KIX 1001: ENGINEERING

MATHEMATICS 1
WEEK 8: INTEGRATION

WEEK 9: ENGINEERING APPLICATIONS OF INTEGRALS

WEEK 10: MULTIPLE INTEGRALS

WEEK 11: DIFFERENTIAL EQUATIONS

WEEK 12: SOLUTION TO HOMOGENEOUS & NONHOMOGENEOUS

WEEK 13: POWER SERIES SOLUTION FOR DIFFERENTIAL EQUATION

WEEK 14: ENGINEERING APPLICATIONS OF DIFFERENTIAL EQUATION

2017/2018
CONTENTS

WEEK 8: INTEGRATION............................................................................................................................... 4
8.1 terminology and basic integration rules.................................................................................................. 4
8.2 techniques of integration ....................................................................................................................... 4
8.2.1 integration BY PARTS ......................................................................................................................... 4
8.2.2 Trigonometric SUBSTITUTION ........................................................................................................... 9
8.2.3 PARTIAL FRACTIONS ....................................................................................................................... 11
8.2.3.1 case 1: Q(X) is a product of distinct linear factors ............................................................................ 13
8.2.3.2 case 2: Q(X) is a product of distinct linear factors, some of which are repeated .............................. 14
8.2.3.3 case 3: Q(X) contains irreducible quadratic factors, none of which is repeated................................ 16
8.2.3.4 case 4: Q(X) contains A repeated irreducible quadratic factor..........................................................17
8.3 improper integrals ................................................................................................................................ 18
8.3.1 Type 1: Infinite intervals ..................................................................................................................... 18
8.3.2 Type 2: discontinuous integrands ...................................................................................................... 19
8.3.3 A comparison test for improper integrals........................................................................................... 21
WEEK 9: ENGINEERING APPLICATIONS OF INTEGRALS ......................................................................... 23
9.1 ARC length ........................................................................................................................................... 23
9.1.1 The ARC length function .................................................................................................................... 27
9.2 area of a surface of revolution .............................................................................................................. 28
9.3 hydrostatic pressure and force ............................................................................................................. 34
9.4 moments and centres of mass .............................................................................................................. 36
WEEK 10 : MULTIPLE INTEGRALS ............................................................................................................. 43
10.1 double integrals over rectangles ......................................................................................................... 43
10.1.1 Review of definite integrals .............................................................................................................. 43
10.1.2 volumes and double integrals .......................................................................................................... 43
10.1.3 iterated integrals.............................................................................................................................. 47
10.1.5 average value ................................................................................................................................... 50
10.2 double integrals over general regions ................................................................................................. 51
10.2.1 properties of double integrals .......................................................................................................... 56
10.3 double integrals in polar coordinates .................................................................................................. 57
10.4 triple integrals .................................................................................................................................... 62

1
10.4.1 triple integrals over a general bounded region E .............................................................................. 65
10.4.2 triple integrals in cylindrical coordinates.......................................................................................... 68
10.4.3 cylindrical coordinates ..................................................................................................................... 68
10.4.4 evaluating triple integrals with cylindrical coordinates .................................................................... 70
WEEK 11: DIFFERENTIAL EQUATIONS ..................................................................................................... 73
11.1 INTRODUCTION TO DIFFERENTIAL EQUATION ............................................................................... 73
11.2 THE CLASSIFICATION/TYPE OF DIFFERENTIAL EQUATIONS ........................................................... 75
11.3 SOLUTION TO DIFFERENTIAL EQUATION ........................................................................................ 85
11.4 VERIFICATION OF DIFFERENTIAL EQUATION’S SOLUTION ............................................................. 87
11.5 STRATEGY TO SOLVE 1ST ORDER DIFFERENTIAL EQUATION .......................................................... 91
11.5.1 Exact Differential Equation........................................................................................................... 91
11.5.2 Linear Differential Equation ......................................................................................................... 93
11.5.3 Separable Differential Equation ................................................................................................... 98
11.5.4 Bernoulli’s Differential Equation ................................................................................................ 100
11.5.5 Differential Equation of Homogeneous 𝑑𝑦/𝑑𝑥 = 𝑓(𝑥, 𝑦)/𝑔(𝑥, 𝑦) form ..................................... 105
11.5.6 Differential Equation of Nonhomogeneous 𝑑𝑦/𝑑𝑥 = 𝑓(𝑥, 𝑦)/𝑔(𝑥, 𝑦) Form ..............................107
WEEK 12: SOLUTION TO HOMOGENEOUS & NON-HOMOGENEOUS .................................................. 108
12.1 INTRODUCTION TO 2ND ORDER DIFFERENTIAL EQUATION .......................................................... 108
12.2 SOLUTION TO 2ND ORDER LINEAR DIFFERENTIAL EQUATION ...................................................... 109
12.3 GENERAL THEORY OF LINEARITY PRINCIPLE & LINEAR DEPENDENCY ....................................... 110
12.4 STRATEGY TO SOLVE 2ND ORDER DIFFERENTIAL EQUATION ........................................................115
12.4.1 Homogenous Linear Differential Equation with Constant Coefficients 𝑎, 𝑏, 𝑐 ..............................115
12.4.2 Homogeneous Linear Differential Equation with Non-Constant Coefficients 𝑥2, 𝑎𝑥 (Known as
Euler-Cauchy Differential Equation) ..................................................................................................... 121
12.4.3 Nonhomogeneous Linear Differential Equation with Constant Coefficients 𝑎, 𝑏, 𝑐 In The Form of
𝑟𝑥 = 𝑒𝛼𝑥𝑃𝑛(𝑥).................................................................................................................................... 124
12.4.4 Solving All Types of Nonhomogeneous Linear Differential Equation with Method of Variation of
Parameters .......................................................................................................................................... 142
WEEK 13: POWER SERIES SOLUTIONS FOR DIFFERENTIAL EQUATIONS ..............................................147
13.1 POWER SERIES METHOD .................................................................................................................147
13.1.1 Basic Concepts of Power Series .................................................................................................. 148
13.1.2 Test for Convergence ................................................................................................................. 150

2
 13.1.3 Operations of Power Series .........................................................................................................151
13.1.4 Vanishing All Coefficients – A CONDITION THAT IS A BASIC TOOL OF THE POWER SERIES
METHOD ............................................................................................................................................. 152
13.1.5 Idea of The Power Series Method ............................................................................................... 154
13.2 FROBENIUS METHOD ...................................................................................................................... 164
WEEK 14: ENGINEERING APPLICATIONS OF DIFFERENTIAL EQUATION .............................................. 169
14.1 A LIQUID SYSTEM ............................................................................................................................ 169
14.2 A MIXTURE PROBLEM ......................................................................................................................170
14.3 AN LCR CIRCUIT ................................................................................................................................172
14.4 VIBRATING SPRINGS ........................................................................................................................172
14.5 DAMPED VIBRATIONS ......................................................................................................................174
14.6 FORCED VIBRATIONS.......................................................................................................................178
14.7 ELECTRIC CIRCUIT ............................................................................................................................179
Appendix 11.1 MATHEMATICAL MODELING AND ENGINEERING PROBLEM SOLVING ........................ 183
Appendix 11.2 CONSERVATION LAWS AND ENGINEERING ................................................................... 186

3
INTEGRATION
WEEK 8: INTEGRATION
8.1 TERMINOLOGY AND BASIC INTEGRATION RULES

There are two types of integrals: Indefinite and Definite. Indefinite integrals are those with no limits and
definite integrals have limits. When dealing with indefinite integrals, you need to add a constant of
integration. For example, if integrating a function 𝑓(𝑥) with respect to x:

∫ 𝑓(𝑥) 𝑑𝑥 = 𝑔(𝑥) + 𝐶,

Where 𝑔(𝑥) is the integrated function. C is an arbitrary constant called the constant of integration and 𝑑𝑥
indicated the variable with respect to which we are integrating, in this case,𝑥. The function being integrated,
𝑓(𝑥), is called the integrand.

The integral of many functions are well known, and there are useful rules to work out the integral of more
complication functions, which are shown in Figure 8.1 below. The summary of the common procedures for
fitting integrands to the basic integration rules is given in Figure 8.2.

8.2 TECHNIQUES OF INTEGRATION

This section will discussed in more detail three methods of integration: Integration by parts, the substitution
method and partial fractions.

8.2.1 INTEGRATION BY PARTS

One of the important integration techniques is called integration by parts. This technique can be applied to
a wide variety of functions and is particularly useful for integrands involving products of algebraic and
transcendental functions.

If 𝑓 and 𝑔 are differentiable functions, then the Product Rule yields

𝑑
𝑑𝑥
[𝑓(𝑥)𝑔(𝑥)] = 𝑓(𝑥)𝑔′ (𝑥) + 𝑔(𝑥)𝑓 ′ (𝑥) ……..(1)

In the form of indefinite integrals, Eq (1) becomes

∫[𝑓(𝑥)𝑔′ (𝑥) + 𝑔(𝑥)𝑓 ′ (𝑥)]𝑑𝑥 = 𝑓(𝑥)𝑔(𝑥)

Then, rearranging the equation yields

∫ 𝑓(𝑥)𝑔′ (𝑥) 𝑑𝑥 = 𝑓(𝑥)𝑔(𝑥) − ∫ 𝑔(𝑥)𝑓 ′ (𝑥) 𝑑𝑥 ……(2)

Eq (2) gives the formula for integration by parts.

4
Figure 8.1: Basic Integration Rules (𝑎 > 0)

5
 Figure 8.2: Procedures for Fitting Integrands to Basic Rule

It is perhaps easier to remember in the following notation.

Let 𝑢 = 𝑓(𝑥) and 𝑣 = 𝑔(𝑥). Then the differentials are 𝑑𝑢 = 𝑓 ′ (𝑥) 𝑑𝑥 and 𝑑𝑣 = 𝑔′ (𝑥) 𝑑𝑥. So, by the
Substitution Rule, the formula for integration by parts becomes

∫ 𝑢 𝑑𝑣 = 𝑢𝑣 − ∫ 𝑣 𝑑𝑢

Example

1. Find ∫ 𝑥 sin 𝑥 𝑑𝑥

Solution

First method
Suppose we choose 𝑓(𝑥) = 𝑥 and 𝑔′ (𝑥) = sin 𝑥. Then 𝑓 ′ (𝑥) = 1 and 𝑔(𝑥) = − cos 𝑥. Note that for 𝑔, we
can choose any antiderivative of 𝑔′ . Thus, using the formula in Eq 2,

∫ 𝑥 sin 𝑥 𝑑𝑥 = 𝑓(𝑥)𝑔(𝑥) − ∫ 𝑔(𝑥)𝑓 ′ (𝑥)𝑑𝑥

= 𝑥(−cos 𝑥) − ∫(− cos 𝑥) 𝑑𝑥

= −𝑥 cos 𝑥 + ∫ cos 𝑥 𝑑𝑥

= −𝑥 cos 𝑥 + sin 𝑥 + 𝐶

6
Second method

Let:

𝑢=𝑥 𝑑𝑣 = sin 𝑥 𝑑𝑥

𝑑𝑢 = 𝑑𝑥 𝑣 = − cos 𝑥

u dv

∫ 𝑥 sin 𝑥 𝑑𝑥 = ∫ 𝑥 sin 𝑥 𝑑𝑥

u v u du

= 𝑥(−cos 𝑥) − ∫(− cos 𝑥) 𝑑𝑥

= −𝑥 cos 𝑥 + ∫ cos 𝑥 𝑑𝑥

= −𝑥 cos 𝑥 + sin 𝑥 + 𝐶

We can evaluate definite integrals by parts. By evaluating both sides of Eq 2 (formula for integration by
parts) between 𝑎 and 𝑏, assuming 𝑓 ′ and 𝑔′ are continuous, and using the Fundamental Theorem of
Calculus, we get
𝑏 𝑏 𝑏
′ (𝑥)𝑑𝑥
∫ 𝑓(𝑥)𝑔 = 𝑓(𝑥)𝑔(𝑥)] − ∫ 𝑔(𝑥) 𝑓 ′ (𝑥) 𝑑𝑥
𝑎 𝑎 𝑎

We could also use trigonometric identities to integrate certain combinations of trigonometric functions.

Example

2. Find ∫ sin5 𝑥 cos 2 𝑥 𝑑𝑥

Solution

We could convert cos 2 𝑥 to 1 − sin2 𝑥, but we would be left with an expression in terms of sin 𝑥 with no
extra cos 𝑥 factor. Therefore, we could separate a single sine factor and rewrite the remaining sin4 factor
in terms of cos 𝑥 factor.

sin5 𝑥 cos 2 𝑥 = ( sin2 𝑥 )2 cos2 𝑥 sin 𝑥

= (1 − cos2 𝑥 )2 cos2 𝑥 sin 𝑥

Substituting 𝑢 = cos 𝑥, then 𝑑𝑢 = − sin 𝑥 𝑑𝑥

∫ sin5 𝑥 cos2 𝑥 𝑑𝑥 = ∫(1 − cos 2 𝑥 )2 cos2 𝑥 sin 𝑥 𝑑𝑥

7
 = ∫(1 − 𝑢2 )2 𝑢2 (−𝑑𝑢)

= − ∫(𝑢2 − 2𝑢4 + 𝑢6 )𝑑𝑢

𝑢3 𝑢5 𝑢7
= −( − 2 + )+ 𝐶
3 5 7

1 2 1
= − cos 3 𝑥 + cos 5 𝑥 − cos7 𝑥 + 𝐶
3 5 7

We can use a similar strategy to evaluate integrals of the form ∫ tan𝑚 𝑥 sec 𝑛 𝑥 𝑑𝑥, where 𝑚 > 0, 𝑛 > 0 are
integers.
𝑑
Since (𝑑𝑥) tan 𝑥 = sec 2 𝑥, we could separate a sec 2 𝑥 factor and convert the remaining (even) power of
secant to an expression involving tangent using the following identity:

sec 2 𝑥 = 1 + tan2 𝑥

Another way is to separate a sec 𝑥 tan 𝑥 factor and convert the remaining (even) power of tangent to
𝑑
secant since ( ) sec 𝑥 = sec 𝑥 tan 𝑥 .
𝑑𝑥

As for other cases, the use of identities, integration by parts, and occasionally a little ingenuity may come
handy. The following formulas and trigonometric identities are also useful:

∫ tan 𝑥 𝑑𝑥 = ln|sec 𝑥| + 𝐶

∫ sec 𝑥 𝑑𝑥 = ln | sec 𝑥 + tan 𝑥 | + 𝐶

1
sin 𝐴 cos 𝐵 = [sin(𝐴 − 𝐵) + sin(𝐴 + 𝐵)]
2
1
sin 𝐴 sin 𝐵 = [cos(𝐴 − 𝐵) − cos(𝐴 + 𝐵)]
2
1
cos 𝐴 cos 𝐵 = [cos(𝐴 − 𝐵) + cos(𝐴 + 𝐵)]
2

8
8.2.2 TRIGONOMETRIC SUBSTITUTION

In finding the area of a circle or an ellipse, an integral of the form ∫ √𝑎2 − 𝑥 2 𝑑𝑥 arises, where 𝑎 > 0. If it
were ∫ 𝑥√𝑎2 − 𝑥 2 𝑑𝑥, the substitution 𝑢 = 𝑎2 − 𝑥 2 would be effective. However,

∫ √𝑎2 − 𝑥 2 𝑑𝑥 … … (3)

Eq (3) would be more challenging. If we change the variable from 𝑥 to 𝜃 by the substitution of 𝑥 = 𝑎 sin 𝜃.
Then, the root sign of Eq (3) can be removed by making use of the identity 1 − sin2 𝜃 = cos2 𝜃. This is
shown as below:

√𝑎2 − 𝑥 2 = √𝑎2 − 𝑎2 sin2 𝜃

= √𝑎2 (1 − sin2 𝜃

= √𝑎2 cos2 𝜃

= 𝑎| cos 𝜃 |

Notice the difference between the substitution 𝑢 = 𝑎2 − 𝑥 2 (in which the new variable is a function of the
old one) and the substitution 𝑥 = 𝑎 sin 𝜃 (the old variable is a function of the new one).

In general, we can make a substitution of the form 𝑥 = 𝑔(𝑡) by using the Substitution Rule in reverse. To
make our calculations simpler, we assume that g has an inverse function; that is, g is one-to-one.

In this case, if we replace 𝑢 by 𝑥 and 𝑥 by 𝑡 in the Substitution Rule, we get

∫ 𝑓(𝑥) 𝑑𝑥 = ∫ 𝑓(𝑔(𝑡))𝑔′ (𝑡)𝑑𝑡

This type of substitution is called the inverse substitution. The inverse substitution of 𝑥 = 𝑎 sin 𝜃 can be
made provided that it defines a one-to-one function. This can be accomplished by restricting 𝜃 to lie in the
𝜋 𝜋
interval [− , ]
2 2

Table 8.1 below shows a list of trigonometric substitutions which are effective for the given radical
expressions because of the specified trigonometric identities. The restriction on 𝜃 is imposed in each of the
cases shown in Table 8.1 to ensure that the function that defines the substitution is one-to-one.

9
 Table 8.1: Table of Trigonometric Substitution

Example

3. Evaluate
√9 − 𝑥 2
∫ 𝑑𝑥
𝑥2

Solution
Let 𝑥 = 3 sin 𝜃, where −𝜋/2 ≤ 𝜃 ≤ 𝜋/2. Then, 𝑑𝑥 = 3 cos 𝜃 𝑑𝜃.

√9 − 𝑥 2 = √9 − 9sin2 𝜃

= √9 cos 2 𝜃

= 3| cos 𝜃|

= 3 cos 𝜃

(Note that cos 𝜃 ≥ 0 because −𝜋/2 ≤ 𝜃 ≤ 𝜋/2.) Thus using the Inverse Substitution Rule:

√9 − 𝑥 2 3 cos 𝜃
∫ 2
𝑑𝑥 = ∫ 3 cos 𝜃 𝑑𝜃
𝑥 9sin2 𝜃

cos2 𝜃
=∫ 𝑑𝜃
sin2 𝜃

= ∫ cot 2 𝜃 𝑑𝜃

= ∫(csc 2 𝜃 − 1)𝑑𝜃

= − cot 𝜃 − 𝜃 + 𝐶

10
Since this is an indefinite integral, we must return to the original variable x. This can be done either by
using trigonometric identities to express cot 𝜃 in terms of sin 𝜃 = 𝑥/3 or by drawing a diagram, as in Figure
8.3, where 𝜃 is interpreted as an angle of a right triangle.

Figure 8.3: sin 𝜃 = 𝑥/3

Based on the Pythagorean Theorem, the length of the adjacent side can be expressed as

√9 − 𝑥 2

Then, we can simply read the value of cot 𝜃 from the figure:

√9 − 𝑥 2
cot 𝜃 =
𝑥
(Although 𝜃 > 0 in the diagram, this expression for cot 𝜃 is valid even when 𝜃 < 0). Since sin 𝜃 = 𝑥/3, then
𝜃 = sin−1 (𝑥/3). Therefore

√9 − 𝑥 2 √9 − 𝑥 2 −1
𝑥
∫ 𝑑𝑥 = − − sin ( )+𝐶
𝑥2 𝑥 3

8.2.3 PARTIAL FRACTIONS

This section demonstrates a method to integrate any rational function (a ratio of polynomials) by expressing
it as a sum of simpler fractions, called partial fractions, that we already know how to integrate.
2 1
To illustrate the method, observe that by taking the fractions and to a common denominator,
(𝑥−1) (𝑥+2)
the expression becomes

2 1 2(𝑥 + 2) − (𝑥 − 1)
− =
(𝑥 − 1) (𝑥 + 2) (𝑥 − 1)(𝑥 + 2)
𝑥+5
=
𝑥2 +𝑥−2
If we reverse the procedure, we see how to integrate the function on the right side of the following equation.

11
 𝑥+5 2 1
∫ 𝑑𝑥 = ∫ ( − ) 𝑑𝑥
𝑥2 +𝑥−2 (𝑥 − 1) (𝑥 + 2)

= 2 ln|𝑥 − 1| − ln|𝑥 + 2| + 𝐶

In order to illustrate the method, let

𝑃(𝑥)
𝑓(𝑥) =
𝑄(𝑥)

Be a rational function where 𝑃(𝑥) and 𝑄(𝑥) are polynomials. The function 𝑓(𝑥) can be expressed as a sum
of simpler fractions provided that the degree of P is less than the degree of Q. Such a rational function is
called proper.

If

𝑃(𝑥) = 𝑎𝑛 𝑥 𝑛 + 𝑎𝑛−1 𝑥 𝑛−1 + ⋯ + 𝑎1 𝑥 + 𝑎0

Where 𝑎𝑛 ≠ 0, then the degree of P is n and we write deg(P) = n. If f is improper, that is, deg(P)  deg(Q),
then we must take the preliminary step of dividing Q into P (by long division) until a remainder R (x) is
obtained such that deg(R) < deg(Q).

The division statement is

𝑃(𝑥) 𝑅(𝑥)
𝑓(𝑥) = = 𝑆(𝑥) + … . . (4)
𝑄(𝑥) 𝑄(𝑥)

where S and R are also polynomials.

As the next example illustrates, sometimes this preliminary step is all that is required.

Example

4. Find
𝑥3 + 𝑥
∫ 𝑑𝑥
𝑥−1

Solution

Since the degree of the numerator is greater than the degree of the denominator, we first perform the long
division.

𝑥3 + 𝑥 2
∫ 𝑑𝑥 = ∫ (𝑥 2 + 𝑥 + 2 + ) 𝑑𝑥
𝑥−1 𝑥−1

𝑥3 𝑥2
= + + 2𝑥 + 2 ln|𝑥 − 1| + 𝐶
3 2

12
From Eq (4), if the denominator is more complicated, then the next step is to factor the denominator Q
(x) as far as possible. It can be shown that any polynomial Q can be factored as a product of linear factors
(of the form ax + b) and irreducible quadratic factors (of the form 𝑎𝑥 2 + 𝑏𝑥 + 𝑐, where 𝑏 2 − 4𝑎𝑐 < 0.

For example, if 𝑄(𝑥) = 𝑥 4 − 16, we could factor it as

𝑄(𝑥) = (𝑥 2 − 4)(𝑥 2 + 4) = (𝑥 − 2)(𝑥 + 2)(𝑥 2 + 4)

Then, the next step is to express the proper rational function 𝑅(𝑥)/𝑄(𝑥) in Eq (4) as a sum of partial
fractions of the following form

𝐴 𝐴𝑥 + 𝐵
𝑜𝑟
(𝑎𝑥 + 𝑏)𝑖 (𝑎𝑥 2 + 𝑏𝑥 + 𝑐)𝑗

A theorem in algebra guarantees that it is always possible to do this. We explain the details for the four
cases that occur:

a) The denominator 𝑄(𝑥) is a product of distinct linear factors

b) 𝑄(𝑥) is a product of linear factors, some of which are repeated
c) 𝑄(𝑥) contains irreducible quadratic factors, none of which is repeated
d) 𝑄(𝑥) contains a repeated irreducible quadratic factor

8.2.3.1 CASE 1: Q(X) IS A PRODUCT OF DISTINCT LINEAR FACTORS

For this case, we could expressed 𝑄(𝑥) as

𝑄(𝑥) = (𝑎1 𝑥 + 𝑏1 )(𝑎2 𝑥 + 𝑏2 ) … (𝑎𝑘 𝑥 + 𝑏𝑘 ),

Where no factor is repeated and no factor is a constant multiple of another. Hence, in this case, the partial
fraction theorem states that there exist constant 𝐴1 , 𝐴2 , … , 𝐴𝑘 such that

𝑅(𝑥) 𝐴1 𝐴2 𝐴𝑘
= + +⋯+ … . (5)
𝑄(𝑥) 𝑎1 𝑥 + 𝑏1 𝑎2 𝑥 + 𝑏2 𝑎𝑘 𝑥 + 𝑏𝑘

Example

5. Evaluate
𝑥 2 + 2𝑥 − 1
∫ 𝑑𝑥
2𝑥 3 + 3𝑥 2 − 2𝑥

Solution

Since the degree of the numerator is less than the degree of the denominator, we don’t need to divide. We
factor the denominator as

2𝑥 3 + 3𝑥 2 − 2𝑥 = 𝑥(2𝑥 2 + 3𝑥 − 2)

= 𝑥(2𝑥 − 1)(𝑥 + 2)

13
Then,

𝑥 2 + 2𝑥 − 1 𝐴 𝐵 𝐶
= + +
𝑥(2𝑥 − 1)(𝑥 + 2) 𝑥 2𝑥 − 1 𝑥 + 2

In order to determine the constant 𝐴, 𝐵 and C, multiply both sides of the equation by the product of the
denominators to give

𝑥 2 + 2𝑥 − 1 = 𝐴(2𝑥 − 1)(𝑥 + 2) + 𝐵𝑥(𝑥 + 2) + 𝐶𝑥(2𝑥 − 1)

= (2𝐴 + 𝐵 + 2𝐶)𝑥 2 + (3𝐴 + 2𝐵 − 𝐶)𝑥 − 2𝐴

System of equations:

2𝐴 + 𝐵 + 2𝐶 = 1

3𝐴 + 2𝐵 − 𝐶 = 2

−2𝐴 = −1
1 1 1
Thus, 𝐴 = , 𝐵 = , 𝐶 = −
2 5 10

𝑥 2 + 2𝑥 − 1 11 1 1 1 1
∫ 3 2
𝑑𝑥 = ∫ ( + − ) 𝑑𝑥
2𝑥 + 3𝑥 − 2𝑥 2 𝑥 5 2𝑥 − 1 10 𝑥 + 2
1 1 1
= ln|𝑥| + ln|2𝑥 − 1| − ln|𝑥 + 2| + 𝐾
2 10 10

Note that in integrating the middle term, the following substitutions have been made:

1
𝑢 = 2𝑥 − 1, 𝑑𝑢 = 2𝑑𝑥, then 𝑑𝑥 = 𝑑𝑢
2

8.2.3.2 CASE 2: Q(X) IS A PRODUCT OF DISTINCT LINEAR FACTORS, SOME OF WHICH ARE
REPEATED

Suppose the first linear factor (𝑎1 𝑥 + 𝑏1 ) is repeated 𝑟 times, that is, (𝑎1 𝑥 + 𝑏1 )𝑟 occurs in the factorization
of 𝑄(𝑥). Then, instead of the single term in Eq (5), we could use

𝐴1 𝐴2 𝐴𝑟
+ 2
+⋯+ … . (6)
𝑎1 𝑥 + 𝑏1 (𝑎1 𝑥 + 𝑏1 ) (𝑎1 𝑥 + 𝑏1 )𝑟

14
Example

6. Find
𝑥 4 − 2𝑥 2 + 4𝑥 + 1
∫ 𝑑𝑥
𝑥3 − 𝑥2 − 𝑥 + 1

Solution

The first step is to divide using long division, which gives

𝑥 4 − 2𝑥 2 + 4𝑥 + 1 4𝑥
3 2
=𝑥+1+ 3 2
𝑥 −𝑥 −𝑥+1 𝑥 −𝑥 −𝑥+1

The next step is to factor out the denominator

𝑥 3 − 𝑥 2 − 𝑥 + 1 = (𝑥 − 1)(𝑥 2 − 1)

= (𝑥 − 1)(𝑥 − 1)(𝑥 + 1)

= (𝑥 − 1)2 (𝑥 + 1)

Hence,

4𝑥 𝐴 𝐵 𝐶
2
= + 2
+
(𝑥 − 1) (𝑥 + 1) 𝑥 − 1 (𝑥 − 1) 𝑥+1

4𝑥 = 𝐴(𝑥 − 1)(𝑥 + 1) + 𝐵(𝑥 + 1) + 𝐶(𝑥 − 1)2

Equate coefficients, we obtain 𝐴 = 1, 𝐵 = 2, 𝐶 = −1. Hence,

𝑥 4 − 2𝑥 2 + 4𝑥 + 1 1 2 1
∫ 3 2
𝑑𝑥 = ∫ [𝑥 + 1 + + 2
− ] 𝑑𝑥
𝑥 −𝑥 −𝑥+1 𝑥 − 1 (𝑥 − 1) 𝑥+1

𝑥2 2
= + 𝑥 + ln|𝑥 − 1| − − ln|𝑥 + 1| + 𝐾
2 𝑥−1
𝑥2 2 𝑥−1
= +𝑥− + ln | | +𝐾
2 𝑥−1 𝑥+1

15
8.2.3.3 CASE 3: Q(X) CONTAINS IRREDUCIBLE QUADRATIC FACTORS, NONE OF WHICH IS
REPEATED

If 𝑄(𝑥) has the factor 𝑎𝑥 2 + 𝑏𝑥 + 𝑐, where 𝑏 2 − 4𝑎𝑐 < 0, then, in addition to the partial fractions in Eq (5)
and (6), the expression for 𝑅(𝑥)/𝑄(𝑥) will have a term of the form

𝐴𝑥 + 𝐵
… . . (7)
𝑎𝑥 2 + 𝑏𝑥 + 𝑐
Where A and B are constants to be determined. The term in Eq (7) can be integrated by completing squares
(if necessary) and using the following formula

𝑑𝑥 1 𝑥
∫ = tan−1 ( ) + 𝐶
𝑥2 +𝑎 2 𝑎 𝑎

Example

7. Evaluate

4𝑥 2 − 3𝑥 + 2
∫ 𝑑𝑥
4𝑥 2 − 4𝑥 + 3

Solution

Since the degree of the numerator is not less than the degree of the denominator, we divide the
expression, which yield

4𝑥 2 − 3𝑥 + 2 𝑥−1
2
=1+ 2
4𝑥 − 4𝑥 + 3 4𝑥 − 4𝑥 + 3
Note that the quadratic 4𝑥 2 − 4𝑥 + 3 is irreducible because its discriminant 𝑏 2 − 4𝑎𝑐 = −32 < 0. Hence,
we complete the square

4𝑥 2 − 4𝑥 + 3 = (2𝑥 − 1)2 + 2

Let

1
𝑢 = 2𝑥 − 1, 𝑑𝑢 = 2 𝑑𝑥, 𝑥 = (𝑢 + 1)
2
Hence,

4𝑥 2 − 3𝑥 + 2 𝑥−1
∫ 2
𝑑𝑥 = ∫ (1 + 2 ) 𝑑𝑥
4𝑥 − 4𝑥 + 3 4𝑥 − 4𝑥 + 3
1
1 2 (𝑢 + 1) − 1
=𝑥+ ∫ 𝑑𝑢
2 𝑢2 + 2
1 𝑢−1
=𝑥+ ∫ 2 𝑑𝑢
4 𝑢 +2

16
 1 𝑢 1 1
=𝑥+ ∫ 2 𝑑𝑢 − ∫ 2 𝑑𝑢
4 𝑢 +2 4 𝑢 +2
1 1 1 𝑢
= 𝑥 + ln(𝑢2 + 2) − . tan−1 ( ) + 𝐶
8 4 √2 √2
1 1 2𝑥 − 1
= 𝑥 + ln(4𝑥 2 − 4𝑥 + 3) − tan−1 ( )+𝐶
8 4√2 √2

8.2.3.4 CASE 4: Q(X) CONTAINS A REPEATED IRREDUCIBLE QUADRATIC FACTOR

If 𝑄(𝑥) has the factor (𝑎𝑥 2 + 𝑏𝑥 + 𝑐)𝑟 , where 𝑏 2 − 4𝑎𝑐 < 0, then

𝐴1 𝑥 + 𝐵1 𝐴2 𝑥 + 𝐵2 𝐴𝑟 𝑥 + 𝐵𝑟
+ + ⋯ … . . (8)
𝑎𝑥 2 + 𝑏𝑥 + 𝑐 (𝑎𝑥 2 + 𝑏𝑥 + 𝑐)2 (𝑎𝑥 2 + 𝑏𝑥 + 𝑐)𝑟

occurs in the partial fraction decomposition of 𝑅(𝑥)/𝑄(𝑥). Each of the terms in Eq (8) can be integrated
using a substitution or by first completing the square if necessary.

Example

8. Evaluate

1 − 𝑥 + 2𝑥 2 − 𝑥 3
∫ 𝑑𝑥
𝑥(𝑥 2 + 1)2

Solution

The form of the partial fraction decomposition is

1 − 𝑥 + 2𝑥 2 − 𝑥 3 𝐴 𝐵𝑥 + 𝐶 𝐷𝑥 + 𝐸
2 2
= + 2 + 2
𝑥(𝑥 + 1) 𝑥 𝑥 + 1 (𝑥 + 1)2

Then

−𝑥 3 + 2𝑥 2 − 𝑥 + 1 = 𝐴(𝑥 2 + 1)2 + (𝐵𝑥 + 𝐶)𝑥(𝑥 2 + 1) + (𝐷𝑥 + 𝐸)𝑥

= (𝐴 + 𝐵)𝑥 4 + 𝐶𝑥 3 + (2𝐴 + 𝐵 + 𝐷)𝑥 2 + (𝐶 + 𝐸)𝑥 + 𝐴

Equating coefficients, we obtain 𝐴 = 1, 𝐵 = −1, 𝐶 = −1, 𝐷 = 1, 𝐸 = 0.

1 − 𝑥 + 2𝑥 2 − 𝑥 3 1 𝑥+1 𝑥
∫ 2 2
𝑑𝑥 = ∫ ( − 2 + 2 ) 𝑑𝑥
𝑥(𝑥 + 1) 𝑥 𝑥 + 1 (𝑥 + 1)2

𝑑𝑥 𝑥 𝑑𝑥 𝑥 𝑑𝑥
=∫ −∫ 2 𝑑𝑥 − ∫ 2 +∫ 2
𝑥 𝑥 +1 𝑥 +1 (𝑥 + 1)2

1 1
= ln|𝑥| − ln(𝑥 2 + 1) − tan−1 𝑥 − 2
+𝐾
2 2(𝑥 + 1)
17
8.3 IMPROPER INTEGRALS

In this sub section, extend the concept of a definite integral to the case where the interval is infinite and
also to the case where f has an infinite discontinuity in [a, b]. In either case the integral is called an
improper integral

8.3.1 TYPE 1: INFINITE INTERVALS

Consider the infinite region 𝒮 that lies under the curve 𝑦 = 1/𝑥 2 , above the x-axis, and to the right of line
𝑥 = 1. This is shown in Figure 8.4.

Figure 8.4: 𝑦 = 1/𝑥 2

You might think that since 𝒮 is infinite in extent, its area must be infinite, but let’s take a closer look. From
Figure 8.4, we can see that the area of the part of 𝒮 that lies to the left of the line 𝑥 = 𝑡 (shaded area) is
𝑡
1 1𝑡
𝐴(𝑡) = ∫ 2 𝑑𝑥 = − ]
1 𝑥 𝑥1

1
=1−
𝑡
Notice that 𝐴(𝑡) < 1 no matter how large 𝑡 is chosen. We also observe that

1
lim 𝐴(𝑡) = lim (1 − ) = 1
𝑡→∞ 𝑡→∞ 𝑡
This is shown in Figure 8.5.

Figure 8.5: Area under the curve as 𝑡 → ∞

18
Using this example as a guide, we define the integral of f over an infinite interval as the limit of integrals
over finite intervals. Figure 8.6 shows the definition of an improper integral of Type 1.

Figure 8.6: Definition of an improper integral of Type 1

Example
∞ 1
9. Determine whether the integral ∫1 (𝑥) 𝑑𝑥 is convergent or divergent

Solution
∞
1
∫ ( ) 𝑑𝑥 = lim ln|𝑥|]1𝑡 = lim (ln 𝑡 − ln 1)
1 𝑥 𝑡→∞ 𝑡→∞

= lim ln 𝑡 = ∞
𝑡→∞

The limit does not exist as a finite number and so the improper integral is divergent.

8.3.2 TYPE 2: DISCONTINUOUS INTEGRANDS

Suppose that 𝑓 is a positive continuous function defined on a finite interval [a, b) but has a vertical asymptote
at b. Let S be the unbounded region under the graph of f and above the x-axis between a and b. (For Type 1
integrals, the regions extended indefinitely in a horizontal direction. Here the region is infinite in a vertical
direction.)

19
The area of the part of S between a and t (the shaded region in Figure 8.7) is
𝑡
𝐴(𝑡) = ∫ 𝑓(𝑥)𝑑𝑥
𝑎

Figure 8.7: Area of the part of S

If it happen that 𝐴(𝑡) approaches a definite number 𝐴 as 𝑡 → 𝑏 − , then we say that the area of the region S
is 𝐴 and we could write
𝑏 𝑏
∫ 𝑓(𝑥)𝑑𝑥 = lim− ∫ 𝑓(𝑥)𝑑𝑥
𝑎 𝑡→𝑏 𝑎

We use this equation to define an improper integral of Type 2 even when 𝑓 is not a positive function, no
matter what type of discontinuity 𝑓 has at b. Figure 8.8 below presents the definition of an improper
Integral of Type 2.

Figure 8.8: Definition of an improper integral of Type 2

20
Example

10. Find
5
1
∫ 𝑑𝑥
2 √𝑥 − 2

Solution
1
The given integral is improper because 𝑓(𝑥) = 𝑥−2
has the vertical asymptote at 𝑥 = 2. Since the infinite
√
discontinuity occurs at the left endpoint of [2, 5], we use part (b) of the definition in Figure 8.8.
5 5
1 𝑑𝑥
∫ 𝑑𝑥 = lim+ ∫
2 √𝑥 − 2 𝑡→2 𝑡 √𝑥 − 2
5
= lim+2√𝑥 − 2]𝑡 = lim+(√3 − √𝑡 − 2)
𝑡→2 𝑡→2

= 2√3

Thus the given improper integral is convergent and, since the integrand is positive, we can interpret the
value of the integral as the area of the shaded region in Figure 8.9.

1
Figure 8.9: 𝑦 =
√𝑥−2

8.3.3 A COMPARISON TEST FOR IMPROPER INTEGRALS

Sometimes it is impossible to find the exact value of an improper integral and yet it is important to know
whether it is convergent or divergent.

In such cases the following theorem is useful. Although we state it for Type 1 integrals, a similar theorem is
true for Type 2 integrals.

21
We omit the proof of the Comparison Theorem, but Figure 8.10 makes it seem plausible.

Figure 8.9: Area under the curve

If the area under the top curve y = f (x) is finite, then so is the area under the bottom curve y = g (x).

If the area under y = g (x) is infinite, then so is the area under y = f (x). [Note that the reverse is not
∞ ∞ ∞
necessarily true: If ∫𝑎 𝑔(𝑥)𝑑𝑥 is convergent, ∫𝑎 𝑓(𝑥)𝑑𝑥 may or may not be convergent, and if ∫𝑎 𝑓(𝑥)𝑑𝑥
∞
is divergent, ∫𝑎 𝑔(𝑥)𝑑𝑥 may or may not be divergent.]

22
ENGINEERING APPLICATIONS OF
INTEGRALS
WEEK 9: ENGINEERING APPLICATIONS OF INTEGRALS
9.1 ARC LENGTH

What do we mean by the length of a curve? We might think of fitting a piece of string to the curve in Figure
9.1 and then measuring the string against a ruler. But that might be difficult to do with much accuracy if we
have a complicated curve.

We need a precise definition for the length of an arc of a curve, in the same spirit as the definitions we
developed for the concepts of area and volume.

Figure 9.1: Fitting a string to a curve

If the curve is a polygon, we can easily find its length; we just add the lengths of the line segments that form
the polygon. (We can use the distance formula to find the distance between the endpoints of each
segment).

We are going to define the length of a general curve by first approximating it by a polygon and then taking a
limit as the number of segments of the polygon is increased. This process is familiar for the case of a circle,
where the circumference is the limit of lengths of inscribed polygons (see Figure 9.2).

Figure 9.2: Circle with different segments of polygon

Suppose that a curve 𝐶 is defined by the equation 𝑦 = 𝑓(𝑥), where 𝑓 is continuous and 𝑎 ≤ 𝑥 ≤ 𝑏. We
obtain a polygonal approximation to 𝐶 by dividing the interval [a, b] into n subintervals with endpoints
𝑥0 , 𝑥1 , 𝑥2 , … . , 𝑥𝑛 and equal width ∆𝑥.

If 𝑦𝑖 = 𝑓(𝑥𝑖 ), then the point 𝑃𝑖 (𝑥𝑖 𝑦𝑖 ) lies on 𝐶 and the polygon with vertices 𝑃0 , 𝑃1 , … . , 𝑃𝑛 illustrated in
Figure 9.3, is an approximation to 𝐶

23
 Figure 9.3: Polygon with vertices 𝑃0 , 𝑃1 , … . , 𝑃𝑛

The length 𝐿 of 𝐶 is approximately the length of this polygon and the approximation gets better as we let n
increase. (See Figure 9.4, where the arc of the curve between 𝑃𝑖−1 and 𝑃𝑖 has been magnified and
approximations with successively smaller values of ∆𝑥 are shown.)

Figure 9.4: Arc of the curve between 𝑃𝑖−1 and 𝑃𝑖 has been magnified

Therefore we define the length 𝐿 of the curve 𝐶 with equation, 𝑦 = 𝑓(𝑥), 𝑎 ≤ 𝑥 ≤ 𝑏 as the limit of the
lengths of these inscribed polygons ( if the limit exists):
𝑛

𝐿 = lim ∑|𝑃𝑖−1 𝑃𝑖 | … . . (1)

𝑛→∞
𝑖=1

Note that the procedure for defining arc length is very similar to the procedure we used for defining area
and volume: We divided the curve into a large number of small parts. We then found the approximate
lengths of the small parts and added them. Finally, we took the limit as 𝑛 → ∞.

The definition of arc length given by Eq. (1) is not very convenient for computational purposes, but we can
derive an integral formula for 𝐿 in the case where 𝑓 has a continuous derivative. [Such a function 𝑓 is called
smooth because a small change in 𝑥 produces a small change in 𝑓 ′ (𝑥).]

24
Let ∆𝑦𝑖 = 𝑦𝑖 − 𝑦𝑖−1 . Then

|𝑃𝑖−1 𝑃𝑖 | = √(𝑥𝑖 − 𝑥𝑖−1 )2 + (𝑦𝑖 − 𝑦𝑖−1 )2 = √(∆𝑥)2 + (∆𝑦𝑖 )2

Applying the Mean Value Theorem of 𝑓 on the interval [𝑥𝑖−1 , 𝑥𝑖 ], we find that there is a number 𝑥𝑖∗
between 𝑥𝑖−1 and 𝑥𝑖 such that

𝑓(𝑥𝑖 ) − 𝑓(𝑥𝑖−1 ) = 𝑓 ′ (𝑥𝑖∗ )(𝑥𝑖 − 𝑥𝑖−1 )

∆𝑦𝑖 = 𝑓 ′ (𝑥𝑖∗ )∆𝑥

Thus, we have
|𝑃𝑖−1 𝑃𝑖 | = √(∆𝑥)2 + (∆𝑦𝑖 )2 = √(∆𝑥)2 + [𝑓 ′ (𝑥𝑖∗ )∆𝑥]2

= √1 + [𝑓 ′ (𝑥𝑖∗ )]2 √(∆𝑥)2 = √1 + [𝑓 ′ (𝑥𝑖∗ )]2 ∆𝑥,

Since ∆𝑥 > 0. Thus, by using the definition in Eq. (1),

𝑛 𝑛

𝐿 = lim ∑|𝑃𝑖−1 𝑃𝑖 | = lim ∑ √1 + [𝑓 ′ (𝑥𝑖∗ )]2 ∆𝑥

𝑛→∞ 𝑛→∞
𝑖=1 𝑖=1

We recognize this expression as being equal to

𝑏
∫ √1 + [𝑓 ′ (𝑥)]2 𝑑𝑥
𝑎

by the definition of a definite integral. We know that this integral exists because the function 𝑔(𝑥) =
√1 + [𝑓 ′ (𝑥)]2 is continuous. Thus we have proved the following theorem:

If we use Leibniz notation for derivatives, we can write the arc length formula as follows:

𝑏
𝑑𝑦 2
𝐿 = ∫ √1 + ( ) 𝑑𝑥 … … (3)
𝑎 𝑑𝑥

25
Example

1. Find the length of the arc of the semicubical parabola 𝑦 2 = 𝑥 3 between the points (1, 1) and (4, 8).
See Figure 9.5

Figure 9.5: 𝑦 2 = 𝑥 3

Solution

For the top half of the curve we have

𝑑𝑦 3
𝑦 = 𝑥 3/2 = 𝑥 1/2
𝑑𝑥 2

and so the arc length formula gives

4
𝑑𝑦 2 4
9
√
𝐿 = ∫ 1 + ( ) 𝑑𝑥 = ∫ √1 + 𝑥 𝑑𝑥
1 𝑑𝑥 1 4

9 9
If we substitute 𝑢 = 1 + 4 𝑥 , then 𝑑𝑢 = 4 𝑑𝑥

13
When x = 1, 𝑢 = 4
; when x = 4, u = 10. Therefore

10
4 10 4 2
𝐿 = ∫ √𝑢 𝑑𝑢 = ∙ 𝑢3/2 |
9 13/4 9 3 13/4

8 13 3/2
= [103/2 − ( ) ]
27 4

1
= (80√10 − 13√13)
27

If a curve has the equation x = g (y), c  y  d, and g  (y) is continuous, then by interchanging the roles of x
and y in Formula 2 or Eq. (3), we obtain the following formula for its length:

26
9.1.1 THE ARC LENGTH FUNCTION

We will find it useful to have a function that measures the arc length of a curve from a particular starting
point to any other point on the curve.

Thus if a smooth curve C has the equation, y = f (x), a  x  b let s (x) be the distance along C from the initial
point P0(a, f (a)) to the point Q (x, f (x)). Then s is a function, called the arc length function, and, by Formula
2,
𝑥
𝑠(𝑥) = ∫ √1 + [𝑓 ′ (𝑡)]2 𝑑𝑡 … (5)
𝑎

(We have replaced the variable of integration by t so that x does not have two meanings.) We can use the
Fundamental Theorem of Calculus to differentiate Eq. (5) (since the integrand is continuous):

𝑑𝑠 𝑑𝑦 2
= √1 + [𝑓 ′ (𝑥)]2 = √1 + ( ) … (6)
𝑑𝑥 𝑑𝑥

Eq. (6) shows that the rate of change of s with respect to x is always at least 1 and is equal to 1 when f(x),
the slope of the curve, is 0. The differential of arc length is

𝑑𝑦 2
𝑑𝑠 = √1 + ( ) 𝑑𝑥 … (7)
𝑑𝑥

and this equation is sometimes written in the symmetric form

(𝑑𝑠)2 = (𝑑𝑥)2 + (𝑑𝑦)2 … (8)

The geometric interpretation of Eq. (8) is shown in Figure 9.6. It can be used as a mnemonic
device for remembering both of the Formulas 3 and 4.

Figure 9.6: Geometric interpretation of Eq. (6)

If we write 𝐿 = ∫ 𝑑𝑠, then from Eq. (8) either we can solve to get (7), which gives (3), or we can solve to get

27
 𝑑𝑥 2
𝑑𝑠 = √1 + ( ) 𝑑𝑦
𝑑𝑦

which gives (4).

Example

Find the arc length function for the curve 𝑦 = 𝑥 2 − 1⁄8 ln 𝑥 taking P0(1, 1) as the starting point.

