 𝑥
or
ℎ′ (𝑦) = 0

and ℎ = 0. Hence,
𝑦
𝐹 (𝑥, 𝑦) =
𝑥
and the general solution is
𝑦
𝐹 (𝑥, 𝑦) = =𝑐
𝑥
Equivalently,
𝑦 = 𝑐𝑥.

49 Lecture Notes for MA451

 4.1 Integrating Factors for Non-exact Equations

 Exercise 4.2 Consider the equation

(𝑥 − 𝑦) d 𝑥 + 𝑥 d 𝑦 = 0.
𝜕𝑀 𝜕𝑁
𝜕𝑦
− 𝜕𝑥
1. Show the equation is not exact but depends only on 𝑥.
𝑁
2. Reduce the equation to an exact equation.
3. Solve the equation

Solution

1. 𝑀 = 𝑥 − 𝑦, 𝑁 = 𝑥 and

𝜕𝑀
=−1
𝜕𝑦
𝜕𝑁
=1
𝜕𝑥
The equation is not exact but
𝜕𝑀 𝜕𝑁
𝜕𝑦
− 𝜕𝑥 −2
=
𝑁 𝑥
depends only on 𝑥.
2. By Proposition 4.1, we look for an integrating factor of type I,
𝜕𝑀 − 𝜕𝑁
𝜕𝑦 𝜕𝑥
d𝑥 −2
d𝑥 1
𝜇 = 𝑒∫ 𝑁 = 𝑒∫ 𝑥 = .
𝑥2
Multiplying the equation by 𝜇 implies

1
((𝑥 − 𝑦) d 𝑥 + 𝑥 d 𝑦) = 0
𝑥2
which is
𝑥 −𝑦 1
2
d𝑥 + d𝑦 = 0
𝑥 𝑥
This is an exact equation because

𝜕 𝑥 −𝑦 −1
=
𝜕𝑦 ( 𝑥 2 ) 𝑥 2
𝜕 1 −1
( ) = 2
𝜕𝑥 𝑥 𝑥
3. We look for 𝐹 such that
𝜕𝐹 𝑥 − 𝑦
= 2
𝜕𝑥 𝑥
𝜕𝐹 1
=
𝜕𝑦 𝑥

50 Lecture Notes for MA451

 4.2 Existence and Uniqueness*

Integrating the first equation with respect to 𝑥 implies

𝑥 −𝑦 1 𝑦 𝑦
𝐹 (𝑥, 𝑦) = ∫ 2
d𝑥 = ∫ − 2 d 𝑥 = ln 𝑥 + + ℎ(𝑦)
𝑥 𝑥 𝑥 𝑥

Since 𝐹 has to satisfy the second equation and

𝜕𝐹 𝜕 𝑦 1
= ln 𝑥 + + ℎ(𝑦) = + ℎ′ (𝑦),
𝜕𝑦 𝜕𝑦 ( 𝑥 ) 𝑥

the function ℎ(𝑦) satisfies the equation

1 1
+ ℎ′ (𝑦) =
𝑥 𝑥
or
ℎ′ (𝑦) = 0

and ℎ = 0. Hence,
𝑦
𝐹 (𝑥, 𝑦) = ln 𝑥 +
𝑥
and the general solution is
𝑦
ln 𝑥 + = 𝑐,
𝑥
or
𝑦 = 𝑐𝑥 − 𝑥 ln 𝑥.

4.2 Existence and Uniqueness*

Solving differential equations can be very complicated. It is impossible to find useful
formulas for the solutions of most differential equations. However, knowing the existence
and uniqueness can helps us looking for solutions.
Theorem 4.1 (Picard’s Existence and Uniqueness)
1. If 𝑓 is a function continuous on an open rectangle 𝑅 ∶ {𝑎 < 𝑥 < 𝑏, 𝑐 < 𝑦 < 𝑑} that
contains (𝑥0 , 𝑦0 ), then the initial value problem

𝑦 ′ = 𝑓 (𝑥, 𝑦), 𝑦(𝑥0 ) = 𝑦0

has at least one solution on some open subinterval of (𝑎, 𝑏) that contains 𝑥0 .
2. If both 𝑓 and 𝑓𝑦 are continuous on 𝑅 then the initial value problem

𝑦 ′ = 𝑓 (𝑥, 𝑦), 𝑦(𝑥0 ) = 𝑦0

51 Lecture Notes for MA451

 4.2 Existence and Uniqueness*

has a unique solution on some open subinterval of (𝑎, 𝑏) that contains 𝑥0

♥

A key to the proof of this theorem is the fact that that a continuously differential function
𝑦(𝑥) is a solution of the differential equation if and only if it satisfies the following integral
equation:
𝑥
𝑦(𝑥) = 𝑦0 + ∫ 𝑓 (𝑡, 𝑦(𝑡)) d 𝑡.
𝑥0

To prove the theorem, the French mathematician Éile Picard used a sequence of approxi-
mations with 𝑦0 (𝑥) = 𝑦0 and
𝑥
𝑦𝑛 (𝑥) = 𝑦0 + ∫ 𝑓 (𝑡, 𝑦𝑛−1 (𝑡)) d 𝑡.
𝑥0

This approximation is known as Picard’s method of successive approximations. An interactive

demonstration can be found on GeoGebra: Picard’s Method of Successive Approximations.

The theorem can be proved by showing that 𝑦𝑛 (𝑥) converges uniformly to a function 𝑦(𝑥)
which is a solution. We refer the reader to (Simmons 2016, Chapter 13) for a proof.

Example 4.3 Consider the initial value problem

𝑦 ′ = 3𝑦 2/3 , 𝑦(0) = 0.

Show that both 𝑦 = 0 and 𝑦 = 𝑥 3 are solutions.

Does it contradict Picard’s Existence and Uniqueness Theorem?

Solution It is clear that 𝑦 = 0 is a solution. The function 𝑦 = 𝑥 3 is also a solution because 𝑦(0) =
03 = 0 and
2 2
𝑦 ′ = 3𝑥 2 = 2(𝑥 3 ) 3 = 3𝑦 3 .

This example does not contradict the uniqueness of the Theorem. Because the partial deriva-
1
tive 𝑓𝑦 = 2𝑦 − 3 is not continuous along the line (𝑥, 0). ■
Example 4.4 Consider the initial value problem
1
𝑦 ′ = 3𝑥𝑦 3 , 𝑦(𝑥0 ) = 𝑦0 .

1. For what points (𝑥0 , 𝑦0 ) does Picard’s Existence and Uniqueness Theorem imply that this
initial value problem has a solution?
2. For what points (𝑥0 , 𝑦0 ) does Picard’s Existence and Uniqueness Theorem imply that this
initial value problem has a unique solution on some open interval that contains 𝑥0 ?
1
Solution Because 𝑓 (𝑥, 𝑦) = 3𝑥𝑦 3 is continuous for all points (𝑥, 𝑦). The theorem implies that the
initial value problem has a solution.

52 Lecture Notes for MA451

 4.3 Applications of First Order Differential Equations

The partial derivative

𝜕 1 𝜕 1 2
𝑓𝑦 = ( 3𝑥𝑦 3 ) = 3𝑥 𝑦 3 = 𝑥𝑦 − 3
𝜕𝑦 𝜕𝑦
is undefined when 𝑦 = 0. For any point (𝑥, 𝑦) such that 𝑦 ≠ 0, 𝑓𝑦 is continuous. Therefore, by the
theorem, the initial value problem has a unique solution for 𝑦 ≠ 0.

Indeed, when 𝑦0 = 0, the initial value problem has a trivial solution 𝑦 = 0 and an implicit
2 2
solution 𝑦 3 = 𝑥 2 − 𝑥02 . If 𝑦0 ≠ 0, only 𝑦 3 = 𝑥 2 − 𝑥02 is a solution. ■
 Exercise 4.3 Find all (𝑥0 , 𝑦0 ) for which Picard’s Existence and Uniqueness Theorem implies
that the initial value problem
𝑥
𝑦 ′ = , 𝑦(𝑥0 ) = 𝑦0
𝑦
has 1. a solution and 2. a unique solution on some open interval that contains 𝑥0 .

Solution Since 𝑓 (𝑥, 𝑦) = 𝑦𝑥 is continuous only when 𝑦 ≠ 0. The theorem implies that the initial
value problem has a solution when 𝑦0 ≠ 0.

The partial derivative

𝜕 𝑥 𝜕 1 𝑥
𝑓𝑦 = =𝑥 =− 2
𝜕𝑦 ( 𝑦 ) 𝜕𝑦 ( 𝑦 ) 𝑦
is discontinuous again when 𝑦 = 0. Therefore, by the theorem, the initial value problem has a
unique solution for 𝑦 ≠ 0. ■

4.3 Applications of First Order Differential Equations

Some examples of applications of differential equations are mentioned in the introduction.
In this section, we will discuss a few more examples.

4.3.1 Exponential Growth

When modeling population growth, Malthus’s exponential model is frequently used:

𝑃 ′ = 𝑟𝑃,

where 𝑟 is the constant.

As a separable equation, the general solution of this model is

𝑃(𝑡) = 𝑐𝑒 𝑟𝑡 .

With an initial condition 𝑃(𝑡0 ) = 𝑃0 , the solution is

𝑃(𝑡) = 𝑃0 𝑒 𝑟(𝑡−𝑡0 ) .

53 Lecture Notes for MA451

 4.3 Applications of First Order Differential Equations

In this model, normally, 𝑃 > 0. Therefore, if 𝑎 > 0, then 𝑃 is increasing without upper
bound. If 𝑎 < 0, then 𝑃 is decreasing with the lower bound 0.

Example 4.5 A bacteria culture starts with 10 bacteria and grows to 90 bacteria after 2 hour.
Assume it grows at a rate proportional to its size.

1. Express the population after 𝑡 hours as a function of 𝑡.

2. What is the population after 9 hours?
3. How long it will take for the population to reach 2500?

Solution

1. Since the growth rate is constantly proportional to its size, the population satisfies the expo-
nential model
𝑃 ′ (𝑡) = 𝑟𝑃(𝑡).

Solving it yields the general solution

𝑃(𝑡) = 𝑒 𝑟𝑡+𝑐1 = 𝑒 𝑐1 𝑒 𝑟𝑡 = 𝑐𝑒 𝑟𝑡 ,

where 𝑐 is a constant to be determined by an initial condition.

In this model, both 𝑐 and 𝑎 are to be determined. In the statement of the question, the sen-
tence “A bacteria culture starts with 10 bacteria and grows to 90 bacteria after 2 hour” means

𝑃(0) = 10 and 𝑃(2) = 90.

Those two conditions implies that 𝑎 and 𝑐 satisfy the following system of equations
⎧
⎪
⎪
⎪ 𝑁 (0) = 𝑐 ⋅ 𝑒 0 = 10
⎨
⎪
⎪
⎪ 𝑁 (2) = 𝑐 ⋅ 𝑒 2𝑎 = 90
⎩

The first equation implies that 𝑐 = 10. Plugging it in to second equation yields

𝑒 2𝑎 =9
2𝑎 = ln 9
1
𝑎 = ln 9
2
𝑎 = ln 3.

Hence the population function is

𝑃(𝑡) = 10𝑒 𝑡 ln 3 .

54 Lecture Notes for MA451

 4.3 Applications of First Order Differential Equations

2. The population after 9 hours is

𝑃(9) = 10𝑒 9 ln 3 ≈ 196830.

3. The time that it will takes the culture to 2500 satisfies the equation

𝑁 (𝑡) = 2500.

Solving the equation yields

10𝑒 𝑡 ln 3 =2500
𝑒 𝑡 ln 3 =250
𝑡 ln 3 = ln(250)
ln(250)
𝑡=
ln 3
𝑡 ≈5.
So it takes about 5 hours for the bacteria culture to grow to 2500.

■
In the previous population model, when 𝑎 > 0, the population grows exponentially with-
out a limit. In reality, the growth is limited by the environment capacity 𝛼1 . A refined model is
Verhulst’s logistic population model,

𝑃 ′ = 𝑟𝑃(1 − 𝛼𝑃),

1
where 𝑎 is the growth rate when the capacity has no or minimal impact on the growth, and 𝛼
is capacity, i.e. the limit of the population in the environment.

Note that the equation is also an separable (indeed, autonomous) equation. The equation

55 Lecture Notes for MA451

 4.3 Applications of First Order Differential Equations

can be solved using partial fraction decomposition:

𝑃 ′ =𝑟𝑃(1 − 𝛼𝑃)
𝑃′
=𝑟
𝑃(1 − 𝛼𝑃)
𝑃′ 𝛼𝑃 ′
+ =𝑟
𝑃 1 − 𝛼𝑃
(ln(𝑃))′ − (ln(1 − 𝛼𝑃))′ =𝑟
′
𝑃
ln =𝑟
( ( 1 − 𝛼𝑃 ))
𝑃
=𝑒 𝑟𝑡
1 − 𝛼𝑃
𝑃 =𝑒 𝑟𝑡 (1 − 𝛼𝑃)
𝑃 + 𝑒 𝑟𝑡 𝛼𝑃 =𝑐𝑒 𝑟𝑡
𝑐𝑒 𝑟𝑡
𝑃= .
1 + 𝑐𝛼𝑒 𝑟𝑡

Note that the limit of 𝑃(𝑡) as 𝑡 → ∞ is nothing but the capacity 𝛼1 .

Example 4.6 One hundred rabbits were released in a forest. . It is observed that the population
after 𝑡 years develops according to Verhulst’s logistic population model. The carrying capacity
is estimated to be 10, 000.

Suppose the growth rate is 𝑟 = 2. What is the population size after 5 years.

Solution Since the carrying capacity is 𝛼1 = 10, 000, the value 𝛼 is 0.00001. The logistic model is then

𝑃 ′ (𝑡) = 2𝑃(𝑡)(1 − 0.00001𝑃(𝑡)).

The discussion of the logistic model above shows that the general solution is

𝑐𝑒 2𝑡
𝑃(𝑡) = .
1 + 0.00001𝑐𝑒 2𝑡
Since 𝑃(0) = 100, solving for 𝑐 yields 𝑐 ≈ 100.

So the population of rabbits after 𝑡 years is

100𝑒 2𝑡
𝑃(𝑡) = .
1 + 0.001𝑒 2𝑡

Therefore, after 10 years the population will be

100𝑒 10
𝑃(15) = ≈ 95657.
1 + 0.001𝑒 10

56 Lecture Notes for MA451

 4.3 Applications of First Order Differential Equations

■
 Exercise 4.4 A bread dough increases in volume at a rate proportional to the volume 𝑉
present. Suppose the initial volume is 𝑉0 . After 2 hours, the volume increases to 1.5𝑉0 . How
long will it take for the volume to increase to 2𝑉0 ?

Solution Suppose the proportional factor is 𝑘. Then 𝑉 ′ (𝑡) = 𝑘𝑉 (𝑡) after 𝑡 hours. Since the
initial condition is 𝑉 (0) = 𝑉0 , it follows that 𝑉 (𝑡) = 𝑉0 𝑒 −𝑘𝑡 .

Because 𝑉 (2) = 1.5𝑉0 . The constant 𝑘 satisfies the equation

1.5𝑉0 = 𝑉0 𝑒 2𝑘 .

Solving the equation yields 𝑘 = 21 ln(1.5).

The time needed for the volume to 2𝑉0 satisfies the equation
ln(1.5)
𝑉0 𝑒 𝑡 2 = 2𝑉0 .

Solving for 𝑡 from the equation gives

2 ln 2
𝑡= ≈ 3.4.
ln(1.5)

4.3.2 Exponential decay

If a quantity decreases at a rate proportional to its current value, then we say it is subject
to exponential decay. Suppose the quantity is 𝑁 (𝑡) after a time 𝑡 and the rate of decreas-
ing, called the exponential decay rate, is 𝑘 > 0. Then the quantity 𝑁 (𝑡) satisfies the following
equation
𝑑𝑁
= −𝑘𝑁 .
𝑑𝑡

Solving the equation yields that

𝑁 (𝑡) = 𝑁0 ⋅ 𝑒 −𝑘𝑡 ,

where 𝑁0 = 𝑁 (0) is the initial quantity.

Exponential decay applies in a wide variety of situations, particularly in natural science.

For example, the quantity of radioactive material decays exponentially.

In a radioactive decay model, the decay rate 𝑘 can be determined by the half-life, that is
the time required to decay the quantity to one half of its initial value. Suppose the half-life of

57 Lecture Notes for MA451

 4.3 Applications of First Order Differential Equations

a radioactive material is 𝜏 . Then the exponential decay rate 𝑘 satisfies the following equation

1
𝑁0 𝑒 −𝑘𝜏 = 𝑁0 .
2
Solving for 𝑘 implies that
ln 2
𝑘=
𝜏
and the exponential decay model becomes
𝑡 𝑡
𝑁 (𝑡) = 𝑁0 𝑒 − ln 2⋅ 𝜏 = 𝑁0 2− 𝜏 .

Example 4.7 A radioactive substance has a half-life of 40 days. Suppose its mass is now 300
g (grams).

After how long will the amount present be 200 g.

Solution Applying the radioactive decay model with 𝜏 = 40 and 𝑁0 = 300 implies that
𝑡
𝑁 (𝑡) = 300 ⋅ 2 40 .

When 𝑁 (𝑡) = 200, the time 𝑡 satisfies the equation

𝑡
200 = 300 ⋅ 2− 40 .

Solving the equation for 𝑡 yields

𝑡
200 =300 ⋅ 2− 40
2 − 40𝑡
=2
3
2 𝑡
ln ( ) = ln 2 ⋅ (− )
3 40
40(ln 3 − ln 2)
𝑡=
ln 2
𝑡 ≈23.
So it takes about 23 days for the mass decrease to 200 g. ■
Example 4.8 Mixed Growth and Decay A radioactive substance has a half-life of 100 days.
Suppose its mass is now 500 mg and additional amounts are added at the rate of 4 mg per
day. Suppose the mass after 𝑡 days is 𝑄(𝑡).

1. Find a formula for the rate of change 𝑄 ′ (𝑡) of the mass in terms of 𝑡.
2. Find 𝑄(𝑡) in terms of 𝑡.

Solution

1. The rate of change composes of two rates: the rate of decreasing and rate of increasing.

58 Lecture Notes for MA451

 4.3 Applications of First Order Differential Equations
ln 2
By exponential decay model, the The rate of decreasing is −𝑘𝑄(𝑡), where 𝑘 = 100
is the expo-
nential decay rate determined by the half-life.
The rate of increasing is the fixed rate 5 mg/day.
Therefore, the rate of change is

ln(2)
𝑄 ′ (𝑡) = − 𝑄(𝑡) + 5.
100
2. Since the right hand side of the above equation is a linear expression of 𝑄(𝑡), a linear substi-
tution 𝑣 = − ln
100
2
𝑄(𝑡) + 5 will reduce the differential equation to the separable equation

ln 2
𝑣′ = − 𝑣.
100
Therefore,
ln 2
𝑣 = 𝑐𝑒 − 100 𝑡

and
100 ln 2
𝑄(𝑡) = − (𝑐𝑒 − 100 𝑡 − 5) .
ln 2
The initial condition 𝑄(0) = 500 implies

100
− (𝑐 − 5) =500
ln 2
(𝑐 − 5) = − 5 ln 2
𝑐 = − 5 ln 2 + 5.

Therefore,
𝑡
(500 ln 2 − 500)2− 100 + 500
𝑄(𝑡) = .
ln 2
■
 Exercise 4.5 Living cells maintain a consistent level of carbon-14. However, when the cell
dies, carbon-14 start decaying exponentially at a constant rate. It is known that the half-life
of carbon-14 is about 5570 years.

An archaeologist investigating the site of an ancient village finds a burial ground where
the amount of carbon-14 present in individual remains is about 63% of the amount present in
live individuals. Estimate the age of the village.

Solution Let 𝑄 = 𝑄(𝑡) be the quantity of carbon-14 remained in individuals 𝑡 years after
death, and let 𝑄0 be the quantity that would be present in live individuals. Since carbon-14
decays exponentially with half-life 5570 years, its exponential decay rate is

ln 2
𝑘= .
5570

59 Lecture Notes for MA451

 4.3 Applications of First Order Differential Equations

Therefore,
𝑄 = 𝑄0 𝑒 −𝑡(ln 2)/5570 .

Since the quantity remained in individuals now is 𝑄(𝑡) = 0.63𝑄0 , the age 𝑡 satisfies the
equation
𝑄0 𝑒 −𝑡(ln 2)/5570 = 0.63𝑄0

Solving the equation for 𝑡 yields

ln(0.63)
𝑡 = −5570 ≈ 3713.
ln 2

Therefore, the village is about 3713 years old. ■

4.3.3 Newton’s law of cooling

Newton’s law of cooling states that the temperature of an object changes at a rate pro-
portional to the difference between its temperature and the temperature of its surrounding.

Let 𝑇 (𝑡) be the temperature of the object and 𝑇𝑚 (𝑡) the temperature of the surrounding
medium at the time 𝑡. Then Newton’s law of cooling can be state as a differential equation:

𝑇 ′ (𝑡) = −𝑘(𝑇 (𝑡) − 𝑇𝑚 (𝑡)).

Here 𝑘 is a positive constant, called the temperature decay constant. The reason that 𝑘 is
positive is because the temperature of the object must decrease if 𝑇 (𝑡) > 𝑇𝑚 (𝑡), or increase if
𝑇 (𝑡) < 𝑇𝑚 (𝑡).

For simplicity, assume that the medium is maintained at a constant temperature 𝑇𝑚 . This
model apply to many situations but not all. For example, to cool down a cup of hot coffee in
the room temperature, the change of room temperature is neglectful. However, to cool down
a cup of hot coffee is a small pot of cold water, the temperature of the water will change
accordingly. In the later case, the heat transfer law in thermodynamics will be needed.

In this section, the surrounding temperature will be assumed to be constant. In this case,
note that 𝑇 ′ (𝑡) = (𝑇 (𝑡) − 𝑇𝑚 )′ , then Newton’s law of cooling can be re-formulated as

(𝑇 (𝑡) − 𝑇𝑚 )′ = −𝑘(𝑇 (𝑡) − 𝑇𝑚 ).

That is, 𝑇 − 𝑇𝑚 decays exponentially, with decay constant 𝑘. Therefore,

𝑇 (𝑡) − 𝑇𝑚 = (𝑇0 − 𝑇𝑚 )𝑒 −𝑘𝑡

60 Lecture Notes for MA451

 4.3 Applications of First Order Differential Equations

or equivalently
𝑇 (𝑡) = (𝑇0 − 𝑇𝑚 )𝑒 −𝑘𝑡 + 𝑇𝑚 ,

where 𝑇0 is the initial temperature of the object.

Example 4.9 A ceramic insulator is baked at 400◦ C and cooled in a room in which the tempera-
ture is 25◦ C. After 4 minutes the temperature of the insulator is 200◦ C. What is its temperature
after 8 minutes?

Solution Here 𝑇0 = 400 and 𝑇𝑚 = 25, so the temperature function of the ceramic insulator is

𝑇 = 25 + 375𝑒 −𝑘𝑡 .

Since 𝑇 (4) = 200, which determines 𝑘, then

200 = 25 + 375𝑒 −4𝑘 .

Solving for 𝑘 yields

1 7
𝑘 = − ln ( ) .
4 15

Substituting this into 𝑇 (𝑡) yields

𝑡 7
𝑇 = 25 + 375𝑒 − 4 ln( 15 ) .

Therefore, the temperature of the insulator after 8 minutes is

7
𝑇 (8) = 25 + 375𝑒 −2 ln( 15 )
7 2
= 25 + 375 ( ≈ 107◦ C.
15 )
■
 Exercise 4.6 A metal bar at a temperature of 200◦ F is placed in a room at a constant
temperature of 50◦ F. If after 20 minutes, the temperature of the bar is 90◦ F, find

1. Find the formula of temperature as a function of 𝑡.

2. Find the time it will take the bar to reach a temperature of 60◦ F;
3. Find the limit of the temperature as time goes to infinity.

Solution

1. Since 𝑇𝑚 = 50 and 𝑇0 = 200, the temperature function is

𝑇 (𝑡) = 50 + 150𝑒 −𝑘𝑡 .

61 Lecture Notes for MA451

 4.3 Applications of First Order Differential Equations

Because 𝑇 (20) = 90, then 𝑘 is determined by the equation

90 = 50 + 150 ⋅ 𝑒 −20𝑘 .

Solving the equation yields

40 =150 ⋅ 𝑒 −20𝑘
4
𝑒 −20𝑘 =
15
4
−20𝑘 = ln ( )
15
1 4
𝑘 =− ln ( ) .
20 15
Hence
1 4
𝑇 (𝑡) = 50 + 150𝑒 20 ln( 15 )𝑡 .

2. The time that it will take the bar to 60◦ satisfies the equation

50 + 150 ⋅ 𝑒 −𝑘𝑡 = 60,

where 𝑘 = − 201 ln ( 154 ).

Solving the equation yields

150 ⋅ 𝑒 −𝑘𝑡 =10

1
𝑒 −𝑘𝑡 =
15
1
−𝑘𝑡 = ln( )
15
1 1
𝑡 = − ln( )
𝑘 15
1 1
𝑡 =− 1 4 ln( )
− 20 ln( 15 ) 15
ln(1/15)
𝑡 =20 ⋅
ln(4/15)
𝑡 ≈41.

So it takes about 41 minutes for the bar to cool down to 60◦ F.

3. When 𝑡 goes to infinity, because 𝑘 is positive, the limit lim 𝑒 −𝑘𝑡 = 0. Hence
𝑡→∞

1 4
lim 𝑇 (𝑡) = 50 + 150 ⋅ lim 𝑒 20 ln( 15 )𝑡 = 50.
𝑡→∞ 𝑡→∞

The answer is the room temperature.

62 Lecture Notes for MA451

 4.3 Applications of First Order Differential Equations

4.3.3.1 Mixing Problems

Example 4.10 A tank initially contains 40 pounds of salt dissolved in 600 gallons of water.
Starting at 𝑡0 = 0, water that contains 1/2 pound of salt per gallon is poured into the tank at
the rate of 4 gal/min and the mixture is drained from the tank at the same rate.

1. Find a differential equation for the quantity 𝑄(𝑡) of salt in the tank at time 𝑡 > 0, and
solve the equation to determine 𝑄(𝑡).
2. Find lim 𝑄(𝑡).
𝑡→∞

Solution To find a differential equation for the quantity of salt 𝑄, because the given information
is about the rate of change of the quantity of salt 𝑄 ′ , we will find an equation for 𝑄 ′ . is the rate of
change of the quantity of salt in the tank changes with respect to time; The rate of change composes
of two parts
𝑄 ′ = in-rate − out-rate.

The in-rate is
1
( 2 lb/gal) × (4 gal/min) = 2 lb/min.

Since the concentration changes, the determine the out-rate, we need know the concentration
at 𝑡. Because the in-flow and out-flow rates are the same, the volume of the mixture is constant
which is 600 gal. Therefore, the concentration at time 𝑡 is 𝑄(𝑡)
600
, and the out-rate is then

𝑄(𝑡) 𝑄(𝑡)
( concentration) × ( rate of flow out) = ×4= .
00 150

Therefore,
𝑄
𝑄′ = 2 − ,
150
which is a first order linear equation that can be rewritten as

𝑄
𝑄′ + = 2.
150
The integrating factor is
1 𝑡
d𝑡
𝑟(𝑡) = 𝑒 ∫ 150 = 𝑒 150 .

Multiplying the integrating factor to both sides and integrating both sides yields

𝑡 𝑡
𝑄(𝑡)𝑒 150 = ∫ 2𝑒 150 d 𝑡
𝑡 𝑡
𝑄(𝑡)𝑒 150 =300𝑒 150 + 𝑐
𝑡
𝑄(𝑡) =300 + 𝑐𝑒 − 150 .

63 Lecture Notes for MA451

 4.3 Applications of First Order Differential Equations

Since 𝑄(0) = 40, 𝑐 = −260 and

𝑄 = 300 − 260𝑒 −𝑡/150 .

The limit of 𝑄(𝑡) as 𝑡 → ∞ is

lim 𝑄(𝑡) = lim (300 − 260𝑒 −𝑡/150 ) = 300.

𝑡→∞ 𝑡→∞

This is intuitively reasonable. Because, the incoming solution contains 1/2 pound of salt per gallon
and there are always 600 gallons of water in the tank. ■
Example 4.11 A 500-liter tank initially contains 10 g of salt dissolved in 200 liters of water.
Starting at 𝑡0 = 0, water that contains 1/4 g of salt per liter is poured into the tank at the rate
of 4 liters/min and the mixture is drained from the tank at the rate of 2 liters/min. Find a
differential equation for the quantity 𝑄(𝑡) of salt in the tank at time 𝑡 prior to the time when
the tank overflows and find the concentration 𝐾 (𝑡) (g/liter) of salt in the tank at any such time.

Solution The difference between this example and the above equation is that the volume of the
mixture changes in this example. Let 𝑊 (𝑡) of solution in the tank at any time 𝑡 prior to overflow.
Since 𝑊 (0) = 200, in-flow rate 4 liters/min and the out-flow rate 2 liters/min, the net flow rate is
then 2 liters/min. Therefore, the volume is
𝑡
𝑊 (𝑡) = 200 + ∫ 2 d 𝑥 = 200 + 2𝑡.
0

Since the volume of tank is 500 liter, and 𝑊 (150) = 500, this formula is valid for when the time
𝑡 is from 0 to 150 min, i.e. 0 ≤ 𝑡 ≤ 150.

Now let 𝑄(𝑡) be the number of grams of salt in the tank at time 𝑡, where 0 ≤ 𝑡 ≤ 150. As in
above example
𝑄 ′ = rate in − rate out
1 2
= ( g/liter) × (4 liters/min) + 𝑄(𝑡) ⋅ g/min
4 2𝑡 + 200
1
=1 g/min + 𝑄(𝑡) ⋅ g/min
𝑡 + 100
𝑄(𝑡)
= 1+ g/min.
( 𝑡 + 100 )
Therefore, the salt flow rate is
𝑄
𝑄′ = 1 − ,
𝑡 + 100
or
1
𝑄′ + 𝑄 = 1.
𝑡 + 100
The integrating factor is
1
d𝑡
𝑟(𝑥) = 𝑒 ∫ 𝑡+100 = 𝑡 + 100.

64 Lecture Notes for MA451

 4.3 Applications of First Order Differential Equations

Multiplying the equation by 𝑟(𝑥) and integrating both sides yields

1 𝑡 + 100 𝑐
𝑄(𝑡) = ∫ (𝑡 + 100) d 𝑡 = + .
𝑡 + 100 2 𝑡 + 100

Since 𝑄(0) = 10, solving for 𝑐 implies

𝑐 = −4000.

Hence,
𝑡 + 100 4000
𝑄(𝑡) = − .
2 𝑡 + 100

Now let 𝐾 (𝑡) be the concentration of salt at time 𝑡. Then

𝑄(𝑡) 1 2000
𝐾 (𝑡) = = − .
𝑊 (𝑡) 4 (𝑡 + 100)2

4.3.4 Motion Under Gravity in a Resisting Medium*

In this section, we will consider an object with constant mass 𝑚 moving vertically in a
resisting medium near Earth’s surface under a force 𝐹 (𝑡). We will take the upward direction
to be positive. Let 𝑦 = 𝑦(𝑡) be the displacement of the object from a reference point above
the ground at time 𝑡. Let 𝑣 = 𝑣(𝑡) and 𝑎 = 𝑎(𝑡) be the velocity and acceleration of the object at
time 𝑡. Then 𝑎(𝑡) = 𝑣 ′ (𝑡) and 𝑣(𝑡) = 𝑦 ′ (𝑡).

Newton’s second law of motion asserts that the force 𝐹 equals the product of the mass
𝑚 and the acceleration 𝑎:
𝐹 (𝑡) = 𝑚𝑎(𝑡).

When an object moves vertically in a resisting medium, two main forces are the gravita-
tional force −𝑚𝑔 and the medium resistive force. Here, 𝑔 is the acceleration due to gravity.

Example 4.12 An object with mass 𝑚 and initial velocity 𝑣(0) = 𝑣0 moves under constant
gravitational force 𝑚𝑔 through a medium that exerts a resistance with magnitude proportional
to the speed |𝑣| of the object.

1. Find the velocity of the object as a function of 𝑡.

2. Find the limit lim 𝑣(𝑡).
𝑡→∞

Solution

Taking the upward as the positive direction, the total force acting on the object is

𝐹 = −𝑚𝑔 + 𝐹1 ,

65 Lecture Notes for MA451

 4.3 Applications of First Order Differential Equations

where −𝑚𝑔 is the force due to gravity and 𝐹1 is the resisting force of the medium. Since, the resisting
force 𝐹1 has the magnitude proportional to |𝑣|. There is a positive constant 𝑘 such that |𝐹1 | = 𝑚𝑘|𝑣|.
Because the resistance force is always opposite the direction of the velocity. If the object is moving
downward, that is 𝑣 ≤ 0, the resisting force is upward. So

𝐹1 = 𝑚𝑘(−𝑣) = −𝑚𝑘𝑣 > 0.

If the object is moving upward, that is 𝑣 ≥ 0, the resisting force is downward. So

𝐹1 = −𝑚𝑘𝑣 < 0

Therefore, the total force 𝐹 is

𝐹 = −𝑚𝑔 − 𝑚𝑘𝑣,

regardless of the sign of the velocity.

From Newton’s second law of motion,

𝐹 = 𝑚𝑎 = 𝑚𝑣 ′ .

So
𝑚𝑣 ′ = −𝑚𝑔 − 𝑚𝑘𝑣,

or equivalently,
𝑣 ′ = −𝑘𝑣 − 𝑔.

This equation can be solved using multiple methods. Because the coefficients are constants, we
may use the linear substitution 𝑣 = 𝑦 − 𝑘𝑔 . Plugging it into the above equation yields

𝑦 ′ = −𝑘𝑦,

which has the general solution 𝑦 = 𝑐𝑒 −𝑘𝑡 . Hence,

𝑔
𝑣 = 𝑐𝑒 −𝑘𝑡 − .
𝑘
Since 𝑣(0) = 𝑣0 , the constant 𝑐 is determined by
𝑔
𝑐− = 𝑣0 .
𝑘

Therefore, 𝑐 = 𝑣0 + 𝑘𝑔 .

So the velocity function is

𝑔 𝑔
𝑣 = (𝑣0 + ) 𝑒 −𝑘𝑡 −
𝑘 𝑘

66 Lecture Notes for MA451

 4.3 Applications of First Order Differential Equations

Letting 𝑡 → ∞ here shows that the limit of 𝑣(𝑡) is

𝑔
lim 𝑣(𝑡) = − .
𝑡→∞ 𝑘
■
We see that under reasonable assumptions on the resisting force, the velocity approaches
a limit as 𝑡 → ∞, which is called the terminal velocity. You will find an object reaches its terminal
velocity when its acceleration is 0.

Example 4.13 A 10 kg mass is given an initial velocity 𝑣(0) = 𝑣0 ≤ 0 near Earth’s surface.
The only forces acting on it are gravity and atmospheric resistance proportional to the square
of the speed. Assuming that the resistance is 8 Newton (N for short) when the speed is 2
meters/second (m/s for short).

1. Find the velocity of the object as a function of 𝑡.

2. Find the terminal velocity.

Solution Since the object is falling, the resistance is in the upward direction which is assumed to
be the positive direction. Hence,
𝑚𝑣 ′ = −𝑚𝑔 + 𝑘𝑣 2 ,

where 𝑘 is a constant. Since the magnitude of the resistance is 8 N when 𝑣 = 2 m/s, that is,

𝑘(22 ) = 8.

So 𝑘 = 2 N-s2 / m2 , where 𝑁 − 𝑠 means the product of 𝑁 and 𝑠. Since 𝑚 = 10 kg and 𝑔 = 9.8 m/ s2 ,

the velocity satisfies the following equation.

10𝑣 ′ = −98 + 2𝑣 2 = 2(𝑣 2 − 49).

Note that the solutions 𝑣 = −7 or 𝑣 = 7 of the equation 𝑣 2 − 49 = 0 are solutions to the above
differential equation. Since 𝑣0 ≤ 0 and the object is falling, the velocity 𝑣 has a negative direction.
So 𝑣 = 7 can not be a solution under the assumption that 𝑣0 ≤ 0. Moreover, 𝑣 = −7 is a solution
only when 𝑣0 = −7.

Now assume 𝑣0 ≠ −7. Separate variables implies

1 ′ 1
( 𝑣 2 − 49 ) 𝑣 = 5 .

67 Lecture Notes for MA451

 4.3 Applications of First Order Differential Equations

Integrating both sides using partial fraction decomposition yields

1 ′ 1
( 𝑣 2 − 49 ) 𝑣 = 5
1 1
=
( (𝑣 − 7)(𝑣 + 7) ) 5
1 1 1 1
( − ) 𝑣′ =
14 𝑣 − 7 𝑣 + 7 5
𝑣′ 𝑣′ 14
− =
𝑣−7 𝑣+7 5
′
𝑣 𝑣′ 14
∫ 𝑣 − 7 d𝑡 − ∫ 𝑣 + 7 d𝑡 =∫ 5 d𝑡
1 1 14𝑡
∫ 𝑣 − 7 d(𝑣 − 7) − ∫ 𝑣 + 7 d(𝑣 + 7) = 5 + 𝑐
14𝑡
ln |𝑣 − 7| − ln |𝑣 + 7| = +𝑐
5
|𝑣 − 7|
ln || | = 14𝑡 + 𝑐
(| 𝑣 + 7 ||) 5
|𝑣 − 7| 14𝑡
| |
| 𝑣 + 7 | =𝑐𝑒 5
| |

Since 𝑣(0) = 𝑣0 ,
| 𝑣0 − 7 |
𝑐 = || |
|
| 𝑣0 + 7 |

If 𝑣0 ≤ −7, then 𝑣 ′ (0) > 0 and 𝑣(𝑡) will be increasing until 𝑣 ′ = 0. So 𝑣0 ≤ 𝑣(𝑡) ≤ −7. Then

|𝑣 − 7| 𝑣 − 7
| |
|𝑣 + 7| = 𝑣 + 7
| |

and
𝑣0 − 7
𝑐= .
𝑣0 + 7
So
𝑣−7 14𝑡 𝑣0 − 7
= 𝑒 15 ⋅ .
𝑣+7 𝑣0 + 7

If 𝑣0 ≥ −7, then 𝑣 ′ (0) < 0 and 𝑣(𝑡) will be decreasing until 𝑣 ′ = 0. So −7 ≤ 𝑣(𝑡) ≤ 𝑣0 . Then

|𝑣 − 7| 𝑣−7
| |
|𝑣 + 7| = −𝑣 + 7
| |

and
𝑣0 − 7
𝑐=− .
𝑣0 + 7
Again
𝑣−7 14𝑡 𝑣0 − 7
= 𝑒 15 ⋅ .
𝑣+7 𝑣0 + 7

68 Lecture Notes for MA451

 4.3 Applications of First Order Differential Equations

Solving for 𝑣 yields

14𝑡
(𝑣0 − 7)𝑒 5 + 𝑣0 + 7
𝑣 = −7 14𝑡 .
(𝑣0 − 7)𝑒 5 − 𝑣0 − 7

Since 𝑣0 ≤ 0, 𝑣 is defined and negative for all 𝑡 > 0. The terminal velocity is

lim 𝑣(𝑡) = −7 m/s,

𝑡→∞

independent of 𝑣0 . ■
In examples above, we have been using international metric system for force. There are
other metric systems (see Wiki page on Newton_(unit) for details). One of them is the British
metric system which uses foot (lb) for length, pound-force (lb) for force, and slug (sl) for mass.

 Exercise 4.7 A 320-lb object is given an initial upward velocity of 20 ft/s near the surface
of Earth. The atmosphere resists the motion with a force of 2 lb for each ft/s of speed.
Assuming that the only other force acting on the object is the gravity 𝑔 = 32 ft/ s2 .

Find its velocity 𝑣 as a function of 𝑡, and find its terminal velocity.

Solution Note that here 320-lb means the gravitational force is 320 lb. Since 𝑚𝑔 = 320 and
𝑔 = 32, 𝑚 = 320/32 = 10 lb-s2 / ft. Suppose the velocity function is 𝑣 = 𝑣(𝑡) ft/s after 𝑡 Then
the atmospheric resistance is −2𝑣 lb.

By Newton’s second law of motion,

10𝑣 ′ = −320 − 2𝑣,

equivalently,
5𝑣 ′ = −160 − 𝑣,

Solving the equation by a linear substitution 𝑦 = 𝑣 + 1600 yields

5𝑣 ′ = − 160 − 𝑣
5𝑦 ′ = − 𝑦
𝑦′
=−5
𝑦
𝑦′
∫ 𝑦 d 𝑡 = ∫ −5 d 𝑡
ln |𝑦| = − 5𝑡 + 𝑐
|𝑦| =𝑐𝑒 −5𝑡
|𝑣 + 160| =𝑐𝑒 −5𝑡 .

Since 𝑣(0) = 20 and 𝑣 ′ (0) < 0, the velocity 𝑣 will decrease. Note that 𝑣 ′ = 0 when 𝑣 = −160.

69 Lecture Notes for MA451

 4.3 Applications of First Order Differential Equations

Then −160 ≤ 𝑣(𝑡) ≤ 20, and |𝑣 + 160| = 𝑣 + 160. Therefore,

𝑣 = −160 + 𝑐𝑒 −5𝑡 .

Since the initial velocity is 𝑣(0) = 20, the constant 𝑐 satisfies the equation

20 = −160 + 𝑐.

Thus, 𝑐 = 180 and the velocity function is

𝑣 = −160 + 180𝑒 −𝑡/10 ft/ s.

The terminal velocity is

lim 𝑣(𝑡) = lim −160 + 180𝑒 −𝑡/10 = −160 ft/ s.

𝑡→∞ 𝑡→∞

4.3.5 Orthogonal Trajectories*

We’ve seen that general solutions to first order differential equations are in the form

𝐹 (𝑥, 𝑦, 𝑐) = 0,

where 𝑐 is a constant taking varies real values. The graph of those solutions are known as
one-parameter families of curves. We will simply call them families of curves.

Two curves 𝐶1 and 𝐶2 are said to be orthogonal at a point of intersection (𝑥0 , 𝑦0 ) if they
have perpendicular tangents at (𝑥0 , 𝑦0 ).

A curve is said to be an orthogonal trajectory of a given family of curves if it is orthogonal

to every curve in the family.

Suppose 𝐹 (𝑥, 𝑦, 𝑐) = 0 is a family of integral curves of the differential equation

𝑦 ′ = −𝑓 (𝑥, 𝑦).

From the definition, an orthogonal trajectory of 𝐹 (𝑥, 𝑦, 𝑐) = 0 has perpendicular tangents. So

slopes of tangent lines of an orthogonal trajectory are determined by

1
𝑦′ = .
𝑓 (𝑥, 𝑦)

Therefore, families of integral curves of the differential equations 𝑦 ′ = 𝑓 (𝑥, 𝑦) and 𝑦 ′ =

70 Lecture Notes for MA451

 4.3 Applications of First Order Differential Equations
1
− 𝑓 (𝑥,𝑦) are orthogonal trajectories to each other.

That suggests a method for finding orthogonal trajectories of a family of curves:

Step 1: Find a differential equation 𝑦 ′ = 𝑓 (𝑥, 𝑦) without 𝑐 for the given family 𝐹 (𝑥, 𝑦, 𝑐).
1
Step 2: Solve the differential equation 𝑦 ′ = − to find the family of orthogonal trajecto-
𝑓 (𝑥, 𝑦)
ries.

Example 4.14 Find the family of orthogonal trajectories to the family of curves defined by

𝑦 = 𝑐𝑥 2 , 𝑐 ≠ 0.

Solution Taking derivative implies

𝑦 ′ = 2𝑐𝑥.
𝑦
Since 𝑦 = 𝑐𝑥 2 , solving for 𝑐 yields 𝑐 = 𝑥2
. So the above differential equation can be re-written as

2𝑦
𝑦′ = .
𝑥
Therefore, orthogonal trajectories satisfy the equation
𝑥
𝑦′ = − .
2𝑦

Solving the equation yields that the ellipse defined by

1
𝑦 2 + 𝑥 2 = 𝑐,
2
is a family of orthogonal trajectories. ■
 Exercise 4.8 Find the orthogonal trajectories of the family of hyperbolas

𝑥𝑦 = 𝑐 𝑐 ≠ 0.

Solution Differentiating the equation implicitly with respect to 𝑥 yields

𝑦 + 𝑥𝑦 ′ = 0.

So
𝑦
𝑦′ = − .
𝑥
Thus, the integral curves of
𝑥
𝑦′ =
𝑦
are orthogonal trajectories of the given family. Separating variables and solving equations

71 Lecture Notes for MA451

 4.A Integrating factor methods for linear first order equations revisited

yields
𝑥
𝑦′ =
𝑦
𝑦 ′ 𝑦 =𝑥
′
∫ 𝑦 𝑦 d𝑥 =∫ 𝑥 d𝑥
𝑦 2 =𝑥 2 + 𝑘
𝑦 2 − 𝑥 2 =𝑘.
So an orthogonal trajectory is either hyperbola (if 𝑘 ≠ 0), or the union of the lines 𝑦 = 𝑥 and
𝑦 = −𝑥 (if 𝑘 = 0). ■

4.A Integrating factor methods for linear first order equations

revisited

4.A.1 The integrating factor method for linear first order equations
Consider the differential equation

d𝑦
+ 𝑝(𝑥)𝑦 = 𝑞(𝑥).
d𝑥
Recall an integrating factor for this equation is

𝑝(𝑥)d𝑥
𝑟(𝑥) = ∫ 𝑒 ∫ d𝑥.

Multiplying the equation by 𝑟(𝑥) yields

d𝑦
𝑟(𝑥) + 𝑟(𝑥)𝑝(𝑥)𝑦 =𝑟(𝑥)𝑞(𝑥)
d𝑥
d
(𝑟(𝑥)𝑦) =𝑟(𝑥)𝑞(𝑥)
d𝑥
𝑟(𝑥)𝑦 = ∫ 𝑟(𝑥)𝑞(𝑥)d𝑥
∫ 𝑟(𝑥)𝑞(𝑥)d𝑥
𝑦=
𝑟(𝑥)
∫ (𝑞(𝑥)𝑒 ∫ 𝑝(𝑥)d𝑥 )d𝑥
𝑦= .
𝑒 ∫ 𝑝(𝑥)d𝑥

72 Lecture Notes for MA451

 4.A Integrating factor methods for linear first order equations revisited

4.A.2 Linear first order equations viewed as non-exact equations

This integrating method can be considered as a special case of the integrating method
discussed today. Note that the differential equation can be rewritten as

(𝑝(𝑥)𝑦 − 𝑞(𝑥))d𝑥 + d𝑦 = 0.

Let 𝑀(𝑥, 𝑦) = 𝑝(𝑥)𝑦 − 𝑞(𝑥) and 𝑁 (𝑥, 𝑦) = 1. Then

𝜕𝑀 𝜕𝑁
= 𝑝(𝑥) and = 0.
𝜕𝑦 𝜕𝑥

Therefore,
𝜕𝑀 𝜕𝑁
−
𝜕𝑦 𝜕𝑥
= 𝑝(𝑥).
𝑁
From the proposition discussed today, an integrating factor for the equation (𝑝(𝑥)𝑦−𝑞(𝑥))d𝑥+d𝑦
is
𝜇(𝑥) = 𝑒 ∫ 𝑝(𝑥)d𝑥 .

From the above calculation, you may notice that 𝑟(𝑥) = 𝜇(𝑥).

Now, to solve the equation (𝑝(𝑥)𝑦 − 𝑞(𝑥))d𝑥 + d𝑦 = 0, multiplying both sides by 𝜇(𝑥) yields
an exact equation
𝑒 ∫ 𝑝(𝑥)d𝑥 (𝑝(𝑥)𝑦 − 𝑞(𝑥))d𝑥 + 𝑒 ∫ 𝑝(𝑥)d𝑥 d𝑦 = 0.

