Integral Calculus and Differential Equations

MAT 120

Lecture Notes
 Preface and Acknowledgements

This lecture note compilation has been prepared for aiding the students who
are taking the course MAT 120 (Integral Calculus & Differential Equations) that
is offered by BRAC University. These notes are a compilation of parts taken from
two other books (listed below) that has been shortened down and altered so that
it is adequate for the students taking this course. These notes were created under
the strict supervision of eminent mathematician, Dr. Syed Hasibul Hasan Chowd-
hury. The main goal of this compilation is to help keep things organized for the
students and ensure continuity of the course content among every section. We are
highly indebted to the senior students of MNS Department, notably Mishaal Hai
and Shaikot Jahan Shuvo, for helping us out in making this possible. Since this is
the first version of this compilation, there may be some errors and typing mistakes.
If any mistakes are find please report them to ahmed.rakin@bracu.ac.bd.

Ahmed Rakin Kamal

Project Coordinator:

Dr. Syed Hasibul Hasan Chowdhury

Ahmed Rakin Kamal

Typeset by:

Rahul Pal (Editor & Template Designer)

Ahmed Nafis Farhan
Ruhan Habib
Md. Sakib Hasan

Reference Books:
• Calculus of One Variable by Keith E. Hirst
• Schaum’s Outline of Differential Equations, 3rd Edition"
Contents
1 Integration 3
1.1 Integration as Summation . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Some Basic Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 The Logarithmic Integral . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 Integrals with Variable Limits . . . . . . . . . . . . . . . . . . . . . . 10
1.5 Infinite Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.6 Improper Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2 Integration by Parts 18
2.1 The Basic Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2 Reduction Formulae . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.3 The Gamma Function . . . . . . . . . . . . . . . . . . . . . . . . . . 22

3 Integration by Substitution 26
3.1 Some Simple Substitutions . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2 Inverse Substitutions . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.3 Square Roots of Quadratics . . . . . . . . . . . . . . . . . . . . . . . 30
3.4 Rational Functions of Cos & Sin . . . . . . . . . . . . . . . . . . . . . 32

4 Integration of Rational Functions 35

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.2 Partial Fractions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.3 The Integration Process . . . . . . . . . . . . . . . . . . . . . . . . . 39

5 Geometrical Applications of Integration 42

5.1 Arc Length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
5.2 Surface Area of Revolution . . . . . . . . . . . . . . . . . . . . . . . . 46
5.3 Volumes by Slicing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
5.4 Volumes of Revolution . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.4.1 The Disc Method . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.4.2 The Cylindrical Method . . . . . . . . . . . . . . . . . . . . . 52

6 First Order Differential Equations 55

6.1 Basic Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
6.2 Classifications of First Order Differential Equations . . . . . . . . . . 59
6.3 Separable First Order Differential Equations . . . . . . . . . . . . . . 61
6.4 Exact First Order Differential Equations . . . . . . . . . . . . . . . . 65
6.5 Linear First Order Differential Equations . . . . . . . . . . . . . . . . 70
6.6 Applications of First Order Differential Equations . . . . . . . . . . . 72

1
BRAC
University MAT 120
6.7 Graphical & Numerical Methods of Solving First Order Differential
Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

7 Higher Order Differential Equations 81

7.1 Linear Differential Equations . . . . . . . . . . . . . . . . . . . . . . . 81
7.2 Second Order Linear Homogeneous Differential Equations . . . . . . . 85
7.3 nth order Linear Homogeneous Differential Equations . . . . . . . . . 88
7.4 Method of Undetermined Coefficients . . . . . . . . . . . . . . . . . . 90
7.5 Initial Value Problems for Linear Differential Equations . . . . . . . . 94
7.6 Application of Second Order Linear Differential Equations . . . . . . 96
7.7 Reduction of Linear Differential Equation to First Order Equations . 103

Appendices 110

2 of 110
Chapter 1

Integration
Ideas about integration have been around much longer than those of differentiation. The
Greek mathematician Archimedes (3rd Century B.C.) knew how to calculate the area of a
segment of a parabola by “quadrature”, which involved approximation by regions of known
area such as quadrilaterals or triangles. Integration as the reverse of differentiation was a
much later idea, after the invention of the differential calculus in the time of Newton (17th
Century A.D.). Bringing the two views of integration together was the work of mathemati-
cians in the 19th Century in particular, culminating in the work of Bernhard Riemann
(1826–1866), whose name is associated with the theory of integration which is often stud-
ied in Real Analysis courses (see Howie Chapter 5 for example). In this chapter we shall
review some of the basic ideas about integration, briefly revise some standard results from
school calculus, and develop some aspects of integration using limits.

1.1 Integration as Summation

The basic idea is that of “area under a graph”, where this phrase is
commonly used to signify the area of a plane region bounded by the
graph of a non-negative function y = f (x), the x−axis, and the lines
x = a and x = b. We approximate this area using rectangles, from
below and from above. These ideas are illustrated in Figure 1.1. For
functions that we commonly use, with continuous graphs for example,
it turns out that the approximations from below and from above can be
made as close as we wish. As these approximations get better and better
they “home in” on a definite numerical value which corresponds to the
area of the region. This is a limiting process, but a more complicated
one than we discussed in Chapter 2.

We generate the approximations by subdividing the interval a ≤ x ≤

b by means of an increasing sequence of points

a = x1 < x2 < ... < xn−1 < xn = b

3
BRAC
University MAT 120

Figure 1.1: The integral as a sum

The sum of the areas of the rectangles lying below the graph (as in
the top diagram) is called the lower sum corresponding to the sub-
division, and the sum of the areas of the rectangles enclosing the area
underneath the graph (as in the bottom diagram) is called the upper
sum corresponding to the subdivision. We introduce the notation for
these as follows.
n
X n
X
s= mi (xi − xi−1 ); S= Mi (xi − xi−1 ).
i=1 i=1

In these sums each term represents the area of a rectangle, where

(xi − xi−1 ) represents the width and mi or Mi represents the height, in
the first and second sum respectively.

The fundamental idea is that as the lengths of the intervals of the

subdivision all tend to zero, both the upper and lower sum tend to a
common limit, the area under the curve.

The theoretical underpinning of this idea is quite extensive, and be-

yond the scope of this book. It can be shown that for “respectable”

4 of 110
BRAC
University MAT 120

functions (given by standard formulae for example) we can use any con-
venient approximating sum in this way, where we replace Mi or mi with
any number between, usually of the form f (ci ), where xi−1 ≤ ci ≤ xi .
This gives the sum
n
X
f (ci )(xi − xi−1 ).
i=1

This sum also represents a sum of areas of rectangles. The theory

shows that all such sums have a common limit as the number of intervals
in the subdivision increases, and their lengths all tend to zero. This
common limit is called the integral, denoted by
Z b
f (x)dx,
a

and a function for which this common limit exists is said to be in-
tegrable over the interval a ≤ x ≤ b. The function f (x) which we are
integrating is referred to as the integrand. We shall utilise the idea of
the integral as the limit of a sum in Chapter 11, applied to geometrical
quantities such as area and volume.

The connection between this process and differentiation was estab-

lished by means of the Fundamental Theorem of Calculus, which says
that under appropriate conditions
Z b
f 0 (x)dx = f (b) − f (a).
a

The first clear statement of this theorem is attributed to Isaac Newton

in some unpublished work written in 1666.

5 of 110
BRAC
University MAT 120

1.2 Some Basic Integrals

In the remainder of this chapter, and in Chapters 8, 9 and 10, we shall
consider integration as the reverse of differentiation. Given a function
f (x), the problem is to find another function F (x) whose derivative is
f (x). Such a function F (x) is called an indefinite integral, denoted by
R Rb
f (x)dx. In contrast, the integral a f (x)dx discussed above is called
a definite integral.
d d
We first note that if dx F (x) = f (x) then dx (F (x) + C) = f (x) for
any real number C.

C is known as the constant of integration, and strictly speaking

should be included whenever we evaluate an indefinite integral.

Basic integration is normally first encountered in school mathemat-

ics, and it is assumed that readers are familiar with a small number of
indefinite integrals, as in the following table, where we have omitted the
constant of integration, as we shall do throughout this book.
R
f (x) f (x)dx

xα+1
xα (α 6= −1) α+1

x−1 ln |x|

cosx sinx

sinx −cosx

sec2 x tanx

ekx
ekx k

Various rules and identities can be used to reduce many integrals to

the basic ones above. There are some standard methods for doing this
which are applicable to various classes of functions, and we shall consider
these in later chapters. The examples in this section illustrate this idea
with some elementary integrals. The algebraic rules of integration used
in this chapter are as follows

6 of 110
BRAC
University MAT 120

Z Z Z
(Cf1 (x) + Df2 (x))dx = C f1 (x)dx + D f2 (x)dx,

where C and D are constants (the addition rule), and

Z b Z c Z b
f (x)dx = f (x)dx + f (x)dx.
a a c

R
Example 1.2.1. Evaluate cos 2xdx.
The table above suggests that the answer will involve sin 2x, and we
d
can check by differentiation. Now dx sin 2x = 2 cos 2x, and so we can
see that we have to compensate for the factor of 2. We can therefore
d sin 2x
write dx 2 = cos 2x, and so,
Z
sin 2x
cos 2x dx =
2

R
Example 1.2.2. Evaluate ax dx(a > 0).
We know that ax = ex ln a . Using this, and realising that ln a is simply
a constant, we can write

ex ln a ax
Z Z
ax dx = ex ln a dx = =
ln a ln a
R3
Example 1.2.3. Evaluate −3 ln (1 + x2 ) sin (x)dx.
The function looks complicated to integrate as an indefinite integral,
so we have to look at it another way, in this case geometrically. We notice
that the function is an odd function, and the interval of integration is
symmetric about the origin. The answer is therefore zero. This is clear
from Figure 1.2, with the interpretation explained above, that areas
below the x-axis correspond to a negative answer for the integral. We
have implicitly used the rule of integration telling us that,

Z 3 Z 0 Z 3
2 2
ln (1 + x ) sin (x)dx = ln (1 + x ) sin (x)dx+ ln (1 + x2 ) sin (x)dx
−3 −3 0

7 of 110
BRAC
University MAT 120

The value of the first integral on the right-hand side is then minus that
of the second integral, so they cancel to zero.