Solution
1
If 𝑓(𝑥) = 𝑥 2 − 1⁄8 ln 𝑥 , then 𝑓 ′ (𝑥) = 2𝑥 − 8𝑥

1 2 1 1
1 + [𝑓 ′ (𝑥)]2 = 1 + (2𝑥 − ) = 1 + 4𝑥 2 − +
8𝑥 2 64𝑥 2

1 1 1 2
= 4𝑥 2 + + = (2𝑥 + )
2 64𝑥 2 8𝑥
1
√1 + [𝑓 ′ (𝑥)]2 = 2𝑥 + ,
8𝑥
Since 𝑥 > 0. Thus the arc length function is given by
𝑥
𝑠(𝑥) = ∫ √1 + [𝑓 ′ (𝑡)]2 𝑑𝑡
1

𝑥
1 𝑥
= ∫ (2𝑡 + ) 𝑑𝑡 = 𝑡 2 + 1⁄8 ln 𝑡|
1 8𝑡 1

= 𝑥 2 + 1⁄8 ln 𝑥 − 1

For instance, the arc length along the curve from (1, 1) to (3, f (3)) is

ln 3
𝑠(3) = 32 + 1⁄8 ln 3 − 1 = 8 + ≈ 8.1373
8

9.2 AREA OF A SURFACE OF REVOLUTION

A surface of revolution is formed when a curve is rotated about a line. Such a surface is the lateral boundary
of a solid of revolution.

We want to define the area of a surface of revolution in such a way that it corresponds to our intuition.

If the surface area is A, we can imagine that painting the surface would require the same amount of paint
as does a flat region with area A.

28
Let’s start with some simple surfaces. The lateral surface area of a circular cylinder with radius r and height
h is taken to be A = 2 rh because we can imagine cutting the cylinder and unrolling it (as in Figure 9.7) to
obtain a rectangle with dimensions 2 r and h.

Figure 9.7: Lateral surface area of circular cylinder

Likewise, we can take a circular cone with base radius r and slant height l, cut it along the dashed line in
2𝜋𝑟
Figure 9.8, and flatten it to form a sector of a circle with radius l and central angle 𝜃 = 𝑙
.

Figure 9.8: Circular cone

1 2
In general, the area of a sector of a circle with radius l and angle  is 2
𝑙 𝜃 and so in this case the area is

1 1 2𝜋𝑟
𝐴 = 𝑙2 𝜃 = 𝑙2 ( ) = 𝜋𝑟𝑙
2 2 𝑙

Therefore we define the lateral surface area of a cone to be 𝐴 = 𝜋𝑟𝑙. What about more complicated
surfaces of revolution? If we follow the strategy we used with arc length, we can approximate the original
curve by a polygon. When this polygon is rotated about an axis, it creates a simpler surface whose surface
area approximates the actual surface area.

By taking a limit, we can determine the exact surface area. The approximating surface, then, consists of a
number of bands, each formed by rotating a line segment about an axis. To find the surface area, each of
these bands can be considered a portion of a circular cone, as shown in Figure 9.9.
29
 Figure 9.8: A portion of circular cone

The area of the band (or frustum of a cone) with slant height l and upper and lower radii r1 and r2 is found
by subtracting the areas of two cones:

𝐴 = 𝜋𝑟2 (𝑙1 + 𝑙) − 𝜋𝑟1 𝑙1 = 𝜋[(𝑟2 − 𝑟1 )𝑙1 + 𝑟2 𝑙] … (1)

From similar triangles we have

𝑙1 𝑙1 + 𝑙
=
𝑟1 𝑟2

Which gives

𝑟2 𝑙1 = 𝑟1 𝑙1 + 𝑟1 𝑙 or (𝑟2 − 𝑟1 )𝑙1 = 𝑟1 𝑙

Putting this in Equation 1, we get

𝐴 = 𝜋(𝑟1 𝑙 + 𝑟2 𝑙) or

where r = ½ (r1 + r2) is the average radius of the band. Now we apply this formula to our strategy. Consider
the surface shown in Figure 9.9, which is obtained by rotating the curve y = f (x), a  x  b, about the x-axis,
where f is positive and has a continuous derivative.

30
 (a) Surface of revolution (b) Approximating band

Figure 9.9

In order to define its surface area, we divide the interval [a, b] into n subintervals with endpoints x0, x1, . . . ,
xn and equal width x, as we did in determining arc length. If yi = f (xi ), then the point Pi(xi, yi ) lies on the
curve.

The part of the surface between xi – 1 and xi is approximated by taking the line segment Pi – 1Pi and rotating it
about the x-axis. The result is a band with slant height l = | Pi – 1Pi | and average radius r = 1/2(yi – 1 + yi) so,
by Formula 2, its surface area is

𝑦𝑖−1 + 𝑦𝑖
2𝜋 |𝑃𝑖−1 𝑃𝑖 |
2
As in the proof, We have

2
|𝑃𝑖−1 𝑃𝑖 | = √1 + [𝑓 ′ (𝑥𝑖∗ )] ∆𝑥

Where 𝑥𝑖∗ is some number in [xi – 1, xi ]. When x is small, we have yi = f (xi)  f (xi) and also yi – 1 = f (xi – 1)  f
(xi), since f is continuous. Therefore

𝑦𝑖−1 + 𝑦𝑖 2
2𝜋 |𝑃𝑖−1 𝑃𝑖 | ≈ 2𝜋𝑓(𝑥𝑖∗ )√1 + [𝑓 ′ (𝑥𝑖∗ )] ∆𝑥
2
and so an approximation to what we think of as the area of the complete surface of revolution is
𝑛
2
∑ 2𝜋𝑓(𝑥𝑖∗ )√1 + [𝑓 ′ (𝑥𝑖∗ )] ∆𝑥 … (3)
𝑖=1

This approximation appears to become better as n  ∞ and, recognizing (3) as a Riemann sum for the
function 𝑔(𝑥) = 2𝜋𝑓(𝑥)√1 + [𝑓 ′ (𝑥)]2 , we have
𝑛 𝑏
2
lim ∑ 2𝜋𝑓(𝑥𝑖∗ )√1 + [𝑓 ′ (𝑥𝑖∗ )] ∆𝑥 = ∫ 2𝜋𝑓(𝑥)√1 + [𝑓 ′ (𝑥)]2 𝑑𝑥
𝑛→∞ 𝑎
𝑖=1

31
Therefore, in the case where f is positive and has a continuous derivative, we define the surface area of the
surface obtained by rotating the curve y = f (x), a  x  b, about the x-axis as

With the Leibniz notation for derivatives, this formula becomes

If the curve is described as x = g(y), c  y  d, then the formula for surface area becomes

Now both Formulas 5 and 6 can be summarized symbolically, using the notation for arc length, as

For rotation about the y-axis, the surface area formula becomes

where, as before, we can use either

𝑑𝑦 2 𝑑𝑥 2
𝑑𝑠 = √1 + ( ) 𝑑𝑥 or 𝑑𝑠 = √1 + ( ) 𝑑𝑦
𝑑𝑥 𝑑𝑦

These formulas can be remembered by thinking of 2 y or 2 x as the circumference of a circle traced out
by the point (x, y) on the curve as it is rotated about the x-axis or y-axis, respectively (see Figure 9.10).

32
 (a) Rotation about x-axis: S = ∫ 2 y ds (b) Rotation about y-axis: S = ∫ 2 x ds

Figure 9.10

Example

The curve 𝑦 = √4 − 𝑥 2 , − 1 ≤ 𝑥 ≤ 1, is an arc of the circle 𝑥 2 + 𝑦 2 = 4. Find the area of the surface
obtained by rotating this arc about the x-axis. (The surface is a portion of a sphere of radius 2. See Figure
9.11.)

Figure 9.11: Rotating arc of the circle about the x-axis

Solution

We have

𝑑𝑦 1 1 −𝑥
= (4 − 𝑥 2 )−2 (−2𝑥) =
𝑑𝑥 2 √4 − 𝑥 2

and so, by Formula 5, the surface area is

1
𝑑𝑦 2
√
𝑆 = ∫ 2𝜋𝑦 1 + ( ) 𝑑𝑥
−1 𝑑𝑥

1
𝑥2
= 2𝜋 ∫ √4 − 𝑥 2 √1 + 𝑑𝑥
−1 4 − 𝑥2
33
 1
4 − 𝑥2 + 𝑥2
= 2𝜋 ∫ √4 − 𝑥 2 √ 𝑑𝑥
−1 4 − 𝑥2

1 1
2
= 2𝜋 ∫ √4 − 𝑥 2 𝑑𝑥 = 4𝜋 ∫ 1 𝑑𝑥
−1 √4 − 𝑥 2 −1

= 4𝜋(2) = 8𝜋

9.3 HYDROSTATIC PRESSURE AND FORCE

Deep-sea divers realize that water pressure increases as they dive deeper. This is because the weight of the
water above them increases.

In general, suppose that a thin horizontal plate with area A square meters is submerged in a fluid of density
 kilograms per cubic meter at a depth d meters below the surface of the fluid as in Figure 9.12.

Figure 9.12: Surface of fluid

The fluid directly above the plate has volume V = Ad, so its mass is m = V =  Ad. The force exerted by the
fluid on the plate is therefore

𝐹 = 𝑚𝑔 = 𝜌𝑔𝐴𝑑

where g is the acceleration due to gravity. The pressure P on the plate is defined to be the force per unit
area:

𝐹
𝑃= = 𝜌𝑔𝑑
𝐴
The SI unit for measuring pressure is a newton per square meter, which is called a pascal
(abbreviation: 1 N/m2 = 1 Pa).

Since this is a small unit, the kilopascal (kPa) is often used. For instance, because the density of water is
 = 1000 kg/m3, the pressure at the bottom of a swimming pool 2 m deep is

P =  gd = 1000 kg/m3  9.8 m/s2  2 m

= 19,600 Pa

= 19.6 kPa
34
An important principle of fluid pressure is the experimentally verified fact that at any point in a liquid the
pressure is the same in all directions. (A diver feels the same pressure on nose and both ears.). Thus the
pressure in any direction at a depth d in a fluid with mass density  is given by

𝑃 = 𝜌𝑔𝑑 = 𝛿𝑑 … (9)

This helps us determine the hydrostatic force against a vertical plate or wall or dam in a fluid. This is not a
straightforward problem because the pressure is not constant but increases as the depth increases.

Example

A dam has the shape of the trapezoid shown in Figure 9.13. The height is 20 m and the width is 50 m at the
top and 30 m at the bottom. Find the force on the dam due to hydrostatic pressure if the water level is 4 m
from the top of the dam.

Figure 9.13: A dam with shape of trapezoid

Solution

We choose a vertical x-axis with origin at the surface of the water and directed downward as in Figure 3(a).

Figure 9.14: A dam with shape of trapezoid

The depth of the water is 16 m, so we divide the interval [0, 16] into subintervals of equal length with
endpoints xi and we choose xi*  [xi – 1, xi]. The ith horizontal strip of the dam is approximated by a
rectangle with height x and width wi, where, from similar triangles in Figure 9.14,

𝑎 10
∗ =
16 − 𝑥𝑖 20

Or

16 − 𝑥𝑖∗ 𝑥𝑖∗
𝑎= =8−
2 2

35
and so wi = 2(15 + a)

𝑥𝑖∗
𝑤𝑖 = 2(15 + 𝑎) = 2 (15 + 8 − )
2

= 46 − 𝑥𝑖∗

If Ai is the area of the i th strip, then

𝐴𝑖 ≈ 𝑤𝑖 ∆𝑥 = (46 − 𝑥𝑖∗ )∆𝑥

If x is small, then the pressure Pi on the i th strip is almost constant and we can use Eq. (9) to write

𝑃𝑖 ≈ 1000𝑔𝑥𝑖∗

The hydrostatic force Fi acting on the i th strip is the product of the pressure and the area:

𝐹𝑖 = 𝑃𝑖 𝐴𝑖 ≈ 1000𝑔𝑥𝑖∗ (46 − 𝑥𝑖∗ )∆𝑥

Adding these forces and taking the limit as n  ∞ , we obtain the total hydrostatic force on the dam:
𝑛 16
𝐹 = lim ∑ 1000𝑔𝑥𝑖∗ (46 − 𝑥𝑖∗ )∆𝑥 = ∫ 1000𝑔𝑥(46 − 𝑥)𝑑𝑥
𝑛→∞ 0
𝑖=1

16 16
2 )𝑑𝑥
𝑥3
= 1000(9.8) ∫ (46𝑥 − 𝑥 = 9000 [23𝑥 − ]| ≈ 4.43 𝑥 107 𝑁
2
0 3 0

9.4 MOMENTS AND CENTRES OF MASS

Our main objective here is to find the point P on which a thin plate of any given shape balances horizontally
as in Figure 9.15. This point is called the center of mass (or center of gravity) of the plate.

Figure 9.15: Centre of mass of the plate

We first consider the simpler situation illustrated in Figure 9.16, where two masses m1 and m2 are attached
to a rod of negligible mass on opposite sides of a fulcrum and at distances d1 and d2 from the fulcrum.

36
 Figure 9.16: Simple illustration of the centre of mass of the plate

The rod will balance if

𝑚1 𝑑1 = 𝑚2 𝑑2 … . (10)

This is an experimental fact discovered by Archimedes and called the Law of the Lever. (Think of a lighter
person balancing a heavier one on a seesaw by sitting farther away from the center.)

Now suppose that the rod lies along the x-axis with m1 at x1 and m2 at x2 and the center of mass at 𝑥̅ . If we
compare Figures 9.16 and 9.17, we see that 𝑑1 = 𝑥̅ − 𝑥1 and 𝑑2 = 𝑥2 − 𝑥̅ and so Eq. (10) gives

𝑚1 (𝑥̅ − 𝑥1 ) = 𝑚2 (𝑥2 − 𝑥̅ )

𝑚1 𝑥̅ + 𝑚2 𝑥̅ = 𝑚1 𝑥1 + 𝑚2 𝑥2
𝑚1 𝑥1 + 𝑚2 𝑥2
𝑥̅ = … . (11)
𝑚1 + 𝑚2

Figure 9.17: Simple illustration of the centre of mass of the plate

The numbers m1x1 and m2x2 are called the moments of the masses m1 and m2 (with respect to the origin),
and Eq. (11) says that the center of mass 𝑥̅ is obtained by adding the moments of the masses and dividing
by the total mass m = m1 + m2.

In general, if we have a system of n particles with masses m1, m2, . . . , mn located at the points x1, x2, . . . , xn
on the x-axis, it can be shown similarly that the center of mass of the system is located at

∑𝑛
𝑖=1 𝑚𝑖 𝑥𝑖 ∑𝑛
𝑖=1 𝑚𝑖 𝑥𝑖
𝑥̅ = ∑𝑛
= … . (12)
𝑖=1 𝑚𝑖 𝑚

where m =  mi is the total mass of the system, and the sum of the individual moments

37
 𝑛

𝑀 = ∑ 𝑚𝑖 𝑥𝑖
𝑖=1

is called the moment of the system about the origin. Then Eq. (12) could be rewritten as 𝑚𝑥̅ = 𝑀, which
says that if the total mass were considered as being concentrated at the center of mass 𝑥̅ , then its moment
would be the same as the moment of the system. Now we consider a system of n particles with masses m1,
m2, . . . , mn located at the points (x1, y1), (x2, y2), . . . , (xn, yn) in the xy-plane as shown
in Figure 9.18.

Figure 9.18: A system of n particles with masses m1, m2, . . . , mn

By analogy with the one-dimensional case, we define the moment of the system about the y-axis to be
𝑛

𝑀𝑦 = ∑ 𝑚𝑖 𝑥𝑖 … (13)
𝑖=1

and the moment of the system about the x-axis as

𝑛

𝑀𝑥 = ∑ 𝑚𝑖 𝑦𝑖 … (14)
𝑖=1

Then My measures the tendency of the system to rotate about the y-axis and Mx measures the tendency to
rotate about the x-axis. As in the one-dimensional case, the coordinates (𝑥,
̅ 𝑦̅) of the center of mass are
given in terms of the moments by the formulas

𝑀𝑦 𝑀𝑥
𝑥̅ = 𝑦̅ = … (15)
𝑚 𝑚

where m =  mi is the total mass. Since 𝑚𝑥̅ = 𝑀𝑦 and 𝑚𝑦̅ = 𝑀𝑥 , the center of mass (𝑥,
̅ 𝑦̅) is the point
where a single particle of mass m would have the same moments as the system.

Example

Find the moments and center of mass of the system of objects that have masses 3, 4, and 8 at the points (–
1, 1), (2, –1), and (3, 2), respectively.

38
Solution

We use Eq. (13) and (14) to compute the moments:

𝑀𝑦 = 3(−1) + 4(2) + 8(3) = 29

𝑀𝑥 = 3(1) + 4(−1) + 8(2) = 15

Since m = 3 + 4 + 8 = 15, we use Equations 15 to obtain

𝑀𝑦 29
𝑥̅ = =
𝑚 15
𝑀𝑥 15
𝑦̅ = = =1
𝑚 15
14
Thus the center of mass is (1 , 1) (See Figure 9.19)
15

14
Figure 9.19: Centre of mass (1 15 , 1)

Next we consider a flat plate (called a lamina) with uniform density  that occupies a region R of the plane.
We wish to locate the center of mass of the plate, which is called the centroid of R.

In doing so we use the following physical principles: The symmetry principle says that if R is symmetric
about a line l, then the centroid of R lies on l. (If R is reflected about l, then R remains the same so its
centroid remains fixed. But the only fixed points lie on l.). Thus the centroid of a rectangle is its center.

Moments should be defined so that if the entire mass of a region is concentrated at the center of mass,
then its moments remain unchanged. Also, the moment of the union of two nonoverlapping regions should
be the sum of the moments of the individual regions.

Suppose that the region R is of the type shown in Figure 9.20 (a) ; that is, R lies between the lines x = a and
x = b, above the x-axis, and beneath the graph of f, where f is a continuous function.

Figure 9.20 (a): Region R

39
We divide the interval [a, b] into n subintervals with endpoints x0, x1, . . . , xn and equal width x. We choose
the sample point xi* to be the midpoint 𝑥,̅ of the i th subinterval, that is, 𝑥,
̅ = (𝑥𝑖−1 + 𝑥𝑖 )/2.

This determines the polygonal approximation to R shown in Figure 9.20(b).

Figure 9.20 (b): Polygonal approximation to R

1
The centroid of the i th approximating rectangle Ri is its center 𝐶𝑖 (𝑥̅𝑖 , 2 𝑓(𝑥̅ )). Its area is 𝑓(𝑥̅𝑖 )∆𝑥, so its
mass is

𝜌𝑓(𝑥̅𝑖 )∆𝑥

The moment of Ri about the y-axis is the product of its mass and the distance from Ci to the y-axis, which is
𝑥̅𝑖 . Thus

𝑀𝑦 (𝑅𝑖 ) = [𝜌𝑓(𝑥̅𝑖 )∆𝑥]𝑥̅𝑖 = 𝜌𝑥̅𝑖 𝑓(𝑥̅𝑖 )∆𝑥

Adding these moments, we obtain the moment of the polygonal approximation to R, and then by taking the
limit as n  ∞, we obtain the moment of R itself about the y-axis:

In a similar fashion we compute the moment of Ri about the x-axis as the product of its mass and the
distance from Ci to the x-axis:

1 1
𝑀𝑥 (𝑅𝑖 ) = [𝜌𝑓(𝑥̅𝑖 )∆𝑥] 𝑓(𝑥̅𝑖 ) = 𝜌 ∙ [𝑓(𝑥̅𝑖 )]2 ∆𝑥
2 2
Again we add these moments and take the limit to obtain the moment of R about the x-axis:

40
Just as for systems of particles, the center of mass of the plate is defined so that 𝑚𝑥̅ = 𝑀𝑦 and 𝑚𝑦̅ = 𝑀𝑥 .
But the mass of the plate is the product of its density and its area:
𝑏
𝑚 = 𝜌𝐴 = 𝜌 ∫ 𝑓(𝑥)𝑑𝑥
𝑎

And so

Notice the cancellation of the ’s. The location of the center of mass is independent of the density. In
summary, the center of mass of the plate (or the centroid of R) is located at the point (𝑥̅ , 𝑦̅), where

… (16)

If the region R lies between two curves y = f (x) and y = g (x), where f (x)  g (x), as illustrated in Figure 9.21,
then the same sort of argument that led to Formulas 16 can be used to show that the centroid of R is (𝑥̅ , 𝑦̅),
where

…. (17)

Figure 9.21: Region R lies between two curves

41
We end this section by showing a surprising connection between centroids and volumes of revolution.

Example

A torus is formed by rotating a circle of radius r about a line in the plane of the circle that is a distance R (>
r) from the center of the circle. Find the volume of the torus.

Solution

The circle has area A =  r2. By the symmetry principle, its centroid is its center and so the distance traveled
by the centroid during a rotation is d = 2 R.

Therefore, by the Theorem of Pappus, the volume of the torus is

𝑉 = 𝐴𝑑 = (2𝜋𝑅)(𝜋𝑟 2 ) = 2𝜋 2 𝑟 2 𝑅

42
MULTIPLE INTEGRALS
WEEK 10 : MULTIPLE INTEGRALS
10.1 DOUBLE INTEGRALS OVER RECTANGLES
10.1.1 REVIEW OF DEFINITE INTEGRALS

First let’s recall the basic facts concerning definite integrals of functions of a single variable. If f (x) is defined
for a  x  b, we start by dividing the interval [a, b] into n subintervals [xi – 1, xi] of equal width
x = (b – a)/n and we choose sample points 𝑥𝑖∗ in these subintervals. Then we form the Riemann sum
𝑛

∑ 𝑓(𝑥𝑖∗ )∆𝑥 … (1)

𝑖=1

and take the limit of such sums as n ∞ to obtain the definite integral of f from a to b:

𝑏 𝑛

∫ 𝑓(𝑥)𝑑𝑥 = lim ∑ 𝑓(𝑥𝑖∗ )∆𝑥 … (2)

𝑎 𝑛→∞
𝑖=1

In the special case where f (x)  0, the Riemann sum can be interpreted as the sum of the areas of the
𝑏
approximating rectangles in Figure 10.1, and ∫𝑎 𝑓(𝑥)𝑑𝑥 represents the area under the curve y = f (x) from a
to b.

Figure 10.1: Riemann sum approximation

10.1.2 VOLUMES AND DOUBLE INTEGRALS

In a similar manner we consider a function f of two variables defined on a closed rectangle

𝑅 = [𝑎, 𝑏] × [𝑐, 𝑑] = {(𝑥, 𝑦) ∈ ℝ2 | 𝑎 ≤ 𝑥 ≤ 𝑏, 𝑐 ≤ 𝑦 ≤ 𝑑}

and we first suppose that f (x, y)  0. The graph of f is a surface with equation z = f (x, y).

43
Let S be the solid that lies above R and under the graph of f, that is,

𝑆 = {(𝑥, 𝑦, 𝑧) ∈ ℝ3 | 0 ≤ 𝑧 ≤ 𝑓(𝑥, 𝑦), (𝑥, 𝑦) ∈ ℝ (See Figure 10.2.)

Figure 10.2: f is a surface with equation z = f (x, y).

Our goal is to find the volume of S. The first step is to divide the rectangle R into subrectangles. We
accomplish this by dividing the interval [a, b] into m subintervals [xi – 1, xi] of equal width x = (b – a)/m and
dividing [c, d ] into n subintervals [yj – 1, yj] of equal width y = (d – c)/n.

By drawing lines parallel to the coordinate axes through the endpoints of these subintervals, as in Figure 3,
we form the subrectangles

𝑅𝑖𝑖 = [𝑥𝑖−1 , 𝑥𝑖 ] × [𝑦𝑖−1 , 𝑦𝑖 ] = {(𝑥, 𝑦) |𝑥𝑖−1 ≤ 𝑥 ≤ 𝑥𝑖 , 𝑦𝑖−1 ≤ 𝑦 ≤ 𝑦𝑖 }

each with area A = x y.

Figure 10.3: Dividing R into subrectangles

If we choose a sample point (𝑥𝑖𝑗∗ , 𝑦𝑖𝑗∗ ) in each Rij, then we can approximate the part of S that lies above
each Rij by a thin rectangular box (or “column”) with base Rij and height (𝑥𝑖𝑗∗ , 𝑦𝑖𝑗∗ ) as shown in Figure 10.4.

44
The volume of this box is the height of the box times the area of the base rectangle:

𝑓(𝑥𝑖𝑗∗ , 𝑦𝑖𝑗∗ )∆𝐴

Figure 10.4: Approximation by a thin rectangular box

If we follow this procedure for all the rectangles and add the volumes of the corresponding boxes, we get
an approximation to the total volume of S:

This double sum means that for each subrectangle we evaluate f at the chosen point and multiply by the
area of the subrectangle, and then we add the results. (See Figure 10.5.)

Figure 10.5: Approximation to the total volume of S

45
Our intuition tells us that the approximation given in Eq. (3) becomes better as m and n become larger and
so we would expect that

We use the expression in Eq. (4) to define the volume of the solid S that lies under the graph of f and above
the rectangle R. Limits of the type that appear in Equation 4 occur frequently, not just in finding volumes
but in a variety of other situations even when f is not a positive function. So we make the following
definition.

The precise meaning of the limit in Definition 5 is that for every number ε > 0 there is an integer N such that

𝑚 𝑛

|∬ 𝑓(𝑥, 𝑦) 𝑑𝐴 − ∑ ∑ 𝑓(𝑥𝑖𝑗∗ , 𝑦𝑖𝑗∗ ) ∆𝐴| < 𝜀

𝑅 𝑖=1 𝑗=1

for all integers m and n greater than N and for any choice of sample points (𝑥𝑖𝑗∗ , 𝑦𝑖𝑗∗ ) in Rij. A function f is
called integrable if the limit in Definition 5 exists.

It is shown in courses on advanced calculus that all continuous functions are integrable. In fact, the double
integral of f exists provided that f is “not too discontinuous.” In particular, if f is bounded on R, [that is,
there is a constant M such that | f (x, y) |  M for all (x, y) in R], and f is continuous there, except on a finite
number of smooth curves, then f is integrable over R.

The sample point (𝑥𝑖𝑗∗ , 𝑦𝑖𝑗∗ ) can be chosen to be any point in the subrectangle Rij, but if we choose it to be
the upper right-hand corner of Rij [namely (xi, yj), see Figure 10.3], then the expression for the double
integral looks simpler:

By comparing Definitions 4 and 5, we see that a volume can be written as a double integral:

46
The sum in Definition 5,
𝑚 𝑛

∑ ∑ 𝑓(𝑥𝑖𝑗∗ , 𝑦𝑖𝑗∗ ) ∆𝐴
𝑖=1 𝑗=1

is called a double Riemann sum and is used as an approximation to the value of the double integral. [Notice
how similar it is to the Riemann sum in (1) for a function of a single variable.]. If f happens to be a positive
function, then the double Riemann sum represents the sum of volumes of columns, as in Figure 10.5, and is
an approximation to the volume under the graph of f.

10.1.3 ITERATED INTEGRALS

Suppose that f is a function of two variables that is integrable on the rectangle R = [a, b]  [c, d ].
𝑑
We use the notation ∫𝑐 𝑓(𝑥, 𝑦)𝑑𝑦 to mean that x is held fixed and f (x, y) is integrated with respect to y
from y = c to y = d. This procedure is called partial integration with respect to y. (Notice its similarity to
partial differentiation.)
𝑑
Now ∫𝑐 𝑓(𝑥, 𝑦)𝑑𝑦 is a number that depends on the value of x, so it defines a function of x:

𝑑
𝐴(𝑥) = ∫ 𝑓(𝑥, 𝑦)𝑑𝑦
𝑐

If we now integrate the function A with respect to x from x = a to x = b, we get

The integral on the right side of Eq. (7) is called an iterated integral. Usually the brackets are omitted. Thus

means that we first integrate with respect to y from c to d and then with respect to x from a to b. Similarly,
the iterated integral

means that we first integrate with respect to x (holding y fixed) from x = a to x = b and then we integrate
the resulting function of y with respect to y from y = c to y = d. Notice that in both Eq. (8) and (9) we work
from the inside out.

47
The following theorem gives a practical method for evaluating a double integral by expressing it as an
iterated integral (in either order).

In the special case where f (x, y) can be factored as the product of a function of x only and a function of y
only, the double integral of f can be written in a particularly simple form. To be specific, suppose that f (x, y)
= g (x)h (y) and R = [a, b]  [c, d]. Then Fubini’s Theorem gives

In the inner integral, y is a constant, so h(y) is a constant and we can write

𝑏
since ∫𝑎 𝑔(𝑥)𝑑𝑥 is a constant. Therefore, in this case, the double integral of f can be written as the product
of two single integrals:

48
Example

Find the volume of the solid that lies above the square R = [0, 2]  [0, 2] and below the elliptic paraboloid z
= 16 – x2 – 2y2, as shown in Figure 10.6.

Figure 10.6: Volume of the approximating rectangular boxes

Solution

Solving using iterated integrals – Method 1

2 2
Sum of volume = ∫0 ∫0 16 − 𝑥 2 − 2𝑦 2 𝑑𝑥 𝑑𝑦

2 2
𝑥3
= ∫ [16𝑥 − − 2𝑥𝑦 2 ] 𝑑𝑦
0 3 0

2 2 2
88 88 4𝑦 3
= ∫ [ − 4𝑦 2 ] 𝑑𝑦 = [ 𝑦 − ] = 48 unit 3
0 3 0 3 3 0

Solving using iterated integrals – Method 2

2 2
Sum of volume = ∫ ∫ 16 − 𝑥 2 − 2𝑦 2 𝑑𝑦 𝑑𝑥
0 0

2 2
2
= ∫ [16𝑦 − 𝑥 2 𝑦 − 𝑦 3 ] 𝑑𝑥
0 3 0

2 2 2
80 80 2𝑥 3
=∫ [ − 2𝑥 2 ] 𝑑𝑥 = [ 𝑥 − ] = 48 unit 3
0 3 0 3 3 0

49
Both method based on iterated integrals give the same answer.

Example

Evaluate the iterated integral

3 2
∫ ∫ 𝑥 2 𝑦 𝑑𝑦 𝑑𝑥
0 1

Solution

Regarding x as a constant, we obtain

2 𝑦=2
2
𝑦2 2
∫ 𝑥 𝑦 𝑑𝑦 = [𝑥 ]
1 2 𝑦=1

22 12 3
= 𝑥2 ( ) − 𝑥2 ( ) = 𝑥2
2 2 2
3
Thus the function A in the preceding discussion is given by 𝐴(𝑥) = 𝑥 2 in this example. We now integrate
2
this function of x from 0 to 3:
3 2 3 2 3
3 2
∫ ∫ 𝑥 2 𝑦 𝑑𝑦 𝑑𝑥 = ∫ [∫ 𝑥 2 𝑦 𝑑𝑦] 𝑑𝑥 = ∫ 𝑥 𝑑𝑥
0 1 0 1 0 2
3
𝑥3 27
= | =
2 0 2

10.1.5 AVERAGE VALUE

We know that the average value of a function f of one variable defined on an interval [a, b] is
𝑏
1
𝑓𝑎𝑣𝑒 = ∫ 𝑓(𝑥) 𝑑𝑥
𝑏−𝑎 𝑎

In a similar fashion we define the average value of a function f of two variables defined on a rectangle R to
be

1
𝑓𝑎𝑣𝑒 = ∬ 𝑓(𝑥, 𝑦) 𝑑𝐴
𝐴(𝑅)
𝑅

where A(R) is the area of R. If f (x, y)  0, the equation

50
 𝐴(𝑅) × 𝑓𝑎𝑣𝑒 = ∬ 𝑓(𝑥, 𝑦) 𝑑𝐴
𝑅

says that the box with base R and height fave has the same volume as the solid that lies under the graph of f.
[If z = f (x, y) describes a mountainous region and you chop off the tops of the mountains at height fave, then
you can use them to fill in the valleys so that the region becomes completely flat. See Figure 10.7.]

Figure 10.7: Mountainous region

10.2 DOUBLE INTEGRALS OVER GENERAL REGIONS

For single integrals, the region over which we integrate is always an interval. But for double integrals, we
want to be able to integrate a function f not just over rectangles but also over regions D of more general
shape, such as the one illustrated in Figure 10.8.

Figure 10.8: Regions D

51
We suppose that D is a bounded region, which means that D can be enclosed in a rectangular region R as in
Figure 10.9.

Figure 10.9: Regions D enclosed in a rectangular region R

Then we define a new function F with domain R by

… (12)

If F is integrable over R, then we define the double integral of f over D by

…. (13)

Where F is given by Eq. (12). Definition in Eq. (13) makes sense because R is a rectangle and so R F (x, y) dA
has been previously defined. The procedure that we have used is reasonable because the values of F (x, y)
are 0 when (x, y) lies outside D and so they contribute nothing to the integral.

This means that it doesn’t matter what rectangle R we use as long as it contains D. In the case where f (x, y)
 0, we can still interpret D f (x, y) dA as the volume of the solid that lies above D and under the surface z =
f (x, y) (the graph of f ).

You can see that this is reasonable by comparing the graphs of f and F in Figures 10.10 and remembering
that R F (x, y) dA is the volume under the graph of F.

Figure 10.10: Comparison between graphs of f and F

52
Figure 10.10 also shows that F is likely to have discontinuities at the boundary points of D. Nonetheless, if f
is continuous on D and the boundary curve of D is “well behaved”, then it can be shown that R F (x, y) dA
exists and therefore D f (x, y) dA exists. In particular, this is the case for type I and type II regions.

A plane region D is said to be of type I if it lies between the graphs of two continuous functions of x, that is,

D = {(x, y) | a  x  b, g1 (x)  y  g2 (x)}

where g1 and g2 are continuous on [a, b].

Some examples of type I regions are shown in Figure 10.11.

Figure 10.11: Examples of type I regions

In order to evaluate D f (x, y) dA when D is a region of type I, we choose a rectangle R = [a, b]  [c, d ] that
contains D, as in Figure 10.12, and we let F be the function given by Eq. (12); that is, F agrees with f on D
and F is 0 outside D.

Figure 10.12: Examples of type I regions

Then, by Fubini’s Theorem,

𝑏 𝑑
∬ 𝑓(𝑥, 𝑦) 𝑑𝐴 = ∬ 𝐹(𝑥, 𝑦) 𝑑𝐴 = ∫ ∫ 𝐹(𝑥, 𝑦)𝑑𝑦 𝑑𝑥
𝑎 𝑐
𝐷 𝑅

53
Observe that F (x, y) = 0 if y < g1 (x) or y > g2 (x) because (x, y) then lies outside D. Therefore
𝑑 𝑔2 (𝑥) 𝑔2 (𝑥)
∫ 𝐹(𝑥, 𝑦)𝑑𝑦 = ∫ 𝐹(𝑥, 𝑦)𝑑𝑦 = ∫ 𝑓(𝑥, 𝑦)𝑑𝑦
𝑐 𝑔1 (𝑥) 𝑔1 (𝑥)

because F (x, y) = f (x, y) when g1 (x)  y  g2 (x). Thus we have the following formula that enables us to
evaluate the double integral as an iterated integral.

.. (14)

The integral on the right side of Eq. (14) is an iterated integral, except that in the inner integral we regard x
as being constant not only in f (x, y) but also in the limits of integration, g1(x) and g2(x).

We also consider plane regions of type II, which can be expressed as

D = {(x, y) | c  y  d, h1(y)  x  h2(y)} … (15)

where h1 and h2 are continuous. Two such regions are illustrated in Figure 10.13.

Figure 10.13: Examples of type II regions

Using the same methods that were used in establishing Eq. (14), we can show that

2 𝑑 ℎ (𝑦)
∬𝐷 𝑓(𝑥, 𝑦) 𝑑𝐴 = ∫𝑐 ∫ℎ (𝑦) 𝑓(𝑥, 𝑦)𝑑𝑥 𝑑𝑦 … (16)
1

Where D is a type II region given by Eq. (15).

54
Example

Evaluate D (x + 2y) dA, where D is the region bounded by the parabolas y = 2x2 and y = 1 + x2.

Solution

The parabolas intersect when 2x2 = 1 + x2, that is, x2 = 1, so x = 1.

We note that the region D, sketched in Figure 10.14, is a type I region but not a type II region and we can
write D = {(x, y) | –1  x  1, 2x2  y  1 + x2}

Figure 10.14: Region D

Since the lower boundary is y = 2x2 and the upper boundary is y = 1 + x2, Eq. (14) gives

1 1+𝑥 2
∬(𝑥 + 2𝑦) 𝑑𝐴 = ∫ ∫ (𝑥 + 2𝑦)𝑑𝑦 𝑑𝑥
−1 2𝑥 2
𝐷

1 2
𝑦=1+𝑥
= ∫ [𝑥𝑦 + 𝑦 2 ]𝑦=2𝑥 2 𝑑𝑥
−1

1
= ∫ [𝑥(1 + 𝑥 2 ) + (1 + 𝑥 2 )2 − 𝑥(2𝑥 2 ) − (2𝑥 2 )2 ]𝑑𝑥
−1

1
32
= ∫(−3𝑥 4 − 𝑥 3 + 2𝑥 2 + 𝑥 + 1)𝑑𝑥 =
15
−1

55
10.2.1 PROPERTIES OF DOUBLE INTEGRALS

We assume that all of the following integrals exist. For rectangular regions D the first three properties can
be proved in the same manner. And then for general regions the properties follow from definition in Eq.
(13).

…(17)

…(18)

where c is a constant. If f (x, y)  g (x, y) for all (x, y) in D, then

…(19)

The next property of double integrals is similar to the property of single integrals given by the equation

If D = D1 U D2, where D1 and D2 don’t overlap except perhaps on their boundaries (see Figure 10.15), then

…(20)

Figure 10.15: Region D= D1 U D2

Property in Eq. (20) can be used to evaluate double integrals over regions D that are neither type I nor type
II but can be expressed as a union of regions of type I or type II.

56
Figure 10.16 illustrates this procedure.

(a): D is neither type I nor type II (b): D = D1 U D2, D1 is type I, D2 is type II

Figure 10.16

The next property of integrals says that if we integrate the constant function f (x, y) = 1 over a region D, we
get the area of D:

The last property states that

10.3 DOUBLE INTEGRALS IN POLAR COORDINATES

Suppose that we want to evaluate a double integral  R f (x, y) dA, where R is one of the regions shown in
Figure 10.17. In either case the description of R in terms of rectangular coordinates is rather complicated,
but R is easily described using polar coordinates.

Figure 10.17

57
From Figure 10.18, the polar coordinates (r,  ) of a point are related to the rectangular coordinates (x, y) by
the equations

𝑟2 = 𝑥2 + 𝑦2 𝑥 = 𝑟 cos 𝜃 𝑦 = 𝑟 sin 𝜃

Figure 10.18: Polar coordinates of a point related to rectangular coordinates

The regions in
Figure 10.17 are special cases of a polar rectangle

R = {(r,  ) | a  r  b,      }

which is shown in Figure 10.19.

Figure 10.19: Polar rectangle

In order to compute the double integral  R f (x, y) dA, where R is a polar rectangle, we divide the interval [a,
b] into m subintervals [ri – 1, ri] of equal width r = (b – a)/m and we divide the interval [,  ] into n
subintervals [j – 1, j] of equal width  = ( – )/n. Then the circles r = ri and the rays  = j divide the polar
rectangle R into the small polar rectangles Rij shown in Figure 10.20.

58
 Figure 10.20: Dividing R into polar subrectangles

The “center” of the polar subrectangle

Rij = {(r,  ) | ri – 1  r  ri, j – 1    j}

has polar coordinates

1 1
𝑟𝑖∗ = 2 (𝑟𝑖−1 + 𝑟𝑖 ) 𝜃𝑗∗ = 2 (𝜃𝑗−1 + 𝜃𝑗 )

We compute the area of Rij using the fact that the area of a sector of a circle with radius r and central angle
1
 is 2 𝑟 2 𝜃. Subtracting the areas of two such sectors, each of which has central angle  = j – j – 1, we find
that the area of Rij is

1 1 2 1
∆𝐴𝑖 = 𝑟𝑖2 ∆𝜃 − 𝑟𝑖−1 ∆𝜃 = (𝑟𝑖2 − 𝑟𝑖−1
2
)∆𝜃
2 2 2
1
= (𝑟𝑖 + 𝑟𝑖−1 )(𝑟𝑖 − 𝑟𝑖−1 )∆𝜃 = 𝑟𝑖∗ ∆𝑟 ∆𝜃
2

Although we have defined the double integral  R f (x, y) dA in terms of ordinary rectangles, it can be shown
that, for continuous functions f, we always obtain the same answer using polar rectangles. The rectangular
coordinates of the center of Rij are (𝑟𝑖∗ cos 𝜃𝑗∗ , 𝑟𝑖∗ sin 𝜃𝑗∗ , ), so a typical Riemann sum is

… (21)

If we write g(r,  ) = r f (r cos , r sin  ), then the Riemann sum in Eq. (21) can be written as
𝑚 𝑛

∑ ∑ 𝑔(𝑟𝑖∗ , 𝜃𝑗∗ )∆𝑟 ∆𝜃

𝑖=1 𝑗=1

which is a Riemann sum for the double integral

59
 𝛽 𝑏
∫ ∫ 𝑔(𝑟, 𝜃) 𝑑𝑟 𝑑𝜃
𝛼 𝑎

Therefore we have

𝑚 𝑛

∬ 𝑓(𝑥, 𝑦)𝑑𝐴 = lim ∑ ∑ 𝑓(𝑟𝑖∗ cos 𝜃𝑗∗ , 𝑟𝑖∗ sin 𝜃𝑗∗ ) ∆𝐴𝑖
𝑚,𝑛→∞
𝑅 𝑖=1 𝑗=1

𝑚 𝑛 𝛽 𝑏
= lim ∑ ∑ 𝑔(𝑟𝑖∗ , 𝜃𝑗∗ )∆𝑟 ∆𝜃 = ∫ ∫ 𝑔(𝑟, 𝜃) 𝑑𝑟 𝑑𝜃
𝑚,𝑛→∞ 𝛼 𝑎
𝑖=1 𝑗=1

𝛽 𝑏
= ∫ ∫ 𝑓(𝑟 cos 𝜃 , 𝑟 sin 𝜃) 𝑟 𝑑𝑟 𝑑𝜃
𝛼 𝑎

…(22)

The formula in Eq. (22) says that we convert from rectangular to polar coordinates in a double integral by
writing x = r cos  and y = r sin , using the appropriate limits of integration for r and , and replacing dA by
r dr d.

Be careful not to forget the additional factor r on the right side of Formula in Eq. (22). A classical method for
remembering this is shown in Figure 10.21, where the “infinitesimal” polar rectangle can be thought of as
an ordinary rectangle with dimensions r d and dr and therefore has “area” dA = r dr d.

Figure 10.21: Illustration of the “infinitesimal” polar rectangle

60
Example

Evaluate  R (3x + 4y2) dA, where R is the region in the upper half-plane bounded by the circles x2 + y2 = 1
and x2 + y2 = 4.

Solution

The region R can be described as

R = {(x, y) | y  0, 1  x2 + y2  4}

It is the half-ring shown in Figure 10.22 and in polar coordinates it is given by 1  r  2, 0    .

Figure 10.22: Half-ring

Therefore, by the formula in Eq. (22),

𝜋 2
2 )𝑑𝐴
∬(3𝑥 + 4𝑦 = ∫ ∫ (3𝑟 cos 𝜃 + 4𝑟 2 sin2 𝜃) 𝑟 𝑑𝑟 𝑑𝜃
0 1
𝑅

𝜋 2 𝜋
= ∫ ∫ (3𝑟 2 cos 𝜃 + 4𝑟 3 sin2 𝜃) 𝑑𝑟 𝑑𝜃 = ∫ [𝑟 3 cos 𝜃 + 𝑟 4 sin2 𝜃]𝑟=2
𝑟=1 𝑑𝜃
0 1 0

𝜋 𝜋
15
= ∫ (7 cos 𝜃 + 15sin2 𝜃 ) 𝑑𝜃 = ∫ [7 cos 𝜃 + (1 − cos 2𝜃) ] 𝑑𝜃
0 0 2
𝜋
15𝜃 15 15𝜋
= 7 sin 𝜃 + − sin 2𝜃] =
2 4 0 2

What we have done so far can be extended to the more complicated type of region shown in Figure 10.23.
In fact, by combining formula in Eq. (22) with

𝑑 ℎ2 (𝑦)
∬ 𝑓(𝑥, 𝑦)𝑑𝐴 = ∫ ∫ 𝑓(𝑥, 𝑦)𝑑𝑥 𝑑𝑦
𝑐 ℎ1 (𝑦)
𝐷

where D is a type II region, we obtain the following formula

61
 … (23)

Figure 10.23: 𝐷 = {(𝑟, 𝜃)|𝛼 ≤ 𝜃 ≤ 𝛽, ℎ1 (𝜃) ≤ 𝑟 ≤ ℎ2 (𝜃)}

In particular, taking f (x, y) = 1, h1( ) = 0, and h2( ) = h( ) in this formula, we see that the area of the region
D bounded by  = ,  = , and r = h( ) is

𝛽 ℎ(𝜃)
𝐴𝐷 = ∬ 1 𝑑𝐴 = ∫ ∫ 𝑟 𝑑𝑟 𝑑𝜃
𝛼 0
𝐷

𝛽 ℎ(𝜃)
𝑟2 𝛽
1
=∫ [ ] 𝑑𝜃 = ∫ [ℎ(𝜃)]2 𝑑𝜃
𝛼 2 0 𝛼 2

10.4 TRIPLE INTEGRALS

We have defined single integrals for functions of one variable and double integrals for functions of two
variables, so we can define triple integrals for functions of three variables. Let’s first deal with the simplest
case where f is defined on a rectangular box:

B = {(x, y, z) | a  x  b, c  y  d, r  z  s} … (24)

The first step is to divide B into sub-boxes. We do this by dividing the interval [a, b] into l subintervals [xi–1,
xi] of equal width x, dividing [c, d ] into m subintervals of width y, and dividing [r, s] into n subintervals of
width z.

The planes through the endpoints of these subintervals parallel to the coordinate planes divide the box B
into lmn sub-boxes

Bijk = [xi – 1, xi]  [yj – 1, yj]  [zk – 1, zk]

62
which are shown in Figure 10.24. Each sub-box has volume V = x y z.

Figure 10.24: Rectangular box

Then we form the triple Riemann sum

𝑙 𝑚 𝑛
∗ ∗ ∗
∑ ∑ ∑ 𝑓(𝑥𝑖𝑗𝑘 , 𝑦𝑖𝑗𝑘 , 𝑧𝑖𝑗𝑘 ) ∆𝑉 … (25)
𝑖=1 𝑗=1 𝑘=1

∗ ∗ ∗
where the sample point (𝑥𝑖𝑗𝑘 , 𝑦𝑖𝑗𝑘 , 𝑧𝑖𝑗𝑘 ) is in Bijk. By analogy with the definition of a double integral, we
define the triple integral as the limit of the triple Riemann sums in Eq. (25).

…(26)

Again, the triple integral always exists if f is continuous. We can choose the sample point to be any point in
the sub-box, but if we choose it to be the point (xi, yj, zk) we get a simpler-looking expression for the triple
integral:

63
Just as for double integrals, the practical method for evaluating triple integrals is to express them as
iterated integrals as follows.

…(27)

The iterated integral on the right side of Fubini’s Theorem means that we integrate first with respect to x
(keeping y and z fixed), then we integrate with respect to y (keeping z fixed), and finally we integrate with
respect to z.