The expected solution is an implicitly function defined by 𝐹 (𝑥, 𝑦) = 𝑐, that is 𝐹 is a function

such that
𝜕𝐹 𝜕𝐹
d𝑥 + d𝑦 = [𝑒 ∫ 𝑝(𝑥)d𝑥 (𝑝(𝑥)𝑦 − 𝑞(𝑥))] d𝑥 + 𝑒 ∫ 𝑝(𝑥)d𝑥 d𝑦.
𝜕𝑥 𝜕𝑦
To find 𝐹 , we take the following steps:
𝑝(𝑥)d𝑥
1. Integrate 𝑒 ∫ d𝑦. Denote the resulting function by 𝐹 (𝑥, 𝑦). Then

𝑝(𝑥)d𝑥
𝐹 (𝑥, 𝑦) = 𝑦𝑒 ∫ + 𝑔(𝑥).

𝜕
The reason to add a 𝑔(𝑥) is because 𝜕𝑦
𝑔(𝑥) = 0.

2. Solve for 𝑔 from the equation

𝜕𝐹 𝑝(𝑥)d𝑥
= 𝑒∫ (𝑝(𝑥)𝑦 − 𝑞(𝑥)).
𝜕𝑥
Note that
𝜕𝐹 𝜕 𝑝(𝑥)d𝑥 𝜕
=𝑦 𝑒 ∫ + 𝑔(𝑥)
𝜕𝑥 𝜕𝑥 𝜕𝑥
𝑝(𝑥)d𝑥 d
=𝑦𝑝(𝑥)𝑒 ∫ + 𝑔(𝑥).
d𝑥

73 Lecture Notes for MA451

 4.A Integrating factor methods for linear first order equations revisited

Then the function 𝑔(𝑥) satisfies the following equation

𝑝(𝑥)d𝑥 d 𝑝(𝑥)d𝑥
𝑦𝑝(𝑥)𝑒 ∫ + 𝑔(𝑥) = 𝑒 ∫ (𝑝(𝑥)𝑦 − 𝑞(𝑥)),
d𝑥
or equivalently,
d 𝑝(𝑥)d𝑥
𝑔(𝑥) = −𝑞(𝑥)𝑒 ∫ .
d𝑥
Therefore,
𝑝(𝑥)d𝑥
𝑔(𝑥) = ∫ (−𝑞(𝑥)𝑒 ∫ ) d𝑥,
and
𝑝(𝑥)d𝑥 𝑝(𝑥)d𝑥
𝐹 (𝑥, 𝑦) = 𝑦𝑒 ∫ − ∫ (𝑞(𝑥)𝑒 ∫ ) d𝑥.
Set 𝐹 (𝑥, 𝑦) = 0 and solve for 𝑦, we find a solution

∫ (𝑞(𝑥)𝑒 ∫ 𝑝(𝑥)d𝑥 ) d𝑥
𝑦= ,
𝑒 ∫ 𝑝(𝑥)d𝑥

which is the seem as the solution obtained using the first method.

Remark Instead of treating the whole equation as an non-exact equation and solve it, one can
also find an integrating factor 𝜇(𝑥) of 𝑦𝑞(𝑥)d𝑥 + d𝑦 and find 𝐹 so that

𝜕𝐹 𝜕𝐹
d𝑥 + d𝑦 = 𝑞(𝑥)d𝑥 + d𝑦.
𝜕𝑥 𝜕𝑦

Then the equation (𝑦𝑞(𝑥) − 𝑞(𝑥))d𝑥 + d𝑦 = 0 can be expressed as

d𝐹 (𝑥, 𝑦) = 𝜇(𝑥)𝑞(𝑥)d𝑥.

Integrating both sides will yield the solution.

74 Lecture Notes for MA451

 Week 5: Linear Second Order Equations

10/4–10/7

5.1 Homogeneous Linear Second Order Equations

5.1.1 Basic concepts

The standard form of a second order equation is

𝑦 ′′ (𝑥) = 𝑓 (𝑥, 𝑦(𝑥), 𝑦 ′ (𝑥)).

We often write it simply as

𝑦 ′′ = 𝑓 (𝑥, 𝑦, 𝑦 ′ )

when it is understood that 𝑥 is the variable and 𝑦 is the unknown.

For a 2nd order equation, the initial value problem is of the form
⎧
⎪
⎪ 𝑦 ′′ = 𝑓 (𝑥, 𝑦, 𝑦 ′ ).
⎪
⎪
⎪
⎨ 𝑦(𝑥0 ) = 𝑦0 .
⎪
⎪
⎪
⎪ ′
⎩ 𝑦 (𝑥0 ) = 𝑦1
⎪

That is, in additional to the equation, the values of the function and its first derivative are
given at a point.

A second order differential equation is said to be linear if it can be written in the following
form:
𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥)𝑦 = 𝑓 (𝑥).

We say that a linear second order equation is homogeneous if 𝑓 ≡ 0, that is, the equation
can be written in the following form.

𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥)𝑦 = 0.

Example 5.1 Determine the whether the following equations are linear and/or homogeneous:

1. 2𝑥𝑦 ′′ + 𝑥 2 𝑦 ′ − (sin 𝑥)𝑦 = 𝑥 3 ;

2. 𝑦𝑦 ′′ + 𝑥𝑦 ′ = 𝑥 2 ;
 5.1 Homogeneous Linear Second Order Equations

3. 2𝑒 𝑥 𝑦 ′′ + 𝑦 = 0.

Solution

1. Linear but non-homogeneous;

2. Nonlinear;
3. Linear and homogeneous.

5.1.2 Existence and uniqueness of solutions

The good thing about linear equations is the existence and uniqueness question has a
nice and simple answer.
Theorem 5.1
Suppose 𝑝 and 𝑞 are continuous on an open interval (𝑎, 𝑏), let 𝑥0 be any point in (𝑎, 𝑏), and let
𝑘0 and 𝑘1 be arbitrary real numbers. Then the initial value problem

𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥)𝑦 = 𝑓 (𝑥), 𝑦(𝑥0 ) = 𝑘0 , 𝑦 ′ (𝑥0 ) = 𝑘1

has a unique solution on (𝑎, 𝑏).

♥

The proof of this theorem also relies on Picard’s method of successive approximations.
We refer the reader to (Simmons 2016, Chapter 13) for a proof.

Example 5.2 Consider the equation

𝑦 ′′ − 𝑦 = 0.

1. Find the general solution.

2. Given that 𝑦(0) = 1 and 𝑦 ′ (0) = 3, find the specific solution to this initial value problem.
3. Determine if this example verifies the existence and uniqueness theorem.

Solution

1. The equation can be solved using substitutions. First let 𝑧 = 𝑦 ′ + 𝑦, then

𝑦 ′′ − 𝑦 =0
𝑧 ′ − 𝑦 ′ − 𝑦 =0 where 𝑧 = 𝑦 ′ + 𝑦
𝑧 ′ − 𝑧 =0
𝑧 ′ =𝑧
𝑧′
=1
𝑧
𝑧 =2𝑐1 𝑒 𝑥

76 Lecture Notes for MA451

 5.1 Homogeneous Linear Second Order Equations

Here, we use 2𝑐1 so that in the next substitution no fraction will appear. Now, let 𝑢 = 𝑦 − 𝑐1 𝑒 𝑥 ,
then
𝑦 ′ + 𝑦 =2𝑐1 𝑒 𝑥
𝑢′ = − 𝑢 where 𝑢 = 𝑦 − 𝑐1 𝑒 𝑥
𝑢′
=−1
𝑢
𝑢 =𝑐2 𝑒 −𝑥
𝑦 − 𝑐1 𝑒 𝑥 =𝑐2 𝑒 −𝑥
𝑦 =𝑐1 𝑒 𝑥 + 𝑐2 𝑒 −𝑥
So the general solution is 𝑦 = 𝑐1 𝑒 𝑥 + 𝑐2 𝑒 −𝑥 .
2. The constants 𝑐1 and 𝑐2 are determined by the initial conditions 𝑦(0) = 1 and 𝑦 ′ (0) = 3. Setting
𝑥 = 0 in the functions 𝑦 and 𝑦 ′ yields

𝑐1 + 𝑐2 = 1
𝑐1 − 𝑐2 = 3.

Solving the system of equations yields 𝑐1 = 2 and 𝑐2 = −1. Therefore, 𝑦 = 2𝑒 𝑥 −𝑒 −𝑥 is the unique
solution of the equation 𝑦 ′′ − 𝑦 = 0 on (−∞, ∞).
3. Because the coefficient of 𝑦 ′ and 𝑦 are constant, hence continuous on (−∞, ∞). The theo-
rem asserts that there is an unique solution on (−∞, ∞) which is also confirmed by the above
solution.

■
Note that in the above example, both 𝑦 = 𝑒 𝑥 and 𝑦 = 𝑒 −𝑥 are solution of the equation
𝑦 ′′ − 𝑦 = 0. This observation suggests an approach of solving linear second order equations.

5.1.3 The general solution

If you know linear algebra, you will find some similarity between solutions of linear sys-
tems and differential equations.
Theorem 5.2
Let 𝑦1 and 𝑦2 be two solutions of the homogeneous and linear second order equation

𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥)𝑦 = 0.

Then, for any real numbers 𝑐1 and 𝑐2 , the linear combination 𝑦 = 𝑐1 𝑦1 + 𝑐2 𝑦2 is also a solution.
♥

The theorem indeed follows from the fact that differentiations are linear.

77 Lecture Notes for MA451

 5.1 Homogeneous Linear Second Order Equations

Proof By the linearity of differentiation, the first and second derivative of 𝑦 = 𝑐1 𝑦1 + 𝑐2 𝑦2 are

𝑦 ′ =𝑐1 𝑦1′ + 𝑐2 𝑦2′

𝑦 ′′ =𝑐1 𝑦1′′ + 𝑐2 𝑦2′′ .

Since 𝑦1 and 𝑦2 are solutions, the following equalities hold true

𝑦1′′ + 𝑝(𝑥)𝑦1′ + 𝑞(𝑥)𝑦1 =0, 𝑦2′′ + 𝑝(𝑥)𝑦2′ + 𝑞(𝑥)𝑦2 = 0.

Therefore,
𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥)𝑦
=𝑐1 𝑦1′′ + 𝑐2 𝑦2′′ + 𝑝(𝑥)(𝑐1 𝑦1′ + 𝑐2 𝑦2′ ) + 𝑞(𝑥)(𝑐1 𝑦1 + 𝑐2 𝑦2 )
=𝑐1 (𝑦1′′ + 𝑝(𝑥)𝑦1′ + 𝑞(𝑥)𝑦1 ) + 𝑐2 (𝑦2′′ + 𝑝(𝑥)𝑦2′ + 𝑞(𝑥)𝑦2 )
=0,
that is, 𝑦 = 𝑐1 𝑦1 + 𝑐2 𝑦2 is a solution too. ■
Bases on the above theorem, you may wonder whether every solution of 𝑦 ′′ + 𝑝(𝑥)𝑦 ′ +
𝑞(𝑥)𝑦 = 0 can be written in the form 𝑦 = 𝑐1 𝑦1 + 𝑐2 𝑦2 . The answer is affirmative if 𝑦1 and 𝑦2 are
linearly independent. Two functions 𝑦1 and 𝑦2 are said to be linearly independent if there is no
constant number 𝑐 such that 𝑦1 = 𝑐𝑦2 or 𝑦2 = 𝑐𝑦1 .
Theorem 5.3
If 𝑦1 and 𝑦2 are two linearly independent solutions of the homogeneous equation

𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥)𝑦 = 0,

then the general solution of the above equation is of the form

𝑦 = 𝑐1 𝑦1 + 𝑐2 𝑦2
♥

For a complete proof, the reader may read Section 15 in the book “Differential equations
with applications and historical notes" by George F. Simmons. The following is a rough idea.

By the uniqueness of the solution (Theorem 5.1), to show that any solution 𝑦 can be
written as 𝑐1 𝑦1 + 𝑐2 𝑦2 , it suffices to show that, for some 𝑥0 , the following system of equations
for 𝑐1 and 𝑐2 is solvable
⎧
⎪
⎪
⎪𝑐1 𝑦1 (𝑥0 ) + 𝑐2 𝑦2 (𝑥0 ) = 𝑦(𝑥0 )
⎨
⎪
⎪
⎪ 𝑐1 𝑦 ′ (𝑥0 ) + 𝑐2 𝑦2′ (𝑥0 ) = 𝑦 ′ (𝑥0 ).
⎩ 1
From linear algebra, or college algebra, we know that the system is solvable if the determinant

|𝑦 (𝑥 ) 𝑦 (𝑥 )|
| 1 0 2 0 |
| ′ | ∶= 𝑦1 (𝑥0 )𝑦2′ (𝑥0 ) − 𝑦2 (𝑥0 )𝑦1′ (𝑥0 ) (5.1)
|𝑦1 (𝑥0 ) 𝑦2′ (𝑥0 )|
| |

78 Lecture Notes for MA451

 5.1 Homogeneous Linear Second Order Equations

is nonzero.

To show that the linear independence implies the nonzero of the determinant, we will
need to study the following function.

The function of 𝑥 defined by

|𝑦 𝑦 |
| 1 2|
𝑊 (𝑦1 , 𝑦2 ) = | ′ ′ | = 𝑦1 𝑦2′ − 𝑦2 𝑦1′
|𝑦1 𝑦2 |
| |

is called the Wronskian determinant or simply Wronskian of 𝑦1 and 𝑦2 .

Wronskians has the following properties.

Lemma 5.1
Suppose 𝑝 and 𝑞 are continuous on (𝑎, 𝑏), let 𝑦1 and 𝑦2 be solutions of

𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥)𝑦 = 0

on (𝑎, 𝑏).
Let 𝑥0 be any point in (𝑎, 𝑏). Then
𝑥
𝑊 (𝑥) = 𝑊 (𝑥0 )𝑒 − ∫𝑥0 𝑝(𝑡) d 𝑡 , 𝑎 < 𝑥 < 𝑏,

which is called Abel’s formula.

Therefore, the Wronskian 𝑊 is either identically zero or never zeros on (𝑎, 𝑏).
♥

Proof Let 𝑦1 and 𝑦2 be two solutions of the homogeneous and linear second order equation.
Then
𝑦1′′ + 𝑝(𝑥)𝑦1′ + 𝑞(𝑥)𝑦1 = 0, (5.2)

𝑦2′′ + 𝑝(𝑥)𝑦2′ + 𝑞(𝑥)𝑦2 = 0. (5.3)

Multiplying equation 5.2 by 𝑦2 and equation 5.3 by 𝑦1 and taking the difference of the resulting
equations yields
𝑦1′′ 𝑦2 − 𝑦2′′ 𝑦1 + 𝑝(𝑥)𝑦1′ 𝑦2 − 𝑦2′ 𝑦1 = 0.

Note that
(𝑦1′ 𝑦2 − 𝑦2′ 𝑦1 )′
=(𝑦1′′ 𝑦2 + 𝑦1′ 𝑦2′ ) − (𝑦1′′ 𝑦2 + 𝑦1′ 𝑦2′ )
=𝑦1′′ 𝑦2 − 𝑦2′′ 𝑦1 .
Let 𝑊 = 𝑦1 𝑦2′ − 𝑦2 𝑦1′ . Then
−𝑊 ′ − 𝑝(𝑥)𝑊 = 0.

79 Lecture Notes for MA451

 5.1 Homogeneous Linear Second Order Equations

Solving the equation yields that the general solution is

−𝑝(𝑥) d 𝑥
𝑊 = 𝑒∫ .

If 𝑊 (𝑥0 ) is given, then a specific solution is

𝑥
𝑊 = 𝑊 (𝑥0 )𝑒 ∫𝑥0 −𝑝(𝑡) d 𝑡 .

𝑥
Since 𝑒 ∫𝑥0 𝑝(𝑡) d 𝑡 in nonzero for all 𝑥, 𝑊 (𝑥) is identically zero if 𝑊 (𝑥0 ) = 0, or 𝑊 (𝑥) is never
zero if 𝑊 (𝑥0 ) ≠ 0. ■
The above lemma also implies the following useful result.
Corollary 5.1
The following are equivalent
1. 𝑊 (𝑥) is identically zero.
2. 𝑊 (𝑥0 ) is zero for some 𝑥0 .
♥

With the above properties of Wronskian, one can show that linear dependence is equiv-
alent to that 𝑊 (𝑥) ≡ 0.
Lemma 5.2
If 𝑦1 and 𝑦2 are two solutions of equation 𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥)𝑦 = 0, then they are linearly
dependent if and only if their Wronskian 𝑊 (𝑦1 , 𝑦2 ) = 𝑦1 𝑦2′ − 𝑦2 𝑦1′ is identically zero.
♥

Theorem 5.3 follows from the above lemma. Because linear independence implies that
𝑊 (𝑥0 ) ≠ 0 for some 𝑥0 , which implies the system of equations (5.1) has a nontrivial solution.

Example 5.3 Consider the equation

𝑦 ′′ − 3𝑦 ′ + 2𝑦 = 0

Determine whether
𝑦 = 𝑐1 𝑒 𝑥 + 𝑐2 𝑒 2𝑥

is the general solution, where 𝑐1 and 𝑐2 are arbitrary constants.

Solution To see if 𝑦 = 𝑐1 𝑒 𝑥 +𝑐2 𝑒 2𝑥 is the general solution, it suffices to check that 𝑦1 = 𝑒 𝑥 and 𝑦2 = 𝑒 2𝑥
are two linearly independent solutions. Since

(𝑒 𝑥 )′′ − 3(𝑒 𝑥 )′ + 2𝑒 𝑥 = 𝑒 𝑥 − 3𝑒 𝑥 + 2𝑒 𝑥 = 0

and
(𝑒 2𝑥 )′′ − 3(𝑒 2𝑥 )′ + 2𝑒 2𝑥 = 4𝑒 2𝑥 − 6𝑒 2𝑥 + 2𝑒 2𝑥 = 0,

80 Lecture Notes for MA451

 5.1 Homogeneous Linear Second Order Equations

both 𝑦1 and 𝑦2 are solutions.

As there is no constant number 𝑐 such that the equations 𝑒 𝑥 = 𝑐𝑒 2𝑥 and 𝑐𝑒 𝑥 = 𝑒 2𝑥 hold true for
all 𝑥, the functions 𝑦1 and 𝑦2 are linearly independent.

Therefore, 𝑦 = 𝑐1 𝑒 𝑥 + 𝑐2 𝑒 2𝑥 is the general solution. ■

Note the equation 𝑦 ′′ − 3𝑦 ′ + 2𝑦 = 0 again can be solve using substitutions (try it as an
exercise).

Example 5.4 Verify Abel’s formula for the following differential equations and the correspond-
ing solutions.
𝑦 ′′ − 𝑦 = 0; 𝑦1 = 𝑒 𝑥 , 𝑦2 = 𝑒 −𝑥 .

Solution Since 𝑝 ≡ 0, the integral ∫ 𝑝(𝑥) d 𝑥 is a constant. A direct calculation shows that

|𝑒 𝑥 𝑒 −𝑥 |
| |
𝑊 (𝑥) = | 𝑥 | = 𝑒 𝑥 (−𝑒 −𝑥 ) − 𝑒 𝑥 𝑒 −𝑥 = −2
|𝑒 −𝑒 −𝑥 |
| |

for all 𝑥. This verifies Abel’s formula. ■

 Exercise 5.1 Verify Abel’s formula for the following differential equations and the corre-
sponding solutions,
𝑦 ′′ + 𝑦 = 0; 𝑦1 = cos 𝑥, 𝑦2 = sin 𝑥.

Solution Since 𝑝 ≡ 0, we can verify Abel’s formula by showing that 𝑊 is constant. A direct
calculation shows that
| cos 𝑥 sin 𝑥 |
| |
𝑊 (𝑥) = | |
|− sin 𝑥 cos 𝑥 |
| |
2
= cos 𝑥 − sin 𝑥(− sin 𝑥)
=1
for all 𝑥. ■

5.1.4 The reduction of order method*

Lemma 5.1 also provides a way of finding the another solution 𝑦2 given a solution 𝑦1 .
Proposition 5.1
Suppose 𝑦1 is a solution to 𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥)𝑦 = 0. Then a function 𝑦2 is a solution if and only
if
𝑒 − ∫ 𝑝(𝑥) d 𝑥
𝑦2 (𝑥) = 𝑦1 (𝑥) ∫ 2 d 𝑥.
𝑦
( 1 ) (𝑥)
♠

81 Lecture Notes for MA451

 5.1 Homogeneous Linear Second Order Equations

Proof Let 𝑦1 is a solution. If 𝑦2 is another solution, then

𝑦1 𝑦2′ − 𝑦1′ 𝑦2 = 𝑊 = 𝑒 ∫ 𝑝(𝑥) d 𝑥

.
′
Note that 𝑦1 𝑦2′ − 𝑦1′ 𝑦2 = 𝑦12 ( 𝑦𝑦12 ) . This equation can be solved by a linear substitution 𝑧 = 𝑦2
𝑦1
.
′ ′
Indeed, since 𝑧 ′ = 𝑦2 𝑦1𝑦−𝑦2 2 𝑦1 , the equation can be rewritten as
1

𝑦12 𝑧 ′ = 𝑒 ∫ 𝑝(𝑥) d 𝑥
.

Dividing both sides by 𝑦12 and integrating directly yields

𝑦2 𝑒 − ∫ 𝑝(𝑥) d 𝑥
=𝑧=∫ d 𝑥.
𝑦1 ( 𝑦12 )

Hence,
𝑒− ∫ 𝑝(𝑥) d 𝑥
𝑦2 (𝑥) = 𝑦1 ∫ 2 d 𝑥.
(𝑦1 )

Conversely, by the Fundamental Theorem of Calculus,

𝑒− ∫ 𝑝(𝑥) d 𝑥
𝑒− ∫ 𝑝(𝑥) d 𝑥
𝑦2′ = 𝑦1′ ∫ 2 d𝑥 +
(𝑦1 ) 𝑦1

and
𝑒− ∫ 𝑝(𝑥) d 𝑥
𝑒− ∫ 𝑝(𝑥) d 𝑥
−𝑝(𝑥)𝑒 − ∫ 𝑝(𝑥) d 𝑥
𝑦1 − 𝑦1′ 𝑒 − ∫ 𝑝(𝑥) d 𝑥
𝑦2′′ =𝑦1′′ ∫ 2 d 𝑥 + 𝑦1′ 2 + .
(𝑦1 ) (𝑦1 ) 𝑦12
𝑒− ∫ 𝑝(𝑥) d 𝑥
𝑝(𝑥)𝑒 − ∫ 𝑝(𝑥) d 𝑥
=𝑦1′′ ∫ 2 d𝑥 − .
(𝑦1 ) 𝑦1

Therefore,

𝑦2′′ + 𝑝(𝑥)𝑦2′ + 𝑦2
𝑒− ∫ 𝑝(𝑥) d 𝑥
𝑝(𝑥)𝑒 − ∫ 𝑝(𝑥) d 𝑥
=𝑦1′′ ∫ 2 d𝑥 −
(𝑦1 ) 𝑦1

𝑒− ∫ 𝑝(𝑥) d 𝑥
𝑒− ∫ 𝑝(𝑥) d 𝑥
𝑒− ∫ 𝑝(𝑥) d 𝑥
+ 𝑝(𝑥) 𝑦1′ ∫ 2 d𝑥 + + 𝑞(𝑥)𝑦1 ∫ 2 d𝑥
( (𝑦1 ) 𝑦1 ) (𝑦1 )
𝑒− ∫ 𝑝(𝑥) d 𝑥
𝑒− ∫ 𝑝(𝑥) d 𝑥
𝑒− ∫ 𝑝(𝑥) d 𝑥
=𝑦1′′ ∫ 2 d 𝑥 + 𝑝(𝑥)𝑦1′ ∫ 2 d 𝑥 + 𝑞(𝑥)𝑦1 ∫ 2 d𝑥
(𝑦1 ) (𝑦1 ) (𝑦1 )
𝑒− ∫ 𝑝(𝑥) d 𝑥
=(𝑦1′′ + 𝑝(𝑥)𝑦1′ + 𝑞(𝑥)𝑦1 ) ∫ 2 d𝑥
(𝑦1 )
=0.

82 Lecture Notes for MA451

 5.1 Homogeneous Linear Second Order Equations

So 𝑦2 is also a solution. ■
Example 5.5 Consider the equation

𝑥 2 𝑦 ′′ + 𝑥𝑦 ′ − 𝑦 = 0.

1. Guess a solution and find another solution that is linearly independent to it.
2. Find the general solution.

Solution Let 𝑦1 = 𝑥. Then it is a solution. Rewrite the equation so that the leading coefficient is 1:

𝑦′ 𝑦
𝑦 ′′ + − = 0.
𝑥 𝑥2

So 𝑝(𝑥) = 𝑥1 . By the proposition, another solution can be calculated as follows

𝑒− ∫ 𝑝(𝑥) d 𝑥
𝑦2 =𝑦1 ∫ d𝑥
( 𝑦12 )
1
𝑒− ∫ 𝑥 d 𝑥
=𝑥 ∫ d𝑥
( 𝑥2 )
𝑥 −1
=𝑥 ∫ d𝑥
( 𝑥2 )
1
=𝑥 ∫ ( 3 ) d 𝑥
𝑥
1
=𝑥 (− 2 )
2𝑥
1
=− .
2𝑥
1
It’s not so hard to check that 𝑥 and 2𝑥
are linearly independent.

Therefore, the function 𝑦 = 𝑐1 𝑥 + 𝑐2

𝑥
is the general solution. ■
 Exercise 5.2 Consider the equation

𝑥𝑦 ′′ − 𝑦 ′ = 0.

1. Guess a solution and find another solution that is linearly independent to it.
2. Find the general solution.

Solution Clearly, the constant function 𝑦1 = 1 is a solution. Rewrite the equation so that the
leading coefficient is 1:
𝑦′
𝑦 ′′ − = 0.
𝑥

83 Lecture Notes for MA451

 5.2 Constant Coefficient Homogeneous Equations

So 𝑝(𝑥) = − 𝑥1 . By the proposition, another solution can be calculated as follows

𝑒− ∫ 𝑝(𝑥) d 𝑥
𝑦2 =𝑦1 ∫ d𝑥
( 𝑦12 )
1
d𝑥
= ∫ 𝑒∫ 𝑥 d𝑥

= ∫ 𝑒 ln 𝑥 d 𝑥

=∫ 𝑥 d𝑥
𝑥2
=
2
2
If there are numbers such that 𝑐1 𝑦1 + 𝑐2 𝑦2 =, then 𝑐1 + 𝑐2 𝑥2 = 0 for all 𝑥, which can only hold
true if 𝑐1 = 𝑐2 = 0. Therefore, 𝑦1 and 𝑦2 are linearly independent.

Therefore, the function 𝑦 = 𝑐1 + 𝑐2 𝑥 2 is the general solution. ■

5.2 Constant Coefficient Homogeneous Equations

A linear second order differential equation is said to be a constant coefficient equation if
it can be written as
𝑦 ′′ + 𝑎𝑦 ′ + 𝑏𝑦 = 𝑓 (𝑥),

where 𝑎 and 𝑏 are constant real numbers.

In this section, we consider the homogeneous constant coefficient equation

𝑦 ′′ + 𝑎𝑦 ′ + 𝑏𝑦 = 0.

How to solve this type of equations? Recall, by Theorem 5.3, if 𝑦1 and 𝑦2 are two solutions
not proportional to each other, then 𝑦 = 𝑐1 𝑦1 + 𝑐2 𝑦2 is the general solution. The question is how
to find two linearly independent solutions.

Since 𝑎 and 𝑏 are constants, a solution function has to have the same "degree" as its
derivatives. We know such a function, 𝑦 = 𝑒 𝑟 𝑥 . Plugging it into the equation 𝑦 ′′ + 𝑎𝑦 ′ + 𝑏𝑦 = 0
yields
(𝑒 𝑟 𝑥 )′′ + 𝑎(𝑒 𝑟 𝑥 )′ + 𝑏𝑒 𝑟 𝑥 =0
𝑟(𝑒 𝑟 𝑥 )′ + 𝑎𝑟𝑒 𝑟 𝑥 + 𝑏𝑒 𝑟 𝑥 =0
𝑟 2 𝑒 𝑟 𝑥 + 𝑎𝑟𝑒 𝑟 𝑥 + 𝑏𝑒 𝑟 𝑥 =0
𝑟 2 + 𝑎𝑟 + 𝑏 =0
So 𝑟 is a solution of the quadratic equation 𝑥 2 + 𝑎𝑥 + 𝑏 = 0.

84 Lecture Notes for MA451

 5.2 Constant Coefficient Homogeneous Equations

We call 𝑝(𝑟) = 𝑟 2 + 𝑎𝑟 + 𝑏 the characteristic polynomial of the equation 𝑦 ′′ + 𝑎𝑦 ′ + 𝑏𝑦 = 0.

The quadratic equation 𝑝(𝑟) = 0 is called the characteristic equation.

Example 5.6 Consider the equation

𝑦 ′′ − 𝑦 ′ − 2𝑦 = 0.

Find the general solution.

Solution We expect that 𝑦 = 𝑒 𝑟 𝑥 to be a solution for some 𝑟. Since

𝑦 ′ =𝑟𝑒 𝑟 𝑥
𝑦 ′′ =𝑟 2 𝑒 𝑟 𝑥

Then
𝑦 ′′ − 𝑦 ′ − 2𝑦 = 𝑟 2 𝑒 𝑟 𝑥 − 𝑟𝑒 𝑟 𝑥 − 2𝑒 𝑟 𝑥 = (𝑟 2 − 𝑟 − 2)𝑒 𝑟 𝑥 .

Therefore, 𝑦 = 𝑒 𝑟 𝑥 is a solution if and only if

(𝑟 2 − 𝑟 − 2)𝑒 𝑟 𝑥 = 0.

Equivalently,
𝑟 2 − 𝑟 − 2 = 0.

Solving the equation by the factorization

𝑟 − 𝑟 − 2𝑟 = (𝑟 + 1)(𝑟 − 2)

yields, 𝑟 = −1 or 𝑟 = 2. Therefore, 𝑦 = 𝑒 −𝑥 and 𝑦 = 𝑒 2𝑥 are both solutions. Moreover, if there are

numbers 𝑑1 and 𝑑2 such that
𝑑1 𝑒 −𝑥 + 𝑑2 𝑒 2𝑥 = 0

Then
𝑑1 + 𝑑2 𝑒 3𝑥 = 0

equivalently, 𝑑1 = 0 and 𝑑2 = 0. Therefore, the two solutions are linearly independent. Hence the
general solution is
𝑦 = 𝑐1 𝑒 −𝑥 + 𝑐2 𝑒 2𝑥 .

■
The methods used in this example works well if 𝑟1 and 𝑟2 are two distinct real root. If
𝑟1 = 𝑟2 , then the method of reduction of order will be needed. If 𝑟1 and 𝑟2 are two conjugate
complex solutions, the Euler’s formula and the fact that 𝑎 + i𝑏 = 0 if and only if 𝑎 = 0 and 𝑏 = 0
will be need.

85 Lecture Notes for MA451

 5.2 Constant Coefficient Homogeneous Equations

The equation can also be solved using decomposition and substitution as follows. Let 𝑟1
and 𝑟2 be two roots of the characteristic equation. Then

𝑦 ′′ + 𝑎𝑦 ′ + 𝑏𝑦 = (𝑦 ′ − 𝑟1 𝑦)′ + 𝑟2 (𝑦 ′ − 𝑟1 𝑦).

Using this decomposition, we can reduced the equation 𝑦 ′′ +𝑎𝑦 ′ +𝑏𝑦 = 0 to first order equations
by the substitution 𝑧 = 𝑦 ′ − 𝑟1 𝑦.
Theorem 5.4
Let 𝑝(𝑟) = 𝑟 2 + 𝑎𝑟 + 𝑏 be the characteristic polynomial of

𝑦 ′′ + 𝑎𝑦 ′ + 𝑏𝑦 = 0.

1. If 𝑝(𝑟) = 0 has distinct real roots 𝑟1 and 𝑟2 , then the general solution of the equation is

𝑦 = 𝑐1 𝑒 𝑟1 𝑥 + 𝑐2 𝑒 𝑟2 𝑥 .

2. If 𝑝(𝑟) = 0 has a repeated root 𝑟 then the general solution of the equation is

𝑦 = 𝑒 𝑟 𝑥 (𝑐1 + 𝑐2 𝑥).

3. If 𝑝(𝑟) = 0 has complex conjugate roots 𝑟1 = 𝛼 + i𝛽 and 𝑟2 = 𝛼 − i𝛽, where 𝛽 > 0, then
the general solution of the equation is

𝑦 = 𝑒 𝛼𝑥 (𝑐1 cos 𝛽𝑥 + 𝑐2 sin 𝛽𝑥).

♥

Proof Let 𝑟1 and 𝑟2 be two roots of the characteristic equation

𝑟 2 + 𝑎𝑟 + 𝑏 = 0.

Then
𝑦 ′′ + 𝑎𝑦 ′ + 𝑏𝑦 =𝑦 ′′ − (𝑟1 + 𝑟2 )𝑦 ′ + 𝑟1 𝑟2 𝑦
=(𝑦 ′′ − 𝑟1 𝑦 ′ ) − 𝑟2 (𝑦 ′ − 𝑟1 𝑦)
=(𝑦 ′ − 𝑟1 𝑦)′ − 𝑟2 (𝑦 ′ − 𝑟1 𝑦)
Therefore, the equation 𝑦 ′′ + 𝑎𝑦 ′ + 𝑏𝑦 can be reduced to first order differential equations by

86 Lecture Notes for MA451

 5.2 Constant Coefficient Homogeneous Equations

the substitution 𝑧 = 𝑦 ′ − 𝑟1 𝑦:

𝑦 ′′ + 𝑎𝑦 ′ + 𝑏𝑦 =0
(𝑦 ′ − 𝑟1 𝑦)′ − 𝑟2 (𝑦 ′ − 𝑟1 𝑦) =0
𝑧 ′ − 𝑟2 𝑧 =0
𝑧′
=𝑟2
𝑧
𝑧 =𝑐1 𝑒 𝑟2 𝑥 .

Thus,
𝑦 ′ − 𝑟1 𝑦 = 𝑐1 𝑒 𝑟2 𝑥 .

Since the coefficients of 𝑦 ′ and 𝑦 are 1 and −𝑟1 respectively, the integrating factor is

𝑟(𝑥) = 𝑒 ∫ (−𝑟1 ) d 𝑥 = 𝑒 −𝑟1 𝑥 .

Therefore,
𝑦 ′ − 𝑟1 𝑦 =𝑐1 𝑒 𝑟2 𝑥
𝑒 −𝑟1 𝑥 𝑦 ′ − 𝑟1 𝑒 −𝑟1 𝑥 𝑦 =𝑐1 𝑒 −𝑟1 𝑥 𝑒 𝑟2 𝑥
(𝑦𝑒 −𝑟1 𝑥 )′ =𝑐1 𝑒 −𝑟1 𝑥 𝑒 𝑟2 𝑥

𝑦𝑒 −𝑟1 𝑥 =𝑐1 ∫ 𝑒 (𝑟2 −𝑟1 )𝑥 d 𝑥

𝑦 =𝑐1 𝑒 𝑟1 𝑥 ∫ 𝑒 (𝑟2 −𝑟1 )𝑥 d 𝑥

If 𝑟1 = 𝑟2 = 𝑟, then
(𝑟2 −𝑟1 )𝑥
∫ 𝑒 d 𝑥 = ∫ d 𝑥 = 𝑥 + 𝑐.

Therefore, the general solution is

𝑦 = 𝑐1 𝑒 𝑟 𝑥 + 𝑐2 𝑥𝑒 𝑟 𝑥

If 𝑟1 ≠ 𝑟2 , then
(𝑟2 −𝑟1 )𝑥 𝑒 (𝑟2 −𝑟1 )𝑥
∫ 𝑒 d 𝑥 = +𝑐
𝑟2 − 𝑟1
Therefore, the general solution is
𝑦 = 𝑐1 𝑒 𝑟1 𝑥 + 𝑐2 𝑒 𝑟2 𝑥 .

If 𝑟1 and 𝑟2 , are real numbers, then 𝑦 = 𝑐1 𝑒 𝑟1 𝑥 + 𝑐2 𝑒 𝑟2 𝑥 is a real-valued function and is the

general solution.

If 𝑟1 = 𝛼 + i𝛽 and 𝑟2 = 𝛼 − i𝛽 are complex numbers, then 𝑦 = 𝑐1 𝑒 𝑟1 𝑥 + 𝑐2 𝑒 𝑟2 𝑥 is a solution but

it is a complex-valued function. We want to find a real-valued function from it.

87 Lecture Notes for MA451

 5.2 Constant Coefficient Homogeneous Equations

Euler’s formula says

𝑒 i𝑥 = cos 𝑥 + i sin 𝑥 𝑒 −i𝑥 = cos 𝑥 − i sin 𝑥.

Then
𝑐1 𝑒 𝑟1 𝑥 + 𝑐2 𝑒 𝑟2 𝑥
=𝑐1 𝑒 𝛼𝑥+i𝛽𝑥 + 𝑐2 𝑒 𝛼𝑥−i𝛽𝑥
=𝑐1 𝑒 𝛼𝑥 (cos(𝛽𝑥) + i sin(𝛽𝑥)) + 𝑐2 𝑒 𝛼𝑥 (cos(𝛽𝑥) − i sin(𝛽𝑥))
=𝑐1 𝑒 𝛼𝑥 cos(𝛽𝑥) + 𝑐2 𝑒 𝛼𝑥 (cos(𝛽𝑥) + i (𝑐1 𝑒 𝛼𝑥 sin(𝛽𝑥) − 𝑐2 𝑒 𝛼𝑥 sin(𝛽𝑥))
=(𝑐1 + 𝑐2 )𝑒 𝛼𝑥 cos(𝛽𝑥) + i(𝑐1 − 𝑐2 )𝑒 𝛼𝑥 sin(𝛽𝑥)

Note that a complex-valued function 𝑤(𝑥) = 𝑢(𝑥) + i𝑣(𝑥) is a solution of 𝑦 ′′ + 𝑎𝑦 ′ + 𝑏𝑦 = 0 if

and only if both 𝑢(𝑥) and 𝑣(𝑥) are solutions. Therefore, by Theorem 5.3, the real-valued linear
combination
𝑦 = 𝐶1 𝑒 𝛼𝑥 cos(𝛽𝑥) + 𝐶2 𝑒 𝛼𝑥 sin(𝛽𝑥)

is the general solution. ■

Example 5.7 Solve the initial value problem

𝑦 ′′ + 6𝑦 ′ + 5𝑦 = 0, 𝑦(0) = 3, 𝑦 ′ (0) = −1.

Solution The characteristic equation of the differential equation is

𝑟 2 + 6𝑟 + 5 = 0.

Solving it by factoring yields two distinct real roots: 𝑟1 = −1 and 𝑟2 = −5. By Theorem 5.4, the
general solution of the differential equation is

𝑦 = 𝑐1 𝑒 −𝑥 + 𝑐2 𝑒 −5𝑥 .

Since 𝑦 satisfies the initial conditions 𝑦(0) = 3 and 𝑦 ′ (0) = −1, and the first derivative of 𝑦 is

𝑦 ′ = −𝑐1 𝑒 −𝑥 − 5𝑐2 𝑒 −5𝑥 ,

the constants 𝑐1 and 𝑐2 satisfy the following system of equations

𝑐1 + 𝑐2 =3
−𝑐1 − 5𝑐2 = − 1.

88 Lecture Notes for MA451

 5.2 Constant Coefficient Homogeneous Equations

The solution of this system is 𝑐1 = 27 , 𝑐2 = − 12 . Therefore, the solution of the initial value problem is

7 1
𝑦 = 𝑒 −𝑥 − 𝑒 −5𝑥 .
2 2
■
Example 5.8 Find the general solution of

𝑦 ′′ − 2𝑦 ′ + 𝑦 = 0.

Solution The characteristic equation is

𝑟 2 − 2𝑟 + 1 = 0

which is the same as

(𝑟 − 1)2 = 0

Hence, 𝑟 = 1 is the repeated root. By Theorem 5.4, the general solution is

𝑦 = (𝑐1 + 𝑐2 𝑥)𝑒 𝑥 .

■
Example 5.9 Solve the initial value problem

𝑦 ′′ − 2𝑦 ′ + 2𝑦 = 0, 𝑦(0) = 1, 𝑦 ′ (0) = 2.

Solution The characteristic equation is

𝑟 2 − 2𝑟 + 2 = 0

which is the same as

(𝑟 − 1)2 + 1 = 0

Hence, 𝑟 = 1 ± i are the solutions. By Theorem 5.4, the general solution is

𝑦 = 𝑒 𝑥 (𝑐1 cos 𝑥 + 𝑐2 sin 𝑥).

Since 𝑦(0) = 1, 𝑦 ′ (0) = 2 and

𝑦 ′ = 𝑒 𝑥 (𝑐1 cos 𝑥 + 𝑐2 sin 𝑥) + 𝑒 𝑥 (𝑐2 cos 𝑥 − 𝑐1 sin 𝑥),

89 Lecture Notes for MA451

 5.2 Constant Coefficient Homogeneous Equations

the constants 𝑐1 and 𝑐2 satisfy the system of equations

⎧
⎪
⎪𝑐1 = 1
⎨
⎪
⎪ 𝑐 +𝑐 =2
⎩1 2

Therefore, 𝑐1 = 1 and 𝑐2 = 1, and the solution of the initial value problem is

𝑦 = 𝑒 𝑥 (cos 𝑥 + sin 𝑥).

■
 Exercise 5.3 Find the general solution of the differential equation

𝑦 ′′ − 3𝑦 ′ + 2𝑦 = 0.

Solution Solving the characteristic equation

𝑟 2 − 3𝑟 + 2 = 0

yields 𝑟1 = 1 and 𝑟2 = 2. By Theorem 5.4, the general solution is

𝑦 = 𝑐1 𝑒 𝑥 + 𝑐2 𝑒 2𝑥 .

■
 Exercise 5.4 Solve the initial value problem

𝑦 ′′ − 4𝑦 ′ + 4𝑦 = 0, 𝑦(0) = 1, 𝑦 ′ (0) = −1.

Solution Solving the characteristic equation

𝑟 2 − 4𝑟 + 4 = 0

yields a double root 𝑟 = 2. By Theorem 5.4, the general solution is

𝑦 = (𝑐1 + 𝑐2 𝑥)𝑒 2𝑥 .

Since 𝑦(0) = 1, 𝑦 ′ (0) = −1, and 𝑦 ′ = 𝑐2 𝑒 2𝑥 + 2(𝑐1 + 𝑐2 𝑥)𝑒 2𝑥 , the constants 𝑐1 and 𝑐2 satisfy
⎧
⎪
⎪𝑐1 = 1
⎨
⎪
⎪2𝑐 + 𝑐 = −1.
⎩ 1 2

90 Lecture Notes for MA451

 5.3 Non-Homogeneous Linear Equations

Therefore, 𝑐1 = 1, 𝑐2 = −3, and the solution of the initial value problem is

𝑦 = (1 − 3𝑥)𝑒 2𝑥 .

■
 Exercise 5.5 Find the general solution of the equation

𝑦 ′′ + 4𝑦 = 0.

Solution Solving the characteristic equation

𝑟2 + 4 = 0

yields two complex roots 𝑟1 = 2i and 𝑟2 = −2i. By Theorem 5.4, the general solution is

𝑦 = 𝑐1 cos(2𝑥) + 𝑐2 sin(2𝑥).

5.3 Non-Homogeneous Linear Equations

In this section, we consider the nonhomogeneous linear second order equation

𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥)𝑦 = 𝑓 (𝑥).

We will assume that 𝑝, 𝑞, and 𝑓 are continuous on a interval (𝑎, 𝑏).

Like first order equation, to find the general solution, it is necessary to find the general
solution of the associated homogeneous equation

𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥)𝑦 = 0,

which is called the complementary equation.

5.3.1 The form of the general solution

If we have two linearly independent solutions of the complementary equation, which is
also known as a fundamental set, and a particular solution of the nonhomogeneous equation,
then the general solution of the nonhomogeneous equation is a linear combination of those
solutions.

91 Lecture Notes for MA451

 5.3 Non-Homogeneous Linear Equations

Theorem 5.5
Suppose 𝑝, 𝑞, and 𝑓 are continuous on (𝑎, 𝑏). Let 𝑦𝑝 be a particular solution of

𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥)𝑦 = 𝑓 (𝑥) (5.4)

on (𝑎, 𝑏), and let {𝑦1 , 𝑦2 } be a fundamental set of solutions of the complementary equation

𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥)𝑦 = 0

on (𝑎, 𝑏). Then 𝑦 is a solution of (5.4) on (𝑎, 𝑏) if and only if

𝑦 = 𝑦𝑝 + 𝑐1 𝑦1 + 𝑐2 𝑦2 ,

where 𝑐1 and 𝑐2 are constants.

♥

Proof Suppose that 𝑦 = 𝑦𝑝 + 𝑐1 𝑦1 + 𝑐2 𝑦2 . Since 𝑦1 and 𝑦2 are solutions of the complementary

equation, and 𝑦𝑝 is a solution of Equation (5.4), we see that

𝑦1′′ + 𝑝(𝑥)𝑦1′ + 𝑞(𝑥)𝑦1 =0

𝑦2′′ + 𝑝(𝑥)𝑦2′ + 𝑞(𝑥)𝑦2 =0
𝑦𝑝′′ + 𝑝(𝑥)𝑦𝑝′ + 𝑞(𝑥)𝑦𝑝 =𝑓 (𝑥)

Then
𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥)𝑦
=(𝑦𝑝 + 𝑐1 𝑦2 + 𝑦2 )′′ + 𝑝(𝑥)(𝑦𝑝 + 𝑐1 𝑦2 + 𝑦2 )′ + 𝑞(𝑥)(𝑦𝑝 + 𝑐1 𝑦2 + 𝑦2 )
=(𝑦𝑝′′ + 𝑝(𝑥)𝑦𝑝′ + 𝑞(𝑥)𝑦𝑝 ) + 𝑐1 (𝑦1′′ + 𝑝(𝑥)𝑦1′ + 𝑞(𝑥)𝑦1 ) + 𝑐2 (𝑦2′′ + 𝑝(𝑥)𝑦2′ + 𝑞(𝑥)𝑦2 )
=𝑓 (𝑥).
Therefore, 𝑦 is a solution of Equation (5.4).

Conversely, suppose that 𝑦 is a solution of Equation (5.4). Then

𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥)𝑦 = 𝑓 (𝑥)

and
(𝑦 − 𝑦𝑝 )′′ + 𝑝(𝑥)(𝑦 − 𝑦𝑝 )′ + 𝑞(𝑥)(𝑦 − 𝑦𝑝 )
=(𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥)𝑦) − (𝑦𝑝′′ + 𝑝(𝑥)𝑦𝑝′ + 𝑞(𝑥)𝑦𝑝 )
=𝑓 (𝑥) − 𝑓 (𝑥)
=0.
Therefore, 𝑦 − 𝑦𝑝 is a solution of the complementary equation. So there exists constants 𝑐1 and
𝑐2 such that
𝑦 − 𝑦𝑝 = 𝑐1 𝑦1 + 𝑐2 𝑦2 ,

92 Lecture Notes for MA451

 5.3 Non-Homogeneous Linear Equations

or
𝑦 = 𝑦𝑝 + 𝑐1 𝑦1 + 𝑐2 𝑦2 .

Thus, 𝑦 is a solution of Equation (5.4) if and only if 𝑦 = 𝑦𝑝 + 𝑐1 𝑦2 + 𝑐2 𝑦2 . ■

Example 5.10 Find the general solution of the equation

𝑦 ′′ − 𝑦 ′ − 6𝑦 = 12.

Solution We first solve the complementary equation

𝑦 ′′ − 𝑦 ′ − 6𝑦 = 0.

Its characteristic equation is

𝑟2 − 𝑟 − 6 = 0

which has two solutions 𝑟1 = −2 and 𝑟2 = 3. Therefore, the fundamental set is {𝑒 −2𝑥 , 𝑒 3𝑥 }.