Figure 1.2: Integral of an odd function

1.3 The Logarithmic Integral

One of the items
R 1 in the table of integrals at the beginningR 1 of Section 1.2
states that x dx = ln |x|. Some textbooks say that x dx = ln x + C
. Neither of these is strictly correct, and in this section we shall discuss
this integral further. The problem is caused by the fact that neither
1/x nor ln |x| is defined for x = 0. We approach the problem through
graphs.

Figure 1.3: The logarithmic integral

8 of 110
BRAC
University MAT 120

In Figure 1.3 the thicker graph is that of y = x1 , and the other graphs
are of ln x+C for x > 0, and ln (−x)+D for x < 0. All these logarithmic
graphs have derivative 1/x. For x > 0 this is a standard result. Now
1/x is an odd function, and so its gradient at −x will be the negative
of the gradient at x. This means that for x < 0, 1/x is the gradient for
the logarithmic function with x replaced by −x, i.e., ln (−x). For x < 0
we can also verify this result using the chain rule.

d 1 d 1 1
(ln (−x) + D) = (−x) = .(−1) =
dx −x dx −x x

Finally we have to consider the constant of integration. We can see

that we could choose an arbitrary function of the form ln x+C for x > 0,
and an arbitrary function of the form ln (−x) + D for x < 0. There is
no reason why C should be the same as D, since the two “halves” do not
join together, because of discontinuity at x = 0. So the most complete
description of the set of functions whose derivative is 1/x is,
(
ln (x) + C, (x > 0),
F (x) =
ln (−x) + D, (x < 0).

To write ln |x| + C is a convenient abbreviation, but it conceals the

fact that C and D can be different.

9 of 110
BRAC
University MAT 120

1.4 Integrals with Variable Limits

Rb
So far when we have evaluated a definite integral of the form a f (x)dx,
the limits of integration a and b have been constants. There is no reason
however why they should not involve a variable.
Z sin t
Example 1.4.1. Evaluate the integral (x2 − 2x + 3)dx.
t−2

Z sin t 3
 sin t
x
(x2 − 2x + 3)dx = − x2 + 3x
t−2 3 t−2
3 3
sin (t) (t − 2)
= − sin2 t + 3 sin t − + (t − 2)2 − 3(t − 2)
3 3
3 3
sin (t) t 38
= − sin2 t + 3 sin t − + 3t2 − 11t +
3 3 3
This example shows that when such an integral is evaluated the an-
swer involves the variable which is present in the limits of integration.
Now if we want to find the derivative of this expression we can evaluate
the integral, as we have done in the above example, and then differenti-
ate the answer. However we can find the derivative without integrating
first, as follows. Suppose that we know the indefinite integral of f (x,
i.e., that we know a function F (x) satisfying F 0 (x) = f (x). We then
have,
Z b(t) Z b(t)
G(t) = f (x)dx = F 0 (x)dx = F (b(t)) − F (a(t))
a(t) a(t)

We can therefore differentiate using chain rule to obtain,

G0 (t) = F 0 (b(t))b0 (t) − F 0 (a(t))a0 (t) = f (b(t))b0 (t) − f (a(t))a0 (t)

Example 1.4.2. Find the derivative with respect to t of the function

Z t3 +4t2
defined by G(t) = cos (x2 )dx
t2 −3t

In this example we cannot find the indefinite integral, but we can still
use the formula for the derivative. This tells us that the derivative is,

G0 (t) = (3t2 + 8t) cos((t3 + 4t2 )2 ) − (2t − 3)cos((t2 − 3t)2 )

10 of 110
BRAC
University MAT 120

1.5 Infinite Integrals

Rb
So far we have considered definite integrals of the form a f (x)dx, where
a and b are real numbers, and for a non-negative function the integral
corresponds to the area between the graph and the x−axis. In effect we
are considering a function whose domain is limited to lie between a and
b. But many functions have as their domain the set of all real numbers,
or the set of all positive real numbers, or some other unbounded set. In
this section we shall consider how to interpret the idea of the area of the
region between such a graph and the x − axis. This leads to the idea
of an infinite integral. Suppose we have a function f (x) whose domain
R ∞the set of all real x satisfying x ≥ a. We want to give ameaning to
is
a f (x)dx.

Imagine that we are going to paint such a region and that we want
to know whether we can do it with a finite amount of paint. What we
can do is to start at x = a and paint the region up as far as x = t. We
can then measure the amount of paint used. This will depend on t, and
so can be expressed as a function F (t). Using the ideas of limits from
Chapter 2 we can then investigate lim F (t). If this is finite it would
y→∞
appear that we can paint the complete region using a finite amount of
paint. This somewhat far-fetched analogy motivates the definition.

Definition 1.5.1. If f (x) is a function continuous for x ≥ a we define

the infinite integral of f by
Z ∞ Z t
f (x)dx = lim f (x)dx.
a t→∞ a

If this limit exists and is finite we say that the infinite integral con-
verges. Otherwise the infinite integral diverges. We can similarly
define
Z b Z b
f (x)dx = lim f (x)dx.
−∞ t→−∞ t

11 of 110
BRAC
University MAT 120

Example 1.5.1. Investigate the convergence or otherwise of the infinite

integral int∞ α
1 x dx.

We will begin with two numerical cases by way of illustration.

Firstly, let α = −2. We then have
Z t  t
1 1 1
dx = − = 1 − → 1ast → ∞
1 x2 x 1 t
The integral therefore converges and we can write
Z ∞
1
dx = 1
1 x2
Now consider the case α = −1. We now have,
Z t
1
dx = [ln x]t1 = ln t → ∞ast → ∞
1 x
Therefore the integral in this case diverges. In general suppose that
α 6= −1. Then,

t t
xα+1 xα+1
Z 
α 1
x = = −
1 α+1 1 α+1 α+1
Now if α + 1 > 0 the right-hand expression tends to infinity, and so
the corresponding infinite integral diverges. If α + 1 < 0 then,

xα+1 1 1
lim − =−
t→∞ α + 1 α+1 α+1
Therefore the integral converges, and we can write
Z ∞
1
xα = −
1 α+1
This is verified with the case α = −2 we considered above.

12 of 110
BRAC
University MAT 120

Z ∞
Example 1.5.2. Show that the infinite integral e−x converges and
0
find its value.
Evaluating the integral over the finite interval 0 ≤ x ≤ t gives,
Z t
e−x = [−e−x ]t0 = 1 − e−t → 1ast → ∞
0

Therefore the integral converges and its value is 1.

R∞ 2
Example 1.5.3. Show that the integral 1 e−x converges.
In this case, unlike Example 1.5.2, we cannot evaluate the indefi-
nite integral explicitly and so we cannot investigate the required limit
directly. What we do is to compare this integral with one which we
already know to converge. To return to the area interpretation, if we
2
can show that the region between the graph of e−x and the x − axis is
smaller than one for which we already know that the area is finite then
it too must enclose a finite area. The argument proceeds as follows in
2
this case. For all x ≥ 1 we know that 0 ≤ e−x ≤ e−x . We therefore
have, using the result of Example 1.5.2,
Z t Z t
2
−x
e dx ≤ e−x dx ≤ 1
0 0

The value of the left-hand integral increases as t increases, because

the integrand is positive for all x, and so it tends to a finite limit as
t → ∞. Therefore the integral converges. Because we cannot evaluate
the indefinite integral we do not know the value of the infinite integral.
What we have showed is that
Z ∞
2
e−x dx ≤ 1
0

The procedure used in this example generalises, as in the following

theorem.

13 of 110
BRAC
University MAT 120

Theorem 1.5.1. Comparison Test for Infinite Integral

Suppose that f (x) and g(x) are continuous, and R ∞ that 0 ≤ g(x) ≤
f (x), for all x ≥ a. Then ifRthe infinite integral a f (x)dx converges,
∞
so does the infinite integral a g(x)dx, and
Z ∞ Z ∞
g(x)dx ≤ f (x)dx
a a

Proof. An analytical proof is outside the scope of this book. We shall

give an intuitive geometrical explanation, which generalises the argu-
ment in Example 1.5.3.

Figure 1.4: Integral comparison test

We can see from Figure 1.4 that the area between the graph of f (x)
and the x − axis is greater than the area between the graph of g(x) and
the x − axis. Therefore
Z t Z t
g(x)dx ≤ f (x)dx
a a

From the figure we can also see that these areas increase as t in-
creases, and because the functions are both non-negative it follows that
the corresponding integrals also increase
R∞ as t increases. If we denote the
value of the convergent integral a f (x)dx by K we deduce that
Z t
g(x)dx ≤ K
a

14 of 110
BRAC
University MAT 120

Rt
for all t ≥ a. So a g(x)dx is an increasing function of t which is
bounded above. It therefore has a limit H ≤ K. (This last result would
be proved in a course on Real Analysis, for example in Howie Chapter
3).

So far we have restricted attention to non-negative functions. Defi-

nition 1.5.1 does not require this, and in the next example we consider
a function taking both signs.
R ∞ sin x
Example 1.5.4. Show that the integral 2π x2 dx converges
From Example 1.5.1 (with α = −2) we know that the area between
the graph of 1/x2 and the x − axis is finite. We also know that

1 sin x 1
− 2
≤ 2 ≤ 2
x x x
We can therefore see, in Figure 1.5 , that the areas contained by those
parts of the graph of sin x
x2 above the x − axis will be finite in total. The
same will be true below the x − axis, so that the total area between the
graph of sin x
x2 and the x − axis will be finite. Now the integral is found
by subtracting the total area below the axis from the total area above
the axis, and this will therefore be finite. In other words the integral
R ∞ sin x
2π x2 dx will converge.