There are five other possible orders in which we can integrate, all of which give the same value. For
instance, if we integrate with respect to y, then z, and then x, we have

𝑏 𝑠 𝑑
∭ 𝑓(𝑥, 𝑦, 𝑧) 𝑑𝑉 = ∫ ∫ ∫ 𝑓(𝑥, 𝑦, 𝑧)𝑑𝑦 𝑑𝑧 𝑑𝑥
𝑎 𝑟 𝑐
𝐵

Example

Evaluate the triple integral B xyz2 dV, where B is the rectangular box given by

B = {(x, y, z) | 0  x  1, –1  y  2, 0  z  3}

Solution

We could use any of the six possible orders of integration. If we choose to integrate with respect to x, then
y, and then z, we obtain

3 2 1 3 2 𝑥=1
2 2
𝑥 2 𝑦𝑧 2
∭ 𝑥𝑦𝑧 𝑑𝑉 = ∫ ∫ ∫ 𝑥𝑦𝑧 𝑑𝑥 𝑑𝑦 𝑑𝑧 = ∫ ∫ [ ] 𝑑𝑦 𝑑𝑧
0 −1 0 0 −1 2 𝑥=0
𝐵

3 2 𝑦=2
𝑦𝑧 2 3
𝑦2𝑧2 3
3𝑧 2
=∫ ∫ 𝑑𝑦 𝑑𝑧 = ∫ [ ] 𝑑𝑧 = ∫ 𝑑𝑧
0 −1 2 0 4 𝑦=−1 0 4

3
𝑧3 27
= ] =
4 0 4

64
10.4.1 TRIPLE INTEGRALS OVER A GENERAL BOUNDED REGION E

Now we define the triple integral over a general bounded region E in three-dimensional space (a solid) by
much the same procedure that we used for double integrals. We enclose E in a box B of the type given by
Eq. (24). Then we define F so that it agrees with f on E but is 0 for points in B that are outside E. By
definition

∭ 𝑓(𝑥, 𝑦, 𝑧)𝑑𝑉 = ∭ 𝐹(𝑥, 𝑦, 𝑧) 𝑑𝑉

𝐸 𝐸

This integral exists if f is continuous and the boundary of E is “reasonably smooth”. The triple integral has
essentially the same properties as the double integral. We restrict our attention to continuous functions f
and to certain simple types of regions.

A solid region E is said to be of type 1 if it lies between the graphs of two continuous functions of x and y,
that is,

E = {(x, y, z) | (x, y)  D, u1(x, y)  z  u2(x, y)} …(28)

where D is the projection of E onto the xy-plane as shown in Figure 10.25.

Figure 10.25: A type I solid region

Notice that the upper boundary of the solid E is the surface with equation z = u2(x, y), while the lower
boundary is the surface z = u1(x, y). By the same sort of argument, it can be shown that if E is a type 1 region
given by Eq. (28), then

…(29)

The meaning of the inner integral on the right side of Eq. (29) is that x and y are held fixed, and therefore
u1(x, y) and u2(x, y) are regarded as constants, while f (x, y, z) is integrated with respect to z.

65
In particular, if the projection D of E onto the xy-plane is a type I plane region (as in Figure 10.26).

Figure 10.26: A type I solid region where the projection D is a type I plane region

Then

E = {(x, y, z) | a  x  b, g1(x)  y  g2(x), u1(x, y)  z  u2(x, y)}

and Eq. (29) becomes

…(30)

If, on the other hand, D is a type II plane region (as in Figure 10.27), then

E = {(x, y, z) | c  y  d, h1(y)  x  h2(y), u1(x, y)  z  u2(x, y)}

and Eq. (29) becomes

…(31)

Figure 10.27: A type I solid region with type II projection

66
A solid region E is of type 2 if it is of the form

E = {(x, y, z) | (y, z)  D, u1(y, z)  x  u2(y, z)}

where, this time, D is the projection of E onto the yz-plane (see Figure 10.28).

Figure 10.28: A type II region

The back surface is x = u1(y, z), the front surface is x = u2(y, z), and we have

..(32)

Finally, a type 3 region is of the form

E = {(x, y, z) | (x, z)  D, u1(x, z)  y  u2(x, z)}

where D is the projection of E onto the xz-plane, y = u1(x, z) is the left surface, and y = u2(x, z) is the right
surface (see Figure 10.29).

Figure 10.29: A type III region

67
For this type of region we have

…(33)

In each of Eq. (32) and (33) there may be two possible expressions for the integral depending on whether D
is a type I or type II plane region (and corresponding to Eq. (30) and (31)).

10.4.2 TRIPLE INTEGRALS IN CYLINDRICAL COORDINATES

In plane geometry the polar coordinate system is used to give a convenient description of certain curves
and regions. Figure 10.30 enables us to recall the connection between polar and Cartesian coordinates.

Figure 10.30: Connection between polar and Cartesian

coordinates

If the point P has Cartesian coordinates (x, y) and polar coordinates (r,  ), then, from the figure,

x = r cos  y = r sin 
𝑦
r2 = x2 + y2 tan = 𝑥

In three dimensions there is a coordinate system, called cylindrical coordinates, that is similar to polar
coordinates and gives convenient descriptions of some commonly occurring surfaces and solids. As we will
see, some triple integrals are much easier to evaluate in cylindrical coordinates.

10.4.3 CYLINDRICAL COORDINATES

In the cylindrical coordinate system, a point P in three-dimensional space is represented by the ordered
triple (r, , z), where r and  are polar coordinates of the projection of P onto the xy-plane and z is the
directed distance from the xy-plane to P. (See Figure 10.31)

68
 Figure 10.31: The cylindrical coordinates of a point

To convert from cylindrical to rectangular coordinates, we use the equations

…(34)

whereas to convert from rectangular to cylindrical coordinates, we use

…(35)

Example

Plot the point with cylindrical coordinates (2, 2 /3, 1) and find its rectangular coordinates.

Solution

The point with cylindrical coordinates (2, 2 /3, 1) is plotted in Figure 10.32. From Eq. (34), its rectangular
coordinates are

2𝜋 1
𝑥 = 2 cos = 2 (− ) = −1
3 2

2𝜋 √3
𝑦 = 2 sin = 2 ( ) = √3
3 2

𝑧=1

So the point is (–1, √3, 1) in rectangular coordinates.

69
 Figure 10.32: Cylindrical coordinates (2, 2 /3, 1)

10.4.4 EVALUATING TRIPLE INTEGRALS WITH CYLINDRICAL COORDINATES

Suppose that E is a type 1 region whose projection D onto the xy-plane is conveniently described in polar
coordinates (see Figure 10.33).

Figure 10.33: Type I region

In particular, suppose that f is continuous and

E = {(x, y, z) | (x, y)  D, u1(x, y)  z  u2(x, y)}

where D is given in polar coordinates by

D = {(r,  ) |     β , h1( )  r  h2( )}

We know

… (36)

But to evaluate double integrals in polar coordinates, we have the formula

..(37)
70
Formula in Eq. (37) is the formula for triple integration in cylindrical coordinates.

It says that we convert a triple integral from rectangular to cylindrical coordinates by writing x = r cos , y =
r sin , leaving z as it is, using the appropriate limits of integration for z, r, and , and replacing dV by r dz dr
d. (Figure 10.34 shows how to remember this.)

Figure 10.34: Volume element in cylindrical coordinates: dV = r dz dr d

It is worthwhile to use this formula when E is a solid region easily described in cylindrical coordinates, and
especially when the function f (x, y, z) involves the expression x2 + y2.

Example

A solid E lies within the cylinder x2 + y2 = 1, below the plane z = 4, and above the paraboloid z = 1 – x2 – y2.
(See Figure 10.35.) The density at any point is proportional to its distance from the axis of the cylinder. Find
the mass of E.

Solution

In cylindrical coordinates the cylinder is r = 1 and the paraboloid is z = 1 – r2, so we can write

E = {(r, , z) | 0    2, 0  r  1, 1 – r2  z  4}

Since the density at (x, y, z) is proportional to the distance from the z-axis, the density function is

𝑓(𝑥, 𝑦, 𝑧) = 𝐾√𝑥 2 + 𝑦 2 = 𝐾𝑟

where K is the proportionality constant.

71
Therefore, the mass of E is

2𝜋 1 4
𝑚 = ∭ 𝐾√𝑥 2 + 𝑦 2 𝑑𝑉 = ∫ ∫ ∫ (𝐾𝑟)𝑟 𝑑𝑧 𝑑𝑟 𝑑𝜃
0 0 1−𝑟 2
𝐸

2𝜋 1 2𝜋 1
2 [4 2
=∫ ∫ 𝐾𝑟 − (1 − 𝑟 )] 𝑑𝑟 𝑑𝜃 = 𝐾 ∫ 𝑑𝜃 ∫ (3𝑟 2 + 𝑟 4 )𝑑𝑟
0 0 0 0

𝑟5 1 12𝜋𝐾
= 2𝜋𝐾 [𝑟 3 + ] =
5 0 5

Figure 10.35: Solid E lies within the

cylinder x2 + y2 = 1

72
DIFFERENTIAL EQUATIONS
WEEK 11: DIFFERENTIAL EQUATIONS
11.1 INTRODUCTION TO DIFFERENTIAL EQUATION

In creating a mathematical model of a physical system, we frequently involve differential equation/ integral
equation / integro-differential equations to express relationships, such as ‘the force acting on a falling object
is proportional to its acceleration’, ‘voltage drop across a resistor is proportional to the current’, etc.

Table 12.2: Three types of equations of a mathematical model: (i) Differential equation (ii) Integral
equation (iii) Integro-differential equation.
(i) Differential equation (ii) Integral equation (iii) Integro-differential equation

Equations which involve Equations which involve Equations which involve both
derivatives of the variables in integrals of the variables in the derivatives and integrals of the
the model. model. variables in the model.

For example: RC Electrical Circuit

What are the amounts of charge and current flow in an electric circuit that consists a generator (E volt), a
resistance (R ohms) and a capacitor (C capacitance)? The RC electrical circuit is shown below.

From experiments, we know that the voltage loss through a resistor and capacitor is proportional to the
current and charge respectively, where Δ𝑉𝑟𝑒𝑠𝑖𝑠𝑡𝑜𝑟 (𝑡) ∝ 𝑖(𝑡) and Δ𝑉𝑐𝑎𝑝𝑎𝑐𝑖𝑡𝑜𝑟 (𝑡) ∝ 𝑞(𝑡). Hence, Δ𝑉𝑅 (𝑡) =
1
𝑅𝑖(𝑡) and Δ𝑉𝐶 (𝑡) = 𝑞(𝑡).
𝐶

According to Kirchhoff’s voltage law, the summation of voltage in a closed loop is equal to zero. Thus,
(𝑉𝐵 − 𝑉𝐴 ) + (𝑉𝐹 − 𝑉𝐵 ) + (𝑉𝐷 − 𝑉𝐹 ) + (𝑉𝐴 − 𝑉𝐷 ) = 0
1
𝑅𝑖(𝑡) + 𝑞(𝑡) + 0 + (−𝐸(𝑡)) = 0
𝐶
1
𝑅𝑖(𝑡) + 𝑞(𝑡) = 𝐸(𝑡)
𝐶

73
 From definition, the current is equal to the rate of charge flow or the charge is the integral of the current
over time.
𝑑𝑞(𝑡)
𝑖(𝑡) = 𝑑𝑡
or 𝑞(𝑡) = ∫ 𝑖(𝑡)𝑑𝑡

Rearrange it, we obtain three different forms of mathematical model as shown below, to solve the desired
variables (i.e. charge and current). Note that different methods and strategies are used to solve these
equations. In this study, we will focus on the topic of differential equation.

𝑑𝑞(𝑡) 1 1 𝑑𝑞(𝑡) 1
(i) 𝑅 + 𝐶 𝑞(𝑡) = 𝐸(𝑡) (ii) 𝑅𝑖(𝑡) + 𝐶 ∫ 𝑖(𝑡)𝑑𝑡 = 𝐸(𝑡) (iii) 𝑅 + 𝐶 ∫ 𝑖(𝑡)𝑑𝑡 = 𝐸(𝑡)
𝑑𝑡 𝑑𝑡

-involves derivative,
𝑑𝑞(𝑡)
. -involves integrals, ∫ 𝑖(𝑡)𝑑𝑡. -involves both derivative,
𝑑𝑞(𝑡)
and
𝑑𝑡 𝑑𝑡
-This is known as integral integrals, ∫ 𝑖(𝑡)𝑑𝑡.
-This is known as differential
equation.
equation. -This is known as integro-differential
equation.

Differential equation (DE) plays a fundamental role in engineering because many physical
phenomena are best formulated mathematically in terms of their rate of change. By solving a differential
equation, we can gain a deeper understanding of the physical processes that these equations are describing.
Some examples of fundamental laws that are written in terms of the rate of change of variables are shown
in the table below.

Table 12.3: Examples of fundamental laws written in the differential equation.

Fundamental Law Mathematical Expression Variables and Parameter
𝑑𝑣(𝑡) 𝛴𝐹(𝑡) Velocity (𝑣);
Newton’s 2nd Law of Motion
= Force (𝐹);
𝑑𝑡 𝑚
Mass (𝑚).
𝑑𝑖(𝑡) Voltage drop (Δ𝑣𝐿 (𝑡));
Faraday’s Law (Voltage drop
Δ𝑣𝐿 (𝑡) = 𝐿 Inductance (𝐿)
across an inductor) 𝑑𝑡
Current (𝑖).
𝑑𝑇(𝑥) Heat flux (𝑞);
Fourier’s Heat Law
𝑞(𝑥) = −𝑘′ Thermal conductivity (𝑘′);
𝑑𝑥
Temperature (𝑇).
𝑑𝐶(𝑥) Mass flux (𝐽);
Fick’s law of diffusion
𝐽(𝑥) = −𝐷 Diffusion coefficient (𝐷);
𝑑𝑥
Concentration(𝐶).
Hint: The use of differential equations may empower us to make precise predictions about the future
behaviour of our models/ system. Sometimes, even if we can’t completely solve a differential equation
(especially when it deals with nonlinear case), we may still be able to determine useful properties about its
solution (qualitative information). The motivation and implementation of the mathematical modelling with
differential equation in the engineering problem solving is illustrated in Appendices 11.1 & 11.2.
74
11.2 THE CLASSIFICATION/TYPE OF DIFFERENTIAL EQUATIONS

Different types of differential equations may require different strategies to solve the problem. Thus, it is
important for the user to understand, recognize and classify the correct categories of differential equations.

(i) Independent and dependent variables

A differential equation expresses such that the dependent variable(s) depends on the independent variable.

Dependent variable Independent variable

It is the variable(s) that is differentiated. It is the variable(s) with respect to which differentiation occurs.

Example (1):
𝑑2 𝑦 𝑑𝑦
𝑑𝑥 2
− 4𝑥 𝑑𝑥 = 𝑐𝑜𝑠2𝑥
This differential equation has dependent variable of 𝑦 and independent variable of 𝑥.
Note that the variable 𝑦 is in the function of 𝑥, i.e. 𝑦 = 𝑦(𝑥). In other words, 𝑦 is changed with respect to
𝑥.

Example (2):
𝑑2 𝑥 𝑑𝑥
𝑑𝑡 2
− 4𝑥 𝑑𝑡 = 𝑐𝑜𝑠2𝑡
This differential equation has dependent variable of 𝑥 and independent variable of 𝑡.
Note that the variable 𝑥 is in the function of 𝑡, i.e. 𝑥 = 𝑥(𝑡). In other words, 𝑥 is changed with respect to
𝑡.

75
(ii) Ordinary Differential Equation (ODE) versus Partial Differential Equation (PDE)

Differential equation can be categorized into 2 cases: ODE & PDE. The classification of ODE and PDE
depends on the number of independent variable, regardless of the number of dependent variables.

CASE 1: ODE
Those equations that involve ordinary derivatives (i.e. 𝑑 symbol) are called ODE. ODE has only one
independent variable. It can be separated into the ODE problem or system of ODE problem depends on the
number of dependent variable.

(i) One dependent variable (ii) More than one dependent variable
For example: Brine mixture problem For example: Population of rabbit & fox
𝑑𝑥 𝑥 𝑑𝑥
=2− , = 𝑎𝑥 − 𝑏𝑥𝑦,
𝑑𝑡 5 𝑑𝑡

where x = concentration of salt. 𝑑𝑦

𝑑𝑡
= −𝑐𝑦 + 𝑑𝑥𝑦,
where x = rabbit; y = fox.

-This is an ODE problem: - This is a system of ODE problem:

(i) One independent variable (t) (i) One independent variable (t)
(ii) One dependent variable (x) (ii) More than one dependent variable (x & y)
Comment: ODE is the main focus of this study.

CASE 2: PDE
Those equations that involve partial derivatives (i.e. ∂ symbol) are called PDE. PDE has two independent
variables or more. It can be separated into the PDE problem or system of PDE problem depends on the
number of dependent variable.

(i) One dependent variable (ii) More than one dependent variable
For example: Transient heat equation For example: Incompressible Navier-Stokes
∂𝑇(𝑥,𝑡) ∂2 𝑇(𝑥,𝑡)
equation for pipe flow
−𝛼 ∂𝑥 2
= 0,
∂𝑡 ∂𝑢𝑖 ∂𝑢 ∂2 𝑢𝑖 ∂ω
∂𝑡
+ 𝑢𝑗 ∂𝑥 𝑖 − 𝑣 ∂𝑥 + ∂𝑥 = 𝑔𝑖 ,
where 𝑇 = temperature; 𝛼 = thermal diffusivity. 𝑗 𝑗 ∂𝑥𝑗 𝑖

where 𝑢 = flow velocity; ω = elevation

-This is a PDE problem: - This is a system of PDE problem:

(i) More than one independent variables (x & t) (i) More than one independent variables (t , 𝑥𝑗 & 𝑥𝑗 )

(ii) One dependent variable (T) (ii) More than one dependent variables (𝑢𝑖 & ω)
Comment: PDE is out of scope in this study. It will be covered in KIX1002. At current stage, students
should know how to classify between ODE and PDE.
76
(iii) Order of a differential equation

The order of a differential equation is the degree of the highest derivative that occurs in the equation. The
approach to find the order of ODE and PDE is the same as illustrated below.

Case 1: Order of ODE

1st order ODE: The order of highest derivative 𝑑 is 1
Example (1):
𝑑𝑥 𝑑𝑦
4 𝑑𝑡 − 3 𝑑𝑡 − 𝑥 + 𝑦 = cos(2𝑡)

Example (2):
𝑑𝑥 𝑑𝑦
( )2 − 3 − 𝑥 + 𝑦 = cos(2𝑡)
𝑑𝑡 𝑑𝑡

2nd order ODE: The order of highest derivative 𝑑2 is 2

Example (1):
𝑑2 𝑓 𝑑𝑓
𝑑𝑥 2
− 4𝑥 𝑑𝑥 = cos(2𝑥)
Example (2):
𝑑2 𝑓 𝑑𝑓 4
𝑑𝑥 2
− 4𝑥 (𝑑x ) = cos(2𝑥)

Case 2: Order of PDE

1st order PDE: The order of highest derivative ∂ is 1
Example (1):
𝜕𝑓 𝜕𝑓
4 −3 − 𝑥 + 𝑦 = cos(2𝑡)
𝜕𝑡 𝜕𝑡

Example (2):
𝜕𝑥 𝜕𝑥
𝜕𝑡
+ (𝜕𝑦)4 = cos(2𝑡) + 2𝑦

2nd order PDE: The order of highest derivative ∂2 is 2

Example (1):
𝜕𝑓 ∂2 𝑓
4 −3 − 𝑥 + 𝑦 = cos(2𝑡)
𝜕𝑥 ∂y2

Example (2):
∂2 𝑓 𝜕𝑓
+ ( )3 = cos(2𝑡) + 2𝑦
∂t ∂y 𝜕𝑦

77
Note: The order of an equation is not affected by any power to which the derivatives may be raised.

Moreover, degree of a differential equation is the power of the highest order derivative.

Example (1):

𝑑𝑥 𝑑𝑦
( )2 − 3 − 𝑥 + 𝑦 = cos(2𝑡)
𝑑𝑡 𝑑𝑡
The differential equation above is a 1st order ODE with degree 2.

Prove: (i) 1st order because the order of highest derivative 𝑑 is 1

(ii) ODE because it has only one independent variable (𝑡)

𝑑𝑥
(iii) Degree 2 because the power of the highest order derivative ( 𝑑𝑡 )2 is 2

Example (2):

∂2 𝑓 𝜕𝑓
+ ( )3 = cos(2𝑡) + 2𝑦
∂t ∂y 𝜕𝑦

The differential equation above is a 2nd order PDE with degree 1.

Prove: (i) 2nd order because the order of highest derivative ∂2 is 2

(ii) PDE because it has more than one independent variables (𝑡 & 𝑦)

∂2 𝑓
(iii) Degree 1 because the power of the highest order derivative ∂t ∂y is 1

Example (3):

𝜕𝑥 𝜕𝑥
+ ( )4 = cos(2𝑡) + 2𝑦
𝜕𝑡 𝜕𝑦

The differential equation above is a 1st order PDE with degree 4.

Prove: (i) 1st order because the order of highest derivative ∂ is 1

(ii) PDE because it has more than one independent variables (𝑡 & 𝑦)
𝜕𝑥
(iii) Degree 4 because the power of the highest order derivative (𝜕𝑦)4 is 4

78
(iv) Linear and nonlinear differential equations

We used to plot a linear graph of 𝑦1 versus 𝑦2 using 𝑦1 = 𝑚𝑦2 + 𝑐 (Linear Algebraic Eqn), where 𝑚 & 𝑐 are
the slope and the intercept respectively. In this case, 𝑦1 is in the function of 𝑦2 . Therefore, 𝑦1 is the
dependent variable while 𝑦2 is the independent variable. Rearrange it, we obtain the general form of linear
equation as follows: 𝑎1 𝑦1 + 𝑎0 𝑦2 = 𝑐 where 𝑎1 = 1 and 𝑎0 = −𝑚.
By using similar approach, we get 1st order linear ODE where 𝑎1 (𝑥)𝑦 ′ + 𝑎0 (𝑥)𝑦 = 𝑔(𝑥). In this case,
𝑦′ is the first derivative of 𝑦. Linear ODE has the properties of 𝑓(𝑦1 + 𝑦2 ) = 𝑓(𝑦1 ) + 𝑓(𝑦2 ).

In general, a linear ODE of order nth has the following form:

𝑎𝑛 (𝑥)𝑦 (𝑛) + 𝑎𝑛−1 (𝑥)𝑦 (𝑛−1) + ⋯ + 𝑎2 (𝑥)𝑦 ′′ + 𝑎1 (𝑥)𝑦 ′ + 𝑎0 (𝑥)𝑦 = 𝑔(𝑥),
where
𝑎𝑖 (𝑥) is a function of independent variable (x) for 𝑖 = 0,1, … , 𝑛.
𝑔(𝑥) is a function of independent variable (x)
𝑦 is the dependent variable

Any equation of ODE that does not follow the linear format as equation above is known as nonlinear ODE.
For example:
𝑎𝑛 (𝑥, 𝑦)(𝑦 (𝑛) ) 𝐴 + 𝑎𝑛−1 (𝑥, 𝑦)(𝑦 (𝑛−1) )𝐵 + ⋯ + 𝑎1 (𝑥, 𝑦)(𝑦 ′ )𝐶 + 𝑎0 (𝑥, 𝑦)(𝑦)𝐷 = 𝑔(𝑥, 𝑦).
where the power 𝐴, 𝐵, 𝐶 & 𝐷 ≠ 1

Examples of the linear and nonlinear ODEs are given as follow.

Case 1: Linear ODE
General format of 1st order linear ODE: 𝑎1 (𝑥)𝑦 ′ + 𝑎0 (𝑥)𝑦 = 𝑔(𝑥)
For example:
𝑑𝑦
1st order linear ODE: −4 𝑑𝑥 − 𝑥 2 = 0,
𝑑𝑦
Rearrange it into the general format: −4 𝑑𝑥 = 𝑥 2
where
𝑑𝑦
𝑦′ = 𝑑𝑥 (𝑦 = 𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒 & 𝑥 = 𝑖𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒)
𝑎1 (𝑥) = −4;
𝑎0 (𝑥) = 0;
𝑔(𝑥) = 𝑥 2 .
∴ It is a linear ODE since it follows the linear format: 𝑎1 (𝑥)𝑦 ′ + 𝑎0 (𝑥)𝑦 = 𝑔(𝑥)

79
General format of 2nd order linear ODE: 𝑎2 (𝑥)𝑦 ′′ + 𝑎1 (𝑥)𝑦 ′ + 𝑎0 (𝑥)𝑦 = 𝑔(𝑥)
For example:
𝑑2 𝑓 𝑑𝑓
2nd order linear ODE: 𝑑𝑥 2 − 4𝑥 𝑑𝑥 − 𝑐𝑜𝑠2𝑥 − 3 = 0,
𝑑2 𝑓 𝑑𝑓
Rearrange it into the general format: 𝑑𝑥2 − 4𝑥 𝑑𝑥 = 𝑐𝑜𝑠2𝑥 + 3,
where
𝑑2 𝑓
𝑓′′ = 𝑑𝑥 2 ; (𝑓 = 𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒 & 𝑥 = 𝑖𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒)
𝑑𝑓
𝑓′ = ;
𝑑𝑥

𝑎2 (𝑥) = 1;
𝑎1 (𝑥) = −4𝑥;
𝑎0 (𝑥) = 0;
𝑔(𝑥) = 𝑐𝑜𝑠2𝑥 + 3
∴ It is a linear ODE since it follows the linear format: 𝑎2 (𝑥)𝑓 ′′ + 𝑎1 (𝑥)𝑓 ′ + 𝑎0 (𝑥)𝑓 = 𝑔(𝑥)

General format of 3rd order linear ODE: 𝑎3 (𝑥)𝑦 ′′′ + 𝑎2 (𝑥)𝑦 ′′ + 𝑎1 (𝑥)𝑦 ′ + 𝑎0 (𝑥)𝑦 = 𝑔(𝑥)
For example:
𝑑3𝑥 𝑑𝑥
3rd order linear ODE: 4 +3 + 2𝑥 + cos(𝑡) = 2,
𝑑𝑡 3 𝑑𝑡
𝑑3 𝑥 𝑑𝑥
Rearrange it into the general format: 4 𝑑𝑡 3 + 3 𝑑𝑡 + 2𝑥 = 2 − cos(𝑡),
where
𝑑3 𝑥
𝑥′′′ = 𝑑𝑡 3
; (𝑥 = 𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒 & 𝑡 = 𝑖𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒)
𝑑𝑥
𝑥′ = 𝑑𝑡
;
𝑎3 (𝑡) = 4;
𝑎2 (𝑡) = 0;
𝑎1 (𝑡) = 3;
𝑎0 (𝑡) = 2;
𝑔(𝑡) = 2 − cos(𝑡)
∴ It is a linear ODE since it follows the linear format: 𝑎3 (𝑡)𝑥 ′′′ + 𝑎2 (𝑡)𝑥 ′′ + 𝑎1 (𝑡)𝑥 ′ + 𝑎0 (𝑡)𝑥 = 𝑔(𝑡)

80
 Case 2: Nonlinear ODE
General format of 1st order linear ODE: 𝑎1 (𝑥)𝑦 ′ + 𝑎0 (𝑥)𝑦 = 𝑔(𝑥)

For example:
𝑑𝑦 2
1st order nonlinear ODE: −4 (𝑑𝑥 ) = 𝑥 2 ,

Rearrange it into the general format: Same Eqn.

∴ It is a nonlinear ODE because it does not obey linear equation: 𝑎1 (𝑥)𝑦 ′ + 𝑎0 (𝑥)𝑦 = 𝑔(𝑥) as the
𝑑𝑦 2 𝑑𝑦
derivative 𝑦′ ≠ ( ) where the is squared.
𝑑𝑥 𝑑𝑥

General format of 2nd order linear ODE: 𝑎2 (𝑥)𝑦 ′′ + 𝑎1 (𝑥)𝑦 ′ + 𝑎0 (𝑥)𝑦 = 𝑔(𝑥)

For example:
𝑑2 𝑓 𝑑𝑓
2nd order nonlinear ODE: − 4𝑓 + 𝑐𝑜𝑠2𝑥 = 0,
𝑑𝑥 2 𝑑𝑥
𝑑2 𝑓 𝑑𝑓
Rearrange it into the general format: 𝑑𝑥2 − 4𝑓 𝑑𝑥 = −𝑐𝑜𝑠2𝑥

∴ It is a nonlinear ODE because it does not obey linear equation: 𝑎2 (𝑥)𝑓 ′′ + 𝑎1 (𝑥)𝑓 ′ + 𝑎0 (𝑥)𝑓 = 𝑔(𝑥) as
the 𝑎1 (𝑥) ≠ −4𝑓, where 𝑎1 (𝑥) should not be in the function of dependent variable 𝑓.

General format of 3rd order linear ODE: 𝑎3 (𝑥)𝑦 ′′′ + 𝑎2 (𝑥)𝑦 ′′ + 𝑎1 (𝑥)𝑦 ′ + 𝑎0 (𝑥)𝑦 = 𝑔(𝑥)

For example:
𝑑3 𝑥 𝑑𝑥
3rd order nonlinear ODE: 4 𝑑𝑡 3 + 3 𝑑𝑡 + 2sin(𝑥) + cos(𝑡) = 2,
𝑑3 𝑥 𝑑𝑥
Rearrange it into the general format: 4 𝑑𝑡 3 + 3 𝑑𝑡 + 2 sin(𝑥) = 2 − cos(𝑡),

∴ It is a nonlinear ODE because it does not obey linear equation: 𝑎3 (𝑡)𝑥 ′′′ + 𝑎2 (𝑡)𝑥 ′′ + 𝑎1 (𝑡)𝑥 ′ +
𝑎0 (𝑡)𝑥 = 𝑔(𝑡) as the 𝑥 ≠ sin(𝑥), where it has nonlinear sinusoidal function of dependent function 𝑥.

Hint: Most of the time, nonlinear ODE has nonlinear components such as coefficient 𝑎𝑖 (𝑥, 𝑦) in the function
𝑑𝑦
of dependent variable 𝑦 and the 𝑦 or its derivative have degree more than one, i.e. 𝑦 2 & (𝑑𝑥 )3 . For linear
differential equations, there are no products of the dependent variable and its derivatives and neither the
derivative occur to any power other than the first power.

81
 For your additional knowledge, many of the nonlinear equations that occur in engineering cannot
be solved easily as they stand, but can be solved, for practical engineering purpose, by the process of
replacing them with linear equations that are a close approximation – at least in some region of interest.

(v) Homogeneous and nonhomogeneous equations of linear equation

This is applied to the case of linear differential equation only. Arrange the linear equation in standard format,
where all terms containing dependent variable occur on left-hand side (LHS), while terms containing only
the independent variable and constant occur on the right-hand side (RHS). Linear ODE can be categorized
into homogenous and non-homogeneous equation by evaluating the RHS term as follows.

Case 1: Homogeneous equation

RHS term is equal to zero in the standard format
(i) Linear homogeneous ODE

Example (1):
𝑑𝑥
𝑑𝑡
+ 4𝑥 = 0
Example (2):
𝑑2 𝑥 𝑑𝑥
+3 + 𝑥𝑠𝑖𝑛(𝑡) = 0
𝑑𝑡 2 𝑑𝑡

(ii) Linear homogeneous PDE

Example (1):
𝜕𝑓 𝜕𝑓
𝜕𝑥
+ 𝜕𝑦 = 0

Example (2):
𝜕2 𝑓 𝜕𝑓
𝜕𝑥𝜕𝑦
+ 𝜕𝑦 = 0

82
 Case 2: Nonhomogeneous equation
RHS term is equal to nonzero in the standard format
(i) Linear nonhomogeneous ODE

Example (1):
𝑑𝑥
𝑑𝑡
+ 4𝑥 = 5
Example (2):
𝑑2 𝑥 𝑑𝑥
𝑑𝑡 2
+ 3 𝑑𝑡 + 𝑥𝑠𝑖𝑛(𝑡) = cos(𝑡)

(ii) Linear nonhomogeneous PDE

Example (1):
𝜕𝑓 𝜕𝑓
𝜕𝑥
+ 𝜕𝑦 = 4𝑥 2 + 2𝑦

Example (2):
𝜕2 𝑓 𝜕𝑓
+ = 5𝑦
𝜕𝑥𝜕𝑦 𝜕𝑦

Note: Other method is used to classify the homogeneity of nonlinear ODE as shown in section (vi).

𝑑𝑦 𝑓(𝑥,𝑦)
(vi) Homogeneous and nonhomogeneous equations of (𝑑𝑥 = 𝑔(𝑥,𝑦))

This section is particular important especially when we deal with 1st order nonlinear ODE problem. It is
worthwhile to mention that there is another method to classify the homogeneous and nonhomogeneous
groups in ODE. For 1st order ODE equation, the classification is given below.
First of all, the descriptions of the homogeneous and nonhomogeneous functions are given:

A function 𝑓(𝑥, 𝑦) is said to be homogeneous of degree 𝑛 if we get 𝑓(𝜆𝑥, 𝜆𝑦) = 𝜆𝑛 𝑓(𝑥, 𝑦) for all arbitrary
constant 𝜆.

Any function that does not follow the homogeneous format as equation above is known as a
nonhomogeneous function.

83
Example (1): Check the homogeneous degree for the function 𝑥 4 + 𝑥𝑦 3 .

In this case, 𝑓(𝑥, 𝑦) = 𝑥 4 + 𝑥𝑦 3 .

Applying 𝑓(𝜆𝑥, 𝜆𝑦), we get 𝑓(𝜆𝑥, 𝜆𝑦) = (𝜆𝑥)4 + (𝜆𝑥)(𝜆𝑦)3 = 𝜆4 (𝜆 + 𝑥𝑦 3 )

Since 𝑓(𝜆𝑥, 𝜆𝑦) = 𝜆𝑛 𝑓(𝑥, 𝑦), thus we say the function 𝑥 4 + 𝑥𝑦 3 is homogeneous of degree, 𝑛 = 4.

Example (2): Check the homogeneous degree for the function 𝑦 2 − 𝑥𝑦 + 1.

In this case, 𝑓(𝑥, 𝑦) = 𝑦 2 − 𝑥𝑦 + 1.

Applying 𝑓(𝜆𝑥, 𝜆𝑦), we get 𝑓(𝜆𝑥, 𝜆𝑦) = (𝜆𝑦)2 − (𝜆𝑥)(𝜆𝑦) + 1.

Since 𝑓(𝜆𝑥, 𝜆𝑦) ≠ 𝜆𝑛 𝑓(𝑥, 𝑦) or (𝜆𝑦)2 − (𝜆𝑥)(𝜆𝑦) + 1 ≠ 𝜆2 (𝑦 2 − 𝑥𝑦 + 1) , thus we say the function
(𝑦)2 − (𝑥𝑦) + 1 is nonhomogeneous.

Then, the homogeneous and nonhomogeneous differential equation of 1st order ODE are given:

𝒅𝒚 𝒇(𝒙,𝒚)
Case (1): Homogeneous differential equation of =
𝒅𝒙 𝒈(𝒙,𝒚)
𝑑𝑦 𝑓(𝑥,𝑦)
𝑑𝑥
= 𝑔(𝑥,𝑦) is a homogeneous differential equation if 𝑓(𝑥, 𝑦) and 𝑔(𝑥, 𝑦) are homogeneous of the same
degree.

For example:
𝑑𝑦 𝑦 2 −𝑥 2
𝑑𝑥
= 2𝑥𝑦

It is a 1st order nonlinear homogeneous differential equation where the functions at numerator and
denominator are homogeneous of degree 2.

𝒅𝒚 𝒇(𝒙,𝒚)
Case (2): Nonhomogeneous differential equation of 𝒅𝒙
= 𝒈(𝒙,𝒚)
𝑑𝑦 𝑓(𝑥,𝑦)
= is a nonhomogeneous differential equation if 𝑓(𝑥, 𝑦) and 𝑔(𝑥, 𝑦) are nonhomogeneous or
𝑑𝑥 𝑔(𝑥,𝑦)
they have homogeneity with different degrees.

For example:
𝑑𝑦 2𝑥−4𝑦+5
𝑑𝑥
= 𝑥−2𝑦+3

It is a 1st order nonlinear nonhomogeneous differential equation where the functions at numerator and
denominator are nonhomogeneous.

84
 To avoid confusion, we use the term “homogeneous/nonhomogeneous” for all the linear ODE case,
𝑑𝑦 𝑓(𝑥,𝑦)
while we use the term “homogeneous/ nonhomogeneous of 𝑑𝑥 = 𝑔(𝑥,𝑦) form” for the 1st order nonlinear
ODE case.

Note: The homogeneity of 2nd and higher order nonlinear ODE is out of scope in this study.

11.3 SOLUTION TO DIFFERENTIAL EQUATION

The difference between the solution of algebraic equation and differential equation is shown in table below:

Case (1): Solution to Algebraic Equation

(i) We expect the solution to be a number
For example: Solution of the equation 𝑥 + 7 = 10 is 𝑥 = 3

Or, perhaps,

(ii) A set of real & complex numbers

For example: Solution of the polynomial equation 𝑥 3 − 5𝑥 2 + 8𝑥 = 12 are 𝑥1 = 3.7162 ; 𝑥2 =
0.6419 + 𝑖1.6784

Or, perhaps,

(iii) A set of vector or matrix

𝑥
For example: Solution of two simultaneous equations 𝑥 − 5𝑦 = 3 and 3𝑥 + 9𝑦 = 12 is a vector {𝑦} =
3.625
{ }.
0.125

Case (2): Solution to Differential Equation

The solution of differential equation is not the same as the case of algebraic equation.

(i) We expect the solution to be a function

𝑑2𝑥
For example: Solution of the differential equation 𝑑𝑡 2 + 25𝑥 = 0 is 𝑥(𝑡) = 𝐴𝑠𝑖𝑛(5𝑡) + 𝐵𝑐𝑜𝑠(5𝑡)

Or, perhaps,

(ii) A family of functions

𝑑𝑥 𝑑𝑦
For example: Solution of multiple differential equations 𝑑𝑡
= 6𝑥 − 2𝑦 and 𝑑𝑡
= −3𝑦 + 5𝑥 are
𝑡(7.2419) 𝑡(−0.4419) 𝑡(4.7016) 𝑡(−1.7016)
𝑥(𝑡) = 𝐴𝑒 − 𝐵𝑒 ; 𝑦(𝑡) = 𝐴𝑒 − 𝐵𝑒

85
 The solution of differential equation can be further divided into two types: (a) General Solution (b)
Particular Solution.

Case (1): General solution

The most general function that will satisfy the differential equation contains one or more arbitrary
constants. Normally the number of arbitrary constants equal to the order of the differential equation.

𝑑𝑥
For example: The general solution of the differential equation 𝑑𝑡
= −4𝑥 is 𝑥(𝑡) = 𝐴𝑒 −4𝑡 where any
arbitrary constant 𝐴 can satisfy the equation.

Case (2): Particular solution

Giving particular numerical values to the constants in the general solution results in a particular solution
of the equation. Normally, particular solution can be obtained by knowing the initial or boundary
condition.

𝑑𝑥
For example: Previously, let the general solution of the differential equation = −4𝑥 is 𝑥(𝑡) = 𝐴𝑒 −4𝑡
𝑑𝑡
where any arbitrary constant 𝐴 can satisfy the equation. Given initial condition where 𝑥(0) = 2.5, the
𝑑𝑥
particular solution of the differential equation = −4𝑥 which has the value 2.5 when 𝑡 = 0 is 𝑥(𝑡) =
𝑑𝑡
2.5𝑒 −4𝑡 .
Here, only a specific constant 𝐴 = 2.5 can satisfy the equation.

Note: In fact, general solution & particular solution indicates that there are an infinite number of solutions
to the differential equation unless we are given the specific condition of the problem. The actual solution to
a differential solution is the specific solution that not only satisfies the differential equation, but also satisfies
the given initial conditions.

The particular solution can be obtained from either “Boundary-value problem” or “Initial-value
problem”. It depends on the given specific condition about the value of the solution at a particular point, in
addition to the differential equation.

Case (1): Boundary-value problem

All conditions are specified at different values of the independent variable, usually at extreme points or
boundaries of a system.
𝑑2 𝑥
For example: The particular solution of the differential equation 𝑑𝑡 2 + 25𝑥 = 0
which has the boundary conditions 𝑥(0) = 4 & 𝑥(10) = 7
is (𝑡) = 5.78 sin(5𝑡) + 4cos(5𝑡) .

86
 Case (2): Initial-value problem
All conditions are specified at the same value of the independent variable.

𝑑2 𝑥
For example: The particular solution of the differential equation 𝑑𝑡 2
+ 25𝑥 = 0
which has the initial condition 𝑥(0) = 4 & 𝑥 ′ (0) = 8
is 𝑥(𝑡) = 1.6 𝑠𝑖𝑛(5𝑡) + 4𝑐𝑜𝑠(5𝑡) .

Note: For 1st order differential equation, the condition can be treated as initial/ boundary condition. For
higher order differential equation, the distinction becomes obvious.

11.4 VERIFICATION OF DIFFERENTIAL EQUATION’S SOLUTION

A solution to an ODE is a function which is differentiable and which satisfies the given equation. This is true
for both explicit and implicit solutions.

Case (1): Explicit solution

An explicit solution is any solution that is given in the form 𝑦 = 𝑄(𝑡) where 𝑦 is the dependent variable;
𝑄(𝑡) is the function of independent variable.

The explicit solution is valid in the interval 𝐼: 𝛼 < 𝑡 < 𝛽 I if the following conditions are meet:
𝑑𝑄(𝑡) 𝑑 𝑛−1 𝑄(𝑡)
(i) 𝑄(𝑡), 𝑑𝑡 , … , 𝑑𝑡 𝑛−1 is differentiable
(ii) 𝑄(𝑡) can satisfy the differential equation

Case (2): Implicit solution

An implicit solution is any solution that isn’t in explicit form (𝑦 = 𝑄(𝑡)).

87
For example:
𝑑𝑦
Find implicit and explicit solution to the first order differential equation 𝑦 𝑑𝑡 = 𝑡 , 𝑦(2) = −1.

Solution:
𝑑𝑦
𝑦 𝑑𝑡 = 𝑡

>> ∫ 𝑦𝑑𝑦 = ∫ 𝑡𝑑𝑡

𝑦2 𝑡2
>> 2
= 2
+𝐶 [Comment: This is general implicit solution]

Applying the initial condition 𝑦(2) = −1, we get

(−1)2 (2)2
>> = +𝐶
2 2

3
>> 𝐶 = −
2

∴ 𝑦2 = 𝑡2 − 3 [Comment: This is particular implicit solution]

Generally, we arrange the implicit solution in the form G(t,y)=0, i.e. 𝑦 2 − 𝑡 2 = −3.

Rearrange the implicit solution and let LHS to be the dependent variable, we get

>> 𝑦 = ±√𝑡 2 − 3 [Comment: This is particular explicit solution]

In this case, it is found that there are two explicit solutions:

>> 𝑦1 = +√𝑡 2 − 3

>>𝑦2 = −√𝑡 2 − 3

To check which one is the true solution, reapply the initial condition, 𝑦(2) = −1.

>>> 𝑦1 = +√𝑡 2 − 3 = +√22 − 3 = 1 [Comment: This is not the true explicit solution]

>>> 𝑦2 = −√𝑡 2 − 3 = −√22 − 3 = −1 [Comment: This is the true explicit solution]

88
Verification of solution:
𝑑𝑦
(i) To verify the 𝑦 2 = 𝑡 2 − 3 is the implicit solution for the differential equation 𝑑𝑡
= 𝑡 . We try to
deduce the differential equation from it.

For example:
𝑦2 = 𝑡2 − 3
𝑑 𝑑
>> 𝑑𝑡 [𝑦 2 ] = 𝑑𝑡 [𝑡 2 − 3]
𝑑𝑦
>> 2𝑦 𝑑𝑡 = 2𝑡
𝑑𝑦
>> 𝑦 =𝑡
𝑑𝑡
𝑑𝑦
∴ Thus, it is proven that 𝑦 2 = 𝑡 2 − 3 is the implicit solution for the 𝑦 𝑑𝑡 = 𝑡.

𝑑𝑦
(ii) To verify the 𝑦2 = −√𝑡 2 − 3 is the explicit solution for the differential equation 𝑑𝑡
= 𝑡 . We
differentiate and substitute it to the equation.

For example:
𝑦2 = −√𝑡 2 − 3
𝑑 𝑑
>> 𝑑𝑡 [𝑦2 ] = 𝑑𝑡 [−√𝑡 2 − 3]
1
𝑑 1
>> 𝑑𝑡 [𝑦2 ] = − 2 (𝑡 2 − 3)2−1 . 2𝑡]
𝑑 𝑡
>> 𝑑𝑡 [𝑦2 ] = −
√(𝑡 2 −3)

𝑑𝑦
Substitute the derivative to the differential equation 𝑦 =𝑡
𝑑𝑡
LHS we get,
𝑑𝑦 𝑡
𝑦 𝑑𝑡 = 𝑦2 (− )
√(𝑡 2 −3)

Substitute 𝑦2 = −√𝑡 2 − 3 into the equation, we get

𝑑𝑦 𝑡
𝑦 = −√𝑡 2 − 3 (− )
𝑑𝑡 √(𝑡 2 −3)

𝑑𝑦
>> 𝑦 =𝑡
𝑑𝑡

∴ Since LHS=RHS, thus it is proven that 𝑦2 = −√𝑡 2 − 3 is the explicit solution for the differential
𝑑𝑦
equation 𝑦 𝑑𝑡 = 𝑡.

(iii) Exercise: Try to prove that 𝑦1 = +√𝑡 2 − 3 is the explicit solution for the differential equation.

89
Note: It will not always be possible to find an explicit solution.

For example:
𝑑𝑦
For differential equation 3𝑦 3 𝑒 3𝑥𝑦 − 1 + (2𝑦𝑒 3𝑥𝑦 + 3𝑥𝑦 2 𝑒 3𝑥𝑦 ) 𝑑𝑥 = 0 , initial condition, 𝑦(0) = 1 . You
suspect that 𝑦 2 𝑒 3𝑥𝑦 − 𝑥 = 1 is the implicit solution to the differential equation. Please verify the solution.

Verification of the implicit solution:

𝑦 2 𝑒 3𝑥𝑦 − 𝑥 = 1
𝑑 𝑑 𝑑
>> 𝑑𝑥 [𝑦 2 𝑒 3𝑥𝑦 ] − 𝑑𝑥 [𝑥] = 𝑑𝑥 [1]
𝑑 𝑑
>> 𝑑𝑥 [𝑦 2 ]. 𝑒 3𝑥𝑦 + 𝑦 2 . 𝑑𝑥 [𝑒 3𝑥𝑦 ] − 1 = 0
𝑑𝑦 𝑑
>> 2𝑦 𝑑𝑥 . 𝑒 3𝑥𝑦 + 𝑦 2 . 𝑑𝑥 [3𝑥𝑦]. 𝑒 3𝑥𝑦 − 1 = 0
𝑑𝑦 𝑑 𝑑
>> (2𝑦𝑒 3𝑥𝑦 ) 𝑑𝑥 + 𝑦 2 . (𝑑𝑥 [3𝑥]. 𝑦 + 3𝑥. 𝑑𝑥 [𝑦]). 𝑒 3𝑥𝑦 − 1 = 0
𝑑𝑦 𝑑𝑦
>> (2𝑦𝑒 3𝑥𝑦 ) + 𝑦 2 . (3𝑦+3𝑥 ). 𝑒 3𝑥𝑦 −1=0
𝑑𝑥 𝑑𝑥
𝑑𝑦 𝑑𝑦
>> (2𝑦𝑒 3𝑥𝑦 ) 𝑑𝑥 + (3𝑦 3 +3𝑥𝑦 2 𝑑𝑥 ). 𝑒 3𝑥𝑦 − 1 = 0
𝑑𝑦
>> 3𝑦 3 𝑒 3𝑥𝑦 − 1 + (2𝑦𝑒 3𝑥𝑦 + 3𝑥𝑦 2 𝑒 3𝑥𝑦 ) 𝑑𝑥 = 0 [proven]

There is no way to rearrange the implicit solution for 𝑦 = 𝑦(𝑥) and get an explicit solution. This mostly
happen for nonlinear case.