Note that 𝑦 = 2 is a particular solution. Then the general solution of the equation is

𝑦 = 2 + 𝑐1 𝑒 −2𝑥 + 𝑐2 𝑒 3𝑥 .

■
Example 5.11 Find the general solution of

𝑦 ′′ − 3𝑦 ′ + 2𝑦 = 𝑥 2 + 1

Solution The complementary homogenous equation is

𝑦 ′′ − 3𝑦 ′ + 2𝑦 = 0.

Since its characteristic equation

𝑟 2 − 3𝑟 + 2 = 0

has two real roots 𝑟1 = 1 and 𝑟2 = 2, the general solution to the complementary equation is

𝑦ℎ = 𝑐1 𝑒 𝑥 + 𝑐2 𝑒 2𝑥 .

Next we need to find a particular solution of

𝑦 ′′ − 3𝑦 ′ + 2𝑦 = 𝑥 2 + 1.

Since the right hand side is a degree 2 polynomial and taking derivatives decreases degrees, we may

93 Lecture Notes for MA451

 5.3 Non-Homogeneous Linear Equations

assume that a particular solution is a quadratic function

𝑦𝑝 = 𝑎𝑥 2 + 𝑏𝑥 + 𝑐.

Since
𝑦𝑝′ =2𝑎𝑥 + 𝑏
𝑦𝑝′′ =2𝑎,
the undetermined coefficients 𝑎, 𝑏, and 𝑐 satisfy the following equation for any value of 𝑥

𝑦𝑝′′ − 3𝑦𝑝′ + 2𝑦𝑝 =𝑥 2 + 1

2𝑎 − 6𝑎𝑥 − 3𝑏 + 2𝑎𝑥 2 + 2𝑏𝑥 + 2𝑐 =𝑥 2 + 1
2𝑎𝑥 2 + (2𝑏 − 6𝑎)𝑥 + (2𝑎 − 3𝑏 + 2𝑐) =𝑥 2 + 1

By comparing coefficients of powers of 𝑥, we get a system of linear equations

⎧
⎪
⎪2𝑎 =1
⎪
⎪
⎨2𝑏 − 6𝑎 =0
⎪
⎪
⎪
⎪2𝑎 − 3𝑏 + 2𝑐 = 1.
⎩

Solving this system implies that 𝑎 = 21 , 𝑏 = 3

2
and 𝑐 = 49 . Hence

2𝑥 2 + 6𝑥 + 9
𝑦𝑝 =
4
is a particular solution.

Therefore, the general solution is

2𝑥 2 + 6𝑥 + 9
𝑦= + 𝑐1 𝑒 𝑥 + 𝑐2 𝑒 2𝑥
4
■
The method used to find a particular solution in the above example is known as the
method of undetermined coefficients. We will revisit this method in the next section.

 Exercise 5.6 Find the general solution of the equation

𝑦 ′′ − 𝑦 = 𝑥.

Solution The complementary equation is

𝑦 ′′ − 𝑦 = 0

94 Lecture Notes for MA451

 5.3 Non-Homogeneous Linear Equations

whose general solution is 𝑦ℎ = 𝑐1 𝑒 −𝑥 + 𝑐2 𝑒 𝑥 .

We may assume that 𝑦𝑝 = 𝑎𝑥 + 𝑏 is a particular solution. Then 𝑎 and 𝑏 satisfies the

following equation for all 𝑥
−(𝑎𝑥 + 𝑏) = 𝑥.

Therefore, 𝑎 = −1, 𝑏 = 0, and the particular solution is 𝑦𝑝 = −𝑥.

Hence, the general solution of the equation is

𝑦 = −𝑥 + 𝑐1 𝑒 −𝑥 + 𝑐2 𝑒 𝑥 .

95 Lecture Notes for MA451

 Week 6: Linear Second Order Equations II

10/12–10/21

6.1 The Method of Undetermined Coefficients

In this section, we will study how to solve some linear second order constant coefficient
nonhomogeneous equations
𝑦 ′′ + 𝑎𝑦 ′ + 𝑏𝑦 = 𝑓 (𝑥).

Recall that the general solution of the equation can be written as

𝑦 = 𝑐1 𝑦1 + 𝑐2 𝑦2 + 𝑦𝑝 ,

where 𝑦1 and 𝑦2 are non-proportional solutions of the complementary equation 𝑦 ′′ +𝑎𝑦 ′ +𝑏𝑦 = 0,
and 𝑦𝑝 is a particular solution of the nonhomogeneous equation.

We already knew how to find a general solution 𝑐1 𝑦1 +𝑐2 𝑦2 of the complementary equation.
A particular solution may be found by guessing. The method of undetermined coefficients
provides an approach to find a particular solution.

Because the derivative a polynomial function is still a polynomial, the derivative of an

exponential function is still an exponential, and the derivative of a trigonometric function is
still a trigonometric function, we often expect a same type of function as a particular solution.

6.1.1 Non-linear with basic functions

6.1.1.1 General cases

If 𝑓 (𝑥) is a polynomial, then a particular solution is often a polynomial of the same degree.
If 𝑓 (𝑥) = 𝑃𝑒 𝛼𝑥 , then a particular solution is often of the form 𝐴𝑒 𝛼𝑥 .
If 𝑓 (𝑥) = 𝑃 sin 𝛽𝑥 + 𝑄 cos 𝛽𝑥, then a particular solution is often of the form: 𝐴 sin 𝛽𝑥 +
𝐵 cos 𝛽𝑥.

Example 6.1 Find the general solution of the equation

𝑦 ′′ − 𝑦 = 𝑥 2

Solution We first find the general solution of the complementary equation 𝑦 ′′ − 𝑦 = 0. Since the
 6.1 The Method of Undetermined Coefficients

characteristic equation
𝑟2 − 1 = 0

has two distinct solutions 𝑟1 = −1 and 𝑟2 = 1. The general solution of the complementary equation
is
𝑦ℎ = 𝑐1 𝑒 −𝑥 + 𝑐2 𝑒 𝑥 .

Since the differentiations of a function remains of the same type, and the right-hand side is an
polynomial function, we expect a particular solution

𝑦𝑝 = 𝑎𝑥 2 + 𝑏𝑥 + 𝑐.

Plugging it into the equation yields

(𝑎𝑥 2 + 𝑏𝑥 + 𝑐)′′ − (𝑎𝑥 2 + 𝑏𝑥 + 𝑐) =𝑥 2

2𝑎 − (𝑎𝑥 2 + 𝑏𝑥 + 𝑐) =𝑥 2
−𝑎𝑥 2 − 𝑏𝑥 + (𝑎 − 𝑐) =𝑥 2 .

Comparing coefficients of powers of 𝑥, we see that 𝑎 = −1, 𝑏 = 0 and 𝑐 = 𝑎 = −1.

Therefore, a particular solution is 𝑦𝑝 = −𝑥 2 − 1 and the general solution is

𝑦 = 𝑐1 𝑒 −𝑥 + 𝑐2 𝑒 𝑥 − 𝑥 2 − 1.

■
Example 6.2 Find the general solution of the equation

𝑦 ′′ − 𝑦 ′ − 2𝑦 = 7𝑒 3𝑥

Solution We first find the general solution of the complementary equation 𝑦 ′′ − 𝑦 ′ − 2𝑦. Since the
characteristic equation
𝑟2 − 𝑟 − 2 = 0

has two distinct solutions 𝑟1 = −1 and 𝑟2 = 2. The general solution of the complementary equation
is
𝑦ℎ = 𝑐1 𝑒 −𝑥 + 𝑐2 𝑒 2𝑥 .

Since the differentiations of a function remains of the same type, and the right-hand side is an
exponential function, we expect a particular solution

𝑦𝑝 = 𝑐𝑒 3𝑥 .

97 Lecture Notes for MA451

 6.1 The Method of Undetermined Coefficients

Plugging it into the equation yields

(𝑐𝑒 3𝑥 )′′ − (𝑐𝑒 3𝑥 ) + 𝑐𝑒 3𝑥 =7𝑒 3𝑥

(3𝑐𝑒 3𝑥 )′ − 3𝑐𝑒 3𝑥 + 𝑐𝑒 3𝑥 =7𝑒 3𝑥
9𝑐𝑒 3𝑥 − 2𝑐𝑒 3𝑥 =7𝑒 3𝑥
7𝑐𝑒 3𝑥 =7𝑒 3𝑥
𝑐 =1.

Therefore, a particular solution is 𝑦𝑝 = 𝑒 3𝑥 and the general solution is

𝑦 = 𝑐1 𝑒 −𝑥 + 𝑐2 𝑒 2𝑥 + 𝑒 3𝑥 .

■
Example 6.3 Find the general solution of the equation

𝑦 ′′ − 3𝑦 ′ + 2𝑦 = sin 𝑥.

Solution Since the characteristic equation 𝑟 2 − 3𝑟 + 2 has two solutions 𝑟1 = 1 and 𝑟2 = 2. Then the
general solution of the complementary equation is

𝑦ℎ = 𝑐1 𝑒 −𝑥 + 𝑐2 𝑒 𝑥 .

Since the right-hand side is sin 𝑥 whose higher derivatives are either sin 𝑥 or cos 𝑥, we expect
a particular solution
𝑦𝑝 = 𝐴 cos 𝑥 + 𝐵 sin 𝑥.

Since (sin 𝑥)′′ = − sin 𝑥 and (cos 𝑥)′′ = − cos 𝑥 Plugging it into the equation yields

(𝐴 cos 𝑥 + 𝐵 sin 𝑥)′′ − 3(𝐴 cos 𝑥 + 𝐵 sin 𝑥)′ + 2(𝐴 cos 𝑥 + 𝐵 sin 𝑥) = sin 𝑥
(−𝐴 cos 𝑥 − 𝐵 sin 𝑥) − 3(−𝐴 sin 𝑥 + 𝐵 cos 𝑥) + 2(𝐴 cos 𝑥 + 𝐵 sin 𝑥) = sin 𝑥
(3𝐴 + 𝐵) sin 𝑥 + (𝐴 − 3𝐵) cos 𝑥 = sin 𝑥.

Then 𝐴 and 𝐵 satisfy the following system of equations

⎧
⎪
⎪3𝐴 + 𝐵 = 1
⎨
⎪
⎪𝐴 − 3𝐵 = 0.
⎩

Solving the equation by elimination method yields

3 1
𝐴= and 𝐵= .
10 10

98 Lecture Notes for MA451

 6.1 The Method of Undetermined Coefficients

Therefore, a particular solution is

3 1
𝑦𝑝 = cos 𝑥 + sin 𝑥
10 10
and the general solution is

3 1
𝑦 = 𝑐1 𝑒 −𝑥 + 𝑐2 𝑒 𝑥 + cos 𝑥 + sin 𝑥.
10 10
■

6.1.1.2 Exceptional cases

In the above examples, the right-hand side function contains no factor which is a solu-
tion of the complementary equation. In some exceptional cases, we can adjust the particular
solution to be used.

Example 6.4 Find a particular solution 𝑦𝑝 of

𝑦 ′′ + 2𝑦 ′ − 3𝑦 = 4𝑒 𝑥 .

Solution Since the characteristic equation 𝑟 2 + 2𝑟 − 3 = 0 has two distinct root, the complementary
equation has a general solution
𝑦ℎ = 𝑐1 𝑒 𝑥 + 𝑐2 𝑒 −3𝑥 .

Since 𝑒 𝑥 is a solution, plugging 𝐴𝑒 𝑥 into the equation will give

0 = 𝑒𝑥 .

So 𝐴𝑒 𝑥 can not be a particular solution for any 𝐴.

Remember, when the characteristic equation has a repeated root, an extra solution can be
taking in the form 𝑥𝑒 𝑟 𝑥 . Here, we can try

𝑦𝑝 = 𝐴𝑥𝑒 𝑥 .

Plugging 𝑦𝑝 into the equation yields

(𝐴𝑥𝑒 𝑥 )′′ + 2(𝐴𝑥𝑒 𝑥 )′ − 3𝐴𝑥𝑒 𝑥 =4𝑒 𝑥

𝐴(𝑒 𝑥 + 𝑥𝑒 𝑥 )′ + 2𝐴(𝑒 𝑥 + 𝑥𝑒 𝑥 ) − 3𝐴𝑥𝑒 𝑥 =4𝑒 𝑥
𝐴(𝑒 𝑥 + 𝑒 𝑥 + 𝑥𝑒 𝑥 ) + 2𝐴(𝑒 𝑥 + 𝑥𝑒 𝑥 ) − 3𝐴𝑥𝑒 𝑥 =4𝑒 𝑥
4𝐴𝑒 𝑥 =𝑒 𝑥
𝐴 =1

99 Lecture Notes for MA451

 6.1 The Method of Undetermined Coefficients

Therefore, a particular solution is 𝑦𝑝 = 𝑥𝑒 𝑥 , and the general solution is

𝑦 = 𝑐1 𝑒 𝑥 + 𝑐2 𝑒 −3𝑥 + 𝑥𝑒 𝑥 .

■
Indeed, there is no surprise that the 𝑥𝑒 𝑟 𝑥 terms disappear. The 𝑛-th derivatives has exactly
one term with a factor 𝑥, the term is 𝑟 𝑛 𝑥𝑒 𝑟 𝑥 . Because 𝑟 is a root of the equation 𝑟 2 + 𝑎𝑟 + 𝑏 = 0,
then the sum of those terms equals zero.

What if 𝑥𝑒 𝑟 𝑥 is also a solution? Well, we raise the power of 𝑥 by 1.

Example 6.5 Find the general solution of the equation

𝑦 ′′ − 2𝑦 ′ + 𝑦 = 2𝑒 𝑥 .

Solution The characteristic equation 𝑟 2 − 2𝑟 + 1 = 0 has a repeated root 𝑟 = 1. Then the comple-
mentary equation has a general solution

𝑦ℎ = 𝑐1 𝑒 𝑥 + 𝑐2 𝑥𝑒 𝑥 .

Since both 𝑒 𝑥 ane 𝑥𝑒 𝑥 are solutions, we try 𝑦𝑝 = 𝐴𝑥 2 𝑒 𝑥 .

Plugging 𝑦𝑝 = 𝐴𝑥 2 𝑒 𝑥 into the equation yields

(𝐴𝑥 2 𝑒 𝑥 )′′ − 2(𝐴𝑥 2 𝑒 𝑥 )′ + 𝐴𝑥 2 𝑒 𝑥 =2𝑒 𝑥

𝐴(2𝑥𝑒 𝑥 + 𝑥 2 𝑒 𝑥 )′ − 2𝐴(2𝑥𝑒 𝑥 + 𝑥 2 𝑒 𝑥 ) + 𝐴𝑥 2 𝑒 𝑥 =2𝑒 𝑥
𝐴(2𝑒 𝑥 + 2𝑥𝑒 𝑥 + 2𝑥𝑒 𝑥 + 𝑥 2 𝑒 𝑥 )′ − 2𝐴(2𝑥𝑒 𝑥 + 𝑥 2 𝑒 𝑥 ) + 𝐴𝑥 2 𝑒 𝑥 =2𝑒 𝑥
𝐴(𝑥 2 𝑒 𝑥 − 2𝑥 2 𝑒 𝑥 + 𝑥 2 𝑒 𝑥 ) + 𝐴(4𝑥𝑒 𝑥 − 4𝑥𝑒 𝑥 ) + 2𝐴𝑒 𝑥 =2𝑒 𝑥
2𝐴𝑒 𝑥 =2𝑒 𝑥
𝐴 =1

Therefore, a particular solution is 𝑦𝑝 = 2𝑥 2 𝑒 𝑥 , and the general solution is

𝑦 = 𝑐1 𝑒 𝑥 + 𝑐2 𝑥𝑒 𝑥 + 𝑥 2 𝑒 𝑥 .

■
When the characteristic equation has complex roots, the methods shown in the above
examples still work.

100 Lecture Notes for MA451

 6.1 The Method of Undetermined Coefficients

Example 6.6 Find a particular solution 𝑦𝑝 of

𝑦 ′′ + 𝑦 = 2 sin 𝑥

Solution Since the characteristic equation is 𝑟 2 +1 = 0 which has two conjugate roots ±𝕚, the general
solution of the complementary equation is 𝑦ℎ = 𝑐1 cos 𝑥 + 𝑐2 sin 𝑥. Taking 𝑦𝑝 = 𝐴 sin 𝑥 + 𝐵 cos 𝑥 won’t
work because plugging it into the left-hand side of the equation yields

𝑦𝑝′′ + 𝑦𝑝 = 0.

However, the right-hand sides is 2 sin 𝑥. So 𝐴 sin 𝑥 + 𝐵 cos 𝑥 can not be a solution.

Let’s try
𝑦𝑝 = 𝑥(𝐴 sin 𝑥 + 𝐵 cos 𝑥).

Differentiating 𝑦𝑝 yields

𝑦𝑝′ =𝐴𝑥 cos 𝑥 + 𝐴 sin 𝑥 − 𝐵𝑥 sin 𝑥 + 𝐵 cos 𝑥

𝑦𝑝′′ = − 𝐴𝑥 sin 𝑥 + 2𝐴 cos 𝑥 − 𝐵𝑥 cos 𝑥 − 2𝐵 sin 𝑥,

and
𝑦𝑝′′ + 𝑦𝑝
= − 𝐴𝑥 sin 𝑥 + 2𝐴 cos 𝑥 − 𝐵𝑥 cos 𝑥 − 2𝐵 sin 𝑥 + 𝐴𝑥 sin 𝑥 + 𝐵𝑥 cos 𝑥
=2𝐴 cos 𝑥 − 2𝐵 sin 𝑥
Hence, 𝑦𝑝 = 𝐴𝑥 sin 𝑥 + 𝐵𝑥 cos 𝑥 is a solution if

2𝐴 cos 𝑥 − 2𝐵 sin 𝑥 = 2 sin 𝑥.

Equivalently,
⎧
⎪
⎪2𝐴 = 0
⎨
⎪
⎪ −2𝐵 = 2
⎩
Solving this system of equations yields 𝐴 = 0 and 𝐵 = −1. Therefore, a particular solution is

𝑦𝑝 = −𝑥 cos 𝑥.

The general solutions is

𝑦 = 𝑐1 cos 𝑥 + 𝑐2 sin 𝑥 − 𝑥 cos 𝑥.

■
You will find that a particular solution of 𝑦 ′′ + 𝑎𝑦 ′ = 𝑓 (𝑥) should also have a higher degree
than the polynomial 𝑓 (𝑥). This is because, if 𝑦𝑝 has the same degree as of 𝑓 , then the left-hand

101 Lecture Notes for MA451

 6.1 The Method of Undetermined Coefficients

side will have the degree 1 less due to the differentiation.

Example 6.7 Find a general solution of the equation

𝑦 ′′ + 𝑦 ′ = 2𝑥

Solution The characteristic equation 𝑟 2 + 𝑟 = 0 has two distinct roots 𝑟1 = 0 and 𝑟2 = 1. So the
general solution of the complementary equation is

𝑦ℎ = 𝑐1 + 𝑐2 𝑒 𝑥 .

We are looking for a polynomial as a particular solution. Since differentiating polynomials

decrease it degree by 1 and there is no 𝑦-term in the equation, a particular solution should be 1
degree bigger than 𝑓 (𝑥) = 2𝑥. Indeed, one will find that taking 𝑦𝑝 = 𝑎𝑥 + 𝑏 won’t work.

Let’s try
𝑦𝑝 = 𝑥(𝑎𝑥 + 𝑏) = 𝑎𝑥 2 + 𝑏𝑥.

The first and second derivatives are

𝑦𝑝′ = 2𝑎𝑥 + 𝑏 and 𝑦𝑝′′ = 2𝑎.

Then
𝑦𝑝′′ + 𝑦𝑝′ = 2𝑎 + 2𝑎𝑥 + 𝑏 = 2𝑎𝑥 + (2𝑎 + 𝑏).

Hence, 𝑦𝑝 = 𝑎𝑥 2 + 𝑏𝑥 is a solution given that

2𝑎𝑥 + (2𝑎 + 𝑏) = 2𝑥.

So 𝑎 and 𝑏 satisfy
⎧
⎪
⎪2𝑎 = 1
⎨
⎪
⎪ 2𝑎 + 𝑏 = 0
⎩
Solving the system yields 𝑎 = 1 and 𝑏 = −2 and

𝑦𝑝 = 𝑥 2 − 2𝑥

is a particular solution. Therefore, the general solution is

𝑦 = 𝑐1 + 𝑐2 𝑒 𝑥 + 𝑥 2 − 2𝑥.

102 Lecture Notes for MA451

 6.1 The Method of Undetermined Coefficients

 Exercise 6.1 Find the general solution of

𝑦 ′′ − 4𝑦 = 10𝑒 3𝑥 .

Solution Since the equation 𝑟 2 − 4 = 0 has two distinct roots 𝑟1 = −2 and 𝑟2 = 2, the general
solution of the complementary equation 𝑦 ′′ − 4𝑦 = 0 is

𝑦ℎ = 𝑐1 𝑒 −2𝑥 + 𝑐2 𝑒 2𝑥 .

We expect a particular solution of the form

𝑦𝑝 = 𝑐𝑒 3𝑥 .

Differentiating the function implies 𝑦𝑝′ = 3𝑐𝑒 3𝑥 , 𝑦𝑝′′ = 9𝑐𝑒 3𝑥 , and

𝑦 ′′ − 4𝑦 = 5𝑐𝑒 3𝑥 .

Therefore,
5𝑐𝑒 3𝑥 = 10𝑒 3𝑥

which implies 𝑐 = 2. A particular solution is 𝑦𝑝 = 2𝑒 3𝑥 and the general solution is

𝑦 = 2𝑒 3𝑥 + 𝑐1 𝑒 2𝑥 + 𝑐2 𝑒 −2𝑥 .

■
 Exercise 6.2 Find a general solution of the equation

𝑦 ′′ − 7𝑦 ′ + 12𝑦 = 2𝑒 4𝑥 .

Solution Since the characteristic polynomial 𝑟 2 − 7𝑟 + 12 = 0 has two distinct real solution
𝑟1 = 3 and 𝑟2 = 4, the complementary equation has the general solution

𝑦ℎ = 𝑐1 𝑒 3𝑥 + 𝑐2 𝑒 4𝑥 .

Because 𝑒 4𝑥 is a solution of the complementary equation and right-hand side is 2𝑒 4𝑥 . We

may take a particular solution 𝑦𝑝 = 𝐴𝑥𝑒 4𝑥 .

103 Lecture Notes for MA451

 6.1 The Method of Undetermined Coefficients

Plugging 𝑦𝑝 into the equation yields

(𝐴𝑥𝑒 4𝑥 )′′ − 7(𝐴𝑥𝑒 4𝑥 )′ + 12𝐴𝑥𝑒 4𝑥 =2𝑒 4𝑥

𝐴(𝑒 4𝑥 + 4𝑥𝑒 4𝑥 )′ − 7𝐴(𝑒 4𝑥 + 4𝑥𝑒 4𝑥 ) + 12𝐴𝑥𝑒 4𝑥 =2𝑒 4𝑥
𝐴(𝑒 4𝑥 + 4𝑒 4𝑥 + 16𝑥𝑒 4𝑥 ) − 7𝐴(𝑒 4𝑥 + 4𝑥𝑒 4𝑥 ) + 12𝐴𝑥𝑒 4𝑥 =2𝑒 4𝑥
−2𝐴𝑒 4𝑥 =2𝑒 4𝑥
𝐴=−1

Therefore, a particular solution is

𝑦𝑝 = −𝑥𝑒 4𝑥

and the general solution is

𝑦 = 𝑐1 𝑒 3𝑥 + 𝑐2 𝑒 4𝑥 − 𝑥𝑒 4𝑥 .

6.1.2 The Principle of Superposition

When the function 𝑓 (𝑥) in the equation 𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥)𝑦 = 𝑓 (𝑥) is a sum of several
functions, we can solve the equation by break it into several equation with few terms on the
right-hand side. Indeed, Theorem 5.5 is such an application: if 𝑦ℎ is a solution of 𝑦 ′′ + 𝑝(𝑥)𝑦 ′ +
𝑞(𝑥)𝑦 = 0 and 𝑦𝑝 is a solution of the equation 𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥)𝑦 = 𝑓 (𝑥), then 𝑦𝑝 + 𝑦ℎ is a solution
of the equation 𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥)𝑦 = 0 + 𝑓 (𝑥). In general, we have the principle of superposition
which has analogous in linear algebra.
Theorem 6.1 (Principle of Superposition)
Suppose 𝑦𝑝1 is a particular solution of

𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥)𝑦 = 𝑓1 (𝑥)

and 𝑦𝑝2 is a particular solution of

𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥)𝑦 = 𝑓2 (𝑥).

Then
𝑦𝑝 = 𝑦𝑝1 + 𝑦𝑝2

is a particular solution of

𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥)𝑦 = 𝑓1 (𝑥) + 𝑓2 (𝑥).

♥

104 Lecture Notes for MA451

 6.1 The Method of Undetermined Coefficients

Proof Since
𝑦𝑝′′1 + 𝑝(𝑥)𝑦𝑝′1 + 𝑞(𝑥)𝑦𝑝1 =𝑓1 (𝑥)
𝑦𝑝′′2 + 𝑝(𝑥)𝑦𝑝′2 + 𝑞(𝑥)𝑦𝑝2 =𝑓2 (𝑥),
taking the sum of those two equations yields

(𝑦𝑝1 + 𝑦𝑝2 )′′ + 𝑝(𝑥)(𝑦𝑝1 + 𝑦𝑝2 )′ + 𝑞(𝑥)(𝑦𝑝1 + 𝑦𝑝2 ) = 𝑓1 (𝑥) + 𝑓2 (𝑥).

Therefore, 𝑦𝑝 = 𝑦𝑝1 + 𝑦𝑝2 is a solution of the equation 𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥)𝑦 = 𝑓1 (𝑥) + 𝑓2 (𝑥). ■
Example 6.8 Find a particular solution 𝑦𝑝 of

𝑦 ′′ + 𝑦 ′ + 𝑦 = cos 𝑥 + 𝑥 + 1.

Solution To find a particular solution, we may first find particular solutions for

𝑦 ′′ + 𝑦 ′ + 𝑦 = cos 𝑥

and
𝑦 ′′ + 𝑦 ′ + 𝑦 = 𝑥 + 1.

For the equation 𝑦 ′′ + 𝑦 ′ + 𝑦 = cos(𝑥), since the derivation of cos 𝑥 is − sin 𝑥, we may assume a
solution is 𝑦𝑝1 = 𝑎 sin 𝑥 + 𝑏 cos 𝑥. Then 𝑎 and 𝑏 satisfy the equation

(−𝑎 sin 𝑥 − 𝑏 cos 𝑥) + (𝑎 cos 𝑥 − 𝑏 sin 𝑥) + (𝑎 sin 𝑥 + 𝑏 cos 𝑥) = cos 𝑥

for all 𝑥. Hence, 𝑎 = 1, 𝑏 = 0 and 𝑦𝑝1 = sin 𝑥 is a particular solution.

For the equation 𝑦 ′′ + 𝑦 ′ + 𝑦 = 𝑥, we may assume a solution is 𝑦𝑝2 = 𝑐𝑥 + 𝑑. Then 𝑐 and 𝑑 satisfy

𝑐 + 𝑐𝑥 + 𝑑 = 𝑥 + 1

for all 𝑥. Hence, 𝑐 = 1, 𝑑 = 0, and 𝑦𝑝2 = 𝑥 is a solution.

Therefore, by Theorem 6.1, a particular solution of the original equation is

𝑦 = 𝑦𝑝1 + 𝑦𝑝2 = sin 𝑥 + 𝑥.

■
 Exercise 6.3 Find a particular solution 𝑦𝑝 of

𝑦 ′′ − 𝑦 ′ + 𝑦 = 𝑒 𝑥 + 𝑥.

105 Lecture Notes for MA451

 6.1 The Method of Undetermined Coefficients

Solution To find a particular solution, we may first find particular solutions for

𝑦 ′′ − 𝑦 ′ + 𝑦 = 𝑒 𝑥

and
𝑦 ′′ − 𝑦 ′ + 𝑦 = 𝑥.

For the equation 𝑦 ′′ − 𝑦 ′ + 𝑦 = 𝑒 𝑥 , we may assume a solution is 𝑦𝑝1 = 𝑐𝑒 𝑥 . Then 𝑐 satisfies

the equation
𝑐𝑒 𝑥 − 𝑐𝑒 𝑥 + 𝑐𝑒 𝑥 = 𝑒 𝑥

for all 𝑥. Hence, 𝑐 = 1 and 𝑦𝑝1 = 𝑒 𝑥 is a particular solution.

For the equation 𝑦 ′′ − 𝑦 ′ + 𝑦 = 𝑥, we may assume a solution is 𝑦𝑝2 = 𝑎𝑥 + 𝑏. Then 𝑎 and 𝑏

satisfy
−𝑎 + 𝑎𝑥 + 𝑏 = 𝑥

for all 𝑥. Hence, 𝑎 = 1, 𝑏 = 1, and 𝑦𝑝2 = 𝑥 + 1 is a solution.

Therefore, by Theorem 6.1, a particular solution of the original equation is

𝑦 = 𝑦𝑝1 + 𝑦𝑝2 = 𝑒 𝑥 + 𝑥 + 1.

6.1.3 Nonlinear with Product Functions

When the function 𝑓 (𝑥) is a product of those basic functions, we can expect a same type
of function as a particular solution.

Example 6.9 Find a general solution of the equation

𝑦 ′′ + 2𝑦 = 𝑒 𝑥 sin 𝑥.

Solution The general solution of the complementary equation is 𝑦ℎ = 𝑐1 + 𝑐2𝑒 −2𝑥

We try 𝑦𝑝 = 𝐴𝑒 𝑥 cos 𝑥 + 𝐵𝑒 𝑥 sin 𝑥. The derivatives of 𝑦𝑝 are

𝑦𝑝′ =𝐴𝑒 𝑥 cos 𝑥 − 𝐴𝑒 𝑥 sin 𝑥 + 𝐵𝑒 𝑥 sin 𝑥 + 𝐵𝑒 𝑥 cos 𝑥

=(𝐴 + 𝐵)𝑒 𝑥 cos 𝑥 + (𝐵 − 𝐴)𝑒 𝑥 sin 𝑥

and
𝑦𝑝′′ = 2𝐵𝑒 𝑥 cos 𝑥 − 2𝐴𝑒 𝑥 sin 𝑥.

106 Lecture Notes for MA451

 6.1 The Method of Undetermined Coefficients

Plugging 𝑦𝑝 into the left-hand side of the equation implies

𝑦𝑝′′ + 2𝑦𝑝
=2𝐵𝑒 𝑥 cos 𝑥 − 2𝐴𝑒 𝑥 sin 𝑥 + 2𝐴𝑒 𝑥 cos 𝑥 + 2𝐵𝑒 𝑥 sin 𝑥
=(2𝐵 + 2𝐴)𝑒 𝑥 cos 𝑥 + (2𝐵 − 2𝐴)𝑒 𝑥 sin 𝑥

Hence, 𝑦𝑝 is a solution if

(2𝐵 + 2𝐴)𝑒 𝑥 cos 𝑥 + (2𝐵 − 2𝐴)𝑒 𝑥 sin 𝑥 = 𝑒 𝑥 sin 𝑥

or if
⎧
⎪
⎪2𝐵 + 2𝐴 = 0
⎨
⎪
⎪2𝐵 − 2𝐴 = 1.
⎩
Solving the system yields 𝐴 = − 41 and 𝐵 = 14 . Hence, a particular solution is

1 1
𝑦𝑝 = − 𝑒 𝑥 cos 𝑥 + 𝑒 𝑥 sin 𝑥.
4 4

The general solution is

1 1
𝑦 = 𝑐1 + 𝑐2 𝑒 −2𝑥 − 𝑒 𝑥 cos 𝑥 + 𝑒 𝑥 sin 𝑥.
4 4
■
Example 6.10 Find the general solution of the equation

𝑦 ′′ − 𝑦 = 𝑥 sin 𝑥

Solution The general solution of the complementary equation is

𝑦ℎ = 𝑐1 + 𝑐2 𝑒 −𝑥 .

For a particular solution, we try

𝑦𝑝 = 𝐴𝑥 sin 𝑥 + 𝐵𝑥 cos 𝑥 + 𝐶 sin 𝑥 + 𝐷 cos 𝑥.

The derivatives of 𝑦𝑝 are

𝑦𝑝′ =𝐴𝑥 cos 𝑥 + 𝐴 sin 𝑥 − 𝐵𝑥 sin 𝑥 + 𝐵 cos 𝑥 + 𝐶 cos 𝑥 − 𝐷 sin 𝑥

=𝐴𝑥 cos 𝑥 + (𝐴 − 𝐷) sin 𝑥 − 𝐵𝑥 sin 𝑥 + (𝐵 + 𝐶) cos 𝑥,

107 Lecture Notes for MA451

 6.1 The Method of Undetermined Coefficients

and
𝑦𝑝′ =𝐴 cos 𝑥 − 𝐴𝑥 sin 𝑥 + (𝐴 − 𝐷) cos 𝑥 − 𝐵 sin 𝑥 − 𝐵𝑥 cos 𝑥 − (𝐵 + 𝐶) sin 𝑥
= − 𝐴𝑥 sin 𝑥 + (2𝐴 − 𝐷) cos 𝑥 − 𝐵𝑥 cos 𝑥 − (2𝐵 + 𝐶) sin 𝑥,

Therefore,

𝑦𝑝′′ − 𝑦𝑝
= − 𝐴𝑥 sin 𝑥 + (2𝐴 − 𝐷) cos 𝑥 − 𝐵𝑥 cos 𝑥 − (2𝐵 + 𝐶) sin 𝑥
− (𝐴𝑥 cos 𝑥 + (𝐴 − 𝐷) sin 𝑥 − 𝐵𝑥 sin 𝑥 + (𝐵 + 𝐶) cos 𝑥)
= − 2𝐴𝑥 sin 𝑥 + (−𝐴 − 𝐵 − 2𝐶 + 𝐷) sin 𝑥 + (2𝐴 − 𝐵 − 𝐶 − 𝐷) cos 𝑥

Hence, 𝑦𝑝 is a solution if

−2𝐴𝑥 sin 𝑥 + (−𝐴 − 𝐵 − 2𝐶 + 𝐷) sin 𝑥 + (2𝐴 − 𝐵 − 𝐶 − 𝐷) cos 𝑥 = 𝑥 sin 𝑥

or if
⎧
⎪
⎪ 𝐴=1
⎪
⎪
⎪
⎪𝐵 = 0
⎨
⎪
⎪ 2𝐴 + 𝐷 = 0
⎪
⎪
⎪
⎪ −2𝐵 + 𝐶 = 0
⎩
Consequently, 𝐴 = 1, 𝐵 = 0, 𝐶 = 0, 𝐷 = −2 and

𝑦𝑝 = 𝑥 sin 𝑥 − 2 cos 𝑥

is a particular solution.

The general solution is

𝑦 = 𝑐1 + 𝑐2 𝑒 𝑥

■
 Exercise 6.4 Find a particular solution of

𝑦 ′′ − 3𝑦 ′ + 2𝑦 = 𝑒 3𝑥 (2𝑥 + 1).

Solution We can try a particular solution 𝑦𝑝 = 𝑒 3𝑥 (𝑎𝑥 + 𝑏). Differentiating 𝑦𝑝 yields

𝑦𝑝′ =3𝑒 3𝑥 (𝑎𝑥 + 𝑏) + 𝑎𝑒 3𝑥

=3𝑎𝑥𝑒 3𝑥 + (𝑎 + 3𝑏)𝑒 3𝑥

𝑦𝑝′′ =3𝑎𝑒 3𝑥 + 9𝑎𝑥𝑒 3𝑥 + 3(𝑎 + 3𝑏)𝑒 3𝑥

=9𝑎𝑥𝑒 3𝑥 + (6𝑎 + 9𝑏)𝑒 3𝑥 .

108 Lecture Notes for MA451

 6.1 The Method of Undetermined Coefficients

Plugging 𝑦𝑝 into the equation implies

9𝑎𝑥𝑒 3𝑥 + (6𝑎 + 9𝑏)𝑒 3𝑥 − 3(3𝑎𝑥𝑒 3𝑥 + (𝑎 + 3𝑏)𝑒 3𝑥 ) + 2(𝑎𝑥𝑒 3𝑥 + 𝑏𝑒 3𝑥 ) =𝑒 3𝑥 (2𝑥 + 1)

9𝑎𝑥𝑒 3𝑥 + (6𝑎 + 9𝑏)𝑒 3𝑥 − 9𝑎𝑥𝑒 3𝑥 − (3𝑎 + 9𝑏)𝑒 3𝑥 ) + 2𝑎𝑥𝑒 3𝑥 + 2𝑏𝑒 3𝑥 =𝑒 3𝑥 (2𝑥 + 1)
(2𝑎𝑥 + (3𝑎 + 2𝑏))𝑒 3𝑥 =𝑒 3𝑥 (2𝑥 + 1)
2𝑎𝑥 + (3𝑎 + 2𝑏) =(2𝑥 + 1)

So 𝑦𝑝 is a solution if
⎧
⎪
⎪2𝑎 = 2
⎨
⎪
⎪ 3𝑎 + 2𝑏 = 1
⎩
Solving the system yields 𝑎 = 1 and 𝑏 = −1.

Therefore, a particular solution is

𝑦𝑝 = 𝑒 3𝑥 (𝑥 − 1).

■
More generally, the method of undetermined coefficients can be applied to

𝑦 ′′ + 𝑎𝑦 ′ + 𝑏𝑦 = 𝑃(𝑥)𝑒 𝛼𝑥

and
𝑦 ′′ + 𝑎𝑦 ′ + 𝑏𝑦 = 𝑒 𝛼𝑥 (𝑃(𝑥) cos(𝛽𝑥) + 𝑄(𝑥) sin(𝛽𝑥).

Theorem 6.2
Consider the equation
𝑦 ′′ + 𝑎𝑦 ′ + 𝑏𝑦 = 𝑃(𝑥)𝑒 𝛼𝑥

where 𝐺 a polynomial of degree 𝑘. Let 𝐴(𝑥) = 𝑎𝑘 𝑥 𝑘 + ⋯ + 𝑎1 𝑥 + 𝑎0 be a polynomial of degree 𝑘.

If 𝑒 𝛼𝑥 is not a solution of the complementary equation, then a particular root is

𝑦𝑝 = 𝐴(𝑥)𝑒 𝛼𝑥 .

If 𝑒 𝛼𝑥 is a solution but 𝑥𝑒 𝛼𝑥 is not a solution of the complementary equation, then a

particular solution is
𝑦𝑝 = 𝑥𝐴(𝑥)𝑒 𝛼𝑥 .

If 𝑒 𝛼𝑥 and 𝑥𝑒 𝛼𝑥 are both solutions of the complementary equation, then a particular

solution is
𝑦𝑝 = 𝑥 2 𝐴(𝑥)𝑒 𝛼𝑥 .
♥

109 Lecture Notes for MA451

 6.2 Variation of Parameters

Theorem 6.3
Consider the equation

𝑦 ′′ + 𝑎𝑦 ′ + 𝑏𝑦 = 𝑒 𝛼𝑥 (𝑃(𝑥) cos 𝛽𝑥 + 𝑄(𝑥) sin 𝛽𝑥)

with 𝑃(𝑥) and 𝑄(𝑥) polynomials such that the larger degree is 𝑘.
If 𝛼 + 𝕚𝛽 is not a root of the characteristic polynomial 𝑝(𝑟) = 𝑟 2 + 𝑎𝑟 + 𝑏, then a particular
solution is
𝑦𝑝 = 𝑒 𝛼𝑥 (𝐴(𝑥) cos 𝛽𝑥 + 𝐵(𝑥) sin 𝛽𝑥) ,

where 𝐴(𝑥) = 𝑎𝑘 𝑥 𝑘 + ⋯ + 𝑎1 𝑥 + 𝑎0 and 𝐵(𝑥) = 𝑏𝑘 𝑥 𝑘 + ⋯ + 𝑏1 𝑥 + 𝑏0 .

If 𝛼 + 𝕚𝛽 is not a root of the characteristic polynomial, then a particular solution is

𝑦𝑝 = 𝑒 𝛼𝑥 (𝐴(𝑥) cos 𝛽𝑥 + 𝐵(𝑥) sin 𝛽𝑥) ,

where 𝐴(𝑥) = 𝑎𝑘+1 𝑥 𝑘+1 + ⋯ + 𝑎1 𝑥 + 𝑎0 and 𝐵(𝑥) = 𝑏𝑘+1 𝑥 𝑘+1 + ⋯ + 𝑏1 𝑥 + 𝑏0 .

♥

Using substitution 𝑦𝑝 = 𝑢(𝑥)𝑒 𝛼𝑥 , where 𝛼 is allowed to be a complex number, the proof of

the theorem can be reduced to the case that 𝑓 (𝑥) is a polynomial. For example, the equation
𝑦 ′′ + 𝑎𝑦 + 𝑏𝑦 = 𝑃(𝑥)𝑒 𝛼𝑥 can be reduced to 𝑢 ′′ + (𝑎 + 2𝑟)𝑢 ′ + (𝑟 2 + 𝑎𝑟 + 𝑏)𝑢 = 𝑃(𝑥). Further more,
by the principle of superposition, we may assume that 𝑓 (𝑥) = 𝑎𝑛 𝑥 𝑛 . We can even take 𝑎𝑛 = 𝑛!1 .
Taking derivatives of both sides 𝑛 times, we get an equations.

𝑦 (𝑛+2) + 𝑎𝑦 (𝑛+1) + 𝑏𝑦 (𝑛) = 1.

Let 𝑢 = 𝑦 𝑛 , then
𝑢 ′′ + 𝑎𝑢 ′ + 𝑏𝑢 = 1.

If 𝑏 ≠ 0, then a particular solution is 𝑢𝑝 = 1. If 𝑏 = 0 but 𝑎 ≠ 0, then a particular solution is 𝑢𝑝 = 𝑥𝑎 .

2
If 𝑎 = 0 and 𝑏 = 0, then a particular solution is 𝑢𝑝 = 𝑥2 . Then a particular solution 𝑦𝑝 such that
𝑦𝑝𝑛 = 𝑢𝑝 is a polynomial of 𝑛 terms with degree 𝑛, 𝑛 + 1 or 𝑛 + 2.

Two useful references can be found at the following webpages: https://mathoverflow.

net/questions/124694/reference-for-a-nice-proof-of-undetermined-coefficients and
http://www.math.utah.edu/~gustafso/undetermined-coeff.pdf.

6.2 Variation of Parameters

The method of variation of parameters for first order differential equations can also be
applied to second order equations.

Consider the equation

𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥)𝑦 = 𝑓 (𝑥).

110 Lecture Notes for MA451

 6.2 Variation of Parameters

Assume that 𝑦ℎ = 𝑐1 𝑦1 + 𝑐2 𝑦2 is a general solutions of

𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥)𝑦 = 0.

We assume that the equation

𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥)𝑦 = 𝑓 (𝑥)

has a solution 𝑦 in the form

𝑦 = 𝑣1 𝑦1 + 𝑣2 𝑦2 ,

where 𝑣1 and 𝑣2 are undetermined functions of 𝑥. Computing the derivative of 𝑦 yields

𝑦 ′ = (𝑣1′ 𝑦1 + 𝑣1 𝑦1′ ) + (𝑣2′ 𝑦2 + 𝑣2 𝑦2′ ).

The second derivative is

𝑦 ′′ = (𝑣1′ 𝑦1 )′ + 𝑣1′ 𝑦1′ + 𝑣1 𝑦1′′ ) + ((𝑣2′ 𝑦2 )′ + 𝑣2′ 𝑦2 + 𝑣2 𝑦2′′ ).

Plugging 𝑦1 𝑣1 + 𝑦2 𝑣2 into the left hand side of the nonhomogeneous equation yields

𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥)𝑦
=((𝑣1′ 𝑦1 )′ + 𝑣1′ 𝑦1′ + 𝑣1 𝑦1′′ )) + ((𝑣2′ 𝑦2 )′ + 𝑣2′ 𝑦2′ + 𝑣2 𝑦2′′ )
+ 𝑝(𝑥)(𝑣1′ 𝑦1 + 𝑣1 𝑦1′ ) + 𝑝(𝑥)(𝑣2′ 𝑦2 + 𝑣2 𝑦2′ )
+ 𝑞(𝑥)(𝑣1 𝑦1 + 𝑣2 𝑦2 )
=((𝑣1′ 𝑦1 )′ + 𝑝(𝑥)𝑣1′ 𝑦1 + 𝑣1′ 𝑦1′ )
+ (𝑣2′ 𝑦2 )′ + 𝑝(𝑥)𝑣2′ 𝑦2 + 𝑣2′ 𝑦2′
=(𝑣1′ 𝑦1 + 𝑣2′ 𝑦2 )′ + 𝑝(𝑥)(𝑣1′ 𝑦1 + 𝑣2′ 𝑦2 ) + (𝑣1′ 𝑦1′ + 𝑣2′ 𝑦2′ ).

6.2.1 Variation of Parameter

Compare with the right hand side of the nonhomogeneous equation, we may assume
that 𝑣1 and 𝑣2 satisfy the following system of equations.
⎧
⎪
⎪𝑣1′ 𝑦1 + 𝑣2′ 𝑦2 =0 (1)
⎨
⎪
⎪𝑣 ′ 𝑦 ′ + 𝑣2′ 𝑦2′ = 𝑓 (𝑥) (2)
⎩ 1 1

Solving for 𝑣1′ and 𝑣2′ yields

⎧
⎪ −𝑦2 𝑓 (𝑥)
⎪
⎪ 𝑣1′ =
⎪ 𝑊 (𝑦1 , 𝑦2 )
⎨ (6.1)
⎪
⎪ 𝑦1 𝑓 (𝑥)
⎪
⎪ 𝑣2′ = ,
⎩ 𝑊 (𝑦1 , 𝑦2 )

111 Lecture Notes for MA451

 6.2 Variation of Parameters

where 𝑊 (𝑦1 , 𝑦2 ) = 𝑦1 𝑦2′ − 𝑦2 𝑦1′ is the Wronskian of 𝑦1 and 𝑦2 .

Therefore,
⎧
⎪ −𝑦2 𝑓 (𝑥)
⎪
⎪ 𝑣1′ = ∫ d𝑥
⎪ ( 𝑊 (𝑦1 , 𝑦2 ) )
⎨
⎪
⎪ 𝑦1 𝑓 (𝑥)
⎪
⎪ 𝑣2′ = ∫ d 𝑥,
⎩ ( 𝑊 (𝑦2 , 𝑦1 ) )
and a particular solution of 𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥) = 𝑓 (𝑥) is

−𝑦2 𝑓 (𝑥) 𝑦1 𝑓 (𝑥)

𝑦 = 𝑦1 ∫ d 𝑥 + 𝑦2 ∫ d 𝑥.
( 𝑊 (𝑦1 , 𝑦2 ) ) ( 𝑊 (𝑦2 , 𝑦1 ) )

The method of variation of parameters can be generalized to higher order equations.

Example 6.11 Find a particular solution of the equation

𝑒 3𝑥
𝑦 ′′ − 3𝑦 ′ + 2𝑦 = .
1 + 𝑒𝑥

Solution The characteristic polynomial of the complementary equation is

𝑟 2 − 3𝑟 + 2 = (𝑟 − 1)(𝑟 − 2) = 0

which has two distinct real roots, 𝑟1 = 1 and 𝑟2 = 2. So the complementary equation has two linearly
independent solutions 𝑦1 = 𝑒 𝑥 and 𝑦2 = 𝑒 2𝑥 .