Figure 1.5: Graph of example 1.5.4

This is an example of a generalisation of Theorem 1.5.1 which we

state here without proof.

15 of 110
BRAC
University MAT 120

Theorem 1.5.2. Suppose that f (x) and g(x) are continuous, R ∞and that
|g(x)| ≤ f (x), for all x ≥ a. Then ifRthe infinite integral a f (x)dx
∞
converges, so does the infinite integral a g(x)dx, and
Z ∞ Z ∞ Z ∞
g(x)dx ≤ |g(x)|dx ≤ f (x)dx.
a a a

1.6 Improper Integrals

Rb
In this section we consider integrals of the form a f (x)dx, where the
function f is undefined, or has a discontinuity, or is unbounded, at some
point of the interval a ≤ x ≤ b. If this occurs inside the interval at some
point c we can split the integral into two by using the rule of integration
Z b Z c Z b
f (x)dx = f (x)dx + f (x)dx
a a c

This means that we can restrict attention to situations where the

point of discontinuity etc. occurs at an end-point of the interval. We
shall in fact discuss only cases where this happens at the lower end-
point, the case of the upper endpoint being
R 1 exactly similar. So by way
1
of example we can consider the integral 0 x2 dx, where the integrand
is undefined at the end-point x = 0. Such an integral is called an im-
proper integral, and we need to investigate questions of convergence.
As with infinite integrals we are asking whether we can sensibly define
the area of the region between the graph and the x-axis when the region
is unbounded.

Definition 1.6.1. Let f (x) be a function continuous for a ≤ x ≤ b.

Rb
Then the improper integral a f (x)dx is said to converge if the integral
Rb +
c f (x)dx has a limit as c → a .We then define

Z b Z b
f (x)dx = lim+ f (x)dx
a c→a c

16 of 110
BRAC
University MAT 120

R1 1
Example 1.6.1. Investigate convergence of the improper integral 0 x2 dx

Using Definition 1.6.1, we have

Z 1  1
1 1 1
2
dx = − = − 1 → ∞asc → 0+ .
c x x c c
R1
Therefore the integral 0 x12 dx does not converge.

17 of 110
Chapter 2

Integration by Parts

2.1 The Basic Technique

We are considering integration as the reverse of differentiation, and we
should therefore expect that rules of differentiation should relate to tech-
niques of integration. The technique of integration by parts discussed
in this chapter is a consequence of the product rule for differentiation.
That rule tells us that

d dV dU
(U (x)V (x)) = U (x) + V (x)
dx dx dx

If we integrate both sides with respect to x we obtain

Z Z
dV dU
U (x)V (x) = U (x) + V (x)
dx dx

which can be rearranged in the form

Z Z
dV dU
U (x) = U (x)V (x) − V (x)
dx dx

This equation
Z therefore gives us a procedure for evaluating an integral
of the form f (x)g(x)dx. We have to decide which of f and g to
dU
identify with U . If we choose f = U this is usually because is a
dx

18
BRAC
University MAT 120

dV
simpler expression than U . We then have to identify g with , and
dx
we have to be able to find V , i.e., we have to be able to integrate g(x).
R
Example 2.1.1. Evaluate x cos x dx.

The integrand is a product, and so we have an expression for which

integration by parts is a possible technique. So we have to decide which
of the two components (x or cos x) to identify with U . It is clear that if
dU
we choose U = x then = 1 , which is simpler, whereas if we choose
dx
dV x2
= x then v = , which will give a more complicated integral. So
dx 2
dV
we choose = cos x gives V = sin x. The formula for integration by
dx
parts therefore gives

Z Z
x cos x dx = x sin x − sin x.1.dx = x sin x + cos x

which can be checked by differentiation.

R
Example 2.1.2. Evaluate ex sin 2x dx.

In this example it is not immediately clear which of the two com-

ponents of the product to choose for U . In each case the derivative is
no simpler, and we can integrate either expression. In fact we could
choose either U = ex or U = sin 2x in this example. We shall choose
the former, and leave the other choice as an exercise for the reader to
carry out. As in Example 8.3 we find that integrating by parts once
does not leave us with a straightforward final integral, so we repeat the
calculation. We shall find that we end up with the integral we started
with, but in fact this then gives us an equation which we can solve to
find I.

Z   Z
cos 2x cos 2x
I= ex sin 2x dx = ex − + ex dx
2 2

19 of 110
BRAC
University MAT 120

2.2 Reduction Formulae

R
When we evaluate xR2 sinh x dx, we have to integrating by parts twice.
If we were faced with x6 sinh x dx we would have to integrate by parts
six times.
Z Z 1
Example 2.2.1. Find reduction formulae for In = xn ex dx and Jn = xn ex dx.
0

The notation In indicates that the integral involves the integer pa-
rameter n. The answer depends upon the value of n, as for example
with

1 1
xn+1
Z 
n 1
x = = (n 6= −1),
0 n+1 0 n+1
where we can see explicitly that the answer involves n.

In this example we have, in the case of the indefinite integral,

Z Z
In = xn ex dx = xn ex − nxn−1 ex dx

We can see that the final integral is obtained from In by replacing

n by n − 1 and multiplying by n. This is expressed in the reduction
formula
Z
In = xn ex dx = xn ex − nIn−1

In the case of the definite integral the calculations are the same, and
so we obtain

Z 1 Z 1
n x
Jn = x e dx = [xn ex ]1o − nxn−1 ex dx
0 0

which gives the reduction formula

Jn = e − nJn−1

20 of 110
BRAC
University MAT 120

R 1 We can use this formula to work out, for example, the value of J7 =
7 x
0 x e dx

We first apply the reduction formula with n = 7, giving J7 = e − J6 .

We now apply the same reduction formula, but with n = 6, and so

J7 = e − J6 = e − 7(e − 6J5 ) = −6e + 7.6J5 = −6e + 42J5 .

This process continues until we reach J0 , which we can evaluate ex-
plicitly, because
Z 1
J0 = x0 ex dx = e − 1
0
The complete process using the reduction formula for values of n from
7 down to 1 is as follows.

J7 = e − J6
= e − 7(e − 6J5 ) = −6e + 42J5
= −6e + 42(e − 5J4 ) = 36e − 210J4
= 36e − 210(e − 4J3 ) = −174e + 840J3
= −174e + 840(e − 3J2 ) = 666e − 2520J2
= 666e − 2520(e − 2J1 ) = −1854e + 5040J1
= −1854e + 5040(e − 1.J0 ) = −1854e + 5040(e − (e − 1))
= −1854e + 5040.
Rπ
Example 2.2.2. Find a reduction formula for In = 0 xn sin x dx.

We shall see that we need to integrate by parts twice in this case, in

order to return to an integral containing sin x.
Z π Z π
n n
In = x sin x dx = [x (− cos x)]π0
+ nxn−1 cos x dx
0 Z π 0

= πn + nxn−1 cos x dx
0 Z π
n n−1 π
= π + [nx sin x]0 + n(n − 1)xn−2 sin x dx
0
n
= π − n(n − 1)In−2

21 of 110
BRAC
University MAT 120

So the reduction formula is

In = π n − n(n − 1)In−2
In Example 2.2.1 the integer parameter n was reduced by 1 at each
stage. In this example n is reduced by 2 each time. So if we begin with
an odd value of n we shall finish by needing to calculate I1 , and if we
start with an even value of n we shall need the value of I0 . In both cases
these are integrals which are easy to evaluate explicitly.

2.3 The Gamma Function

In this section we consider some elementary properties of the Gamma
function. This is a function encountered in applied mathematics and
in statistics. It brings together integration by parts and infinite and
improper integrals discussed in Sections 1.5 and 1.6.
Definition 2.3.1. The Gamma function is a function of the variable x
defined by the integral

Z ∞
Γ(x) = tx−1 e−t dt (2.1)
0

This is an infinite integral, and so we need to discuss convergence, as

in Section 1.5.

If x−1 < 0 then the integrand contains a negative power of t, which is

undefined at t = 0. In this case we therefore have an improper integral,
and convergence at t = 0 must be investigated, as in Section 1.6.

We shall investigate these two cases of convergence separately by

splitting the integral at t = 1, letting

Z 1 Z ∞
x−1 −t
I1 = t e dt, tx−1 e−t dt.
0 1

In both cases we use comparison to discuss convergence.

Consider I1 . Since 0 ≤ t ≤ 1 we have e−1 ≤ e−t ≤ 1, so

22 of 110
BRAC
University MAT 120

tx−1 e−1 ≤ tx−1 e−t ≤ tx−1

If x > 0 then for 0 < h < 1 we have

Z 1  x 1
t 1 1
tx−1 dt = = (1 − hx ) → as h → 0.
h x h x x
R1 R1
Hence 0 tx−1 dt converges, and so 0 tx−1 e−1 dt converges by compar-
ison. Now if x < 0 then for 0 < h < 1 we have

1  x 1
e−1
Z
−1 x−1 t −1
e t dt = e = (1 − hx ) → ∞ as h → 0.
h x h x
R1 R1
Hence 0 e−1 tx−1 dt converges, and so 0 tx−1 e−1 dt converges by com-
parison. Now if x < 0 then for 0 < h < 1 we have

Now we consider I2. Let n denote the first integer greater than x.
We therefore have,

(n + 1)! (n + 1)!
tx−1 e−t < tn−1 e−1 < tn−1 n+1 =
t t2
R∞
In Example 1.5.1 we saw that 1 t12 dt converges.
R ∞ x−1 Therefore by com-
parison (noting that (n + 1)! is a constant), 1 t e−tdt converges.