90
11.5 STRATEGY TO SOLVE 1 S T ORDER DIFFERENTIAL EQUATION

There are many strategies that have been developed to solve differential equation. We will start with the
most fundamental one. Bear in your mind various strategies can be implemented depends on the
types/forms of differential equation.
First of all, we will start with the strategies to solve 1st order linear differential equation, i.e.
𝑎1 (𝑥)𝑦 ′ + 𝑎0 (𝑥)𝑦 = 𝑔(𝑥). These strategies include
(a) exact differential equation
(b) linear differential equation
(c) separable differential equation
Then we will discuss the strategies used for solving the 1st order nonlinear differential equation, i.e.
𝑎1 (𝑥, 𝑦)(𝑦 ′ )𝐶 + 𝑎0 (𝑥, 𝑦)(𝑦)𝐷 = 𝑔(𝑥, 𝑦). These strategies include
(a) separable differential equation
(b) Bernoulli’s equation
(c) homogeneous differential equation
(d) nonhomogeneous differential equation.
Note: In many cases there are at least one of these strategies which can be used to solve the problem.

11.5.1 EXACT DIFFERENTIAL EQUATION

Some differential equations are of a form that can be solved readily, it would be useful to be able to
recognize them. For example: we know the derivative for:

Product rule Quotient rule

𝑑(𝑥𝑦) 𝑑(𝑦) 𝑑(𝑥) 𝑥 𝑑(𝑥) 𝑑(𝑦)
=𝑥 +𝑦 = 𝑥𝑦 ′ + 𝑦 𝑑( )
𝑦 𝑦 −𝑥 𝑦−𝑥𝑦 ′
𝑑𝑥 𝑑𝑥 𝑑𝑥
𝑑𝑥
= 𝑑𝑥
𝑦2
𝑑𝑥
= 𝑦2

Example: Solve the differential equation 𝑥𝑦 ′ + 𝑦 = 0.

𝑑(𝑥𝑦)
Solution: If we recalled the product rule, we can see that the LHS is equal to 𝑥𝑦 ′ + 𝑦 = . Thus, the
𝑑𝑥
differential equation 𝑥𝑦 ′ + 𝑦 = 0 is of the ‘exact’ form. Hence, we can further solve it as
𝑥𝑦 ′ + 𝑦 = 0
𝑑(𝑥𝑦)
>> =0
𝑑𝑥
𝑑(𝑥𝑦)
>> ∫ 𝑑𝑥
𝑑𝑥 = ∫ 0𝑑𝑥
>> 𝑥𝑦 = 𝐶
𝐶
>> 𝑦 = 𝑥

91
 Identification of exact differential equation would be difficult in most of the time as it involves a lot
of memorization and the identification process become hard, especially when the differential equation is
varies with the original format that you memorize.
For example: Is 𝑥𝑦 ′ + 𝑦 + 4 = 0 an exact differential equation?
Thus, it is crucial for us to know the equation whether it is exact or not before we proceed with some lengthy
process in finding its solution. In general, exact differential equation can be identified by using the following
approach.

Procedure to check the exact differential equation:

First, arrange the 1st order differential equation of the form
𝑀(𝑥, 𝑦)𝑑𝑥 + 𝑁(𝑥, 𝑦)𝑑𝑦 = 0
where 𝑀(𝑥, 𝑦) & 𝑁(𝑥, 𝑦) are two functions of independent variable 𝑥 and dependent variable 𝑦.

𝜕𝑀(𝑥,𝑦) 𝜕𝑁(𝑥,𝑦)
If 𝜕𝑦
= 𝜕𝑥
,

then it is an exact differential equation.

𝜕𝑀(𝑥,𝑦) 𝜕𝑁(𝑥,𝑦)
If 𝜕𝑦
≠ 𝜕𝑥
,

then there is no way for the differential equation to be exact.

Thus, we need to use other strategy to solve the problem

Example: Solve the differential equation 𝑥𝑦 ′ + 𝑦 + 4 = 0.

Solution: Check if we can solve using exact differential equation.

𝑥𝑦 ′ + 𝑦 + 4 = 0
>> 𝑥𝑑𝑦 + (𝑦 + 4)𝑑𝑥 = 0 [Comment: Arrange it in the form of (𝑥, 𝑦)𝑑𝑥 + 𝑁(𝑥, 𝑦)𝑑𝑦 = 0 ]
>> where
𝑀(𝑥, 𝑦) = (𝑦 + 4)
𝑁(𝑥, 𝑦) = 𝑥

𝜕𝑀(𝑥,𝑦) 𝜕𝑁(𝑥,𝑦)
>> Check it with 𝜕𝑦
= 𝜕𝑥
condition.
𝜕𝑀(𝑥,𝑦) 𝜕
LHS: 𝜕𝑦
= 𝜕𝑦 (𝑦 + 4) = 1
𝜕𝑁(𝑥,𝑦) 𝜕
RHS: 𝜕𝑥
= 𝜕𝑥 (𝑥) = 1
𝜕𝑀(𝑥,𝑦) 𝜕𝑁(𝑥,𝑦)
∴ Since 𝜕𝑦
= 𝜕𝑥
is true, we can solve it using exact differential equation.
92
Further solve it using exact approach.
𝑥𝑦 ′ + 𝑦 + 4 = 0
𝑑(𝑥(𝑦+4))
>> 𝑑𝑥
=0
𝑑(𝑥(𝑦+4))
>> ∫ 𝑑𝑥
𝑑𝑥 = ∫ 0𝑑𝑥
>> 𝑥(𝑦 + 4) = 𝐶
𝐶
>> 𝑦 = 𝑥 − 4

11.5.2 LINEAR DIFFERENTIAL EQUATION

Previous section shows that if the differential equation can be in the “exact” form, it can be solved directly.
However, it is undesired to obtain the exact differential equation in most of the time.

Example: Solve the differential equation 𝑥𝑦 ′ − 𝑦 = 0

Solution:
𝜕𝑀(𝑥,𝑦) 𝜕𝑁(𝑥,𝑦)
Again, we are not sure if 𝑥𝑦 ′ − 𝑦 = 0, so we check it with 𝜕𝑦
= 𝜕𝑥
condition.

𝑥𝑦 ′ − 𝑦 = 0
>> 𝑥𝑑𝑦 − 𝑦𝑑𝑥 = 0 [Comment: Arrange it in the form of (𝑥, 𝑦)𝑑𝑥 + 𝑁(𝑥, 𝑦)𝑑𝑦 = 0 ]
>> where
𝑀(𝑥, 𝑦) = −𝑦
𝑁(𝑥, 𝑦) = 𝑥

𝜕𝑀(𝑥,𝑦) 𝜕𝑁(𝑥,𝑦)
>> Check it with 𝜕𝑦
= 𝜕𝑥
condition.
𝜕𝑀(𝑥,𝑦) 𝜕
LHS: 𝜕𝑦
= 𝜕𝑦 (−𝑦) = −1
𝜕𝑁(𝑥,𝑦) 𝜕
RHS: 𝜕𝑥
= 𝜕𝑥 (𝑥) = 1
𝜕𝑀(𝑥,𝑦) 𝜕𝑁(𝑥,𝑦)
∴ Since 𝜕𝑦
= 𝜕𝑥
is not true, we cannot solve it using exact differential equation.

Note: It can’t be solved as it is in “non exact form”. However, there is a method used to convert it to exact
form. It is a strategy involving Integrating Factor and Linear Differential Equation as shown in the
following steps:

93
Procedure to solve 1st linear ODE problem using linear differential equation

Step 1: Arrange the differential equation in the linear form of

𝑑𝑦
𝑑𝑥
+ 𝑝(𝑥)𝑦 = 𝑞(𝑥)
where 𝑥 = 𝑖𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒 and 𝑦 = 𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒

Step 2: Create integrating factor,

𝐼𝐹 = 𝑒 ∫ 𝑝(𝑥)𝑑𝑥

Step 3: Multiply 1st ODE eqn by 𝐼𝐹

𝑑𝑦
𝐼𝐹. ( ) + 𝐼𝐹. (𝑝(𝑥)𝑦) = 𝐼𝐹. (𝑞(𝑥))
𝑑𝑥

Step 4: Recognize the LHS is exact solution,

𝑑𝑦 𝑑
i.e. LHS: 𝐼𝐹. ( ) + 𝐼𝐹. (𝑝(𝑥)𝑦) = (𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒. 𝐼𝐹)
𝑑𝑥 𝑑𝑥

𝑑 𝑑𝑦 𝑑
Prove: 𝑑𝑥 (𝑦. 𝑒 ∫ 𝑝(𝑥)𝑑𝑥 ) = (𝑑𝑥 ) 𝑒 ∫ 𝑝(𝑥)𝑑𝑥 + 𝑦 (𝑑𝑥 𝑒 ∫ 𝑝(𝑥)𝑑𝑥 )
𝑑𝑦 𝑑
= ( ) 𝑒 ∫ 𝑝(𝑥)𝑑𝑥 + 𝑦 ( 𝑝(𝑥)𝑑𝑥) 𝑒 ∫ 𝑝(𝑥)𝑑𝑥
𝑑𝑥 𝑑𝑥
𝑑𝑦
= ( ) 𝑒 ∫ 𝑝(𝑥)𝑑𝑥 + 𝑦(𝑝(𝑥))𝑒 ∫ 𝑝(𝑥)𝑑𝑥
𝑑𝑥
𝑑𝑦
= 𝐼𝐹. (𝑑𝑥 ) + 𝐼𝐹. (𝑝(𝑥)𝑦) [proven]

From steps 3 & 4:

𝑑𝑦
𝐼𝐹. (𝑑𝑥 ) + 𝐼𝐹. (𝑝(𝑥)𝑦) = 𝐼𝐹. (𝑞(𝑥))
𝑑
(𝑦. (𝐼𝐹))𝑑𝑥 = 𝐼𝐹. (𝑞(𝑥))
𝑑𝑥

Step 5:
𝑑
Integrate ∫ 𝑦(𝐼𝐹)𝑑𝑥 = ∫ 𝐼𝐹(𝑞(𝑥))
𝑑𝑥

and obtain the solution in the explicit form of 𝑦 = 𝑦(𝑥)

94
Continue with the previous problem:
Solve 𝑥𝑦 ′ − 𝑦 = 0 [where 𝑥 = 𝑖𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒 and 𝑦 = 𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒]
𝑑𝑦 1 𝑑𝑦
>> 𝑑𝑥 − 𝑥 𝑦 = 0 [Step 1- linear form of 𝑑𝑥
+ 𝑝(𝑥)𝑦 = 𝑞(𝑥) ]
where
1
𝑝(𝑥) = − 𝑥

𝑞(𝑥) = 0
1
>> The integrating factor, 𝐼𝐹 = 𝑒 ∫ −𝑥𝑑𝑥 = 𝑒 −𝑙𝑛𝑥 = 𝑥 −1 [Step 2- IF=𝑒 ∫ 𝑝(𝑥)𝑑𝑥 ]

𝑑𝑦 1
>> 𝑥 −1 ( ) − 𝑥 −1 ( 𝑦) = 𝑥 −1 (0) [Step 3- Multiply]
𝑑𝑥 𝑥
𝑑𝑦 1
>>𝑥 −1 ( ) − ( 𝑦) =0
𝑑𝑥 𝑥2

𝑑
>> 𝑑𝑥 (𝑦. 𝑥 −1 ) = 0 [Step 4- Exact]
𝑑
where LHS= (𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒. 𝐼𝐹)
𝑑𝑥

𝑑
>>∫ 𝑑𝑥 (𝑥 −1 𝑦)𝑑𝑥 = ∫ 0𝑑𝑥 [Step 5- Integrate]

>> 𝑥 −1 𝑦 = 𝐶
>> 𝑦 = 𝐶𝑥

∴ The solution of the first order linear ODE problem 𝑥𝑦 ′ − 𝑦 = 0 is 𝑦 = 𝐶𝑥, where 𝐶 = arbitrary constant.

Exercise: Verify that the solution is correctly determined.

95
 𝑑𝑥
Example: Solve the differential equation 𝑦
+ 2𝑥𝑑𝑦 = 0 by using Linear Differential Equation.

Solution:
𝑑𝑥
𝑦
+ 2𝑥𝑑𝑦 = 0
1 𝑑𝑦
>> 𝑦
+ 2𝑥 𝑑𝑥 = 0 [Comment: This is a nonlinear ODE; Rearrange it to linear ODE]

[where 𝑥 = 𝑖𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒 and 𝑦 = 𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒]

Exercise: What is the nonlinear component for the original eqn.?

Rearrange it (Let 𝑥 = 𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒 and 𝑦 = 𝑖𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒], we get

𝑑𝑥
>> 𝑑𝑦
+ (2𝑦)𝑥 = 0 [Step 1- linear form]

where
𝑝(𝑦) = 2𝑦
𝑞(𝑦) = 0

2
>> The integrating factor, 𝐼𝐹 = 𝑒 ∫ 2𝑦𝑑𝑥 = 𝑒 𝑦 [Step 2- IF]
2 𝑑𝑥 2
>> 𝑒 𝑦 𝑑𝑦
+ 𝑒 𝑦 (2𝑦)𝑥 = 0 [Step 3- Multiply]
𝑑 2
>> 𝑑𝑦 (𝑥. 𝑒 𝑦 ) = 0 [Step 4- Exact]
𝑑 2
>> ∫ 𝑑𝑦 (𝑥. 𝑒 𝑦 )𝑑𝑦 = ∫ 0𝑑𝑦
2
>>(𝑥. 𝑒 𝑦 ) = 𝐶
2
∴ 𝑥 = 𝐶𝑒 −𝑦 [Step 5- Integrate]

96
 𝑑𝑦 𝑦
Example: Solve the initial value problem 𝑑𝑥 = 2𝑥+3𝑦2 −2 , 𝑦(1) = 1

Solution:
𝑑𝑦 𝑦
𝑑𝑥
= 2𝑥+3𝑦2 −2
𝑑𝑦 𝑦
>> − =0 [Comment: This is a nonlinear ODE; Rearrange it to linear ODE]
𝑑𝑥 2𝑥+3𝑦 2 −2

Rearrange it (Let 𝑥 = 𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒 and 𝑦 = 𝑖𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒], we get

𝑑𝑥 2𝑥+3𝑦 2 −2
>> − =0
𝑑𝑦 𝑦

𝑑𝑥 2 3𝑦 2 −2
>> 𝑑𝑦 − (𝑦) 𝑥 = 𝑦
[Step 1- linear form]
2
∫ −(𝑦)𝑑𝑥
>> The integrating factor, 𝐼𝐹 = 𝑒 = 𝑒 −2𝑙𝑛𝑦 = 𝑦 −2 [Step 2- IF]
𝑑𝑥 2 3𝑦 2 −2 3𝑦 2 −2
>> 𝑦 −2 𝑑𝑦 − 𝑦 −2 (𝑦) 𝑥 = 𝑦 −2 ( 𝑦
) =( 𝑦3
) [Step 3- Multiply]
𝑑 3𝑦 2 −2
>> (𝑥. 𝑦 −2 ) =( ) [Step 4- Exact]
𝑑𝑦 𝑦3
𝑑 3𝑦 2 −2
>> ∫ 𝑑𝑦 (𝑥. 𝑦 −2 )𝑑𝑦 = ∫ ( 𝑦3
) 𝑑𝑦 [Step 5- Integrate]
3 −2
>> 𝑥. 𝑦 −2 = ∫ ( ) 𝑑𝑦 + ∫ ( 3 ) 𝑑𝑦
𝑦 𝑦

>> 𝑥. 𝑦 −2 = 3𝑙𝑛𝑦 + 𝑦 −2 + 𝐶, where 𝐶 = arbitrary constant.

Using initial condition, 𝑦(1) = 1 (i.e. when 𝑥 = 1, 𝑦 = 1)

>> (1). (1)−2 = 3ln(1) + (1)−2 + 𝐶
>> 𝐶 = 0

∴ 𝑥 = 3𝑦 2 𝑙𝑛𝑦 + 1

97
 In this study, it shows that integrating factor 𝐼𝐹 = 𝑒 ∫ 𝑝(𝑥)𝑑𝑥 is very useful in solving 1st order linear
ODE. For your additional knowledge, there are other types of integrating factors to solve various ODE
problems, as shown in the table below:

Condition Integrating factor (IF)

If 𝑀(𝑥, 𝑦)𝑑𝑥 + 𝑁(𝑥, 𝑦)𝑑𝑦 = 0 is a homogeneous 𝐼𝐹 = 1 , 𝑀𝑥 + 𝑁𝑦 ≠ 0
𝑀𝑥+𝑁𝑦
equation in 𝑥 & 𝑦.

If 𝑀(𝑥, 𝑦)𝑑𝑥 + 𝑁(𝑥, 𝑦)𝑑𝑦 = 0 is of the form 1

𝐼𝐹 = 𝑀𝑥−𝑁𝑦 , 𝑀𝑥 − 𝑁𝑦 ≠ 0
𝑓(𝑥𝑦)𝑦𝑑𝑥 + 𝛷(𝑥𝑦)𝑑𝑦 + 𝑁(𝑥, 𝑦)𝑑𝑦 = 0

If 𝑀(𝑥, 𝑦)𝑑𝑥 + 𝑁(𝑥, 𝑦)𝑑𝑦 = 0 be a differential 𝐼𝐹 = 𝑒 ∫ 𝑓(𝑥)𝑑𝑥

𝜕𝑀(𝑥,𝑦)⁄𝜕𝑦−𝜕𝑁(𝑥,𝑦)⁄𝜕𝑥
equation. If 𝑁
is a function of 𝑥
only, i.e. 𝑓(𝑥)

If 𝑀(𝑥, 𝑦)𝑑𝑥 + 𝑁(𝑥, 𝑦)𝑑𝑦 = 0 be a differential 𝐼𝐹 = 𝑒 ∫ 𝑓(𝑦)𝑑𝑥

𝜕𝑁(𝑥,𝑦)⁄𝜕𝑥 −𝜕𝑀(𝑥,𝑦)⁄𝜕𝑦
equation. If 𝑀
is a function of 𝑦
only, i.e. 𝑓(𝑦)

For the equation 𝐼𝐹 = 𝑥 ℎ 𝑦 𝑘 , where ℎ & 𝑘 are such that

′ ′
𝑥 𝑎 𝑥 𝑏 (𝑚𝑦𝑑𝑥 + 𝑛𝑥𝑑𝑦) + 𝑥 𝑎 𝑥 𝑏 (𝑚′ 𝑦𝑑𝑥 + 𝑛′ 𝑥𝑑𝑦) = 0 𝑎+ℎ+1 𝑏+𝑘+1
= ;
𝑚 𝑛
𝑎′ + ℎ + 1 𝑏 ′ + 𝑘 + 1
= .
𝑚′ 𝑛′

Note: This is for your extra knowledge and other types of integrating factors are not included in this study.

11.5.3 SEPARABLE DIFFERENTIAL EQUATION

A separable differential equation can be written in the form of

𝑔(𝑦)𝑑𝑦 = 𝑓(𝑥)𝑑𝑥
where the variables are separable
(LHS = function in 𝑦 variable; RHS = function in 𝑥 variable).

Then, we can solve the problem by integrating both sides. This is particularly useful in solving certain linear
& nonlinear differential equation which is separable.

98
 𝑑𝑦
For example: Solve the equation 9𝑦 𝑑𝑥 − 4𝑥 = 0.

Solution:
Recognize that this is a nonlinear equation that is difficult to be solved by using Linear or Exact Differential
Equations. However, it can be solved easily by using the Separable Differential Equation.
𝑑𝑦
9𝑦 𝑑𝑥 − 4𝑥 = 0 [Comment- Nonlinear form]

Think: What is the nonlinear component?

>> 9𝑦𝑑𝑦 = 4𝑥𝑑𝑥 [Step 1- Separable form]

>> ∫ 9𝑦𝑑𝑦 = ∫ 4𝑥𝑑𝑥 [Step 2- Integrate both sides]
9 2
∴ 2
𝑦 = 2𝑥 2 + 𝐶

For example: Solve the equation 𝑦 ′ = 1 + 𝑦 2 .

Solution:
𝑑𝑦
𝑑𝑥
− 𝑦2 = 1 [Comment- Nonlinear form]
𝑑𝑦
>> = 1 + 𝑦2
𝑑𝑥
𝑑𝑦
>> = 𝑑𝑥 [Step 1- Separable form]
1+𝑦 2
𝑑𝑦
>> ∫ = ∫ 𝑑𝑥 [Step 2- Integrate both sides]
1+𝑦 2

>> 𝑡𝑎𝑛−1 𝑦 = 𝑥 + 𝐶
∴ 𝑦 = 𝑡𝑎𝑛(𝑥 + 𝐶)

𝑑𝑦 𝑦+1
For example: Solve the equation 𝑑𝑥 = 𝑥−4 , 𝑦(6) = 0

Solution:
𝑑𝑦 𝑦+1
𝑑𝑥
= 𝑥−4 [Comment- Nonlinear form]
1 1
>>∫ 𝑦+1 𝑑𝑦 = ∫ 𝑥−4 𝑑𝑥

>> 𝑙𝑛|𝑦 + 1| = 𝑙𝑛|𝑥 − 4| + 𝐶

Apply initial condition, 𝑦(6) = 0 to solve the unknown 𝐶

99
 >> 𝑙𝑛|0 + 1| = 𝑙𝑛|6 − 4| + 𝐶
>> 𝐶 = −𝑙𝑛|2|
>> 𝑙𝑛|𝑦 + 1| = 𝑙𝑛|𝑥 − 4| − 𝑙𝑛|2|
|𝑥−4|
>> 𝑙𝑛|𝑦 + 1| = 𝑙𝑛 |2|
𝑥−4
>> 𝑦 + 1 = 2
𝑥
∴𝑦 = 2−3

11.5.4 BERNOULLI’S DIFFERENTIAL EQUATION

There are some cases cannot be solved by the previous strategies due to its non-linear and non-separable
characteristics, however, if it is in the form of Bernoulli’s equation, it can be converted to the linear
differential equation and hence solved by using the Integrating Factor strategy.

A Bernoulli’s differential equation has the form of

𝑑𝑦
+ 𝑝(𝑥)𝑦 = 𝑟(𝑥)𝑦 𝑛
𝑑𝑥
Note: The component 𝑦 𝑛 decides whether it is linear or nonlinear differential equation.

𝑑𝑦
For 𝑛 = 0, 𝑑𝑥
+ 𝑝(𝑥)𝑦 = 𝑟(𝑥) [linear differential equation]
𝑑𝑦
For 𝑛 = 1, 𝑑𝑥
+ (𝑝(𝑥) − 𝑟(𝑥))𝑦 = 0 [linear differential equation]
𝑑𝑦
For 𝑛 > 1, 𝑑𝑥
+ 𝑝(𝑥)𝑦 = 𝑟(𝑥)𝑦 𝑛 [nonlinear differential equation]

To convert the nonlinear Bernoulli’s equation into linear form, we need two important properties:
[i] Let 𝑣(𝑥) = {𝑦(𝑥)}1−𝑛
where 𝑦 = 𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒 & 𝑥 = 𝑖𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒

1 𝑑𝑣(𝑥) 𝑑𝑦(𝑥)
[ii] The derivative, (1−𝑛) 𝑑𝑥
= {𝑦(𝑥)}−𝑛 𝑑𝑥

Prove:
𝑑𝑣(𝑥) 𝑑𝑦(𝑥)
= (1 − 𝑛){𝑦(𝑥)}(1−𝑛)−1
𝑑𝑥 𝑑𝑥
1 𝑑𝑣(𝑥) 𝑑𝑦(𝑥)
>> (1−𝑛) 𝑑𝑥
= {𝑦(𝑥)}−𝑛 𝑑𝑥

Substitute eqns. [i] & [ii] to the nonlinear Bernoulli’s equation:

100
 𝑑𝑦
For 𝑛 > 1, + 𝑝(𝑥)𝑦 = 𝑟(𝑥)𝑦 𝑛 [Comment: nonlinear Bernoulli’s equation]
𝑑𝑥

where 𝑦 = 𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒 & 𝑥 = 𝑖𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒

𝑑𝑦
Divide by 𝑦 𝑛 and we get 𝑦 −𝑛 𝑑𝑥 + 𝑝(𝑥)𝑦1−𝑛 = 𝑟(𝑥) [Step 1: Divide by 𝑦 𝑛 ]

Substitute the eqns.

(i) 𝑣(𝑥) = {𝑦(𝑥)}1−𝑛 ;
1 𝑑𝑣(𝑥) −𝑛 𝑑𝑦(𝑥)
(ii) (1−𝑛) 𝑑𝑥
= {𝑦(𝑥)} . 𝑑𝑥

We get [Step 2: Substitution method]

1 𝑑𝑣(𝑥)
>> (1−𝑛) 𝑑𝑥
+ 𝑝(𝑥)𝑣(𝑥) = 𝑟(𝑥) [Comment: linear Bernoulli’s equation]

where 𝑣 = 𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒 & 𝑥 = 𝑖𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒

Hence, the 1st order linear differential equation can be solved using the previous strategy. Once you obtain
the solution 𝑣. Back substitute it into eqn. 𝑣(𝑥) = {𝑦(𝑥)}1−𝑛 to obtain the solution 𝑦.

𝑑𝑦 3
For example: Solve the equation 𝑑𝑥 + 3𝑥 2 𝑦 = 𝑥𝑒 𝑥 𝑦 2

Solution:
𝑑𝑦 3
+ 3𝑥 2 𝑦 = 𝑥𝑒 𝑥 𝑦 2
𝑑𝑥

Think: What is the nonlinear component.

𝑑𝑦
Recognize that it is nonlinear Bernoulli eqn. 𝑑𝑥 + 𝑝(𝑥)𝑦 = 𝑟(𝑥)𝑦 𝑛 ,
3
With 𝑝(𝑥) = 3𝑥 2 , 𝑟(𝑥) = 𝑥𝑒 𝑥 , 𝑛 = 2

[Step 1: Divide by 𝑦 𝑛 ]
𝑑𝑦 3
>> 𝑦 −2 𝑑𝑥 + 3𝑥 2 𝑦 −1 = 𝑥𝑒 𝑥

[Step 2: Substitution method]

>> Let 𝑣(𝑥) = {𝑦(𝑥)}1−2 = {𝑦(𝑥)}−1 ;
1 𝑑𝑣(𝑥) 𝑑𝑦(𝑥) 𝑑𝑣(𝑥)
>> Let (1−𝑛) 𝑑𝑥
= {𝑦(𝑥)}−2 𝑑𝑥
=− 𝑑𝑥
.

101
 𝑑𝑦 3
𝑦 −2 + 3𝑥 2 𝑦 −1 = 𝑥𝑒 𝑥
𝑑𝑥
𝑑𝑣(𝑥) 3
>> − 𝑑𝑥
+ 3𝑥 2 𝑣(𝑥) = 𝑥𝑒 𝑥 [Comment: linear Bernoulli’s equation]

[Step 3: Linear Differential equation]

𝑑𝑣(𝑥) 3
− 𝑑𝑥
+ 3𝑥 2 𝑣(𝑥) = 𝑥𝑒 𝑥
𝑑𝑣(𝑥) 3
>> − 3𝑥 2 𝑣(𝑥) = −𝑥𝑒 𝑥 [Step 1- Linear Form]
𝑑𝑥

where
𝑣 = 𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒
𝑥 = 𝑖𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒
𝑝(𝑥)=−3𝑥 2
3
𝑟(𝑥) = −𝑥𝑒 𝑥

2 𝑑𝑥 3
>> The integrating factor, 𝐼𝐹 = 𝑒 ∫ 𝑝(𝑥)𝑑𝑥 = 𝑒 ∫ −3𝑥 = 𝑒 −𝑥 [Step 2- IF]
3 𝑑𝑣(𝑥) 3 3 3
>> 𝑒 −𝑥 𝑑𝑥
− 𝑒 −𝑥 (3𝑥 2 𝑣(𝑥)) = 𝑒 −𝑥 (−𝑥𝑒 𝑥 ) [Step 3- Multiply]
3 𝑑𝑣(𝑥) 3
>> 𝑒 −𝑥 𝑑𝑥
− 𝑒 −𝑥 (3𝑥 2 𝑣(𝑥)) = −𝑥

𝑑 −𝑥 3
>> 𝑑𝑥
(𝑒 . 𝑣(𝑥)) = −𝑥 [Step 4- Exact]
3
where 𝐼𝐹 = 𝑒 −𝑥 & 𝑣 = 𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒

𝑑 3
>> ∫ 𝑑𝑥 (𝑒 −𝑥 . 𝑣(𝑥)) 𝑑𝑥 = ∫(−𝑥)𝑑𝑥 [Step 5- Integrate]
3 −𝑥 2
>> 𝑒 −𝑥 . 𝑣(𝑥) = 2
+𝐶

[Step 4: Back Substitution]

3 −𝑥 2
>> Previously we obtained 𝑣(𝑥) = {𝑦(𝑥)}−1 and 𝑒 −𝑥 . 𝑣(𝑥) = 2
+𝐶
3 −𝑥 2
>> 𝑒 −𝑥 . {𝑦(𝑥)}−1 = +𝐶
2
3 1 2
∴ 𝑦(𝑥) = 𝑒 −𝑥 . −𝑥2 = 3 , where 𝐶 = arbitrary constant.
+𝐶 𝑒 𝑥 (2𝐶−𝑥 2 )
2

102
 𝑑𝑦
For example: Solve the equation 3 𝑑𝑥 + 𝑦 = (1 − 2𝑥)𝑦 4 .

Solution:
𝑑𝑦
3 + 𝑦 = (1 − 2𝑥)𝑦 4
𝑑𝑥

Think: What is the nonlinear component.

𝑑𝑦
Recognize that it is nonlinear Bernoulli eqn. 𝑑𝑥 + 𝑝(𝑥)𝑦 = 𝑟(𝑥)𝑦 𝑛 ,
1 (1−2𝑥)
With 𝑝(𝑥) = , 𝑟(𝑥) = ,𝑛 =4
3 3

[Step 1: Divide by 𝑦 𝑛 ]
𝑑𝑦 𝑦 (1−2𝑥) 4
𝑑𝑥
+3= 3
𝑦
𝑑𝑦 𝑦 −3 (1−2𝑥)
>> 𝑦 −4 𝑑𝑥 + 3
= 3

[Step 2: Substitution method]

>> Let 𝑣(𝑥) = {𝑦(𝑥)}1−4 = {𝑦(𝑥)}−3 ;
1 𝑑𝑣(𝑥) 𝑑𝑦(𝑥) 1 𝑑𝑣(𝑥)
>> Let = {𝑦(𝑥)}−4 = .
(1−𝑛) 𝑑𝑥 𝑑𝑥 −3 𝑑𝑥

𝑑𝑦 𝑦 −3 (1−2𝑥)
𝑦 −4 𝑑𝑥 + 3
= 3
1 𝑑𝑣(𝑥) 𝑣(𝑥) (1−2𝑥)
>> −3 𝑑𝑥
+ 3
= 3
𝑑𝑣(𝑥)
>> − 𝑑𝑥
+ 𝑣(𝑥) = (1 − 2𝑥) [Comment: linear Bernoulli’s equation]

[Step 3: Linear Differential equation]

𝑑𝑣(𝑥)
− + 𝑣(𝑥) = (1 − 2𝑥)
𝑑𝑥
𝑑𝑣(𝑥)
>> 𝑑𝑥
− 𝑣(𝑥) = (2𝑥 − 1) [Step a- Linear Form]

where
𝑣 = 𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒
𝑥 = 𝑖𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒
𝑝(𝑥)=−1
𝑟(𝑥) = (2𝑥 − 1)

103
>> The integrating factor, 𝐼𝐹 = 𝑒 ∫ 𝑝(𝑥)𝑑𝑥 = 𝑒 ∫ −1𝑑𝑥 = 𝑒 −𝑥 [Step b- IF]
𝑑𝑣(𝑥)
>> 𝑒 −𝑥 𝑑𝑥
− 𝑒 −𝑥 𝑣(𝑥) = 𝑒 −𝑥 (2𝑥 − 1) [Step c- Multiply]

𝑑
>> 𝑑𝑥
(𝑒 −𝑥 . 𝑣(𝑥)) = 𝑒 −𝑥 (2𝑥 − 1) [Step d- Exact]
where 𝐼𝐹 = 𝑒 −𝑥 & 𝑣 = 𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒

𝑑
>> ∫
𝑑𝑥
(𝑒 −𝑥 . 𝑣(𝑥))𝑑𝑥 = ∫ 𝑒 −𝑥 (2𝑥 − 1)𝑑𝑥 [Step e- Integrate]

>> 𝑒 −𝑥 . 𝑣(𝑥) = ∫ 𝑒 −𝑥 (2𝑥) − 𝑒 −𝑥 𝑑𝑥

Where ∫ 𝑒 −𝑥 (2𝑥) − 𝑒 −𝑥 𝑑𝑥 = ∫ 𝑒 −𝑥 (2𝑥) − 2𝑒 −𝑥 + 𝑒 −𝑥 𝑑𝑥

= ∫ 𝑒 −𝑥 (2𝑥) − 2𝑒 −𝑥 𝑑𝑥 + ∫ 𝑒 −𝑥 𝑑𝑥
= −2𝑥𝑒 −𝑥 + (−𝑒 −𝑥 ) + 𝐶
Hint: Product rule, i.e. ∫ 𝑓 ′ (𝑥)𝑔(𝑥) + 𝑓(𝑥)𝑔′ (𝑥) 𝑑𝑥 = 𝑓(𝑥)𝑔(𝑥) + 𝐶
Think: Can we use integration by part, i.e. ∫ 𝑢 𝑑𝑣 = 𝑢𝑣 − ∫ 𝑣 𝑑𝑢 to solve the integration problem?

>> 𝑒 −𝑥 . 𝑣(𝑥) = −2𝑥𝑒 −𝑥 + (−𝑒 −𝑥 ) + 𝐶

[Step 4: Back Substitution]

>> Previously we obtained 𝑣(𝑥) = {𝑦(𝑥)}−3 and 𝑒 −𝑥 . 𝑣(𝑥) = −2𝑥𝑒 −𝑥 − 𝑒 −𝑥 + 𝐶
>> 𝑒 −𝑥 . {𝑦(𝑥)}−3 = −2𝑥𝑒 −𝑥 − 𝑒 −𝑥 + 𝐶
1 1
>> {𝑦(𝑥)}3 = 𝑒 −𝑥 . −2𝑥𝑒 −𝑥 −𝑒 −𝑥 +𝐶 = −2𝑥−1+𝐶𝑒 𝑥

3 1
∴ 𝑦(𝑥) = √−2𝑥−1+𝐶𝑒 𝑥 , where 𝐶 = arbitrary constant.

104
11.5.5 DIFFERENTIAL EQUATION OF HOMOGENEOUS 𝑑𝑦/𝑑𝑥 = 𝑓(𝑥, 𝑦)/𝑔(𝑥, 𝑦) FORM
𝑑𝑦 𝑓(𝑥,𝑦)
Previously, we use the 𝑑𝑥 = 𝑔(𝑥,𝑦) form to check the homogeneity of 1st order nonlinear ODE. This section
𝑑𝑦 𝑓(𝑥,𝑦)
introduces the strategy used to solve the homogeneous 𝑑𝑥 = 𝑔(𝑥,𝑦).

A homogeneous differential equation has the form of

𝑑𝑦 𝑓(𝑥,𝑦)
𝑑𝑥
= 𝑔(𝑥,𝑦)

if 𝑓(𝑥, 𝑦) and 𝑔(𝑥, 𝑦) are homogeneous of the same degree, where

𝑓(𝜆𝑥, 𝜆𝑦) = 𝜆𝑛1 𝑓(𝑥, 𝑦)
𝑔(𝜆𝑥, 𝜆𝑦) = 𝜆𝑛2 𝑔(𝑥, 𝑦)
& 𝑛1 = 𝑛2

𝑑𝑦
For example: Solve 2𝑥𝑦 − 𝑦 2 = −𝑥 2
𝑑𝑥

Think: What is the nonlinear component? Is it in Bernoulli eqn. form?

Hint: If it is non-linear and can’t be in Bernoulli eqn. form, we need to use other method to solve the 1st order
nonlinear ODE equation.

𝑑𝑦 𝑓(𝑥,𝑦)
Rearrange it to the form of = , we get
𝑑𝑥 𝑔(𝑥,𝑦)

𝑑𝑦 𝑦 2 −𝑥 2
>> =
𝑑𝑥 2𝑥𝑦

Check its homogeneous degree

𝑓(𝜆𝑥,𝜆𝑦) (𝜆𝑦)2 −(𝜆𝑥)2 𝜆2 ((𝑦)2 −(𝑥)2 ) 𝜆2 𝑓(𝑥,𝑦)
>> 𝑔(𝜆𝑥,𝜆𝑦) = 2(𝜆𝑥)(𝜆𝑦)
= 𝜆2 (2(𝑥𝑦))
= 𝜆2 𝑔(𝑥,𝑦)
𝑑𝑦 𝑓(𝑥,𝑦)
∴ It is a 1st order nonlinear ODE with homogeneous 𝑑𝑥 = 𝑔(𝑥,𝑦) form of degree 2.

𝑑𝑦 𝑓(𝑥,𝑦)
The 1st order nonlinear ODE with homogeneous 𝑑𝑥 = 𝑔(𝑥,𝑦) form can be solved by using substitution
method as shown in the table below. In general, this form is very useful to convert the non-separable
differential equation into separable differential equation.

105
 𝑑𝑦 𝑓(𝑥,𝑦)
To convert the 1st order nonlinear ODE with homogeneous 𝑑𝑥 = 𝑔(𝑥,𝑦) form into separable form, we need
two important properties:
𝑦
[i] Let 𝑣(𝑥) = 𝑥 or 𝑦 = 𝑣𝑥
where 𝑦 = 𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒 & 𝑥 = 𝑖𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒

𝑑𝑦 𝑑
[ii] The derivative, 𝑑𝑥 = 𝑥 𝑑𝑥 [𝑣] + 𝑣

𝑑𝑦
For example: Solve 2𝑥𝑦 − 𝑦 2 = −𝑥 2
𝑑𝑥

Solution:
𝑑𝑦
2𝑥𝑦 𝑑𝑥 − 𝑦 2 = −𝑥 2

𝑑𝑦 𝑦 2 −𝑥 2
>> 𝑑𝑥 = 2𝑥𝑦

𝑑𝑦 𝑓(𝑥,𝑦)
[Comment: 1st order nonlinear ODE with homogeneous 𝑑𝑥 = 𝑔(𝑥,𝑦) form; non-separable form]

Step 1: Substitution method

𝑦 = 𝑣𝑥
𝑑𝑦 𝑑
=𝑥 [𝑣] + 𝑣
𝑑𝑥 𝑑𝑥

𝑑𝑦 (𝑣𝑥)2 −𝑥 2 𝑑
>> = =𝑥 [𝑣] + 𝑣
𝑑𝑥 2𝑥(𝑣𝑥) 𝑑𝑥

(𝑣)2 −1 𝑑𝑣
>> −𝑣 = 𝑥 [Comment: Solve using separable differential equation]
2(𝑣) 𝑑𝑥

1 1
>> 𝑥 𝑑𝑥 = (𝑣)2−1 𝑑𝑣
−𝑣
2(𝑣)

1 1
>> 𝑥 𝑑𝑥 = 𝑣2 −1−2𝑣2
𝑑𝑣
2(𝑣)

1 2𝑣
>> 𝑥 𝑑𝑥 = −𝑣 2 −1 𝑑𝑣 [Step a- Separable form]

1 2𝑣
>>∫ 𝑥 𝑑𝑥 = ∫ −𝑣2 −1 𝑑𝑣 [Step b- Integrate both sides]

>> 𝑙𝑛|𝑥| = −𝑙𝑛|−𝑣 2 − 1| + 𝐶

2𝑣
Hint: Solve ∫ −𝑣 2 −1 𝑑𝑣 using substitution method, let 𝑢 = −𝑣 2 − 1

106
Step 2: Back Substitution

>> Previously we obtained 𝑦 = 𝑣𝑥 & 𝑙𝑛|𝑥| = −𝑙𝑛|−𝑣 2 − 1| + 𝐶

𝑦 2
>> 𝑙𝑛|𝑥| = −𝑙𝑛 |− (𝑥 ) − 1| + 𝐶

𝑦 2
>> 𝑙𝑛|𝑥| + 𝑙𝑛 |− (𝑥 ) − 1| = +𝐶

𝑦2
>> 𝑙𝑛 |− − 𝑥| = +𝐶
𝑥

𝑦2
>> − 𝑥
− 𝑥 = 𝑒𝐶

>> 𝑦 2 = −𝑥 2 − 𝑥𝑒 𝐶

∴ 𝑦 = ±√−𝑥 2 − 𝑥𝑒 𝐶

11.5.6 DIFFERENTIAL EQUATION OF NONHOMOGENEOUS 𝑑𝑦/𝑑𝑥 = 𝑓(𝑥, 𝑦)/𝑔(𝑥, 𝑦) FORM

𝑑𝑦 𝑓(𝑥,𝑦)
Previous strategies/ methods are not sufficient to solve the nonhomogeneous 𝑑𝑥 = 𝑔(𝑥,𝑦) problem such as
𝑑𝑦 2𝑥−4𝑦+5
𝑑𝑥
= 𝑥−2𝑦+3
. It involves other types of substitution to solve the problem. For example, for the
𝑑𝑦 𝑓(𝑥,𝑦) 𝑎1 𝑥+𝑏1 𝑦+𝑐1
nonhomogeneous = = , we use the substitution as follows:
𝑑𝑥 𝑔(𝑥,𝑦) 𝑎2 𝑥+𝑏2 𝑦+𝑐2

Case I Case II
𝑎1 𝑏1 𝑎 𝑏1
| |=0 | 1 |≠0
𝑎2 𝑏2 𝑎2 𝑏2

Let 𝑣 = 𝑎1 𝑥 + 𝑏1 𝑦 𝑥 =𝑋+ℎ
Let
𝑦 = 𝑌+𝑘
where the pair constant (ℎ, 𝑘) can be obtained by
solving the simultaneous equations:
𝑎1 ℎ + 𝑏1 𝑘 + 𝑐1 = 0
𝑎2 ℎ + 𝑏2 𝑘 + 𝑐2 = 0
𝑑𝑦 𝑓(𝑥,𝑦)
The substitution converts the nonhomogeneous = to homogeneous form and hence it can be
𝑑𝑥 𝑔(𝑥,𝑦)
solved by method introduced in section 11.5.5.

𝑑𝑦 𝑓(𝑥,𝑦)
The nonlinear nonhomogeneous 𝑑𝑥 = 𝑔(𝑥,𝑦) has a lot of branches and it may require various types
of substitution to convert it to homogeneous form. The topic is wide and thus it is not covered in this study.

107
SOLUTION TO HOMOGENEOUS & NON-
HOMOGENEOUS
WEEK 12: SOLUTION TO HOMOGENEOUS & NON-HOMOGENEOUS
12.1 INTRODUCTION TO 2 ND ORDER DIFFERENTIAL EQUATION

As discussed earlier, the differential equation is important, especially in mathematical modelling for
engineering application. Previously, we have discussed several strategies and methods to solve the 1st order
differential equation. However, the 1st order differential equation is only able to model certain engineering
problem.

For example,

1st order differential equation is sufficient to model a 𝑅𝐶 electrical circuit,

𝑑𝑞(𝑡) 1
i.e. 𝑅 𝑑𝑡
+ 𝐶 𝑞(𝑡) = 𝐸(𝑡).

However, this differential equation is insufficient to describe a 𝑅𝐿𝐶 circuit as illustrated below:

where the RLC electric circuit consists a generator (𝐸 volt), a resistance (𝑅 ohms), a inductance (𝐿 Henry), a
capacitor (C capacitance).

Example: 𝑅𝐿𝐶 Circuit

We know that the voltage loss through a resistor, capacitor and inductor is proportional to the current,
charge and rate of charge change respectively, where

Δ𝑉𝑟𝑒𝑠𝑖𝑠𝑡𝑜𝑟 (𝑡) ∝ 𝑖(𝑡)

Δ𝑉𝑐𝑎𝑝𝑎𝑐𝑖𝑡𝑜𝑟 (𝑡) ∝ 𝑞(𝑡)

Δ𝑉𝑖𝑛𝑑𝑢𝑐𝑡𝑜𝑟 (𝑡) ∝ 𝑑𝑖 ⁄𝑑𝑡

108
 According to Kirchhoff’s Law, the sum of all the voltages around any closed network/ loop is equal
to zero. Based on the theory, a mathematical model involving differential equation can be derived.
(𝑉𝐵 − 𝑉𝐴 ) + (𝑉𝐹 − 𝑉𝐵 ) + (𝑉𝐷 − 𝑉𝐹 ) + (𝑉𝐴 − 𝑉𝐷 ) = 0

The voltage drops across various components are provided as follow:

Components Voltage Drop
Resistor 𝑑𝑞(𝑡)
𝛥𝑉(𝑡) = (𝑉𝐵 − 𝑉𝐴 ) = 𝑅𝑖(𝑡) = 𝑅
𝑑𝑡

Capacitor 1
𝛥𝑉(𝑡) = (𝑉𝐹 − 𝑉𝐵 ) = 𝐶 𝑞(𝑡)

Inductor 𝑑𝑖(𝑡) 𝑑 2 𝑞(𝑡)

𝛥𝑉(𝑡) = (𝑉𝐷 − 𝑉𝐹 ) = 𝐿 𝑑𝑡
=𝐿 𝑑𝑡 2

Battery/ Generator 𝛥𝑉(𝑡) = (𝑉𝐴 − 𝑉𝐷 ) = −𝐸(𝑡)

𝑑𝑞(𝑡) 1 𝑑 2 𝑞(𝑡)
>> 𝑅 𝑑𝑡
+ 𝐶 𝑞(𝑡) + 𝐿 𝑑𝑡 2
− 𝐸(𝑡) = 0.

Rearrange it, we get the 2nd order linear differential equation of

𝑑 2 𝑞(𝑡) 𝑑𝑞(𝑡) 1
>> 𝐿 +𝑅 + 𝑞(𝑡) = 𝐸(𝑡).
𝑑𝑡 2 𝑑𝑡 𝐶

In this chapter, we will discuss about the strategies used to solve the 2nd order differential equation.

12.2 SOLUTION TO 2 ND ORDER LINEAR DIFFERENTIAL EQUATION

Previously, we know that 2nd order linear differential equation can be categorized into two forms:

𝑑 2 𝑦(𝑥) 𝑑𝑦(𝑥)
(i) Homogeneous (i.e. 𝑎 𝑑𝑥 2
+𝑏 𝑑𝑥
+ 𝑐𝑦(𝑥) = 0.)
Solution: Known as complementary solution, 𝑦𝑐

Example:
𝑑2 𝑦 𝑑𝑦(𝑥)
The solution to 𝑑𝑥 2 − 4 𝑑𝑥
+ 3𝑦(𝑥) = 0 is 𝑦𝑐 = 𝑐1 𝑒 3𝑥 + 𝑐2 𝑒 𝑥

𝑑 2 𝑦(𝑥) 𝑑𝑦(𝑥)
(ii) Nonhomogeneous (i.e. 𝑎 𝑑𝑥 2
+𝑏 𝑑𝑥
+ 𝑐𝑦(𝑥) = 𝑟(𝑥)).
Solution: Combination of complementary solution, 𝑦𝑐 and particular solution, 𝑦𝑝

Example:
𝑑2 𝑦 𝑑𝑦(𝑥) 2
The solution to 𝑑𝑥 2 − 4 𝑑𝑥
+ 3𝑦(𝑥) = 10𝑒 −2𝑥 is 𝑦𝑐 + 𝑦𝑝 = 𝑐1 𝑒 3𝑥 + 𝑐2 𝑒 𝑥 + 3 𝑒 −2𝑥

109
Note: We assume 𝑦 = dependent variable and 𝑥 = independent variable in the above example. It is
worthwhile to mention that the methods to obtain the yc is same for both homogeneous and
nonhomogeneous cases. Just that we have an additional solution y p for nonhomogeneous case.