A particular solution of non-homogeneous equations can be taken in the form

𝑦𝑝 = 𝑣1 𝑒 𝑥 + 𝑣2 𝑒 2𝑥 ,

where 𝑣1 and 𝑣2 are functions of 𝑥 that satisfy the following system of equations

𝑣1′ 𝑒 𝑥 + 𝑣2′ 𝑒 2𝑥 =0
𝑒 3𝑥
𝑣1′ 𝑒 𝑥 + 2𝑣2′ 𝑒 2𝑥 = .
1 + 𝑒𝑥

Subtracting the first equation from the second one implies

𝑒 3𝑥
𝑣2′ 𝑒 2𝑥 =
1 + 𝑒𝑥

112 Lecture Notes for MA451

 6.2 Variation of Parameters

So
𝑒𝑥
𝑣2′ =
1 + 𝑒𝑥
𝑒𝑥
𝑣2 = ∫ d𝑥
( 1 + 𝑒𝑥 )
𝑣2 = ln(1 + 𝑒 𝑥 )
The first equation together with 𝑣2 yields

𝑣1′ = − 𝑣2′ 𝑒 𝑥
𝑒 2𝑥
𝑣1′ = −
1 + 𝑒𝑥
𝑒 2𝑥
𝑣1 = − ∫ d𝑥
( 1 + 𝑒𝑥 )
𝑒𝑥
𝑣1 = − ∫ 𝑒𝑥 − d𝑥
( 1 + 𝑒𝑥 )
𝑣1 = − 𝑒 𝑥 + ln(1 + 𝑒 𝑥 ).

Therefore
𝑦𝑝 = 𝑣1 𝑒 𝑥 + 𝑣2 𝑒 2𝑥
= (𝑒 𝑥 + ln(1 + 𝑒 𝑥 ))𝑒 𝑥 + ln(1 + 𝑒 𝑥 )𝑒 2𝑥 .
■
Example 6.12 Find a particular solution of the equation

𝑥(𝑦 ′′ + 𝑦 ′ ) − 2(𝑦 ′ + 𝑦) = 𝑥 2

Solution The complementary equation 𝑥(𝑦 ′′ + 𝑦 ′ ) − 2(𝑦 ′ + 𝑦) = 0 can be solved by substitution.

Plugging 𝑢 = 𝑦 ′ + 𝑦 into the complementary equation and solving for 𝑢 yields

𝑥(𝑦 ′′ + 𝑦 ′ ) − 2(𝑦 ′ + 𝑦) =0
𝑥𝑢 ′ − 2𝑢 =0
𝑢′ 2
=
𝑢 𝑥
𝑢′ 2
∫ 𝑢 d𝑥 =∫ 𝑥 d𝑥
ln(𝑢) = ln(𝑥 2 ) + 𝑐
𝑢 =𝑐1 𝑥 2

Then
𝑦 ′ + 𝑦 = 𝑐1 𝑥 2 .

113 Lecture Notes for MA451

 6.2 Variation of Parameters

The integrating factor is 𝑟(𝑥) = 𝑒 𝑥 . Therefore, applying the method of integration by parts implies

1
𝑦ℎ = 𝑐1 (𝑥 2 𝑒 𝑥 )) d 𝑥
𝑒𝑥 ∫
1
= 𝑥 (𝑐1 𝑒 𝑥 (𝑥 2 − 2𝑥 + 2) + 𝑐2 )
𝑒
=𝑐1 (𝑥 2 − 2𝑥 + 2) + 𝑐2 𝑒 −𝑥

Let 𝑦1 = 𝑥 2 − 2𝑥 + 2 and 𝑦2 = 𝑒 −𝑥 . Both of them are solutions of the complementary equation.

Since 𝑦1′ = 2𝑥 − 2, 𝑦2′ = −𝑒 −𝑥 and the Wronskian is

𝑊 (𝑦1 , 𝑦2 ) =𝑦1 𝑦2′ − 𝑦1′ 𝑦2

= − (𝑥 2 − 2𝑥 + 2)𝑒 −𝑥 − (2𝑥 − 2)𝑒 −𝑥
=(−𝑥 2 − 4𝑥 + 2)𝑒 −𝑥 ≠ 0,

they are linearly independent.

A particular solution of non-homogeneous equations can be taken in the form

𝑦𝑝 = 𝑣1 (𝑥 2 − 2𝑥 + 2) + 𝑣2 𝑒 −𝑥 ,

where 𝑣1 and 𝑣2 are functions of 𝑥 that satisfy the following system of equations

𝑣1′ (𝑥 2 − 2𝑥 + 2) + 𝑣2′ 𝑒 −𝑥 =0
𝑣1′ (2𝑥 − 2) − 𝑣2′ 𝑒 −𝑥 =𝑥 2 .

The sum of the two equations yields

𝑥 2 𝑣1′ =𝑥 2
𝑣1′ =1
𝑣1 =𝑥
Then 𝑣2 satisfies
𝑣2′ 𝑒 −𝑥 = − 𝑣1′ (𝑥 2 − 2𝑥 + 2)
𝑣2′ = − (𝑥 2 − 2𝑥 + 2)𝑒 𝑥

𝑣2 = − ∫ ((𝑥 2 − 2𝑥 + 2)𝑒 𝑥 ) d 𝑥

𝑣2 = − ∫ ((𝑥 2 − 2𝑥 + 2)𝑒 𝑥 ) d 𝑥

𝑣2 = − 𝑒 −𝑥 (𝑥 2 − 4𝑥 + 6).

Therefore,

Therefore,
𝑦𝑝 = 𝑥(𝑥 2 − 2𝑥 + 2) − 𝑒 −𝑥 𝑒 𝑥 (𝑥 2 − 4𝑥 + 6) = 𝑥 3 − 3𝑥 2 + 6𝑥 − 6

114 Lecture Notes for MA451

 6.2 Variation of Parameters

is a particular solution. ■
 Exercise 6.5 Find a particular solution of

𝑥 2 𝑦 ′′ − 𝑥𝑦 ′ + 𝑦 = 𝑥 3 ,

using variation of parameter. It is known that 𝑦1 = 𝑥 and 𝑦2 = 𝑥 ln 𝑥 are linearly independent

solutions of the complementary equation.

Solution We first write the equation in standard form:

1 1
𝑦 ′′ − 𝑦 ′ + 2 𝑦 = 𝑥
𝑥 𝑥
Since 𝑦1 = 𝑥 and 𝑦2 = 𝑥 ln 𝑥 are linearly independent solutions of the complementary equation,
a particular solution of the equation can be written as

𝑦𝑝 = 𝑣1 𝑥 + 𝑣2 𝑥 ln 𝑥,

where 𝑣1 and 𝑣2 are functions of 𝑥.

We will find 𝑣1 and 𝑣2 under the following conditions:

⎧
⎪
⎪𝑣1′ 𝑥 + 𝑣2′ 𝑥 ln 𝑥 =0 (1)
⎨
⎪
⎪ 𝑣 ′ + 𝑣2′ (1 + ln 𝑥) = 𝑥 (2)
⎩ 1

The first derivatives 𝑣1′ and 𝑣2′ can be solved by the elimination method. Subtracting the
product of Equation (1) with (1 + ln 𝑥) from the product of Equation (2) with 𝑥 ln 𝑥 implies

𝑣1′ = −𝑥 ln 𝑥

Applying integration by parts yields

1 2 𝑥2
𝑣1 = − 𝑥 ln 𝑥 + .
2 4

Plugging 𝑣1′ into Equation (1) implies

𝑣2′ = 𝑥.

Then
1
𝑣2 = 𝑥 2 .
2

115 Lecture Notes for MA451

 6.2 Variation of Parameters

Therefore, a particular solution is

1 𝑥2 1 𝑥3
𝑦𝑝 = (− 𝑥 2 ln 𝑥 + )𝑥 + 𝑥 3 ln 𝑥 = .
2 4 2 4
■
 Exercise 6.6 Find the general solution to

𝑦 ′′ − 𝑦 = 𝑒 𝑥

using variation of parameter.

Solution The complementary homogenous equation

𝑦 ′′ − 𝑦 = 0,

has the general solution is

𝑦 = 𝑐1 𝑒 𝑥 + 𝑐2 𝑒 −𝑥 .

We look for a particular solution of the form

𝑦𝑝 = 𝑣1 (𝑥)𝑒 𝑥 + 𝑣2 (𝑥)𝑒 −𝑥 .

Consider the following system of equations of 𝑣1 and 𝑣2 :

⎧
⎪
⎪𝑣1′ 𝑒 𝑥 + 𝑣2′ 𝑒 −𝑥 =0
⎨
⎪
⎪ 𝑣 ′ 𝑒 𝑥 + (−1)𝑣2′ 𝑒 −𝑥 = 𝑒𝑥
⎩ 1

Solving for 𝑣1′ and 𝑣2′ yields

1 1
𝑣1′ = and 𝑣2′ = − 𝑒 2𝑥 .
2 2
Direct integrations implies that

1 1
𝑣1 = 𝑥 and 𝑣2 = − 𝑒 2𝑥 .
2 4
Therefore, a particular solution is

1 1 1 1
𝑦𝑝 = 𝑥𝑒 𝑥 + (− 𝑒 2𝑥 )𝑒 −𝑥 = 𝑥𝑒 𝑥 − 𝑒 𝑥 .
2 4 2 4
By Theorem 5.5, the general solution of the original equation is

1 1
𝑦 = 𝑥𝑒 𝑥 − 𝑒 𝑥 + 𝑐1 𝑒 𝑥 + 𝑐2 𝑒 −𝑥 .
2 4

116 Lecture Notes for MA451

 6.2 Variation of Parameters

6.2.2 Reduction of Order

You may wonder if there is another way to find 𝑣1 and 𝑣2 other than using the system of
equation (6.1). The answer is yes. We may take 𝑣2 = 0, then 𝑣1 will be solution of the following
equation
(𝑣1′ 𝑦1 )′ + 𝑝(𝑥)𝑣1′ 𝑦1 + 𝑣1′ 𝑦1′ = 𝑓 (𝑥).

Setting 𝑧 = 𝑣1′ . The equation can be reduce to a linear first order equation of 𝑧:

𝑧 ′ 𝑦1 + 𝑧(𝑝(𝑥)𝑦1 + 2𝑦1′ ) = 𝑓 (𝑥),

or equivalently (assuming 𝑦1 ≢ 0)

𝑓 (𝑥)
𝑧 ′ + 𝑧(𝑝(𝑥) + 2 ln(𝑦1 )) = .
𝑦1

By the method of integrating factor,

𝑓 (𝑥) ∫ (𝑝(𝑥)+2 ln(𝑦1 )) d 𝑥

𝑧 = 𝑒 − ∫ (𝑝(𝑥)+2 ln(𝑦1 )) d 𝑥 ∫ 𝑒 d𝑥
(( 𝑦1 ) )

Hence,
𝑓 (𝑥) ∫ (𝑝(𝑥)+2 ln(𝑦1 )) d 𝑥
𝑣1 = ∫ 𝑒 − ∫ (𝑝(𝑥)+2 ln(𝑦1 )) d 𝑥 ∫ 𝑒 d 𝑥 d 𝑥.
( (( 𝑦1 ) ) )

To summarize, we may use the variation of parameter with a particular solution of the
complementary equation of a liner second order equation to find a particular solution.

However, this method only works for second order equations.

Example 6.13 Find a particular solution of the equation

𝑦 ′′ + 𝑦 = tan 𝑥.

Solution The complementary equation 𝑦 ′′ + 𝑦 = 0 has a solution 𝑦 = cos 𝑥. Assume 𝑦 = 𝑢 cos 𝑥 is

a solution of 𝑦 ′′ + 𝑦 = 𝑡𝑎𝑛𝑥. Then, the function 𝑢 of 𝑥 satisfies the following equation

(𝑢 cos 𝑥)′′ + 𝑢 cos 𝑥 = tan 𝑥

(𝑢 ′ cos 𝑥 − 𝑢 sin 𝑥)′ + 𝑢 cos 𝑥 = tan 𝑥
(𝑢 ′′ cos 𝑥 − 𝑢 ′ sin 𝑥) − (𝑢 ′ sin 𝑥 + 𝑢 cos 𝑥) + 𝑢 cos 𝑥 = tan 𝑥
𝑢 ′′ cos 𝑥 − 2𝑢 ′ sin 𝑥 = tan 𝑥

117 Lecture Notes for MA451

 6.2 Variation of Parameters

Let 𝑧 = 𝑢 ′ , then 𝑧 satisfies

𝑧 ′ cos 𝑥 − 2𝑧 sin 𝑥 = tan 𝑥,

or
sin 𝑥 sin 𝑥
𝑧 ′ − 2𝑧 = .
cos 𝑥 cos2 𝑥

As a linear first order equation, it can be solved by the method of integrating factor. Since the
coefficient of 𝑧 is − 2cos
sin 𝑥
𝑥
, an integrating factor is
−2 sin 𝑥
d𝑥
𝑟(𝑥) = 𝑒 ∫ cos 𝑥 = cos2 𝑥.

Multiplying both sides of the equation of 𝑧 yields

sin 𝑥
(𝑧 cos2 𝑥)′ = cos2 𝑥 ⋅
cos2 𝑥
𝑧 cos2 𝑥 = ∫ sin 𝑥 d 𝑥

𝑧 cos2 𝑥 = − cos 𝑥
1
𝑧=−
cos 𝑥

Therefore,
𝑢 ′ =𝑧
1
𝑢 = ∫ (− d𝑥
cos 𝑥 )
1
𝑢 =−∫ d(sin 𝑥)
cos2 𝑥
1
𝑢 =−∫ d(sin 𝑥)
1 − sin2 𝑥
1 1 1
𝑢 =− ∫ ( + d(sin 𝑥)
2 1 − sin 𝑥 1 + sin 𝑥 )
1 1 + sin 𝑥
𝑢 = − ln
2 ( 1 − sin 𝑥 )
Thus, the original equation has a solution

1 1 + sin 𝑥
𝑦 = − cos 𝑥 ln .
2 ( 1 − sin 𝑥 )

■
 Exercise 6.7 Find a particular solution of the equation

𝑥 2 𝑦 ′′ − 𝑥𝑦 ′ + 𝑦 = 𝑥 3 .

118 Lecture Notes for MA451

 6.2 Variation of Parameters

Solution The complementary equation

𝑥 2 𝑦 ′′ − 𝑥𝑦 ′ + 𝑦 = 0

has a solution 𝑦 = 𝑥. This can be found by guessing, or by a decomposition similar to the

proof of Theorem 5.4.

Suppose a particular solution is of the form

𝑦 = 𝑥𝑢.

Then
𝑦 ′ = 𝑥𝑢 ′ + 𝑢

and
𝑦 ′′ = 𝑥𝑢 ′′ + 2𝑢 ′ .

Hence,
𝑥 2 𝑦 ′′ − 𝑥𝑦 ′ + 𝑦 = 𝑥 2 (𝑥𝑢 ′′ + 2𝑢 ′ ) − 𝑥(𝑥𝑢 ′ + 𝑢) + 𝑥𝑢 = 𝑥 3 𝑢 ′′ + 𝑥 2 𝑢 ′

Therefore, 𝑦 = 𝑥𝑢 is a solution if
𝑥 3 𝑢 ′′ + 𝑥 2 𝑢 ′ = 𝑥 3 ,

or
1
𝑢 ′′ + 𝑢 ′ = 1.
𝑥

To solve this second order equation for 𝑢, set 𝑣 = 𝑢 ′ . The equation become a first order
linear equation of 𝑣
1
𝑣 ′ + 𝑣 = 1.
𝑥
The integrating factor is
1
𝑟(𝑥) = 𝑒 ∫ 𝑥 d 𝑥 = 𝑒 ln 𝑥 = 𝑥.

Then
1 𝑥 1
𝑣= ∫ 𝑥 d 𝑥 = + 𝑐1
𝑥 2 𝑥
Therefore,
𝑥 1
𝑢 ′ = + 𝑐1
2 𝑥
𝑥2
𝑢 = + 𝑐1 ln 𝑥 + 𝑐2
4
and
𝑥3
𝑦 = 𝑥𝑢 = + 𝑐1 𝑥 ln 𝑥 + 𝑐2 𝑥.
4
■

119 Lecture Notes for MA451

 Week 7: Applications of Linear Second Order
Equations

10/25–10/28

7.1 Vibrating Springs

We consider the motion of an object with mass attached to an end of a spring with neg-
ligible mass that is vertically attached to a fixed object.

The spring–mass system is in equilibrium when the object is at rest and the forces acting
on it sum to zero. The position of the object in this case is the equilibrium position.

Denote by 𝑦 the displacement of the object from its equilibrium position at time 𝑡, mea-
sured positive upward.

𝑚
𝑦>0
Equilibrium
0
position 𝑚
𝑦<0

Figure 7.1: Spring-mass system

Hooke’s Law - Wikipedia says that if the length of the spring is changed by an amount
Δ𝐿 from its natural length, if it is then the spring exerts a force 𝐹𝑠 whose magnitude that is
proportional to Δ𝐿, that is |𝐹𝑠 | = 𝑘Δ𝐿, where 𝑘 is a positive number called the spring constant.
Since we take upward as the positive direction, the force is 𝐹𝑠 = 𝑘Δ𝐿 if the spring is stretched
or 𝐹𝑠 = −𝑘Δ𝐿 if the spring is compressed.,

Besides Earth’s gravitational force and the force of the spring, there can be other forces.
The system may have a damping force 𝐹𝑑 = −𝑐𝑦 ′ that resists the motion with a force propor-
tional to the velocity of the object. It may be due to resistance or friction. We say that the
 7.1 Vibrating Springs

Natural length 𝐿

𝐹𝑠 = 𝑘Δ𝐿
Change in length Δ𝐿

𝑚
Figure 7.2: Hooke’s Law of Spring

motion is undamped if 𝑐 = 0, or damped if 𝑐 > 0.

Figure 7.3: A spring system with damping

It may have an external force 𝐹 , other than the force due to gravity, that may vary with
𝑡, but is independent of displacement and velocity. We say that the motion is free if 𝐹 ≡ 0, or
forced if 𝐹 ≢ 0.

By Newton’s second law of motion, we have

𝑚𝑎 = −𝑚𝑔 + 𝐹𝑑 + 𝐹𝑠 + 𝐹 = −𝑚𝑔 − 𝑐𝑦 ′ + 𝐹𝑠 + 𝐹 ,

where 𝑎 = 𝑦 ′′ is the acceleration.

Let Δ𝑙 be the change of length when the system reaches equilibrium. Then 𝑚𝑔 = 𝑘Δ𝑙.
When the displacement of the object is 𝑦, the change in length of the spring is |𝑦 − Δ𝑙| and
𝐹𝑠 = 𝑘(Δ𝑙 − 𝑦).

Therefore,
−𝑚𝑔 + 𝐹𝑠 = −𝑘𝑦.

121 Lecture Notes for MA451

 7.1 Vibrating Springs

𝑦
𝐹𝑠𝑐 =−𝑘Δ𝐿

Natural Length 𝐿
=𝑘(Δ𝑙 − 𝑦)

𝑚
𝐹𝑠𝑒 = 𝑘Δ𝑙
Δ𝐿 +
𝐹𝑔 = −𝑚𝑔
𝑦 = Δ𝑙 + Δ𝐿

=
0
Δ𝑙
Equilibrium
0
position 𝑚

Figure 7.4: Relation between displacement and change in length

So the displacement 𝑦 of a object attached to a spring satisfies the following equation

𝑐 ′ 𝑘 𝐹 (𝑡)
𝑦 ′′ + 𝑦 + 𝑦= . (7.1)
𝑚 𝑚 𝑚

Note that for a horizontal spring motion, Equation (7.1) is still valid.

7.1.1 Simple harmonic motion

Assume there is no damping force or other external force, that is, 𝐹𝑑 ≡ 0 and 𝐹 ≡ 0. Then,
by Newton’s second law of motion, the displacement 𝑦 satisfies the following equation

𝑘
𝑦 ′′ + 𝑦 = 0.
𝑚
This motion is known as the simple harmonic motion.
√
Let 𝜔 = 𝑚𝑘 . Then the general solution of the equation 𝑚𝑦 ′ + 𝑘𝑦 = 0 is

𝑦 = 𝐶1 cos(𝜔𝑡) + 𝐶2 sin(𝜔𝑡).
√
If further, we let 𝑅 = (𝐶1 )2 + (𝐶2 )2 and 𝜙 the angle in (−𝜋, 𝜋) such that cos 𝜙 = 𝐶1
𝑅
and sin 𝜙 = − 𝐶𝑅2 ,
then the general solution can be written as

𝑦 = 𝑅 cos(𝜔𝑡 + 𝜙).

Here, the angle 𝜙 is known as the phase angle, 𝑅 is the amplitude of the oscillation and the
solution 𝑦 = 𝑅 cos(𝜔𝑡 + 𝜙) is called the amplitude-phase form of the displacement. The constant

122 Lecture Notes for MA451

 7.1 Vibrating Springs

𝜔 is the frequency of the motion. If the time 𝑡 is measured in seconds, then, the frequency is
measured in cycle per second or, in international unit, Hertz (Hz for short). Because the period
𝑇 of the amplitude-phase form displacement is

2𝜋
𝑇 = .
𝜔

Example 7.1 An object stretches a spring 5 m in equilibrium.

1. Find the displacement of the object after 𝑡 seconds if it is initially displaced 18 m above
equilibrium and given a downward velocity of 7 m/s.
2. Find the frequency, amplitude, the phase angle of this motion, and the amplitude-phase
form of the displacement.

Solution

1. The motion of the object is undamped and free, that is 𝐹𝑑 ≡ 0 and 𝐹 ≡ 0. So the displacement
𝑦 away from the equilibrium position satisfies the differential equation

𝑘
𝑦 ′′ + 𝑦 = 0,
𝑚
where 𝑚 is the mass and 𝑘 is the spring constant.
Since the spring stretches 5 m to reach the equilibrium position, the spring constant 𝑘 is de-
termined by the equation
𝑚𝑔 = 𝑘 ⋅ 6.

Since the units for the displacement is in meters, we take the gravitational acceleration to be
𝑔 = 9.8 m/s2 and then
𝑘 9.8 49𝑚
= = .
𝑚 5 25
Therefore, the displacement 𝑦 satisfies

49𝑚
𝑚𝑦 ′′ + 𝑦 =0
25
49
𝑦 ′′ + 𝑦 =0.
25
49
The associated characteristic equation 𝑟 2 + 25
= 0 has two complex solutions 𝑟 = ± 7i5 . There-
fore, the general solution is

7 7
𝑦 = 𝐶1 cos ( 𝑡 ) + 𝐶2 sin ( 𝑡 ) .
5 5
Because the object is initially displaced 18 m above equilibrium and given a downward velocity

123 Lecture Notes for MA451

 7.1 Vibrating Springs

of 7 m/s. In terms of mathematical equalities, we have

𝑦(0) = 18 and 𝑦 ′ (0) = −7.

Note that
7 7 7 7
𝑦 ′ = − 𝐶1 sin ( 𝑡 ) + 𝐶2 cos ( 𝑡 )
5 5 5 5
Then 𝐶1 = 18, 𝐶2 = −5, and the displacement 𝑦 at time 𝑡 is

7 7
𝑦 = 18 cos ( 𝑡 ) − 5 sin ( 𝑡 ) m.
5 5

2. The frequency is 𝜔 = 75 rad/s.

The amplitude is
√ √
𝑅= 182 + (−5)2 = 349rad/s.

Since
𝐶1 18 18
cos 𝜙 = =√ =√
𝑅 2
18 + (−5) 2 349
−𝐶2 −5 5
sin 𝜙 = = −√ =√
𝑅 182 + (−5)2 349
−𝐶2 5
tan 𝜙 = = ,
𝐶1 18
the angle 𝜙 is in the first quadrant, and

5
𝜙 = arctan ( rad.
18 )
The amplitude-phase form of the displacement is about
√ 7 5
𝑦= 349 cos ( 𝑡 + arctan ( )) m.
5 18
■
 Exercise 7.1 The natural length of a spring is 1 m. An object is attached to it and the length
of the spring increases to 1.2 m when the object is in equilibrium. Then the object is initially
displaced downward 0.4 m and given an upward velocity of 2.1 m/s. Find the amplitude-phase
form of the displacement for 𝑡 > 0.

Solution Since the gravitational acceleration is 𝑔 = 9.8 m/s2 and the change in length at the
equilibrium position is Δ𝑙 = 1.2 − 1 = 0.2 m, the ratio of the spring constant and the mass is

𝑘 𝑔 9.8
= = = 49,
𝑚 Δ𝑙 0.2

124 Lecture Notes for MA451

 7.2 Free Vibrations with Damping

and the frequency is √

𝑘 √
𝜔= = 49 = 7 Hz.
𝑚

Therefore, the displacement 𝑦 satisfies the initial value problem

𝑦 ′′ + 49𝑦 = 0, 𝑦(0) = −0.4, 𝑦 ′ (0) = 1.4.

The general solution of the differential equation is

𝑦 = 𝐶1 cos(𝜔𝑡) + 𝐶2 sin(𝜔𝑡) = 𝐶1 cos(7𝑡) + 𝐶2 sin(7𝑡).

So
𝑦 ′ = −7𝐶1 sin(7𝑡) + 7𝐶2 cos(7𝑡).

Substituting the initial conditions into 𝑦 and 𝑦 ′ yields 𝐶1 = −0.4 and 𝐶2 = 0.3. Hence, the
displacement is
𝑦 = −0.4 cos(7𝑡) + 0.3 sin(7𝑡)

The amplitude is √ √
𝑅= (𝐶1 )2 + (𝐶2 )2 = (−0.4)2 + 0.32 = 0.5

The phase angle is determined by

𝐶1 −0.4 4 𝐶2 0.3 3
cos 𝜙 = = =− and sin 𝜙 = − =− =− .
𝑅 0.5 5 𝑅 0.5 5
Therefore, 𝜙 is in the third quadrant and

4
𝜙 = − arccos (− ) rad.
5

So the amplitude-phase form of the displacement is

𝑦 = 0.5 cos(7𝑡 − 0.64).

7.2 Free Vibrations with Damping

Assume that the damping force is proportional to the velocity of the mass and acts in the
direction opposite to the motion, that is

𝐹𝑑 = −𝑐𝑦 ′ ,

125 Lecture Notes for MA451

 7.2 Free Vibrations with Damping

where 𝑐 is a positive constant, called the damping constant. Then the displacement satisfies
the equation
𝑐 𝑘
𝑦 ′′ + 𝑦 ′ + 𝑦 = 0.
𝑚 𝑚
The solution of this equation depends on the value of 𝑐, or more precisely, the value of
√
𝑐 2 − 4𝑘𝑚. The reason is that the associated characteristic equation

𝑐 𝑘
𝑟2 + 𝑟+ = 0,
𝑚 𝑚
√
−𝑐± 𝑐 2 −4𝑘𝑚
whose solutions 𝑟 = 2𝑚
may be two complex roots, a double root, or two distinct real
roots.

Underdamping
The motion is said to be underdamped, if 𝑐 2 < 4𝑚𝑘. In this case, the solutions are complex
roots √ √
−𝑐 ± 𝑐 2 − 4𝑚𝑘 −𝑐 ± i 4𝑚𝑘 − 𝑐 2
𝑟= = .
2𝑚 2𝑚
√
4𝑚𝑘−𝑐 2
Let 𝜔1 = 2𝑚
. Then the general solution is
𝑐𝑡
𝑦 = 𝑒 − 2𝑚 (𝐶1 cos(𝜔1 𝑡) + 𝐶2 sin(𝜔1 𝑡)).

Again, we may derive an amplitude-phase form of the displacement

𝑐𝑡
𝑦 = 𝑅𝑒 − 2𝑚 cos(𝜔1 𝑡 + 𝜙),

where
√ 𝐶1 𝐶2
𝑅= 𝐶12 + 𝐶22 , cos 𝜙 = , sin 𝜙 = − , −𝜙 < 𝜙 < 𝜋.
𝑅 𝑅
𝑐𝑡
In the amplitude-phase form of the displacement, the factor 𝑅𝑒 − 2𝑚 is called the time–varying
amplitude of the motion, the quantity 𝜔1 is called the frequency, and 𝑇 = 2𝜋 𝜔1
is called the
quasi–period of the displacement.

Example 7.2 An object with mass 2 kg is hanging at one end of a spring with the spring
constant 𝑘 = 50 kg/s2 . The spring is subject to a damping force with the damping constant
𝑐 = 12 kg/s. Suppose the object is initially displaced downward 4 m with a downward velocity
8 m/s. Find the displacement of the object.

Solution If the time 𝑡 is measured in seconds, then from the given information, the displacement
𝑦, measured in meters, is the solution of the following initial value problem

2𝑦 ′′ + 12𝑦 ′ + 50𝑦 = 0, 𝑦(0) = −4, 𝑦 ′ (0) = −8,

126 Lecture Notes for MA451

 7.2 Free Vibrations with Damping

𝑣(𝑡)

𝑐𝑡
𝑦 = 𝑅𝑒 − 𝑚
𝑡
𝑐𝑡
𝑦 = 𝑅𝑒 − 𝑚

Figure 7.5: Graph of the displacement of an undamped motion

or equivalently
𝑦 ′′ + 6𝑦 ′ + 25𝑦, 𝑦(0) = −4, 𝑦 ′ (0) = −8.

The characteristic equation

𝑟 2 + 6𝑟 + 25 = 0

has two complex solutions √

−6 ± 62 − 252
𝑟= = −3 ± 4i.
2⋅2
Therefore, the general solution is

𝑦 = 𝑒 −3𝑡 (𝐶1 cos(4𝑡) + 𝐶2 sin(4𝑡)).

Since
𝑦 ′ = 𝑒 −3𝑡 ((−3𝐶1 + 4𝐶2 ) cos(4𝑡) + (−4𝐶1 − 3𝐶2 ) sin(4𝑡)),

plugging the initial conditions into 𝑦 and 𝑦 ′ yields

𝐶1 = −4 and 𝐶2 = −5.

So the displacement 𝑦 is given by

𝑦 = −𝑒 −5𝑡 (4 cos(4𝑡) + 5 sin(4𝑡)).

127 Lecture Notes for MA451

 7.2 Free Vibrations with Damping

Overdamping
The motion is said to be overdamped if 𝑐 2 > 4𝑚𝑘. In this case, the solutions of the charac-
teristic equation are distinct real roots
√
−𝑐 ± 𝑐 2 − 4𝑚𝑘
𝑟= .
2𝑚
The general solution is √ √
(−𝑐+ )
𝑐 2 −4𝑚𝑘 𝑡 (−𝑐− )
𝑐 2 −4𝑚𝑘 𝑡
𝑦 = 𝐶1 𝑒 2𝑚 + 𝐶2 𝑒 2𝑚 .

Example 7.3 An object with mass 2 kg is hanging at one end of a spring with the spring
constant 𝑘 = 50 kg/s2 . The spring is subject to a damping force with the damping constant
𝑐 = 52 kg/s. Suppose the object is initially displaced 6 m above the equilibrium position with a
downward velocity 30 m/s. Find the displacement of the object.

Solution If the time 𝑡 is measured in seconds, then from the given information, the displacement
𝑦, measured in meters, is the solution of the following initial value problem

2𝑦 ′′ + 52𝑦 ′ + 50𝑦 = 0, 𝑦(0) = 6, 𝑦 ′ (0) = −30

or equivalently
𝑦 ′′ + 26𝑦 ′ + 25𝑦, 𝑦(0) = 6, 𝑦 ′ (0) = −30

The characteristic equation

𝑟 2 + 26𝑟 + 25 = 0

has two real solutions

𝑟1 = −1 and 𝑟2 = −25.

Therefore, the general solution is

𝑦 = 𝐶1 𝑒 −𝑡 + 𝐶2 𝑒 −25𝑡 .

Since
𝑦 ′ = −𝐶1 𝑒 −𝑡 − 25𝐶2 𝑒 −25𝑡 ,

plugging the initial conditions into 𝑦 and 𝑦 ′ yields

𝐶1 = 5 and 𝐶2 = 1.

So the displacement 𝑦 is given by

𝑦 = 5𝑒 −𝑡 + 𝑒 −25𝑡 .

128 Lecture Notes for MA451

 7.3 Forced vibrations

Critically Damping
√ 𝑐
The motion is said to be critically damped if 𝑐 = 4𝑚𝑘. In this case 𝑟1 = 𝑟2 = − 2𝑚 and the
general solution is
𝑐𝑡
𝑦 = 𝑒 − 2𝑚 (𝑐1 + 𝑐2 𝑡).

Example 7.4 An object with mass 2 kg is hanging at one end of a spring with the spring
constant 𝑘 = 50 kg/s2 . The spring is subject to a damping force with the damping constant
𝑐 = 20 kg/s. Suppose the object is initially displaced 1 m below the equilibrium position with a
upward velocity 3 m/s. Find the displacement of the object.

Solution If the time 𝑡 is measured in seconds, then from the given information, the displacement
𝑦, measured in meters, is the solution of the following initial value problem

2𝑦 ′′ + 20𝑦 ′ + 50𝑦 = 0, 𝑦(0) = −1, 𝑦 ′ (0) = 3

or equivalently
𝑦 ′′ + 10𝑦 ′ + 25𝑦, 𝑦(0) = −1, 𝑦 ′ (0) = 3

The characteristic equation

𝑟 2 + 10𝑟 + 25 = 0

has a repeated root

𝑟 = −5.

Therefore, the general solution is

𝑦 = 𝑒 −5𝑡 (𝐶1 + 𝐶2 𝑡).

Since
𝑦 ′ = 𝑒 −5𝑡 ((−5𝐶1 + 𝐶2 ) − 5𝐶2 𝑡),

plugging the initial conditions into 𝑦 and 𝑦 ′ yields

𝐶1 = −1 and 𝐶2 = −2.

So the displacement 𝑦 is given by

𝑦 = −𝑒 −5𝑡 − 2𝑡𝑒 −5𝑡 .

129 Lecture Notes for MA451

 7.3 Forced vibrations

7.3 Forced vibrations

Suppose that the motion of the spring is affected by an external force 𝐹 (𝑡) depending on
the time. Then the displacement 𝑦 satisfies the equation

𝑚𝑦 ′′ + 𝑐𝑦 ′ + 𝑘𝑦 = 𝐹 (𝑡),

which is a non-homogeneous equation.

Forced vibration without damping

In many mechanical problems, a device is subjected to periodic external forces 𝐹 (𝑡) but not
damping, for example, 𝐹 (𝑡) = 𝐹0 cos(𝜔0 𝑡). In this case, the displacement of the object satisfies
the equation
𝑘 𝐹0
𝑦 ′′ + 𝑦 = cos(𝜔0 𝑡).
𝑚 𝑚
√
If 𝜔0 ≠ 𝜔 = 𝑚𝑘 , solving this equation using the method of undetermined coefficients and using
the identity 𝑘 = 𝑚𝜔 2 yields the general solution

𝐹0 cos(𝜔0 𝑡)
𝑦 = 𝐶1 cos(𝜔𝑡) + 𝐶2 sin(𝜔𝑡) + .
𝑚(𝜔 2 − 𝜔02 )

If 𝜔0 = 𝜔, the general solution is

𝐹0 𝑡 sin(𝜔𝑡)
𝑦 =𝐶1 cos(𝜔𝑡) + 𝐶2 sin(𝜔𝑡) +
2𝑚𝜔
𝐹0 𝑡
=𝐶1 cos(𝜔𝑡) + 𝐶2 + sin(𝜔𝑡).
( 2𝑚𝜔 )

In this case, the amplitude increases as time goes. This phenomenon is known as the reso-
nance.

Example 7.5 A 2 kg object is attached to a spring with constant 𝑘 = 275 kg/s and subjected to
an external force 𝐹 (𝑡) = 32 cos 8𝑡 kg − m/s2 . The object begins at rest in its equilibrium position.
Find its displacement for 𝑡 > 0, with 𝑦(𝑡) measured positive upward.

Solution From the given information, the displacement satisfies

2𝑦 ′′ + 128𝑦 = 32 cos(8𝑡), 𝑦(0) = 0, 𝑦 ′ (0) = 0,

equivalently,
𝑦 ′′ + 64𝑦 = 16 cos(8𝑡), 𝑦(0) = 0, 𝑦 ′ (0) = 0.

Solving the equation by the method of undetermined coefficients yields the general solution to the

130 Lecture Notes for MA451

 7.3 Forced vibrations

homogeneous equation is

𝑦(𝑡) = 𝐶1 cos(8𝑡) + 𝐶2 sin(8𝑡) + 𝑡 sin(8𝑡).

Since 𝑦(0) = 0 and 𝑦 ′ (0) = 0, the undetermined coefficients are

𝐶1 = 0 𝐶2 = 0.

Therefore, the displacement is

𝑦 = 𝑡 sin(8𝑡).

Forced vibrations with damping

Suppose the damping constant 𝑐 is nonzero and the external force 𝐹 (𝑡) = 𝐹0 cos(𝜔0 𝑡). The
general solution of the equation

𝑚𝑦 ′′ + 𝑐𝑦 ′ + 𝑘𝑦 = 𝐹0 cos(𝜔0 𝑡)

is in the form
𝑦 = 𝑦ℎ + 𝑦𝑝 ,

where 𝑦ℎ is the general solution of the complementary equation and 𝑦𝑝 is a particular solution
in the form
𝑦𝑝 = 𝐴 cos(𝜔0 𝑡) + 𝐵 sin(𝜔0 𝑡),

where 𝐴 and 𝐵 can be determined by plugging 𝑦𝑝 in the equation.

Differentiating 𝑦𝑝 yields

𝑦𝑝′ = − 𝐴𝜔 sin(𝜔0 𝑡) + 𝐵𝜔 cos(𝜔0 𝑡).𝑦𝑝′ = −𝐴𝜔 2 cos(𝜔0 𝑡) − 𝐵𝜔 2 sin(𝜔0 𝑡).

Plugging them into 𝑚𝑦 ′′ + 𝑐𝑦 ′ + 𝑘𝑦 yields

𝑚𝑦 ′′ + 𝑐𝑦 ′ + 𝑘𝑦
=(−𝑚𝐴𝜔02 + 𝑐𝐵𝜔0 + 𝑘𝐴) cos(𝜔0 𝑡) + (−𝑚𝐵𝜔02 − 𝑐𝐴𝜔0 + 𝑘𝐵) sin(𝜔0 𝑡)

So 𝑦𝑝 is a particular solution if
{
(𝑘 − 𝑚𝜔02 )𝐴 + 𝑐𝐵𝜔0 =𝐹0
−𝑐𝐴𝜔0 + (𝑘 − 𝑚𝜔02 )𝐵 =0.

131 Lecture Notes for MA451

 7.3 Forced vibrations

Solving 𝐴 and 𝐵 yields

(𝑘 − 𝑚𝜔02 )𝐹0
𝐴=
(𝑘 − 𝑚𝜔02 )2 + 𝑐 2 𝜔02
(𝑐𝜔0 )𝐹0
𝐵= .
(𝑘 − 𝑚𝜔02 )2 + 𝑐 2 𝜔02

Therefore,
𝐹0 2
𝑦𝑝 = [(𝑘 − 𝑚𝜔0 ) cos 𝜔0 𝑡 + 𝑐𝜔0 sin 𝜔0 𝑡 ]
(𝑘 − 𝑚𝜔02 )2 + 𝑐 2 𝜔02
which can be written in the amplitude-phase form as

𝑦𝑝 = 𝑅 cos(𝜔0 𝑡 + 𝜙),

where
𝐹0
𝑅=√
(𝑘 − 𝑚𝜔02 )2 + 𝑐 2 𝜔02
and the phase angle 𝜙 is determined by

𝑘 − 𝑚𝜔02 𝑐𝜔0
cos 𝜙 = √ and sin 𝜙 = − √ .
(𝑘 − 𝑚𝜔02 )2 + 𝑐 2 𝜔02 (𝑘 − 𝑚𝜔02 )2 + 𝑐 2 𝜔02

When the motion is underdamped, that is, 𝑐 2 < 4𝑘𝑚, the general form of the displacement
is
𝑐𝑡
𝑦 = 𝑒 − 2𝑚 (𝐶1 cos(𝜔1 𝑡) + 𝐶2 sin(𝜔1 𝑡)) + 𝑅 cos(𝜔0 𝑡 + 𝜙),
4𝑘𝑚−𝑐 2
where 𝜔1 = 2𝑚
.

When the motion is overdamped, that is, 𝑐 2 > 4𝑘𝑚, the general form of the displacement
is √ √
(−𝑐+ )
𝑐 2 −4𝑚𝑘 𝑡 (−𝑐− )
𝑐 2 −4𝑚𝑘 𝑡
𝑦 = 𝐶1 𝑒 2𝑚 + 𝐶2 𝑒 2𝑚 ) + 𝑅 cos(𝜔0 𝑡 + 𝜙).

When the motion is critically damped, that is, 𝑐 2 = 4𝑘𝑚, the general form of the displace-
ment is
𝑐𝑡
𝑦 = 𝑒 − 2𝑚 (𝐶1 + 𝐶2 𝑡)) + 𝑅 cos(𝜔0 𝑡 + 𝜙).

Since the exponential summand approaches to 0 as 𝑡 goes to the infinity, for large 𝑡, the
displacement 𝑦 is closely approximated by the particular solution 𝑦𝑝 :

𝑦 ≈ 𝑦𝑝 = 𝑅 cos(𝜔0 𝑡 + 𝜙).

For this reason, we say that 𝑦ℎ is the transient component of the solution 𝑦, and 𝑦𝑝 is the
steady state component of 𝑦. Thus, for large 𝑡 the motion is like simple harmonic motion at the
frequency of the external force.

132 Lecture Notes for MA451

 7.3 Forced vibrations

An interesting question is to find the value of 𝜔0 so that the amplitude of the steady state
component is maximal. Let

𝜌(𝑥) = (𝑘 − 𝑚𝑥)2 + 𝑐 2 𝑥 = 𝑚2 𝑥 2 + (𝑐 2 − 2𝑘𝑚)𝑥 + 𝑘 2 , 𝑥 ≥ 0.

Then 𝑅 reaches its maximum 𝑅max when 𝜌(𝜔02 ) attains its minimum.
√
If 𝑐 > 2𝑘𝑚, then the minimum of 𝜌(𝜔02 ) is at 𝜔0 = 0 and

𝐹0
𝑅max = .
𝑘

√
If 𝑐 < 2𝑘𝑚, then the minimum of 𝜌(𝜔02 ) is at
√
2𝑘𝑚 − 𝑐 2
𝜔=
2𝑚2

and
2𝑚𝐹0
𝑅max = √ .
𝑐 4𝑘𝑚 − 𝑐 2

Example 7.6 A 1 kg object is attached to a spring with spring constant 𝑘 = 8 kg/s, and subjected
to an damping force with the constant 𝑐 = 4 kg/s and an external force 𝐹 (𝑡) = 3 cos(5𝑡) kg m/s2 .
Find the general solution and also the steady periodic solution

Solution The displacement 𝑦 satisfies the equation

𝑦 ′′ + 4𝑦 ′ + 8𝑦 = 3 cos(5𝑡).

Solving the characteristic equation

𝑟 2 + 4𝑟 + 8 =0
(𝑟 + 2)2 + 4 =0
yields
𝑟 = −2 ± 2𝑖.

Then the general solution of the complementary equation is

𝑦ℎ = 𝑒 −2𝑡 (𝐶1 cos 2𝑡 + 𝐶2 sin 2𝑡)

A particular solution can be taken in the form

𝑦𝑝 = 𝐴 cos 5𝑡 + 𝐵 sin 5𝑡,

where 𝐴 and 𝐵 can be determined by the method of undetermined coefficients. Differentiating 𝑦𝑝

133 Lecture Notes for MA451

 7.3 Forced vibrations

yields
𝑦𝑝′ =5𝐵 cos 5𝑡 − 5𝐴 sin 5𝑡
𝑦𝑝′′ = − 25𝐴 cos 5𝑡 + −25𝐵 sin 5𝑡.
Hence
𝑦𝑝′′ + 4𝑦𝑝′ + 8𝑦𝑝 = (−17𝐴 + 20𝐵) cos 5𝑡 + (−17𝐵 − 20𝐴) sin 5𝑡.

For 𝑦𝑝 to be a solution, the above expression must equal to 3 cos(5𝑡), which implies
{
−17𝐴 + 20𝐵 =3
20𝐴 + 17𝐵 =0

Solving this system of equations yields

−51 60
𝐴= and 𝐵= .
689 689

So the general solution is:

51 60
𝑦 = 𝑦ℎ + 𝑦𝑝 = 𝑒 −2𝑡 (𝐶1 cos 2𝑡 + 𝐶2 sin 2𝑡) − cos 5𝑡 + sin 5𝑡.
689 689

The steady periodic solution is

51 60
𝑦=− cos 5𝑡 + sin 5𝑡.
689 689

134 Lecture Notes for MA451

 Week 8: Series Solutions

11/01–11/04

Most functions seen in calculus belongs to a class known as the elementary functions.
The class of elementary functions consists of polynomials, rational functions, radical functions,
trigonometric functions, inverse trigonometric functions, exponential functions, logarithmic
functions and all others that can be constructed from those by adding, subtracting, multiplying,
dividing, or composing. In applications, many second order differential equations cannot be
solved in terms of elementary functions. For example, Airy’s Equation

𝑦 ′′ − 𝑥𝑦 = 0

does not have any solution that is an elementary function. How do we know that? This ques-
tion turns out to be complicated. A proof employs differential Galois theory. Interested reader
many find an argument in the paper An algorithm for solving second order linear homogeneous
differential equations by Kovacic.

One approach to solve equations that have no elementary solutions is to use power series.
We can then define special functions using power series and study their properties.

8.1 Review of Power Series

Convergence
A power series in (𝑥 − 𝑥0 ) is an infinite sum of the form
∞
∑ 𝑎𝑛 (𝑥 − 𝑥0 )𝑛 = 𝑎0 + 𝑎1 (𝑥 − 𝑥0 ) + 𝑎2 (𝑥 − 𝑥0 )2 + ⋯ .
𝑛=0

A power series in (𝑥 − 𝑥0 ) is said to converge at 𝑥 if the limit

𝑚
lim ∑(𝑥 − 𝑥0 )
𝑚→∞
𝑛=0

exists.
∞
Clearly, the power series ∑ 𝑎𝑛 (𝑥 − 𝑥0 )𝑛 converges at 0. If it also converges at another point,
𝑛=0
then either it converges for all 𝑥, or converges over an interval (𝑥0 − 𝑅, 𝑥0 + 𝑅) for some positive
number 𝑅 and diverges over the interval (−∞, 𝑥0 − 𝑅) ∪ (𝑥0 + 𝑅, ∞).
 8.1 Review of Power Series

Theorem 8.1
For any power series
∞
∑ 𝑎𝑛 (𝑥 − 𝑥0 )𝑛 ,
𝑛=0

exactly one of the following statements holds true:

1. The power series converges only for 𝑥 = 𝑥0 .
2. The power series converges for all values of 𝑥.
3. There’s a positive number 𝑅 such that the power series converges if |𝑥 − 𝑥0 | < 𝑅 and
diverges if |𝑥 − 𝑥0 | > 𝑅.
♥

The number 𝑅 in the third, case 3., is called the radius of convergence of the power series.
For convenience, we set 𝑅 = 0 in the first case and 𝑅 = ∞ in the second case. The interval
(𝑥0 − 𝑅, 𝑥0 + 𝑅) is called the interval of convergence.

In calculus, several methods of finding the radius of convergence are given. One of them
is the following theorem.
Theorem 8.2
Suppose there’s an integer 𝑁 such that 𝑎𝑛 ≠ 0 if 𝑛 ≥ 𝑁 and

| 𝑎𝑛+1 |
lim || | = 𝐿,
𝑛→∞ | 𝑎𝑛 ||

where 0 ≤ 𝐿 ≤ ∞. Then the radius of convergence of ∑∞ 𝑛 1

𝑛=0 𝑎𝑛 (𝑥 − 𝑥0 ) is 𝑅 = 𝐿 , where the the
convention that 𝑅 = 0 if 𝐿 = ∞, or 𝑅 = ∞ if 𝐿 = 0 is used.
♥

Example 8.1 Find the radius of convergence of the series

∞
𝑥𝑛
∑ (−1)𝑛 .
𝑛=10
𝑛!

| |
Solution When 𝑛 goes to infinity, the limit lim | 𝑎𝑎𝑛+1 | is
𝑛→∞ | 𝑛 |

| (−1)𝑛+1 |
| 𝑎𝑛+1 | | |
lim || | = lim | (𝑛+1)!𝑛 |
𝑛→∞ | 𝑎𝑛 | 𝑛→∞ | (−1) ||
| |
| 𝑛! |
1
= lim
𝑛→∞ 𝑛 + 1

=0.