We would like to integrate by parts, but we have an infinite (and

possibly improper) integral, so this needs careful consideration. Suppose
that x > 1, so that the integral is not improper at x = 0. We then
integrate by parts as follows.

Z k
[−tx e−t ]k0 + xtx−1 e−tdt
0

lrtting k → ∞ then gives

Z k Z ∞
x −t
t e dt = x tx−1 e−1 dt, i.e., Γ(x + 1) = xΓ(x)
0 0

23 of 110
BRAC
University MAT 120

This argument will generalise to show that we can integrate conver-

gent infinite integrals by parts. We can deal with the case 0 < x ≤ 1,
when the integral is improper, in a similar fashion, using the limit def-
inition of an improper integral as we did for an infinite integral above.
So we can say that Γ(x + 1) = xΓ(x) for x > 0.
This looks like a reduction formula (see Section 2.2), and if we let
x = n, a positive integer, we can see that

Γ(n + 1) = nΓ(n) = n(n − 1)Γ(n − 1)

= n(n − 1)(n − 2)Γ(n − 2)
= n(n − 1)(n − 2)...2.1.Γ(1).

R∞
Now Γ(1) = 1 t0 e−t dt = [−e−t ]∞ 0 = 1, so Γ(n + 1) = n! The
Gamma function can therefore be seen as a generalisation of the factorial
function for non-integer values.
Example 2.3.1. A Strange Example

1 dV
We integrate tan x by parts, using U = , = sin x.
cos x dx
Z Z
sin x
tan x dx = dx
cos x Z
− cos x − sin x
= + − cos x. dx
cos xZ cos2 x
= −1 + tan x dx.
R
Cancelling tan x dx therefore gives 0 = −1.

So how do we explain this apparent paradox? Is it because of the

constant of integration? Well if we were to include that we would get
Z Z
tan x dx + C = −1 + tan x dx + C,

suggesting that C = −1+C, which again is wrong. If the constant of

integration is arbitrary perhaps we should not have the same constant
C on both sides. We would then obtain

24 of 110
BRAC
University MAT 120

Z Z
tan x dx + C = −1 + tan x dx + D.

This would imply that C = −1 + D, but if C and D are arbitrary

constants why should they be related? What we need to do is to interpret
C and D as representing the set of all possible constants, so that if C
and D represent all possible real numbers then the set of all numbers
of the form −1 + D is also the set of all real numbers, the same as C.
From another perspective it means that we have to think carefully about
what an indefinite integral is. This example suggests that an indefinite
integral is not a function, but a set of functions, and so the two sets of
functions on either side of the equation

Z Z
tan x dx + C = −1 + tan x dx + C
R
are identical, and there is no question of cancelling tan x dx.

When we introduced indefinite integrals in Section 1.2 we used the

indefinite article, and described a function F (x) whose derivative is f (x)
as an indefinite integral (and not the indefinite integral). This is normal
usage, and it would be very complicated to develop the procedures of
integration in terms of the language of sets of functions. We shall not
therefore change our approach, but simply be aware that occasionally
we may need to think more precisely about the definition of an indefinite
integral if we encounter an apparent paradox.

25 of 110
Chapter 3

Integration by Substitution

The theoretical basis for integration by substitution is the chain rule for
differentiation, which says that

d
f (g(x)) = f 0 (g(x))g 0 (x)
dx

Regarding integration as the reverse of differentiation leads us to

integrate both sides of this equation to write

Z
f 0 (g(x))g 0 (x) dx = f (g(x))

Given an integral of this form we can transform it by means of the

du
substitution u = g(x). We then have = g 0 (x) and so
dx
Z Z Z
0 0 0 du d
f (g(x))g (x) dx = f (u) dx = (f (u)) dx = f (u) = f (g(x))
dx dx

where we have used the chain rule with the intermediate variable u,
du d
to recognise that f 0 (u) = (f (u)).
dx dx
du
In fact we implement this process symbolically by rewriting =
dx
g 0 (x) in the form du = g 0 (x) dx. The integration procedure then appears
in the form

26
BRAC
University MAT 120

Z Z
0 0
f (g(x))g (x) dx = f 0 (u)du = f (u) = f (g(u)) (3.1)

The underlying idea is that the substitution gives rise to a simpler

integral involving the variable u. After having evaluated this integral
we then replace u in the answer by g(x), so as to present the answer in
terms of the original variable x.

This can all be made analytically rigorous. The details are beyond
the scope of this book. In the remainder of this chapter we shall con-
centrate therefore on applying this technique of integration in a variety
of circumstances.

3.1 Some Simple Substitutions

Z
cos x
Example 3.1.1. Evaluate the indefinite integral dx
(1 + sin x)3
The theory above
Z looks straightforward, but if we are presented with
cos x
an integral like dx, how are we to find an appropriate
(1 + sin x)3
substitution which will transform the integral into a simpler one that
we can recognise? We need to be able to discern what should play the
role of f and what Zshould play the role of g. We note that the general
integral expression f 0 (g(x))g 0 (x) dx involves both g and its derivative,
so what we need to look for is one part of the integrand which is the
derivative of another part of the integrand. In this example we can see
that the integrand involves sin x and also cos x, which is the derivative
of sin x. This suggests using the substitution u = sin x. We then have
du = cos x dx, and so the integral transforms as
Z Z
cos x 1
dx = du
(1 + sin x)3 (1 + u)3

The latter is an integral we should know how to do, but if not we can
simplify it still further with a linear transformation w = 1 + u, giving
dw = du, and therefore

27 of 110
BRAC
University MAT 120

Z Z Z
cos x 1 1 1
dx = du = dw = −
(1 + sin x)3 (1 + u)3 w3 2w2

The answer has to be given in terms of the original variable x, so we

have

Z
cos x 1 1 1
dx = − = − = −
(1 + sin x)3 2w2 2(1 + u)2 2(1 + sin x)2
ex
Z
Example 3.1.2. Evaluate the indefinite integral dx
1 + e2x
We first note that e2x = (ex )2 , so that the numerator is the derivative
of part of the denominator. This suggests the substitution u = ex , giving
du = ex dx. We therefore have

ex
Z Z
1
dx = du = tan−1 u = tan−1 (ex )
1 + e2x 1+u2

3.2 Inverse Substitutions

In the previous section the substitutions used replaced part of the inte-
grand by a single variable, as in Example 3.1.1 where we replaced sin x
by u. In this section we consider substitutions which work in the reverse
direction, so that x itself is replaced by an expression in another vari-
able, for example we may substitute x = sin u. We can regard this as
equivalent to u = sin−1 x, hence the term inverse substitution.
x2
Z
Example 3.2.1. Evaluate the indefinite integral I = √ dx
x+2
It is the square root term which makes this integral slightly awkward,
so we choose an inverse substitution to remove it, the obvious one being
x = u2 . We then have dx = 2u du, and so

28 of 110
BRAC
University MAT 120

u4 2u5
Z Z
I= · 2u du = du
u
 + 2 u+2 
Z
64
= 2u4 − 4u3 + 8u2 − 16u + 32 − du
u+2
(using polynomial division)
2u5 8u3
= − u4 + − 8u2 + 32u − 64 ln|u + 2|
5 3
5 3
2x 2 2 8x 2 √ √
= −x + − 8x + 32 x − 64 ln| x + 2|
5 3

Because x = u2 is a simple inverse transformation it has a straight-

√
forward direct equivalent, namely u = x. Applying this direct substi-
dx
tution, however, is slightly more awkward, because we obtain du = √ ,
x
which still involves the square root, and for which the algebraic manip-
ulation needed is slightly more involved.
Z
1
Example 3.2.2. Evaluate the indefinite integral I = √ √ dx
x( 3 x + 2)
Here we need to look for a substitution which will eliminate both the
square and the cube roots, and x = u6 will achieve this. So we have
dx = 6u5 du and therefore

6u5
Z Z
1
I= √ √ dx = du
x( 3 x + 2) u3 (u2 + 2)
u2
Z Z  
2
=6 du = 6 1− 2 du
u2 + 2 u +2
√
√ √ √
    
−1 u −1
6
x
= 6 u − 2 tan √ =6 6
x − 2 tan √
2 2

In the remainder of this chapter we shall concentrate on some inverse

substitutions which work for classes of integrals. They will be illustrated
through particular examples, but discussed in a way which emphasises
their general applicability.

29 of 110
BRAC
University MAT 120

3.3 Square Roots of Quadratics

These can generally be simplified by means of trigonometric or hyper-
bolic substitutions. A knowledge of trigonometric identities, and of the
algebraic technique of completing the square, is a necessity therefore.
The examples are chosen to involve only a single square root term,
but the techniques are equally applicable to integrals of algebraic frac-
tions where the numerator or the de- nominator is the square root of a
quadratic.
Z p
Example 3.3.1. Evaluate the indefinite integral I = 4x2 − 16x + 52 dx

We shall break this example down into a number of steps, each of

which is important in evaluating integrals of this type.

STEP 1 We make the coefficient of x2 equal to 1, so

Z p
I=2 x2 − 4x + 13 dx

(Note that if the coefficient of x2 is negative then we make its coeffi-

cient equal to −1 and then proceed with the following steps.)

STEP 2 Complete the square, giving

Z p
I=2 (x − 2)2 + 9 dx

STEP 3 Write the constant as a square number.