12.3 GENERAL THEORY OF LINEARITY PRINCIPLE & LINEAR DEPENDENCY

Before we proceed to the strategy in solving the 2nd order differential equation, we should know about the
theory of linearity principle and linearly dependency.

The importance of the theory is illustrated in the following example:

𝑑 2 𝑦(𝑥) 𝑑𝑦(𝑥)
It is given that the solution to 𝑑𝑥 2
−4 𝑑𝑥
+ 3𝑦(𝑥) = 0 is 𝑦𝑐 = 𝑐1 𝑒 3𝑥 + 𝑐2 𝑒 𝑥

To prove that it is the correct solution, let us do the verification:

Case (1): Assume the solution is 𝑦1 = 𝑒 3𝑥

Verification:
𝑦1 = 𝑒 3𝑥
𝑑𝑦1
>> 𝑑𝑥
= 3𝑒 3𝑥
𝑑 2 𝑦1
>> 𝑑𝑥 2
= 9𝑒 3𝑥

Substitute to LHS
𝑑 2 𝑦(𝑥) 𝑑𝑦(𝑥)
𝑑𝑥 2
−4 𝑑𝑥
+ 3𝑦(𝑥) = 9𝑒 3𝑥 − 4(3𝑒 3𝑥 ) + 3𝑒 3𝑥 = 0
>> LHS = RHS=0
𝑑 2 𝑦(𝑥) 𝑑𝑦(𝑥)
∴ 𝑦1 = 𝑒 3𝑥 is proven to be the solution of 𝑑𝑥 2
−4 𝑑𝑥
+ 3𝑦(𝑥) = 0

Case (2): Assume the solution is 𝑦2 = 𝑒 𝑥

Verification:
𝑦2 = 𝑒 𝑥
𝑑𝑦2
>> 𝑑𝑥
= 𝑒𝑥
𝑑 2 𝑦2
>> = 𝑒𝑥
𝑑𝑥 2

110
 Substitute to LHS
𝑑 2 𝑦(𝑥) 𝑑𝑦(𝑥)
−4 + 3𝑦(𝑥) = 𝑒 𝑥 − 4(𝑒 𝑥 ) + 3𝑒 𝑥 = 0
𝑑𝑥 2 𝑑𝑥

>> LHS = RHS=0

𝑑 2 𝑦(𝑥) 𝑑𝑦(𝑥)
∴ 𝑦2 = 𝑒 𝑥 is proven to be the solution of 𝑑𝑥 2
−4 𝑑𝑥
+ 3𝑦(𝑥) = 0

Case (3): Assume the solution is 𝑦𝑐 = 𝑐1 𝑒 3𝑥 + 𝑐2 𝑒 𝑥

Verification:
𝑦𝑐 = 𝑐1 𝑒 3𝑥 + 𝑐2 𝑒 𝑥
𝑑𝑦𝑐
>> = 3𝑐1 𝑒 3𝑥 + 𝑐2 𝑒 𝑥
𝑑𝑥
𝑑 2 𝑦𝑐
>> 𝑑𝑥 2
= 9𝑐1 𝑒 3𝑥 + 𝑐2 𝑒 𝑥

Substitute to LHS
𝑑 2 𝑦(𝑥) 𝑑𝑦(𝑥)
𝑑𝑥 2
−4 𝑑𝑥
+ 3𝑦(𝑥) = (9𝑐1 𝑒 3𝑥 + 𝑐2 𝑒 𝑥 ) − 4(3𝑐1 𝑒 3𝑥 + 𝑐2 𝑒 𝑥 ) + 3(𝑐1 𝑒 3𝑥 + 𝑐2 𝑒 𝑥 ) = 0
>> LHS = RHS=0
𝑑 2 𝑦(𝑥) 𝑑𝑦(𝑥)
∴ 𝑦𝑐 = 𝑐1 𝑒 3𝑥 + 𝑐2 𝑒 𝑥 is proven to be the solution of 𝑑𝑥 2
−4 𝑑𝑥
+ 3𝑦(𝑥) = 0

From the verification, it was shown that all the three solutions satisfy the ODE equation and thus they are
true solutions. The complete solution is formed according to Linearity Principle/ Principle of Superposition
as follows:

Linearity Principle/ Principle of Superposition:

If 𝑦1 & 𝑦2 are both solutions
of the homogeneous linear differential equation.
Then so is the solution 𝑦𝑐 = 𝑐1 𝑦1 + 𝑐2 𝑦2
where 𝑐1 & 𝑐2 are arbitrary constants.

𝑑 2 𝑦(𝑥) 𝑑𝑦(𝑥)
So, the solution of the 2nd order differential equation ( 𝑑𝑥 2
−4 𝑑𝑥
+ 3𝑦(𝑥) = 0) is not purely
3𝑥 𝑥
𝑦1 = 𝑒 or 𝑦2 = 𝑒 . If a 2 order linear differential equation is encountered, the complete solution will be
nd

equal to 𝑦 = 𝑐1 𝑦1 (𝑥) + 𝑐2 𝑦2 (𝑥) = 𝑐1 𝑒 3𝑥 + 𝑐2 𝑒 𝑥 .

111
 In conclusion, the general complementary solution of the homogeneous linear differential
equation is equal to

𝑦 = 𝑐1 𝑦1 (𝑥) + 𝑐2 𝑦2 (𝑥)

where 𝑦1 (𝑥) & 𝑦2 (𝑥) are also the solution for the equation and they are linearly independent to each other.

Frankly speaking, if two solutions are linearly dependent, it means these two solutions are
redundant and hence they do not represent two independent solutions instead of one. Thus, linearly
independent solutions of 𝑦1 (𝑥) & 𝑦2 (𝑥) are desired. We can use the following two methods to check
whether two solutions are linearly independent to each other or not: (a) Linear Dependency Theorem (b)
Wronskian Method.

Let 𝑥 = independent variable; 𝑦 = dependent variable

Method 1
(Linear Dependency Theorem)
(i) Linearly dependent 𝑦1 (𝑥) & 𝑦2 (𝑥) are linearly dependent
[Undesired solutions for if 𝑐1 𝑦1 (𝑥) + 𝑐2 𝑦2 (𝑥) = 0
ODE]
where 𝑐1 & 𝑐2 ≠ 0

Or in other words,
𝑦1 (𝑥) & 𝑦2 (𝑥) are proportional to each other

For example: If 𝑦1 = 𝑒 𝑥 & 𝑦2 = 2𝑒 𝑥 are linearly dependent, 𝑐1 𝑒 𝑥 + 𝑐2 (2𝑒 𝑥 ) = 0 , where

𝑐1 & 𝑐2 ≠ 0.
Check if 𝑦1 = 𝑒 𝑥 & 𝑦2 =
2𝑒 𝑥 are linearly
dependent or not.
It was found that when 𝑐1 = −2 & 𝑐2 = 1,
𝑐1 𝑒 𝑥 + 𝑐2 (2𝑒 𝑥 ) = −2𝑒 𝑥 + 2𝑒 𝑥 = 0

∴ Thus, 𝑦1 = 𝑒 𝑥 & 𝑦2 = 2𝑒 𝑥 are linearly dependent.

(ii) Linearly independent 𝑦1 (𝑥) & 𝑦2 (𝑥) are linearly independent

[Desired solutions of ODE] if 𝑐1 𝑦1 (𝑥) + 𝑐2 𝑦2 (𝑥) = 0
only when 𝑐1 & 𝑐2 = 0

Or in other words,
𝑦1 (𝑥) & 𝑦2 (𝑥) are not proportional to each other

112
For example: If 𝑦1 = 𝑒 𝑥 & 𝑦2 = 𝑒 3𝑥 are linearly independent, 𝑐1 𝑒 𝑥 + 𝑐2 (2𝑒 𝑥 ) = 0 , only
when 𝑐1 & 𝑐2 = 0.
Check if 𝑦1 = 𝑒 𝑥 & 𝑦2 =
𝑒 3𝑥 are linearly
dependent or not.
It was found that only when 𝑐1 = 0 & 𝑐2 = 0,
𝑐1 𝑒 𝑥 + 𝑐2 (𝑒 3𝑥 ) = 0𝑒 𝑥 + 0𝑒 3𝑥 = 0

∴ Thus, 𝑦1 = 𝑒 𝑥 & 𝑦2 = 𝑒 3𝑥 are linearly independent.

Note:
If linearly dependent solutions are obtained,
i.e. 𝑦1 & 𝑦2 are linearly dependent,
we do not obtained a complete solution of
𝑦 = 𝑐1 𝑦1 (𝑥) + 𝑐2 𝑦2 (𝑥).

Thus, extra effort/ treatment should be continue

to obtain another solution which is linearly independent,
i.e. 𝑦1 & 𝑦2 are linearly independent.

Method 2
(Wronskian, 𝑊(𝑦1 , 𝑦2 ))
(i) Linearly dependent 𝑦1 (𝑥) & 𝑦2 (𝑥) are linearly dependent if
[Undesired solutions for ODE] 𝑦1 𝑦2
𝑑𝑦
𝑊(𝑦1 , 𝑦2 ) = | 1 𝑑𝑦 2| = 0
𝑑𝑥 𝑑𝑥
For example: 𝑦1 = 𝑒 𝑥
Check if 𝑦1 = 𝑒 𝑥 & 𝑦2 = 2𝑒 𝑥 >> 𝑑𝑦1 = 𝑒 𝑥
𝑑𝑥
are linearly dependent or not.

𝑦2 = 2𝑒 𝑥
𝑑𝑦
>> 𝑑𝑥2 = 2𝑒 𝑥

𝑦1 𝑦2
𝑒𝑥 2𝑒 𝑥
𝑊(𝑦1 , 𝑦2 ) = |𝑑𝑦1 𝑑𝑦2 | =| 𝑥 | = 𝑒 𝑥 (2𝑒 𝑥 ) − 𝑒 𝑥 (2𝑒 𝑥 ) = 0
𝑑𝑥 𝑑𝑥 𝑒 2𝑒 𝑥

∴ Thus, 𝑦1 = 𝑒 𝑥 & 𝑦2 = 2𝑒 𝑥 are linearly dependent.

113
 (ii) Linearly independent 𝑦1 (𝑥) & 𝑦2 (𝑥) are linearly independent if
[Desired solutions of ODE] 𝑦1 𝑦2
𝑊(𝑦1 , 𝑦2 ) = |𝑑𝑦1 𝑑𝑦2 | ≠ 0
𝑑𝑥 𝑑𝑥

For example: 𝑦1 = 𝑒 𝑥
𝑑𝑦1
Check if 𝑦1 = 𝑒 𝑥 & 𝑦2 = 𝑒 3𝑥 >> = 𝑒𝑥
𝑑𝑥
are linearly dependent or
not.
𝑦2 = 𝑒 3𝑥
𝑑𝑦2
>> = 3𝑒 3𝑥
𝑑𝑥

𝑦1 𝑦2 𝑥
𝑊(𝑦1 , 𝑦2 ) = |𝑑𝑦1 𝑑𝑦2 | = |𝑒 𝑥 𝑒 3𝑥 | = 𝑒 𝑥 (3𝑒 3𝑥 ) − 𝑒 𝑥 (𝑒 3𝑥 ) = 2𝑒 4𝑥
𝑑𝑥 𝑑𝑥 𝑒 3𝑒 3𝑥
Since 𝑊(𝑦1 , 𝑦2 ) ≠ 0
∴ Thus, 𝑦1 = 𝑒 𝑥 & 𝑦2 = 𝑒 3𝑥 are linearly independent.

Note:
If linearly dependent solutions are obtained,
i.e. 𝑦1 & 𝑦2 are linearly dependent,
we do not obtained a complete solution of
𝑦 = 𝑐1 𝑦1 (𝑥) + 𝑐2 𝑦2 (𝑥).

Thus, extra effort/ treatment should be continue

to obtain another solution which is linearly independent,
so that 𝑦1 & 𝑦2 are linearly independent.

114
12.4 STRATEGY TO SOLVE 2 N D ORDER DIFFERENTIAL EQUATION

In this study, we will discuss several strategies to solve the 2nd order linear differential equations that are
given in the following form:

(i) Homogeneous linear differential equation with constant coefficients 𝑎, 𝑏, 𝑐

𝑑2 𝑦 𝑑𝑦
𝑎 𝑑𝑥 2 + 𝑏 𝑑𝑥 + 𝑐𝑦 = 0 [Strategy: Let 𝑦(𝑥) = 𝑒 𝑚𝑥 and use characteristic eqn]

(ii) Homogeneous linear differential equation with non-constant coefficients 𝑥 2 , 𝑎𝑥

(Known as Euler-Cauchy Differential Equation)
𝑑2 𝑦 𝑑𝑦
𝑥 2 𝑑𝑥 2 + 𝑎𝑥 𝑑𝑥 + 𝑏𝑦 = 0 [Strategy: Let solution to be 𝑥 = 𝑒 𝑡 & convert to (i)]

(iii) Nonhomogeneous linear differential equation with constant coefficients 𝑎, 𝑏, 𝑐, in the form of
𝑟(𝑥) = 𝑒 𝛼𝑥 𝑃𝑛 (𝑥)
𝑑2 𝑦 𝑑𝑦
𝑎 +𝑏 + 𝑐𝑦 = 𝑟(𝑥) [Strategy: Method of underdetermined coefficients]
𝑑𝑥 2 𝑑𝑥

(iv) All types of nonhomogeneous linear differential equation

𝑑2 𝑦 𝑑𝑦
𝑑𝑥 2
+ 𝑝(𝑥) 𝑑𝑥 + 𝑞(𝑥)𝑦 = 𝑟(𝑥) [Strategy: Method of variation of parameters]

12.4.1 HOMOGENOUS LINEAR DIFFERENTIAL EQUATION WITH CONSTANT COEFFICIENTS

𝑎, 𝑏, 𝑐

Over the years, scientist and engineer have found that the solution to 2nd order differential equation:
𝑑2 𝑦 𝑑𝑦
Solution of 𝑎 𝑑𝑥 2 + 𝑏 𝑑𝑥 + 𝑐𝑦 = 0

to be 𝑦(𝑥) = 𝑒 𝑚𝑥 ≠ 0

However, this is not a complete solution because 2nd order problem should have 2 linearly independent
solutions. To find the complete solution to the problem, we therefore follow the strategy utilizing the
characteristic/auxiliary equation.
𝑑2 𝑦 𝑑𝑦
To solve 𝑑𝑥 2 + 𝑏 𝑑𝑥 + 𝑐𝑦 = 0 ,

(i) First, form the characteristic equation 𝑎𝑚2 + 𝑏𝑚 + 𝑐 = 0

(ii) Solve the characteristic eqn. and find its roots, 𝑚1 & 𝑚2
−𝑏±√𝑏 2 −4𝑎𝑐
-This can be easily obtained by quadratic formula, 𝑚1 & 𝑚2 = 2𝑎
.
(iii) Check if these two roots are:
(a) Real & distinct root, 𝑚1 ≠ 𝑚2
(b) A pair of complex conjugates roots, 𝑚1 = 𝑚 + 𝑖𝛽 & 𝑚2 = 𝑚 − 𝑖𝛽
(c) Repeated real root, 𝑚 = 𝑚1 = 𝑚2
(iv) Check the table below for the complete solution.

115
Prove: To obtain the characteristic equation: 𝑎𝑚2 + 𝑏𝑚 + 𝑐 = 0

𝑑2 𝑦 𝑑𝑦
𝑎 𝑑𝑥 2 + 𝑏 𝑑𝑥 + 𝑐𝑦 = 0
Assume solution: 𝑦(𝑥) = 𝑒 𝑚𝑥 to be solution of 2nd order ODE
>> Its derivative: 𝑦 ′ = 𝑚𝑒 𝑚𝑥 ; 𝑦′′ = 𝑚2 𝑒 𝑚𝑥
𝑑2 𝑦 𝑑𝑦
Substitute into equation 𝑎 𝑑𝑥 2 + 𝑏 𝑑𝑥 + 𝑐𝑦 = 0 , we get

>> 𝑎𝑚2 𝑒 𝑚𝑥 + 𝑏𝑚𝑒 𝑚𝑥 + 𝑐𝑒 𝑚𝑥 = 0

>> 𝑒 𝑚𝑥 (𝑎𝑚2 + 𝑏𝑚 + 𝑐) = 0
>> Since 𝑒 𝑚𝑥 ≠ 0, we obtain 𝑎𝑚2 + 𝑏𝑚 + 𝑐 = 0, i.e. the characteristic equation.

Recall: Complex conjugate has same magnitude but opposite sign for the imaginary part. For example, the
complex conjugate for a complex number 𝑚1 = (5 + 6𝑖) is 𝑚2 = (5 − 6𝑖).

The summary of the complete solution are listed below. The detail description will be provided next.

Type of (a) Real and distinct (b) A pair of complex conjugates (c) Repeated real root
Roots roots roots 𝑚 = 𝑚1 = 𝑚2
𝑚1 & 𝑚2 𝑚1 = 𝑚 + 𝑖𝛽 & 𝑚2 = 𝑚 − 𝑖𝛽
Indicator 𝑏 2 − 4𝑎𝑐 > 0 𝑏 2 − 4𝑎𝑐 < 0 𝑏 2 − 4𝑎𝑐 = 0
Complete
𝑦(𝑥) = 𝑐1 𝑒 𝑚1 𝑥 + 𝑐2 𝑒 𝑚2 𝑥 𝑦(𝑥) = 𝑐1 𝑒 𝑚1 𝑥 + 𝑐2 𝑥𝑒 𝑚2 𝑥 𝑦(𝑥) = 𝑐1 𝑒 𝑚1 𝑥 + 𝑐2 𝑥𝑒 𝑚2 𝑥
solution
Or

𝑦(𝑥) = 𝑒 𝑚𝑥 (𝐴𝑐𝑜𝑠𝛽𝑥 + 𝐵𝑠𝑖𝑛𝛽𝑥)

where 𝐴 = 𝑐1 + 𝑐2 ;
𝐵 = 𝑖(𝑐1 − 𝑐2 )
𝐴 & 𝐵 are arbitrary constants
Comment -Complementary solution in 𝑒 𝑚𝑥 form -Complementary solution in
-No treatment is needed 𝑒 𝑚𝑥 & 𝑥𝑒 𝑚𝑥 form
-Treatment is needed to
avoid linearly dependent
solution.
Note: Euler formula: 𝑒 ±𝑖𝑥 = 𝑐𝑜𝑠𝑥 ± 𝑖(𝑠𝑖𝑛𝑥); 𝑖 = √−1 = 𝑖𝑚𝑎𝑔𝑖𝑛𝑎𝑟𝑦.

116
 Case (a): Real and distinct roots
𝑚1 ≠ 𝑚2
Characteristic equation:
𝑎𝑚2 + 𝑏𝑚 + 𝑐 = 0

Indicator:
>> 𝑏 2 − 4𝑎𝑐 > 0
Comment: If 𝑏 2 − 4𝑎𝑐 is greater than 0, it indicates that the roots, 𝑚1 & 𝑚2 are real and distinct.

Complete solution:
>> 𝑦(𝑥) = 𝑐1 𝑒 𝑚1 𝑥 + 𝑐2 𝑒 𝑚2 𝑥
where 𝑚1 ≠ 𝑚2

Example of the case for real and distinct roots:

𝑑2 𝑦 𝑑𝑦
𝑑𝑥 2
+ 2 𝑑𝑥 − 3𝑦 = 0

Let 𝑦(𝑥) = 𝑒 𝑚𝑥 , we obtain Characteristic equation: 𝑚2 + 2𝑚 − 3 = 0

Indicator: 𝑏 2 − 4𝑎𝑐 = 22 − 4(1)(−3) = 16

Since 𝑏 2 − 4𝑎𝑐 > 0, it is the case of real and distinct roots.

Solution of Characteristic equation: (𝑚 − 1)(𝑚 + 3) = 0

>> 𝑚1 = 1, 𝑚3 = −3

Complete solution:

∴ 𝑦(𝑥) = 𝑐1 𝑒 𝑥 + 𝑐2 𝑒 −3𝑥

117
 Case (b): A pair of complex conjugates roots
𝑚1 = 𝑚 + 𝑖𝛽 & 𝑚2 = 𝑚 − 𝑖𝛽
Characteristic equation:
𝑎𝑚2 + 𝑏𝑚 + 𝑐 = 0

Indicator:
>> 𝑏 2 − 4𝑎𝑐 < 0
Comment: If 𝑏 2 − 4𝑎𝑐 is less than 0, it indicates that the roots, 𝑚1 & 𝑚2 are a pair of complex conjugates.

Complete solution:
>> 𝑦(𝑥) = 𝑐1 𝑒 𝑚1 𝑥 + 𝑐2 𝑒 𝑚2 𝑥
where 𝑚1 ≠ 𝑚2 ;
𝑚1 = 𝑚 + 𝑖𝛽 & 𝑚2 = 𝑚 − 𝑖𝛽;
𝑖 = √−1 = 𝑖𝑚𝑎𝑔𝑖𝑛𝑎𝑟𝑦
or
>> 𝑦(𝑥) = 𝑒 𝑚𝑥 (𝐴𝑐𝑜𝑠𝛽𝑥 + 𝐵𝑠𝑖𝑛𝛽𝑥)
where 𝐴 = 𝑐1 + 𝑐2 ;
𝐵 = 𝑖(𝑐1 − 𝑐2 )

Note: In this case, the complete solution can be either 𝑦(𝑥) = 𝑐1 𝑒 𝑚1 𝑥 + 𝑐2 𝑒 𝑚2 𝑥 and 𝑦(𝑥) =
𝑒 𝑚𝑥 (𝐴𝑐𝑜𝑠𝛽𝑥 + 𝐵𝑠𝑖𝑛𝛽𝑥). Both are the same equation but in different format. See the proof below.
𝑦(𝑥) = 𝑐1 𝑒 𝑚1 𝑥 + 𝑐2 𝑒 𝑚2 𝑥
>> 𝑦(𝑥) = 𝑐1 𝑒 (𝑚+𝑖𝛽)𝑥 + 𝑐2 𝑒 (𝑚−𝑖𝛽)𝑥
= 𝑐1 𝑒 (𝑚)𝑥 𝑒 (𝑖𝛽)𝑥 + 𝑐2 𝑒 (𝑚)𝑥 𝑒 (−𝑖𝛽)𝑥
= 𝑒 (𝑚)𝑥 ( 𝑐1 𝑒 (𝑖𝛽)𝑥 + 𝑐2 𝑒 (−𝑖𝛽)𝑥 )

Given Euler formula: 𝑒 𝑖𝑥 = 𝑐𝑜𝑠𝑥 + 𝑖𝑠𝑖𝑛𝑥 ; 𝑒 −𝑖𝑥 = 𝑐𝑜𝑠𝑥 − 𝑖(𝑠𝑖𝑛𝑥)

>> 𝑦(𝑥) = 𝑒 𝑚𝑥 ( 𝑐1 (𝑐𝑜𝑠𝛽𝑥 + 𝑖(𝑠𝑖𝑛𝛽𝑥)) + 𝑐2 (𝑐𝑜𝑠𝛽𝑥 − 𝑖(𝑠𝑖𝑛𝛽𝑥)))
= 𝑒 𝑚𝑥 (𝑐𝑜𝑠𝛽𝑥( 𝑐1 + 𝑐2 ) + 𝑖(𝑠𝑖𝑛𝛽𝑥)( 𝑐1 − 𝑐2 ))
= 𝑒 𝑚𝑥 (𝐴𝑐𝑜𝑠𝛽𝑥 + 𝐵𝑠𝑖𝑛𝛽𝑥) [proven]

118
Example of the case of a pair of complex conjugates roots:

𝑑2 𝑦 𝑑𝑦
4 + 16 + 17𝑦 = 0
𝑑𝑥 2 𝑑𝑥

Let 𝑦(𝑥) = 𝑒 𝑚𝑥 , we obtain Characteristic equation: 4𝑚2 + 16𝑚 + 17 = 0

Indicator: 𝑏 2 − 4𝑎𝑐 = 42 − 4(4)(17) = −256

Since 𝑏 2 − 4𝑎𝑐 < 0, it is the case of pair of complex conjugates roots

Solution of Characteristic equation:

Completing the square we get

𝑚2 + 4𝑚 + 17⁄4 = 0

>> (𝑚 + 2)2 − 4 + 17⁄4 = 0

>> (𝑚 + 2)2 = − 1⁄4

1 1
>> 𝑚1 = −2 + 𝑖, 𝑚2 = −2 − 𝑖
2 2

Complete solution:
1 1
∴ 𝑦(𝑥) = 𝑐1 𝑒 (−2+2𝑖)𝑥 + 𝑐2 𝑒 (−2−2𝑖)𝑥

Or
1 1
𝑦(𝑥) = 𝑒 −2𝑥 (𝐴𝑐𝑜𝑠 𝑥 + 𝐵𝑠𝑖𝑛 𝑥)
2 2

119
 Case (c): Repeated real root
𝑚 = 𝑚1 = 𝑚2
Characteristic equation:
𝑎𝑚2 + 𝑏𝑚 + 𝑐 = 0

Indicator:
>> 𝑏 2 − 4𝑎𝑐 = 0
Comment: If 𝑏 2 − 4𝑎𝑐 is equal to 0, it indicates that the roots, 𝑚1 & 𝑚2 are repeated real root.

Complete solution:
Previous solution is not valid here
>> 𝑦(𝑥) = 𝑐1 𝑒 𝑚1 𝑥 + 𝑐2 𝑒 𝑚2 𝑥
where 𝑚1 = 𝑚2 and
thus 𝑐1 𝑒 𝑚1 𝑥 & 𝑐2 𝑒 𝑚2 𝑥 are linearly dependent solution (undesired situation) in this case.

Treatment is needed as follows:

>> 𝑦(𝑥) = 𝑐1 𝑒 𝑚1 𝑥 + 𝑐2 𝑥𝑒 𝑚2 𝑥
where 𝑥 is added so that 𝑐1 𝑒 𝑚1 𝑥 & 𝑐2 𝑥𝑒 𝑚2 𝑥 are linearly independent solution (desired situation).

Example of the case of a pair of complex conjugates roots:

𝑑2 𝑦 𝑑𝑦
𝑑𝑥 2
− 4 𝑑𝑥 + 4𝑦 = 0

Let 𝑦(𝑥) = 𝑒 𝑚𝑥 , we obtain Characteristic equation: 𝑚2 − 4𝑚 + 4 = 0

Indicator: 𝑏 2 − 4𝑎𝑐 = (−4)2 − 4(1)(4) = 0

Since 𝑏 2 − 4𝑎𝑐 < 0, it is the case of repeated real root.

Solution of Characteristic equation:

𝑚2 − 4𝑚 + 4 = 0

>> (𝑚 − 2)(𝑚 − 2) = 0

>> 𝑚1 = 2, 𝑚2 = 2

Complete solution:

∴ 𝑦(𝑥) = 𝑐1 𝑒 2𝑥 + 𝑐2 𝑥𝑒 2𝑥

120
Overall comment:

1. Suppose that the roots of the characteristic equation are 𝑚1 & 𝑚2 , then 𝑒 𝑚1 𝑥 & 𝑒 𝑚2 𝑥 are the
solution of the differential equation.
𝑑2 𝑦 𝑑𝑦
2. Since 𝑎 𝑑𝑥 2 + 𝑏 𝑑𝑥 + 𝑐𝑦 = 0 is a linear homogeneous equation, by using the linear independency
and linearity principle, the general solution must be
(i) 𝑦(𝑥) = 𝑐1 𝑒 𝑚1 𝑥 + 𝑐2 𝑒 𝑚2 𝑥 for real and distinct root,
𝑚1 ≠ 𝑚2 .
(ii) 𝑦(𝑥) = 𝑐1 𝑒 𝑚1 𝑥 + 𝑐2 𝑒 𝑚2 𝑥 for a pair of complex conjugate roots, 𝑚1 = 𝑚 + 𝑖𝛽 & 𝑚2 = 𝑚 −
𝑖𝛽.
(iii) However, if there is repeated real roots, 𝑚 = 𝑚1 = 𝑚2 , we get linearly dependent solution
𝑦(𝑥) = 𝑐1 𝑒 𝑚𝑥 + 𝑐2 𝑒 𝑚𝑥 as proven earlier. In this case, 𝑦(𝑥) = 𝑥𝑒 𝑚𝑥 is proven as one of the
solution of 2nd order ODE and it is linearly independent with 𝑒 𝑚𝑥 . Thus a complete solution
𝑦(𝑥) = 𝑐1 𝑒 𝑚1 𝑥 + 𝑐2 𝑥𝑒 𝑚2 𝑥 is obtained which satisfy the linearly independency property.

12.4.2 HOMOGENEOUS LINEAR DIFFERENTIAL EQUATION WITH NON-CONSTANT

COEFFICIENTS 𝑥 2 , 𝑎𝑥 (KNOWN AS EULER-CAUCHY DIFFERENTIAL EQUATION)

In previous case, we discussed about homogeneous linear differential equation with constant coefficient, i.e.
𝑑2 𝑦 𝑑𝑦
𝑎 𝑑𝑥 2 + 𝑏 𝑑𝑥 + 𝑐𝑦 = 0 and let the solution to be 𝑦(𝑥) = 𝑒 𝑚𝑥 𝑜𝑟 𝑦(𝑥) = 𝑥𝑒 𝑚𝑥 depending on the roots of
characteristic equation.

𝑑2 𝑦
However this methods is inefficient to solve the Euler-Cauchy differential equation, i.e. 𝑥 2 +
𝑑𝑥 2
𝑑𝑦
𝑎𝑥 𝑑𝑥 + 𝑏𝑦 = 0 where the coefficients are not constant. The strategy to solve this type of differential
equation is to convert the non-constant coefficient into constant form. This can be achieved by
substitution (let 𝑥 = 𝑒 𝑡 ).

Two important properties used to convert non-constant coefficient to constant coefficient:

𝑑2 𝑦 𝑑2𝑦 𝑑𝑦
(i) 𝑥 2 𝑑𝑥 2 = 𝑑𝑡 2
− 𝑑𝑡
𝑑𝑦 𝑑𝑦
(ii) 𝑥 =
𝑑𝑥 𝑑𝑡

121
 The detail description and proof is provided in the table below.

𝒅𝒚 𝒅𝒚
(i) Convert non-constant coefficient (𝒙 ) to constant coefficient ( )
𝒅𝒙 𝒅𝒕

𝑥 = 𝑒𝑡
>> 𝑙𝑛|𝑥| = 𝑡
1 𝑑𝑡
>> =
𝑥 𝑑𝑥
1 𝑑𝑡
>> 𝑥 = 𝑑𝑥

𝑑𝑦 𝑑𝑦 𝑑𝑡
Using chain rule, 𝑑𝑥
= 𝑑𝑡 𝑑𝑥
𝑑𝑦 𝑑𝑦 1
>> = ( )
𝑑𝑥 𝑑𝑡 𝑥
𝑑𝑦 𝑑𝑦
>> 𝑥 𝑑𝑥 = 𝑑𝑡

𝒅𝟐 𝒚 𝒅𝟐 𝒚 𝒅𝒚
(ii) Convert non-constant coefficient (𝒙𝟐 𝟐 ) to constant coefficient ( 𝟐 − )
𝒅𝒙 𝒅𝒕 𝒅𝒕

𝑑𝑦 𝑑𝑦
𝑥 𝑑𝑥 = 𝑑𝑡
𝑑 𝑑𝑦 𝑑 𝑑𝑦
>> 𝑑𝑥 (𝑥 𝑑𝑥 ) = 𝑑𝑥 ( 𝑑𝑡 )
𝑑2 𝑦 𝑑𝑦 𝑑 𝑑𝑦
>> 𝑥 𝑑𝑥 2 + 𝑑𝑥 = 𝑑𝑥 ( 𝑑𝑡 )

𝑑 𝑑𝑦 𝑑 𝑑𝑦 𝑑𝑡 1 𝑑𝑡
Using chain rule, 𝑑𝑥 ( 𝑑𝑡 ) = 𝑑𝑡 ( 𝑑𝑡 ) (𝑑𝑥) where 𝑥 = 𝑑𝑥
𝑑 𝑑𝑦 𝑑2 𝑦 1
>> 𝑑𝑥 ( 𝑑𝑡 ) = ( 𝑑𝑡 2 ) (𝑥)

𝑑2 𝑦 𝑑𝑦 𝑑2 𝑦 1
Combining the equations, we get 𝑥 𝑑𝑥 2 + 𝑑𝑥 = ( 𝑑𝑡 2 ) (𝑥)
𝑑2 𝑦 𝑑𝑦 𝑑2 𝑦 𝑑𝑦 𝑑𝑦
>> 𝑥 2 𝑑𝑥 2 + 𝑥 𝑑𝑥 = 𝑑𝑡 2
where 𝑥 𝑑𝑥 = 𝑑𝑡
𝑑2 𝑦 𝑑2 𝑦 𝑑𝑦
>> 𝑥 2 𝑑𝑥 2 = 𝑑𝑡 2
− 𝑑𝑡

122
 𝑑2 𝑦 𝑑𝑦
For example: Solve 2𝑥 2 𝑑𝑥 2 − 3𝑥 𝑑𝑥 − 3𝑦 = 0.

Solution: Let 𝑥 = 𝑒 𝑡 , then we get

𝑑2 𝑦 𝑑2𝑦 𝑑𝑦
(i) 𝑥2 2 = 2 −
𝑑𝑥 𝑑𝑡 𝑑𝑡
𝑑𝑦 𝑑𝑦
(ii) 𝑥 =
𝑑𝑥 𝑑𝑡

𝑑2𝑦 𝑑𝑦 𝑑2 𝑦 𝑑𝑦
2𝑥 2 𝑑𝑥 2 − 3𝑥 𝑑𝑥 − 3𝑦 = 0 [Euler-Cauchy differential equation, 𝑥 2 𝑑𝑥 2 + 𝑎𝑥 𝑑𝑥 + 𝑏𝑦 = 0]

𝑑2 𝑦 𝑑𝑦 𝑑𝑦
>> 2 ( 𝑑𝑡 2 − 𝑑𝑡 ) − 3 ( 𝑑𝑡 ) − 3𝑦 = 0

𝑑2 𝑦 𝑑𝑦
>> 2 ( 𝑑𝑡 2 ) − 5 ( 𝑑𝑡 ) − 3𝑦 = 0 [2nd order linear homogeneous DE with Constant coefficient]

>> 2(𝑚2 ) − 5(𝑚) − 3 = 0 [Characteristic equation]

where

𝑏 2 − 4𝑎𝑐 = (−5)2 − 4(2)(−3) = 49 > 0, thus it is Case (a) 𝑚1 ≠ 𝑚2

>> (2𝑚 + 1)(𝑚 − 3) = 0

>>𝑚1 = −0.5, 𝑚2 = 3 [Real and distinct roots case]

The complete solution: 𝑦(𝑡) = 𝑐1 𝑒 𝑚1 𝑡 + 𝑐2 𝑒 𝑚2 𝑡

>> 𝑦(𝑡) = 𝑐1 𝑒 −0.5𝑡 + 𝑐2 𝑒 3𝑡

𝑑2 𝑦 𝑑𝑦
Back substitution, we get the complementary solution to the 2𝑥 2 𝑑𝑥 2 − 3𝑥 𝑑𝑥 − 3𝑦 = 0.

>> 𝑥 = 𝑒 𝑡

>> 𝑦(𝑥) = 𝑐1 𝑥 −0.5 + 𝑐2 𝑥 3 where 𝑐1 & 𝑐2 = 𝑎𝑟𝑏𝑖𝑡𝑟𝑎𝑟𝑦 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡𝑠.

123
12.4.3 NONHOMOGENEOUS LINEAR DIFFERENTIAL EQUATION WITH CONSTANT
COEFFICIENTS 𝑎, 𝑏, 𝑐 IN THE FORM OF 𝑟(𝑥) = 𝑒 𝛼𝑥 𝑃𝑛 (𝑥)

So far we have discussed two strategies to solve homogeneous problem, now we will continue with the
𝑑2 𝑦 𝑑𝑦
nonhomogeneous problem, i.e. 𝑎 𝑑𝑥 2 + 𝑏 𝑑𝑥 + 𝑐𝑦 = 𝑟(𝑥).

If the RHS components, 𝑟(𝑥) are in the form of exponential, polynomial, sine and cosine functions,
we can implement the method of undetermined coefficient by let the RHS components to be equal to
𝑒 𝛼𝑥 𝑃𝑛 (𝑥) as following:

𝑑2 𝑦 𝑑𝑦
𝑎 𝑑𝑥 2 + 𝑏 𝑑𝑥 + 𝑐𝑦 = 𝑒 𝛼𝑥 𝑃𝑛 (𝑥)

where 𝑃𝑛 (𝑥) 𝑖𝑠 𝑡ℎ𝑒 𝑝𝑜𝑙𝑦𝑛𝑜𝑚𝑖𝑎𝑙 𝑓𝑢𝑛𝑐𝑡𝑖𝑜𝑛 𝑤𝑖𝑡ℎ 𝑑𝑒𝑔𝑟𝑒𝑒 𝑛

Hence, we can propose the possible particular solution of 𝑦𝑝 = 𝑒 𝛼𝑥 𝑄𝑛 (𝑥). The general procedure to
solve the 2nd order nonhomogeneous linear differential equation using the method of undetermined
coefficient is given below.

124
 General procedure for the method of undetermined coefficient:

2nd order nonhomogeneous linear differential equation:

𝑑2 𝑦 𝑑𝑦
𝑎 𝑑𝑥 2 + 𝑏 𝑑𝑥 + 𝑐𝑦 = 𝑟(𝑥)

Step 1: Solve the homogenous part first

𝑑2 𝑦 𝑑𝑦
𝑎 𝑑𝑥 2 + 𝑏 𝑑𝑥 + 𝑐𝑦 = 0
Complementary solution:
𝑦𝑐 (𝑥) = 𝑐1 𝑒 𝑚1 𝑥 + 𝑐2 𝑒 𝑚2 𝑥 or 𝑦𝑐 (𝑥) = 𝑐1 𝑒 𝑚1 𝑥 + 𝑐2 𝑥𝑒 𝑚2 𝑥

Step 2: Solve the nonhomogeneous part next

𝑑2 𝑦 𝑑𝑦
𝑎 +𝑏 + 𝑐𝑦 = 𝑒 𝛼𝑥 𝑃𝑛 (𝑥)
𝑑𝑥 2 𝑑𝑥

Possible particular solution:

𝑦𝑝 = 𝑒 𝛼𝑥 𝑄𝑛 (𝑥)
where 𝑄𝑛 (𝑥) = 𝑔𝑒𝑛𝑒𝑟𝑎𝑙 𝑝𝑜𝑙𝑦𝑛𝑜𝑚𝑖𝑎𝑙 𝑤𝑖𝑡ℎ 𝑠𝑎𝑚𝑒 𝑑𝑒𝑔𝑟𝑒𝑒 𝑜𝑓 𝑛 𝑤𝑖𝑡ℎ 𝑃𝑛 (𝑥) ,
e.g. 𝑃2 (𝑥) = 5𝑥 2 𝑤ℎ𝑒𝑟𝑒 𝑛 = 2, 𝑡ℎ𝑒𝑛 𝑄2 = 𝐴𝑥 2 + 𝐵𝑥 + 𝐶

Step 3: If 𝑦𝑝 & 𝑦𝑐 are linearly dependent,

give treatment/cure to 𝑦𝑝 to obtain linearly independent solution.
Proposed particular solution after cure:
𝑦𝑝 = 𝑥𝑒 𝛼𝑥 𝑄𝑛 (𝑥) or 𝑦𝑝 = 𝑥 2 𝑒 𝛼𝑥 𝑄𝑛 (𝑥)

Step 4: Solve the undetermined coefficient of the particular solution

by comparing the coefficient on both sides of the equation.

Step 5: The total solution for the 2nd order nonhomogeneous linear differential equation:
𝑦𝑡𝑜𝑡𝑎𝑙 = 𝑦𝑐 + 𝑦𝑝
Hint: Always solve the complementary solution first before proposing the particular solution.

The detail description for the “Step 1: Solve the homogenous part first” can be found in the previous section.
Now, we will discuss on the “Step 2: Solve the nonhomogenous part next”.

The method of undetermined coefficient is only applicable to 2nd order nonhomogeneous linear
differential equation, where the RHS components, 𝑟(𝑥) are restricted for exponential, polynomial, sine and
cosine functions, i.e. 𝑒 𝛼𝑥 𝑃𝑛 (𝑥).
125
 The exponential function is related directly to the 𝑒 𝛼𝑥 and polynomial function is related directly to
the 𝑃𝑛 (𝑥). Moreover, the exponential function, 𝑒 𝛼𝑥 is related indirectly to sine & cosine functions through
Euler’s Formula: 𝑒 ±𝑖𝑥 = 𝑐𝑜𝑠𝑥 ± 𝑖(𝑠𝑖𝑛𝑥).

For example:

𝑒 −𝑖(10𝑥) = cos(10𝑥) − 𝑖𝑠𝑖𝑛(10𝑥).

Thus, imaginary part of 𝑒 −𝑖(10𝑥) , i.e. 𝐼𝑚(𝑒 −𝑖(10𝑥) ) = −sin(10𝑥)

Real part of 𝑒 −𝑖(10𝑥) , i.e. 𝑅𝑒(𝑒 −𝑖(10𝑥) ) = cos(10𝑥)

Depends on the RHS function, the possible particular solution is proposed for the 2nd order
nonhomogeneous linear differential equation as shown in table below.

RHS function The form of Possible Particular Comment

𝑟(𝑥) = 𝑒 𝛼𝑥 𝑃𝑛 (𝑥) Solution
𝑦𝑝 = 𝑒 𝛼𝑥 𝑄𝑛 (𝑥)
(i) Pure Exponential 𝑟(𝑥) = 𝑒 −3𝑥 𝑃0 (𝑥) 𝑦𝑝 = 𝑒 −3𝑥 𝑄𝑂 Sine, Cosine,
Function, e.g. = 𝐴𝑒 −3𝑥 Exponential are
𝑟(𝑥) = 𝑒 −3𝑥 where related to each other
𝛼 = −3 where 𝑄𝑂 = in euler formula:
& general polynomial 𝑒 ±𝑖𝑥 = 𝑐𝑜𝑠𝑥 ±
𝑃𝑛 (𝑥) = 1 with with degree 𝑛 = 0 𝑖(𝑠𝑖𝑛𝑥) , thus the
degree 𝑛 = 0 mixture of them can
be represented by the
(ii) Pure Sine Function, 𝑟(𝑥) = 𝐼𝑚(𝑒 (5𝑖)𝑥 )𝑃0 (𝑥) 𝑦𝑝 = 𝑒 (5𝑖)𝑥
𝑄𝑂 exponential function.
e.g. = 𝐴𝑒 5𝑖𝑥 Hence, the method of
𝑟(𝑥) = 𝑠𝑖𝑛5𝑥 where undertermined
𝛼 = 5𝑖 coefficient can be
Given & 𝑦𝑝,𝑎𝑐𝑡𝑢𝑎𝑙 = 𝐼𝑚(𝑦𝑝 ) implemented to solve
𝑃𝑛 (𝑥) = 1 with the problem.
𝑒 𝑖(5𝑥) = 𝑐𝑜𝑠5𝑥 + 𝑖(𝑠𝑖𝑛5𝑥)
degree 𝑛 = 0
𝐼𝑚(𝑒 𝑖(5𝑥) ) = 𝑠𝑖𝑛5𝑥 For your extra
information, there are
other alternative to
solve the possible
(iii) Pure Cosine 𝑟(𝑥) = 𝑅𝑒(𝑒 (6𝑖)𝑥 )𝑃0 (𝑥) 𝑦𝑝 = 𝑒 (6𝑖)𝑥 𝑄𝑂 particular solution for
Function, e.g. = 𝐴𝑒 6𝑖𝑥 the RHS sine and
𝑟(𝑥) = 𝑐𝑜𝑠6𝑥 where cosine functions by let
𝛼 = 6𝑖 𝑦𝑝 = 𝐶𝑐𝑜𝑠𝑥 + 𝐷𝑠𝑖𝑛𝑥
& 𝑦𝑝,𝑎𝑐𝑡𝑢𝑎𝑙 = 𝑅𝑒(𝑦𝑝 )
Given
𝑃𝑛 (𝑥) = 1 with For example:
𝑒 𝑖(6𝑥) = 𝑐𝑜𝑠6𝑥 + 𝑖(𝑠𝑖𝑛6𝑥) (i) 𝑟(𝑥) = 𝑠𝑖𝑛5𝑥
degree 𝑛 = 0
𝑦𝑝
𝑅𝑒(𝑒 𝑖(6𝑥) ) = 𝑐𝑜𝑠6𝑥
= 𝐶𝑐𝑜𝑠5𝑥 + 𝐷𝑠𝑖𝑛5𝑥

126
 (iv) Mixture of 𝑟(𝑥) 𝑦𝑝 = 𝑒 (6𝑖−3)𝑥 𝑄𝑂
(6𝑖)𝑥
Exponential & = 𝑅𝑒(𝑒 )(𝑒 (−3𝑥) )𝑃0 (𝑥) = 𝐴𝑒 (6𝑖−3)𝑥 (ii) 𝑟(𝑥) = 𝑐𝑜𝑠6𝑥
Cosine Function, = 𝑅𝑒(𝑒 (6𝑖−3)𝑥
)𝑃0 (𝑥) 𝑦𝑝
e.g. = 𝐶𝑐𝑜𝑠6𝑥 + 𝐷𝑠𝑖𝑛6𝑥
𝑟(𝑥) = 𝑒 −3𝑥 𝑐𝑜𝑠6𝑥 where 𝑦𝑝,𝑎𝑐𝑡𝑢𝑎𝑙 = 𝑅𝑒(𝑦𝑝 )
𝛼 = 6𝑖 − 3 (iii) 𝑟(𝑥) = 𝑒 −3𝑥 𝑐𝑜𝑠6𝑥
& 𝑦𝑝
Given 𝑃𝑛 (𝑥) = 1 with = 𝑒 −3𝑥 (𝐶𝑐𝑜𝑠6𝑥
+ 𝐷𝑠𝑖𝑛6𝑥)
𝑒 𝑖(6𝑥) = 𝑐𝑜𝑠6𝑥 + 𝑖(𝑠𝑖𝑛6𝑥) degree 𝑛 = 0
𝑅𝑒(𝑒 𝑖(6𝑥) ) = 𝑐𝑜𝑠6𝑥

(v) Pure Polynomial 𝑦𝑝 = 𝑒 (0)𝑥 𝑄3

𝑟(𝑥) = 𝑒 (0𝑥) 𝑃3 (𝑥) Nil
Function, e.g. = 𝐴𝑥 3 + 𝐵𝑥 2
𝑟(𝑥) = 6𝑥 3 + 4𝑥 2 + 5 +𝐶𝑥 + 𝐷
Where
𝛼=0
&
𝑃3 (𝑥) = 6𝑥 3 + 4𝑥 2 + 5 is
the polynomial function of
degree 𝑛 = 3

(vi) Mixture of
𝑟(𝑥) = 𝑒 −3𝑥 𝑃1 (𝑥) 𝑦𝑝 = 𝑒 (−3)𝑥 𝑄1 Nil
Polynomial &
Exponential = 𝑒 (−3)𝑥 (𝐴𝑥 + 𝐵)
Function in where
multiplication, eg.
𝛼 = −3
𝑟(𝑥) = 6𝑥𝑒 −3𝑥 &
𝑃1 (𝑥) = 6𝑥

(vii) Mixture of
For polynomial function, 𝑦𝑝,1 = 𝑒 (0)𝑥 𝑄3 Can be solved
Polynomial & = 𝐴𝑥 3 + 𝐵𝑥 2 separately and then
Exponential 𝑟(𝑥) = 𝑒 (0𝑥) 𝑃3 (𝑥) +𝐶𝑥 + 𝐷 combine the result
Function in ‘+’ where 𝛼 = 0 & 𝑃3 (𝑥) = −3𝑥
later.
6𝑥 3 + 4𝑥 2 + 5 is the 𝑦𝑝,2 = 𝑒 𝑄𝑂
𝑟(𝑥) = 𝑒 −3𝑥 + 6𝑥 3 polynomial function of = 𝐸𝑒 −3𝑥
+4𝑥 2 + 5 degree 𝑛 = 3
For exponential function, 𝑦𝑝 = 𝑦𝑝,1 + 𝑦𝑝,2
𝑟(𝑥) = 𝑒 −3𝑥 𝑃0 (𝑥)
where 𝛼 = −3 & 𝑃𝑛 (𝑥) = 1
with degree 𝑛 = 0
Note: 𝑒 ±𝑖𝑥 = 𝑐𝑜𝑠𝑥 ± 𝑖(𝑠𝑖𝑛𝑥); 𝑄𝑛 (𝑥) & 𝑃𝑛 (𝑥) are two polynomial functions with same degree.
127
Now, we will discuss on the “Step 3: To check the linear dependency and give treatment to particular solution
if needed”. The possible particular solution is proposed according to the RHS function, however, further
treatment will be needed to obtain a linearly independent solution by comparing the complementary
solution. In fact, the actual particular solution 𝑦𝑝 can be separated into 3 cases depending on

(i) The possible particular solution, 𝑦𝑝 = 𝑒 𝛼𝑥 𝑄𝑛 (𝑥) and

(ii) The complementary solution that in the function of roots 𝑚1 & 𝑚2 , i.e.
𝑐1 𝑒 𝑚1 𝑥 + 𝑐2 𝑒 𝑚2 𝑥 , 𝑤ℎ𝑒𝑟𝑒 𝑚1 ≠ 𝑚2
𝑦𝑐 = {𝑐1 𝑒 (𝑚+𝑖β)𝑥 + 𝑐2 𝑒 (𝑚−𝑖β)𝑥 , 𝑤ℎ𝑒𝑟𝑒 𝑚1 ≠ 𝑚2 }
𝑐1 𝑥𝑒 𝑚𝑥 + 𝑐2 𝑒 𝑚𝑥 , 𝑤ℎ𝑒𝑟𝑒 𝑚1 = 𝑚2

The proposed particular solution is illustrated below.