Therefore, the radius of convergence is 𝑅 = ∞. ■

136 Lecture Notes for MA451

 8.1 Review of Power Series

 Exercise 8.1 Find the radius of convergence of the series

∞
𝑥𝑛
∑(−1)𝑛+1 .
𝑛=5
2𝑛

| |
Solution When 𝑛 goes to infinity, the limit lim | 𝑎𝑎𝑛+1 | is
𝑛→∞ | 𝑛 |

| 𝑎𝑛+1 | | (−1)𝑛+1𝑛+2 |
| | | 2 |
lim | | = lim | (−1) 𝑛+1 |
𝑛→∞ | 𝑎𝑛 | 𝑛→∞ | |
| 2𝑛 |
1
= lim
𝑛→∞ 2
1
= .
2

Therefore, the radius of convergence is 𝑅 = 1

1
2
= 2. ■

Differentiations of power series

For a continuous function 𝑓 that has derivatives of all orders for |𝑥| < 𝑅, the Taylor expan-
sion ∞
𝑓 (𝑛) (𝑥) 𝑛
𝑓 (𝑥) = ∑ 𝑥
𝑛=0
𝑛!
is a power series known as the Maclaurin series of 𝑓 . For example,
∞
𝑥𝑛
𝑒𝑥 = ∑ , −∞ < 𝑥 < ∞;
𝑛=0
𝑛!

∞
𝑥 2𝑛+1
sin 𝑥 = ∑(−1)𝑛 , −∞ < 𝑥 < ∞;
𝑛=0
(2𝑛 + 1)!
∞
𝑥 2𝑛
cos 𝑥 = ∑(−1)𝑛 − ∞ < 𝑥 < ∞;
𝑛=0
(2𝑛)!
∞
1
= ∑ 𝑥𝑛 − 1 < 𝑥 < 1.
1 − 𝑥 𝑛=0

∞
Backwards, suppose the power series ∑ 𝑎𝑛 (𝑥 − 𝑥0 )𝑛 has a positive radius of convergence
𝑛=0
𝑅. Then we can define a function
∞
𝑓 (𝑥) = ∑ 𝑎𝑛 (𝑥 − 𝑥0 )𝑛
𝑛=0

on its open interval of convergence (𝑥0 − 𝑅, 𝑥0 + 𝑅). Such a function is called an analytic function
and has very good properties.

137 Lecture Notes for MA451

 8.1 Review of Power Series

Definition 8.1
A function 𝑓 is analytic at 𝑥0 if
∞
𝑓 (𝑥) = ∑ 𝑎𝑛 (𝑥 − 𝑥0 )𝑛
𝑛=0

and the series is convergent to 𝑓 (𝑥) for all 𝑥 in an open interval containing 𝑥0 .
A function is analytic on an open interval if it is analytic at all points in the interval.
♣

The set of analytic functions is closed under addition and multiplication.

Theorem 8.3
Suppose the power series ∑∞ 𝑛
𝑛=0 𝑎𝑛 (𝑥 − 𝑥0 ) has a positive radius 𝑅 and a function 𝑓 is defined
by
∞
𝑓 (𝑥) = ∑ 𝑎𝑛 (𝑥 − 𝑥0 )𝑛 .
𝑛=0

Then
1. 𝑓 has derivatives of all orders in the open interval (𝑥0 − 𝑅, 𝑥0 + 𝑅),
2. successive derivatives of 𝑓 can be obtained by repeatedly differentiating the power se-
ries term by term, that is
∞
𝑓 ′ (𝑥) =∑ 𝑛𝑎𝑛 (𝑥 − 𝑥0 )𝑛−1 ,
𝑛=1
∞
𝑓 ′′ (𝑥) =∑ 𝑛(𝑛 − 1)𝑎𝑛 (𝑥 − 𝑥0 )𝑛−2 ,
𝑛=2

⋮
∞
𝑓 (𝑘) (𝑥) =∑ 𝑛(𝑛 − 1) ⋯ (𝑛 − 𝑘 + 1)𝑎𝑛 (𝑥 − 𝑥0 )𝑛−𝑘 ,
𝑛=𝑘

3. all of these series have the same radius of convergence 𝑅.

♥

As a corollary, we see that if 𝑓 is a function defined by a power series, then the Taylor
series of the function is exactly the power series.
Corollary 8.1
Let 𝑓 be a function defined by
∞
𝑓 (𝑥) = ∑ 𝑎𝑛 (𝑥 − 𝑥0 )𝑛 ,
𝑛=0

where the power series has a positive radius of convergence 𝑅. Then

𝑓 (𝑛) (𝑥0 )
𝑎𝑛 = ,
𝑛!
that is, the power series ∑∞ 𝑛
𝑛=0 𝑎𝑛 (𝑥 − 𝑥0 ) is the Taylor series of 𝑓 about 𝑥0 .
♥

138 Lecture Notes for MA451

 8.1 Review of Power Series
∞ 2𝑛+1
𝑥
Example 8.2 Consider the power series ∑ (−1)𝑛 (2𝑛+1)! which converges for all real number 𝑥.
𝑛=0
Show that the derivative of the power series defines a function which is cos 𝑥.

Solution The derivative of a power series is the infinite sum of derivatives of terms. So

∞ 2𝑛+1 ∞
d 𝑛 𝑥 d 𝑥 2𝑛+1
∑(−1) = ∑(−1)𝑛
d 𝑥 ( 𝑛=0 (2𝑛 + 1)! ) 𝑛=0 d 𝑥 ( (2𝑛 + 1)! )
∞
𝑥 2𝑛
= ∑(−1)𝑛
𝑛=0
(2𝑛)!

Comparing with the Maclaurin series of cos 𝑥, we see that the derivative is cos 𝑥. Indeed, the func-
tion defined by the power series is noting but sin 𝑥. ■
∞
𝑥𝑛
 Exercise 8.2 Consider the power series ∑ 𝑛!
which converges for all real number 𝑥. Show
𝑛=0
that the second derivative of the power series defines a function which is 𝑒 𝑥 .

Solution The first derivative is

∞ ∞
d 𝑥𝑛 d 𝑥𝑛
∑ =∑
d 𝑥 ( 𝑛=0 𝑛! ) 𝑛=0 d 𝑥 ( 𝑛! )
∞
𝑥 𝑛−1
=∑
𝑛=1
(𝑛 − 1)!
∞
𝑥𝑛
=∑
𝑛=0
𝑛!

which is the power series itself. So is the second derivative. Comparing with the Maclaurin
series of 𝑒 𝑥 , we see that the power series, its first derivative and the second derivative are
the same function 𝑒 𝑥 . ■

Shifting indices
You probably know that the summation index is called a dummy index, and can changed
to any other name. For a power series, whose 𝑛-th term is not a multiple of (𝑥 − 𝑥0 )𝑛 , one can
use a substitution and rename the index to shift exponent to 𝑛. For example
∞
𝑓 ′ (𝑥) = ∑ 𝑛𝑎𝑛 (𝑥 − 𝑥0 )𝑛−1
𝑛=1
∞
= ∑(𝑘 + 1)𝑎𝑘+1 (𝑥 − 𝑥0 )𝑘 , substituting 𝑘 = 𝑛 − 1
𝑘=0
∞
= ∑(𝑛 + 1)𝑎𝑛+1 (𝑥 − 𝑥0 )𝑛 , renaming
𝑛=0

139 Lecture Notes for MA451

 8.1 Review of Power Series

In general, for any integer 𝑘, the power series

∞
∑ 𝑎𝑛 (𝑥 − 𝑥0 )𝑛−𝑘
𝑛=𝑛0

can be rewritten as ∞
∑ 𝑎𝑛+𝑘 (𝑥 − 𝑥0 )𝑛 ,
𝑛=𝑛0 −𝑘

that is, replacing 𝑛 by 𝑛 + 𝑘 in the general term and 𝑛0 by 𝑛0 − 𝑘 in the lower limit of summation
leaves the series unchanged.

Example 8.3 Given that

∞
𝑦(𝑥) = ∑ 𝑎𝑛 𝑥 𝑛 ,
𝑛=0

write the function 𝑥𝑦 as a power series in which the general term is a constant multiple of 𝑥 𝑛 .

Solution Since ∞
𝑥𝑦(𝑥) = ∑ 𝑎𝑛 𝑥 𝑛+1
𝑛=0

replacing 𝑛 by 𝑛 − 1 and 0 by 0 − (−1) = 1 yields

∞
𝑥𝑦(𝑥) = ∑ 𝑎𝑛−1 𝑥 𝑛 .
𝑛=1

■
 Exercise 8.3 Given that ∞
𝑦(𝑥) = ∑ 𝑎𝑛 𝑥 𝑛 ,
𝑛=0

write the function 𝑦 ′′ as a power series in which the general term is a constant multiple of
𝑥 𝑛.

Solution Since ∞
𝑦 ′′ (𝑥) = ∑ 𝑛(𝑛 − 1)𝑎𝑛 𝑥 𝑛−2 ,
𝑛=2

replacing 𝑛 by 𝑛 + 2 and 2 by 2 − 2 = 0 yields

∞
𝑦 (𝑥) = ∑(𝑛 + 2)(𝑛 + 1)𝑎𝑛+2 𝑥 𝑛 .
′′

𝑛=0

140 Lecture Notes for MA451

 8.1 Review of Power Series

Linear combination of power series

From calculus, particular, the section on Riemann sums, we know that the linear combina-
∞
tion of infinite sums is the infinite sums of linear combinations. More precisely, if 𝑓 (𝑥) = ∑ 𝑎𝑛 𝑥 𝑛
𝑛=0
∞
𝑛
with the convergence radius 𝑅1 and 𝑔(𝑥) = ∑ 𝑏𝑛 𝑥 with the convergence radius 𝑅2 , then
𝑛=0

∞
𝐴𝑓 (𝑥) ± 𝐵𝑔(𝑥) = ∑(𝐴𝑎𝑛 + 𝐵𝑏𝑛 )𝑥 𝑛
𝑛=0

which has the convergence radius 𝑅 = min{𝑅1 , 𝑅2 }.

Note that power series can also be multiplied like polynomials:

∞ 𝑛
𝑓 (𝑥)𝑔(𝑥) = ∑ ∑ 𝑎𝑘 𝑏𝑛−𝑘 𝑥 𝑛 .
𝑛=0 ( 𝑘=0 )

Let’s end this section by finding a power series solution for the Airy equation 𝑦 ′′ − 𝑥𝑦 = 0.

Example 8.4 Find the first 3 terms of the power series solution for the Airy equation 𝑦 ′′ −𝑥𝑦 = 0
with initial conditions 𝑦(0) = 0 and 𝑦 ′ (0) = 1.
∞
Solution Suppose that 𝑦(𝑥) = ∑ 𝑎𝑛 𝑥 𝑛 is a solution. Using the linear combination property, and the
𝑛=0
technique of shifting indices, we see that
∞ ∞
′′ 𝑛
𝑦 − 𝑥𝑦 = ∑(𝑛 + 2)(𝑛 + 1)𝑎𝑛+2 𝑥 − 𝑥 ∑ 𝑎𝑛 𝑥 𝑛
𝑛=0 𝑛=0
∞ ∞
= ∑(𝑛 + 2)(𝑛 + 1)𝑎𝑛+2 𝑥 𝑛 − ∑ 𝑎𝑛 𝑥 𝑛+1
𝑛=0 𝑛=0
∞ ∞
= ∑(𝑛 + 2)(𝑛 + 1)𝑎𝑛+2 𝑥 𝑛 − ∑ 𝑎𝑛−1 𝑥 𝑛
𝑛=0 𝑛=1

=(0 + 2)(0 + 1)𝑎0+2 𝑥 + ∑((𝑛 + 2)(𝑛 + 1)𝑎𝑛+2 − 𝑎𝑛−1 )𝑥 𝑛

0

𝑛=1

=2𝑎2 + ∑((𝑛 + 2)(𝑛 + 1)𝑎𝑛+2 − 𝑎𝑛−1 )𝑥 𝑛 .

𝑛=1

If 𝑦(𝑥) is solution of the Airy equation, then

2𝑎2 =0
(𝑛 + 2)(𝑛 + 1)𝑎𝑛+2 − 𝑎𝑛−1 =0 for 𝑛 = 1, 2, … .

Since 𝑦(0) = 0 and 𝑦 ′ (0) = 1, we see that 𝑎0 = 0 and 𝑎1 = 1. Note that 𝑎2 = 0 too. Then 𝑎𝑛 can be

141 Lecture Notes for MA451

 8.1 Review of Power Series

solve recursively using the second formula. For example,

𝑎1−1 𝑎0
𝑎3 = 𝑎1+2 = = = 0,
(1 + 2)(1 + 1) 3 ⋅ 2

𝑎2−1 𝑎1 1
𝑎4 = 𝑎2+2 = = = ,
(2 + 2)(2 + 1) 4 ⋅ 3 12
𝑎3−1 𝑎2
𝑎5 = 𝑎3+2 = = = 0.
(3 + 2)(3 + 1) 5 ⋅ 4
𝑎4−1 𝑎3
𝑎6 = 𝑎4+2 = = = 0.
(4 + 2)(4 + 1) 6 ⋅ 5
𝑎5−1 𝑎4 1 1
𝑎7 = 𝑎5+2 = = = = .
(5 + 2)(5 + 1) 7 ⋅ 6 7 ⋅ 6 ⋅ 4 ⋯ 3 45360

Indeed, the power series solution of the Airy equation has the form

1 4 1 7 1 1
𝑦(𝑥) = 𝑥 + 𝑥 + 𝑥 + 𝑥 10 + ⋯ + 𝑘
𝑥 3𝑘+1 + ⋯ .
12 504 45360
∏ 3𝑗 ⋅ (3𝑗 + 1)
𝑗=1

■
Remark When the recurrence relation is in the form

𝑎𝑛+2 = 𝑐1 𝑎𝑛 + 𝑐2 𝑎𝑛−1 ,

where 𝑐1 and 𝑐2 are constant, the solution, known as the generating function of the sequence
𝑎𝑛 , may be expressed as a rational function. See the wiki page Recurrence relation for more
information.

 Exercise 8.4 Find the first 4 terms of the power series solution for the initial value problem

𝑦 ′′ − 𝑥𝑦 ′ = 0, 𝑦(0) = 1, 𝑦 ′ (0) = 1.

∞
Solution Suppose that 𝑦(𝑥) = ∑ 𝑎𝑛 𝑥 𝑛 is a solution. Then
𝑛=0

∞ ∞
𝑦 ′′ − 𝑥𝑦 ′ = ∑(𝑛 + 2)(𝑛 + 1)𝑎𝑛+2 𝑥 𝑛 − 𝑥 ∑ 𝑛𝑎𝑛 𝑥 𝑛−1
𝑛=0 𝑛=1
∞ ∞
= ∑(𝑛 + 2)(𝑛 + 1)𝑎𝑛+2 𝑥 𝑛 − ∑ 𝑛𝑎𝑛 𝑥 𝑛
𝑛=0 𝑛=1
∞
=2𝑎2 + ∑((𝑛 + 2)(𝑛 + 1)𝑎𝑛+2 − 𝑛𝑎𝑛 )𝑥 𝑛
𝑛=1

142 Lecture Notes for MA451

 8.2 Series Solutions Near an Ordinary Point

If 𝑦(𝑥) is solution of the Airy equation, then

2𝑎2 =0
(𝑛 + 2)(𝑛 + 1)𝑎𝑛+2 − 𝑛𝑎𝑛 =0 for 𝑛 = 1, 2, … .

Since 𝑦(0) = 1 and 𝑦 ′ (0) = 1, we see that 𝑎0 = 1 and 𝑎1 = 1. Note that 𝑎2 = 0 too. Then

𝑎1 1
𝑎3 = = ,
6 6
2𝑎2
𝑎4 = = 0.
12
3𝑎3 1
𝑎5 = = .
20 40
So the power series solution has the form

𝑥3 𝑥5
𝑦(𝑥) = 1 + 𝑥 + + + ⋯.
6 40
■

8.2 Series Solutions Near an Ordinary Point

In this section, we will study the homogeneous linear second order equation of the form

𝑃0 𝑦 ′′ + 𝑃1 𝑦 ′ + 𝑃2 𝑦 = 0, (8.1)

where the coefficient functions 𝑃0 (𝑥), 𝑃1 (𝑥) and 𝑃2 (𝑥) are polynomials with no common factor
and 𝑃0 is not identically zero.

A point 𝑥0 is called an ordinary point of Equation 8.1 if 𝑃0 (𝑥0 ) ≠ 0, otherwise, it is called a

singular point.

Example 8.5 Find ordinary points of Legendre’s equation

(1 − 𝑥 2 )𝑦 ′′ − 2𝑥𝑦 ′ + 𝛼(𝛼 + 1)𝑦 = 0.

Solution Solving the equation 𝑃0 (𝑥) = 0 will gives singular points. In this case, the equation

1 − 𝑥2 = 0

has two solutions 𝑥 = 1 and 𝑥 = −1 which are singular points of Legendre’s equation. All other
points are ordinary point of the equation. ■
𝑃1
Since polynomials are analytic functions, it can be shown that the rational function 𝑃0
and

143 Lecture Notes for MA451

 8.2 Series Solutions Near an Ordinary Point
𝑃2
𝑃0
are analytic at any ordinary point. Near an ordinary point, we can rewrite Equation 8.1 as

𝑃1 ′ 𝑃2
𝑦 ′′ + 𝑦 + 𝑦=0
𝑃0 𝑃0

which is called the normalized equation. Since 𝑃𝑃01 and 𝑃𝑃20 are analytic, using properties of power
series, we can find a power series solution 𝑦(𝑥) in 𝑥 − 𝑥0 which is valid near 𝑥0 .
Theorem 8.4
Let 𝑥0 be an ordinary point of the equation

𝑃0 𝑦 ′′ + 𝑃1 𝑦 ′ + 𝑃2 𝑦 = 0

and let 𝑎0 and 𝑎1 be arbitrary constants. Then there exists a unique solution 𝑦(𝑥) = ∑∞ 𝑛=0 𝑎𝑛 (𝑥 −
𝑥0 )𝑛 near 𝑥0 such that 𝑦(𝑥0 ) = 𝑎0 and 𝑦 ′ (𝑥0 ) = 𝑎1 . Moreover, if the power series expansions of
𝑃1 (𝑥)
𝑃0 (𝑥)
and 𝑃𝑃02 (𝑥)
(𝑥)
converge on an open interval (𝑥0 − 𝑅, 𝑥0 + 𝑅), then the power series solution also
converges on the same interval.
♥

The first part of this theorem can be proved by solving recursive formula of 𝑎𝑛 . The proof
of the same convergence interval is a little bit involved. As we mainly focus on solving the
equation, we will not discuss the proof of the theorem. We refer the reader to (Simmons
2016, Section 28) for a proof in more general setting.

This theorem, together with the existence and uniqueness of the solution of a linear sec-
ond order equation, implies that every solution of Equation 8.1 can be represented by a power
series. We can such a solution a power series solution.

The basic idea to find a power series solution is similar to the undetermined coefficient
method. To simplify notations in calculation, we define a differential operator

d2 d
𝐿 = 𝑃0 2
+ 𝑃1 + 𝑃2
d𝑥 d𝑥
which acts on 𝑦 by

d2 d
𝐿𝑦 = 𝑃0 2
𝑦 + 𝑃1 𝑦 + 𝑃2 𝑦 = 𝑃0 𝑦 ′′ + 𝑃1 𝑦 ′ + 𝑃2𝑦.
d𝑥 d𝑥
∞
Suppose 𝑦(𝑥) = ∑ 𝑎𝑛 (𝑥 − 𝑥0 )𝑛 is a power series solution satisfies the initial conditions 𝑦(0) = 𝑎0
𝑛=0
and 𝑦 ′ (0) = 𝑎1 . Then
∞
𝐿𝑦 = ∑ 𝑏𝑛 (𝑥 − 𝑥0 )𝑛 ,
𝑛=0

where 𝑏𝑛 are expressions in terms of coefficients of 𝑃0 , 𝑃1 , and 𝑃2 , and 𝑎0 , 𝑎1 , . . . , 𝑎𝑛+𝑁 for some
positive integer 𝑁 . Then 𝑦 is a solution if and only if 𝑏𝑛 = 0 for all 𝑛 ≥ 0. The coefficients 𝑎2 , 𝑎3 ,
. . . , can be determined recursively using relations 𝑏𝑛 = 0.

144 Lecture Notes for MA451

 8.2 Series Solutions Near an Ordinary Point

You will find in calculations of power series, the product of a sequence frequent appear.
To simplify calculation, we denote the product of a sequence 𝑎𝑚 , 𝑎𝑚+1 , 𝑎𝑚+2 , … , 𝑎𝑛 , where 𝑛 > 𝑚,
by
𝑛
∏ 𝑎𝑘 = 𝑎𝑚 ⋯ 𝑎𝑚+1 ⋯ ⋯ ⋅ 𝑎𝑛 .
𝑘=𝑚

For convenience, we define

𝑛
∏ 𝑎𝑘 = 1 if 𝑛 < 𝑚.
𝑘=𝑚

Example 8.6 Find the power series in 𝑥 for the general solution of

(1 + 2𝑥 2 )𝑦 ′′ + 6𝑥𝑦 ′ + 2𝑦 = 0.

Solution

Let
𝐿𝑦 = (1 + 2𝑥 2 )𝑦 ′′ + 6𝑥𝑦 ′ + 2𝑦 = 𝑦 ′′ + 2𝑥 2 𝑦 ′′ + 6𝑥𝑦 ′ + 2𝑦.

If
∞
𝑦 = ∑ 𝑎𝑛 𝑥 𝑛 ,
𝑛=0

then ∞ ∞
𝑦 ′ = ∑ 𝑛𝑎𝑛 𝑥 𝑛−1 and 𝑦 ′′ = ∑ 𝑛(𝑛 − 1)𝑎𝑛 𝑥 𝑛−2 .
𝑛=1 𝑛=2

So ∞ ∞ ∞ ∞
𝑛−2 2 𝑛−2 𝑛−1
𝐿𝑦 = ∑ 𝑛(𝑛 − 1)𝑎𝑛 𝑥 + 2𝑥 ∑ 𝑛(𝑛 − 1)𝑎𝑛 𝑥 + 6𝑥 ∑ 𝑛𝑎𝑛 𝑥 + 2 ∑ 𝑎𝑛 𝑥 𝑛
𝑛=2 𝑛=2 𝑛=1 𝑛=0
∞ ∞ ∞ ∞
= ∑ 𝑛(𝑛 − 1)𝑎𝑛 𝑥 𝑛−2 + 2 ∑ 𝑛(𝑛 − 1)𝑎𝑛 𝑥 𝑛 + 6 ∑ 𝑛𝑎𝑛 𝑥 𝑛 + 2 ∑ 𝑎𝑛 𝑥 𝑛
𝑛=2 𝑛=2 𝑛=1 𝑛=0
∞ ∞ ∞ ∞
= ∑(𝑛 + 2)(𝑛 + 1)𝑎𝑛+2 𝑥 + 2 ∑ 𝑛(𝑛 − 1)𝑎𝑛 𝑥 + 6 ∑ 𝑛𝑎𝑛 𝑥 + 2 ∑ 𝑎𝑛 𝑥 𝑛
𝑛 𝑛 𝑛

𝑛=0 𝑛=0 𝑛=0 𝑛=0

∞
= ∑ [(𝑛 + 2)(𝑛 + 1)𝑎𝑛+2 + (2𝑛(𝑛 − 1) + 6𝑛 + 2)𝑎𝑛 ] 𝑥 𝑛
𝑛=0
∞
= ∑ [(𝑛 + 2)(𝑛 + 1)𝑎𝑛+2 + 2(𝑛 + 1)2 𝑎𝑛 ] 𝑥 𝑛
𝑛=0

If 𝑦 is a solution, then 𝐿𝑦 = 0 which implies that coefficients of the power series expression of
𝑦 satisfy the recurrence relation

𝑛+1
𝑎𝑛+2 = −2 𝑎𝑛 , 𝑛 ≥ 0.
𝑛+2
Since the indices on the left and right differ by two, we write the recurrence relation separately for

145 Lecture Notes for MA451

 8.2 Series Solutions Near an Ordinary Point

𝑛 = 2𝑚 and 𝑛 = 2𝑚 + 1. Then

2𝑚 + 1 2𝑚 + 1
𝑎2𝑚+2 = − 2 𝑎2𝑚 = − 𝑎2𝑚 , 𝑚 ≥ 0,
2𝑚 + 2 𝑚+1

and
2𝑚 + 2 𝑚+1
𝑎2𝑚+3 = − 2 𝑎2𝑚+1 = −4 𝑎2𝑚+1 , 𝑚 ≥ 0.
2𝑚 + 3 2𝑚 + 3
Computing the coefficients of even powers of 𝑥 from the recurrence relation yields

1
𝑎2 = − 𝑎0 ,
1
3 3 1 1⋅3
𝑎4 = − 𝑎2 = (− ) (− ) 𝑎0 = 𝑎0 ,
2 2 1 1⋅2
5 5 1⋅3 1⋅3⋅5
𝑎6 = − 𝑎4 = − ( 𝑎 = − 𝑎0 ,
3 1 ⋅ 2)
0
3 1⋅2⋅3
7 7 1⋅3⋅5 1⋅3⋅5⋅7
𝑎8 = − 𝑎6 = − (− ) 𝑎0 = 𝑎0 .
4 4 1⋅2⋅3 1⋅2⋅3⋅4

In general,
∏𝑚
𝑘=1 (2𝑘 − 1)
𝑎2𝑚 = (−1)𝑚 𝑎0 , 𝑚 ≥ 0.
𝑚!

Computing the coefficients of odd powers of 𝑥 yields

1
𝑎3 = − 4 ⋅ 𝑎1 ,
3
2 2 1 1⋅2
𝑎5 = − 4 ⋅ 𝑎3 = −4 ⋅ ( −4 ) 𝑎1 = 42 𝑎1 ,
5 5 3 3⋅5
3 3 21 ⋅ 2 1⋅2⋅3
𝑎7 = − 4 ⋅ 𝑎5 = −4 ⋅ ( 4 ) 𝑎1 = −43 𝑎1 ,
7 7 3⋅5 3⋅5⋅7
4 4 31 ⋅ 2 ⋅ 3 1⋅2⋅3⋅4
𝑎9 = − 4 ⋅ 𝑎7 = −4 ⋅ ( 4 ) 𝑎1 = 44 𝑎1 .
9 9 3⋅5⋅7 3⋅5⋅7⋅9
In general,
(−1)𝑚 4𝑚 𝑚!
𝑎2𝑚+1 = 𝑎1 , 𝑚 ≥ 0.
∏𝑚𝑘=1 (2𝑘 + 1)

Therefore, the power series form of the general solution is

∞ 𝑚 ∞
𝑚 ∏𝑗=1
(2𝑗 − 1) 2𝑚 4𝑚 𝑚!
𝑦 = 𝑎0 ∑(−1) 𝑥 + 𝑎1 ∑(−1)𝑚 𝑥 2𝑚+1 .
𝑚=0
𝑚! 𝑚=0 ∏𝑚
𝑗=1 (2𝑗
+ 1)

■
Using the method shown in the above example, we find the power series solution of the
more general equation
(1 + 𝛼(𝑥 − 𝑥0 )2 )𝑦 ′′ + 𝛽(𝑥 − 𝑥0 )𝑦 ′ + 𝛾 𝑦 = 0.

146 Lecture Notes for MA451

 8.2 Series Solutions Near an Ordinary Point

Theorem 8.5
The coefficients {𝑎𝑛 } in any solution 𝑦 = ∑∞ 𝑛
𝑛=0 𝑎𝑛 (𝑥 − 𝑥0 ) of the equation

2 ′′ ′
(1 + 𝛼(𝑥 − 𝑥0 ) ) 𝑦 + 𝛽(𝑥 − 𝑥0 )𝑦 + 𝛾 𝑦 = 0

satisfy the recurrence relation

𝑝(𝑛)
𝑎𝑛+2 = − 𝑎𝑛 , 𝑛 ≥ 0,
(𝑛 + 2)(𝑛 + 1)

where
𝑝(𝑛) = 𝛼𝑛(𝑛 − 1) + 𝛽𝑛 + 𝛾 .

Moreover, the coefficients of the even and odd powers of 𝑥 − 𝑥0 can be computed separately
as
𝑝(2𝑚)
𝑎2𝑚+2 = − 𝑎2𝑚 , 𝑚 ≥ 0
(2𝑚 + 2)(2𝑚 + 1)
𝑝(2𝑚 + 1)
𝑎2𝑚+3 = − 𝑎2𝑚+1 , 𝑚≥0
(2𝑚 + 3)(2𝑚 + 2)
where 𝑎0 and 𝑎1 are arbitrary.
♥
∞
Example 8.7 Compute 𝑎0 , 𝑎1 , … , 𝑎7 in the power series solution 𝑦 = ∑ 𝑎𝑛 𝑥 𝑛 of the initial value
𝑛=0
problem
(1 + 2𝑥 2 )𝑦 ′′ + 10𝑥𝑦 ′ + 8𝑦 = 0, 𝑦(0) = 2, 𝑦 ′ (0) = −3.

Solution Since 𝛼 = 2, 𝛽 = 10, and 𝛾 = 8 in the equation, we have

𝑝(𝑛) = 2𝑛(𝑛 − 1) + 10𝑛 + 8 = 2(𝑛 + 2)2 .

Therefore,
(𝑛 + 2)2 𝑛+2
𝑎𝑛+2 = −2 𝑎𝑛 = −2 𝑎𝑛 , 𝑛 ≥ 0.
(𝑛 + 2)(𝑛 + 1) 𝑛+1
For 𝑛 = 2𝑚, we have
(2𝑚 + 2) 𝑚+1
𝑎2𝑚+2 = −2 𝑎2𝑚 = −4 𝑎2𝑚 , 𝑚≥0
2𝑚 + 1 2𝑚 + 1

For 𝑛 = 2𝑚 + 1, we have

2𝑚 + 3 2𝑚 + 3
𝑎2𝑚+3 = −2 𝑎2𝑚+1 = − 𝑎2𝑚+1 , 𝑚 ≥ 0.
2𝑚 + 2 𝑚+1

147 Lecture Notes for MA451

 8.2 Series Solutions Near an Ordinary Point

Since 𝑎0 = 𝑦(0) = 2, we have

1
𝑎2 = −4 ⋅ 2 = −8,
1
2 64
𝑎4 = −4 ⋅ (−8) = ,
3 3
3 64 256
𝑎6 = −4 ⋅ ( ) = − .
5 3 5

Since 𝑎1 = 𝑦 ′ (0) = −3, we have

3
𝑎3 = − (−3) = 9,
1
5 45
𝑎5 = − 9 = − ,
2 2
7 45 105
𝑎7 = − (− ) = .
3 2 2

Therefore, the solution in power series form is

64 4 45 5 256 6 105 7
𝑦 = 2 − 3𝑥 − 8𝑥 2 + 9𝑥 3 + 𝑥 − 𝑥 − 𝑥 + 𝑥 + ⋯.
3 2 5 2
■
 Exercise 8.5 Let 𝑥0 be an arbitrary real number. Find the power series in (𝑥 − 𝑥0 ) for the
general solution of
𝑦 ′′ + 𝑦 = 0.

Solution Let
𝐿𝑦 = 𝑦 ′′ + 𝑦.

Suppose
∞
𝑦 = ∑ 𝑎𝑛 (𝑥 − 𝑥0 )𝑛 ,
𝑛=0

then ∞
𝑦 ′′ = ∑ 𝑛(𝑛 − 1)𝑎𝑛 (𝑥 − 𝑥0 )𝑛−2 .
𝑛=2

So ∞ ∞
𝑛−2
𝐿𝑦 = ∑ 𝑛(𝑛 − 1)𝑎𝑛 (𝑥 − 𝑥0 ) + ∑ 𝑎𝑛 (𝑥 − 𝑥0 )𝑛
𝑛=2 𝑛=0
∞ ∞
= ∑(𝑛 + 2)(𝑛 + 1)𝑎𝑛+2 (𝑥 − 𝑥0 )𝑛 + ∑ 𝑎𝑛 (𝑥 − 𝑥0 )𝑛
𝑛=0 𝑛=0
∞
= ∑((𝑛 + 2)(𝑛 + 1)𝑎𝑛+2 + 𝑎𝑛 ).
𝑛=0

148 Lecture Notes for MA451

 8.2 Series Solutions Near an Ordinary Point

Therefore 𝐿𝑦 = 0 if and only if

𝑎𝑛
𝑎𝑛+2 = − , 𝑛 ≥ 0,
(𝑛 + 2)(𝑛 + 1)

where 𝑎0 and 𝑎1 are arbitrary.

Since the indices on the left and right sides of the recurrence relation differ by two, we
write the recurrence relation separately for 𝑛 even 𝑛 = 2𝑚 and 𝑛 odd 𝑛 = 2𝑚 + 1, where
𝑚 = 0, 1, 2, … . Then
(−1)𝑎2𝑚
𝑎2𝑚+2 = , 𝑚≥0
(2𝑚 + 2)(2𝑚 + 1)
and
−𝑎2𝑚+1
𝑎2𝑚+3 = , 𝑚 ≥ 0.
(2𝑚 + 3)(2𝑚 + 2)

Computing the coefficients of the even powers of 𝑥 − 𝑥0 using the recurrence relation
yields
𝑎0
𝑎2 = −
2⋅1
𝑎2 1 𝑎0 𝑎0
𝑎4 = − =− ( − ) = ,
4⋅3 4⋅3 2⋅1 4⋅3⋅2⋅1
𝑎4 1 𝑎0 𝑎0
𝑎6 = − =− = − ,
6⋅5 6 ⋅ 5 (4 ⋅ 3 ⋅ 2 ⋅ 1) 6⋅5⋅4⋅3⋅2⋅1
and, in general,
𝑎0
𝑎2𝑚 = (−1)𝑚 , 𝑚 ≥ 0.
(2𝑚)!

Similarly, computing the coefficients of the odd powers of 𝑥 − 𝑥0 yields

𝑎1
𝑎3 = −
3⋅2
𝑎3 1 𝑎1 𝑎1
𝑎5 = − =− ( − ) = ,
5⋅4 5⋅4 3⋅2 5⋅4⋅3⋅2
𝑎5 1 𝑎1 𝑎1
𝑎7 = − =− ( ) =− ,
7⋅6 7⋅6 5⋅4⋅3⋅2 7⋅6⋅5⋅4⋅3⋅2
and, in general,
(−1)𝑚 𝑎1
𝑎2𝑚+1 = 𝑚 ≥ 0.
(2𝑚 + 1)!
Therefore, the general solution can be written as
∞ ∞
(𝑥 − 𝑥0 )2𝑚 (𝑥 − 𝑥0 )2𝑚+1
𝑦 = 𝑎0 ∑(−1)𝑚 + 𝑎1 ∑(−1)𝑚 .
𝑚=0
(2𝑚)! 𝑚=0
(2𝑚 + 1)!

Recall from calculus that

∞ ∞
(𝑥 − 𝑥0 )2𝑚 (𝑥 − 𝑥0 )2𝑚+1
∑(−1)𝑚 = cos(𝑥 − 𝑥0 ) and ∑(−1)𝑚 = sin(𝑥 − 𝑥0 ).
𝑚=0
(2𝑚)! 𝑚=0
(2𝑚 + 1)!

149 Lecture Notes for MA451

 8.2 Series Solutions Near an Ordinary Point

Then the solution is indeed,

𝑦 = 𝑎0 cos(𝑥 − 𝑥0 ) + 𝑎1 sin(𝑥 − 𝑥0 ).

■
 Exercise 8.6 Find the coefficients 𝑎0 , 𝑎1 , . . . , 𝑎7 of the power series solution of the initial
value problem
𝑦 ′′ + 𝑥𝑦 ′ + 𝑦 = 0, 𝑦(0) = 1, 𝑦 ′ (0) = 1.

∞
Solution Let 𝐿𝑦 = 𝑦 ′′ + 𝑥𝑦 ′ + 𝑦 and 𝑦 = ∑ 𝑎𝑛 𝑥 𝑛 . Then
𝑛=0

∞
′
𝑦 = ∑ 𝑛𝑎𝑛 𝑥 𝑛−1 ,
𝑛=1

∞
𝑦 ′′ = ∑ 𝑛(𝑛 − 1)𝑎𝑛 𝑥 𝑛−2 .
𝑛=2

Therefore,
∞ ∞ ∞
𝐿𝑦 = ∑ 𝑛(𝑛 − 1)𝑎𝑛 𝑥 𝑛−2 + 𝑥 ∑ 𝑛𝑎𝑛 𝑥 𝑛−1 + ∑ 𝑎𝑛 𝑥 𝑛
𝑛=2 𝑛=1 𝑛=0
∞ ∞ ∞
= ∑(𝑛 + 2)(𝑛 + 1)𝑎𝑛+2 𝑥 𝑛 + ∑ 𝑛𝑎𝑛 𝑥 𝑛 + ∑ 𝑎𝑛 𝑥 𝑛
𝑛=0 𝑛=1 𝑛=0
∞ ∞ ∞
= ∑(𝑛 + 2)(𝑛 + 1)𝑎𝑛+2 𝑥 𝑛 + ∑ 𝑛𝑎𝑛 𝑥 𝑛 + ∑ 𝑎𝑛 𝑥 𝑛
𝑛=0 𝑛=0 𝑛=0
∞
= ∑[(𝑛 + 2)(𝑛 + 1)𝑎𝑛+2 + (𝑛 + 1)𝑎𝑛 ]𝑥 𝑛
𝑛=0

If 𝑦 is a solution, then 𝐿𝑦 = 0 which yields the recurrence relation

𝑎𝑛
𝑎𝑛+2 = − .
𝑛+2
When 𝑛 = 2𝑚 is even,
𝑎2𝑚−2
𝑎2𝑚 = − , 𝑚 ≥ 1.
2𝑚
When 𝑛 = 2𝑚 + 1 is odd, then
𝑎2𝑚−1
𝑎2𝑚+1 = − , 𝑚 ≥ 1.
2𝑚 + 1

150 Lecture Notes for MA451

 8.3 Method of Frobenius and Euler Equations

Since 𝑎0 = 𝑦(0) = 1, we see that

𝑎0 1
𝑎2 = −
=− ,
2 2
𝑎2 1 1 1
𝑎4 = − = ⋅ = ,
4 2 4 8
𝑎4 1 1 1
𝑎6 = − = − ⋅ = − .
6 8 6 48

Since 𝑎1 = 𝑦 ′ (0) = 1, we see that

𝑎1 1
𝑎3 = − =− ,
3 3
𝑎3 1 1 1
𝑎5 = − = ⋅ = ,
5 3 5 15
𝑎5 1 1 1
𝑎7 = − = − ⋅ = − .
7 15 7 105

Therefore, the power series solution is in the form

1 1 1 1 1 1 7
𝑦 = 1 + 𝑥 − 𝑥2 − 𝑥3 + 𝑥4 + 𝑥5 − 𝑥6 − 𝑥 + ⋯.
2 3 8 15 48 105
■

8.3 Method of Frobenius and Euler Equations

Note that differentiating a power series increases the lowest exponent. At a regular point,
1
𝑃0
is also analytic. Hence, the equation

𝑃1 ′ 𝑃2
𝑦 ′′ + 𝑦 + 𝑦=0
𝑃0 𝑃0

may have a power series solution.

At a singular point, the rational function 𝑃10 is no longer analytic, that is, it may not have a
power series expression. For example, in the hypergeometric equation

𝑥(1 − 𝑥)𝑦 ′′ + [𝑐 − (𝑎 + 𝑏 + 1)𝑥]𝑦 ′ − 𝑎𝑏𝑦 = 0,

where 𝑎, 𝑏, and 𝑐 are constants, the coefficient 𝑃0 = 𝑥(1 − 𝑥), then

∞
1 1 1
= = 𝑥 −1 = 𝑥 −1 ∑ 𝑥 𝑛 .
𝑃0 𝑥(1 − 𝑥) 1−𝑥 𝑛=0

∞
If 𝑦(𝑥) = ∑ 𝑎𝑛 𝑥 𝑛 is a solution, then the lowest exponent of 𝑦 ′′ may be strictly greater than
𝑛=0
those of the other two terms. As a consequence, at 𝑥 = 0, there may be no power series

151 Lecture Notes for MA451

 8.3 Method of Frobenius and Euler Equations

solution.

Here is another concrete example, the equation

2 2
𝑦 ′′ + 𝑦 ′ − 2 𝑦 = 0
𝑥 𝑥

has two solutions, 𝑦1 = 𝑥, and 𝑦2 = 𝑥 −2 , where 𝑦2 = 𝑥 −2 does not have a power series expression
∑∞ 𝑛
𝑛=0 𝑎𝑛 𝑥 .

Mathematicians don’t give up, when one method does not work, they look for other
methods. When the equation 𝑦 ′′ + 𝑃𝑃10 𝑦 ′ + 𝑃𝑃02 𝑦 = 0 is not too singular at a singular point 𝑥0 , in the
sense that 𝑃0 = (𝑥 − 𝑥0 )2 𝐴(𝑥), where 𝐴(𝑥) is a polynomial and 𝐴(𝑥0 ) ≠ 0, German mathematician,
Ferdinand Frobenius developed the method of finding series solution in the form
∞
𝑦(𝑥) = 𝑥 𝑟 ∑ 𝑎𝑛 𝑥 𝑛 .
𝑛=0

Definition 8.2
Let 𝑃0 , 𝑃1 , and 𝑃2 be polynomials with no common factor and suppose 𝑃0 (𝑥0 ) = 0. Then 𝑥0 is a
regular singular point of the equation

𝑃0 (𝑥)𝑦 ′′ + 𝑃1 (𝑥)𝑦 ′ + 𝑃2 (𝑥)𝑦 = 0

if
(𝑥 − 𝑥0 )𝑃1 (𝑥 − 𝑥0 )2 𝑃2
and
𝑃0 𝑃0
are analytic at 𝑥0 .
Otherwise, 𝑥0 is called an irregular singular point of the equation.
♣

Example 8.8 Bessel’s equation

The equation
𝑥 2 𝑦 ′′ + 𝑥𝑦 ′ + (𝑥 2 − 𝜈 2 )𝑦 = 0,

has the singular point 𝑥0 = 0. Determine if 𝑥0 = 0 is a regular singular point.

Solution Since
𝑥𝑃1 𝑥 2 𝑥 2 𝑃2 𝑥 2 (𝑥 2 − 𝜈 2 )
= 2 = 1 and = 2
= (𝑥 2 − 𝜈 2 )
𝑃0 𝑥 𝑃0 𝑥
are analytic at 0, the point 𝑥0 = 0 is regular singular. ■
 Exercise 8.7 Legendre’s equation

The equation
(1 − 𝑥 2 )𝑦 ′′ − 2𝑥𝑦 ′ + 𝛼(𝛼 + 1)𝑦 = 0

152 Lecture Notes for MA451

 8.3 Method of Frobenius and Euler Equations

has the singular points 𝑥0 = ±1. Determine if 𝑥0 = ±1 are regular singular points.

Solution Since
(𝑥 ± 1)𝑃1 −(𝑥 ± 1)2𝑥 −2𝑥
= =
𝑃0 𝑥2 − 1 𝑥 ∓1
and
(𝑥 ± 1)2 𝑃2 𝛼(𝛼 + 1)(𝑥 ± 1)2 𝛼(𝛼 + 1)(𝑥 ± 1)
= =
𝑃0 𝑥2 − 1 𝑥 ∓1
are both analytic at ±1, the points 𝑥0 = ±1 are regular singular points of the equation. ■
At this stage, the only second order linear equation we can solve completely near a sin-
gular point is the Euler equation.
Definition 8.3 (Euler Equation)
An Euler equation is an equation that can be written in the form

𝑎𝑥 2 𝑦 ′′ + 𝑏𝑥𝑦 ′ + 𝑐𝑦 = 0,

where 𝑎, 𝑏, and 𝑐 are real constants and 𝑎 ≠ 0.

♣

From the existence theorem (Theorem 5.1), we know that Euler equation has solutions
defined on (0, ∞) and (−∞, 0). Since the two intervals are symmetric, by a substitution 𝑡 = −𝑥
when 𝑥 < 0, we may and will restrict ourself to the interval (0, ∞).

The normalized form equation

𝑝 𝑞
𝑦 ′′ + 𝑦 ′ + 2 𝑦 = 0,
𝑥 𝑥
𝑏
where 𝑝 = 𝑎
and 𝑞 = 𝑎𝑐 , suggest that a solution is in the form

∞ ∞
𝑦 = 𝑥 ∑ 𝑎𝑛 𝑥 = ∑ 𝑎𝑛 𝑥 𝑛+𝑟 .
𝑟 𝑛

𝑛=0 𝑛=0

We can then determine 𝑟 and 𝑎𝑛 ’s by plugging the series into the equation. Indeed, for Euler
equation, we can take 𝑎0 = 1 and 𝑎𝑛 = 0 for 𝑛 = 1, 2, 3, … .

Differentiating 𝑦 yields
𝑦 ′ = ∑(𝑛 + 𝑟)𝑎𝑛 𝑥 𝑛+𝑟−1 ,
𝑛=0

𝑦 ′′ = ∑(𝑛 + 𝑟)(𝑛 + 𝑟 − 1)𝑎𝑛 𝑥 𝑛+𝑟−2 ,

𝑛=0

153 Lecture Notes for MA451

 8.3 Method of Frobenius and Euler Equations

and
𝑝 𝑞
𝑦 ′′ + 𝑦 ′ + 2
𝑥 𝑥
=𝑦 + 𝑝𝑥 𝑦 + 𝑞𝑥 −2
′′ −1 ′

∞
= ∑(𝑛 + 𝑟)(𝑛 + 𝑟 − 1)𝑎𝑛 𝑥 𝑛+𝑟−2 + 𝑝 ∑(𝑛 + 𝑟)𝑎𝑛 𝑥 𝑛+𝑟−2 + 𝑞 ∑ 𝑎𝑛 𝑥 𝑛+𝑟−2
𝑛=0 𝑛=0 𝑛=0
𝑛+𝑟−2
= ∑[(𝑛 + 𝑟)(𝑛 + 𝑟 − 1) + 𝑝(𝑛 + 𝑟) + 𝑞]𝑎𝑛 𝑥 .
𝑛=0

Therefore, 𝑦 is a solution if

[(𝑛 + 𝑟)(𝑛 + 𝑟 − 1) + 𝑝(𝑛 + 𝑟) + 𝑞]𝑎𝑛 = 0 for 𝑛 = 0, 1, 2, … .

A sufficient condition is that

𝑟(𝑟 − 1) + 𝑝𝑟 + 𝑞 = 0 and 𝑎𝑛 = 0 for 𝑛 = 1, 2, … .

The equation
𝑟(𝑟 − 1) + 𝑝𝑟 + 𝑞 = 0

or equivalently,
𝑎𝑟(𝑟 − 1) + 𝑏𝑟 + 𝑐 = 0

is called the indicial equation of the differential equation

𝑎𝑥 2 𝑦 ′′ + 𝑏𝑥𝑦 ′ + 𝑐𝑦 = 0.

Example 8.9 Find the general solution of

𝑥 2 𝑦 ′′ − 𝑥𝑦 ′ − 8𝑦 = 0

on (0, ∞).

Solution The indicial equation is

𝑟(𝑟 − 1) − 𝑟 − 8 = 0.

Equivalently,
𝑟 2 − 2𝑟 − 8 = 0

Solving the equation yields 𝑟 = −2 or 𝑟 = 4. Then 𝑦1 = 𝑥 4 and 𝑦2 = 𝑥 −2 are solutions of the equation
on (0, ∞).