Z p
I=2 (x − 2)2 + 33 dx

STEP 4 Use a linear substitution to replace the variable square term

with a single square, using u = x − 2 in this case.
Z p
I=2 u2 + 32 du

30 of 110
BRAC
University MAT 120

This succession of steps will always transform the square root of a

quadratic into a sum of squares, as in this example, or a difference
of squares. The following steps
√ are therefore applicable to integrals
involving a term of the form u2 + a2

STEP 5 Make an appropriate trigonometric substitution. We need

to find a substitution which makes use of a trigonometric identity reduc-
ing a sum of two squares to a single square term, which will then enable
us to remove the square root. In this√case the appropriate identity is
tan2 t + 1 = sec2 t. However we have u2 + 32 , and so we need to take
the 32 into account, using u = 3 tan t. We then have du = 3 sec2 t dt,
and so the substitution gives

Z p Z
I=2 32 tan2 t + 32 · 3 sec2 t dt = 2 · 3 · 3 sec3 t dt

STEP 6 We now have to evaluate the trigonometric integral. This

involves integration by parts, and one of the exercises in Chapter 2
describes how to do this. We find that

 
1 1
I = 18 sec t tan t + ln|sec t + tan t|
2 2

STEP 7 We now have to express the result in terms of x, firstly by

finding the trigonometric functions in terms of u. The substitution gives
us one of them, because tan t = u3 . To find sec t (or any other trigono-
metric function which might arise is integrals of this type) a helpful
technique is to draw a right-angled triangle in which tan t = u3 . Using
Pythagoras’ Theorem gives the third side of the triangle, and we can
then read off any trigonometric ratio that we need.

31 of 110
BRAC
University MAT 120

Figure 3.1: Trigonometric substitutions

√
u2 + 32
We can now see from the diagram that sec t = . So we can
3
express in terms of u as

√ √
u2 + 32 u u2 + 32 u
I=9 + 9 ln +
3 3 3 3

STEP 8 We replace u, using u = x − 2 from STEP 4, and undertake

some algebraic simplification, to give the answer

p p
I = (x − 2) (x − 2)2 + 32 + 9 ln (x − 2)2 + 32 + (x − 2) − 9 ln 3

The −9 ln 3 term can be absorbed into the constant of integration,

giving finally

p p
I = (x − 2) (x − 2) + 3 + 9 ln (x − 2)2 + 32 + (x − 2)
2 2

3.4 Rational Functions of Cos & Sin

A rational function, is a quotient of polynomials, where the numera-
tor and denominator involve only terms containing non-negative integer
powers of the variable x. By a rational function of cos and sin we mean
a quotient where the numerator and denominator involve only terms
containing non-negative integer powers of cos and sin, for example

32 of 110
BRAC
University MAT 120

cos3 t sin t + sin2 t + 3 − 2 cos2 t

3 sin t − 4 + 5 sin2 t cos5 t − 2 cos t

 a function we use the so-called half-angle substi-

To integrate such
t
tution x = tan . This will always lead to a rational function of x.
2
In Chapter 4 we shall consider methods of integrating rational functions
in general, so the example in this section will be a straightforward one.

Because such expressions often involve both cos t and sin t we need
to use identities which express these in terms of x. We can do this using
a right-angled triangle together with basic trigonometric identities.

Figure 3.2: Half-angle substitution

   
t 1 t
In the right-angled triangle we have x = tan . So dx = sec2 dt,
2 2 2
which we can rearrange in the form
 
t 2
dt = 2 cos2 dx = dx
2 1 + x2
 
t 2
where we obtain the value of cos from the triangle. We can
2
also use the triangle to see that
 
t x
sin =√
2 1 + x2

33 of 110
BRAC
University MAT 120

We therefore have

1 − x2
 
t 2
cos t = 2 cos2 −1= − 1 =
2 1 + x2 1 + x2
   
t t x 1 2x
sin t = 2 sin cos = 2√ √ =
2 2 1 + x2 1 + x2 1 + x2

We can therefore see that each term in the integral of a rational func-
tion of cos t and sin t (including dt) will be transformed into a rational
fraction of x, and so we shall be left with a rational function of x to
integrate.
Z
1
Example 3.4.1. Evaluate the indefinite integral I = dt
2 + sin t
x
Using the substitution x = tan with the identities above gives
2
Z Z
1 1 2
I= dt = 2x 2
dx
2 + sin t 2 + 1+x 2 1 + x
Z
1
= 2+x+1
dx (after simplification)
Z x
1
= (completing the square)
1 2
(x + 2 ) + 34
 
1  
1 x + 2 2x + 1
= q tan−1  q 2  = √ tan−1
 
3 3 3 3
4 4

2 tan( 2t ) + 1
 
2
= √ tan−1 √
3 3

34 of 110
Chapter 4

Integration of Rational Functions

4.1 Introduction
P (x)
A rational function is one of the form R(x) = Q(x) ,where are both P (x)
and Q(x) are polynomials in the variable x, for example
2x3 + 3x2 − 4x + 1
x2 − 3x + 2
In this chapter we shall explain the steps involved in a procedure
which will enable us to integrate any rational function, provided the
algebra is not too horrible! One of the algebraic tools needed is the de-
composition of rational functions into partial fractions, and we discuss
this in the next section. In Section 4.3 we describe the process of inte-
grating a rational function, split into a sequence of steps. We explain
what happens at each step, using different examples at each stage to
illustrate some degree of generality.

Apart from partial fractions, the main algebraic prerequisite is poly-

nomial division.

4.2 Partial Fractions

P (x)
The partial fraction decomposition expresses a rational function Q(x) as a
sum of simpler algebraic fractions. The denominators of these fractions
are determined by the factorisation of the denominator Q(x). When
a real polynomial is completely factorised into real factors, the factors
will either be linear or quadratic, and some factors may occur more than
once.

35
BRAC
University MAT 120

Dealing with partial fractions where the denominator factorises into

linear factors only, none of which are repeated, is a familiar area of school
mathematics. A common method used is that of equating coefficients,
as in the following example.
Example 4.2.1. Find the partial fraction decomposition of
x3 + 2x2 − x + 4
(x − 1)(x + 2)(x − 3)(x + 1)
The decomposition is of the form
x3 + 2x2 − x + 4 A B C D
= + + +
(x − 1)(x + 2)(x − 3)(x + 1) x − 1 x + 2 x − 3 x + 1
If we now put the right hand side over the common denominator

(x − 1)(x + 2)(x − 3)(x + 1)

the numerator will be

A(x + 2)(x − 3)(x + 1) + B(x − 1)(x − 3)(x + 1)

+ C(x − 1)(x + 2)(x + 1) + D(x − 1)(x + 2)(x − 3)

So equating the numerators gives

x3 + 2x2 − x + 4 ≡ A(x + 2)(x − 3)(x + 1) + B(x − 1)(x − 3)(x + 1)

+C(x − 1)(x + 2)(x + 1)
+D(x − 1)(x + 2)(x − 3)

where the use of the identity symbol ≡ emphasises that this is true for
all values of x. One method of finding A, B, C, D is to multiply out the
right-hand side, giving

x3 + 2x2 − x + 4 ≡ (A + B + C + D)x3 + (−3B + 2C − 2D)x2

+(−7A − B − C − 5D)x
+(−6A + 3B − 2C + 6D)

Because this is an identity the coefficient of each power of x on both sides

must be the same. This gives the following system of four equations to

36 of 110
BRAC
University MAT 120

solve for A, B, C, D.
A+B+C +D = 1,
−3B + 2C − 2D = 2,
−7A − B − C − 5D = −1,
−6A + 3B − 2C + 6D =4

An alternative method is to substitute specific values of x in the identity,

chosen so as to make one of the linear factors zero. This has the effect of
making all but one of the terms zero on the right hand side of the first
identity above. This enables us to determine the unknown coefficients
one at a time. The results are given in the following table.
x = 1 6 = −12A A = −1/2
x = −2 6 = −15B B = −2/5
x = 3 46 = 40C C = 23/20
x = −1 6 = 8D D = 3/4
We therefore have
x3 + 2x2 − x + 4 1 2 23
= − +
(x − 1)(x + 2)(x − 3)(x + 1) 2(x − 1) 5(x + 2) 20(x − 3)
3
+
4(x + 1)
x2 − 2x + 3
Example 4.2.2. Decompose into partial fractions.
(x + 1)3

The decomposition will be of the form

x2 − 2x + 3 A B C
= + +
(x + 1)3 x + 1 (x + 1)2 (x + 1)3
We have to find A, B, C and as usual we arrange the right-hand side
as a single fraction over the common denominator, (x + 1)3 . This gives
rise to the numerator A(x + 1)2 + B(x + 1) + C, which must be equal
to x2 − 2x + 3. We could find A, B, C, by comparing coefficients, as in
Example 10.1, or by compensation. Here we shall demonstrate another
method, involving differentiation, reminiscent of the discussion on Taylor
Polynomials.

37 of 110
BRAC
University MAT 120

So in the equation A(x + 1)2 + B(x + 1) + C = x2 − 2x + 3, we

substitute x = −1 to give C = 6. We then differentiate both sides of
the equation to give 2A(x + 1) + B = 2x − 2. Putting x = −1 now
gives B = −4. Differentiating once more gives 2A = 2, and so A = 1.
Therefore the decomposition is
x2 − 2x + 3 1 4 6
= − +
(x + 1)3 x + 1 (x + 1)2 (x + 1)3

2x2 − x + 4
Example 4.2.3. Decompose 2 into partial fractions.
(x + x + 1)2

Ax + B Cx + D
The decomposition will be of the form + 2
(x + x + 1) (x + x + 1)2
2

Arranging this as a single fraction over a common denominator gives

the numerator (Ax + B)(x2 + x + 1) + Cx + D, which must be equal
to 2x2 − x + 4.

We can’t use the differentiation method in the same way as in Ex-

ample 4.5, because there is a quadratic involved, and indeed there is no
real value of x for which it is zero. We can adapt the differentiation
method however, using x = 0 in each case to simplify the calculations.
So we have to solve the identity (Ax + B)(x2 + x + 1) + Cx + D ≡
2x2 − x + 4 for A, B, C. We can see immediately that A = 0 because
there is no x3 term on the right hand side. So the identity simplifies to
B(x2 + x + 1) + Cx + D ≡ 2x2 − x + 4. Putting x = 0 gives B + D = 4.