Case 1 Case 2 Case 3

(𝛼 ≠ 𝑚1 & 𝑚2 ) (𝛼 = 𝑚1 𝑜𝑟 𝑚2 , (𝛼 = 𝑚1 = 𝑚2 )
𝑚1 ≠ 𝑚2 )

Definition Coefficient 𝛼 is not equal to

Coefficient 𝛼 is equal to one Coefficient 𝛼 is equal
coefficients 𝑚1 & 𝑚2 of the coefficient 𝑒. 𝑔. 𝑚1 to both coefficients
and different with another 𝑚1 & 𝑚2
coefficient 𝑚2
Possible 𝑦𝑐 𝑦𝑐 𝑦𝑐
complementary 𝑐1 𝑒 𝑚1 𝑥 + 𝑐2 𝑒 𝑚2 𝑥 𝑐1 𝑒 𝑚1 𝑥 + 𝑐2 𝑒 𝑚2 𝑥 = 𝑐1 𝑥𝑒 𝑚𝑥 + 𝑐2 𝑒 𝑚𝑥
= { }
solution = {𝑐1 𝑒 (𝑚+𝑖β)𝑥 + 𝑐2 𝑒 (𝑚−𝑖β)𝑥 } 𝑐1 𝑒 (𝑚+𝑖β)𝑥 + 𝑐2 𝑒 (𝑚−𝑖β)𝑥
𝑚𝑥 𝑚𝑥
for homogeneous 𝑐1 𝑥𝑒 + 𝑐2 𝑒
ODE
Proposed 𝑦𝑝 = 𝑒 𝛼𝑥 𝑄𝑛 (𝑥) 𝑦𝑝 = 𝑥𝑒 𝛼𝑥 𝑄𝑛 (𝑥) 𝑦𝑝 = 𝑥 2 𝑒 𝛼𝑥 𝑄𝑛 (𝑥)
particular
solution
for non-
homogeneous
ODE
Observation If 𝛼 ≠ 𝑚1 & 𝑚2 , If 𝛼 = 𝑚1 𝑜𝑟 𝑚2 , If 𝛼 = 𝑚1 = 𝑚2 , we
then we have no concern to 𝑚1 ≠ 𝑚2 , can’t use
apply 𝑦𝑝 = 𝑒 𝛼𝑥 𝑄𝑛 (𝑥) we can’t use 𝑦𝑝 = 𝑒 𝛼𝑥 𝑄𝑛 (𝑥) 𝑦𝑝 = 𝑒 𝛼𝑥 𝑄𝑛 (𝑥)
because because or 𝑦𝑝 = 𝑥𝑒 𝛼𝑥 𝑄𝑛 (𝑥)
𝑦𝑝 has various forms as 𝑦𝑐 this form has the similar because these forms
(i.e. 𝑦𝑝 & 𝑦𝑐 are linearly form as 𝑦 𝑐 and cause zero have the similar form
independent) RHS function. as 𝑦𝑐 and cause zero
RHS function.
Hint: Multiply the independent variable, 𝑥 or 𝑥 2 to the particular solution if you found the complementary
solution has the similar exponential function as the proposed particular solution. This is known as the cure/
treatment to the particular solution.

Step 4 & 5 are quite straight forward, the solution of nonhomogeneous differential equation consists of
complementary solution and particular solution (i.e. 𝑦𝑡𝑜𝑡𝑎𝑙 = 𝑦𝑐 + 𝑦𝑝 ). Example of solving the 2nd order
nonhomogeneous linear using the method of undetermined coefficient is given below.
128
 𝑑2 𝑦 𝑑𝑦
For example: Solve 𝑑𝑥 2
− 3 𝑑𝑥 + 2𝑦 = 𝑒 𝑥 [RHS - Pure Exponential Function]

Step 1: Homogeneous Part Step 2: Nonhomogeneous Part

𝑑2 𝑦 𝑑𝑦 𝑑2 𝑦 𝑑𝑦
i.e. 𝑑𝑥 2
− 3 𝑑𝑥 + 2𝑦 = 0 i.e. 𝑑𝑥 2
− 3 𝑑𝑥 + 2𝑦 = 𝑒 𝑥

Characteristic equation: The method of undetermined coefficient:

𝑚2 − 3𝑚 + 2 = 0 RHS: 𝑟(𝑥) = 𝑒 𝛼𝑥 𝑃𝑛 (𝑥)
(𝑚 − 2)(𝑚 − 1) = 0 where 𝛼 = 1, 𝑛 = 0
𝑚1 = 2 & 𝑚2 = 1 Possible particular solution:
𝑦𝑝 = 𝑒 𝑥 𝑄0 (𝑥) = 𝐴𝑒 𝑥
Comment: Real & distinct roots Since 𝛼 ≠ 𝑚1 and 𝛼 = 𝑚2 , treatment is necessary:
𝑦𝑝 = 𝐴𝑥𝑒 𝑥
Complementary solution: Comment:
2𝑥 𝑥
𝑦𝑐 = 𝑐1 𝑒 + 𝑐2 𝑒 (i) 𝑦𝑝 = 𝐴𝑒 𝑥 & 𝑦𝑐 = 𝑐1 𝑒 2𝑥 + 𝑐2 𝑒 𝑥 are linearly dependent.
(ii) 𝑦𝑝 = 𝐴𝑥𝑒 𝑥 & 𝑦𝑐 = 𝑐1 𝑒 2𝑥 + 𝑐2 𝑒 𝑥 are linearly independent.

Solve the coefficient for the particular solution:

𝑦𝑝 = 𝐴𝑥𝑒 𝑥
𝑑𝑦𝑝
Differentiate it, we get: 𝑑𝑥
= 𝐴𝑥𝑒 𝑥 + 𝐴𝑒 𝑥
𝑑 2 𝑦𝑝
𝑑𝑥 2
= 𝐴𝑥𝑒 𝑥 + 2𝐴𝑒 𝑥
𝑑2 𝑦 𝑑𝑦
Substitute to the ODE equation: −3 + 2𝑦 = 𝑒 𝑥
𝑑𝑥 2 𝑑𝑥

>> (𝐴𝑥𝑒 𝑥 + 2𝐴𝑒 𝑥 ) − 3(𝐴𝑥𝑒 𝑥 + 2𝐴𝑒 𝑥 ) + 2(𝐴𝑥𝑒 𝑥 ) = 𝑒 𝑥

>> −𝐴𝑒 𝑥 = 𝑒 𝑥
Comparing the coefficients,
>> 𝑒 𝑥 : 𝐴 = −1
The actual particular solution:
𝑦𝑝 = −𝑥𝑒 𝑥
𝑑2 𝑦 𝑑𝑦
The complete/ general solution to −3 + 2𝑦 = 𝑒 𝑥 is
𝑑𝑥 2 𝑑𝑥

𝑦𝑡𝑜𝑡𝑎𝑙 = 𝑦𝑐 + 𝑦𝑝 = 𝑐1 𝑒 2𝑥 + 𝑐2 𝑒 𝑥 − 𝑥𝑒 𝑥

129
 𝑑2 𝑦 𝑑𝑦
For example: Solve 𝑑𝑥 2 − 5 𝑑𝑥 + 6𝑦 = 4𝑠𝑖𝑛2𝑥 [RHS - Pure Sine Function]
Step 1: Homogeneous Part Step 2: Nonhomogeneous Part
𝑑2 𝑦 𝑑𝑦 𝑑2 𝑦 𝑑𝑦
i.e. 𝑑𝑥 2
− 5 𝑑𝑥 + 6𝑦 = 0 i.e. 𝑑𝑥 2 − 5 𝑑𝑥 + 6𝑦 = 4𝑠𝑖𝑛2𝑥

Characteristic equation: The method of undetermined coefficient:

𝑚2 − 5𝑚 + 6 = 0
(𝑚 − 2)(𝑚 − 3) = 0
𝑚1 = 2 & 𝑚2 = 3
Comment: Real & distinct roots
Complementary solution:
𝑦𝑐 = 𝑐1 𝑒 2𝑥 + 𝑐2 𝑒 3𝑥
RHS: 𝑟(𝑥) = 4𝑠𝑖𝑛2𝑥
From Euler’s formula: 𝑒 (2i𝑥) = cos(2𝑥) + 𝑖𝑠𝑖𝑛(2𝑥)
Thus, 𝐼𝑚[𝑒 (2i𝑥) ] = 𝑠𝑖𝑛(2𝑥)
RHS: 𝑟(𝑥) = 𝑒 𝛼𝑥 𝑃𝑛 (𝑥) = 4𝐼𝑚[𝑒 (2i𝑥) ]
where 𝛼 = 2𝑖, 𝑛 = 0
Possible particular solution:
𝑦𝑝 = 𝑒 2𝑖𝑥 𝑄0 (𝑥) = 𝐴𝑒 2𝑖𝑥
Since 𝛼 ≠ 𝑚1 and 𝑚2 , treatment is not needed.
𝑦𝑝 = 𝐴𝑒 2𝑖𝑥
𝑦𝑝,𝑎𝑐𝑡𝑢𝑎𝑙 = 𝐼𝑚(𝑦𝑝 )
Comment:
(i) 𝑦𝑝 = 𝐴𝑒 2𝑖𝑥 & 𝑦𝑐 = 𝑐1 𝑒 2𝑥 + 𝑐2 𝑒 3𝑥 are linearly independent.
(ii) Common practice use the 𝑦𝑝 in the calculation instead of 𝐼𝑚(𝑦𝑝 )
for the ease of calculation. Once the 𝑦𝑝 is solved, then we can
determine the actual 𝑦𝑝 using 𝑦𝑝,𝑎𝑐𝑡𝑢𝑎𝑙 = 𝐼𝑚(𝑦𝑝 ).

Solve the coefficient for the particular solution:

𝑦𝑝 = 𝐴𝑒 2𝑖𝑥
𝑑𝑦𝑝
Differentiate it, we get: 𝑑𝑥
= 2𝑖𝐴𝑒 2𝑖𝑥
𝑑 2 𝑦𝑝
𝑑𝑥 2
= −4𝐴𝑒 2𝑖𝑥
𝑑2 𝑦 𝑑𝑦
Substitute to the ODE equation: −5 + 6𝑦 = 4𝑠𝑖𝑛2𝑥
𝑑𝑥 2 𝑑𝑥

>> (−4𝐴𝑒 2𝑖𝑥 ) − 5(2𝑖𝐴𝑒 2𝑖𝑥 ) + 6(𝐴𝑒 2𝑖𝑥 ) = 4𝑒 (2i𝑥)

>> 𝑖(−10𝐴𝑒 2𝑖𝑥 ) + 2(𝐴𝑒 2𝑖𝑥 ) = 4𝑒 (2i𝑥)
Comparing the coefficients,
>> 𝑒 2𝑖𝑥 : 2𝐴 − 10𝐴𝑖 = 4
4 2
>> 𝐴 = 2−10𝑖 = 1−5𝑖

130
 The particular solution:
2
𝑦𝑝 = 1−5𝑖 𝑒 2𝑖𝑥
2 1+5𝑖
>> 𝑦𝑝 = 1−5𝑖 (1+5𝑖) (𝑐𝑜𝑠2𝑥 + 𝑖𝑠𝑖𝑛2𝑥)
2(1+5𝑖)
>> 𝑦𝑝 = (𝑐𝑜𝑠2𝑥 + 𝑖𝑠𝑖𝑛2𝑥)
26
(1+5𝑖)
>> 𝑦𝑝 = (𝑐𝑜𝑠2𝑥 + 𝑖𝑠𝑖𝑛2𝑥)
13
(𝑐𝑜𝑠2𝑥−5𝑠𝑖𝑛2𝑥)+𝑖(5𝑐𝑜𝑠2𝑥+𝑠𝑖𝑛2𝑥)
>> 𝑦𝑝 = 13

The actual particular solution:

𝑦𝑝,𝑎𝑐𝑡𝑢𝑎𝑙 = 𝐼𝑚(𝑦𝑝 )
(𝑐𝑜𝑠2𝑥−5𝑠𝑖𝑛2𝑥)+𝑖(5𝑐𝑜𝑠2𝑥+𝑠𝑖𝑛2𝑥) (5𝑐𝑜𝑠2𝑥+𝑠𝑖𝑛2𝑥)
>> 𝑦𝑝,𝑎𝑐𝑡𝑢𝑎𝑙 = 𝐼𝑚 ( 13
)= 13

𝑑2 𝑦 𝑑𝑦
The complete/ general solution to −5 + 6𝑦 = 4𝑠𝑖𝑛2𝑥 is
𝑑𝑥 2 𝑑𝑥
(5𝑐𝑜𝑠2𝑥+𝑠𝑖𝑛2𝑥)
𝑦𝑡𝑜𝑡𝑎𝑙 = 𝑦𝑐 + 𝑦𝑝 = 𝑐1 𝑒 2𝑥 + 𝑐2 𝑒 3𝑥 + 13

Note: Similar procedure as the case of RHS - Pure Cosine Function

131
 𝑑2 𝑦 𝑑𝑦
For example: Solve 𝑑𝑥 2 + 4 𝑑𝑥 + 5𝑦 = 𝑒 −2𝑥 𝑐𝑜𝑠𝑥 [RHS - Mixture of Exponential & Cosine Function]

Step 1: Homogeneous Part Step 2: Nonhomogeneous Part

𝑑2 𝑦 𝑑𝑦 𝑑2 𝑦 𝑑𝑦
i.e. 𝑑𝑥 2 + 4 𝑑𝑥 + 5𝑦 = 0 𝑑𝑥 2
+ 4 𝑑𝑥 + 5𝑦 = 𝑒 −2𝑥 𝑐𝑜𝑠𝑥

Characteristic equation: The method of undetermined coefficient:

𝑚2 + 4𝑚 + 5 = 0 RHS: 𝑟(𝑥) = 𝑒 −2𝑥 𝑐𝑜𝑠𝑥
(𝑚 + 2)2 − 4 + 5 = 0 From Euler’s formula: 𝑒 (i𝑥) = cos(𝑥) + 𝑖𝑠𝑖𝑛(𝑥)
𝑚 = −2 ± √−1 Thus, 𝑅𝑒[𝑒 (i𝑥) ] = 𝑐𝑜𝑠(𝑥)
𝑚1 = −2 + i & 𝑚2 = −2 − i RHS: 𝑟(𝑥) = 𝑒 𝛼𝑥 𝑃𝑛 (𝑥) = 𝑒 −2𝑥 𝑅𝑒[𝑒 (i𝑥) ] = 𝑅𝑒[𝑒 (−2+i)𝑥 ]
where 𝛼 = −2 + 𝑖, 𝑛 = 0
Comment: A pair of complex Possible particular solution:
conjugates roots
𝑦𝑝 = 𝑒 (−2+𝑖)𝑥 𝑄0 (𝑥) = 𝐴𝑒 (−2+𝑖)𝑥
Since 𝛼 = 𝑚1 and 𝛼 ≠ 𝑚2 , treatment is needed.
Complementary solution:
(−2+i)𝑥 (−2−i)𝑥
𝑦𝑝 = 𝐴𝑥𝑒 (−2+𝑖)𝑥
𝑦𝑐 = 𝑐1 𝑒 + 𝑐2 𝑒
𝑦𝑝,𝑎𝑐𝑡𝑢𝑎𝑙 = 𝑅𝑒(𝑦𝑝 )
Comment:
Comment:
(i) For your extra info., the (i) 𝑦𝑝 = 𝐴𝑒 (−2+𝑖)𝑥 and 𝑦𝑐 = 𝑐1 𝑒 (−2+i)𝑥 + 𝑐2 𝑒 (−2−i)𝑥 are linearly
complementary solution can dependent
be converted to 𝑦𝑐 = (ii) 𝑦𝑝 = 𝐴𝑥𝑒 (−2+𝑖)𝑥 and 𝑦𝑐 = 𝑐1 𝑒 (−2+i)𝑥 + 𝑐2 𝑒 (−2−i)𝑥 are linearly
𝑒 −2𝑥 (𝐴𝑐𝑜𝑠𝑥 + 𝐵𝑠𝑖𝑛𝑥) independent
(iii) Common practice use the 𝑦𝑝 in the calculation instead of 𝑅𝑒(𝑦𝑝 )
for the ease of calculation. Once the 𝑦𝑝 is solved, then we can
solve the actual 𝑦𝑝 using 𝑦𝑝,𝑎𝑐𝑡𝑢𝑎𝑙 = 𝑅𝑒(𝑦𝑝 ).

Solve the coefficient for the particular solution:

𝑦𝑝 = 𝐴𝑥𝑒 (−2+𝑖)𝑥
Differentiate it, we get:
𝑑𝑦𝑝
𝑑𝑥
= (−2 + 𝑖)𝐴𝑥𝑒 (−2+𝑖)𝑥 + 𝐴𝑒 (−2+𝑖)𝑥
𝑑 2 𝑦𝑝
= (3 − 4𝑖)𝐴𝑥𝑒 (−2+𝑖)𝑥 + (−4 + 2𝑖)𝐴𝑒 (−2+𝑖)𝑥
𝑑𝑥 2

𝑑2 𝑦 𝑑𝑦
Substitute to the ODE equation: 𝑑𝑥 2 + 4 𝑑𝑥 + 5𝑦 = 𝑒 −2𝑥 𝑐𝑜𝑠𝑥

>> ((3 − 4𝑖)𝐴𝑥𝑒 (−2+𝑖)𝑥 + (−4 + 2𝑖)𝐴𝑒 (−2+𝑖)𝑥 ) + 4 ((−2 +

𝑖)𝐴𝑥𝑒 (−2+𝑖)𝑥 + 𝐴𝑒 (−2+𝑖)𝑥 ) + 5(𝐴𝑥𝑒 (−2+𝑖)𝑥 ) = 𝑒 (−2+i)𝑥

>> 2𝑖𝐴𝑒 (−2+𝑖)𝑥 = 𝑒 (−2+i)𝑥

132
 Comparing the coefficients,
>> 𝑒 (−2+𝑖)𝑥 : 2𝑖𝐴 = 1
1
>> 𝐴 = 2𝑖
The particular solution:
1
𝑦𝑝 = 𝐴𝑥𝑒 (−2+𝑖)𝑥 = 2𝑖 𝑥𝑒 (−2+𝑖)𝑥
1 𝑖
>> 𝑦𝑝 = 2𝑖 𝑖 𝑥𝑒 (−2)𝑥 𝑒 (𝑖)𝑥
𝑖
>> 𝑦𝑝 = − 2 𝑥𝑒 −2𝑥 (𝑐𝑜𝑠𝑥 + 𝑖𝑠𝑖𝑛𝑥)
1
>> 𝑦𝑝 = − 2 𝑥𝑒 −2𝑥 (𝑖𝑐𝑜𝑠𝑥 − 𝑠𝑖𝑛𝑥)
The actual particular solution:
𝑦𝑝,𝑎𝑐𝑡𝑢𝑎𝑙 = 𝑅𝑒(𝑦𝑝 )
1 1
>> 𝑦𝑝,𝑎𝑐𝑡𝑢𝑎𝑙 = 𝑅𝑒 (− 2 𝑥𝑒 −2𝑥 (𝑖𝑐𝑜𝑠𝑥 − 𝑠𝑖𝑛𝑥)) = 2 𝑥𝑒 −2𝑥 (𝑠𝑖𝑛𝑥)

𝑑2 𝑦 𝑑𝑦
The complete/ general solution to 𝑑𝑥 2 + 4 𝑑𝑥 + 5𝑦 = 𝑒 −2𝑥 𝑐𝑜𝑠𝑥 is
1
𝑦𝑡𝑜𝑡𝑎𝑙 = 𝑦𝑐 + 𝑦𝑝 = 𝑐1 𝑒 (−2+𝑖)𝑥 + 𝑐2 𝑒 (−2−𝑖)𝑥 + 2 𝑥𝑒 −2𝑥 (𝑠𝑖𝑛𝑥)

There is another alternative to solve the 2nd order nonhomogeneous linear ODE problem with RHS sine and
cosine functions by using 𝑦𝑝 = 𝐶𝑐𝑜𝑠𝑥 + 𝐷𝑠𝑖𝑛𝑥 . Students are allowed to use any of them to solve the
problem.

𝑑2 𝑦 𝑑𝑦
2nd Alternative Method is used to solve the same example: Solve 𝑑𝑥 2 + 4 𝑑𝑥 + 5𝑦 = 𝑒 −2𝑥 𝑐𝑜𝑠𝑥

Step 1: Homogeneous Part Step 2: Nonhomogeneous Part

𝑑2 𝑦 𝑑𝑦 𝑑2 𝑦 𝑑𝑦
i.e. 𝑑𝑥 2 + 4 𝑑𝑥 + 5𝑦 = 0 𝑑𝑥 2
+ 4 𝑑𝑥 + 5𝑦 = 𝑒 −2𝑥 𝑐𝑜𝑠𝑥

Characteristic equation: The method of undetermined coefficient:

𝑚2 + 4𝑚 + 5 = 0 RHS: 𝑟(𝑥) = 𝑒 −2𝑥 𝑐𝑜𝑠𝑥
(𝑚 + 2)2 − 4 + 5 = 0 Possible particular solution:
𝑚 = −2 ± √−1 𝑦𝑝 = 𝑒 −2𝑥 (𝐶𝑐𝑜𝑠𝑥 + 𝐷𝑠𝑖𝑛𝑥)
𝑚1 = −2 + i & 𝑚2 = −2 − i Since 𝑦𝑐 = 𝑒 −2𝑥 (𝐴𝑐𝑜𝑠𝑥 + 𝐵𝑠𝑖𝑛𝑥) & 𝑦𝑝 = 𝑒 −2𝑥 (𝐶𝑐𝑜𝑠𝑥 + 𝐷𝑠𝑖𝑛𝑥)
are linearly dependent, treatment is needed.

Comment: A pair of complex Actual particular solution:

conjugates roots 𝑦𝑝 = 𝑥𝑒 −2𝑥 (𝐶𝑐𝑜𝑠𝑥 + 𝐷𝑠𝑖𝑛𝑥)

133
Complementary solution: Solve the coefficient for the particular solution:
𝑦𝑐 = 𝑐1 𝑒 (−2+i)𝑥 + 𝑐2 𝑒 (−2−i)𝑥 𝑦𝑝 = 𝑥𝑒 −2𝑥 (𝐶𝑐𝑜𝑠𝑥 + 𝐷𝑠𝑖𝑛𝑥)
Differentiate it, we get:
Using Euler equation:
𝑑𝑦𝑝
𝑦𝑐 = 𝑒 −2𝑥 (𝑐1 𝑒 𝑖𝑥 + 𝑐2 𝑒 −𝑖𝑥 ) = 𝑒 −2𝑥 (𝐶𝑐𝑜𝑠𝑥 + 𝐷𝑠𝑖𝑛𝑥) + 𝑥(−2𝑒 −2𝑥 )(𝐶𝑐𝑜𝑠𝑥 + 𝐷𝑠𝑖𝑛𝑥)
𝑑𝑥

= 𝑒 −2𝑥 (𝑐1 (𝑐𝑜𝑠𝑥 + 𝑖𝑠𝑖𝑛𝑥) +𝑥𝑒 −2𝑥 (−𝐶𝑠𝑖𝑛𝑥 + 𝐷𝑐𝑜𝑠𝑥)

+(𝑐2 (𝑐𝑜𝑠𝑥 − 𝑖𝑠𝑖𝑛𝑥)) = 𝑒 −2𝑥 (𝐶𝑐𝑜𝑠𝑥 + 𝐷𝑠𝑖𝑛𝑥)
= 𝑒 −2𝑥 ((𝑐1 + 𝑐2 )𝑐𝑜𝑠𝑥) +𝑥𝑒 −2𝑥 ((−2𝐶 + 𝐷)𝑐𝑜𝑠𝑥 + (−2𝐷 + 𝐶)𝑠𝑖𝑛𝑥)
+((𝑐1 𝑖 − 𝑐2 𝑖)𝑠𝑖𝑛𝑥)

Let 𝐴 = 𝑐1 + 𝑐2 & 𝐵 = 𝑐1 𝑖 + 𝑐2 𝑖 𝑑 2 𝑦𝑝
𝑑𝑥 2
= (−2𝑒 −2𝑥 )(𝐶𝑐𝑜𝑠𝑥 + 𝐷𝑠𝑖𝑛𝑥) + 𝑒 −2𝑥 (−𝐶𝑠𝑖𝑛𝑥 + 𝐷𝑐𝑜𝑠𝑥)
+𝑒 −2𝑥 ((−2𝐶 + 𝐷)𝑐𝑜𝑠𝑥 + (−2𝐷 + 𝐶)𝑠𝑖𝑛𝑥)
−2𝑥 (𝐴𝑐𝑜𝑠𝑥
𝑦𝑐 = 𝑒 + 𝐵𝑠𝑖𝑛𝑥) +𝑥(−2𝑒 −2𝑥 )((−2𝐶 + 𝐷)𝑐𝑜𝑠𝑥 + (−2𝐷 + 𝐶)𝑠𝑖𝑛𝑥)
+𝑥𝑒 −2𝑥 ((−2𝐶 + 𝐷)(−𝑠𝑖𝑛𝑥) + (−2𝐷 + 𝐶)𝑐𝑜𝑠𝑥)
= 𝑒 −2𝑥 ((−4𝐶 + 2𝐷)𝑐𝑜𝑠𝑥 + (−4𝐷 − 2𝐶)𝑠𝑖𝑛𝑥)
+𝑥𝑒 −2𝑥 ((3𝐶 − 4𝐷)𝑐𝑜𝑠𝑥 + (3𝐷 + 4𝐶)𝑠𝑖𝑛𝑥)

𝑑2 𝑦 𝑑𝑦
Substitute to the ODE equation: 𝑑𝑥 2 + 4 𝑑𝑥 + 5𝑦 = 𝑒 −2𝑥 𝑐𝑜𝑠𝑥

>> (𝑒 −2𝑥 ((−4𝐶 + 2𝐷)𝑐𝑜𝑠𝑥 + (−4𝐷 − 2𝐶)𝑠𝑖𝑛𝑥) + 𝑥𝑒 −2𝑥 ((3𝐶 −

4𝐷)𝑐𝑜𝑠𝑥 + (3𝐷 + 4𝐶)𝑠𝑖𝑛𝑥) ) + 4 (𝑒 −2𝑥 (𝐶𝑐𝑜𝑠𝑥 + 𝐷𝑠𝑖𝑛𝑥) +
𝑥𝑒 −2𝑥 ((−2𝐶 + 𝐷)𝑐𝑜𝑠𝑥 + (−2𝐷 + 𝐶)𝑠𝑖𝑛𝑥)) + 5(𝑥𝑒 −2𝑥 (𝐶𝑐𝑜𝑠𝑥 +
𝐷𝑠𝑖𝑛𝑥)) = 𝑒 −2𝑥 𝑐𝑜𝑠𝑥
>> 𝑒 −2𝑥 ((2𝐷)𝑐𝑜𝑠𝑥 + (−2𝐶)𝑠𝑖𝑛𝑥) = 𝑒 −2𝑥 𝑐𝑜𝑠𝑥
Comparing the coefficients,
>>𝑒 −2𝑥 𝑐𝑜𝑠𝑥: 2𝐷 = 1
1
𝐷=2

>> 𝑒 −2𝑥 𝑠𝑖𝑛𝑥: −2𝐶 = 0

𝐶=0
The particular solution:
1 1
𝑦𝑝 = 𝑥𝑒 −2𝑥 (0𝑐𝑜𝑠𝑥 + 𝑠𝑖𝑛𝑥) = 𝑥𝑒 −2𝑥 𝑠𝑖𝑛𝑥
2 2

𝑑2 𝑦 𝑑𝑦
The complete/ general solution to 𝑑𝑥 2 + 4 𝑑𝑥 + 5𝑦 = 𝑒 −2𝑥 𝑐𝑜𝑠𝑥 is
1
𝑦𝑡𝑜𝑡𝑎𝑙 = 𝑦𝑐 + 𝑦𝑝 = 𝑒 −2𝑥 (𝐴𝑐𝑜𝑠𝑥 + 𝐵𝑠𝑖𝑛𝑥) + 2 𝑥𝑒 −2𝑥 (𝑠𝑖𝑛𝑥)

[Comment: Same answer as previous example.]

134
 𝑑2 𝑦
For example: Solve 𝑑𝑥 2 + 4𝑦 = 8𝑥 2 [RHS - Pure Polynomial Function]

Step 1: Homogeneous Part Step 2: Nonhomogeneous Part

𝑑2 𝑦 𝑑2 𝑦
i.e. 𝑑𝑥 2 + 4𝑦 = 0 𝑑𝑥 2
+ 4𝑦 = 8𝑥 2

Characteristic equation: The method of undetermined coefficient:

𝑚2 + 4 = 0 RHS: 𝑟(𝑥) = 𝑒 𝛼𝑥 𝑃𝑛 (𝑥) = 8𝑥 2
𝑚 = ±√−4 where 𝛼 = 0, 𝑛 = 2
𝑚1 = 2i & 𝑚2 = −2i Possible particular solution:
𝑦𝑝 = 𝑒 (0)𝑥 𝑄2 (𝑥) = 𝐴2 𝑥 2 + 𝐴1 𝑥 + 𝐴0
Comment: A pair of complex conjugates Since 𝛼 ≠ 𝑚1 & 𝑚2 , treatment is not needed.
roots

Comment:
Complementary solution:
(i) 𝑦𝑝 = 𝐴2 𝑥 2 + 𝐴1 𝑥 + 𝐴0 and 𝑦𝑐 = 𝑐1 𝑒 (2i)𝑥 + 𝑐2 𝑒 (−2i)𝑥 are
(2i)𝑥 (−2i)𝑥
𝑦𝑐 = 𝑐1 𝑒 + 𝑐2 𝑒 linearly independent

Solve the coefficient for the particular solution:

𝑦𝑝 = 𝐴2 𝑥 2 + 𝐴1 𝑥 + 𝐴0
Differentiate it, we get:
𝑑𝑦𝑝
𝑑𝑥
= 2𝐴2 𝑥 + 𝐴1
𝑑 2 𝑦𝑝
𝑑𝑥 2
= 2𝐴2
𝑑2 𝑦
Substitute to the ODE equation: 𝑑𝑥 2 + 4𝑦 = 8𝑥 2

>> 2𝐴2 + 4(𝐴2 𝑥 2 + 𝐴1 𝑥 + 𝐴0 ) = 8𝑥 2

>> 4𝐴2 𝑥 2 + 4𝐴1 𝑥 + 4𝐴0 + 2𝐴2 = 8𝑥 2
Comparing the coefficients,
>> 𝑥 2 : 𝐴2 = 2
>> 𝑥: 𝐴1 = 0
𝐴2
>> 𝑥 0 : 𝐴0 = − 2
= −1
The actual particular solution:
𝑦𝑝 = 2𝑥 2 − 1

𝑑2 𝑦
The complete/ general solution to 𝑑𝑥 2 + 4𝑦 = 8𝑥 2 is

𝑦𝑡𝑜𝑡𝑎𝑙 = 𝑦𝑐 + 𝑦𝑝 = 𝑐1 𝑒 (2i)𝑥 + 𝑐2 𝑒 (−2i)𝑥 + 2𝑥 2 − 1

135
 𝑑2 𝑦 𝑑𝑦
For example: Solve 𝑑𝑥 2 − 4 𝑑𝑥 − 12𝑦 = 𝑥𝑒 4𝑥 [Mixture of Polynomial & Exponential Function in ‘x’]

Step 1: Homogeneous Part Step 2: Nonhomogeneous Part

𝑑2 𝑦 𝑑𝑦 𝑑2 𝑦 𝑑𝑦
i.e. 𝑑𝑥 2 − 4 𝑑𝑥 − 12𝑦 = 0 i.e. 𝑑𝑥 2 − 4 𝑑𝑥 − 12𝑦 = 𝑥𝑒 4𝑥

Characteristic equation: The method of undetermined coefficient:

2
𝑚 − 4𝑚 − 12 = 0 RHS: 𝑟(𝑥) = 𝑒 𝛼𝑥 𝑃𝑛 (𝑥) = 𝑥𝑒 4𝑥
(𝑚 − 6)(𝑚 + 2) = 0 where 𝛼 = 4, 𝑛 = 1
𝑚1 = 6 & 𝑚2 = −2 Possible particular solution:
𝑦𝑝 = 𝑒 (4)𝑥 𝑄1 (𝑥) = (𝐴𝑥 + 𝐵)𝑒 4𝑥
Comment: Real and distinct Since 𝛼 ≠ 𝑚 & 𝑚 , treatment is not needed.
1 2
roots

Comment:
Complementary solution:
(i) 𝑦𝑝 = (𝐴𝑥 + 𝐵)𝑒 4𝑥 and 𝑦𝑐 = 𝑐1 𝑒 6𝑥 + 𝑐2 𝑒 −2𝑥 are linearly
6𝑥 −2𝑥
𝑦𝑐 = 𝑐1 𝑒 + 𝑐2 𝑒 independent
Solve the coefficient for the particular solution:
𝑦𝑝 = (𝐴𝑥 + 𝐵)𝑒 4𝑥
Differentiate it, we get:
𝑑𝑦𝑝
= 4(𝐴𝑥 + 𝐵)𝑒 4𝑥 + (𝐴)𝑒 4𝑥
𝑑𝑥
𝑑 2 𝑦𝑝
𝑑𝑥 2
= 16(𝐴𝑥 + 𝐵)𝑒 4𝑥 + 8(𝐴)𝑒 4𝑥
𝑑2 𝑦 𝑑𝑦
Substitute to the ODE equation: 𝑑𝑥 2 − 4 𝑑𝑥 − 12𝑦 = 𝑥𝑒 4𝑥

>> (16(𝐴𝑥 + 𝐵)𝑒 4𝑥 + 8(𝐴)𝑒 4𝑥 ) − 4(4(𝐴𝑥 + 𝐵)𝑒 4𝑥 + (𝐴)𝑒 4𝑥 ) −

12(𝐴𝑥 + 𝐵)𝑒 4𝑥 = 𝑥𝑒 4𝑥
>> ((4𝐴 − 12𝐵)𝑒 4𝑥 ) − 12𝐴𝑥𝑒 4𝑥 = 𝑥𝑒 4𝑥
Comparing the coefficients,
1
>> 𝑥𝑒 4𝑥 : 𝐴 = −12
4𝐴 1
>> 𝑒 4𝑥 : 𝐵 = 12
= − 36

The actual particular solution:

1 1
𝑦𝑝 = (−12 𝑥 − 36) 𝑒 4𝑥

𝑑2 𝑦 𝑑𝑦
The complete/ general solution to 𝑑𝑥 2 − 4 𝑑𝑥 − 12𝑦 = 𝑥𝑒 4𝑥 is
1 1
𝑦𝑡𝑜𝑡𝑎𝑙 = 𝑦𝑐 + 𝑦𝑝 = 𝑐1 𝑒 6𝑥 + 𝑐2 𝑒 −2𝑥 + (−12 𝑥 − 36) 𝑒 4𝑥

136
 𝑑2 𝑦 𝑑𝑦
For example: Solve 𝑑𝑥 2 − 4 𝑑𝑥 + 3𝑦 = 3𝑒 2𝑥 + 4𝑥 [RHS - Mixture of Polynomial & Exponential Function in ‘+’]

Step 1: Homogeneous Part Step 2: Nonhomogeneous Part

𝑑2 𝑦 𝑑𝑦 𝑑2 𝑦 𝑑𝑦
i.e. 𝑑𝑥 2 − 4 𝑑𝑥 + 3𝑦 = 0 i.e. 𝑑𝑥 2 − 4 𝑑𝑥 + 3𝑦 = 3𝑒 2𝑥 + 4𝑥

Characteristic equation: The method of undetermined coefficient:

𝑚2 − 4𝑚 + 3 = 0 RHS (1): 𝑟1 (𝑥) = 𝑒 𝛼𝑥 𝑃𝑛 (𝑥) = 3𝑒 2𝑥
(𝑚 − 1)(𝑚 − 3) = 0 where 𝛼 = 2, 𝑛 = 0
𝑚1 = 1 & 𝑚2 = 3 RHS (2): 𝑟2 (𝑥) = 𝑒 𝛼𝑥 𝑃𝑛 (𝑥) = 4𝑥
where 𝛼 = 0, 𝑛 = 1
Comment: Real and distinct Possible particular solution:
roots
𝑦𝑝,1 = 𝑒 (2)𝑥 𝑄0 (𝑥) = 𝐴𝑒 2𝑥
𝑦𝑝,2 = 𝑒 (0)𝑥 𝑄1 (𝑥) = 𝐵𝑥 + 𝐶
Complementary solution:
𝑦𝑝,𝑡𝑜𝑡𝑎𝑙 = 𝑦𝑝,1 + 𝑦𝑝,2 = 𝐴𝑒 2𝑥 + 𝐵𝑥 + 𝐶
𝑦𝑐 = 𝑐1 𝑒 𝑥 + 𝑐2 𝑒 3𝑥
Since 𝛼 ≠ 𝑚1 & 𝑚2 , treatment is not needed.
Comment:
(i) 𝑦𝑝 = 𝐴𝑒 2𝑥 + 𝐵𝑥 + 𝐶 and 𝑦𝑐 = 𝑐1 𝑒 𝑥 + 𝑐2 𝑒 3𝑥 are
linearly independent
Solve the coefficient for the particular solution:
𝑦𝑝 = 𝐴𝑒 2𝑥 + 𝐵𝑥 + 𝐶
Differentiate it, we get:
𝑑𝑦𝑝
= 2𝐴𝑒 2𝑥 + 𝐵
𝑑𝑥
𝑑 2 𝑦𝑝
𝑑𝑥 2
= 4𝐴𝑒 2𝑥
𝑑2 𝑦 𝑑𝑦
Substitute to the ODE equation: 𝑑𝑥 2 − 4 𝑑𝑥 + 3𝑦 = 3𝑒 2𝑥 + 4𝑥

>> (4𝐴𝑒 2𝑥 ) − 4(2𝐴𝑒 2𝑥 + 𝐵) + 3(𝐴𝑒 2𝑥 + 𝐵𝑥 + 𝐶) = 3𝑒 2𝑥 +

4𝑥
>> (−𝐴𝑒 2𝑥 ) + (3𝐵𝑥) + 3𝐶 − 4𝐵 = 3𝑒 2𝑥 + 4𝑥
Comparing the coefficients,
>> 𝑒 2𝑥 : 𝐴 = −3
4
>> 𝑥: 𝐵 = 3
4𝐵 16
>> 𝑥 0 : 𝐶 = 3
= 9

The actual particular solution:

4 16
𝑦𝑝 = −3𝑒 2𝑥 + 𝑥 +
3 9

137
 𝑑2 𝑦 𝑑𝑦
The complete/ general solution to 𝑑𝑥 2
− 4 𝑑𝑥 + 3𝑦 = 3𝑒 2𝑥 + 4𝑥 is
4 16
𝑦𝑡𝑜𝑡𝑎𝑙 = 𝑦𝑐 + 𝑦𝑝 = 𝑐1 𝑒 𝑥 + 𝑐2 𝑒 3𝑥 − 3𝑒 2𝑥 + 3 𝑥 + 9

Hint: The example above illustrates the linearity or superposition principle, where the solutions can be
added directly as illustrated below.
𝑑2 𝑦 𝑑𝑦
>> 𝑦𝑝,1 = −3𝑒 2𝑥 is the particular solution to − 4 + 3𝑦 = 3𝑒 2𝑥 ;
𝑑𝑥 2 𝑑𝑥
4 16 𝑑2 𝑦 𝑑𝑦
>> 𝑦𝑝,2 = 3 𝑥 + 9 is the particular solution to 𝑑𝑥 2 − 4 𝑑𝑥 + 3𝑦 = 4𝑥;
𝑑2 𝑦 𝑑𝑦
>> 𝑦𝑝,𝑡𝑜𝑡𝑎𝑙 = 𝑦𝑝,1 + 𝑦𝑝,2 is the total particular solution to 2 − 4 + 3𝑦 = 3𝑒 2𝑥 + 4𝑥.
𝑑𝑥 𝑑𝑥

Recommendation: In engineering application, learning mathematics tool such as the strategy/method to

solve ODE problems should not be the main focus, as this can be done perfectly with the use of computer &
algorithm. However, it is important to engineer to evaluate and justify the appropriateness of the selected
mathematical tool in solving certain engineering problem. Furthermore, engineer should learn to interpret
the result and data analysis. For example, we can obtain the solutions of charge and current from the 2nd
order ODE/ 𝑅𝐿𝐶 circuit problem. We should plot it and do further analysis to know the characteristic and
behaviour of the system. This enables us to design things/systems in scientific approach (i.e. analytical study
in this case) instead of trial and error.

For example: A 𝑅𝐿𝐶 Circuit is provided as shown in the figure below where there are an inductor of 𝐿 =
50 ℎ𝑒𝑛𝑟𝑦𝑠, a resistor of 𝑅 = 5 𝑜ℎ𝑚𝑠 and a capacitor of 𝐶 = 8 𝑓𝑎𝑟𝑎𝑑𝑠. At 𝑡 = 0, the switch is closed. Given
𝑑 2 𝑞(𝑡) 𝑑𝑞(𝑡) 1
the 2nd order ODE for the system is 𝐿 +𝑅 + 𝑞(𝑡) = 𝐸(𝑡).
𝑑𝑡 2 𝑑𝑡 𝐶

Find the charge and current at any time 𝑡 > 0 if (a) The voltage is supplied by a DC battery, i.e. 𝐸 = 40 𝑣𝑜𝑙𝑡𝑠

138
 (a) The voltage is supplied by a DC battery, i.e. 𝐸 = 40 𝑣𝑜𝑙𝑡𝑠

Step 1: Homogeneous Part Step 2: Nonhomogeneous Part

𝑑 2 𝑞(𝑡) 𝑑𝑞(𝑡) 1 𝑑 2 𝑞(𝑡) 𝑑𝑞(𝑡) 1
i.e. 50 𝑑𝑡 2 + 5 𝑑𝑡 + 8
𝑞(𝑡) =0 i.e. 50 𝑑𝑡 2
+5 𝑑𝑡
+ 8 𝑞(𝑡) = 40

Characteristic equation: The method of undetermined coefficient:

50𝑚2 + 5𝑚 + 8 = 0
1 RHS : 𝑟(𝑥) = 𝑒 𝛼𝑥 𝑃𝑛 (𝑥) = 40
where 𝛼 = 0, 𝑛 = 0
400𝑚2 + 40𝑚 + 1 = 0
(20𝑚 + 1)2 = 0 Possible particular solution:

𝑚1 = 𝑚2 = −0.05 𝑞𝑝 = 𝑒 (0)𝑥 𝑄0 (𝑥) = 𝐴

Comment: Repeated real root Since 𝛼 ≠ 𝑚1 & 𝑚2 , treatment is not needed.

Complementary solution: Comment:

𝑞𝑐 = 𝑐1 𝑒 −0.05𝑡 + 𝑐2 𝑡𝑒 −0.05𝑡 (i) 𝑞𝑝 = 𝐴 and 𝑞𝑐 = 𝑐1 𝑒 −0.05𝑡 + 𝑐2 𝑡𝑒 −0.05𝑡 are linearly

independent

Comment: Solve the coefficient for the particular solution:

−0.05𝑡 −0.05𝑡
(i) 𝑞𝑐,1 = 𝑒 and 𝑞𝑐,2 = 𝑒 are 𝑞𝑝 = 𝐴
linearly dependent.
(ii) Treatment is done so that 𝑞𝑐,1 = Differentiate it, we get:
𝑒 −0.05𝑡 and 𝑞𝑐,2 = 𝑡𝑒 −0.05𝑡 are 𝑑𝑞𝑝
𝑑𝑡
=0
linearly independent.
𝑑 2 𝑞𝑝
𝑑𝑡 2
=0
𝑑 2 𝑞(𝑡) 𝑑𝑞(𝑡)
Substitute to the ODE equation: 50 +5 +
𝑑𝑡 2 𝑑𝑡
1
𝑞(𝑡) = 40
8
1
>> 50(0) + 5(0) + 𝐴 = 40
8

>>𝐴 = 320
The actual particular solution:
𝑞𝑝 = 320

𝑑 2 𝑞(𝑡) 𝑑𝑞(𝑡) 1
The complete/ general solution to 50 𝑑𝑡 2 + 5 𝑑𝑡 + 8 𝑞(𝑡) = 40 is
(i) Charge solution,
𝑞𝑡𝑜𝑡𝑎𝑙 = 𝑞𝑐 + 𝑞𝑝 = 𝑐1 𝑒 −0.05𝑡 + 𝑐2 𝑡𝑒 −0.05𝑡 + 320

(ii) Differentiate it, we obtain the current solution

𝑑𝑞𝑡𝑜𝑡𝑎𝑙
𝑖𝑡𝑜𝑡𝑎𝑙 = 𝑑𝑡 = 𝑐2 𝑒 −0.05𝑡 − 0.05𝑒 −0.05𝑡 (𝑐1 + 𝑐2 𝑡)

139
Solution to Initial value problem

Note:
The general solution is not the actual solution to the problem because it has infinite 𝑐1 , 𝑐2 that can
satisfy the problem. Recall that in the initial value problem, we can further solve the constant 𝑐1 , 𝑐2
in the system if the initial condition of the problem is known in advance.