Because the Wronskian is

𝑊 (𝑦1 , 𝑦2 ) = 𝑦1 𝑦2′ − 𝑦1′ 𝑦2 = 𝑥 4 ⋅ (𝑥 −2 )′ − (𝑥 4 )′ ⋅ 𝑥 −2 = −6𝑥 ≢ 0.

154 Lecture Notes for MA451

 8.3 Method of Frobenius and Euler Equations

Therefore, 𝑦1 and 𝑦2 are linearly independent and the general solution on (0, ∞) is
𝑐2
𝑦 = 𝑐1 𝑥 4 + .
𝑥2
■
 Exercise 8.8 Find the general solution of

6𝑥 2 𝑦 ′′ + 5𝑥𝑦 ′ − 𝑦 = 0

on (0, ∞).

Solution The indicial equation is

6𝑟(𝑟 − 1) + 5𝑟 − 1 = 0.

Solving the equation yields

1 1
𝑟= or 𝑟 =− .
2 3

Therefore, the general solution is

𝑦 = 𝑐1 𝑥 1/2 + 𝑐2 𝑥 −1/3 .

■
When the indicial equation has a repeated solution or complex solutions, to find the gen-
eral solution, we can using the Wronskian (see Proposition 5.1) and Euler’s formula

𝑥 𝑖𝜔 = 𝑒 i𝜔 ln 𝑥 = cos(𝜔 ln 𝑥) + i sin(𝜔 ln 𝑥).

The general result is summarized in the following theorem.

Theorem 8.6
Suppose the roots of the indicial equation

𝑎𝑟(𝑟 − 1) + 𝑏𝑟 + 𝑐 = 0

are 𝑟1 and 𝑟2 . Then the general solution of the Euler equation

𝑎𝑥 2 𝑦 ′′ + 𝑏𝑥𝑦 ′ + 𝑐𝑦 = 0

on (0, ∞) is

𝑦 = 𝑐1 𝑥 𝑟1 + 𝑐2 𝑥 𝑟2

155 Lecture Notes for MA451

 8.3 Method of Frobenius and Euler Equations

if 𝑟1 and 𝑟2 are distinct real numbers;

𝑦 = 𝑥 𝑟 (𝑐1 + 𝑐2 ln 𝑥)

if 𝑟1 = 𝑟2 = 𝑟;

𝑦 = 𝑥 𝜆 [𝑐1 cos (𝜔 ln 𝑥) + 𝑐2 sin (𝜔 ln 𝑥)]

if 𝑟1 , 𝑟2 = 𝜆 ± i𝜔 with 𝜔 > 0.
♥

Here we present a shorter proof using the substitution 𝑧 = ln 𝑥.

Proof Let 𝑧 = ln 𝑥, or equivalently 𝑥 = 𝑒 𝑧 . Let 𝑢(𝑧) = 𝑦(𝑒 𝑧 ). By the chain rule, we get

𝑢 ′ (𝑧) = 𝑦 ′ (𝑒 𝑧 )𝑒 𝑧 = 𝑥𝑦 ′ (𝑥),

𝑢 ′′ (𝑧) = (𝑦 ′ (𝑒 𝑧 )𝑒 𝑧 )′ = 𝑦 ′′ (𝑒 𝑧 )(𝑒 𝑧 )2 + 𝑦 ′ (𝑒 𝑧 )𝑒 𝑧 = 𝑥 2 𝑦 ′′ (𝑥) + 𝑥𝑦 ′ (𝑥)

Therefore,

𝑎𝑥 2 𝑦 ′′ (𝑥) + 𝑏𝑥𝑦 ′ (𝑥) + 𝑐𝑦(𝑥) = 𝑎𝑢 ′′ (𝑧) + (𝑏 − 𝑎)𝑢 ′ (𝑧) + 𝑐𝑢(𝑧).

The equation
𝑎𝑢 ′′ + (𝑏 − 𝑎)𝑢 ′ + 𝑐𝑢 = 0

has the characteristic equation

𝑎𝑟 2 + (𝑏 − 𝑎)𝑟 + 𝑐 = 𝑎𝑟(𝑟 − 1) + 𝑏𝑟 + 𝑐 = 0.

Because 𝑟1 and 𝑟2 are solutions of this characteristic equation.

If 𝑟1 and 𝑟2 are two distinct real root, then the general solution of the Euler equation is

𝑦(𝑥) = 𝑢(𝑧) = 𝑐1 𝑒 𝑟1 𝑧 + 𝑐2 𝑒 𝑟2 𝑧 = 𝑐1 𝑥 𝑟1 + 𝑐2 𝑥 𝑟2 .

If 𝑟1 = 𝑟2 = 𝑟, then the general solution of the Euler equation is

𝑦(𝑥) = 𝑢(𝑧) = 𝑒 𝑟𝑧 (𝑐1 + 𝑐2 𝑧) = 𝑥 𝑟 (𝑐1 + 𝑐2 ln 𝑥).

If 𝑟1 , 𝑟2 = 𝜆 + i𝜔, then the general solution of the Euler equation is

𝑦(𝑥) = 𝑢(𝑧) = 𝑒 𝛼𝑧 (𝑐1 cos(𝜔𝑧) + 𝑐2 sin(𝜔𝑧)) = 𝑥 𝛼 (𝑐1 cos(𝜔 ln 𝑥) + 𝑐2 sin(𝜔 ln 𝑥)).

156 Lecture Notes for MA451

 8.3 Method of Frobenius and Euler Equations

■
Example 8.10 Find the general solution of

𝑥 2 𝑦 ′′ − 5𝑥𝑦 ′ + 9𝑦 = 0

on (0, ∞).

Solution The indicial equation is

𝑟(𝑟 − 1) − 5𝑟 + 9 = 0

which has a repeated root 𝑟 = 3.

Therefore, the general solution of the equation on (0, ∞) is

𝑦 = 𝑥 3 (𝑐1 + 𝑐2 ln 𝑥).

■
Example 8.11 Find the general solution of

𝑥 2 𝑦 ′′ + 3𝑥𝑦 ′ + 2𝑦 = 0

on (0, ∞).

Solution The indicial equation is

𝑟(𝑟 − 1) + 3𝑟 + 2 = 0

which has two complex solutions 𝑟1 , 𝑟2 = −1 ± i

Therefore, the general solution is

1
𝑦= [𝑐1 cos(ln 𝑥) + 𝑐2 sin(ln 𝑥)] .
𝑥
■
 Exercise 8.9 Find the general solution of

𝑥 2 𝑦 ′′ + 5𝑥𝑦 ′ + 4𝑦 = 0

on (0, ∞).

Solution The indicial equation is

𝑟(𝑟 − 1) + 5𝑟 + 4 = 0

which has a repeated root 𝑟 = −2.

157 Lecture Notes for MA451

 8.3 Method of Frobenius and Euler Equations

Therefore, the general solution of the equation on (0, ∞) is

𝑦 = 𝑥 −2 (𝑐1 + 𝑐2 ln 𝑥).

■
 Exercise 8.10 Find the general solution of

𝑥 2 𝑦 ′′ + 𝑥𝑦 ′ + 4𝑦 = 0

on (0, ∞).

Solution The indicial equation is

𝑟(𝑟 − 1) + 𝑟 + 4 = 0

which has two complex solutions 𝑟1 , 𝑟2 = ±2i

Therefore, the general solution is

𝑦 = [𝑐1 cos(2 ln 𝑥) + 𝑐2 sin(2 ln 𝑥)] .

158 Lecture Notes for MA451

 Week 9: Introduction to Laplace Transforms

11/08–11/11

9.1 Laplace Transform

Using infinite series to solve differential equations provides an new idea. Transforming a
solution into a series can reduce the differential equation into algebraic equations of coeffi-
cients which are much easier to solve. From calculus, we know that power series are closely
∞ ∞
related to improper integrals. Indeed, the improper integrals ∫0 𝑎(𝑡)𝑥 𝑡 d 𝑡 = ∫0 𝑎(𝑡)𝑒 −𝑠𝑡 d 𝑡 may
∞
be considered a continuous analogue of the power series ∑ 𝑎𝑛 𝑥 𝑛 , where −𝑠 = ln 𝑥 and 𝑡 plays
𝑛=0
the role of 𝑛. The interested reader may watch the youtube video by Professor Arthur Mat-
tuck for a detailed explanation. This observation suggests an idea of applying an integral
transformation to reduce a differential equation to algebraic equations. One of such integral
transformations is the Laplace transformation which transform function 𝑓 (𝑡) in real variable is
another function depending on a new variable 𝑠 that can be a complex number in general.

Definition of the Laplace Transform

Definition 9.1
Let 𝑓 (𝑡) be a function in a real varaiable 𝑡. The Laplace transform 𝑓 of 𝑓 is a function of 𝑠
defined by
∞
(𝑓 )(𝑠) ∶= ∫ 𝑓 (𝑡)𝑒 −𝑠𝑡 d𝑡,
0

whose domain consists of all values of 𝑠 such that the improper integral converges.
♣

In the definition, the value 𝑠 may be taken to be complex. For simplicity, we assume that
𝑠 > 0. We often denote (𝑓 )(𝑠) by 𝐹 (𝑠).

By the definition of the improper integral,

∞ 𝑁
−𝑠𝑡
𝐹 (𝑠) = ∫ 𝑓 (𝑡)𝑒 d 𝑡 = lim ∫ 𝑓 (𝑡)𝑒 −𝑠𝑡 d 𝑡.
0 𝑁 →∞ 0

Example 9.1 Find the Laplace transform for the function 𝑓 (𝑡) = 1.

Solution By a substitution 𝑢 = −𝑠𝑡, the improper integral is

∞ 𝑁 −𝑠𝑁
1 𝑢 1 1
(1) = ∫ 𝑒 −𝑠𝑡 d 𝑡 = lim ∫ 𝑒 −𝑠𝑡 d 𝑡 = lim ∫ 𝑒 d 𝑢 = − lim (𝑒 −𝑠𝑁 − 𝑒 0 ) =
0 𝑁 →∞0 𝑁 →∞ 0 −𝑠 𝑠 𝑁 →∞ 𝑠
 9.1 Laplace Transform

if 𝑠 > 0. ■
Example 9.2 Find the Laplace transform for the function 𝑓 (𝑡) = 𝑡 𝑛 .

Solution Applying integration by parts iteratedly, the indefinite integral ∫ 𝑡 𝑛 𝑒 −𝑠𝑡 d 𝑡 is

(𝑡 𝑛 ) = ∫ 𝑡 𝑛 𝑒 −𝑠𝑡 d 𝑡
1 1
= − 𝑡 𝑛 𝑒 −𝑠𝑡 + ∫ 𝑛𝑡 𝑛−1 𝑒 −𝑠𝑡 d 𝑡
𝑠 𝑠
1 𝑛 −𝑠𝑡 1 𝑛−1 −𝑠𝑡 1
= − 𝑡 𝑒 − 2 𝑛𝑡 𝑒 + 2 ∫ 𝑛(𝑛 − 1)𝑡 𝑛−2 𝑒 −𝑠𝑡 d 𝑡
𝑠 𝑠 𝑠
⋮
1 1 𝑛 ⋅ ⋯ ⋅ 2 −𝑠𝑡 𝑛!
= − 𝑡 𝑛 𝑒 −𝑠𝑡 − 2 𝑛𝑡 𝑛−1 𝑒 −𝑠𝑡 − ⋯ − 𝑛
𝑡𝑒 − 𝑛+1 𝑒 −𝑠𝑡 .
𝑠 𝑠 𝑠 𝑠
𝑡𝑘
Since lim 𝑒 −𝑠𝑡 = 0 for any finite 𝑘, the improper integral is
𝑁 →∞

∞ 𝑁
𝑛!
∫ 𝑡 𝑛 𝑒 −𝑠𝑡 d 𝑡 = lim ∫ 𝑡 𝑛 𝑒 −𝑠𝑡 d 𝑡 =
0 𝑁 →∞ 0 𝑠 𝑛+1

if 𝑠 > 0. ■
 Exercise 9.1 Find the Laplace transform for the function 𝑓 (𝑡) = 𝑒 𝑟𝑡 .

Solution By the definition,

∞ ∞
(𝑒 𝑟𝑡 ) = ∫ 𝑒 𝑟𝑡 𝑒 −𝑠𝑡 d 𝑡 = ∫ 𝑒 (𝑟−𝑠)𝑡 d 𝑡.
0 0

From the first example, we know that

∞
1
∫ 𝑒 (𝑟−𝑠)𝑡 d 𝑡 =
0 𝑠−𝑟

if 𝑠 > 𝑟. ■
 Exercise 9.2 Find the Laplace transforms of 𝑓 (𝑡) = sin(𝜔𝑡) and 𝑔(𝑡) = cos(𝜔𝑡) where 𝜔 is a
constant.

Solution Let 𝐹 (𝑠) = (𝑓 )(𝑠) and 𝐺(𝑠) = (𝑔)(𝑠). Assume 𝑠 > 0. Applying integration by parts
yields
∞
𝑒 −𝑠𝑡 |∞ 𝜔 𝜔
𝐹 (𝑠) = − sin(𝜔𝑡)| + ∫ 𝑒 −𝑠𝑡 cos(𝜔𝑡) d 𝑡 = 𝐺(𝑠).
𝑠 | 0 𝑠 0 𝑠

Similarly,
𝑒 −𝑠𝑡 cos(𝜔𝑡) |∞ 𝜔 ∞ −𝑠𝑡 1 𝜔
𝐺(𝑠) = − | − ∫ 𝑒 sin(𝜔𝑡) d 𝑡 = − 𝐹 (𝑠).
𝑠 | 0 𝑠 0 𝑠 𝑠

160 Lecture Notes for MA451

 9.1 Laplace Transform

Therefore,
1 𝜔2
𝐺(𝑠) = − 𝐺(𝑠).
𝑠 𝑠2

Solving for 𝐺 yields

𝑠
𝐺(𝑠) = , 𝑠 > 0.
𝑠2 + 𝜔2

Therefore,
𝜔 𝑠 𝜔
𝐹 (𝑠) = ⋅ 2 2
= 2 , 𝑠 > 0.
𝑠 𝑠 +𝜔 𝑠 + 𝜔2
■

Linearity of the Laplace Transform

Because integration is a linear operator, that is, integration of linear combination is the
linear combination of integrations. The Laplace transform is also a linear operator.
Theorem 9.1
Suppose (𝑓𝑖 ) is defined for 𝑠 > 𝑠𝑖 , 1 ≤ 𝑖 ≤ 2. Let 𝑠0 be the largest of the numbers 𝑠1 and 𝑠2 , and
let 𝑐1 and 𝑐2 constants. Then

(𝑐1 𝑓1 + 𝑐2 𝑓2 ) = 𝑐1 (𝑓1 ) + 𝑐2 (𝑓2 ) for 𝑠 > 𝑠0 .

♥

Proof By the linear combination law of integration, we get

∞ ∞ ∞
(𝑐1 𝑓1 + 𝑐2 𝑓2 ) = ∫ (𝑐1 𝑓1 + 𝑐2 𝑓2 )𝑒 −𝑠𝑡 d 𝑡 = 𝑐1 ∫ 𝑓1 𝑒 −𝑠𝑡 d 𝑡 + 𝑐2 ∫ 𝑓2 𝑒 −𝑠𝑡 d 𝑡 = 𝑐1 (𝑓1 ) + 𝑐2 (𝑓2 ).
0 0 0

The equalities hold if 𝑠 ≥ 𝑠0 . Because both improper integral converge if 𝑠 is no less than
𝑠0 = max{𝑠1 , 𝑠2 }. ■
Recall the the hyperbolic trigonometric functions sinh 𝑥 and cosh 𝑥 are

𝑒 𝑥 − 𝑒 −𝑥 𝑒 𝑥 + 𝑒 −𝑥
sinh 𝑥 = cosh 𝑥 .
2 2

Example 9.3 Find the (sinh(𝑏𝑡))(𝑠).

Solution By the linearity and definitions, we get

𝑒 𝑏𝑡 − 𝑒 −𝑏𝑡 1 1 1 1 𝑏
(sinh(𝑏𝑡))(𝑠) =  = ((𝑒 𝑏𝑡 ) − (𝑒 −𝑏𝑡 )) = ( − = .
( 2 ) 2 2 𝑠 − 𝑏 𝑠 + 𝑏 ) 𝑠2 − 𝑏2

if 𝑠 > 𝑏. ■
 Exercise 9.3 Find the (cosh(𝑏𝑡)).

161 Lecture Notes for MA451

 9.1 Laplace Transform

Solution By the linearity and definitions, we get

𝑒 𝑏𝑡 + 𝑒 −𝑏𝑡 1 1 1 1 𝑠
(cosh(𝑏𝑡))(𝑠) =  = ((𝑒 𝑏𝑡 ) + (𝑒 −𝑏𝑡 )) = ( + = .
( 2 ) 2 2 𝑠 − 𝑏 𝑠 + 𝑏 ) 𝑠2 − 𝑏2

if 𝑠 > 𝑏. ■

𝑠-shifting property
From the definition of Laplace transform, it is not hard to see that the Laplace transform
of the product 𝑒 𝑟𝑡 𝑓 (𝑡) is a shift of the Laplace transform of 𝑓 .
Theorem 9.2 (𝑠-Shifting Property)
Suppose the Laplace transform
∞
(𝑓 )(𝑠) = ∫ 𝑓 (𝑡)𝑒 −𝑠𝑡 d 𝑡
0

is defined for 𝑠 > 𝑠0 . Then (𝑓 )(𝑠 − 𝑟) is the Laplace transform of 𝑒 𝑟𝑡 𝑓 (𝑡) for 𝑠 > 𝑠0 + 𝑟.
♥

Proof By the definition,

∞ ∞
𝑟𝑡 −𝑠𝑡
∫ 𝑒 𝑓 (𝑡)𝑒 d𝑡 = ∫ 𝑓 (𝑡)𝑒 −(𝑠−𝑟)𝑡 d 𝑡 = (𝑓 )(𝑠 − 𝑟).
0 0

■
Example 9.4 Find
(𝑡 𝑛 𝑒 𝑟𝑡 ).

Solution By the shifting theorem,

𝑛!
(𝑡 𝑛 𝑒 𝑟𝑡 )(𝑠) = (𝑡 𝑛 )(𝑠 − 𝑟) = .
(𝑠 − 𝑟)𝑛+1

■
 Exercise 9.4 Find
(𝑠𝑖𝑛(𝜔𝑡)𝑒 𝑟𝑡 ) and (𝑐𝑜𝑠(𝜔𝑡)𝑒 𝑟𝑡 )

Solution By the shifting theorem,

𝜔
(𝑠𝑖𝑛(𝜔𝑡)𝑒 𝑟𝑡 )(𝑠) = (𝑠𝑖𝑛(𝜔𝑡))(𝑠 − 𝑟) = ,
(𝑠 − 𝑟)2 + 𝜔 2
𝑠
(𝑐𝑜𝑠(𝜔𝑡)𝑒 𝑟𝑡 )(𝑠) = (𝑐𝑜𝑠(𝜔𝑡))(𝑠 − 𝑟) = .
(𝑠 − 𝑟)2 + 𝜔 2
■

162 Lecture Notes for MA451

 9.1 Laplace Transform

Remark Using the identities

𝑒 𝑖𝑡 + 𝑒 −𝑖𝑡 𝑒 𝑖𝑡 − 𝑒 −𝑖𝑡
cos 𝑡 = and sin 𝑡 =
2 2
and the linearity theorem, one can also find the Laplace transforms of cos(𝜔𝑡) and sin(𝜔𝑡) using
the shifting theorem.

Change of scale
By a linear substitution 𝑢 = 𝑎𝑡, one can find the Laplace transform (𝑓 (𝑎𝑡)) in terms of
(𝑓 (𝑡)).
Theorem 9.3 (Change of Scale)
Suppose the Laplace transform
∞
(𝑓 (𝑡))(𝑠) = ∫ 𝑓 (𝑡)𝑒 −𝑠𝑡 d 𝑡
0

is defined for 𝑠 > 𝑠0 . Then

1 𝑠
(𝑓 (𝑎𝑡))(𝑠) = (𝑓 ) ( )
𝑎 𝑎
for 𝑠 > 𝑎𝑠0 .
♥

Proof Let 𝑢 = 𝑎𝑡. Then

∞
−𝑠𝑡 1 ∞ 𝑠𝑢 1 𝑠
(𝑓 (𝑎𝑡))(𝑠) = ∫ 𝑓 (𝑎𝑡)𝑒 d 𝑡 = ∫ 𝑓 (𝑢)𝑒 − 𝑎 d 𝑢 = (𝑓 ) ( ) .
0 𝑠 0 𝑠 𝑎

Transform of derivatives
The most important property of Laplace transform for differential equations may be the
relation between (𝑓 ′ ) and (𝑓 ).
Theorem 9.4 (Transform of the first derivative)
Suppose the Laplace transform
∞
(𝑓 (𝑡))(𝑠) = ∫ 𝑓 (𝑡)𝑒 −𝑠𝑡 d 𝑡
0

is defined for 𝑠 > 𝑠0 . Then

(𝑓 ′ (𝑡))(𝑠) = 𝑠(𝑓 )(𝑠) − 𝑦(0)

for 𝑠 > 𝑠0 .
♥

163 Lecture Notes for MA451

 9.2 Existence and Additional Properties*

Proof The definition implies

∞ ∞
|∞
(𝑓 ′ (𝑡))(𝑠) = ∫ 𝑓 ′ (𝑡)𝑒 −𝑠𝑡 d 𝑡 = −𝑓 (𝑡)𝑒 −𝑠𝑡 | + 𝑠 ∫ 𝑓 (𝑡)𝑒 −𝑠𝑡 d 𝑡 = 𝑠(𝑓 )(𝑠) − 𝑓 (0)
0 |0 0

for 𝑠 > 𝑠0 .
∞
Note in the third equality, we used the fact that lim 𝑓 (𝑡)𝑒 −𝑠𝑡 = 0 if ∫0 𝑓 (𝑡)𝑒 −𝑠𝑡 d 𝑡 converges.
𝑁 →∞
■
Example 9.5 Suppose the Laplace transform (𝑓 )(𝑠) is defined for 𝑠 > 𝑠0 . Find the Laplace
transform (𝑓 ′′ ) in terms of (𝑓 ), 𝑓 ′ (0) and 𝑓 (0).

Solution Applying the above theorem repeatedly yields

(𝑓 ′′ ) = 𝑠(𝑓 ′ ) − 𝑓 ′ (0) = 𝑠 2 (𝑓 ) − 𝑠𝑓 (0) − 𝑓 ′ (0).

9.2 Existence and Additional Properties*

Existence of Laplace transforms

Not every function has a Laplace transform. For example,
∞
2
∫ 𝑒 −𝑠𝑡 𝑒 𝑡 𝑑𝑡
0

2
diverges for all 𝑠 because lim 𝑒 −𝑠𝑡 𝑒 𝑡 = ∞. Since definite integrals define for functions that
𝑡→∞
are continuous with at most jumping or removable discontinuities, Laplace transform may be
defined for piece-wise continuous function.
Definition 9.2
A function 𝑓 is piecewise continuous on a finite closed interval [𝑎, 𝑏] if 𝑓 (𝑎+) and 𝑓 (𝑏−) are
finite and 𝑓 is continuous on the open interval (𝑎, 𝑏) except possibly at finitely many points
where 𝑓 may have jump discontinuities or removable discontinuities.
A function 𝑓 is piecewise continuous on the infinite interval [𝑎, ∞) if it is piecewise continu-
ous on [𝑎, 𝑏] for every 𝑏 > 𝑎.
♣

Example 9.6 The unit step function or Heaviside step function defined as follows
⎧
⎪
⎪0 for 𝑡 < 0,
𝑢(𝑡) = ⎨
⎪
⎪ 1 for 𝑡 ≥ 0.
⎩

is piecewise continuous over the infinite interval (−∞, ∞)

164 Lecture Notes for MA451

 9.2 Existence and Additional Properties*

However, piecewise continuity alone does not guarantee the existence of Laplace trans-
form. An addition sufficient condition is that the function grows slower that 𝑒 −𝑠𝑡 for some 𝑠 = 𝑠0
as 𝑡 goes to infinity.
Definition 9.3
A function 𝑓 is of exponential order 𝑠0 if there are constants 𝑀 and 𝑡0 such that

|𝑓 (𝑡)| ≤ 𝑀𝑒 𝑠0 𝑡 , 𝑡 ≥ 𝑡0 .
♣

When value of 𝑠0 is irrelevant, we simply say that 𝑓 is of exponential order.

Theorem 9.5 (Existence of Laplace Transform)
If 𝑓 is piecewise continuous on [0, ∞) and of exponential order 𝑠0 , then (𝑓 ) is defined for 𝑠 > 𝑠0 .
♥
𝑁
Proof Since 𝑓 is piecewise continuous on [0, ∞), the integral ∫0 𝑓 (𝑡)𝑒 −𝑠𝑡 d 𝑡 is defined for all 𝑁 ,
∞
so is the improper integral ∫0 𝑓 (𝑡)𝑒 −𝑠𝑡 d 𝑡 as long as it converges.

Since 𝑓 is of exponential order 𝑠0 , we know that

|𝑓 (𝑡)| ≤ 𝑀𝑒 𝑠0 𝑡 .

Note that the improper integral

∞ ∞
𝑀
∫ 𝑀𝑒 𝑠0 𝑡 𝑒 −𝑠𝑡 d 𝑡 = 𝑀 ∫ 𝑒 −(𝑠−𝑠0 )𝑡 d 𝑡 = .
0 0 𝑠 − 𝑠0
∞
for 𝑠 > 𝑠0 . By the comparison test, |𝐹 (𝑠)| = ∫0 |𝑓 (𝑡)|𝑒 −𝑠𝑡 d 𝑡 converges for 𝑠 > 𝑠0 and so is
∞
∫0 𝑓 (𝑡)𝑒 −𝑠𝑡 d 𝑡.

Therefore, as long as 𝑓 is piecewise continuous and of exponential order, the Laplace

transform of 𝑓 exists. ■
From the proof, we see that lim 𝐹 (𝑠) = 0. Indeed, as long as the Laplace transform of 𝑓
𝑠→∞
exists, it always true that lim 𝐹 (𝑠) = 0.
𝑠→∞

Corollary 9.1
Suppose that the Laplace transform 𝐹 (𝑠) of 𝑓 (𝑡) exists for 𝑠 > 𝑠0 . Then

lim 𝐹 (𝑠) = 0.
𝑠→∞
♥

Proof It suffice to show that 𝐹 (𝑠) cam be expressed as a Laplace transform of a bounded
function.
𝑡
Let 𝑔(𝑡) = ∫0 𝑒 −𝑠1 𝑥 𝑓 (𝑥) d 𝑥, where 𝑠1 > 𝑠0 . Then 𝑔 ′ (𝑡) = 𝑒 −𝑠1 𝑡 𝑓 (𝑡) and 𝑔(𝑡) is bounded because

165 Lecture Notes for MA451

 9.2 Existence and Additional Properties*

|𝑔(𝑡)| < |𝐹 (𝑠1 )|. By integrating by parts,

∞
𝐹 (𝑠) = ∫ 𝑒 −𝑠𝑡 𝑓 (𝑡) d 𝑡
0
∞
=∫ 𝑒 −(𝑠−𝑠1 )𝑡 𝑒 −𝑠1 𝑡 𝑓 (𝑡) d 𝑡
0
∞
=∫ 𝑒 −(𝑠−𝑠1 )𝑡 d 𝑔(𝑡)
0
∞
|∞
=𝑒 −(𝑠−𝑠1 )𝑡 𝑔(𝑡)| + (𝑠 − 𝑠1 ) ∫ 𝑒 −(𝑠−𝑠1 )𝑡 𝑔(𝑡) d 𝑡
|0 0
∞
=(𝑠 − 𝑠1 ) ∫ 𝑒 −(𝑠−𝑠1 )𝑡 𝑔(𝑡) d 𝑡.
0

■
From the corollary, we know that 𝑠 𝑛 , sin 𝑠, cos 𝑠, 𝑒 𝑠 and ln 𝑠 can not be Laplace transforms.
∞
Remark The identity 𝐹 (𝑠) = (𝑠 −𝑠1 ) ∫0 𝑒 −(𝑠−𝑠1 )𝑡 𝑔(𝑡) d 𝑡 can also be used to show that 𝐹 (𝑠) is analytic
(see, for example, Doetsch 1974, Chapter 6).

Example 9.7 Find the Laplace transform of the delayed unit step function 𝑢𝑎 (𝑡) = 𝑢(𝑡 − 𝑎) with
𝑎 > 0, that is
⎧
⎪
⎪0 for 𝑡 < 𝑎,
𝑢𝑎 (𝑡) = ⎨
⎪
⎪1 for 𝑡 ≥ 𝑎.
⎩

Solution The improper integral can be calculated directly:

∞ 𝑎 ∞
1 |∞ 1
∫ 𝑢𝑎 (𝑡)𝑒 −𝑠𝑡 d 𝑡 = ∫ 0 ⋅ 𝑒 −𝑠𝑡 d 𝑡 + ∫ 1 ⋅ 𝑒 −𝑠𝑡 d 𝑡 = − 𝑒 −𝑠𝑡 | = 𝑠𝑎 .
0 0 𝑎 𝑠 |𝑎 𝑠𝑒

Transforms of piecewise functions

Using the unit step function, a piecewise function
⎧
⎪
⎪ 𝑓1 (𝑡) 0 ≤ 𝑡 < 𝑡1
⎪
⎪
⎪
⎪𝑓2 (𝑡) 𝑡1 ≤ 𝑡 < 𝑡2
𝑓 (𝑡) = ⎨
⎪
⎪ ⋮
⎪
⎪
⎪
⎪ 𝑓 (𝑡) 𝑡𝑛−1 ≤ 𝑡
⎩𝑛

can be expresses as

𝑓 (𝑡) = (𝑢(𝑡 − 𝑡1 ) − 𝑢(𝑡))𝑓1 (𝑡) + (𝑢(𝑡 − 𝑡1 ) − 𝑢(𝑡 − 𝑡1 ))𝑓2 (𝑡) + ⋯ + 𝑢(𝑡 − 𝑡𝑛−1 )𝑓𝑛 (𝑡).

166 Lecture Notes for MA451

 9.2 Existence and Additional Properties*

Theorem 9.6 (𝑡-Shifting Theorem)

Let 𝑓 be a function such that (𝑓 (𝑡 + 𝑏))(𝑠) exist for 𝑠 > 𝑠0 . Then the Laplace transform of the
product function 𝑢(𝑡 − 𝑎)𝑓 (𝑡 − 𝑎), where 𝑢 is the unit step function, is

(𝑢(𝑡 − 𝑎)𝑓 (𝑡 − 𝑎)) = 𝑒 −𝑎𝑠 (𝑓 (𝑡))(𝑠).

♥

Proof By the definition of the unit step function,

∞ ∞

∫ 𝑢(𝑡 − 𝑎)𝑓 (𝑡 − 𝑎)𝑒 −𝑠𝑡 d 𝑡 = ∫ 𝑓 (𝑡 − 𝑎)𝑒 −𝑠𝑡 d 𝑡

0 𝑎
∞
−𝑎𝑠
=𝑒 ∫ 𝑓 (𝑡 − 𝑎)𝑒 −𝑠(𝑡−𝑎) d(𝑡 − 𝑎)
𝑎
∞
=𝑒 −𝑎𝑠 ∫ 𝑓 (𝑥)𝑒 −𝑠𝑥 d 𝑥
0
−𝑎𝑠
=𝑒 (𝑓 (𝑡))(𝑠).

■
Example 9.8 Find the Laplace transform

(𝑢(𝑡 − 𝑎)𝑓 (𝑡 + 𝑏))(𝑠).

Solution Let 𝑔(𝑡) = 𝑓 (𝑡 + 𝑎 + 𝑏). Then

(𝑢(𝑡 − 𝑎)𝑓 (𝑡 + 𝑏))(𝑠) = (𝑢(𝑡 − 𝑎)𝑔(𝑡 − 𝑎))(𝑠) = 𝑒 −𝑎𝑠 (𝑔(𝑡))(𝑠) = 𝑒 −𝑎𝑠 (𝑓 (𝑡 + 𝑎 + 𝑏))(𝑠).

■
Example 9.9 Find the Laplace transform

(𝑢(𝑡 − 1)𝑡 2 )(𝑠).

Solution Let 𝑔(𝑡) = (𝑡 + 1)2 = 𝑡 2 + 2𝑡 + 1. Then 𝑔(𝑡 − 1) = 𝑡 2 and

2 1 1
(𝑢(𝑡 − 1)𝑡 2 )(𝑠) = 𝑒 −𝑠 (𝑡 2 + 2𝑡 + 1)(𝑠) = 𝑒 −𝑠 ( + + .
𝑠3 𝑠2 𝑠 )
■
Example 9.10 Find the Laplace transform of the function

⎧
⎪
⎪𝑡 + 2, 0 ≤ 𝑡 < 2,
𝑓 (𝑡) = ⎨
⎪
⎪
⎩3𝑡, 𝑡 ≥ 2.

167 Lecture Notes for MA451

 9.2 Existence and Additional Properties*

Solution Using the unit step function 𝑢, the function 𝑓 can be written as

𝑓 (𝑡) = 𝑢(𝑡)(𝑡 + 2) − 𝑢(𝑡 − 2)(𝑡 + 2) + 3𝑡𝑢(𝑡 − 2) = 𝑢(𝑡)(𝑡 + 2) + 𝑢(𝑡 − 2)(2𝑡 − 2).

Then
(𝑓 (𝑡))(𝑠) =(𝑢(𝑡)(𝑡 + 2))(𝑠) + (𝑢(𝑡 − 2)(2𝑡 − 2))(𝑠)
=(𝑡 + 2)(𝑠) + 𝑒 −2𝑠 (2(𝑡 + 2) − 2)(𝑠)
=(𝑡 + 2)(𝑠) + 2𝑒 −2𝑠 (𝑡 + 1)(𝑠)
1 2 1 1
= ( 2 + ) + 2𝑒 −2𝑠 ( 2 + ) .
𝑠 𝑠 𝑠 𝑠
■

Transforms of integrals
For some differential equations, computing the Laplace transform of an integral may be
necessary.
Theorem 9.7
𝑡
Let 𝑓 be piecewise continuous function of exponential order 𝑠0 and 𝑔(𝑡) = ∫0 𝑓 (𝑥) d 𝑥. Then

1
(𝑔)(𝑠) = (𝑓 )(𝑠).
𝑠 ♥

Proof Note that 𝑔 ′ (𝑡) = 𝑓 (𝑡) and 𝑔(0) = 0 By the existence theorem and the theorem of
transform of derivative, we get

(𝑓 )(𝑠) = 𝑠(𝑔)(𝑠) − 𝑔(0) = 𝑠(𝑔)(𝑠)

which implies
1
(𝑔)(𝑠) = (𝑓 )(𝑠).
𝑠
■

Tables of Laplace Transforms

𝑛! 1
 (𝑡 𝑛 ) (𝑠) =  (𝑒 𝑎𝑡 𝑡 𝑛 ) (𝑠) =
𝑠 𝑛+1 (𝑠 − 𝑎)𝑛+1
𝑠−𝑎 𝑏
 (𝑒 𝑎𝑡 cos(𝑏𝑡)) (𝑠) = 2 2
 (𝑒 𝑎𝑡 sin(𝑏𝑡)) (𝑠) =
(𝑠 − 𝑎) + 𝑏 (𝑠 − 𝑎)2 + 𝑏 2
𝑠−𝑎 𝑏
 (𝑒 𝑎𝑡 cosh(𝑏𝑡)) (𝑠) = 2 2
 (𝑒 𝑎𝑡 sinh(𝑏𝑡)) (𝑠) = 2
(𝑠 − 𝑎) − 𝑏 𝑠 − 𝑏2
Table 9.1: Table of Basic Laplace Transforms

168 Lecture Notes for MA451

 9.2 Existence and Additional Properties*

Linearity (𝑐1 𝑓1 + 𝑐2 𝑓2 ) = 𝑐1 (𝑓1 ) + 𝑐2 (𝑓2 )

𝑠-shifting (𝑒 𝑎𝑡 𝑓 (𝑡))(𝑠) = (𝑓 (𝑡))(𝑠 − 𝑎)
𝑡-shifting (𝑢(𝑡 − 𝑎)𝑓 (𝑡 − 𝑎)) = 𝑒 −𝑎𝑠 (𝑓 (𝑡))(𝑠)
change of scale (𝑓 (𝑎𝑡))(𝑠) = 𝑎1 (𝑓 (𝑡)) ( 𝑎𝑡 )
Transform of 𝑓 ′ (𝑓 ′ )(𝑠) = (𝑓 )(𝑠) − 𝑓 (0)
Transform of 𝑓 ′′ (𝑓 ′′ )(𝑠) = (𝑓 )(𝑠) − 𝑓 ′ (0)𝑠 − 𝑓 (0)
𝑡 (𝑓 )(𝑠)
Transform of integral (∫0 𝑓 (𝑥) d 𝑥)(𝑠) =
𝑠
Table 9.2: Table of Rules of Laplace Transform

Derivatives of Transforms
Since Laplace transforms are analytic, the derivatives exists.
Theorem 9.8
Let 𝑓 be a piecewise continuous function of exponential order 𝑟. Then

d
(𝑓 )(𝑠) =  (−𝑡𝑓 (𝑡)) (𝑠), for 𝑠 > 𝑟
d𝑠 ♥

Proof Since 𝑓 is of exponential order, the improper integral converges absolutely. Hence, the
derivative and the integral are interchangeable:
∞
d
𝐹 ′ (𝑠) = 𝑓 (𝑡)𝑒 −𝑠𝑡 d 𝑡
d 𝑠 ∫0
∞
d
=∫ 𝑓 (𝑡)𝑒 −𝑠𝑡 d 𝑡
0 d𝑠
∞
=∫ −𝑡𝑓 (𝑡)𝑒 −𝑠𝑡 d 𝑡
0

=(−𝑡𝑓 (𝑡))(𝑠).

■
The above theorem holds true without the assumption that 𝑓 is of exponential order.
In this cae, one needs to express 𝐹 (𝑠) in terms of a transform of a bounded function 𝑔(𝑡) =
𝑡
∫0 𝑓 (𝑥)𝑒 −𝑠1 𝑥 d 𝑥. We refer the interested read to (Doetsch 1974, Chapter 6) for this generality.

Example 9.11 Find the Laplace transform

(𝑡 cos 𝑡).

169 Lecture Notes for MA451

 9.3 Inverse Laplace Transforms

Solution Applying the theorem of derivative of transform to 𝑓 (𝑡) = − cos 𝑡 yields

d d 1 2𝑠
(𝑡 cos 𝑡) = (− cos 𝑡) = − ( 2 ) = 2 .
d𝑠 d𝑠 𝑠 + 1 (𝑠 + 1)2

Integrals of Transforms
The integral of a Laplace transform can be calculated using the derivative.
Theorem 9.9
𝑓 (𝑡)
Let 𝑓 be a piecewise continuous function of exponential order 𝑟 such that lim𝑡→0+ 𝑡
exists.
Then ∞
𝑓 (𝑡)
∫ (𝑓 )(𝑥) d 𝑥 =  ( 𝑡 ) (𝑠).
𝑠
♥

Proof Applying the theorem of derivative of transform to  ( 𝑓 (𝑡)

𝑡 )
(𝑠) yields

d 𝑓 (𝑡) 𝑓 (𝑡)
 (𝑠) =  −𝑡 (𝑠) = −(𝑓 )(𝑠).
d𝑠 ( 𝑡 ) ( ( 𝑡 ))

Then by the fundamental theorem of calculus,

𝑠
𝑓 (𝑡)
 (𝑠) = − ∫ (𝑓 )(𝑥) d 𝑥.
( 𝑡 ) 𝑎

Since lim𝑠→∞ (𝑓 )(𝑠) = 0, taking 𝑎 = ∞ yields the equality

∞
𝑓 (𝑡)
 (𝑠) = ∫ (𝑓 )(𝑥) d 𝑥.
( 𝑡 ) 𝑠

■
Example 9.12 Find the Laplace transform

sin 𝑡
 .
( 𝑡 )

Solution
∞ ∞
sin 𝑡 1 𝜋 1
 = (sin 𝑡)(𝑠) d 𝑠 = ∫ d 𝑠 = − tan−1 (𝑠) = cot−1 (𝑠) = tan−1 ( ) .
( 𝑡 ) ∫𝑠 𝑠 𝑠2 +1 2 𝑠

170 Lecture Notes for MA451

 9.3 Inverse Laplace Transforms

9.3 Inverse Laplace Transforms

In order to solve differential equations using Laplace transforms, inverse Laplace trans-
forms must exist and be unique.
Definition 9.4
Given a function 𝐹 (𝑠), the inverse Laplace transform of 𝐹 , denoted by −1 (𝐹 ), is a function
𝑓 such that (𝑓 )(𝑠) = 𝐹 (𝑠).
♣
1 1
For example, −1 ( 𝑠−1 𝑡 𝑡
) = 𝑒 because (𝑒 )(𝑠) = 𝑠−1
.

The uniqueness in general may not be true because two functions with different discon-
tinuities may have the same Laplace transform. For example, (1)(𝑠) = (𝑢)(𝑠), where 𝑢 is the
unit step function.

Fortunately, discontinuities are the only differences that two functions can have if they
have the same Laplace transform. This result is known as Lerch’s theorem.
Theorem 9.10 (Lerch’s Theorem)
Let 𝑓 and 𝑔 be two functions having the same Laplace transform: (𝑓 )(𝑠) = (𝑔)(𝑠). If 𝑓
and 𝑔 are piecewise continuous, then 𝑓 (𝑡) = 𝑔(𝑡) for any 𝑡 where 𝑓 and 𝑔 are continuous. In
particular, if 𝑓 and 𝑔 are continuous, then 𝑓 = 𝑔.
♥

We refer the reader to (Doetsch 1974, Chapter 5) for a proof.

Properties of inverse Laplace transforms

Like calculating antiderivatives, some inverse Laplace tranforms can be calculated using
those basic Laplace transforms and properties of inverse Laplace transforms.

By the uniqueness theorem and linearity of Laplace transform, we see that the inverse
Laplace transform is also a linear operator.
Theorem 9.11
Suppose the inverse Laplace transforms for 𝐹 (𝑠) and 𝐺(𝑠) exist. Then

−1 (𝑐1 𝐹1 (𝑠) + 𝑐2 𝐹2 (𝑠)) = 𝑐1 −1 (𝐹1 (𝑠)) + 𝑐2 −1 (𝐹2 (𝑠)).
♥

Example 9.13 Find the inverse transform

1 1
−1 ( +
𝑠 − 1 𝑠 − 2)

171 Lecture Notes for MA451

 9.3 Inverse Laplace Transforms

Solution
1 1 1 1
−1 ( + ) =−1 ( ) + −1 (
𝑠−1 𝑠−2 𝑠−1 𝑠 − 2)
𝑡 2𝑡
=𝑒 + 𝑒
■
 Exercise 9.5 Find the inverse transform
1 2
−1 ( + 2 )
𝑠 𝑠

Solution
1 2 1 2
−1 ( + 2 ) =−1 ( ) + −1 ( 2 )
𝑠 𝑠 𝑠 𝑠
=1 + 2𝑡.
■
Theorem 9.12 (Shifting Inverse Transforms)
If 𝑓 (𝑡) = −1 (𝐹 (𝑠)), then
1.
−1 (𝐹 (𝑠 − 𝑎)) = 𝑒 𝑎𝑡 𝑓 (𝑡),

2.
−1 (𝑒 −𝑎𝑠 𝐹 (𝑠)) = 𝑢(𝑡 − 𝑎)𝑓 (𝑡 − 𝑎).
♥

Example 9.14 Find the inverse transform

1
−1 .
( (𝑠 + 1)2 )

Solution
1 1
−1 = 𝑒 −𝑡 −1 ( 2 ) = 𝑡𝑒 −𝑡 .
( (𝑠 + 1) )
2 𝑠
■
Example 9.15 Find
𝑠
−1 ( .
𝑠2 + 2𝑠 + 2 )

Solution
𝑠 𝑠
−1 ( ) =−1
𝑠2 + 2𝑠 + 2 ( (𝑠 + 1)2 + 1 )
𝑠+1 1
=−1 − −1
( (𝑠 + 1) + 1 )
2 ( (𝑠 + 1)2 + 1 )
𝑠 1
=𝑒 −𝑡 −1 ( 2 ) − 𝑒 −𝑡 −1 ( 2
𝑠 +1 𝑠 + 1)
=𝑒 −𝑡 cos(𝑠𝑡) − 𝑒 −𝑡 sin(𝑠𝑡).

172 Lecture Notes for MA451

 9.3 Inverse Laplace Transforms

The method of partial fractional decomposition

Using the Laplace transform to solve differential equations often requires finding the in-
verse transform of a rational function
𝑃(𝑠)
𝐹 (𝑠) = .
𝑄(𝑠)

𝑃(𝑠)
By the fundamental theorem of algebra, a rational expression 𝑄(𝑠) , where 𝑃(𝑠) and 𝑄(𝑠) are
polynomials and deg 𝑃(𝑠) < deg 𝑄(𝑠), can always be written as the sum of rational expressions
whose denominator are powers of linear or irreducible quadratic polynomials.
Theorem 9.13
Let 𝑃 and 𝑄 are nonzero polynomials. Assume that deg 𝑃 < deg 𝑄. Write 𝑄 as a produce of
powers of distinct irreducible polynomials of degree at most two:
𝑛
𝑄(𝑠) = ∏ 𝑝𝑖𝑛𝑖 .
𝑖

Then there are unique polynomials 𝑎𝑖𝑗 with deg 𝑎𝑖𝑗 < deg 𝑝𝑖 such that
𝑛 𝑛𝑖
𝑃 𝑎𝑖𝑗
= ∑∑ 𝑗 .
𝑄 𝑖 𝑗 𝑝𝑖
♥

Example 9.16 Find the inverse Laplace transform of

𝑠
𝐹 (𝑠) = .
𝑠2 − 𝑠 − 2

Solution Factoring the denominator yeilds

𝑠
𝐹 (𝑠) = .
(𝑠 + 1)(𝑠 − 2)

By the partial fraction decomposition theorem, 𝐹 has an expression

𝐴 𝐵
𝐹 (𝑠) = + .
𝑠+1 𝑠−2
Note that
𝐵(𝑠 + 1) 𝑠 𝐵(𝑠 + 1)
𝐴 = (𝑠 + 1)𝐹 (𝑠) − = − ,
𝑠−2 𝑠−2 𝑠−2
𝐴(𝑠 − 2) 𝑠 𝐴(𝑠 − 2)
𝐵 = (𝑠 − 2)𝐹 (𝑠) − = − .
𝑠 + 1) 𝑠+1 𝑠 + 1)

173 Lecture Notes for MA451

 9.3 Inverse Laplace Transforms

Because the equities should hold true for all 𝑠. Taking 𝑠 = −1 implies

−1 1
𝐴= = .
−1 − 2 3
Taking 𝑠 = 2 implies
2 2
𝐵= = .
2+1 3
Therefore,
1 1 2 1
𝐹 (𝑠) = + .
3𝑠 +1 3𝑠 −2
By linearity,
1 1 2 1 1 2
−1 (𝐹 (𝑠)) = −1 ( ) + −1 ( ) = 𝑒 −𝑡 + 𝑒 2𝑡 .
3 𝑠+1 3 𝑠−2 3 3
■
Example 9.17 Find the inverse transform of
1
𝐹 (𝑠) = .
𝑠(𝑠 − 1)2

Solution The function 𝐹 can be decomposed at

𝐴 𝐵 𝐶
𝐹 (𝑠) = + + .
𝑠 𝑠 − 1 (𝑠 − 1)2

Applying the evaluation method as in the above example yields

| 1 |
𝐴 = 𝑠𝐹 (𝑠)| = | = 1,
|𝑠=0 (𝑠 − 1)2 |𝑠=0

| 1|
𝐶 = (𝑠 − 1)2 𝐹 (𝑠)| = | = 1.
|𝑠=1 𝑠 |𝑠=1
𝐶
Subtracting (𝑠−1)2
from the decomposition and then applying the evaluation method yields

1 | 1−𝑠 | 1|
𝐵 = ((𝑠 − 1)𝐹 (𝑠) − ) | = | = − | = −1.
𝑠−1 | 𝑠=1 𝑠(𝑠 − 1) | 𝑠=1 𝑠 |𝑠=1

Therefore,
1 1 1
𝐹 (𝑠) = − + .
𝑠 𝑠 − 1 (𝑠 − 1)2
The inverse transform is
−1 (𝐹 (𝑠)) = 1 − 𝑒 𝑡 − 𝑡𝑒 𝑡 .