Differentiating gives 2Bx + B + C ≡ 4x − 1, and putting x = 0 gives

B + C = −1.

Differentiating again gives 2B = 4 and so B = 2. We can now deduce

that C = −3 and D = 2. So the decomposition is
2x2 − x + 4 2 −3x + 2
= +
(x2 + x + 1)2 (x2 + x + 1) (x2 + x + 1)2
Notice that although we expect each numerator to be linear, it is
possible that some of the coefficients may be zero, as has happened with
the first fraction in this case.

38 of 110
BRAC
University MAT 120

In the examples above we have used a variety of methods, chosen to

keep the algebraic manipulation as straightforward as we can.

4.3 The Integration Process

P (x)
Integrating a rational function Q(x) is a process which can be broken
down into a well-defined sequence of steps. In this section we shall de-
scribe the procedure, providing illustrative examples. In the subsequent
section we work through some examples in detail.
STEP 1 Polynomial Division
Is the degree of P (x) greater than or equal to the degree of Q(x)?
P (x)
If the answer is YES then divide Q(x) into P (x) to obtain Q(x) =
A(x) + B(x)
Q(x) , where A(x) and B(x) are polynomials and deg B(x) < deg
Q(x).
STEP 2 Polynomial Division
Following polynomial division we need to factorise the denominator
Q(x). The problem with this step is that there is no general algorithm
which will factorise all polynomials. So in practice this step can only be
carried out if the polynomial Q(x) is relatively straightforward.
STEP 3 Polynomial Division
B(x)
Decompose Q(x) using partial fractions. This was discussed in detail
in Section 4.2.
STEP 4 Polynomial Division
We can now integrate each term in the partial fraction decomposition
separately. Each is a rational function, but only a few different types
of expression occur, as we have seen in Section 4.2, and we discuss each
of them. The first two types involve a linear factor in the denominator,
which may be repeated, and a constant numerator. From the basic
integrals described in Section 1.2 we have the following two results.
Z
A
dx = A ln |x + k|,
x+k
(x + k)−n+1
Z
A
dx = A (n 6= 1)
(x + k)n −n + 1

39 of 110
BRAC
University MAT 120

We now need to consider quadratic denominators. Where the quadratic

R linear
is not a repeated factor the integral will be of the form quadratic
0
q (x) d
We can re-write the integrand in the form linear
q(x) = c q(x) + q(x) , where
c and d are constants, and integrate each term separately.
Z
x+3
Example 4.3.1. Evaluate the integral dx
x2 − x + 4

We rewrite the integrand as explained above to give

Z 1 7
(2x − 1)
Z Z
x+3 2 2
dx = dx + dx
x2 − x + 4 x2 − x + 4 x2 − x + 4
Using the general result
q 0 (x)
Z
dx = ln |q(x)|
q(x)
we can deal with the first integral as follows.
Z 1
2 (2x − 1) 1
2
dx = ln |x2 − x + 4|
x −x+4 2
We deal with the second integral by completing the square of the
denominator, which doesn’t have real roots.
Z 7 Z r r  !
2 7 dx 7 4 −1 4 1
dx = = tan x −
x2 − x + 4 2 (x − 12 )2 + 154
2 15 15 2
R linear
Finally we must deal with integrals of the form (quadratic)n (n > 1)

0
linear q (x) d
We can write the integrand as (q(x))n = c (q(x)) n + (q(x))n , where c and d
are constants.

The following example illustrates the general process for evaluating

such integrals.

Example 4.3.2. Evaluate the integral

2x − 1
Z
dx
(x2 − 2x + 5)2

40 of 110
BRAC
University MAT 120

Splitting the integral as explained above gives

2x − 1 2x − 2
Z Z Z
1
2 2
dx = 2 2
dx + dx
(x − 2x + 5) (x − 2x + 5) (x − 2x + 5)2
2

Evaluating the first integral gives

2x − 2
Z
1
dx = −
(x2 − 2x + 5)2 x2 − 2x + 5
The second integral can be evaluated as follows, using the trigonometric
substitution u = 2 tan t, du = 2 sec2 tdt.
Z Z Z
1 1 1
dx = dx = du
(x2 − 2x + 5)2 ((x − 1)2 + 4)2 (u2 + 4)2
2 sec2 t 2 sec2 t
Z Z
= dt = dt
(4 tan2 t + 4)2 (4 sec2 t)2
Z Z
1 1
= cos2 tdt = (1 + cos 2t)dt
8  16
1 sin 2t 1
= t+ = (t + sin t cos t)
16 2 16
 
1  u  u 2
= tan−1 +√ √
16 2 u 2 + 4 u2 + 4
   
1 −1 x − 1 2(x − 1)
= tan + 2
16 2 x − 2x + 5
where in the penultimate line we have used the right-angled triangle
method.

Adding the two results together then gives

   
2x − 1 −1 x − 1 2(x − 9)
Z
1
dx = tan + 2
(x2 − 2x + 5)2 16 2 x − 2x + 5

41 of 110
Chapter 5

Geometrical Applications of
Integration
In Section 1.1 we discussed integration as summation, and we use that interpretation of
the integral in this chapter to construct integral formulae for some geometrical quantities.
We shall consider length, area and volume, and the notions of centroid and centre of mass.
The most important thing in this chapter is not to remember particular formulae, but to
understand the principles underpinning their construction, so that analogous formulae can
be constructed in other areas of application.

5.1 Arc Length

In this section we derive formulae for the length of a curve. Not all
curves can be given as the graph of a function, for example a circle,
and so we shall deal with the more general situation where a curve is
described parametrically by means of the equations

x = x(t), y = y(t), a ≤ t ≤ b.

We shall assume that we have a smooth curve, for which the functions
x(t) and y(t) have continuous derivatives for a ≤ t ≤ b. On the graph
represented by these parametric equations, we divide the curve into small
pieces by means of a sequence of points P0 , P1 , P2 , . . . , Pn , specified by
the sequence of values of the parameter, given by

a = t0 < t1 < t2 < · · · < tn = b.

This is shown in the left-hand diagram in Figure 5.1. In the right-

hand diagram we have isolated one small piece of the graph lying be-
tween two successive points. The length of that small piece of arc is
denoted conventionally by ds.

42
BRAC
University MAT 120

Figure 5.1: Arc length

In the right-hand diagram, if the piece of arc is very small then the
gradient will not change much along the arc because the parametric
functions are assumed to have continuous derivatives. So the length ds
will be approximately equal to that of the line segment Pi−1 Pi . Using
Pythagoras’ Theorem gives

(Pi−1 Pi )2 = (xi − xi−1 )2 + (yi − yi−1 )2

= (x(ti ) − x(ti−1 ))2 + (y(ti ) − y(ti−1 ))2
= (x0 (ci )(ti − ti−1 ))2 + (y 0 (ci )(ti − ti−1 ))2

using the Mean Value Theorem, we now deduce that,

ds2 ≈ (Pi−1 Pi )2 = (x0 (ci )2 + y 0 (di )2 )(ti − ti−1 )2

Taking square roots gives

p
ds ≈ (x0 (ci )2 + y 0 (di )2 )(ti − ti−1 )

The total arc length is therefore given by

n p
X
L≈ (x0 (ci )2 + y 0 (di )2 )(ti − ti−1 )
i=1

This is an approximating sum to an integral, as outlined in Section

1.1, and so we finally we have the formula
Z bp
L= (x0 (t)2 + y 0 (t)2 )dt
a

43 of 110
BRAC
University MAT 120

As a special case, suppose that the curve is the graph of the func-
tion specified by y = f (x), p ≤ x ≤ q. This can be expressed in the
parametric form x = t, y = f (t), p ≤ t ≤ q. We then have

dy
x0 (t) = 1, y 0 (t) = f 0 (t) = f 0 (x) = .
dx
So the formula becomes
s 2
Z qp Z q 
0 2
dy
L= (1 + f (x) )dx = 1+ dx
p p dx

Example 5.1.1. Find the length of the curve given by x = t2 , y =

2t3 , −1 ≤ t ≤ 1, shown in Figure 5.2.

Figure 5.2: Graph of x = t2 , y = 2t3 , −1 ≤ t ≤ 1

Using the formula derived above gives

Z 1 Z 1  1
p p 2 1 3
L= 2 4
4t + 36t dt = 2t 1 + 9t2 dt = (1 + 9t2 ) 2 =0
−1 −1 3 9 −1

44 of 110
BRAC
University MAT 120

This is clearly wrong. The curve does not have zero length! The
problem is that we were not sufficiently careful with the square root
when we factored out t2 from √under the square root in the first integral.
It is not true in general that t2 = t, especially in this case where the
integrand must be positive for all values of t because it represents a
length. A correct version of the calculations is as follows.
Z 1 p Z p 1
L= 2 4
4t + 36t dt = 2|t| 1 + 9t2 dt
−1 −1
Z 1 p  1
2 1 2 3
=2 2t 1 + 9t2 dt = 2 (1 + 9t ) 2
0 3 9 0
4  3

= 10 2 − 1 .
27

Example 5.1.2. Find the length of the curve given by y = x2 − ln8x (1 ≤

x ≤ 2).
In this example we need to use the cartesian formula for arc length.
s 2 s 2
Z 2  Z 2
dy 1
L= 1+ dx = 1 + 2x − dx
1 dx 1 8x
Z 2r Z 2r
1 1 1 1
= 1 + 4x2 + − dx = 4x2+ 0 dx
1 64x2 2 1 64x2 2
s 2
Z 2  Z 2 
1 1
= 2x + dx = 2x + dx
1 8x 1 8x
 2
2 ln x ln 2
= x + =3+
8 1 8
In passing from line 2 to line 3 we had to be able to spot that 4x2 +
1 1
64x2 + 2 is a perfect square.