Hint for the initial condition:

“At 𝑡 = 0, the switch is closed.”
This shows that charge and current flows only when 𝑡 > 0. Thus, the initial conditions are
𝑞(0) = 0, 𝑖(0) = 0

Apply the initial conditions to the general solution, we obtain:

𝑞𝑡𝑜𝑡𝑎𝑙 = 𝑞𝑐 + 𝑞𝑝 = 𝑐1 𝑒 −0.05𝑡 + 𝑐2 𝑡𝑒 −0.05𝑡 + 320
>> 𝑞(0) = 𝑐1 𝑒 −0.05(0) + 𝑐2 (0)𝑒 −0.05(0) + 320 = 0
>> 𝑐1 = −320

𝑖𝑡𝑜𝑡𝑎𝑙 = 𝑐2 𝑒 −0.05𝑡 − 0.05𝑒 −0.05𝑡 (𝑐1 + 𝑐2 𝑡)

>> 𝑖𝑡𝑜𝑡𝑎𝑙 = 𝑐2 𝑒 −0.05(0) − 0.05𝑒 −0.05(0) (𝑐1 + 𝑐2 (0)) = 0
>> 𝑐2 − 0.05(𝑐1 ) = 0
>> 𝑐2 = 0.05(𝑐1 ) = 0.05(−320) = −16

The actual charge solution to the RLC circuit problem:

𝑞𝑡𝑜𝑡𝑎𝑙 = −320𝑒 −0.05𝑡 − 16𝑡𝑒 −0.05𝑡 + 320

The actual current solution to the RLC circuit problem:

𝑖𝑡𝑜𝑡𝑎𝑙 = −16𝑒 −0.05𝑡 − 0.05𝑒 −0.05𝑡 (−320 − 16𝑡)
𝑖𝑡𝑜𝑡𝑎𝑙 = 𝑒 −0.05𝑡 (16𝑡)

Note: We plot the result of charge and current solutions (for the 𝑅𝐿𝐶 circuit problem with 𝐿 = 50 ℎ𝑒𝑛𝑟𝑦𝑠,
a resistor of 𝑅 = 5 𝑜ℎ𝑚𝑠 and a capacitor of 𝐶 = 8 𝑓𝑎𝑟𝑎𝑑𝑠) as follows. The intention of this example is to
encourage student to link the mathematical result to the actual problem instead of calculating it for nothing.
However, data analysis on specific engineering problem requires certain knowledge on the subject and
hence it is out of the scope in this study.

140
Hint: Relate this to the scenario of charging battery.

Data analysis – Think:

(1) Identify the transient and steady state region in the graph. Why it is important to identify them?
Ans: Transient region happens within around 0 − 150𝑠; while steady state region occurs after that.
This relationship is important to the RLC circuit problem such as charging battery, where we can
estimate how much time is needed to fully charge the battery. Moreover, it shows that the current is
reduced to zero once it is fully charged.

(2) What are the relationships between complementary solution & particular solution with these
transient and steady state region?
Ans: We know that the total solution is a combination of the complementary solution & particular
solution. The complementary solution contributes more within the transient region and its effect
diminishes within the steady state region; while the particular solution contributes significantly within
the steady state region.

(3) Why does the charge behave as such: increase over a time initially and at time approximately 150s,
the charge remains constant afterward? Why does the current behave as such: increase over a time
initially and decrease after it reaches its maximum? The charge decreases to zero at time approximately
150s and remains constant afterward.

Ans: To understand the changes, we need to check the equation, 𝑞𝑡𝑜𝑡𝑎𝑙 = −320𝑒 −0.05𝑡 −
16𝑡𝑒 −0.05𝑡 + 320, where complementary solution of charge, 𝑞𝑐 = −320𝑒 −0.05𝑡 − 16𝑡𝑒 −0.05𝑡 and its
particular solution, 𝑞𝑝 = 320.Since the complementary solution consists the exponential function,
i.e. 𝑒 −0.05𝑡 . As the time increase, this function will approach zero and thus diminish. At the same time,
the particular solution remains all the time. Hence, 𝑞𝑡𝑜𝑡𝑎𝑙 = −320𝑒 −0.05𝑡 − 16𝑡𝑒 −0.05𝑡 + 320
increase over a time initially and at time approximately 150s, the charge remains constant afterward.
𝑑𝑞𝑡𝑜𝑡𝑎𝑙
The current is the rate of change of the charge, i.e. 𝑖𝑡𝑜𝑡𝑎𝑙 = 𝑑𝑡
= 𝑒 −0.05𝑡 (16𝑡). Since it consists of
the exponential function, it illustrates that the charging rate will be higher initially and reduce to zero
afterwards, as this function will approach to zero as the time increase.

Think: In case you want to reduce the charging time or increase the charging capacity, what should you do?

141
12.4.4 SOLVING ALL TYPES OF NONHOMOGENEOUS LINEAR DIFFERENTIAL EQUATION
WITH METHOD OF VARIATION OF PARAMETERS

The previous undetermined coefficients method is only applicable for simple functions such as a mixture of
exponential and polynomial functions. However, it is impractical to solve complicated function other than
exponential and polynomial functions such as tangent function, Mixture of Polynomial & Exponential
Function in ‘÷’, logarithmic function, etc. To solve 2nd order ODE with complicated function in its RHS, we
recommend user to use the method of variation of parameters.

The steps of implementing the method of variation of parameters are as follow:

𝑑2 𝑦 𝑑𝑦
(1) Standard form: 𝑑𝑥 2
+ 𝑝(𝑥) 𝑑𝑥 + 𝑞(𝑥)𝑦 = 𝑟(𝑥)

(2) Compute the complementary solution, 𝑦𝑐 = 𝑐1 𝑦1 (𝑥) + 𝑐2 𝑦2 (𝑥) by using method introduced in section
12.4.1 & 12.4.2, where 𝑐1 & 𝑐2 are arbitrary constants.

(3) Compute the particular solution, 𝑦𝑝 = 𝑢1 (𝑥)𝑦1 (𝑥) + 𝑢2 (𝑥)𝑦2 (𝑥)

𝑟(𝑥)𝑦2 (𝑥) 𝑟(𝑥)𝑦1 (𝑥)
where function 𝑢1 (𝑥) = − ∫ 𝑑𝑥 and 𝑢2 (𝑥) = ∫ 𝑑𝑥;
𝑊(𝑦1 ,𝑦2 ) 𝑊(𝑦1 ,𝑦2 )
𝑦1 𝑦2
Wronskian, 𝑊(𝑦1 , 𝑦2 ) = |𝑑𝑦1 𝑑𝑦2 |
𝑑𝑥 𝑑𝑥

The drawback of this technique is that it is time consuming to complete the integration operation
and there are cases where the integration function cannot be solved analytically (using calculus). In this case,
we may need an advance tool such as numerical method to solve the integration problem.

142
 𝑑2 𝑦 𝑑𝑦
For example: Solve 𝑑𝑥 2 + 2 𝑑𝑥 + 𝑦 = 𝑒 −𝑥 𝑙𝑛𝑥

Step 1: Homogeneous Part Step 2: Nonhomogeneous Part

𝑑2 𝑦 𝑑𝑦 𝑑2 𝑦 𝑑𝑦
i.e. 𝑑𝑥 2 + 2 𝑑𝑥 + 𝑦 = 0, i.e. 𝑑𝑥 2 + 2 𝑑𝑥 + 𝑦 = 𝑒 −𝑥 𝑙𝑛𝑥

Characteristic equation: The method of undetermined coefficient:

𝑚2 + 2𝑚 + 1 = 0 RHS : 𝑟(𝑥) = 𝑒 𝛼𝑥 𝑃𝑛 (𝑥) = 𝑒 −𝑥 𝑙𝑛𝑥
(𝑚 + 1)2 = 0 where 𝛼 = −1, 𝑛 = 𝑛𝑜𝑡 𝑎𝑝𝑝𝑙𝑖𝑐𝑎𝑏𝑙𝑒
𝑚1 = 𝑚2 = −1 Comment: The RHS function is a complicated function that
can’t solved by the method of undetermined coefficient.

Comment: Repeated real root

The method of variation of parameters:
𝑟(𝑥)𝑦 (𝑥)
Complementary solution: 𝑢1 (𝑥) = − ∫ 𝑊(𝑦 2,𝑦 ) 𝑑𝑥
1 2

𝑦𝑐 = 𝑐1 𝑒 −𝑥 + 𝑐2 𝑥𝑒 −𝑥 (𝑒 −𝑥 𝑙𝑛𝑥)(𝑥𝑒 −𝑥 )
= −∫ 𝑒 −2𝑥
𝑑𝑥

= − ∫ 𝑥𝑙𝑛𝑥 𝑑𝑥
Comment: Using integration by part method:
(i) 𝑦𝑐,1 = 𝑒 −𝑥 and 𝑦𝑐,2 = 𝑒 −𝑥 are linearly Let 𝑢 = 𝑙𝑛𝑥; 𝑑𝑣 = 𝑥𝑑𝑥
dependent.
1 𝑥2
(ii) Treatment is done so that 𝑦𝑐,1 = 𝑑𝑢 = 𝑥 𝑑𝑥; 𝑣 = 2
𝑒 −𝑥 and 𝑦𝑐,2 = 𝑥𝑒 −𝑥 are linearly
𝑢1 (𝑥) = − ∫ 𝑥𝑙𝑛𝑥 𝑑𝑥 = −(𝑢𝑣 − ∫ 𝑣𝑑𝑢)
independent.
𝑥2 𝑥2 1
= − (𝑙𝑛𝑥 ( 2 ) − ∫ 2 𝑥
𝑑𝑥)
Compute the Wronskian, 𝑊(𝑦1 , 𝑦2 )
𝑦1 𝑦2 𝑥2 𝑥
= − (𝑙𝑛𝑥 ( 2 ) − ∫ 2 𝑑𝑥)
= |𝑑𝑦1 𝑑𝑦2 |
𝑑𝑥−𝑥 𝑑𝑥 = (−𝑙𝑛𝑥 ( ) +
𝑥2 𝑥2
)
𝑒 𝑥𝑒 −𝑥 2 4
= | −𝑥 |
−𝑒 −𝑥𝑒 −𝑥 + 𝑒 −𝑥 𝑟(𝑥)𝑦1 (𝑥)
=𝑒 −𝑥 (−𝑥𝑒 −𝑥
+𝑒 −𝑥 )
− (𝑥𝑒 −𝑥 )(−𝑒 −𝑥 ) 𝑢2 (𝑥) = ∫ 𝑑𝑥
𝑊(𝑦1 ,𝑦2 )
−2𝑥
=𝑒
(𝑒 −𝑥 𝑙𝑛𝑥)(𝑒 −𝑥 )
=∫ 𝑒 −2𝑥
𝑑𝑥

= ∫ 𝑙𝑛𝑥 𝑑𝑥
Using integration by part method:
1
Let 𝑢 = 𝑙𝑛𝑥; 𝑑𝑣 = 𝑑𝑥, then 𝑑𝑢 = 𝑥 𝑑𝑥; 𝑣 = 𝑥

𝑢2 (𝑥) = ∫ 𝑙𝑛𝑥 𝑑𝑥 = (𝑢𝑣 − ∫ 𝑣𝑑𝑢)

1
= (𝑙𝑛𝑥(𝑥) − ∫ 𝑥 𝑥 𝑑𝑥)

= (𝑙𝑛𝑥(𝑥) − ∫ 1𝑑𝑥)
= (𝑙𝑛𝑥(𝑥) − 𝑥)
143
 The particular solution:
𝑦𝑝 = 𝑢1 (𝑥)𝑦1 (𝑥) + 𝑢2 (𝑥)𝑦2 (𝑥)
𝑥2 𝑥2
= (−𝑙𝑛𝑥 ( 2 ) + 4
) 𝑒 −𝑥 + (𝑙𝑛𝑥(𝑥) − 𝑥)(𝑥𝑒 −𝑥 )
𝑥2 3𝑥 2
= (𝑙𝑛𝑥 ( 2 ) − 4
) 𝑒 −𝑥

𝑑2 𝑦 𝑑𝑦
The complete/ general solution to 𝑑𝑥 2 + 2 𝑑𝑥 + 𝑦 = 𝑒 −𝑥 𝑙𝑛𝑥 is
𝑥2 3𝑥 2
𝑦𝑡𝑜𝑡𝑎𝑙 = 𝑦𝑐 + 𝑦𝑝 = 𝑐1 𝑒 −𝑥 + 𝑐2 𝑥𝑒 −𝑥 + (𝑙𝑛𝑥 ( 2 ) − 4
) 𝑒 −𝑥

144
 𝑑2 𝑦 𝑑𝑦 𝑒𝑥
For example: Solve 𝑑𝑥 2 − 2 𝑑𝑥 + 𝑦 = 𝑥 2 +1

Step 1: Homogeneous Part Step 2: Nonhomogeneous Part

𝑑2 𝑦 𝑑𝑦 𝑑2 𝑦 𝑑𝑦 𝑒𝑥
i.e. 𝑑𝑥 2 − 2 𝑑𝑥 + 𝑦 = 0 i.e. 𝑑𝑥 2 − 2 𝑑𝑥 + 𝑦 = 𝑥 2 +1

Characteristic equation: The method of undetermined coefficient:

𝑚2 − 2𝑚 + 1 = 0 𝑒𝑥
RHS : 𝑟(𝑥) = 𝑒 𝛼𝑥 𝑃𝑛 (𝑥) = 𝑥 2 +1
(𝑚 − 1)2 = 0
where 𝛼 = 1, 𝑛 = 𝑛𝑜𝑡 𝑎𝑝𝑝𝑙𝑖𝑐𝑎𝑏𝑙𝑒
𝑚1 = 𝑚2 = 1
Comment: The RHS function is a complicated function that can’t
solved by the method of undetermined coefficient.
Comment: Repeated real root
The method of variation of parameters:
Complementary solution: 𝑟(𝑥)𝑦2 (𝑥)
𝑢1 (𝑥) = − ∫ 𝑑𝑥
𝑊(𝑦1 ,𝑦2 )
𝑦𝑐 = 𝑐1 𝑒 𝑥 + 𝑐2 𝑥𝑒 𝑥
𝑒𝑥
( 2 )(𝑥𝑒 𝑥 )
= − ∫ 𝑥 +1𝑒 2𝑥 𝑑𝑥
𝑥
Comment: = − ∫ 2 𝑑𝑥
𝑥 +1
(i) 𝑦𝑐,1 = 𝑒 𝑥 and 𝑦𝑐,2 = 𝑒 𝑥 are Using substitution method:
linearly dependent.
Let 𝑢 = 𝑥 2 + 1; 𝑑𝑢 = 2𝑥𝑑𝑥
(ii) Treatment is done so that
𝑦𝑐,1 = 𝑒 𝑥 and 𝑦𝑐,2 = 𝑥𝑒 𝑥 are 𝑢1 (𝑥) = − ∫ 1 𝑑𝑢
𝑢 2
linearly independent.
1
= − 𝑙𝑛|𝑢|
2
1
Compute the Wronskian, = − 𝑙𝑛|𝑥 2 + 1|
2
𝑊(𝑦1 , 𝑦2 ) 𝑟(𝑥)𝑦 (𝑥)
𝑦1 𝑦2 𝑢2 (𝑥) = ∫ 𝑊(𝑦 1,𝑦 ) 𝑑𝑥
1 2
= |𝑑𝑦1 𝑑𝑦2 |
𝑒𝑥
𝑑𝑥 𝑑𝑥 𝑥 ( 2 )(𝑒 𝑥 )
𝑥 +1
𝑒𝑥 𝑥𝑒 =∫ 𝑑𝑥
=| 𝑥 | 𝑒 2𝑥
𝑒 𝑥𝑒 𝑥 + 𝑒 𝑥 1
=𝑒 𝑥 (𝑥𝑒 𝑥
+ 𝑒 − (𝑥𝑒 𝑥 )(𝑒 𝑥 )
𝑥) = ∫ 𝑥 2 +1 𝑑𝑥
2𝑥
=𝑒 𝑑𝑥
Using trigonometric substitution: Let 𝑥 = 𝑡𝑎𝑛𝜃, 𝑑𝜃 = 𝑠𝑒𝑐 2 𝜃
1
𝑢2 (𝑥) = ∫ 𝑡𝑎𝑛2 𝜃+1 𝑠𝑒𝑐 2 𝜃𝑑𝜃

Using 𝑡𝑟𝑖𝑔𝑜𝑛𝑜𝑚𝑒𝑡𝑟𝑖𝑐 identity: 𝑡𝑎𝑛2 𝜃 + 1 = 𝑠𝑒𝑐 2 𝜃

𝑢2 (𝑥) = ∫ 1 𝑑𝜃 = 𝜃
= 𝑡𝑎𝑛−1 𝑥

145
 The particular solution:
𝑦𝑝 = 𝑢1 (𝑥)𝑦1 (𝑥) + 𝑢2 (𝑥)𝑦2 (𝑥)
1
= (− 2 𝑙𝑛|𝑥 2 + 1|) 𝑒 𝑥 + (𝑡𝑎𝑛−1 𝑥)(𝑥𝑒 𝑥 )
1
= (𝑥𝑡𝑎𝑛−1 𝑥 − 2 𝑙𝑛|𝑥 2 + 1|) 𝑒 𝑥

𝑑2 𝑦 𝑑𝑦 𝑒𝑥
The complete/ general solution to 𝑑𝑥 2 − 2 𝑑𝑥 + 𝑦 = 𝑥 2 +1 is
1
𝑦𝑡𝑜𝑡𝑎𝑙 = 𝑦𝑐 + 𝑦𝑝 = 𝑐1 𝑒 𝑥 + 𝑐2 𝑥𝑒 𝑥 + (𝑥𝑡𝑎𝑛−1 𝑥 − 𝑙𝑛|𝑥 2 + 1|) 𝑒 𝑥
2

146
POWER SERIES SOLUTIONS FOR
DIFFERENTIAL EQUATIONS
WEEK 13: POWER SERIES SOLUTIONS FOR DIFFERENTIAL EQUATIONS
13.1 POWER SERIES METHOD

Power Series Method

The power series method is the standard basic method for solving linear differential equations with
variable coefficients. It gives solutions in the form of power series.

Power Series
A power series is an infinite series of the form
∞

∑ 𝑎𝑛 (𝑥 − 𝑥0 )𝑛
𝑛=0
= 𝑎0 + 𝑎1 (𝑥 − 𝑥0 ) + 𝑎2 (𝑥 − 𝑥0 )2 + ⋯ (1)
where a0, a1, a2, ... are real constants, called the coefficients of the series, x0 is a constant, called the center
of the series, and x is a variable.

In particular, if x0 = 0, a power series in powers of x is obtained

∞

∑ 𝑎𝑛 𝑥 𝑛 = 𝑎0 + 𝑎1 𝑥 + 𝑎2 𝑥 2 + 𝑎3 𝑥 3 + ⋯
𝑛=0

Familiar examples of power series:

∞
1
(i) = ∑ 𝑥𝑚 = 1 + 𝑥 + 𝑥2 + ⋯
1−𝑥
𝑚=0

∞
𝑥
𝑥𝑛 𝑥2 𝑥3
(ii) 𝑒 = ∑ = 1 + 𝑥 + + +⋯
𝑛! 2! 3!
𝑛=0

∞
(−1)𝑛 𝑥 2𝑛 𝑥2 𝑥4
(iii) cos 𝑥 = ∑ = 1 − + ±⋯
(2𝑛)! 2! 4!
𝑛=0

∞
(−1)𝑛 𝑥 2𝑛+1 𝑥3 𝑥5
(iv) sin 𝑥 = ∑ = 𝑥 − + ±⋯
(2𝑛 + 1)! 3! 5!
𝑛=0

147
 ∞
(−1)𝑛+1 𝑥 𝑛 𝑥2 𝑥3
(v) ln(1 + 𝑥) = ∑ =𝑥 − + − ⋯
𝑛 2 3
𝑛=1

13.1.1 BASIC CONCEPTS OF POWER SERIES

The nth partial sum of (1) is

𝑠𝑛 (𝑥) = 𝑎0 + 𝑎1 (𝑥 − 𝑥0 ) + 𝑎2 (𝑥 − 𝑥0 )2 + ⋯ + 𝑎𝑛 (𝑥 − 𝑥0 )𝑛 (2)
where n = 0, 1, ... . If the terms of sn are from (1), the remaining expression is
𝑅𝑛 (𝑥) = 𝑎𝑛+1 (𝑥 − 𝑥0 )𝑛+1 + 𝑎𝑛+2 (𝑥 − 𝑥0 )𝑛+2 + ⋯
and is called the remainder of (1) after the term 𝑎𝑛 (𝑥 − 𝑥0 )𝑛 .

Example:
For the geometric series
1 + 𝑥 + 𝑥2 + ⋯ + 𝑥𝑛 + ⋯
Then:
𝑠1 = 1 + 𝑥 𝑅1 = 𝑥 2 + 𝑥 3 + 𝑥 4 + ⋯
𝑠2 = 1 + 𝑥 + 𝑥 2 𝑅2 = 𝑥 2 + 𝑥 3 + 𝑥 4 + ⋯
etc.

If for some x = x1, sn(x) converges, that is, lim 𝑠𝑛 (𝑥1 ) = 𝑠(𝑥1 ) then the series (1) converges, or is called
𝑛→∞
convergent at x = x1; and the number s(x1) is called the value or sum of (1) at x1, and can be written as
∞

𝑠(𝑥1 ) = ∑ 𝑎𝑛 (𝑥1 − 𝑥0 )𝑛
𝑛=0

If the sequence is divergent at x = x1, then the series (1) is said to diverge, or to be divergent at x = x1.

Note:
1. The series (1) converges at x = x0 when all its terms except for the first a0 are zero. In unusual
cases this may be the only x for which (1) converges.

2. If there are further values of x for which the series (1) converges, these value form an interval,
called the convergence interval. If this interval is finite, it has the midpoint x0 so that it is of the
form
|𝑥 − 𝑥0 | < 𝑅

148
 and the series (1) converges for all x such that |𝑥 − 𝑥0 | < 𝑅 and diverges for all x such that
|𝑥 − 𝑥0 | > 𝑅. The number R is called the radius of convergence of (1). It can be obtained from
either of the following formulas:
1 𝑎𝑛+1
(𝑎) 𝑅 = (𝑏) 𝑅 = lim | | (3)
𝑛
lim √|𝑎𝑚 | 𝑛→∞ 𝑎𝑛
𝑛→∞

provided these limits exist and are not zero. [If they are infinite, then (1) converges only at the
center x0.]

3. The convergence interval may sometimes be infinite, that is, (1) converges for all x. For example,
if the limit in (3a) and (3b) is zero. Then 𝑅 = ∞, for convenience.

4. Since power series are functions of x and we know that not every series will in fact exist, it then
makes sense to ask if a power series will exist for all x. This question is answered by looking at the
convergence of the power series. We say that a power series converges for x = c if the series,
∞

∑ 𝑎𝑛 (𝑐 − 𝑥0 )𝑛
𝑛=0

converges. Recall that this series will converge if the limit of partial sums,
𝑁

lim ∑ 𝑎𝑛 (𝑐 − 𝑥0 )𝑛
𝑁→∞
𝑛=0

exists and is finite. In other words, a power series will converge for x = c if
∞

∑ 𝑎𝑛 (𝑐 − 𝑥0 )𝑛
𝑛=0

is a finite number.

5. A power series will always converge if x = x0. In this case the power series will become
∞

∑ 𝑎𝑛 (𝑐 − 𝑥0 )𝑛 = 𝑎0
𝑛=0

With this it is known now that power series are guaranteed to exist for at least one value of x. The
following fact about the convergence of a power series is derived.

Fact
Given a power series, (1), there will exist a number 0 ≤ 𝜌 ≤ ∞ so that the power series will converge
for|𝑥 − 𝑥0 | < 𝜌 and diverge for |𝑥 − 𝑥0 | > 𝜌. This number is called the radius of convergence.

149
13.1.2 TEST FOR CONVERGENCE

1. If lim 𝑢𝑛 = 0 the series may be convergent; and

𝑛→∞

if lim 𝑢𝑛 ≠ 0 the series is certainly divergent.

𝑛→∞

2. Comparison test – useful standard series

1 1 1 1 1 1
𝑝
+ 𝑝 + 𝑝 + 𝑝 + 𝑝 + ⋯+ 𝑝 +⋯
1 2 3 4 5 𝑛
For p > 1, the series converges; for p < 1, the series diverges.

3. D’Alembert’s Ratio Test for positive terms

Let u1 + u2 + u3 + u4 + ... + un + … be a series of positive terms. Find expressions for un and un+1, that
is, the nth term and the (n + 1)th term, respectively, and form the ratio
𝑢𝑛+1
𝑢𝑛
Then, find the limiting value for this ratio,
𝑢𝑛+1
𝜌 = lim
𝑛→∞ 𝑢𝑛

If  < 1, the series converges;

 > 1, the series diverges;
 = 1, the series may converge or diverge and the test gives no definite information.

4. For general series:

(i) if ∑|𝑢𝑛 | converges, ∑ 𝑢𝑛 is absolutely convergent
(ii) if ∑|𝑢𝑛 | diverges, but ∑ 𝑢𝑛 converges, then ∑ 𝑢𝑛 is conditionally convergent

Problem Set 1
Find the radius of convergence of the following series.
∞
𝑘
1. ∑
2𝑘
𝑘=1

Solution:

150
 𝑘 + 1 2𝑘 𝑘+1 1 𝑘+1 1 1 1
𝜌 = lim | 𝑘+1 ∙ | = lim | | = lim | |= lim |1 + | =
𝑘→∞ 2 𝑘 𝑘→∞ 2𝑘 2 𝑘→∞ 𝑘 2 𝑘→∞ 𝑘 2
The series converges

∞
(−1)𝑘 (𝑥 − 3)𝑘
2. ∑
3𝑘 (𝑘 + 1)
𝑘=1

Solution:
(−1)𝑘+1 (𝑥 − 3)𝑘+1 3𝑘 (𝑘 + 1) (−1)(𝑥 − 3)(𝑘 + 1)
𝜌 = lim | 𝑘+1 ∙ | = lim | |
𝑘→∞ 3 (𝑘 + 1 + 1) (−1)𝑘 (𝑥 − 3)𝑘 𝑘→∞ 3(𝑘 + 2)

|(𝑥 − 3)| (−1)(𝑘 + 1) |(𝑥 − 3)| |(𝑥 − 3)|

= lim | | = ∙1=
3 𝑘→∞ (𝑘 + 2) 3 3
Series converges when 𝜌 < 1
|(𝑥 − 3)|
<1
3
|(𝑥 − 3)| < 3
−(𝑥 − 3) < 3 ⇒ 𝑥>0
or
(𝑥 − 3) < 3 ⇒ 𝑥<6
Convergence interval (0, 6)
Radius of convergence 3

13.1.3 OPERATIONS OF POWER SERIES

Three permissible operations on power series: differentiation, addition, and multiplication.

(1) Termwise differentiation
A power series may be differentiated term by term. If
∞

𝑦(𝑥) = ∑ 𝑎𝑛 (𝑥 − 𝑥0 )𝑛
𝑛=0

converges for |𝑥 − 𝑥0 | < 𝑅 where R > 0, then the series obtained by differentiating term by term
also converges for those x and represents the derivative y’ of y for those x, that is,
∞

𝑦′(𝑥) = ∑ 𝑛𝑎𝑛 (𝑥 − 𝑥0 )𝑛−1

𝑛=1

151
 Similarly,
∞

𝑦′′(𝑥) = ∑ 𝑛(𝑛 − 1)𝑎𝑛 (𝑥 − 𝑥0 )𝑛−2

𝑛=2

and so on.

(2) Termwise addition

Two power series may be added term by term. If the series
∞ ∞
𝑛
∑ 𝑎𝑛 (𝑥 − 𝑥0 ) and ∑ 𝑏𝑛 (𝑥 − 𝑥0 )𝑛
𝑛=0 𝑛=0

have positive radii of convergence and their sums are f(x) and g(x), respectively, then the series
∞

∑(𝑎𝑛 + 𝑏𝑛 )(𝑥 − 𝑥0 )𝑛
𝑛=0

converges and represent f(x) + g(x) for each x that lies in the interior of the convergence interval
of each of the given series.
(3) Termwise multiplication
Two power series may be multiplied term by term. Suppose that
∞ ∞
𝑛
∑ 𝑎𝑛 (𝑥 − 𝑥0 ) and ∑ 𝑏𝑛 (𝑥 − 𝑥0 )𝑛
𝑛=0 𝑛=0

have positive radii of convergence and let f(x) and g(x) be their sums, respectively. Then the series
obtained by multiplying each term of the first series by each term of the second series and
collecting like powers of x – x0, that is,
∞

∑(𝑎0 𝑏𝑛 + 𝑎1 𝑏𝑛−1 + ⋯ + 𝑎𝑛 𝑏0 )(𝑥 − 𝑥0 )𝑛

𝑛=0

= 𝑎0 𝑏0 + (𝑎0 𝑏1 + 𝑎1 𝑏0 )(𝑥 − 𝑥0 ) + (𝑎0 𝑏2 + 𝑎1 𝑏1 + 𝑎2 𝑏0 )(𝑥 − 𝑥0 )2 + ⋯

converges and represents f(x)g(x) for each x in the interior of convergence interval of each of the
given series.

13.1.4 VANISHING ALL COEFFICIENTS – A CONDITION THAT IS A BASIC TOOL OF THE

POWER SERIES METHOD

If a power series has a positive radius of convergence and a sum that is identically zero throughout its interval
of convergence, then each coefficient of the series is zero.

152
Sifting summation indices
(1) An index of summation is a dummy and can be changed.
Example:
∞ ∞
3𝑛 𝑛 2 3𝑘 𝑘 2 81
∑ =∑ = 1 + 18 + + ⋯.
𝑛! 𝑘! 2
𝑛=1 𝑘=1

(2) An index of summation can be “shifted”.

If set n = s + 2, then s = n – 2, and
∞ ∞
𝑛−2
∑ 𝑛(𝑛 − 1)𝑎𝑛 (𝑥 − 𝑥0 ) = ∑(𝑠 + 2)(𝑠 + 1)𝑎𝑠+2 𝑥 𝑠 = 2𝑎2 + 6𝑎3 𝑥 + 12𝑎4 𝑥 2 + ⋯
𝑛=2 𝑠=0

When writing the sum of two series,

∞ ∞
2 𝑛−2
𝑥 ∑ 𝑛(𝑛 − 1)𝑎𝑛 𝑥 + 2 ∑ 𝑛𝑎𝑛 𝑥 𝑛−1
𝑛=2 𝑛=1

= 𝑥 2 (2𝑎2 + 6𝑎3 𝑥 + 12𝑎4 𝑥 2 + ⋯ ) + 2(𝑎1 + 2𝑎2 𝑥 + 3𝑎3 𝑥 2 + ⋯

as a single series; firstly, take x2 and 2, respectively, inside the summation, obtaining
∞ ∞

∑ 𝑛(𝑛 − 1)𝑥 + ∑ 2𝑛𝑎𝑛 𝑥 𝑛−1

𝑛

𝑛=2 𝑛=1

and then set n = s and n – 1 = s, respectively, obtaining

∞ ∞

∑ 𝑠(𝑠 − 1)𝑎𝑠 𝑥 + ∑ 2(𝑠 + 1) 𝑎𝑠+1 𝑥 𝑠

𝑠

𝑠=2 𝑠=0

where s = 2 can be replaced by s = 0, so that

∞

∑[𝑠(𝑠 − 1)𝑎𝑠 + 2(𝑠 + 1)𝑎𝑠+1 ] 𝑥 𝑠 = 2𝑎1 + 4𝑎2 𝑥 + (2𝑎2 + 6𝑎3 )𝑥 2 + (6𝑎3 + 8𝑎4 )𝑥 3 + ⋯
𝑠=0

Theorem (Existence of power series solution)

If the functions p, q, and r in the differential equation
(4) 𝑦 ′′ + 𝑝(𝑥)𝑦 ′ + 𝑞(𝑥)𝑦 = 𝑟(𝑥)
are analytic at x =x0, then every solution y(x) of (4) is analytic at x = x0 and can thus be represented by
a power series in powers x – x0 with radius of convergence R > 0.

153
Problem Set 2
Find the series of the following functions.
2
1. 𝑒𝑥
Solution:
∞
𝑥2
(𝑥 2 )𝑚 (𝑥 2 )2 (𝑥 2 )3
𝑒 = ∑ = 1 + 𝑥2 + + +⋯
𝑚! 2! 3!
𝑚=0

𝑥4 𝑥6
= 1 + 𝑥2 + + +⋯
2 6

2. 𝑒 𝑥 + sin 𝑥
Solution:
∞ ∞
𝑥
𝑥𝑚 (−1)𝑚 𝑥 2𝑚+1
𝑒 + sin 𝑥 = ∑ + ∑
𝑚! (2𝑚 + 1)!
𝑚=0 𝑚=0

𝑥2 𝑥3 𝑥3 𝑥5
=1 + 𝑥 + + + ⋯+ 𝑥 − + ±⋯
2! 3! 3! 5!
𝑥2 𝑥4 2𝑥 5
= 1 + 2𝑥 + + + +⋯
2 4! 5!

3. 𝑒 𝑥 (cos 𝑥)
Solution:
∞ ∞
𝑥
𝑥𝑚 (−1)𝑚 𝑥 2𝑚
𝑒 (cos 𝑥) = ( ∑ )(∑ )
𝑚! (2𝑚)!
𝑚=0 𝑚=0
2
𝑥 𝑥3 𝑥2 𝑥4
= (1 + 𝑥 + + + ⋯ ) (1 − + ± ⋯)
2! 3! 2! 4!
𝑥2 𝑥4 𝑥3 𝑥5
=1 − + + …+ 𝑥 − + +⋯
2! 4! 2! 4!

13.1.5 IDEA OF THE POWER SERIES METHOD

Before finding series solutions to differential equations; we need to determine when we can find series
solutions to differential equations with nonconstant coefficients. So, let’s start with the differential
equation,
𝑝(𝑥)𝑦 ′′ + 𝑞(𝑥)𝑦 ′ + 𝑟(𝑥)𝑦 = 0 (5)

154
To this point we’ve only dealt with constant coefficients. However, with series solutions we can now have
nonconstant coefficient differential equations. Also, here we will be dealing only with polynomial
coefficients.
Now, we say that x = x0 is an ordinary point if provided both
𝑞(𝑥) 𝑟(𝑥)
and
𝑝(𝑥) 𝑝(𝑥)
are analytic at x = x0. That is to say that these two quantities have Taylor series around x = x0. Since, we
are only dealing with coefficients that are polynomials so this will be equivalent to saying that
𝑝(𝑥0 ) ≠ 0
for most of the problems.
If a point is not an ordinary point we call it a singular point.
The basic idea to finding a series solution to a differential equation is to assume that we can write the
solution as a power series in the form,
∞

𝑦(𝑥) = ∑ 𝑎𝑛 (𝑥 − 𝑥0 )𝑛 (6)
𝑛=0

and then try to determine what the an’s need to be. We will only be able to do this if the point x = x0, is an
ordinary point. We will usually say that (6) is a series solution around x = x0.

Problem Set 3

1. Find a series solution around x0 = 0 for the following differential equation.

𝑦 ′′ − 𝑥𝑦 = 0
Solution:
In this case, 𝑝(𝑥) = 1; hence for this differential equation every point is an ordinary point.
Assume solution:
∞ ∞

𝑦 = ∑ 𝑎𝑛 (𝑥 − 0) = ∑ 𝑎𝑛 𝑥 𝑛
𝑛

𝑛=0 𝑛=0

Then,
∞

𝑦 = ∑ 𝑛 𝑎𝑛 𝑥 𝑛−1
′

𝑛=1
∞

𝑦 = ∑ 𝑛 (𝑛 − 1)𝑎𝑛 𝑥 𝑛−2
′′

𝑛=2

Step1: Plugging into the differential equation

155
 ∞ ∞
𝑛−2
∑ 𝑛 (𝑛 − 1)𝑎𝑛 𝑥 − 𝑥 ∑ 𝑎𝑛 𝑥 𝑛 = 0
𝑛=2 𝑛=0

Step 2: Get all the coefficients moved into the series.

∞ ∞
𝑛−2
∑ 𝑛 (𝑛 − 1)𝑎𝑛 𝑥 − ∑ 𝑎𝑛 𝑥 𝑛+1 = 0
𝑛=2 𝑛=0

Step 3: Shift the first series down by 2 and the second series up by 1 to get both of the series in
terms of xn
∞ ∞

∑(𝑛 + 2) (𝑛 + 1)𝑎𝑛+2 𝑥 − ∑ 𝑎𝑛−1 𝑥 𝑛 = 0

𝑛

𝑛=0 𝑛=1

Step 4: Get the two series starting at the same value of n. The only way to do that for this problem is to
strip out the n = 0 term
∞ ∞
(2)(1)𝑎2 𝑥 + ∑(𝑛 + 2) (𝑛 + 1)𝑎𝑛+2 𝑥 − ∑ 𝑎𝑛−1 𝑥 𝑛 = 0
0 𝑛

𝑛=1 𝑛=1
∞

2𝑎2 + ∑[(𝑛 + 2)(𝑛 + 1)𝑎𝑛+2 − 𝑎𝑛−1 ]𝑥 𝑛 = 0

𝑛=1

Step 5: Set all the coefficients equal to zero. The n = 0 coefficient is in front of the series and the n
= 1,2,3… are all in the series. So, setting coefficient equal to zero gives,
𝑛=0 2𝑎2 = 0
𝑛 = 1, 2, 3, … (𝑛 + 2)(𝑛 + 1)𝑎𝑛+2 − 𝑎𝑛−1 = 0
Step 6: Solving the first as well as the recurrence relation gives
𝑛=0 𝑎2 = 0
𝑎𝑛−1
𝑛 = 1, 2, 3, … 𝑎𝑛+2 −=
(𝑛 + 2)(𝑛 + 1)
Step 7: Start plugging in values of n
𝑎0
𝑛=1 𝑎3 =
(3)(2)
𝑎1
𝑛=2 𝑎4 =
(4)(3)
𝑎2
𝑛=3 𝑎5 = =0
(5)(4)
𝑎3 𝑎0
𝑛=4 𝑎6 = =
(6)(5) (6)(5)(3)(2)
𝑎4 𝑎1
𝑛=5 𝑎7 = =
(7)(6) (7)(6)(4)(3)

156
 𝑎5
𝑛=6 𝑎8 = =0
(8)(7)

𝑎0
𝑎3𝑘 = 𝑘 = 1, 2, 3, ⋯
(2)(3)(5)(6) ⋯ (3𝑘 − 1)(3𝑘)
𝑎1
𝑎3𝑘+1 = 𝑘 = 1, 2, 3, ⋯
(3)(4)(6)(7) ⋯ (3𝑘)(3𝑘 + 1)
𝑎3𝑘+2 = 0 𝑘 = 0, 1, 2, ⋯
Note: Every third coefficient is zero. The formulas here are somewhat unpleasant and not all that easy
to see the first time around. These formulas will not work for k = 0.

Step 8: Get the solution

𝑦(𝑥) = 𝑎0 + 𝑎1 𝑥 + 𝑎2 𝑥 2 + 𝑎3 𝑥 3 + 𝑎4 𝑥 4 + ⋯ + 𝑎3𝑘 𝑥 3𝑘 + 𝑎3𝑘+1 𝑥 3𝑘+1 + ⋯
𝑎0 𝑎1 𝑎0
= 𝑎0 + 𝑎1 𝑥 + 𝑥 2 + 𝑥 4 + ⋯ + 𝑥 3𝑘
6 12 (2)(3)(5)(6) ⋯ (3𝑘 − 1)(3𝑘)
𝑎1
+ 𝑥 3𝑘+1 + ⋯
(3)(4)(6)(7) ⋯ (3𝑘)(3𝑘 + 1)
Step 9: Collect up the terms that contain the same coefficient, factor the coefficient out and write the
results as a new series
∞ ∞
𝑥 3𝑘 𝑥 3𝑘+1
𝑦(𝑥) = 𝑎0 [1 + ∑ ] + 𝑎1 [𝑥 + ∑ ]
(2)(3)(5)(6) ⋯ (3𝑘 − 1)(3𝑘) (3)(4)(6)(7) ⋯ (3𝑘)(3𝑘 + 1)
𝑘=1 𝑘=1

Note: The series could not start at k = 0 since the general term doesn’t hold for k = 0

2. Find the first four terms in each portion of the series solution around x0 = -2 for the following
differential equation
𝑦 ′′ − 𝑥𝑦 = 0

Solution:
In this case, 𝑝(𝑥) = 1; hence for this differential equation every point is an ordinary point.
Assume solution:
∞ ∞

𝑦 = ∑ 𝑎𝑛 (𝑥 − (−2)) = ∑ 𝑎𝑛 (𝑥 + 2)𝑛
𝑛

𝑛=0 𝑛=0

Then,
∞

𝑦 = ∑ 𝑛 𝑎𝑛 (𝑥 + 2)𝑛−1
′

𝑛=1

157
 ∞

𝑦 = ∑ 𝑛 (𝑛 − 1)𝑎𝑛 (𝑥 + 2)𝑛−2
′′

𝑛=2

Step 1: Plugging into the differential equation

∞ ∞
𝑛−2
∑ 𝑛 (𝑛 − 1)𝑎𝑛 (𝑥 + 2) − 𝑥 ∑ 𝑎𝑛 (𝑥 + 2)𝑛 = 0
𝑛=2 𝑛=0

Step 2: Get all the coefficients moved into the series. There is a difference between this example and
the previous example. In this case we can’t just multiply the x into the second series since in
order to combine with the series it must be x + 2. Therefore we will first need to modify the
coefficient of the second series before multiplying it into the series.
∞ ∞
𝑛−2
∑ 𝑛 (𝑛 − 1)𝑎𝑛 (𝑥 + 2) − (𝑥 + 2 − 2) ∑ 𝑎𝑛 (𝑥 + 2)𝑛 = 0
𝑛=2 𝑛=0
∞ ∞ ∞
𝑛−2
∑ 𝑛 (𝑛 − 1)𝑎𝑛 (𝑥 + 2) − (𝑥 + 2) ∑ 𝑎𝑛 (𝑥 + 2) + 2 ∑ 𝑎𝑛 (𝑥 + 2)𝑛 = 0
𝑛

𝑛=2 𝑛=0 𝑛=0

∞ ∞ ∞
𝑛−2 𝑛+1
∑ 𝑛 (𝑛 − 1)𝑎𝑛 (𝑥 + 2) − ∑ 𝑎𝑛 (𝑥 + 2) + ∑ 2𝑎𝑛 (𝑥 + 2)𝑛 = 0
𝑛=2 𝑛=0 𝑛=0

Note: Now have three series to work with.

Step 3: Need to shift the first series down by 2 and the second series up by 1 to get common exponents
in all the series
∞ ∞ ∞

∑(𝑛 + 2) (𝑛 + 1)𝑎𝑛+2 (𝑥 + 2) − ∑ 𝑎𝑛−1 (𝑥 + 2) + ∑ 2𝑎𝑛 (𝑥 + 2)𝑛 = 0

𝑛 𝑛

𝑛=0 𝑛=1 𝑛=0

Step 4: Combine the series by stripping out the n = 0 terms from both the first and third series
∞ ∞ ∞

2𝑎2 + ∑(𝑛 + 2) (𝑛 + 1)𝑎𝑛+2 (𝑥 + 2) − ∑ 𝑎𝑛−1 (𝑥 + 2) + 2𝑎0 + ∑ 2𝑎𝑛 (𝑥 + 2)𝑛 = 0

𝑛 𝑛

𝑛=1 𝑛=1 𝑛=1

∞

2𝑎2 + 2𝑎0 + ∑[(𝑛 + 2)(𝑛 + 1)𝑎𝑛+2 − 𝑎𝑛−1 + 2𝑎𝑛 ](𝑥 + 2)𝑛 = 0

𝑛=1

Step 5: Set all the coefficients equal to zero.

𝑛=0 2𝑎2 + 2𝑎0 = 0
𝑛 = 1, 2, 3, … (𝑛 + 2)(𝑛 + 1)𝑎𝑛+2 − 𝑎𝑛−1 + 2𝑎𝑛 = 0

158
Step 6: Solve the first as well as the recurrence relation. In the first case there are two options, we can
solve for a2 or we can solve for a0. Out of habit I’ll solve for a0. In the recurrence relation we’ll
solve for the term with the largest subscript
𝑛=0 𝑎2 = −𝑎0
𝑎𝑛−1 − 2𝑎𝑛
𝑛 = 1, 2, 3, … 𝑎𝑛+2 =
(𝑛 + 2)(𝑛 + 1)
Note 1: This example we won’t be having every third term drop out as we did in the previous example.
Note 2: At this point we’ll also acknowledge that the instructions for this problem are different as well.
We aren’t going to get a general formula for the an’s this time so we’ll have to be satisfied with
just getting the first couple of terms for each portion of the solution. This is often the case for
series solutions. Getting general formulas for the an’s is the exception rather than the rule in
these kinds of problems.

Step 7: Start plugging in values of n. To get the first four terms we’ll just start plugging in terms until
we’ve got the required number of terms. Note that we will already be starting with an a0 and
an a1 from the first two terms of the solution so all we will need are three more terms with an a0
in them and three more terms with an a1 in them
𝑛=0 𝑎2 = −𝑎0
𝑎0 − 2𝑎1 𝑎0 𝑎1
𝑛=1 𝑎3 = = −
(3)(2) 6 3
𝑎1 − 2𝑎2 𝑎1 − 2(−𝑎0 ) 𝑎0 𝑎1
𝑛=2 𝑎4 = = = +
(4)(3) (4)(3) 6 12
𝑎2 − 2𝑎3 𝑎0 1 𝑎0 𝑎1 𝑎0 𝑎1
𝑛=3 𝑎5 = = − ( − )=− +
(5)(4) 20 10 6 3 15 30

Step 8: Get the solution

𝑦(𝑥) = 𝑎0 + 𝑎1 (𝑥 + 2) + 𝑎2 (𝑥 + 2)2 + 𝑎3 (𝑥 + 2)3 + 𝑎4 (𝑥 + 2)4 + 𝑎5 (𝑥 + 2)5 + ⋯
𝑎0 𝑎1 𝑎0 𝑎1
= 𝑎0 + 𝑎1 (𝑥 + 2) − 𝑎0 (𝑥 + 2)2 + ( − ) (𝑥 + 2)3 + ( + ) (𝑥 + 2)4
6 3 6 12
𝑎0 𝑎1
+ (− + ) (𝑥 + 2)5 + ⋯
15 30

Step 9: Collect up the terms that contain the same coefficient, factor the coefficient out and write the
results as a new series
1 1 1
𝑦(𝑥) = 𝑎0 {1 − (𝑥 + 2)2 + (𝑥 + 2)3 + (𝑥 + 2)4 − (𝑥 + 2)5 + ⋯ }
6 6 15
1 1 1
+ 𝑎1 {(𝑥 + 2) − (𝑥 + 2) + (𝑥 + 2)4 + (𝑥 + 2)5 + ⋯ }
3
3 12 30

159
Note: That’s the solution for this problem as far as we’re concerned. Notice that this solution looks
nothing like the solution to the previous example. It’s the same differential equation, but
changing x0 completely changed the solution.