174 Lecture Notes for MA451

 9.3 Inverse Laplace Transforms

Example 9.18 Find the inverse Laplace transform of

𝑠2 + 1
𝐹 (𝑠) =
𝑠(𝑠 2 + 2𝑠 + 2)

Solution The function 𝐹 can be written as the decomposition

𝐴 𝐵𝑠 + 𝐶
𝐹 (𝑠) = + 2 .
𝑠 𝑠 + 2𝑠 + 2
Then
| 𝑠2 + 1 | 1
𝐴 = 𝑠𝐹 (𝑠)| = 2 | = .
|𝑠=0 𝑠 + 2𝑠 + 2 |𝑠=0 2
Note that 𝑠 2 + 2𝑠 + 2 = (𝑠 + 1)2 + 1 which has two roots 𝑠 = −1 ± i. Then

| 𝑠2 + 1 | 3 1
𝐵(−1 + i) + 𝐶 = (𝑠 2 + 2𝑠 + 2)𝐹 (𝑠)| = | =− + i (1)
|𝑠=−1+i 𝑠 |𝑠=−1+i 2 2

| 𝑠2 + 1 | 3 1
𝐵(−1 − i) + 𝐶 = (𝑠 2 + 2𝑠 + 2)𝐹 (𝑠)| = | = − + i. (2)
|𝑠=−1−i 𝑠 |𝑠=−1−i 2 2
Solving the liner system implies
1
𝐵= 𝐶 = −1.
2
Therefore,
1 1 1 𝑠−2
𝐹 (𝑠) = ⋅ + .
2 𝑠 2 (𝑠 + 1)2 + 1
To get the inverse transform using the table, the second term needs to be rewritten as

1 𝑠−2 1 𝑠+1 3 1
= − ⋅ .
2 (𝑠 + 1) + 1 2 (𝑠 + 1) + 1 2 (𝑠 + 1)2 + 1
2 2

Then
1 1 1 𝑠+1 3 1
−1 (𝐹 (𝑠)) = −1 ( ) + −1 − −1
2 𝑠 2 ( (𝑠 + 1) + 1 ) 2
2 ( (𝑠 + 1)2 + 1 )
1 1 3
= + 𝑒 −𝑡 cos(𝑡) − 𝑒 −𝑡 sin(𝑡).
2 2 2
■

175 Lecture Notes for MA451

 Week 10: Solving Differential Equations using
Laplace Transforms

11/15–11/18

10.1 Solving IVP using Laplace Transforms

Consider the initial value problem

𝑎𝑦 ′′ + 𝑏𝑦 ′ + 𝑐𝑦 = 𝑓 (𝑡), 𝑦(0) = 𝑦0 , 𝑦 ′ (0) = 𝑦1 .

Applying the Laplace transform to both sides yields

(𝑎𝑦 ′′ + 𝑏𝑦 ′ + 𝑐𝑦)(𝑠) =(𝑓 )(𝑠)

𝑎(𝑦 ′′ )(𝑠) + 𝑏(𝑦 ′ )(𝑠) + 𝑐(𝑦)(𝑠) =(𝑓 )(𝑠)
𝑎[𝑠 2 (𝑦)(𝑠) − 𝑦0 𝑠 − 𝑦1 ] + 𝑏[𝑠(𝑦)(𝑠) − 𝑦(0)] + 𝑐(𝑦)(𝑠) =(𝑓 )(𝑠)
(𝑎𝑠 2 + 𝑏𝑠 + 𝑐)(𝑦)(𝑠) − [𝑎𝑦0 𝑠 − 𝑎𝑦1 − 𝑏𝑦0 ] =(𝑓 )(𝑠)
(𝑓 )(𝑠) + 𝑎𝑦0 𝑠 + 𝑎𝑦1 + 𝑏𝑦0
(𝑦)(𝑠) = .
𝑎𝑠 2 + 𝑏𝑠 + 𝑐

Applying the Inverse Laplace transform yields

(𝑓 )(𝑠) + 𝑎𝑦0 𝑠 + 𝑎𝑦1 + 𝑏𝑦0

𝑦(𝑡) = −1 .
( 𝑎𝑠 2 + 𝑏𝑠 + 𝑐 )

Example 10.1 Use the Laplace transform to solve the initial value problem

𝑦 ′′ − 6𝑦 ′ + 5𝑦 = 3𝑒 2𝑡 , 𝑦(0) = 2, 𝑦 ′ (0) = 3.

Solution Using the formulas

(𝑦 ′ )(𝑠) = 𝑠(𝑦) − 𝑦(0),

(𝑦 ′′ )(𝑠) = 𝑠 2 (𝑦) − 𝑠𝑦(0) − 𝑦 ′ (0),

and applying the Laplace transform to both sides yields

3
(𝑠 2 − 6𝑠 + 5)(𝑦) = + 2(𝑠 − 6) + 3.
𝑠−2
 10.1 Solving IVP using Laplace Transforms

Therefore,
3 2𝑠 − 9
(𝑦) = + .
(𝑠 − 2)(𝑠 − 1)(𝑠 − 5) (𝑠 − 1)(𝑠 − 5)
Applying the partial fractional decomposition method to the right hand side yields

3 2𝑠 − 9 5 1 1 1 1
+ = ⋅ − + ⋅ .
(𝑠 − 2)(𝑠 − 1)(𝑠 − 5) (𝑠 − 1)(𝑠 − 5) 2 𝑠 − 1 𝑠 − 2 2 𝑠 − 5

Therefore,
5 1 1 1 1
𝑦 =−1 ( ⋅ − + ⋅
2 𝑠 − 1 𝑠 − 2 2 𝑠 − 5)
𝑡
5𝑒 2𝑡 𝑒 5𝑡
= −𝑒 + .
2 2
■
Example 10.2 Solve the initial value problem

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

Solution Applying the Laplace transform to the equation and solve for (𝑦) implies

3 𝑠+3
(𝑦) = + .
𝑠(𝑠 2 + 2𝑠 + 2) 𝑠 2 + 2𝑠 + 2

The partial fractional decomposition is

3 𝑠+1 3 1 1 𝑠+1 3 1
+ 2 = ⋅ − ⋅ − ⋅ .
𝑠(𝑠 2 + 2𝑠 + 2) 𝑠 + 2𝑠 + 2 2 𝑠 2 (𝑠 + 1) + 1 2 (𝑠 + 1)2 + 1
2

Therefore,
3 1 1 𝑠+1 1 1
𝑦 =−1 ⋅ − ⋅ + ⋅
( 2 𝑠 2 (𝑠 + 1) + 1 2 (𝑠 + 1)2 + 1 )
2

2 1 1
= − 𝑒 −𝑡 cos 𝑡 + 𝑒 −𝑡 sin 𝑡.
3 2 2
■
 Exercise 10.1 Solve the initial value problem

𝑦 ′′ + 𝑦 = 0, 𝑦(0) = 0, 𝑦 ′ (0) = 1.

Solution Applying the Laplace transform to the equation yields

1
(𝑦) = .
𝑠2 + 1
Then
1
𝑦 = −1 ( = sin 𝑡.
𝑠2 + 1)

177 Lecture Notes for MA451

 10.2 Convolutions*

■
 Exercise 10.2 Solve the initial value problem

𝑦 ′′ + 25𝑦 = 9𝑒 𝑡 , 𝑦(0) = 0, 𝑦 ′ (0) = 2.

Solution Applying the Laplace transform to the equation yields

1 9
(𝑦) = ⋅( + 2) .
𝑠2 + 25 𝑠−1
The partial fractional decomposition of the right hand side is

9 2 9 𝑠 43 5 9 1
+ = − ⋅ + ⋅ + ⋅ .
(𝑠 − 1)(𝑠 2 + 25) 𝑠 + 25 26 𝑠 2 + 25 130 𝑠 2 + 25 26 𝑠 − 1

Then
9 𝑠 43 5 9 1
𝑦 =−1 (− ⋅ 2 + ⋅ 2 + ⋅
26 𝑠 + 25 5 𝑠 + 25 26 𝑠 − 1 )
9 43 9
=− cos(5𝑡) + sin(5𝑡) + 𝑒 𝑡 .
26 130 26
■

10.2 Convolutions*
Let 𝐹 (𝑠) = (𝑓 (𝑡))(𝑠) and 𝐺(𝑠) = (𝑔(𝑡))(𝑠). The inverse transform −1 (𝐹 (𝑠)𝐺(𝑠)) can be
calculated by an integral of 𝑓 and 𝑔.
Theorem 10.1
Let 𝐹 (𝑠) = (𝑓 (𝑡))(𝑠) and 𝐺(𝑠) = (𝑔(𝑡))(𝑠). Then
𝑡
−1
 (𝐹 (𝑠)𝐺(𝑠)) = ∫ 𝑓 (𝑡 − 𝑥)𝑔(𝑥)d𝑥.
0
♥

Proof Since changing the dummy variable of integration won’t change the integral, the func-
tion 𝐹 (𝑠)𝐺(𝑠) is determined by
∞ ∞
−𝑠𝑥
𝐹 (𝑠)𝐺(𝑠) = ∫ 𝑒 𝑓 (𝑥)d𝑥 ∫ 𝑒 −𝑠𝑡 𝑔(𝑦)d𝑦.
0 0

By Fubini’s theorem, the order of integration in this case is interchangeable. Together with

178 Lecture Notes for MA451

 10.2 Convolutions*

change of coordinates, it implies that

∞ ∞ ∞ ∞

∫ 𝑒 −𝑠𝑥 𝑓 (𝑥)d𝑥 ∫ 𝑒 −𝑠𝑦 𝑔(𝑦)d𝑦 = ∫ ∫ 𝑒 −𝑠(𝑥+𝑦) 𝑓 (𝑥)𝑔(𝑦)d𝑥d𝑦

0 0 0 0
∞ ∞
=∫ 𝑒 −𝑠(𝑥+𝑦) 𝑓 (𝑥)d𝑥 𝑔(𝑦)d𝑦
0 (∫0 )
∞ ∞
=∫ 𝑒 −𝑠𝑢 𝑓 (𝑢 − 𝑦)d𝑢 𝑔(𝑦)d𝑦 where 𝑢 = 𝑥 + 𝑦
0 (∫𝑦 )
∞ ∞
=∫ ∫ (𝑒 −𝑠𝑥 𝑓 (𝑥 − 𝑦)𝑔(𝑦)) d𝑥d𝑦
0 𝑦
∞ 𝑥
=∫ ∫ (𝑒 −𝑠𝑥 𝑓 (𝑥 − 𝑦)𝑔(𝑦)) d𝑦d𝑥 change of coordinates
0 0
∞ 𝑥
=∫ 𝑒 −𝑠(𝑥) ∫ 𝑓 (𝑥 − 𝑦)𝑔(𝑦)d𝑦 d𝑥
0 ( 0 )
𝑦
= ∫ 𝑓 (𝑥 − 𝑦)𝑔(𝑦)d𝑥 (𝑠).
( 0 )

That completes the proof. ■

This theorem motivates the definition of convolution.
Definition 10.1
The convolution 𝑓 ∗ 𝑔 of two functions 𝑓 and 𝑔 is defined by
𝑡
(𝑓 ∗ 𝑔)(𝑡) = ∫ 𝑓 (𝑥)𝑔(𝑡 − 𝑥)d𝑥.
0
♣

Applying a substitution 𝑢 = 𝑡 − 𝑥 and changing the dummy variable 𝑢 back to 𝑥 implies

𝑡 0 𝑡
(𝑓 ∗ 𝑔)(𝑡) = ∫ 𝑓 (𝑥)𝑔(𝑡 − 𝑥)d𝑥 = − ∫ 𝑓 (𝑡 − 𝑢)𝑔(𝑢)d𝑢 = ∫ 𝑔(𝑥)𝑓 (𝑡 − 𝑥)d𝑥 = (𝑔 ∗ 𝑓 )(𝑡).
0 𝑡 0

Example 10.3 Let 𝑓 (𝑡) = 𝑒 𝑎𝑡 and 𝑔(𝑡) = 𝑒 𝑏𝑡 . Find (𝑓 ∗ 𝑔).

Solution By the theorem of convolution,

1
(𝑓 ∗ 𝑔) = (𝑒 𝑎𝑡 )(𝑒 𝑏𝑡 ) = .
(𝑠 − 𝑎)(𝑠 − 𝑏)

■
1
Example 10.4 Find −1 ( 𝑠(𝑠−1) 2 ).

Solution From the table of Laplace transform 1𝑠 = (1)(𝑠) and 1

𝑠2
= (𝑡)(𝑠). By the 𝑠-shifting
theorem,
1
2
= (𝑡)(𝑠 − 1) = (𝑡𝑒 𝑡 )(𝑠).
(𝑠 − 1)

179 Lecture Notes for MA451

 10.3 Dirac Delta Function*

Therefore, by the theorem of convolution,

𝑡
1
−1 = (1) ∗ (𝑡𝑒 𝑡
) = 𝑥 𝑥 𝑥
∫ 1 ∗ 𝑥𝑒 d𝑥 = 𝑥𝑒 − 𝑒 + 1.
( 𝑠(𝑠 − 1)2 ) 0

■
 Exercise 10.3 Find −1 ( 𝑠2 (𝑠12 +1) ).
1
Solution Note that 𝑠2
= (𝑡)(𝑠) and (sin(𝑡))(𝑠). Therefore,

1 𝑡 |𝑡 |𝑡
−1 = (𝑡 − 𝑥) sin 𝑥d𝑥 = ((𝑥 − 𝑡) cos 𝑥)| − sin 𝑥 | = 𝑡 − sin 𝑡.
( 𝑠 2 (𝑠 2 + 1) ) ∫0 | |
|0 |0

10.3 Dirac Delta Function*

In real life, there are situations that the systems acted by sudden external forces. The
external forces may or may not apply for a period. For example, an electrical surge caused by
a lightning strike. In general terms, there is an impulse force whose magnitude can be infinitely
large.

In Mathematics, an idea to understand an impulse force is to approximate the force func-

tion by a piecewise continuous function.

Consider the function unit step function

⎧
⎪
⎪1 for 0 < 𝑡 ≤ ℎ
𝛿ℎ (𝑡) = ⎨ ℎ
⎪
⎪0 otherwise,
⎩
1
which represents a force of a magnitude ℎ
during the time period 0 to ℎ.

The limit of this function 𝛿ℎ (𝑡) as ℎ goes to 0, denoted as 𝛿(𝑡), is called the Dirac delta
function (or simply delta function, or unit impulse function). That is
⎧
⎪
⎪∞ 𝑡 = 0
𝛿(𝑡) = lim 𝛿ℎ (𝑡) = ⎨
ℎ→0 ⎪
⎪ 0 otherwise.
⎩

For any real number 𝑥, we define

⎧
⎪
⎪∞ 𝑡 = 𝑥
𝛿(𝑡 − 𝑥) = lim 𝛿ℎ (𝑡 − 𝑥) = ⎨
ℎ→0 ⎪
⎪0 otherwise.
⎩

180 Lecture Notes for MA451

 10.3 Dirac Delta Function*

Remark In defining 𝛿(𝑡) as a limit, the shape of the impulse sequence 𝛿ℎ (𝑡) turns out irrelevant.
For example, Dirac delta function 𝛿(𝑡) can also be defined as the limit of the sequence of
Gaussian functions:
1 𝑥
𝛿(𝑡) = lim √ 𝑒 − ℎ2 .
ℎ→0 ℎ 𝜋

The interested reader is referred to (Bracewell 1999, Chapter 5) for more details.

The Dirac delta function is not a function in the traditional sense as 𝛿(0) = ∞. It is a
generalized function in the sense that it is function from the space of functions to the real
line.

Note that
∞ ℎ

∫ 𝛿ℎ (𝑡) d 𝑡 = ∫ 𝛿ℎ (𝑡) d 𝑡 = 1
−∞ 0

which yields
∞

∫ 𝛿(𝑡) d 𝑡 = 1.
−∞

∞
Note that ∫−∞ lim 𝛿ℎ (𝑡)𝑓 (𝑡) d 𝑡 = 0 for any function. The integral is not meaningful if it
ℎ→0
𝑡
is understood in this way. So the integral ∫−∞ 𝛿(𝑥 − 𝑎) d 𝑥 really should be understood as
𝑡
lim ∫−∞ 𝛿𝑏 (𝑥 − 𝑎) d 𝑥. After all, the function 𝛿(𝑡) is not a real function, the meaning of the in-
𝑏→0
tegration involving 𝛿(𝑡) should be generalized too. In the rest of the section, we shall interpret
operations involving Dirac delta function, such as multiplication, integration, and differentia-
tion in the way of performing the operation first and then taking the limit. This interpretation
of operations involving 𝛿(𝑡) is compatible with other approaches of Dirac delta function. One
of such approaches is to define the Dirac delta function as the derivative of the unit step
function in the following sense:
𝑡
d d
𝛿(𝑡 − 𝑎) = ∫ 𝛿(𝑥 − 𝑎) d 𝑥 = (𝑢(𝑡 − 𝑎)) = 𝑢 ′ (𝑡 − 𝑎).
d𝑡 ( −∞ ) d𝑡

This integration property together with 𝛿(𝑡) = 0 for 𝑡 ≠ 0 gives the formal characterization
of Dirac delta function, that is, the Dirac delta function is a generalized function (or distribu-
tion) with the following properties:

1. 𝛿(𝑡) = 0 if 𝑡 ≠ 0, and
∞
2. ∫−∞ 𝛿(𝑡) d 𝑡 = 1, where the integral may be understand as the Riemann–Stieltjes integral.

An important property which highlights the viewpoint of 𝛿(𝑡) being a function over the
space of functions.

181 Lecture Notes for MA451

 10.3 Dirac Delta Function*

Theorem 10.2 (Sifting Property)

For any function 𝑓 continuous near 𝑥,

𝑏
⎧
⎪
⎪𝑓 (𝑥) for 𝑥 ∈ (𝑎, 𝑏)
∫ 𝛿(𝑡 − 𝑥)𝑓 (𝑡) d 𝑡 = ⎨
𝑎 ⎪
⎪ 0 for 𝑥 ∉ [𝑎, 𝑏].
⎩ ♥

Proof If 𝑥 ∉ [𝑎, 𝑏], then 𝛿(𝑡 − 𝑥) = 0 on [𝑎, 𝑏]. So is the integral. Suppose that 𝑥 is in (𝑎, 𝑏). Using
the fundamental theorem of calculus, we get
𝛿 𝑏

∫ (𝑡 − 𝑥)𝑓 (𝑡) d 𝑡 = ℎ→0

lim ∫ 𝛿ℎ (𝑡 − 𝑥)𝑓 (𝑡) d 𝑡
𝑎 𝑎
𝑥+ℎ
1
= lim ∫ 𝑓 (𝑡) d 𝑡
ℎ→0 𝑥 ℎ
𝑥+ℎ
∫𝑥 𝑓 (𝑡) d 𝑡
= lim
ℎ→0 ℎ
=𝑓 (𝑥).

■
As a consequence, for any 𝑎 > 0, the Laplace transform of 𝛿(𝑡 − 𝑎) is
∞
(𝛿(𝑡 − 𝑎))(𝑠) = ∫ 𝛿(𝑡 − 𝑎)𝑒 −𝑠𝑡 d 𝑡 = 𝑒 −𝑎𝑠 .
0

In particular,
(𝛿(𝑡))(𝑠) = 1.

The Dirac delta function can be used to solve the following initial value problem

𝑎𝑦 ′′ + 𝑏𝑦 ′ + 𝑐𝑦 = 𝑓 (𝑡), 𝑦(0) = 0, 𝑦 ′ (0) = 0,

where 𝑎 > 0. Applying the Laplace transform implies

(𝑓 )(𝑠)
(𝑦)(𝑠) = .
𝑎𝑠 2 + 𝑏𝑠 + 𝑐
1
Write the inverse transform of 𝑎𝑠 + 𝑏𝑠+𝑐
as

1
𝑤(𝑡) = −1 ( ,
𝑎𝑠 + 𝑏𝑠 + 𝑐)

which is called the unit impulse response (or weight function). Then applying the theorem of
convolution yields
𝑡
1
𝑦 = −1 ((𝑓 (𝑡))(𝑠) ⋅ = 𝑓 (𝑡) ∗ 𝑤(𝑡) = ∫ 𝑓 (𝑡 − 𝑥)𝑤(𝑥) d 𝑥.
𝑎𝑠 2 + 𝑏𝑠 + 𝑐 ) 0

182 Lecture Notes for MA451

 10.3 Dirac Delta Function*

Because of the commutativity of convolution, the solution 𝑦 can also be calculated by

𝑡
𝑦=∫ 𝑓 (𝑥)𝑤(𝑡 − 𝑥) d 𝑥.
0

In particular, if 𝑓 (𝑡) = 𝛿(𝑡), then the function 𝑤(𝑡) is nothing but the solution of the initial
value problem
𝑎𝑦 ′′ + 𝑏𝑦 ′ + 𝑐𝑦 = 𝛿(𝑡), 𝑦(0) = 0, 𝑦 ′ (0) = 0.

This is the reason why it is called the unit impulse response.

Example 10.5 Solve the initial value problem

𝑦 ′′ + 𝑦 ′ − 2𝑦 = 3𝑒 5𝑡 , 𝑦(0) = 0, 𝑦 ′ (0) = 0.

Solution The unit impulse function is determined by

1
𝑤(𝑡) =−1 (
+ 𝑠 − 2)
𝑠2
1 1 1
= −1 ( −
3 𝑠 − 1 𝑠 + 2)
1
= (𝑒 𝑡 − 𝑒 −2𝑡 ).
3

Then the solution of the equation is

𝑡
𝑦 = ∫ (𝑒 𝑥 − 𝑒 −2𝑥 )𝑒 5(𝑡−𝑥) d 𝑥
0
𝑡
=𝑒 5𝑡 ∫ (𝑒 −4𝑥 − 𝑒 −7𝑥 ) d 𝑥
0
𝑒 −4𝑥 𝑒 −7𝑥 𝑡
=𝑒 5𝑡 − + |
( 4 7 )0
𝑒 −4𝑡 𝑒 −7𝑡 3
=𝑒 5𝑡 − + −
( 4 7 28 )
−2𝑡 𝑡 5𝑡
𝑒 𝑒 3𝑒
= − − .
7 4 28
■

183 Lecture Notes for MA451

 Week 11: Linear System of Equations

11/22–11/30

11.1 Basics of Linear Algebra

A system of linear equations (or simply a linear system) consists of 𝑚 linear equations in
𝑛 variables:
𝑎11 𝑥1 + 𝑎12 𝑥2 + ⋯ 𝑎1𝑛 𝑥𝑛 =𝑏1
𝑎21 𝑥1 + 𝑎22 𝑥2 + ⋯ 𝑎2𝑛 𝑥𝑛 =𝑏2
(11.1)
⋮
𝑎𝑚1 𝑥1 + 𝑎𝑚2 𝑥2 + ⋯ 𝑎𝑚𝑛 𝑥𝑛 =𝑏𝑚 .

Linear algebra provides an operational way to solve system of linear equations. Here, the
operators are matrix operations.

Matrices
An 𝑚 × 𝑛 matrix 𝐴 is a number array of the form

⎛𝑎11 𝑎12 ⋯ 𝑎1𝑛 ⎞

⎜ ⎟
⎜𝑎21 𝑎22 ⋯ 𝑎2𝑛 ⎟
𝐴=⎜ ⎟.
⎜ ⋮ ⋮ ⋱ ⋮ ⎟
⎜𝑎𝑛1 𝑎𝑛2 ⋯ 𝑎𝑛𝑛 ⎟
⎝ ⎠

The element in the 𝑖-th row and 𝑗-th column is called the 𝑖𝑗-th element (or 𝑖𝑗-th entry).

For simplicity, a matrix is also denoted by its 𝑖𝑗-th element, for example, the matrix 𝐴 can
be denote by 𝐴 = (𝑎𝑖𝑗 ).

An 𝑚 × 1 matrix is called a column vector. A 1 × 𝑛 matrix is called a row vector.

Example 11.1 The matrix

⎛ 𝑥1 ⎞
⎜ ⎟
⎜ 𝑥2 ⎟
𝑋 =⎜ ⎟
⎜⋮⎟
⎜𝑥𝑛 ⎟
⎝ ⎠
is a column vector.

Two matrices 𝐴 = (𝑎𝑖𝑗 ) and 𝐵 = (𝑏𝑖𝑗 ) are said to be equal if 𝑎𝑖𝑗 = 𝑏𝑖𝑗 for 1 ≤ 𝑖 ≤ 𝑚 and
 11.1 Basics of Linear Algebra

1 ≤ 𝑗 ≤ 𝑛.

A matrix whose element are all equal to 0 is denoted by 0.

Using matrices, the linear system (11.1) can be written as

𝐴𝑋 = 0.

The scalar product of a constant 𝑘 and a matrix 𝐴 = (𝑎𝑖𝑗 ) is defined as

𝑘𝐴 = (𝑘𝑎𝑖𝑗 ) .

The sum of two matrices 𝐴 = (𝑎𝑖𝑗 ) and 𝐵 = (𝑏𝑖𝑗 ) is defined as

𝐴 + 𝐵 = (𝑎𝑖𝑗 + 𝑏𝑖𝑗 )

The produce 𝐴𝐵 of two matrices 𝐴 = (𝑎𝑖𝑗 ) and 𝐵 = (𝑏𝑖𝑗 ) is defined if the number of
columns of 𝐴 equals the number of rows of 𝐵. If 𝐴 is a 𝑚 × 𝑝 matrix and 𝐵 is an 𝑝 × 𝑛 matrix,
then the product 𝐴𝐵 is an 𝑚 × 𝑛 matrix defined as
𝑛
𝐴𝐵 = ∑ 𝑎𝑖𝑘 𝑏𝑘𝑗 .
(𝑘=1 )

Note that in general 𝐴𝐵 ≠ 𝐵𝐴.

A square matrix is a matrix with the same number of rows and column. The number of
rows or columns of a square matrix is called the order of the matrix.

The diagonal elements of a square matrix 𝐴 = (𝑎𝑖𝑗 ) are the elements 𝑎𝑖𝑗 with 𝑖 = 𝑗.

A diagonal matrix is a square matrix whose non-diagonal elements are all equal to 0.

An identity matrix, denoted by 𝐼 , is a diagonal matrix whose diagonal elements are all
equal to 1. It is a straightforward result that 𝐼 𝐴 = 𝐴 = 𝐴𝐼 for 𝐴 and 𝐼 have the same order.

Determinant

The determinant (𝐴) of a 2 × 2 matrix

𝑎11 𝑎12
𝐴=
(𝑎21 𝑎22 )

185 Lecture Notes for MA451

 11.1 Basics of Linear Algebra

is defined as
det (𝐴) = 𝑎11 𝑎22 − 𝑎12 𝑎21 .

Geometrically, the determinant represents the signed area of the parallelograms gener-
ated by the column vectors of the matrix. In higher dimension, it represents the signed volume
of the parallelepiped.

The determinant det (𝐴𝑖𝑗 ) of the (𝑛 − 1) × (𝑛 − 1) matrix 𝐴𝑖𝑗 obtained from a 𝑛 × 𝑛 matrix 𝐴
by removing the 𝑝-th row and the 𝑘-th column is called the 𝑖𝑗-th minor of 𝐴.

The number 𝐶𝑖𝑗 = (−1)𝑖+𝑗 det (𝐴𝑖𝑗 ) is called the cofactor of the element 𝑎𝑖𝑗 of 𝐴.

The determinant det (𝐴) of a 𝑛 × 𝑛 matrix 𝐴 = (𝑎𝑖𝑗 ) is defined as

𝑛
det (𝐴) = ∑ 𝑎1𝑘 𝐶1𝑘 .
𝑘=1

It is a fundamental result in linear algebra that

𝑛
det (𝐴) = ∑ 𝑎𝑖𝑘 𝐶𝑖𝑘
𝑘=1

for any 1 ≤ 𝑖 ≤ 𝑛.

A square matrix 𝐴 is said to be nonsingular if the determinant is nonzero. Otherwise, it is

said to be singular.

Example 11.2 Find the determinant of the matrix

1 2
.
(3 1)

Solution From the definition, the determinant is

1 2
det = 1 ⋅ 1 − 2 ⋅ 3 = 1 − 6 = −5.
(3 1)

Inverse Matrices
A square matrix 𝐵 is called an inverse of a square matrix 𝐴 if it satisfies both

𝐴𝐵 = 𝐼 and 𝐵𝐴 = 𝐼 .

186 Lecture Notes for MA451

 11.1 Basics of Linear Algebra

A square matrix is called invertible if it has an inverse.

A main result about matrix inverse is the following theorem.

Theorem 11.1
Let 𝐴 be a square matrix.
1. 𝐴 is invertible if and only if 𝐴 is non-singular.
2. if 𝐴 is invertible, then its inverse is unique, which is usually denoted by 𝐴−1 .
♥

Example 11.3 Determine if the following matrix is invertible:

3 6
𝐴= .
(2 4)

Solution The determinant of the matrix 𝐴 is

3 6
det = 3 ⋅ 4 − 2 ⋅ 6 = 0.
(2 4)

So the matrix is not invertible.

Here is another way to see that it not invertible. Note that

3 6 2 0
= .
(2 4) (−1) (0)

If the matrix 𝐴 is invertible, then multiplying the inverse 𝐴−1 to both sides yields

2 0
=
(−1) (0)

which is a contradiction. ■
This example highlights a more general result.
Theorem 11.2
The determinant of a matrix with two proportional row or columns is equal to 0.
♥

Finding the inverse matrix of a given matrix is essentially to solve linear systems.

Example 11.4 Find the inverse matrix of the matrix

0 1
𝐴= .
(−1 0)

187 Lecture Notes for MA451

 11.1 Basics of Linear Algebra

Solution Since the determinant det (𝐴) = 1, this the inverse matrix 𝐴−1 exists. Write

𝑎 𝑏
𝐴−1 = .
( 𝑐 𝑑)

From the definition, the elements 𝑎, 𝑏, 𝑐 and 𝑑 satisfies the following equations

0 1 𝑎 𝑏 1 0
= .
(−1 0) ( 𝑐 𝑑) (0 1)

Simplifying the product yields

𝑐 𝑑 1 0
= .
(−𝑎 −𝑏) (0 1)
Therefore, 𝑎 = 0, 𝑏 = −1, 𝑐 = 1 and 𝑑 = 0. It can be checked that

0 −1 0 1 1 0
= .
(1 0 ) (−1 0) (0 1)

Therefore, the inverse matrix of 𝐴 is

0 −1
𝐴−1 = .
(1 0 )
■
Applying the method in the above example to the matrix

𝑎11 𝑎12
𝐴= , det (𝐴) ≠ 0
(𝑎21 𝑎22 )

yields
1 𝑎22 −𝑎12
𝐴−1 = .
det (𝐴) ( −𝑎 21 𝑎11 )

The transpose of a matrix 𝐴, denoted by 𝐴𝑇 , is the matrix whose 𝑖𝑗-th element is the 𝑗𝑖-th
element of 𝐴.

For a higher order square matrix, there is also a formula for the inverse.

188 Lecture Notes for MA451

 11.1 Basics of Linear Algebra

Theorem 11.3 (Cofactor formula)

The inverse matrix of 𝐴 = (𝑎𝑖𝑗 ) is

1
𝐴−1 = (𝐶𝑗𝑖 ) ,
det (𝐴)

where 𝐶𝑖𝑗 is the cofactor of the 𝑖𝑗-th element of 𝐴.

♥

Be aware of the transpose to the matrix (𝐶𝑗𝑖 ).

The cofactor formula is better for theoretical applications. In practice, to find the inverse
matrix, the Gauss-Jordan elimination method is frequently used. The idea is to solve 𝑛 equa-
tions together by elimination using row operations:

1. interchanging two rows;

2. multiplying a row by a nonzero number;
3. adding a scalar multiple of one row to another row.

Example 11.5 Find the inverse of the matrix

2 5
.
(1 3)

Solution First form the matrix

2 5 1 0
.
(1 3 0 1)
The goal is two use row operations to change the left square to the identity matrix.

2 5 | 1 0 switch rows 1 3 | 0 1 2row1−row2 1 3 | 0 1 row1−3row2 1 0 | 3 −5

→ → →
(1 3 | 0 1) (2 5 | 1 0) (0 1 | −1 2) (0 1 | −1 2 )

So the inverse matrix is

3 −5
,
(−1 2 )
which is exactly the one given by the cofactor formula. ■
The reason that this method works is that each of the row operations can be viewed as
multiply the matrix from the left by a so-called elementary matrix.

The cofactor formula can be used to deduce the Crammer’s rule for linear systems.

189 Lecture Notes for MA451

 11.1 Basics of Linear Algebra

Theorem 11.4 (Crammer’s rule)

Let 𝐴𝑋 = 𝐵 be a linear system, where 𝑋 and 𝐵 are column vectors and 𝑋 is the vector of
unknowns. If 𝐴 is invertible, then

⎛ det 𝐴(𝑖|𝐵) ⎞
( )⎟
𝑋 =⎜ ,
⎜ det 𝐴 ⎟
( )
⎝ ⎠

where 𝐴(𝑖|𝐵) is the matrix obtained from 𝐴 by replacing the 𝑖-th column by the column vector
𝐵. ♥

Example 11.6 Solve the linear system

𝑥1 − 2𝑥2 =1
𝑥1 + 𝑥2 =4

Solution One can solve this linear system using the Gauss elimination method. Here, let’s try the
Crammer’s rule. Write

1 −2 𝑥1 1
𝐴= , 𝑋 = , 𝐵= .
(1 1 ) (𝑥2 ) (4)

Then
1 −2 1 1
𝐴(1|𝐵) = 𝐴(2|𝐵) = .
(4 1 ) (1 4)
Calculating the determinants of 𝐴, 𝐴(1|𝐵) and 𝐴(2|𝐵) yields

det (𝐴) = 3, det (𝐴(1|𝐵)) = 9, det (𝐴(2|𝐵)) = 3

Therefore,
𝑥1 1 9 3
= = .
(𝑥2 ) 3 (3) (1)
■

Eigenvectors and Eigenvalues

Let 𝐴 be a square matrix. A non-zero vector 𝑣⃗ is called an eigenvector of 𝐴 with the
eigenvalue 𝜆 if they satisfy
𝐴𝑣⃗ = 𝜆𝑣.
⃗

A constant 𝜆 is an eigenvalue with the non-zero eigenvector 𝑣⃗ if and only if

(𝐴 − 𝜆𝐼 )𝑣⃗ = 0

190 Lecture Notes for MA451

 11.1 Basics of Linear Algebra

if and only if
det (𝐴 − 𝜆𝐼 ) = 0.

The polynomial 𝑝(𝑥) = det (𝐴 − 𝑥𝐼 ) is called the characteristic polynomial of the matrix 𝐴.
A eigenvalue 𝜆 is of (algebraic) multiplicity 𝑘 if 𝑝(𝑥) = (𝑥 − 𝜆)𝑘 𝑞(𝑥) where 𝑞(𝑥) is a polynomial
had no more factor 𝑥 − 𝜆.
Theorem 11.5
Let 𝐴 be a square matrix. A constant 𝜆 is an eigenvalue of 𝐴 if and only if it is a root of the
characteristic polynomial of 𝐴.
♥

For a 2 × 2 matrix 𝐴, denote by tr(𝐴) the sum of diagonal elements of 𝐴, which is called
the trace of 𝐴, and det(𝐴) the determinant of 𝐴, then the characteristic polynomial of 𝐴 is

𝑝(𝑥) = 𝑥 2 − tr(𝐴)𝑥 + det(𝐴).

There is also a trick to find an eigenvector of a 2 × 2 matrix 𝐴. Suppose that 𝜆 is an

eigenvalue of 𝐴. If
𝑎 𝑏
𝐴 − 𝜆𝐼 = ,
( 𝑐 𝑑)
then the vector
−𝑏 −𝑑
and
(𝑎) (𝑐)
are both eigenvectors if they not zero vectors. Otherwise, 𝐴 − 𝜆𝐼 = 0 and any vector is an
eigenvector. The reason is that that the determinant of 𝐴 − 𝜆𝐼 is zero implies that (𝑐, 𝑑) is a
multiple of (𝑎, 𝑏).

Example 11.7 Find the eigenvalues and their corresponding eigenvectors of the matrix

1 2
.
(3 2)

Solution The determinant of the matrix is

1 2
det = −4.
(3 2)

The trace of the matrix is

1 + 2 = 3.

191 Lecture Notes for MA451

 11.1 Basics of Linear Algebra

So the characteristic polynomial of the matrix is

𝑝(𝑥) = 𝑥 2 − 3𝑥 − 4.

Solving the equation 𝑝(𝑥) = 0 yields

𝑥 = −1 or 𝑥 = 4.

Therefore, the eigenvalues are −1 and 4.

Since
1 2 1 0 2 2
− (−1) =
(3 2) (0 1) (3 3)
An eigenvector of the eigenvalue −1 can be taken to be

2
𝑣⃗ = ,
(−2)

or
1
𝑣⃗ = .
(−1)

A similar calculation implies that

2
𝑣⃗ =
(3)
is an eigenvector of the eigenvalue 4. ■
The usefulness of eigenvalues and eigenvectors is highlighted by the following theorem.
Theorem 11.6 (Diagonolization Theorem)
Suppose the eigenvalues 𝜆1 , … , 𝜆𝑛 of a 𝑛 × 𝑛 matrix 𝐴 are distinct. Let 𝑣⃗𝑖 be the eigenvector
of 𝜆𝑖 . Write 𝑃 = (𝑣|
⃗ 1 | ⋯ |𝑣⃗𝑛 ). Then

⎛ 𝜆1 ⋯ 0 ⎞
−1 ⎜ ⎟
𝑃𝐴𝑃 = ⎜ ⋮ ⋱ ⋮ ⎟.
⎜0 ⋯ 𝜆 ⎟
𝑛
⎝ ⎠ ♥

When the characteristic polynomial of the matrix 𝐴 has a repeated root 𝜆, more vectors
other than eigenvectors will be needed to form an invertible square matrix 𝑃.

A vector 𝑣⃗ is called a generalized eigenvector associated to the eigenvalue 𝜆 of the matrix

𝐴 if
(𝐴 − 𝜆𝐼 )𝑘 𝑣⃗ = 0 for some 𝑘.

192 Lecture Notes for MA451

 11.1 Basics of Linear Algebra

Theorem 11.7 (Jordan Normal Form)

For a square matrix 𝐴, there exist a matrix 𝑀 consists of independent generalized eigenvectors
such that
⎛ 𝐽1 ⋯ 0 ⎞
⎜ ⎟
𝑀𝐴𝑀 −1 = ⎜ ⋮ ⋱ 0 ⎟ ,
⎜0 ⋯ 𝐽 ⎟
𝑟
⎝ ⎠
where 𝐽𝑖 is a 𝑘 × 𝑘 square matrix in the form

⎛𝜆𝑖 1 ⎞
⎜ ⎟
⎜ 𝜆𝑖 ⋱ ⎟
𝐽𝑖 = ⎜ ⎟,
⎜ ⋱ 1⎟
⎜ 𝜆𝑖 ⎟
⎝ ⎠

and 𝜆𝑖 is an eigenvalue of order 𝑘.

♥

Example 11.8 Find the generalized eigenvectors of the matrix

2 1
(−1 0)

Solution The trace is 2 + 0 = 2 and the determinant is 2 ⋅ 0 − (−1) ⋅ 1 = 1. Therefore, the characteristic
polynomial is 𝑝(𝑥) = 𝑥 2 − 2𝑥 + 1 which has a repeated root 𝑥 = 1.

Subtracting the identity matrix from the given matrix yields

2 1 1 0 1 1
− =
(−1 0) (0 1) (−1 −1)

So the eigenvector associate to the eigenvalue is

−1
𝑢⃗ = .
(1)

A generalized vector in this equation is the a vector 𝑣⃗ that satisfies the following linear system

1 1 −1
𝑣⃗ = .
(−1 −1) (1)

Solving the linear system yields that

0
𝑣⃗ = .
(−1)
■

193 Lecture Notes for MA451

 11.1 Basics of Linear Algebra

For 2 × 2 matrix
𝑎 𝑏
𝐴= ,
( 𝑐 𝑑)

the characteristic polynomial 𝑝(𝑥) = 𝑥 2 − (𝑎 + 𝑑)𝑥 + det(𝐴) has a repeated root if and only if
2 2
𝑥 2 − (𝑎 + 𝑑)𝑥 + det(𝐴) = (𝑥 − 𝑎+𝑑
2 )
. It follows that the repeated root is 𝜆 = 𝑎+𝑑
2
and ( 𝑎−𝑑
2 )
= −𝑏𝑐.
Then
𝑎−𝑑
2
𝑏
𝐴 − 𝜆𝐼 = 𝑑−𝑎
.
( 𝑐 2
)
Then
−𝑏
𝑢⃗ =
( 𝑎−𝑑
2
)
is an eigenvector.

Solving the linear system

𝑎−𝑑
2
𝑏 𝑥 −𝑏
𝑑−𝑎
=
( 𝑐 2
) (𝑦) ( 𝑎−𝑑
2
)

yields that the column vector

0
𝑣⃗ = 𝑣⃗ =
(−1)
is a generalized eigenvector.

Calculus of Matrices
If the elements of a matrix 𝐴 are functions of a variable 𝑡, then 𝐴 is called a matrix of
functions of 𝑡.

A matrix of functions 𝐴(𝑥) = (𝑎𝑖𝑗 (𝑥)) is said to be continuous at 𝑡0 if all 𝑎𝑖𝑗 (𝑥) are continuous
at 𝑡0 . It is differential at 𝑡0 if all 𝑎𝑖𝑗 (𝑥) are differential at 𝑡0 . We write the derivate of 𝐴(𝑥) as

d𝐴
(𝑥) = 𝐴′ (𝑥) = (𝑎𝑖𝑗′ (𝑥)) .
d𝑡

Applying the chain rule to the product of matrix yields

(𝐴(𝑥)𝐵(𝑥))′ = 𝐴′ (𝑥)𝐵(𝑥) + 𝐴(𝑥)𝐵′ (𝑥).

Similarly, the integral of a matrix of functions 𝐴(𝑥) = (𝑎𝑖𝑗 (𝑥)) is defined as

𝑏
𝑏
∫ 𝐴(𝑥) d 𝑡 = (∫𝑎 𝑎𝑖𝑗 (𝑥) d 𝑡 ) .
𝑎

194 Lecture Notes for MA451

 11.2 Linear Systems

The Laplace transform of a matrix function 𝑀(𝑡) can be defined similarly as

∞
(𝑀(𝑡))(𝑠) = ∫ 𝑀(𝑡)𝑒 −𝑠𝑡 d 𝑡.
0

Example 11.9 Let

𝑡 𝑒𝑡
𝐴(𝑥) = .
(sin 𝑡 cos 𝑡)
1
Find 𝐴′ (𝑥), ∫0 𝐴(𝑥) d 𝑡 and (𝐴).

Solution From the definition, direct calculations shows that

(𝑥)′ (𝑒 𝑡 )′ 1 𝑒𝑡
𝐴′ (𝑥) = = ,
((sin 𝑡)′ (cos 𝑡)′ ) (cos 𝑡 − sin 𝑡)

1 1 1 1
∫0 𝑡 d 𝑡 ∫0 𝑒 𝑡 d 𝑡 2
𝑒−1
∫ 𝐴(𝑥) d 𝑡 = 1 1 = ,
0 (∫0 sin 𝑡 d 𝑡 ∫0 cos 𝑡 d 𝑡) (cos 1 − 1 −1 )
and
1 1
(𝑡) (𝑒 𝑡 ) 𝑠2 𝑠−1
(𝐴)(𝑠) = (𝑠) = .
((sin 𝑡) (cos 𝑡)) ( 𝑠2𝑠+1 1
𝑠 2 +1
)
■

11.2 Linear Systems

A system of first-order differential equations of 𝑛 unknown functions of a single indepen-
dent variable 𝑥 is a system of differential equations that can be written as

𝑦1′ =𝑓1 (𝑥, 𝑦1 , 𝑦2 , … , 𝑦𝑛 )

𝑦2′ =𝑓2 (𝑥, 𝑦1 , 𝑦2 , … , 𝑦𝑛 )
⋮
𝑦𝑛′ =𝑓𝑛 (𝑥, 𝑦1 , 𝑦2 , … , 𝑦𝑛 )

A system of first-order differential equations is called a linear system if it can be written

as
𝑦1′ =𝑎11 (𝑥)𝑦1 + 𝑎12 (𝑥)𝑦2 + ⋯ + 𝑎1𝑛 (𝑥)𝑦𝑛 + 𝑓1 (𝑥)
𝑦2′ =𝑎21 (𝑥)𝑦1 + 𝑎22 (𝑥)𝑦2 + ⋯ + 𝑎2𝑛 (𝑥)𝑦𝑛 + 𝑓2 (𝑥)
⋮
𝑦𝑛′ =𝑎𝑛1 (𝑥)𝑦1 + 𝑎𝑛2 (𝑥)𝑦2 + ⋯ + 𝑎𝑛𝑛 (𝑥)𝑦𝑛 + 𝑓𝑛 (𝑥).

195 Lecture Notes for MA451

 11.2 Linear Systems

One mathematical reason for studying systems is that an 𝑛-th order differential equation

𝑦 (𝑛) = 𝑓 (𝑥, 𝑦, 𝑦 ′ , ⋯ , 𝑦 (𝑛−1) )

can always be turned into a linear system. Set

𝑦1 = 𝑦, 𝑦2 = 𝑦 ′ , ⋯ , 𝑦𝑛 = 𝑦 (𝑛) .

Then the 𝑛-th order differential equation is equivalent to the linear system

𝑦1′ =𝑦2
𝑦2′ =𝑦3
⋮
𝑦𝑛′ =𝑓 (𝑥, 𝑦1 , 𝑦2 , … , 𝑦𝑛 ) .

In terms of matrices of functions, a linear system can be written as

𝑦⃗′ = 𝐴(𝑥)𝑦⃗ + 𝑓⃗(𝑥),

which suggests the following analogue of the existence and uniqueness theorem of the linear
first-order differential equations.
Theorem 11.8 (Existence and uniqueness)
In the linear system 𝑦⃗′ = 𝐴(𝑥)𝑦⃗ + 𝑓⃗(𝑥), suppose the coefficient matrix 𝐴(𝑥) and the vector
function 𝑓⃗(𝑥) are continuous on (𝑎, 𝑏), let 𝑥0 be in (𝑎, 𝑏), and let 𝑘⃗ be an arbitrary constant
𝑛-vector. Then the initial value problem

𝑦⃗′ = 𝐴(𝑥)𝑦⃗ + 𝑓⃗(𝑥), ⃗ 0 ) = 𝑘⃗

𝑦(𝑥

has a unique solution on (𝑎, 𝑏).

♥

For convenience and clarity, we focus on linear systems of equations in two unknown
functions: ′
𝑦1 (𝑥) 𝑎1 (𝑥) 𝑏1 (𝑥) 𝑦1 (𝑥) 𝑓1 (𝑥)
= + . (11.2)
(𝑦2 (𝑥)) (𝑎2 (𝑥) 𝑏2 (𝑥)) (𝑦2 (𝑥)) (𝑓2 (𝑥))

A linear system is called a homogeneous linear system if

𝑓1 (𝑥)
=0
(𝑓2 (𝑥))

196 Lecture Notes for MA451

 11.2 Linear Systems

Example 11.10 Consider the following linear system

𝑦1′ =2𝑦1 + 4𝑦2

𝑦2′ =4𝑦1 + 2𝑦2 .