45 of 110
BRAC
University MAT 120

5.2 Surface Area of Revolution

If we take a curve and rotate it about a line we will obtain a curved
surface of revolution. For example if we rotate the semicircle in Figure
5.3 about its base through a complete rotation of 2π we will generate
a sphere. In general of course we consider not just a semicircle but
an arbitrary smooth curve, which we represent in parametric form as
x = x(t), y = y(t), a ≤ t ≤ b. To obtain a formula for such a surface
area we divide the curve into small pieces as we did in Section 1.1.

Figure 5.3: Surface of revolution

In Figure 5.3 we have shown the effect of rotating one such piece of
arc. To find the area generated we imagine cutting and unwrapping the
section of surface shown. This will give us a piece of “ribbon” approxi-
mately rectangular in shape. Its length will be the circumference, 2πr,
of the circle generated by rotating a point on the small piece of arc. Its
width will be the length ds of the piece of arc, for which we found an
approximate formula in Section 5.1. So, using the same notation as in
Section 5.1, the small piece of surface area will be approximately
p
2πr.ds ≈ 2π|y(t)| (x0 (ti )2 ) + y 0 (ti )2 )(ti − ti−1 ).

46 of 110
BRAC
University MAT 120

Note that we have used |y(t)| as the value of r to allow for the fact
that y(t) might be negative for some values of t. So the total surface
area will be given by

n
X p
S≈ 2π|y(t)| (x0 (ti )2 ) + y 0 (ti )2 )(ti − ti−1 ).
i=1

This is an approximating sum for an integral, and so the formula for

the surface area of rotation is
Z b p
S = 2π |y(t)| (x0 (ti )2 ) + y 0 (ti )2 )(ti − ti−1 )dt.
a

If the curve is the graph of a function, y = f (x), p ≤ x ≤ q, then it

can be expressed in the parametric form x = t, y = f (t), p ≤ t ≤ q, as
in Section 5.1. The argument in that section shows that in this case
s  2
Z q Z q
p dy
S = 2π |f (x)| 1 + f 0 (x)2 )dx = 2π |f (x)| 1 + )dx.
p p dx

Example 5.2.1. Find the area of the surface obtained by rotating the
curve y = x3 , 0 ≤ x ≤ 1 about the x − axis.
We use the cartesian formula to obtain

pZ 1 Z 1 p
3
S = 2π x 1 + (3x2 )2 dx = 2π x3 1 + 9x4 dx
0 1 0
1 4 32 π  3

= 2π (1 + 9x ) = 10 − 1 .
2
54 0 27
Note that because x3 ≥ 0 in the interval we do not need the modulus
signs.

47 of 110
BRAC
University MAT 120

Example 5.2.2. Verify the formula for the surface of a sphere of radius
a.
In this example we shall use the parametric formula. The sphere can
be obtained by rotating the semicircle given by x2 + y 2 = a2 , y ≥ 0,
about the x − axis. We parameterise the semicircle by x = a cos t, y =
a sin t, 0 ≤ t ≤ π. Since y(t) ≥ 0 for all t in the interval we do not need
the modulus signs in the formula, and so we have
Z π p
S = 2π a sin t a2 sin2 t + a2 cos2 tdt
0Z
π
2
= 2πa sin tdt = 2πa2 [− cos t]π0 = 4πa2 .
0

5.3 Volumes by Slicing

Consider a prism, whose uniform cross section can be any shape, for
example triangular, rectangular, or circular (a circular prism is of course
a cylinder). Its volume is equal to the cross-sectional area multiplied by
its length. We can extend this idea to solids where the cross section
is not uniform. So suppose we have a solid, contained between planes
x = a and x = b, and that we can calculate the area A(c) of the cross
section made by the plane x = c. A good way to imagine this is to think
of a sliced loaf of bread, where the cross section will vary from one end
to the other. We now slice up the solid by means of a sequence of planes

x = x0 (= a), x = x1 , x = x2 , . . . , x = xn (= b).

Assuming that the slices are sufficiently thin, and that A(x) is a
continuous function of x, the volume of the slice contained between
x = xi−1 and x = xi will be approximately A(xi )(xi − xi−1 ). The total
volume will therefore be approximately the sum of these slice volumes,
so

n
X
V ≈ A(xi )(xi − xi−1 ).
i=1

48 of 110
BRAC
University MAT 120

This is an approximating sum for an integral, as in Section 1.1, and

so we have
Z b
V = A(x)dx.
a

Naturally for this to be useful we have to be able to find A(x) for the
solid we are concerned with, and we do this is the next example.

Example 5.3.1. A tetrahedron is formed by cutting a corner from a

cube by means of a plane. Find its volume.
We place the corner of the cube O at the origin, with the sides of the
cube meeting at that corner along the coordinate axes. Let the cutting
plane meet the coordinate axes at x = a, y = b, z = c respectively.
Figure 5.4 shows the tetrahedron with a triangular cross section P QR
made by a plane x = p. We therefore need to calculate the area of this
triangle.

Figure 5.4: Tetrahedron for Example 5.3.1

Some elementary calculation with similar triangles tells us that

c(a − p) b(a − p)
PR = , PQ = ,
a a
so the area of triangle P QR is given by

P R.P Q bc(a − p)2

A(p) = = .
2 2a2

49 of 110
BRAC
University MAT 120

Therefore the volume of the tetrahedron is given by

a a a
bc(a − x)2 (a − x)3
Z Z 
bc abc
V = A(x)dx = 2
dx = 2 − =
0 0 2a 2a 3 0 6

5.4 Volumes of Revolution

If we imagine rotating a plane region about a line in that plane then a
solid will be generated with circular cross sections. For example if we
rotate a rectangular region about one of its edges we obtain a cylinder. A
right-angled triangle rotated about one of its shorter edges will generate
a cone. A semicircular region rotated about its diameter will generate a
sphere. In this section we shall investigate two methods of calculating
the volume of a solid of revolution.

5.4.1 The Disc Method

If we rotate a rectangular region about a line parallel to one of its edges
which does not intersect the rectangle then we will generate a solid cylin-
der with a cylindrical hole through its centre. This is a simple example
of a general type of solid obtained by rotating the region specified by

f (x) ≤ y ≤ g(x), a ≤ x ≤ b

about the line y = c, shown in Figure 5.5.

As before we calculate the cross sectional area of the solid and use
the integral formula derived above. We subdivide the interval a ≤ x ≤ b
as in Section 1.1, and use this to divide the region into strips parallel to
the y − axis. We have shown one of these in Figure 5.5, together with
the solid obtained by rotating this strip about the line y = c. It looks
like a “washer”, i.e., a disc with a smaller disc removed from its centre.
The cross-sectional area is therefore

πR2 − πr2 = π (g(x) − c)2 − (f (x) − c)2 .

50 of 110
BRAC
University MAT 120

The integral formula tells us that the total volume is given by

Z b
(g(x) − c)2 − (f (x) − c)2 dx


V =π
a

Figure 5.5: Volume of revolution; the disc method

Example 5.4.1. Find the volume of the solid obtained by rotating the
region 0 ≤ y ≤ sin x, 0 ≤ x ≤ π, about (a) the x − axis & (b) the line
y = −1

(a) In this case we have c = 0, f (x) = 0, g(x) = sin x, and so the

volume is given by

π π
π2
Z 
2 π sin 2x
V =π sin x dx = x− =
0 2 x 0 2

(b) Here we have c = −1 and so the volume is given by

Z π
V =π ((sin2 x + 1)2 − (−1)2 ) dx
Z0 π Z π
2 π2
=π sin x dx + π 2 sin x dx = + 4π
0 0 2

51 of 110
BRAC
University MAT 120

5.4.2 The Cylindrical Method

This method applies to the same region

f (x) ≤ y ≤ g(x), a ≤ x ≤ b.

for which we developed the disc method, but this time we rotate
the region about the line x = d, parallel to the y − axis, as shown in
Figure 5.6. In this case the strips produced by subdividing the interval
a ≤ x ≤ b generate cylindrical shells rather than a cross section of the
solid. One has to imagine each of these shells fitting inside the previous
one to form the solid, like some childrens’ toys where plastic beakers fit
inside each other, or like the separate parts of Russian dolls.

In Figure 5.6, suppose that the strip is specified by one of the intervals
of the subdivision, i.e.,

f (x) ≤ y ≤ g(x), xi − 1 ≤ x ≤ xi.

The distance of this strip from the axis of rotation is shown as R,

where we have R = x − d. (In the diagram d is negative, and so
R = x − d > x, as the figure suggests.) On the left of the figure is
shown the cylindrical shell generated by rotating the strip, and we need
to find its volume. If the cylinder is made of some flexible material, we
can cut it parallel to the axis of rotation. We can then open it out and
we will obtain an approximately rectangular slab. Its height will be the
height of the strip, namely g(x) − f (x), its width will be

Figure 5.6: Volume of revolution; the shell method

the circumference of the cylinder, namely 2πR, and its thickness will
be the width of the strip, namely xi − xi−1 . The volume is therefore
approximately

52 of 110
BRAC
University MAT 120

2πR(g(x) − f (x))(xi − xi−1 ) = 2π(x − d)(g(x) − f (x))(xi − xi−1 ).

The total volume is therefore approximately

n
X
2π(x − d)(g(x) − f (x))(xi − xi−1 ),
i=1

which is an approximating sum for an integral, and hence

Z b
V = 2π (x − d)(g(x) − f (x)) dx.
a

Example 5.4.2. Rotate the region in Example 5.4.1 about the line x =
−π, and find the volume of the solid obtained.
The region is 0 ≤ y ≤ sin x, 0 ≤ x ≤ π, and so using the formula
obtained above we have

f (x) = 0, g(x) = sin x, a = 0, b = π, d = −π.