3. Determine a series solution about x0 = 0 for the following initial value problem.
𝑦 ′′ − 2𝑥𝑦 ′ + 𝑦 = 0, 𝑦(0) = 1, 𝑦 ′ (0) = 1
Solution:
Assume solution:
∞

𝑦 = ∑ 𝑎𝑛 𝑥 𝑛
𝑛=0

Then,
∞

𝑦 = ∑ 𝑛 𝑎𝑛 𝑥 𝑛−1
′

𝑛=1
∞

𝑦 = ∑ 𝑛 (𝑛 − 1)𝑎𝑛 𝑥 𝑛−2
′′

𝑛=2

Step 1: Plugging into the differential equation

∞ ∞ ∞
𝑛−2 𝑛−1
∑ 𝑛 (𝑛 − 1)𝑎𝑛 𝑥 − 2𝑥 ∑ 𝑛𝑎𝑛 𝑥 + ∑ 𝑎𝑛 𝑥 𝑛 = 0
𝑛=2 𝑛=1 𝑛=0

Step 2: Get all the coefficients moved into the series.

∞ ∞ ∞
𝑛−2 𝑛
∑ 𝑛 (𝑛 − 1)𝑎𝑛 𝑥 − ∑ 2𝑛𝑎𝑛 𝑥 + ∑ 𝑎𝑛 𝑥 𝑛 = 0
𝑛=2 𝑛=1 𝑛=0

Step 3: Need to shift the first series down by 2 to get common exponents in all the series
∞ ∞ ∞

∑(𝑛 + 2) (𝑛 + 1)𝑎𝑛+2 𝑥 − ∑ 2𝑛𝑎𝑛 𝑥 + ∑ 𝑎𝑛 𝑥 𝑛 = 0

𝑛 𝑛

𝑛=0 𝑛=1 𝑛=0

Step 4: Combine the series by stripping out the n = 0 terms from both the first and third series
∞ ∞ ∞

2𝑎2 𝑥 + ∑(𝑛 + 2) (𝑛 + 1)𝑎𝑛+2 𝑥 − ∑ 2𝑛𝑎𝑛 𝑥 + 𝑎0 𝑥 + ∑ 𝑎𝑛 𝑥 𝑛 = 0

0 𝑛 𝑛 0

𝑛=1 𝑛=1 𝑛=1

∞ ∞

(𝑎0 + 2𝑎2 )𝑥 + ∑[(𝑛 + 2)(𝑛 + 1)𝑎𝑛+2 − 2𝑛𝑎𝑛 + 𝑎𝑛 ] 𝑥 𝑛 = 0 = 0𝑥 0 + ∑[0] 𝑥 𝑛

0

𝑛=1 𝑛=0

Step 5: Set all the coefficients equal to zero

𝑛=0 𝑎0 + 2𝑎2 = 0
𝑛 = 1, 2, 3, … (𝑛 + 2)(𝑛 + 1)𝑎𝑛+2 − 2𝑛𝑎𝑛 + 𝑎𝑛 = 0

160
Step 6: Solve the first as well as the recurrence relation.
𝑎0
𝑛=0 𝑎2 = −
2
(2𝑛 − 1)𝑎𝑛
𝑛 = 1, 2, 3, … 𝑎𝑛+2 =
(𝑛 + 2)(𝑛 + 1)
Step 7: Start plugging in values of n.
𝑎0
𝑛=0 𝑎2 = −
2
𝑎1
𝑛=1 𝑎3 =
6
3𝑎2 𝑎2 𝑎0
𝑛=2 𝑎4 = = =−
12 4 8
5𝑎3 𝑎3 𝑎1
𝑛=3 𝑎5 = = =
20 4 24
Note: Can choose any arbitrary constants for a0 and a1

Step 8: Get the solution

∞

𝑦(𝑥) = ∑ 𝑎𝑛 𝑥 𝑛 = 𝑎0 + 𝑎1 𝑥 + 𝑎2 𝑥 2 + 𝑎3 𝑥 3 + ⋯
𝑛=0
𝑎0 2 𝑎1 3 𝑎0 𝑎1
= 𝑎0 + 𝑎1 𝑥 + (− )𝑥 + 𝑥 + (− ) 𝑥 4 + 𝑥 5 + ⋯
2 6 8 24

Step 9: Collect up the terms that contain the same coefficient, factor the coefficient out and write the
results as a new series
𝑥2 𝑥4 𝑥3 𝑥5
𝑦(𝑥) = 𝑎0 [1 − − + ⋯ ] + 𝑎1 [ 𝑥 + + + ⋯]
2 8 6 24

Step 10: Applying the initial conditions gives values for a0 and a1
𝑦(0) = 1 ⇒ 𝑎0 = 1
𝑦′(0) = 1 ⇒ 𝑎1 = 1

Step 11: Write out the particular solution

𝑥2 𝑥4 𝑥3 𝑥5
𝑦(𝑥) = [1 − − + ⋯] + [𝑥 + + + ⋯]
2 8 6 24

161
Solutions About Singular Points
The power series method for solving linear differential equations with variable coefficients no longer
works when solving the differential equation about a singular point. It appears that some features of the
solutions of such equations of the most importance for applications are largely determined by their
behavior near their singular points. Frobenius method is usually used to solve the differential equation
about a regular singular point. This method does not always yield two infinite series solutions. When only
one solution is found, a certain formula can be used to get the second solution.

The two differential equations

(𝑎) 𝑦 ′′ + 𝑥𝑦 = 0 (𝑏) 𝑥𝑦 ′′ + 𝑦 = 0 (7)
are similar only in that they are both examples of simple linear second-order differential equations with
variable coefficients. For (7a), x = 0 is an ordinary point; hence, there is no problem in finding two distinct
power series solution centered at that point. In contrast, x = 0 is a singular point for (7b), finding two
infinite series solutions about that point becomes more difficult task.

For the homogeneous second-order linear differential equation

𝐴(𝑥)𝑦 ′′ + 𝐵(𝑥)𝑦 ′ + 𝐶(𝑥)𝑦 = 0 (8)
The singular points are simply points where A(x) = 0 if the functions A, B, and C are polynomials having
no common factors.

For example, x = 0 is the only singular point of the Bessel equation of order n,
𝑥 2 𝑦 ′′ + 𝑥𝑦 ′ + (𝑥 2 − 𝑛2 )𝑦 = 0
whereas the Legendre equation of order n,
(1 − 𝑥 2 )𝑦 ′′ − 2𝑥𝑦 ′ + 𝑛(𝑛 + 1)𝑦 = 0
has two singular points x = -1 and x = 1.

Note: Usually, only the case in which x = 0 is a singular point of Equation (7) is considered. A differential
equation having x = a as a singular point is easily transformed by the substitution t = x – a into
one having a corresponding singular point at 0.

Types of Singular Points

A differential equation having a singular point at 0 ordinarily will not have power series solutions of the
form

𝑦(𝑥) = ∑ 𝑐𝑛 𝑥 𝑛

162
So the straightforward method of power series fails in this case.

A singular point x0 of a linear differential equation

𝐴(𝑥)𝑦 ′′ + 𝐵(𝑥)𝑦 ′ + 𝐶(𝑥)𝑦 = 0
is further classified as either regular or irregular. The classification depends on the functions P and Q in
the standard form
𝑦 ′′ + 𝑃(𝑥)𝑦 ′ + 𝑄(𝑥)𝑦 = 0

Definition (Regular or Irregular Singular Points)

A singular point x0 is said to be a regular singular point of the differential equation (8) if the functions
𝑝(𝑥) = (𝑥 − 𝑥0 )𝑃(𝑥) and 𝑞(𝑥) = (𝑥 − 𝑥0 )2 𝑄(𝑥)
are both analytic at x0. A singular point that is not regular is said to be irregular singular point of the
equation.

Quick Visual Check (Regular or Irregular Singular Points)

If 𝑥 − 𝑥0 appears at most to the first power in the denominator of P(x) and at most to the second power
in the denominator of Q(x), then 𝑥 − 𝑥0 is a regular singular point.

Example:
Find the singular point(s) for the differential equation
(𝑥 2 − 4)2 𝑦 ′′ + 3(𝑥 − 2)𝑦 ′ + 5𝑦 = 0
Answer
Divide the equation with
(𝑥 2 − 4)2 = (𝑥 − 2)2 (𝑥 + 2)2
and reduce the coefficients to the lowest terms, produce
3 5
𝑃(𝑥) = and 𝑄(𝑥) =
(𝑥 − 2)(𝑥 + 2)2 (𝑥 − 2)2 (𝑥 + 2)2

Test P(x) and Q(x)

(i) For x = 2 to be a regular point, the factor x – 2 can appear at most to the first power in the
denominator of P(x) and at most to the second power in the denominator of Q(x). A check of
the denominators of P(x) and Q(x) shows that both these conditions are satisfied, so x _ 2 is a
regular singular point. Alternatively, the same conclusion is madeby noting that both rational
functions

163
 3 5
𝑝(𝑥) = (𝑥 − 2)𝑃(𝑥) = and 𝑞(𝑥) = (𝑥 − 2)2 𝑄(𝑥) =
(𝑥 + 2)2 (𝑥 + 2)2
are analytic at x -2.

(ii) Now since the factor x - (-2) = x + 2 appears to the second power in the denominator of P(x), we
can conclude immediately that x = -2 is an irregular singular point of the equation. This also
follows from the fact that
3
𝑝(𝑥) = (𝑥 + 2)𝑃(𝑥) =
(𝑥 − 2)(𝑥 + 2)
is not analytic at x = -2.

13.2 FROBENIUS METHOD

If x = x0 is a singular point of the differential equation (8), then there exists at least one solution of the
form
∞ ∞

𝑦(𝑥) = (𝑥 − 𝑥0 )𝑟 ∑ 𝑐𝑛 (𝑥 − 𝑥0 )𝑛 = ∑ 𝑐𝑛 (𝑥 − 𝑥0 )𝑛+𝑟
𝑛=0 𝑛=0
where the number r is a constant to be determined. The series will converge at least on some interval
0 < x – x0 < R.

An Introduction to The Method of Frobenius

Any differential equation of the form
𝑏(𝑥) ′ 𝑐(𝑥)
𝑦 ′′ + 𝑦 + 2 𝑦=0 (9)
𝑥 𝑥
where the function b(x) and c(x) are analytic at x = 0, has at least one solution that can be represented in
the form
∞ ∞

𝑦(𝑥) = 𝑥 ∑ 𝑎𝑛 𝑥 = ∑ 𝑎𝑛 𝑥 𝑛+𝑟 = 𝑥 𝑟 (𝑎0 + 𝑎1 𝑥 + 𝑎2 𝑥 2 + ⋯ )

𝑟 𝑛 (𝑎 ≠ 0) (10)
𝑛=0 𝑛=0

where the exponent r may be any real or complex number and r is chosen so that a0  0.

The equation also has a second solution such that these two solutions are linearly independent that may
be similar to (10) with different r and different coefficients or may contain a logarithmic term.

164
Solution Procedure:
(i) Write (8) in the following more convenient form
𝑥 2 𝑦 ′′ + 𝑥𝑏(𝑥)𝑦 ′ + 𝑐(𝑥)𝑦 = 0 (11)

(ii) Expand b(x) and c(x) in power series

∞

𝑏(𝑥) = ∑ 𝑏𝑛 𝑥𝑛 = 𝑏0 + 𝑏1 𝑥 + 𝑏2 𝑥 2 + ⋯
𝑛=0
∞

𝑐(𝑥) = ∑ 𝑐𝑛 𝑥𝑛 = 𝑐0 + 𝑐1 𝑥 + 𝑐2 𝑥 2 + ⋯
𝑛=0

(iii) Differentiate (10) term by term, finding

∞

𝑦 ′ (𝑥) = ∑(𝑛 + 𝑟)𝑎𝑛 𝑥 𝑛+𝑟−1 = 𝑥 𝑟−1 [𝑟𝑎0 + (𝑟 + 1)𝑎1 𝑥 + ⋯ ]

𝑛=0
∞
′′ (𝑥)
𝑦 = ∑(𝑛 + 𝑟)(𝑛 + 𝑟 − 1)𝑎𝑛 𝑥 𝑛+𝑟−2 = 𝑥 𝑟−2 [𝑟(𝑟 − 1)𝑎0 + (𝑟 + 1)𝑟𝑎1 𝑥 + ⋯ ]
𝑛=0

(iv) Insert all these series into (11)

∞ ∞ ∞ ∞ ∞
2 𝑛+𝑟−2 𝑛 𝑛+𝑟−1
𝑥 ∑(𝑛 + 𝑟)(𝑛 + 𝑟 − 1)𝑎𝑛 𝑥 + 𝑥 ∑ 𝑏𝑛 𝑥 ∑(𝑛 + 𝑟)𝑎𝑛 𝑥 + ∑ 𝑐𝑛 𝑥 ∑ 𝑎𝑛 𝑥 𝑛+𝑟 = 0
𝑛

𝑛=0 𝑛=0 𝑛=0 𝑛=0 𝑛=0

∞ ∞ ∞

∑(𝑛 + 𝑟)(𝑛 + 𝑟 − 1)𝑎𝑛 𝑥 𝑛+𝑟 + ∑(𝑛 + 𝑟)𝑎𝑛 𝑏𝑛 𝑥 2𝑛+𝑟 + ∑ 𝑎𝑛 𝑐𝑛 𝑥 2𝑛+𝑟 = 0

𝑛=0 𝑛=0 𝑚=0
∞ ∞ ∞
𝑟 𝑛 𝑟 2𝑛
𝑥 ∑(𝑛 + 𝑟)(𝑛 + 𝑟 − 1)𝑎𝑛 𝑥 + 𝑥 ∑(𝑛 + 𝑟)𝑎𝑛 𝑏𝑛 𝑥 + 𝑥 ∑ 𝑎𝑛 𝑐𝑛 𝑥 2𝑛 = 0
𝑟

𝑛=0 𝑛=0 𝑛=0

When n = 0
𝑥 𝑟 [𝑟(𝑟 − 1)𝑎0 + 𝑟𝑎0 𝑏0 + 𝑎0 𝑐0 ] = 0
𝑥𝑟 ≠ 0 ⇒ [𝑟(𝑟 − 1)𝑎0 + 𝑟𝑏0 + 𝑐0 ]𝑎0 = 0
𝑎0 ≠ 0 ⇒ 𝑟(𝑟 − 1)𝑎0 + 𝑟𝑏0 + 𝑐0 = 0
𝑟 2 + (𝑏0 − 1)𝑟 + 𝑐0 = 0 (12)

165
 Equation (12) is called an indicial equation of the differential equation (11), and its two roots
(possibly equal) are exponents of the differential equation at the regular singular point x = 0.
Equation (12) also can be written as
𝑟(𝑟 − 1) + 𝑏0 𝑟 + 𝑐0 = 0 (13)

(v) Find the indicial roots r1, r2 for the indicial equation
If (9) is to be a solution of the differential equation in (10), the exponent r must be one of the roots
r1 and r2 of the indicial equation in (13). If r1  r2, it follows that there are two possible Frobenius
series solutions, whereas if r1 = r2 there is only one possible Frobenius series solution; the second
solution cannot be a Frobenius series.
The exponents r1 and r2 in the possible Frobenius series solutions are determined (using the
indicial equation) by the values b0 = b(0) and c0 = c(0). In practice, particularly when the
coefficients in the differential equation in the original form in (8) are polynomials, the simplest
way of finding b0 and c0 is often to write the equation in the form
𝑏0 + 𝑏1 𝑥 + 𝑏2 𝑥 2 + ⋯ ′ 𝑐0 + 𝑐1 𝑥 + 𝑐2 𝑥 2 + ⋯
𝑦 ′′ + 𝑦 + 𝑦=0
𝑥 𝑥2
Then inspection of the series that appear in the two numerators reveals the constants b0
and c0.

Example
Find the exponents in the possible Frobenius series solutions of the equation
2𝑥 2 (1 + 𝑥)𝑦 ′′ + 3𝑥(1 + 𝑥)3 𝑦 ′ − (1 − 𝑥 2 )𝑦 = 0
Answer:
Divide each term by 2𝑥 2 (1 + 𝑥) to recast the differential equation in the form
3 1
′′ 2
(1 + 2𝑥 + 𝑥 2 ) ′ −
2
(1 − 𝑥)
𝑦 + 𝑦 + 𝑦=0
𝑥 𝑥2
3 1
and get 𝑏0 = 2 and 𝑐0 = − 2 . Hence the indicial equation is
3 1 1 1 1
𝑟(𝑟 − 1) + 𝑟 − = 𝑟 2 + 𝑟 − = (𝑟 + 1) (𝑟 − ) = 0
2 2 2 2 2
1
with roots 𝑟1 = 2 and 𝑟2 = −1. The two possible Frobenius series solutions are then in the forms
∞ ∞
1⁄2 𝑛 −1
𝑦1 (𝑥) = 𝑥 ∑ 𝑎𝑛 𝑥 and 𝑦2 (𝑥) = 𝑥 ∑ 𝑎𝑛 𝑥 𝑛
𝑛=0 𝑛=0

166
Review Materials
1. Taylor Series
If f(x) is an infinitely differentiable function then the Taylor Series of f(x) about x = x0 is,
∞
𝑓 (𝑛) (𝑥0 )
𝑓(𝑥) = ∑ (𝑥 − 𝑥0 )𝑛
𝑛!
𝑛=0

Recall that
𝑓 (0) (𝑥) = 𝑓(𝑥)
𝑓 (𝑛) (𝑥) = 𝑛𝑡ℎ derivative of 𝑓(𝑥)

Example
Determine the Taylor series for 𝑓(𝑥) = 𝑒 𝑥 about x = 0

Solution
This is probably one of the easiest functions to find the Taylor series for. Recall that,
𝑓 (𝑛) (𝑥) = 𝑒 𝑥 𝑛 = 0, 1, 2, ⋯
and get
𝑓 (𝑛) (0) = 1 𝑛 = 0, 1, 2, ⋯
The Taylor series for this example is then
∞
𝑥
𝑥𝑛
𝑒 =∑
𝑛!
𝑛=0

Example
Determine the Taylor series for 𝑓(𝑥) = 𝑒 𝑥 about x = -4

Solution
This problem is virtually identical to the previous problem. In this case we just need to notice that,
𝑓 (𝑛) (−4) = 𝑒 −4 𝑛 = 0, 1, 2, ⋯
The Taylor series for this example is then,
∞
𝑥
𝑒 −4
𝑒 =∑ (𝑥 + 4)𝑛
𝑛!
𝑛=0

167
Definition
A function, f(x), is called analytic at x = a if the Taylor series for f(x) about x = a has a positive radius
of convergence and converges to f(x).

168
ENGINEERING APPLICAIONS OF
DIFFERENTIAL EQUATION
WEEK 14: ENGINEERING APPLICATIONS OF DIFFERENTIAL EQUATION
14.1 A LIQUID SYSTEM

Figure 1 shows a tank of liquid. The tank has a constant cross-sectional area A. The liquid can flow out of
the tank through a valve near the base. As it does, the height or head, h, of liquid in the tank will reduce.
Let q be the rate at which liquid flows out of the tank. Under certain conditions the rate outflow is
proportional to the head, so that q = kh where k is a constant of proportionality. Situation like this arise
frequently in chemical engineering industry. Obtain a differential equation for this system or equivalently,
come out with the mathematical model for the physical system.

Area, A

q = kh
q
Figure 1. Modelling a liquid system

The expression for the volume V of liquid in the tank at any time.
𝑉 =𝐴×ℎ

The volume of liquid in the tank changes because liquid is flowing out. Hence,
the rate at which this volume changes = rate of flow in – rate of flow out
This is the law of conservation of mass. The rate of change of volume is
𝑑𝑉
𝑑𝑡
There is no flow into the tank and liquid flows out at a rate q. Hence
𝑑𝑉
= −𝑞
𝑑𝑡
𝑑ℎ
But, 𝑉 = 𝐴ℎ and A is constant, so the rate of change of volume is 𝐴 𝑑𝑡

169
Therefore
𝑑𝑉 𝑑ℎ
=𝐴 = −𝑞
𝑑𝑡 𝑑𝑡
Also, 𝑞 = 𝑘ℎ, so
𝑑ℎ
𝐴 = −𝑘ℎ
𝑑𝑡
or
𝐴ℎ′ + 𝑘ℎ = 0
This is a first order differential equation with dependent variable h and independent variable t. It is linear
and has constant coefficients. The unknown function to seek (the solution) is h(t). Solve the equation to
find the head, h, at any time, t.

14.2 A MIXTURE PROBLEM

A tank contains 40 gallons of a solution composed of 90 percent water and 10 percent alcohol. A second
solution containing 50 percent water and 50 percent alcohol is added to the tank at a rate of 4
gallon/minute. As the second solution being added, the tank is being drained at the rate of 4
gallon/minute, as shown in Figure 2. Assuming the solution in the tank is stirred constantly, how much
alcohol is in the tank after t minutes, how much alcohol is in the tank after 10 minutes?

Figure 2. Modelling of mixture

Let y(t) = number of gallons of alcohol in the tank at any time t.

This means that the percentage of alcohol in the tank at any time is y/40.
Since 4 gallons of solution is being drained every minute, the rate of change of y is given by
𝑑𝑦
= alcohol draining − alcohol entering
𝑑𝑡
or
𝑑𝑦 y
= −4 ( ) + 2
𝑑𝑡 40

170
where 2 represents the 2 gallons of alcohol entering each minute in the 50% solution.
In standard form,
1
𝑦′ + 𝑦= 2
10
is a first order linear differential equation. The unknown function to seek (the solution) is y(t). Solve the
equation to find the number of gallons of alcohol, y, at any time, t.

Solution
1
To solve the linear equation, let 𝑝(𝑥) =
10

Obtain
1 𝑡
∫ 𝑝(𝑡)𝑑𝑡 = ∫ 𝑑𝑡 =
10 10
Thus, the integrating factor is
𝑡⁄
𝑢(𝑡) = 𝑒 ∫ 𝑝(𝑡)𝑑𝑡 = 𝑒 10

and the general solution is

𝑡 𝑡⁄
𝑦 = 𝑒 − ⁄10 ∫ 2𝑒 10 𝑑𝑡

𝑡 𝑡⁄
= 𝑒 − ⁄10 (20𝑒 10 + 𝐶)
𝑡
= 20 + 𝐶𝑒 − ⁄10
Initial conditions, since y = 4 when t = 0,
4 = 20 + 𝐶 ⇒ 𝐶 = −16
The particular solution is
𝑡
𝑦 = 20 − 16𝑒 − ⁄10
Finally, when t = 10, the amount of alcohol in the tank is
10⁄
𝑦 = 20 − 16𝑒 − 10 = 20 − 16𝑒 −1 ≈ 14.11 gallons

171
14.3 AN LCR CIRCUIT

Figure 3. Modelling an LCR circuit

Figure 3 shows an LCR circuit. This is a circuit comprising an inductor of inductance L, a capacitor of
capacitance C, and a resistor of resistance R placed in a series. When a constant voltage source, V, is
applied, it can be shown that the current, i, through the circuit satisfies the differential equation
𝑑2 𝑖 𝑑𝑖 1
𝐿 2+𝑅 + 𝑖 =0
𝑑𝑡 𝑑𝑡 𝐶
or
1
𝐿 𝑖 ′′ + 𝑅 𝑖 ′ + 𝑖=0
𝐶
This equation can be derived using Kirchhoff’s voltage law, the individual laws for each component.
Because L, R, and C are constants, this is a constant coefficient equation. It is linear and second order.
The unknown function to seek (the solution) is i(t). Solve the equation to find the current in the circuit, i,
at any time, t.

14.4 VIBRATING SPRINGS

(b)

(a)

Figure 4. Modelling of vibrating spring (a) vertical (b) horizontal

172
Consider the motion of an object with mass m at the end of the spring that is either vertical or horizontal
on a level surface. Using Hooke’s Law, which says that if the spring is stretched (or compressed) x units
from its natural length, then it exerts a force that is proportional to x
restoring force = −𝑘𝑥
where k is a positive constant (called the spring constant). If any external resisting forces (due to air
resistance or friction) are ignored, then by Newton’s Second Law, F = ma,
𝑑2 𝑥
𝑚 = −𝑘𝑥
𝑑𝑡 2
or
𝑑2 𝑥
𝑚 + 𝑘𝑥 = 0 (1)
𝑑𝑡 2
This is a second order linear differential equation.

Solution:
The auxiliary equation is
𝑚𝑟 2 + 𝑘 = 0
with roots

𝑘
𝑟 = ±𝜔𝑖 where 𝜔=√
𝑚

Thus, the general solution is

𝑥(𝑡) = 𝑐1 cos 𝜔𝑡 + 𝑐2 sin 𝜔𝑡 = 𝐴 cos(𝜔𝑡 + 𝛿)
where

𝑘
𝜔=√ (frequency)
𝑚

𝐴 = √𝑐12 + 𝑐22 (amplitude)

𝑐1 𝑐2
cos 𝛿 = sin 𝛿 = − (phase angle)
𝐴 𝐴

This type of motion is called simple harmonic motion.

Exercise 1

173
A spring with mass of 2kg has natural length 0.5 m. A force of 25.6 N is required to maintain it stretched
to a length of 0.7 m. If the spring is stretched to a length of 0.7 m and then released with initial velocity 0,
find the position of the mass at any time t.

Solution
From Hooke’s Law, the force required to stretch the spring is
𝑘(0.2) = 25.6
25.6
𝑘= = 128
0.2
Using this value of the spring constant k, together with m = 2 in Equation 1
𝑑2 𝑥
2 + 128𝑥 = 0
𝑑𝑡 2
The solution of this equation is
𝑥(𝑡) = 𝑐1 cos 8𝑡 + 𝑐2 sin 8𝑡 (2)
The initial condition is given as 𝑥(0) = 0.2. Hence from Equation (2), 𝑥(0) = 𝑐1 = 0.2.
Differentiating Equation (2)
𝑥′(𝑡) = −8𝑐1 cos 8𝑡 + 8𝑐2 sin 8𝑡
Since the initial velocity is given as 𝑥′(0) = 0, 𝑐2 = 0
So, the solution is
𝑥(𝑡) = 0.2 cos 8𝑡

14.5 DAMPED VIBRATIONS

Consider the motion of a spring that is subject to a frictional force (in the case of horizontal spring) or a
damping force (in the case where a vertical spring moves through a fluid) An example is the damping
force supplied by a shock absorber in a car or a bicycle.

Figure 5. Modelling of vertical spring in a fluid

174
Assume that the damping force is proportional to the velocity of the mass and acts in the direction
opposite to the motion. Thus,
𝑑𝑥
damping force = −𝑐
𝑑𝑡
where c is a positive constant, called the damping constant. Thus, Newton’s Second Law gives
𝑑2 𝑥 𝑑𝑥
𝑚 2
= restoring force + damping force = −𝑘𝑥 − 𝑐
𝑑𝑡 𝑑𝑡
or
𝑑2 𝑥 𝑑𝑥
𝑚 2
+𝑐 + 𝑘𝑥 = 0 (3)
𝑑𝑡 𝑑𝑡
This is a second order linear differential equation.

Solution:
The auxiliary equation is
𝑚𝑟 2 + 𝑐𝑟 + 𝑘 = 0
with roots

−𝑐 + √𝑐 2 − 4𝑚𝑘 −𝑐 − √𝑐 2 − 4𝑚𝑘
𝑟1 = 𝑟2 = (4)
2𝑚 2𝑚

Case I: 𝑐 2 − 4𝑚𝑘 > 0 (overdamping)

The roots, r1 and r2, are distinct real roots and
𝑥(𝑡) = 𝑐1 𝑒 𝑟1 𝑡 + 𝑐2 𝑒 𝑟2 𝑡

Since c, m, and k are all positive, then √𝑐 2 − 4𝑚𝑘 < 0, and the roots r1 and r2 given by Equations (3) must
be both negative. This shows that 𝑥 → 0 as 𝑡 → ∞. Typical graphs of x as a function of t are shown in
Figure 6. Notice oscillations do not occur. This is because 𝑐 2 > 4𝑚𝑘 means that there is a strong
damping force (high viscosity oil or grease) compared with a weak spring or small mass.

175
Figure 6. Typical graphs for overdamping case

Case II: 𝑐 2 − 4𝑚𝑘 = 0 (critical damping)

This case corresponds to equal roots
𝑐
𝑟1 = 𝑟2 = −
2𝑚
The solution is given by
𝑐
−( )𝑡
𝑥 = (𝑐1 + 𝑐2 𝑡)𝑒 2𝑚

and a typical graph is shown in Figure 7. It is similar to Case I, but the damping is just sufficient to suppress
vibrations. Any decrease in the viscosity of fluid leads to the vibrations of the following case.

Figure 7. Graph for critical damping case

Case III: 𝑐 2 − 4𝑚𝑘 < 0 (underdamping)

Here the roots are complex
𝑐
𝑟1,2 = − ± 𝜔𝑖
2𝑚
where

√4𝑚𝑘 − 𝑐 2
𝜔=
2𝑚
The solution is given by

176
 𝑐
−( )𝑡
𝑥=𝑒 2𝑚 (𝑐1 cos 𝜔𝑡 + 𝑐2 sin 𝜔𝑡)
𝑐 𝑐
−( )𝑡 −( )𝑡
There are oscillations that are damped by the factor 𝑒 2𝑚 . Since c > 0 and m > 0, then 𝑒 2𝑚 < 0 so
𝑐
−( )𝑡
𝑒 2𝑚 → 0 as 𝑡 → ∞ . This implies that 𝑥 → 0 as 𝑡 → ∞ ; that is, the motion decays to 0 as time
increases. A typical graph is shown in Figure 8

Figure 8. Graph for underdamping case

Exercise 2
Suppose that the string in Exercise 1 is immersed in a fluid with damping constant c = 40. Find the position
of the mass at any time t if it starts from the equilibrium position and is given a push to start it with an
initial velocity of 0.6 m/s.

Solution
The mass is m = 2 and the spring constant is k = 128, so the differential equation becomes
𝑑2 𝑥 𝑑𝑥
2 2
+ 40 + 128𝑥 = 0
𝑑𝑡 𝑑𝑡
or
𝑑2 𝑥 𝑑𝑥
+ 20 + 64𝑥 = 0
𝑑𝑡 2 𝑑𝑡
The auxiliary equation is
𝑟 2 + 20𝑟 + 64 = (𝑟 + 4)(𝑟 + 16) = 0
with roots -4 and -16, so the motion is overdamped and the solution is
𝑥(𝑡) = 𝑐1 𝑒 −4𝑡 + 𝑐2 𝑒 −16𝑡
Given 𝑥(0) = 0, so 𝑐1 + 𝑐2 = 0. Differentiating,
𝑥 ′ (𝑡) = −4𝑐1 𝑒 −4𝑡 − 16𝑐2 𝑒 −16𝑡
So

177
 𝑥 ′ (0) = −4𝑐1 − 16𝑐2 = 0.6
Since 𝑐1 = −𝑐2, this gives 12𝑐1 = 0.6 or 𝑐1 = 0.05. Therefore
𝑥(𝑡) = 0.05(𝑒 −4𝑡 − 𝑒 −16𝑡 )

14.6 FORCED VIBRATIONS

Suppose that, in addition to the restoring force and the damping force, the motion of the spring is
affected by an external force F(t). The Newton’s Second Law gives

𝑑2 𝑥 𝑑𝑥
𝑚 2
= restoring force + damping force + external force = −𝑘𝑥 − 𝑐 + 𝐹(𝑡)
𝑑𝑡 𝑑𝑡

Thus, instead of the homogeneous Equation (3), the motion of the spring is now governed by the
following nonhomogeneous differential equation.
𝑑2 𝑥 𝑑𝑥
𝑚 2
+𝑐 + 𝑘𝑥 = 𝐹(𝑡) (5)
𝑑𝑡 𝑑𝑡
The motion of the spring can be determined by the methods such as the method of undetermined
coefficients, the method of variation parameters, etc.
A commonly occurring type of external force is a periodic force function
𝐹(𝑡) = 𝐹0 cos 𝜔0 𝑡
𝑘
where 𝜔0 ≠ 𝜔 = √𝑚

In this case, and in the absence of a damping force (c = 0), using method of undetermined coefficients
show that
𝐹0
𝑥 = 𝑐1 cos 𝜔𝑡 + 𝑐2 sin 𝜔𝑡 + cos 𝜔0 𝑡 (6)
𝑚(𝜔 2 − 𝜔02 )
If 𝜔0 = 𝜔, then the applied frequency reinforces the natural frequency and the result is vibrations of large
amplitude. This is the phenomenon of resonance.

178
14.7 ELECTRIC CIRCUIT

Figure 9. Modelling of electric circuits

Analyze the electric circuit shown in Figure 9. It contains an electromotive force E (supplied by a battery
or generator), a resistor R, an inductor L, and a capacitor C, in series. If the charge on the capacitor at
time t is Q = Q(t), then the current, I, is the rate of change of Q with respect to t,
𝑑𝑄
𝐼=
𝑑𝑡
It is known from physics that the voltage drops across the resistor, inductor, and capacitor are

𝑑𝐼 𝑄
𝑅𝐼 𝐿
𝑑𝑡 𝐶
respectively. Kirchhoff’s voltage law says that the sum of these voltage drops is equal to the supplied
voltage
𝑑𝐼 𝑄
𝐿 + 𝑅𝐼 + = 𝐸(𝑡) (7)
𝑑𝑡 𝐶
𝑑𝑄
Since 𝐼 = , Equation 7 becomes
𝑑𝑡

𝑑2 𝑄 𝑑𝑄 1
𝐿 2
+𝑅 + 𝑄 = 𝐸(𝑡) (8)
𝑑𝑡 𝑑𝑡 𝐶
which is a second order linear differential equation with constants coefficients. If the charge Q0 and the
current I0 are known at time 0, then the initial conditions are

𝑄(0) = 𝑄0 𝑄 ′ = 𝐼(0) = 𝐼0

The initial value problem can be solved by such as the method of undetermined coefficients, the
method of variation parameters, etc.

The differential equation for the current can be obtained by differentiating Equation (*) with respect to
𝑑𝑄
t and using 𝐼 = 𝑑𝑡

179
 𝑑2 𝐼 𝑑𝐼 1
𝐿 2
+ 𝑅 + 𝐼 = 𝐸′(𝑡)
𝑑𝑡 𝑑𝑡 𝐶

Exercise 3

Find the charge and current at time t in the circuit of Figure 9 if R = 40 , L = 1 H, C = 16 × 10-4 F, E(t) =
100 cos 10t, and the initial charge and current are both 0.

Solution

With the given values of L, R, C, and E(t), Equation (8) becomes

𝑑2 𝑄 𝑑𝑄
2
+ 40 + 625𝑄 = 100 cos 10𝑡 (9)
𝑑𝑡 𝑑𝑡

The auxiliary equation is 𝑟 2 + 40𝑟 + 625 = 0 with roots

−40 ± √−900
𝑟= = −20 ± 15𝑖
2
The solution of the complementary equation is

𝑄𝑐 (𝑡) = 𝑒 −20𝑡 (𝑐1 cos 15𝑡 + 𝑐2 sin 15𝑡)

For the method of undetermined coefficients, the particular solution

𝑄𝑝 (𝑡) = 𝐴 cos 10𝑡 + 𝐵 sin 10𝑡

Then

𝑄𝑝′ (𝑡) = −10𝐴 sin 10𝑡 + 10𝐵 cos 10𝑡

𝑄𝑝′′ (𝑡) = −100𝐴 cos 10𝑡 − 100𝐵 sin 10𝑡

Substituting into Equation (9)

(525𝐴 + 400𝐵) cos 10𝑡 + (−400𝐴 + 525𝐵) sin 10𝑡 = 100 cos 10𝑡

Equating coefficients,

525𝐴 + 400𝐵 = 100 − 400𝐴 + 525𝐵 = 0

or

21𝐴 + 16𝐵 = 4 − 16𝐴 + 21𝐵 = 0

The solution is

180
 84 64
𝐴= 𝐵=
697 697
and the particular solution is

1
𝑄𝑝 (𝑡) = (84 cos 10𝑡 + 64 sin 10𝑡)
697
The general solution is

4
𝑄(𝑡) = 𝑄𝑐 (𝑡) + 𝑄𝑝 (𝑡) = 𝑒 −20𝑡 (𝑐1 cos 15𝑡 + 𝑐2 sin 15𝑡) + (21 cos 10𝑡 + 16 sin 10𝑡)
697
Imposing the initial condition Q(0) = 0

84 84
𝑄(0) = 𝑐1 + =0 or 𝑐1 = −
697 697
To impose the other initial condition, first differentiate to find the current

𝑑𝑄 40
𝐼= = 𝑒 −20𝑡 [(−20𝑐1 + 15𝑐2 ) cos 15𝑡 + (−15𝑐1 − 20𝑐2 ) sin 15𝑡] + (−21 sin 10𝑡 + 16 cos 10𝑡)
𝑑𝑡 697

640 464
𝐼(0) = −20𝑐1 + 15𝑐2 + =0 or 𝑐2 = −
697 2091
Thus, the formula for charge is

4 𝑒 −20𝑡
𝑄(𝑡) = [ (−63 cos 15𝑡 − 116 sin 15𝑡) + (21 cos 10𝑡 + 16 sin 10𝑡)]
697 3

and the expression for the current is

1
𝐼(𝑡) = [𝑒 −20𝑡 (−1920 cos 15𝑡 + 13.060 sin 15𝑡) + 120(−21 sin 10𝑡 + 16 cos 10𝑡)]
2091

Note 1:

The solution for Q(t) consists of two parts. Since 𝑒 −20𝑡 → 0 as 𝑡 → ∞ and both cos 15t and sin 15t are
bounded functions

4
𝑄𝑐 (𝑡) = 𝑒 −20𝑡 (−63 cos 15𝑡 − 116 sin 15𝑡) → 0 as 𝑡 → ∞
2091
So, for large values of t,

4
𝑄(𝑡) ≈ 𝑄𝑝 (𝑡) = (21 cos 10𝑡 + 16 sin 10𝑡)
697

181
and, for this reason, Qp(t) is called the steady-state solution. Figure 10 shows how the graph of the steady
state solution compares with the graph of Q.

Figure 10. Graphs of Q(t) and Qp(t)

Note 2:

Physical situations versus mathematically defined for the spring system and electric circuit are given in
Table 1.

Table 1. Physical entities versus mathematical entities

Spring System Electric Circuit

x displacement Q charge
𝑑𝑥 𝑑𝑄
velocity 𝐼= current
𝑑𝑡 𝑑𝑡
m mass L inductance
c damping constant R resistance
k spring constant 1
elastance
𝐶
F(t) external force E(t) electromotive force

182
APPENDIX 11.1 MATHEMATICAL MODELING AND ENGINEERING PROBLEM SOLVING

Mathematical model plays an important role in engineering problem solving. The engineering problem
solving process in parachute design is illustrated as follows:

Problem: Impact velocity upon landing is an important

parameter when designing a parachute. Develop a mathematical
model to solve the velocity for a falling parachutist.

Figure 11.1: Problem solving process in parachute design

To develop the mathematical model for calculating the velocity of a parachute, we can use the
knowledge from the existing theory/physical law (e.g. Newton’s Law) or understand the problem by
empirical means (e.g. by observation and experiment). In this case, Newton’s 2nd Law is applied, where
∑ 𝐹(𝑡) = 𝑚𝑎(𝑡). A free-body diagram is drawn as below. Assume that the drag force due to air resistance
is proportional to the falling velocity, the downward motion is positive, and 𝑔 = 9.81𝑚𝑠 −2. Then, we are
able to develop the mathematical model.

Figure 11.2: Free-body diagram of the falling parachutist

183
 The mathematical modelling of the falling parachutist’s velocity:

𝑑𝑣(𝑡)
∑ 𝐹(𝑡) = 𝑚𝑎(𝑡) = 𝑚 .
𝑑𝑡
𝑑𝑣(𝑡)
𝐹𝑔𝑟𝑎𝑣𝑖𝑡𝑦 + 𝐹𝑑𝑟𝑎𝑔 = 𝑚
𝑑𝑡
𝑑𝑣(𝑡) 𝐹𝑔𝑟𝑎𝑣𝑖𝑡𝑦 + 𝐹𝑑𝑟𝑎𝑔 𝑐𝑣(𝑡)
𝑅𝑎𝑡𝑒 𝑜𝑓 𝑣𝑒𝑙𝑜𝑐𝑖𝑡𝑦 𝑐ℎ𝑎𝑛𝑔𝑒 = = =𝑔−
𝑑𝑡 𝑚 𝑚

Rearrange it, we found that the mathematical model is in the form of first order linear
nonhomogeneous Ordinary Differential Equation (ODE):

1st order ODE format: 𝑎1 (𝑡)𝑣 ′ + 𝑎0 (𝑡)𝑣 = 𝑔(𝑡)

𝑑𝑣(𝑡) 𝑐
Parachutist problem: 𝑑𝑡
+ 𝑚 (𝑣(𝑡)) = 𝑔
Where,
First parameter, 𝑎1 (𝑡) = 1,
𝑐
Second parameter, 𝑎0 (𝑡) = 𝑚,
Forcing function, 𝑔(𝑡) = 𝑔,
Independent variable = time, (𝑡)
Dependent variable = velocity, 𝑣(𝑡)
𝑑𝑣
First derivative, ′ = 𝑑𝑡
.

The classification of the order, linear vs nonlinear, homogeneous vs nonhomogeneous will be covered
later.

In general, a mathematical model can be broadly defined as a formulation or an equation that

expresses the essential features of a physical system or process in mathematical terms. It can be
represented as a functional relationship of the following form and explained in Table 12.1.

𝐷𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒 = 𝑓(𝑖𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠, 𝑝𝑎𝑟𝑎𝑚𝑒𝑡𝑒𝑟𝑠, 𝑓𝑜𝑟𝑐𝑖𝑛𝑔 𝑓𝑢𝑛𝑐𝑡𝑖𝑜𝑛𝑠)

184
 Table 12.1: Elements of mathematical model and its example of falling parachutist problem.

Elements of mathematical model Example: Falling parachutist problem

(i) Dependent variable
-The characteristic that you are looking for, The falling velocity, 𝑣(𝑡) is the dependent
which reflects the behavior of the system variable that we are looking for.

(ii) Independent variable

-indicates the dimension of the examined Time, (𝑡) is the independent variable or the
system, such as time, space dimension of the problem that we are working.

The parameters of this problem are the properties

(iii) Parameters of the system such as the mass of parachutist &
-reflective of the system’s properties the parachute, (𝑚) and drag coefficient of air
resistance, (𝑐).
(iv) Forcing functions The acceleration due to gravity, (𝑔) is the forcing
-external influences acting upon the system function in this problem.
𝑑𝑣(𝑡)
We need to solve the mathematical model (which is in differential equation such as 𝑑𝑡
+
𝑐
𝑣(𝑡) = 𝑔) to obtain the velocity function, 𝑣(𝑡). To do this, two main problem-solving tools can be
𝑚
implemented such as the analytical method and the numerical method. In this study, we will learn how
to implement the analytical method (i.e. calculus & ODE) to solve the differential equation. The desired
𝑔𝑚
solution obtained by analytical method-1st order ODE is given as follows: 𝑣(𝑡) = (1 − 𝑒 −(𝑐⁄𝑚)𝑡 ).
𝑐

With this solution, it helps us to gain intuition about what to expect on the behaviour of the
examined system. For example, what are the suitable mass and damping to reduce the impact velocity
of the parachutist to the ground? This approach is known as the result interpretation and analysis. With
an accurate mathematical model, we can predict the performance of the design without testing with the
real subject. Avoiding the unnecessary physical cycles of ‘modify-and-test’ would save time and money.

In addition, numerical method is used to solve a complicated mathematical model that can’t be
solved by using analytical method. The desired solution obtained by numerical method- Euler method is
𝑐
given as follows: 𝑣(𝑡𝑖+1 ) = 𝑣(𝑡𝑖 ) + [𝑔 − 𝑚 𝑣(𝑡𝑖 )]( 𝑡𝑖+1 − 𝑡𝑖 ). This is for your extra information and you
will learn the numerical method in advanced mathematic class.

185
APPENDIX 11.2 CONSERVATION LAWS AND ENGINEERING

When we employed the Newton’s law to develop a force balance equation for the falling parachutist, i.e.
𝑑𝑣(𝑡) 𝑚𝑔−𝑐𝑣(𝑡)
𝑑𝑡
= 𝑚
, we can eventually boil down it to a simple equation:

𝑑𝑣(𝑡)
(i) Transient problem: [Change=increase-decrease] or [ 𝑑𝑡
≠ 0]

Although simple, it embodies one of the most fundamental ways in which conservation laws are
used in engineering – that is, to predict changes with respect to time. In this case, it is recognized as time-
variant (or transient) problem. (For example, the falling velocity of the parachutist with 𝑚 = 68.1𝑘𝑔, 𝑐 =
12.5𝑘𝑔/𝑠 changes with respect to time is given in Figure 11.2 (a).)

Asides from predicting changes, another way in which conservation laws are applied is for cases
where change is nonexistent. In this case, it is recognized as time-invariant (or steady state) problem (For
example, we would like to know when will the falling velocity becomes constant, i.e. the terminal
velocity).

𝑑𝑣(𝑡)
(ii) Steady state problem: [Change=increase-decrease=0] or [ 𝑑𝑡
= 0]

𝑑𝑣(𝑡)
For falling parachutist case, steady-state conditions would correspond to the case where 𝑑𝑡
=
𝑚𝑔−𝑐𝑣(𝑡) 𝑔𝑚
𝑚
= 0. Figure 11.2 is plotted by using the solution, 𝑣(𝑡) = 𝑐
(1 − 𝑒 −(𝑐⁄𝑚)𝑡 ). From the Figure 11.2
below, it shows that the falling parachutist’s velocity keeps increasing and thus varies with time initially
(transient case). When the velocity reaches to a point where the 𝐹𝑔𝑟𝑎𝑣𝑖𝑡𝑦 , (𝑚𝑔) is equal to 𝐹𝑑𝑟𝑎𝑔 , (𝑐𝑣),
𝑑𝑣(𝑡)
there is no more increase in velocity afterward due to 𝑑𝑡
= 0.

186
 Figure 11.2: (a) Transient and (b) steady state cases for falling parachutist problem.

𝑔𝑚
From Figure 11.2(b), the terminal velocity and terminal time can be computed as 𝑣𝑡𝑒𝑟𝑚𝑖𝑛𝑎𝑙 = 𝑐
𝑚 𝑐
and 𝑡𝑡𝑒𝑟𝑚𝑖𝑛𝑎𝑙 = − 𝑐
ln |1 − 𝑔𝑚 (𝑣𝑡𝑒𝑟𝑚𝑖𝑛𝑎𝑙 )|. Note that velocity will increase when the first derivative,
𝑑𝑣(𝑡) 𝑑𝑣(𝑡) 𝑑𝑣(𝑡)
> 0 , decrease when < 0, and remain unchanged when = 0.
𝑑𝑡 𝑑𝑡 𝑑𝑡

187