Rewrite the linear system into matrix form and verify that the vector of functions

𝑒 6𝑡 𝑒 −2𝑡
𝑦⃗ = 𝑐1 + 𝑐 2
(𝑒 6𝑡 ) (−𝑒 −2𝑡 )

is a solution.

Solution The linear system can be written in matrix form as follows

𝑦1′ 2 4 𝑦1
′
= .
(𝑦2 ) (4 2) (𝑦2 )

The derivative of the vector function 𝑦⃗ is

6𝑒 6𝑡 −2𝑒 −2𝑡
𝑦⃗′ = 𝑐1 + 𝑐 2
(6𝑒 6𝑡 ) ( 2𝑒 −2𝑡 )

Plugging the vector function into the right hand side of the linear system yields

2 4 𝑒 6𝑡 𝑒 −2𝑡
𝑐1 + 𝑐2
(4 2) ( (𝑒 6𝑡 ) (−𝑒 −2𝑡 ))
2 4 𝑒 6𝑡 2 4 𝑒 −2𝑡
=𝑐1 + 𝑐 2
(4 2) (𝑒 6𝑡 ) (4 2) (−𝑒 −2𝑡 )
6𝑒 6𝑡 −2𝑒 −2𝑡
=𝑐1 + 𝑐2
(6𝑒 6𝑡 ) ( 2𝑒 −2𝑡 )
6𝑒 6𝑡 2𝑒 −2𝑡
=𝑐1 + 𝑐2
(6𝑒 6𝑡 ) (−2𝑒 −2𝑡 )

Therefore, the vector function 𝑦⃗ is a solution. ■

Theorems on linear systems

Before discussing how to solve linear systems, we first discuss the structure of general
solutions.

197 Lecture Notes for MA451

 11.2 Linear Systems

Theorem 11.9
If 𝑢⃗ and 𝑣⃗ are two solutions of the homogeneous linear system 𝑦⃗′ = 𝐴𝑦⃗ on [𝑎, 𝑏], then

𝑐1 𝑢⃗ + 𝑐2 𝑣⃗

is also a solution on [𝑎, 𝑏] for any constants 𝑐1 and 𝑐2 .

♥

Proof Since 𝑢⃗ and 𝑣⃗ are solutions,

𝑢⃗′ = 𝐴𝑢⃗ and 𝑣⃗′ = 𝐴𝑣.

⃗

Taking the linear combination yields

⃗ ′ =𝑐1 𝑢⃗′ + 𝑐2 𝑣⃗′

(𝑐1 𝑢⃗ + 𝑐2 𝑣)
=𝑐1 𝐴𝑢⃗ + 𝑐2 𝐴𝑣⃗
=𝐴(𝑐1 𝑢⃗ + 𝑐2 𝑣).
⃗

So 𝑐1 𝑢⃗ + 𝑐2 𝑣⃗ is also a solution. ■
Comparing with linear second-order equations which can be converted to a linear system,
you may wondering if the general solution is a linear combination of two linearly independent
solution. The answer is yes.

Given two vector functions 𝑢⃗ and 𝑣,

⃗ the Wronskian 𝑊 (𝑥) of them is defined to be the
determinant of the matrix (𝑢|
⃗ 𝑣⃗).

Example 11.11 Find the Wronskian of the vector functions

𝑒 2𝑡 2𝑒 𝑡
𝑢⃗ = , 𝑣⃗ = .
(𝑒 2𝑡 ) (−𝑒 𝑡 )

Solution From the definition, the Wronskian is

𝑒 2𝑡 2𝑒 𝑡
det 2𝑡 𝑡
= −𝑒 2𝑡 𝑒 𝑡 − 2𝑒 2𝑡 𝑒 𝑡 = −3𝑒 3𝑡 .
(𝑒 −𝑒 )

■
Theorem 11.10
If two solutions 𝑢⃗ and 𝑣⃗ of the linear system 𝑦⃗′ = 𝐴𝑦⃗ have a non-vanishing Wronskian on
[𝑎, 𝑏], then 𝑐1 𝑢⃗ + 𝑐2 𝑣⃗ is a general solution of the linear system.
♥

The proof is similar to that of the second order differential equations.

Proof By the uniqueness theorem, it sufficiently to show that 𝑐1 and 𝑐2 can be chosen to

198 Lecture Notes for MA451

 11.2 Linear Systems

satisfy arbitrary conditions 𝑦1 (𝑥0 ) = 𝑎 and 𝑦2 (𝑥0 ) = 𝑏. If the Wronskian does not vanish on [𝑎, 𝑏],
then the coefficient matrix of the system of equations

𝑐1 𝑢1 (𝑥0 ) + 𝑐2 𝑣1 (𝑥0 ) =𝑎
𝑐1 𝑢2 (𝑥0 ) + 𝑐2 𝑣2 (𝑥0 ) =𝑏

is invertible and hence −1

𝑐1 𝑢1 (𝑥0 ) 𝑣1 (𝑥0 ) 𝑎
= .
(𝑐2 ) (𝑢2 (𝑥0 ) 𝑣2 (𝑥0 )) (𝑏)
This proves the theorem. ■
Now you may expect more analog properties between Wronskian for linear second-order
differential equation and Wronskian for linear systems.
Theorem 11.11
The Wroskian 𝑊 (𝑥) of two solutions 𝑢⃗ and 𝑣⃗ on [𝑎, 𝑏] of the homogeneous linear system
𝑦⃗′ = 𝐴𝑦⃗ is given by
𝑥
𝑊 (𝑥) = 𝑊 (𝑥0 )𝑒 ∫𝑥0 tr 𝐴(𝑥) d 𝑡 ,

where, tr 𝐴(𝑥) is the sum of the diagonal elements. In particular, the Wronskian is either
identically zero or nonwhere zero.
♥

Proof Suppose

𝑢1 (𝑥) 𝑣1 (𝑥) 𝑎1 (𝑥) 𝑏1 (𝑥)

𝑢(𝑥) = , 𝑣(𝑥) = , 𝐴(𝑥) = .
(𝑢2 (𝑥)) (𝑣2 (𝑥)) (𝑎2 (𝑥) 𝑏2 (𝑥))

Then
𝑢1′ (𝑥) 𝑣1′ (𝑥) 𝑎1 (𝑥) 𝑏1 (𝑥) 𝑢1 (𝑥) 𝑣1 (𝑥)
′ ′
= ,
(𝑢2 (𝑥) 𝑣2 (𝑥)) (𝑎2 (𝑥) 𝑏2 (𝑥)) (𝑢2 (𝑥) 𝑣2 (𝑥))

Applying Jacobi’s formula yields

−1
′ 𝑢1 (𝑥) 𝑣1 (𝑥) 𝑎1 (𝑥) 𝑏1 (𝑥) 𝑢1 (𝑥) 𝑣1 (𝑥)
𝑊 (𝑥) =𝑊 (𝑥) tr
((𝑢2 (𝑥) 𝑣2 (𝑥)) (𝑎2 (𝑥) 𝑏2 (𝑥)) (𝑢2 (𝑥) 𝑣2 (𝑥)))
−1
𝑎1 (𝑥) 𝑏1 (𝑥) 𝑢1 (𝑥) 𝑣1 (𝑥) 𝑢1 (𝑥) 𝑣1 (𝑥)
=𝑊 (𝑥) tr
((𝑎2 (𝑥) 𝑏2 (𝑥)) (𝑢2 (𝑥) 𝑣2 (𝑥)) (𝑢2 (𝑥) 𝑣2 (𝑥)) )
=𝑊 (𝑥) tr(𝐴(𝑥)).

Solving this separable equation implies,

𝑥
𝑊 (𝑥) = 𝑊 (𝑥0 )𝑒 ∫𝑥0 tr 𝐴(𝑡) d 𝑡 .

199 Lecture Notes for MA451

 11.2 Linear Systems

Therefore, 𝑊 (𝑥) ≡ 0 if and only if 𝑊 (𝑥0 ) = 0 for some 𝑥0 which implies the second state-
ment. ■
It follows directly from the above two theorems that the linear combination of two solu-
tions is a general solution if and only if their Wronskian is non-zero.
Theorem 11.12
The linear combination of two solutions 𝑢⃗ and 𝑣⃗ on [𝑎, 𝑏] of the homogeneous linear system
𝑦⃗′ = 𝐴𝑦⃗ is a general solution if and only if the Wronskian is non-zero.
♥

Conceptually, there are equivalent characterization of solutions that form a general solu-
tion.

Two solutions 𝑢⃗ and 𝑣⃗ of 𝑦⃗′ = 𝐴𝑦⃗ are called linearly independent if the only constants 𝑐1
and 𝑐2 such that
𝑐1 𝑢⃗ + 𝑐2 𝑣⃗ = 0

are zero.

A set of solutions {𝑢, ⃗ of 𝑦⃗′ = 𝐴𝑦⃗ is called fundamental set if every solution can be
⃗ 𝑣}
written as the linear combination of those solutions.

The matrix (𝑢⃗ ∣ 𝑣⃗) is called a fundamental matrix for 𝑦⃗′ = 𝐴𝑦⃗ if the column vectors 𝑢⃗ and
𝑣⃗ are linearly independent.

Example 11.12 Consider the linear system

𝑦⃗′ = 𝐴𝑦,
⃗

where
−4 −3
𝐴= .
(6 5)

1. Verify that the vector functions

−𝑒 2𝑡 −𝑒 −𝑡
𝑢⃗ = and 𝑣⃗ =
( 2𝑒 2𝑡 ) ( 𝑒 −𝑡 )

are solutions of the linear system on (−∞, ∞).

2. Compute the Wronskian of 𝑢⃗ and 𝑣. ⃗
3. Find the general solution.

Solution For the first two question, it is convenient to work with the matrix

−𝑒 2𝑡 −𝑒 −𝑡
𝐹 (𝑡) = .
( 2𝑒 2𝑡 𝑒 −𝑡 )

200 Lecture Notes for MA451

 11.3 Constant Coefficient Homogeneous Linear Systems

The derivative of 𝐹 (𝑡) is

−2𝑒 2𝑡 𝑒 −𝑡
𝐹 ′ (𝑡) = .
( 4𝑒 2𝑡 −𝑒 −𝑡 )
The product 𝐴(𝑡)𝐹 (𝑡) is

−4 −3 −𝑒 2𝑡 −𝑒 −𝑡 −2𝑒 2𝑡 𝑒 −𝑡
𝐴(𝑡)𝐹 (𝑡) = =
( 6 5 ) ( 2𝑒 2𝑡 𝑒 −𝑡 ) ( 4𝑒 2𝑡 −𝑒 −𝑡 )

Therefore, 𝐹 ′ (𝑡) = 𝐴(𝑡)𝐹 (𝑡) which verifies that 𝑢⃗ and 𝑣⃗ are solutions.

The Wronskian of 𝑢⃗ and 𝑣⃗ is

−𝑒 2𝑡 −𝑒 −𝑡
𝑊 (𝑡) = 2𝑡 −𝑡
= −𝑒 2𝑡 𝑒 −𝑡 − 2𝑒 2𝑡 (−𝑒 −𝑡 ) = 𝑒 𝑡 .
( 2𝑒 𝑒 )

Since 𝑊 (𝑡) ≢ 0, the solutions 𝑢⃗ and 𝑣⃗ are linearly independent and the linear combination

−𝑒 2𝑡 −𝑒 −𝑡
𝑐1 + 𝑐2
( 2𝑒 2𝑡 ) ( 𝑒 −𝑡 )

is the general solution. ■

We now present the theorem about solutions of non-homogeneous linear systems.
Theorem 11.13
Let 𝑦𝑝 be a particular solution of the linear system the linear system 𝑦⃗′ = 𝐴𝑦⃗ + 𝑓⃗ and 𝑦ℎ be
the general solution of the homogeneous linear system 𝑦⃗′ = 𝐴𝑦. ⃗ Then 𝑦⃗ℎ + 𝑦⃗𝑝 is the general
′ ⃗
solution of 𝑦⃗ = 𝐴𝑦⃗ + 𝑓 .
♥

Proof The statement follows from the fact that the difference of two solutions of 𝑦⃗′ = 𝐴𝑦⃗ + 𝑓⃗
is a solution of 𝑦⃗′ = 𝐴𝑦.
⃗ ■

11.3 Constant Coefficient Homogeneous Linear Systems

Assume the matrix 𝐴 of the homogeneous linear system 𝑦⃗′ = 𝐴𝑦⃗ is a matrix of constant
functions. If you recall that the equation 𝑦 ′ = 𝑐𝑦 has the exponential solution 𝑦 = 𝑦(0)𝑒 𝑐𝑡 , then
⃗ is a solution of the linear system. This is indeed true if we
it is natural to ask where 𝑦⃗ = 𝑒 𝐴𝑡 𝑦(0)
𝑥𝑛
make sense of 𝑒 𝐴𝑡 . Recall that 𝑒 𝑥 = ∑∞ 𝑛=0 𝑛! . The matrix exponential 𝑒
𝐴𝑡
is defined as

∞
𝐴𝑛 𝑡 𝑛
𝑒 𝐴𝑡 = ∑ .
𝑛=0
𝑛!

201 Lecture Notes for MA451

 11.3 Constant Coefficient Homogeneous Linear Systems

Theorem 11.14
⃗ = 𝑦⃗0 has the solution
The linear system 𝑦⃗′ = 𝐴𝑦⃗ with 𝑦(0)

𝑦⃗ = 𝑒 𝐴𝑡 𝑦⃗0 .
♥

The question is how to calculate 𝑒 𝐴𝑡 . In a special case, say 𝐴 is a diagonal matrix:

𝜆1 0
𝐴= ,
( 0 𝜆2 )

where 𝜆1 ≠ 𝜆2 , the power 𝐴𝑛 is also diagonal and

𝜆1𝑛 0
𝐴𝑛 = ,
( 0 𝜆2𝑛 )

which leads to the fact that

𝑒 𝜆1 𝑡 0
𝑒 𝐴𝑡 = .
( 0 𝑒 𝜆2 𝑡 )
Therefore,
𝑦1 𝑦1 (0)𝑒 𝜆1 𝑡
=
(𝑦2 ) (𝑦2 (0)𝑒 𝜆2 𝑡 )
is the solution.

For general cases, there are results from linear algebra that can be used to calculate 𝑒 𝐴𝑡 .
For a constant coefficient homogeneous linear system of two unknown functions, we have
the following results.
Theorem 11.15
Consider the constant coefficient homogeneous linear system of two unknown functions 𝑦⃗′ =
⃗
𝐴𝑦.
Case 1: If 𝐴 has two distinct real roots 𝜆1 and 𝜆2 , then the general solution is

𝑦 = 𝑐1 𝑒 𝜆1 𝑡 𝑣⃗1 + 𝑐2 𝑒 𝜆2 𝑡 𝑣⃗2 ,

where 𝑣⃗1 and 𝑣⃗2 are associated eigenvectors of 𝜆1 and 𝜆2 respectively.

Case 2: If 𝐴 has two distinct complex eigenvalues 𝜆1 , 𝜆2 = 𝛼 + 𝑖𝛽, then the general solution is

𝑦⃗ = 𝑐1 𝑒 𝛼𝑡 (cos(𝛽𝑡)𝑢⃗ − sin(𝛽𝑡)𝑣)
⃗ + 𝑐2 𝑒 𝛼𝑡 (cos(𝛽𝑡)𝑣⃗ + sin(𝛽𝑡)𝑢)
⃗

where 𝑢⃗ ± i𝑣⃗ are eigenvectors of 𝛼 ± i𝛽 respectively.

202 Lecture Notes for MA451

 11.3 Constant Coefficient Homogeneous Linear Systems

Case 3: If 𝐴 has a repeated eigenvalue 𝜆, then the general solution is

𝑦⃗ = 𝑐1 𝑒 𝜆𝑡 𝑢⃗ + 𝑐2 (𝑒 𝜆𝑡 𝑣⃗ + 𝑡𝑒 𝜆𝑡 𝑢),
⃗

where 𝑢⃗ is an eigenvector of 𝜆 and 𝑣⃗ is a vector satisfying (𝐴 − 𝜆𝐼 )𝑣⃗ = 𝑢.

⃗
♥

Example 11.13 Solve the linear system

2 1
𝑦⃗′ = 𝑦⃗
(1 2)

Solution The characteristic polynomial is

𝑝(𝑥) = 𝑥 2 − tr(𝐴)𝑥 + det(𝐴) = 𝑥 2 − 4𝑥 + 3

which has two roots 𝑥 = 1 and 𝑥 = 3.

Since
2 1 1 0 1 1
−1⋅ = ,
(1 2) (0 1) (1 1)
an associated eigenvector is
−1
𝑢⃗ = .
(1)

Since
2 1 1 0 −1 1
−3⋅ = ,
(1 2) (0 1) ( 1 −1)
an associated eigenvector is
1
𝑣⃗ = .
(1)
Therefore, the general solution is

−1 1
𝑦⃗′ = 𝑐1 𝑒 𝑡 + 𝑐2 𝑒 3𝑡 .
(1) (1)

■
Example 11.14 Solve the linear system

4 6
𝑦⃗′ = 𝑦⃗
(−3 −2)

203 Lecture Notes for MA451

 11.3 Constant Coefficient Homogeneous Linear Systems

Solution The characteristic polynomial is

𝑝(𝑥) = 𝑥 2 − tr(𝐴)𝑥 + det(𝐴) = 𝑥 2 − 2𝑥 + 10 = (𝑥 − 1)2 + 9

which has two complex roots 𝑥 = 1 ± 3i.

Since
2 1 1 0 1 − 3i 1
− (1 + 3i) ⋅ = ,
(1 2) (0 1) ( 1 1 − 3i)
an associated eigenvector is

−1 −1 0
𝑢⃗ = = +i .
(1 − 3i) ( 1 ) (−3)

Therefore, the general solution is

−1 0 −1 0
𝑦⃗′ = 𝑐1 𝑒 𝑡 cos(3𝑡) − sin(3𝑡) + 𝑐2 𝑒 𝑡 sin(3𝑡) + cos(3𝑡) .
( (1) (−3)) ( (1) (−3))

■
Example 11.15 Solve the linear system

0 −1
𝑦⃗′ = 𝑦⃗
(1 −2)

Solution The characteristic polynomial is

𝑝(𝑥) = 𝑥 2 − tr(𝐴)𝑥 + det(𝐴) = 𝑥 2 + 2𝑥 + 1

which has a repeated roots 𝑥 = −1.

Since
0 −1 1 0 1 −1
− (−1) ⋅ = ,
(1 −2) (0 1) (1 −1)
an associated eigenvector is
1
𝑢⃗ = .
(1)

Solving the linear system

1 −1 1
𝑣⃗ = ,
(1 −1) (1)

204 Lecture Notes for MA451

 11.4 Variation of Parameters for Nonhomogeneous Linear Systems*

yields
0
𝑣⃗ = .
(−1)
Therefore, the general solution is

1 0 1
𝑦⃗′ = 𝑐1 𝑒 −1 + 𝑐2 𝑒 −𝑡 +𝑡 .
(1) ((−1) (1))

11.4 Variation of Parameters for Nonhomogeneous Linear

Systems*
For nonhomogeneous linear system 𝑦⃗′ = 𝐴𝑦⃗ + 𝑓⃗, the method of variation of parameters
can be used to find a particular solution. Suppose

𝑌 = (𝑦⃗1 |𝑦⃗2 )

be a fundamental matrix of the complementary linear system 𝑦⃗′ = 𝐴𝑦,

⃗ a particular solution 𝑦⃗𝑝
can be expected in the form
𝑦⃗𝑝 = 𝑌 𝑢,
⃗

where 𝑢⃗ is a column vector of functions to be determined. Note that

𝑌 ′ = 𝐴𝑌 .

Plugging 𝑦⃗𝑝 into 𝑦⃗′ = 𝐴𝑦⃗ + 𝑓⃗ and simplifying the equation yields

𝑌 𝑢⃗′ = 𝑓⃗.

It follows that
𝑢⃗ = ∫ 𝑌 −1 𝑓⃗ d 𝑡.

Example 11.16 Consider the linear system

1 2 2𝑒 4𝑡
𝑦⃗′ = 𝑦⃗ + .
(2 1) ( 𝑒 4𝑡 )

1. Find a particular solution of the system.

2. Find the general solution.

205 Lecture Notes for MA451

 11.5 Solving Linear System by Laplace Transform

Solution Solving the complementary homogeneous linear system

1 2
𝑦⃗′ = 𝑦⃗
(2 1)

yields a fundamental matrix

𝑒 −𝑡 𝑒 3𝑡
𝑌 =
(−𝑒 −𝑡 𝑒 3𝑡 .)
Since the determinant of 𝑌 is
det(𝑌 ) = 2𝑒 2𝑡 ,

the inverse matrix of 𝑌 is

1 𝑒 3𝑡 −𝑒 3𝑡 1 𝑒 𝑡 −𝑒 𝑡
𝑌 −1 = = .
2𝑒 2𝑡 (𝑒 −𝑡 𝑒 −𝑡 ) 2 (𝑒 −3𝑡 𝑒 −3𝑡 )

Therefore,
1 𝑒 𝑡 −𝑒 𝑡 2𝑒 4𝑡 1 𝑒 5𝑡
𝑌 −1 𝑓⃗ = =
2 (𝑒 −3𝑡 𝑒 −3𝑡 ) ( 𝑒 4𝑡 ) 2 (3𝑒 𝑡 )

and the parameter vector is

1 5𝑡
1 ∫ 𝑒 5𝑡 d 𝑡 10
𝑒
𝑢⃗ = 𝑡
= 3 𝑡 .
2 (∫ 3𝑒 d 𝑡) ( 2 𝑒 )

It follows that a particular solution is

1 5𝑡 8𝑒 4𝑡
𝑒 −𝑡 𝑒 3𝑡 10
𝑒 5
𝑦⃗𝑝 = 𝑌 𝑢⃗ = −𝑡 3𝑡 3 𝑡
= 7𝑒 4𝑡
.
(−𝑒 𝑒 ) ( 2𝑒 ) ( 4 )

Therefore, the general solution is

8𝑒 4𝑡
𝑒 −𝑡 𝑒 3𝑡 𝑐1 5
𝑦⃗ = 𝑌 𝑐⃗ + 𝑦⃗𝑝 = −𝑡 3𝑡
+ 7𝑒 4𝑡
.
( −𝑒 𝑒 𝑐
) ( 2) ( 4 )

11.5 Solving Linear System by Laplace Transform

Recall that the linear system 𝑦⃗′ = 𝐴𝑦⃗ has the solution 𝑦⃗ = 𝑒 𝐴𝑡 . One way to find 𝑒 𝐴𝑡 is the
use the Jordan normal form of 𝐴. Another method is to use Laplace transform.

Recall that the Laplace transform of the first derivative of a function 𝑦 is

(𝑦 ′ ) = 𝑠(𝑦) − 𝑦(0).

206 Lecture Notes for MA451

 11.5 Solving Linear System by Laplace Transform

Therefore,
((𝑒 𝐴𝑡 )′ )(𝑠) =(𝑒 𝐴𝑡 ) − 𝑒 𝐴⋅0
(𝐴𝑒 𝐴𝑡 )(𝑠) =(𝑒 𝐴𝑡 ) − 𝐼
𝐴(𝑒 𝐴𝑡 )(𝑠) =(𝑒 𝐴𝑡 ) − 𝐼
(𝐴 − 𝑠𝐼 )(𝑒 𝐴𝑡 ) = − 𝐼
(𝑒 𝐴𝑡 ) = − (𝐴 − 𝑠𝐼 )−1
(𝑒 𝐴𝑡 ) =(𝑠𝐼 − 𝐴)−1 .

The matrix 𝑠𝐼 − 𝐴 is known as the characteristic matrix of the matrix 𝐴.

Example 11.17 Find the general solution of the linear system

1 −1
𝑦⃗′ = 𝑦⃗
(0 1 )

Solution The characteristic matrix is

1 0 1 −1 𝑠−1 1
𝑠 = .
(0 1) (0 1 ) ( 0 𝑠 − 1)

The determinant of the characteristic matrix is (𝑠 − 1)2 . So the inverse matrix is

1
𝑠−1 1 𝑠−1
0
= 1 1
.
( 0 𝑠 − 1) (− (𝑠−1) 2 𝑠−1 )

Therefore, the fundamental matrix 𝑒 𝐴𝑡 is

1
0 𝑒𝑡 0
−1 𝑠−1
1 1
=
((− (𝑠−1)2 𝑠−1 )) (−𝑡𝑒 𝑡 𝑒 𝑡 .)

The general solution is

𝑒𝑡 0 𝑐1
𝑦⃗ = 𝑡 𝑡
.
(−𝑡𝑒 𝑒 .) (𝑐2 )
■

207 Lecture Notes for MA451

 Week 12: Bounded Value Problems and Fourier
Series

12/6–12/09

12.1 Introduction to Bounded Value Problems

In studying heat conduction, or wave propagation, the following type of equation will be
employed
𝑦 ′′ + 𝜆𝑦 = 0, 𝑦(0) = 0, 𝑦(𝐿) = 0,

where 𝜆 is a real number and 𝐿 > 0.

More generally, the conditions may be given as

𝑎𝑦(𝑥0 ) + 𝑏𝑦 ′ (𝑥0 ) = 𝛼 𝑐𝑦(𝑥1 ) + 𝑑𝑦 ′ (𝑥1 ) = 𝛽,

which are called boundary conditions. A differential equation with boundary conditions is
called bounded value problems.

Not like linear second-order differential equations with constant coefficients and initial
conditions, with boundary conditions, the existence and number of solutions vary.

Example 12.1 Solve 𝑦 ′′ + 𝑦 = 0 with each of the following boundary conditions

1. 𝑦(0) = 1 and 𝑦(𝜋/2) = 0.

2. 𝑦(0) = 0 and 𝑦(𝜋) = 1.
3. 𝑦(0) = 0 and 𝑦(𝜋) = 0.

Solution Since the characteristic polynomial of the equation is 𝑝(𝑟) = 𝑟 2 + 1 which has two roots
𝑟 = ±𝑖. Then the general solution of the equation is

𝑦 = 𝑐1 cos 𝑥 + 𝑐2 sin 𝑥.

1. The boundary condition 𝑦(0) = 0 implies that 𝑐1 = 1. The boundary condition 𝑦(𝜋/2) = 0
implies that 𝑐2 = 0. So this boundary value problem has a unique solution 𝑦 = cos 𝑥.
2. The boundary condition 𝑦(0) = 0 implies that 𝑐1 = 0. But the boundary condition 𝑦(𝜋) = 1
implies that 𝑐1 = −1. That’s a contradiction, which means this boundary value problem has
no solution.
3. The boundary conditions implies that 𝑐1 = 0 and 𝑐2 can be any numbers. So this boundary
value problem has infinitely many solutions in the form 𝑦 = 𝑐 sin 𝑥.
 12.1 Introduction to Bounded Value Problems

■
The existence and number of solutions also depend on the value of 𝜆.

Example 12.2 Solve 𝑦 ′′ + 14 𝑦 = 0 with boundary conditions 𝑦(0) = 0 and 𝑦(𝜋) = 0.

1
Solution Since the characteristic polynomial of the equation is 𝑝(𝑟) = 𝑟 2 + 4
which has two roots
𝑟 = ± 12 𝑖. Then the general solution of the equation is

𝑥 𝑥
𝑦 = 𝑐1 cos ( ) + 𝑐2 sin ( ) .
2 2
The boundary condition 𝑦(0) = 0 implies that 𝑐1 = 0. The boundary condition 𝑦(𝜋) = 1 implies that
𝑐2 = 0. So this boundary value problem has only a trivial solution 𝑦 = 0. ■
The boundary value problem

𝑦 ′′ + 𝜆𝑦 = 𝑓 (𝑥), 𝑎𝑦(0) + 𝑏𝑦 ′ (0) = 𝛼 𝑐𝑦(𝐿) + 𝑑𝑦 ′ (𝐿) = 𝛽

is called a nonhomogeneous boundary value problem over [0, 𝐿] if either 𝑓 (𝑥), 𝛼 or 𝛽 is nonzero.
The associated homogeneous boundary value problem is given by setting 𝑓 = 0, 𝛼 = 0 and 𝛽 = 0.

Like initial value problems, solutions of nonhomogeneous boundary value problems are
sums of a particular solution and solutions of the associated homogeneous boundary problem.
Lemma 12.1
Let 𝑢 be a solution of the nonhomogeneous boundary value problem

𝑦 ′′ + 𝜆𝑦 = 𝑓 (𝑥), 𝑎𝑦(0) + 𝑏𝑦 ′ (0) = 𝛼 𝑐𝑦(𝐿) + 𝑑𝑦 ′ (𝐿) = 𝛽.

Then any solution 𝑦 of this nonhomogeneous boundary problem is given by

𝑦 =𝑢+𝑧

for some solution 𝑧 of the associated homogeneous boundary problem

𝑦 ′′ + 𝜆𝑦 = 0, 𝑎𝑦(0) + 𝑏𝑦 ′ (0) = 0 𝑐𝑦(𝐿) + 𝑑𝑦 ′ (𝐿) = 0.

♥

Proof It sufficiently to show that 𝑦 − 𝑢 is a solution of the associated boundary value problem.
But that can be verified directly. ■
To find all solutions of a boundary value problem, the key is to study the associated ho-
mogeneous boundary value problem.

The following property of homogeneous boundary value problem can be verified directly.

209 Lecture Notes for MA451

 12.2 Eigenvalues and Eigenfunctions

Theorem 12.1 (Superposition Principle)

Let 𝑢1 and 𝑢2 be two solutions of the homogeneous boundary value problem

𝑦 ′′ + 𝜆𝑦 = 0, 𝑎𝑦(0) + 𝑏𝑦 ′ (0) = 0 𝑐𝑦(𝐿) + 𝑑𝑦 ′ (𝐿) = 0.

Then the linear combination

𝑐1 𝑢2 + 𝑐2 𝑢2

is also a solution. ♥

12.2 Eigenvalues and Eigenfunctions

In this section, we study homogeneous boundary value problems.

Consider the boundary value problem

𝑦 ′′ + 𝜆𝑦 = 0, 𝑎𝑦(0) + 𝑏𝑦 ′ (0) = 0 𝑐𝑦(𝐿) + 𝑑𝑦 ′ (𝐿) = 0.

A value of 𝜆 such that the boundary value problem has a nontrivial solution is called an eigen-
value of the boundary value problem, and the nontrivial solutions are called 𝜆-eigenfunctions,
or eigenfunctions associated to 𝜆.

Example 12.3 Find eigenvalues and eigenfunctions of the boundary value problem

𝑦 ′′ + 𝜆𝑦 = 0, 𝑦(0) = 0 𝑦(𝐿) = 0.

Solution If 𝜆 = 0, then the general solution of the equation 𝑦 ′′ +𝜆𝑦 = 0 is 𝑦 = 𝑐1 +𝑐2 𝑥. The boundary
conditions imply 𝑐1 = 0 and 𝑐2 = 0. So 𝜆 = 0 is not an eigenvalue.

If 𝜆 < 0, then equation 𝑦 ′′ + 𝜆𝑦 = 0 has the general solution

√ √
𝑦 = 𝑐1 𝑒 − −𝜆𝑥
+ 𝑐2 𝑒 −𝜆𝑥
.
√
Note that the general solution can also be written as a linear combination 𝑦 = 𝑐1 cosh( −𝜆𝑥) +
√
𝑐2 sinh( −𝜆𝑥) of the hyperbolic functions:

𝑒 𝑥 + 𝑒 −𝑥 𝑒 𝑥 − 𝑒 −𝑥
cosh 𝑥 = sinh 𝑥 = ,
2 2
which will be convenient to determine 𝑐1 and 𝑐2 from the boundary conditions.

Since cosh(𝑥) > 0 for any 𝑥 and sinh(𝑥) = 0 if and only if 𝑥 = 0, then the boundary condition
𝑦(0) = 0 implies that 𝑐1 = 0 and the boundary condition 𝑦(𝐿) = 0 then implies 𝑐2 = 0. So a negative
𝜆 is not an eigenvalue.

210 Lecture Notes for MA451

 12.2 Eigenvalues and Eigenfunctions

If 𝜆 > 0, the the general solution is

√ √
𝑦 = 𝑐1 cos( 𝜆𝑥) + 𝑐2 sin( 𝜆𝑥).

The boundary conditions imply 𝑐1 and 𝑐2 satisfy the system

⎧
⎪
⎪𝑐1 = 0
⎨
⎪ √
⎪ 𝑐2 sin( 𝜆𝐿) = 0
⎩
√ 𝑛2 𝜋 2
So 𝜆 is an eigenvalue if 𝜆𝐿 = 𝑛𝜋,or equivalently 𝜆 = 𝐿2
, where 𝑛 ≠ 0.

In this case, the function

𝑛𝜋𝑥
𝑦𝑛 = sin (
𝐿 )
is an eigenfunction associated to the eigenvalue 𝜆 = 𝑛2 𝜋 2
𝐿2
. ■
 Exercise 12.1 Find eigenvalues and eigenfunctions of the boundary value problem

𝑦 ′′ + 𝜆𝑦 = 0, 𝑦 ′ (0) = 0 𝑦 ′ (𝐿) = 0.

Solution If 𝜆 = 0, then the general solution of the equation 𝑦 ′′ + 𝜆𝑦 = 0 is 𝑦 = 𝑐1 + 𝑐2 𝑥. The

boundary conditions imply 𝑐2 = 0. So 𝜆 = 0 is an eigenvalue with an eigenfunction 𝑦 = 1.

If 𝜆 < 0, then equation 𝑦 ′′ + 𝜆𝑦 = 0 has the general solution

√ √
𝑦 = 𝑐1 cosh( −𝜆𝑥) + 𝑐2 sinh( −𝜆𝑥).

Since (cosh 𝑥)′ = − sinh 𝑥 and (sinh 𝑥) = cosh 𝑥, the boundary conditions 𝑦 ′ (0) = 0 and 𝑦 ′ (𝐿) = 0
imply that 𝑐1 = 0 and 𝑐2 = 0. So a negative 𝜆 is not an eigenvalue.

If 𝜆 > 0, the the general solution is

√ √
𝑦 = 𝑐1 cos( 𝜆𝑥) + 𝑐2 sin( 𝜆𝑥).

The boundary conditions imply 𝑐1 and 𝑐2 satisfy the system

⎧
⎪
⎪𝑐2 = 0
⎨
⎪ √
⎪ 𝑐1 sin( 𝜆𝐿) = 0
⎩
√ 𝑛2 𝜋 2
So 𝜆 is an eigenvalue if 𝜆𝐿 = 𝑛𝜋,or equivalently 𝜆 = 𝐿2
, where 𝑛 ≠ 0.

In this case, the function

𝑛𝜋𝑥
𝑦𝑛 = cos (
𝐿 )

211 Lecture Notes for MA451

 12.2 Eigenvalues and Eigenfunctions

is an eigenfunction associated to the eigenvalue 𝜆 = 𝑛2 𝜋 2

𝐿2
. ■
Those eigenfunctions has very interesting properties.

Two integrable functions 𝑓 and 𝑔 are said to be orthogonal on an interval [𝑎, 𝑏] if

𝑏

∫ 𝑓 (𝑥)𝑔(𝑥) d 𝑥 = 0.
𝑎

More generally, a collection of integrable functions 𝜙1 , 𝜙2 , ..., 𝜙𝑛 , ..., are orthogonal on [𝑎, 𝑏]
if they are mutually orthogonal.

Example 12.4 The eigenfunctions sin ( 𝑛𝜋𝑥

𝐿 )
, 𝑛 = 1, 2, … , are orthogonal to each other over
[0, 𝐿].

Solution Using the product to sum identity

1
sin 𝛼 sin 𝛽 = (cos(𝛼 − 𝛽) − cos(𝛼 + 𝛽)) ,
2
the integral is calculated by
𝐿
𝑚𝜋𝑥 𝑛𝜋𝑥
∫ sin ( ) sin ( d𝑥
0 𝐿 𝐿 )
𝐿 𝐿
1 (𝑚 − 𝑛)𝜋𝑥 (𝑚 + 𝑛)𝜋𝑥
= ∫ cos d 𝑥 − ∫ cos d𝑥
2( 0 ( 𝐿 ) 0 ( 𝐿 ) )
1 𝐿 (𝑚 + 𝑛)𝜋𝑥 ||𝐿 𝐿 (𝑚 − 𝑛)𝜋𝑥 ||𝐿
= sin − sin
2 ( (𝑚 + 𝑛)𝜋 ( 𝐿 ) ||0 (𝑚 − 𝑛)𝜋 ( 𝐿 ) ||0 )
1 𝐿
= sin((𝑚 + 𝑛)𝜋) − sin((𝑚 − 𝑛)𝜋)
2 ( (𝑚 + 𝑛)𝜋 )
= 0.

Therefore, those eigenfunctions are mutually orthogonal. ■

 Exercise 12.2 Show that the functions
𝜋𝑥 𝜋𝑥 2𝜋𝑥 2𝜋𝑥 𝑛𝜋𝑥 𝑛𝜋𝑥
1, cos ( ) , sin ( ) , cos ( ) , sin ( ) , … , cos ( ) , sin ( ,…
𝐿 𝐿 𝐿 𝐿 𝐿 𝐿 )
are orthogonal on [−𝐿, 𝐿].

Solution From the definition of orthogonality, we need to show that

𝐿

∫ 𝑓 (𝑥)𝑔(𝑥) d 𝑥 = 0,
−𝐿

where 𝑓 and 𝑔 are two distinct functions in the given set of functions.

212 Lecture Notes for MA451

 12.3 Introduction to Fourier Series

For any integer 𝑟 ≠ 0, because 𝑠𝑖𝑛 is an odd function, calculating using the substitution
𝑡 = −𝑥 yields
𝐿 𝐿
𝑟𝜋𝑥 𝑟𝜋𝑡
∫ sin ( ) d 𝑥 = − ∫ sin ( d𝑡
−𝐿 𝐿 −𝐿 𝐿 )
which implies
𝐿
𝑟𝜋𝑥
∫ sin ( d 𝑥 = 0.
−𝐿 𝐿 )

Similarly,
𝐿
𝑚𝜋𝑥 𝑛𝜋𝑥
∫ sin ( ) cos ( d 𝑥 = 0,
−𝐿 𝐿 𝐿 )
because the integrand is an odd function.
𝑟𝜋𝑥
Calculating using the substitution 𝑡 = 𝐿
yields

𝐿
𝑟𝜋𝑥 𝐿 𝑟𝜋
𝐿 |𝑟𝜋
∫ cos ( d𝑥 = cos 𝑡 d 𝑡 = sin 𝑡 || = 0.
−𝐿 𝐿 ) 𝑟𝜋 ∫−𝑟𝜋 𝑟𝜋 |−𝑟𝜋

Therefore, 1 is orthogonal with any function in the set, and sine and cosine functions in
the set are also orthogonal.

It remains to check orthogonality among sine functions and among cosine functions.

The trigonometric identities,

1
sin 𝛼 sin 𝛽 = (cos(𝛼 − 𝛽) − cos(𝛼 + 𝛽)),
2
1
cos 𝛼 cos 𝛽 = (cos(𝛼 + 𝛽) + cos(𝛼 − 𝛽)),
2
and the following result
𝐿
𝑟𝜋𝑥
∫ cos ( d 𝑥 = 0,
−𝐿 𝐿 )
together implies the orthogonality among sine function or among cosine function.

Therefore, the set of functions are orthogonal. ■

12.3 Introduction to Fourier Series

The orthogonality of eigenfunctions has great applications in solving partial differential
equations. Eigenfunctions can be used to express a function 𝑓 as a sum of linear combination
of them if the function 𝑓 is good enough.

213 Lecture Notes for MA451

 12.3 Introduction to Fourier Series

Theorem 12.2
Suppose the functions 𝜙1 , 𝜙2 , 𝜙3 , ..., are orthogonal on [𝑎, 𝑏] and
𝑏

∫ 𝜙𝑛2 (𝑥) d 𝑥 ≠ 0, 𝑛 = 1, 2, 3, … .
𝑎

Let 𝑐1 , 𝑐2 , 𝑐3 ,... be constants such that the partial sums

𝑁
𝑓𝑁 (𝑥) = ∑ 𝑐𝑚 𝜙𝑚 (𝑥)
𝑚=1

satisfy the inequalities

|𝑓𝑁 (𝑥)| ≤ 𝑀, 𝑎 ≤ 𝑥 ≤ 𝑏, 𝑁 = 1, 2, 3, …

for some constant 𝑀 < ∞. Suppose also that the series

∞
𝑓 (𝑥) = ∑ 𝑐𝑚 𝜙𝑚 (𝑥)
𝑚=1

converges and is integrable on [𝑎, 𝑏]. Then

𝑏
∫𝑎 𝑓 (𝑥)𝜙𝑛 (𝑥) d 𝑥
𝑐𝑛 = 𝑏
, 𝑛 = 1, 2, 3, … .
∫𝑎 𝜙𝑛2 (𝑥) d 𝑥
♥
𝑏
Suppose 𝜙1 , 𝜙2 ,..., 𝜙𝑛 , ..., are orthogonal on [𝑎, 𝑏] and ∫ 𝜙𝑛2 (𝑥) d 𝑥 ≠ 0, 𝑛 = 1, 2, 3, … . Let 𝑓
𝑎
be integrable on [𝑎, 𝑏], and define
𝑏
∫𝑎 𝑓 (𝑥)𝜙𝑛 (𝑥) d 𝑥
𝑐𝑛 = 𝑏
, 𝑛 = 1, 2, 3, … .
∫𝑎 𝜙𝑛2 (𝑥) d 𝑥

Then the infinite series ∞

∑ 𝑐𝑛 𝜙𝑛 (𝑥)
𝑛=1

is called the Fourier expansion of 𝑓 in terms of the orthogonal set {𝜙𝑛 }∞

𝑛=1 , and 𝑐1 , 𝑐2 ,..., 𝑐𝑛 ,... are
∞
called the Fourier coefficients of 𝑓 with respect to {𝜙𝑛 }𝑛=1 .

Due to the fact that the Fourier expansion of a function may diverges every where, we
indicate the relationship between 𝑓 and its Fourier expansion by
∞
𝑓 (𝑥) ∼ ∑ 𝑐𝑛 𝜙𝑛 (𝑥), 𝑎 ≤ 𝑥 ≤ 𝑏.
𝑛=1

214 Lecture Notes for MA451

 12.3 Introduction to Fourier Series

Recall that the set of functions

𝜋𝑥 𝜋𝑥 2𝜋𝑥 2𝜋𝑥 𝑛𝜋𝑥 𝑛𝜋𝑥
1, cos ( ) , sin ( ) , cos ( , sin , … , cos , sin ( 𝐿 ),…
𝐿 𝐿 𝐿 ) ( 𝐿 ) ( 𝐿 )

are orthogonal on [−𝐿, 𝐿].

For any piecewise continuous function on [−𝐿, 𝐿], the Fourier expansion

𝑎0 ∞ { 𝑛𝜋𝑥 𝑛𝜋𝑥 }
𝑓 (𝑥) ∼ + ∑ 𝑎𝑛 cos ( + 𝑏𝑛 sin (
2 𝑛=1 𝐿 ) 𝐿 )

is called the Fourier series of 𝑓 , where 𝑎𝑛 and 𝑏𝑛 , for 𝑛 = 0, 1, 2, … , are given by the Euler-Fourier
formulas

1 𝐿 𝑛𝜋𝑥 1 𝐿 𝑛𝜋𝑥
𝑎𝑛 = ∫ 𝑓 (𝑥) cos ( d𝑥 and 𝑏𝑛 = ∫ 𝑓 (𝑥) sin ( d 𝑥.
𝐿 −𝐿 𝐿 ) 𝐿 −𝐿 𝐿 )

Example 12.5 Find the Fourier series of 𝑓 (𝑥) = |𝑥| on [−𝜋, 𝜋].

Solution

Since 𝑓 is even and sin(𝑛𝑥) is odd, the number 𝑏𝑛 = 0 for all 𝑛 and the Fourier series can be
written as
𝑎0 ∞
𝐹 (𝑥) ∶= + ∑ 𝑎𝑛 cos(𝑛𝑥).
2 𝑛=1

Since cos(𝑛𝑥) is also even, applying the methods of substitutions and integration by parts im-
plies
1 𝜋 2 𝜋
𝑎0 = |𝑥| d 𝑥 = 𝑥 d 𝑥 = 𝜋.
𝜋 ∫−𝜋 𝜋 ∫0
1 𝜋
𝑎𝑛 = ∫ |𝑥| cos(𝑛𝑥) d 𝑥
𝜋 −𝜋
2 𝜋
= ∫ 𝑥 cos(𝑛𝑥) d 𝑥
𝜋 0
2 |𝜋 𝜋
= |
𝑥 sin(𝑛𝑥)| − ∫ sin(𝑛𝑥) d 𝑥
𝑛𝜋 ( |0 0 )
| 𝜋
2
= 2 cos(𝑛𝑥)||
𝑛𝜋 |0
𝑛
2((−1) − 1)
=
𝑛2 𝜋
⎧
⎪
⎪0 if 𝑛 = 2(𝑘 + 1)
=⎨ 𝑘 = 0, 1, 2, … .
⎪
⎪ −2 if 𝑛 = 2𝑘 + 1
⎩

215 Lecture Notes for MA451

 12.3 Introduction to Fourier Series

Therefore
𝜋 4 ∞ 1
𝐹 (𝑥) = − ∑ cos[(2𝑘 + 1)𝑥].
2 𝜋 𝑘=0 (2𝑘 + 1)2
■

216 Lecture Notes for MA451

 Bibliography

Bracewell, Ronald (June 1999). The Fourier Transform & Its Applications. 3rd edition. Boston:
McGraw-Hill Science/Engineering/Math (cit. on p. 181).
Doetsch, Gustav (1974). Introduction to the Theory and Application of the Laplace Transformation.
Springer-Verlag, New York-Heidelberg (cit. on pp. 166, 169, 171).
Simmons, George F. (2016). Differential equations with applications and historical notes. English.
Boca Raton, FL: CRC Press, pp. xxi + 740 (cit. on pp. 52, 76, 144).
 Index

𝜆-eigenfunctions, 210 implicit solution of the initial value problem,

16
analytic, 138
initial condition, 9
Bernoulli first order, 29 initial value problem, 9
boundary conditions, 208 integral curve, 12
bounded value problems, 208 integrating factor, 24, 44
integrating factor method, 24
characteristic equation, 85
inverse Laplace transform, 171
characteristic polynomial, 85, 86
irregular singular, 152
complementary equation, 91
complimentary, 27 linear, 23, 75
constant coefficient equation, 84 linear system, 195
linearly independent, 78, 200
delta function, 180
differential equation, 1 nonhomogeneous boundary value problem, 209
differential form, 15 nonlinear, 23
Dirac delta function, 180
order, 7
direct integration problem, 11
ordinary differential equation, 1
direction filed, 13
ordinary point, 143
eigenfunctions associated to 𝜆, 210 orthogonal, 70, 212
eigenvalue of the boundary value problem, 210 orthogonal trajectory, 70
Euler equation, 153
piecewise continuous, 164
exact, 38
power series, 135
exponential order, 165
power series solution, 144
Fourier coefficients, 214
radius of convergence, 136
Fourier expansion, 214
regular singular, 152
fundamental matrix, 200
fundamental set, 200 separable, 16
separation of variables, 16
general solution, 10
singular point, 143
Heaviside step function, 164 slope field, 13
homogeneous, 75 solution, 7
homogeneous boundary value problem, 209 solution curve, 12
homogeneous first order differential equation, solution of an initial value problem, 9
33 standard form, 15
homogeneous linear system, 196
unit impulse function, 180
homogeneous of degree 𝑚, 33
unit impulse response, 182
implicit solution, 16 unit step function, 164
 INDEX

weight function, 182

Wronskian, 79
Wronskian determinant, 79

219 Lecture Notes for MA451