We therefore have
Z π Z π Z π
2
V = 2π (x + π) sin x dx = 2π sin x dx + 2π x sin x dx = 6π 2
0 0 0

Both integrals are straightforward. The second is done by parts, and

the first is a basic integral. This integral is encountered frequently, so
it is worth learning that its value, which corresponds to the area of the
region in this example, is equal to 2.

53 of 110
BRAC
University MAT 120

Example 5.4.3. Let R denote the region contained between the two
√
graphs y = x2 , y = x, 0 ≤ x ≤ 1.Find the volume obtained by
rotating this region about (a) the x − axis, (b) the y − axis.

Figure 5.7: Diagram for Example 5.4.3

The region is shown in Figure 5.7, and we have drawn a strip parallel
to the y − axis. If we rotate the strip round the x − axis this will
generate a washer, and so we can use the disc method. If we rotate it
round the y − axis we will generate a cylinder, so we can use the shell
method.

(a) The volume obtained by rotating round the x − axis is given by

Z 1 1
√
Z
2 2 2
 3π
V =π x − (x ) dx = π (x − x4 ) dx = .
0 0 10

(b) The volume obtained by rotating round the y − axis is given by

Z π √
Z π  3
 3π
V = 2π x( x − x2 ) dx = 2π x −x 2 3
dx = .
0 0 10

Because of the symmetry of the region we can see that the two solids
will in fact be identical in shape and size.

54 of 110
Chapter 6

First Order Differential Equations

6.1 Basic Concepts

DIFFERENTIAL EQUATIONS

A differential equation is an equation involving an unknown function

and Ms derivatives.

The follow ins arc differential equations involving the unknown func-
tion v,

dy
= 5x + 3 (6.1)
dx

2
 2
d y dy
ey 2 + 2 =1 (6.2)
dx dx

d3 y d2 y
4 3 + (sin x) 2 + 5xy = 0 (6.3)
dx dx

3 7 2
d2 y
  
dy dy
+ 3y + y3 = 5x (6.4)
dx2 dx dx

∂ 2y ∂ 2y
−4 2 =0 (6.5)
∂t2 ∂x
A differential equation is an ordinary differential equation (ODH) if
the unknown function depends on only one independent variable. If the

55
BRAC
University MAT 120

unknown function depends on two or more independent variables. the

differential equation is a partial differential equation (PDE). With the
exceptions of Chapters 31 and 34, the primary focus of this hook mil he
ordinary differential equations.

Equations (6.1) through (6.4) are examples, of ordinary differential

equations, since the unknown function Y depends solely on the variable
x Equation (6.5) is a partial differential equation, since Y depends on
both the independent variables t and x.

The order of a differential equation is the order of the highest deriva-

tive appearing in the equation.

Equation (6.1) is a first-order differential equation - (6.2), (6.4), and

(6.5) are second-order differential equations. [Note in (1.4) that the order
of the highest derivative appearing in the equation is two.] Equation
(6.3) is a third order differential equation.

NOTATION

The expressions y 0 , y 00 , y 000 , y (4) , ..., y (n) are often used to represent, re-
spectively, the first, second, third, fourth,..., nth derivatives of y with
respect to the independent variable under consideration. Thus, y 00 rep-
d2 y d2 y
resents if the independent variable is x, but represents 2 if the
dx2 dp
independent variable is p. Observe that parentheses are used in y (n) to
distinguish it from the nth power, y (n) If the independent variable is
time, usually denoted by t, primes are often replaced by dots. Thus,
... dy d2 y d3 y
ẏ, ÿ, and y represent , 2 , and , 3 respectively.
dt dt dt
SOLUTIONS

A solution of a differential equation in the unknown function y and

the independent variable x on the interval I , is a function y(x) that
satisfies the differential equation identically for all x in I .
Example 6.1.1. Is y(x) = c1 sin 2x + c2 cos 2x, where c1 and c2 are
arbitrary constants, a solution of y 00 + 4y = 0 ?
Differentiating y, we find

56 of 110
BRAC
University MAT 120

y 0 = 2c1 cos 2x − 2c2 sin 2x and y 00 = −4c1 sin 2x − 4c2 cos 2x

Hence,
y 00 + 4y = (−4c1 sin 2x − 4c2 cos 2x) + 4(2c1 cos 2x − 2c2 sin 2x)
= (−4c1 + 4c1 ) sin 2x + (−4c2 + 4c2 ) cos 2x
=0

Thus, y(x) = c1 sin 2x + c2 cos 2x satisfies the differential equation

for all values of x and is a solution on the interval (−∞, ∞)
Example 6.1.2. Determine whether y = x2 − 1 is a solution of (y 0 )4 +
y 2 = −1.
Note that the left side of the differential equation must be nonnegative
for every real function y(x) and any x, since it is the sum of terms raised
to the second and fourth powers, while the right side of the equation is
negative. Since no function y(x) will satisfy this equation, the given
differential equation has no solution.

We see that some differential equations have infinitely many solutions

(Example 6.1.1), whereas other differential equations have no solutions
(Example 6.1.2). It is also possible that a differential equation has ex-
actly one solution. Consider (y 0 )4 + y 2 = 0, which for reasons identical
to those given in Example 6.1.2 has only one solution y ≡ 0.

A particular solution of a differential equation is any one solution.

The general solution of a differential equation is the set of all solutions.

Example 6.1.3. The general solution to the differential equation in

Example 6.1.1 can be shown to be (see Chapters 8 and 9) y(x) =
c1 sin 2x + c2 cos 2x. That is, every particular solution of the differ-
ential equation has this general form. A few particular solutions are:
(a)y = 5 sin 2x − 3 cos 2x (choose c1 = 5 and c2 = −3), (b)y = sin 2x
(choose c1 = 1 and c2 = 0), and (c)y ≡ 0 (choose C1 = c2 , = 0).
The general solution of a differential equation cannot always be ex-
pressed by a single formula. As an example consider the differential

57 of 110
BRAC
University MAT 120

equation y 0 + y 2 = 0, which has two particular solutions y = 1/x and

y ≡ 0.

INITIAL-VALUE AND BOUNDARY-VALUE PROBLEMS

A differential equation along with subsidiary conditions on the un-

known function and its derivatives, all given at the same value of the
independent variable, constitutes an initial-value problem. The sub-
sidiary conditions are initial conditions. If the subsidiary conditions are
given at more than one value of the independent variable, the problem is
a boundary-value problem and the conditions are boundary conditions.

A solution to an initial-value or boundary-value problem is a function

y(x) that both solves the differential equation and satisfies all given
subsidiary conditions.

58 of 110
BRAC
University MAT 120

6.2 Classifications of First Order Differential Equa-

tions
MATHEMATICAL MODELS

Mathematical models can he thought of as equations. In this chapter,

and in other parts of this book (sec Chapter 7. Chapter 14 and Chapter
31. for example), we will consider equations which model certain real-
world situations.

For example, when considering a simple direct current (DC) electri-

cal circuit, the equation V = RI models the voltage drop (measured in
volts) across a resistor (measured in ohms), here I is the current (mea-
sured in amperes). This equation is called Ohm’s Law. named in honor
of G. S. Ohm (1787-1854), a German physicist,

Once constructed, some models can he used to predict main physical

situations. for example, weather forecasting, the growth of a tumor, or
the outcome of a roulette wheel, can all be connected with some form
of mathematical modeling.

In this chapter, we consider variables that are continuous and how

differential equations can he used in modeling. Chapter 34 introduces
the idea of difference equations. These are equations in which we con-
sider discrete variables; that is, variables which can lake on only certain
values, such as whole numbers. With few modifications, everything pre-
sented about modeling with differential equations also holds true with
regard to modeling with difference equations.

THE "MODELING CYCLE"

Suppose we have a real-life situation (we want to find the amount

of radio-active material in some element). Research may be able to
model this situation (in the form of a "very difficult" differential equa-
tion). Technology may be used to help us solve the equation (computer
programs give us an answer). The technological answers are then inter-
preted.

QUALITATIVE METHODS

59 of 110
BRAC
University MAT 120

To build a model can be a long and arduous process; it may take many
years of research. Once they are formulated, models may be virtually
impossible to solve analytically. Then the researcher has two options:

• Simplify, or "tweak", the model so that it can be dealt with in a

more manageable way. This is a valid approach, provided the sim-
plification does not overly compromise the "real-world" connection,
and therefore, its usefulness
• Retain the model as is and use other techniques, such as numerical
or graphical methods (see Chapter 18, Chapter 19, and Chapter
20). This represents a qualitative approach. While we do not pos-
sess an exact, analytical solution, we do obtain some information
which can shed some light on the model and its application. Tech-
nological tools can be extremely helpful with this approach (see
Appendix B).

60 of 110
BRAC
University MAT 120

6.3 Separable First Order Differential Equations

GENERAL SOLUTION

The solution to the first-order separable differential equation (see

Chapter 3)

A(x)dx + B(y)dy = 0 (6.6)

is

Z Z
A(x)dx + B(y)dy = c (6.7)

where c represents an arbitrary constant.

The integrals obtained in Eq.(6.7) may be. for all practical purposes,
impossible lo evaluate. In such eases. numerical techniques (see Chap-
ters 18, 14. 20) are used to obtain an approximate solution. Even if
the indicated integrations in (6.7) can be performed, it may not be al-
gebraically possible lo solve for y explicitly in terms of x. In that case,
the solution is left in implicit form.

The solution to the initial-value problem

A(x)dx + B(y)dy = 0; y(x0 ) = y0 (6.8)

can be obtained, as usual. by first using Eq.(6.7) to solve the dif-
ferential equation and then applying The initial condition directly lo
evaluate r.

Alternatively, the solution lo Eq.(6.8) can be obtained from

Z x Z y
A(x)dx + B(y)dy = 0 (6.9)
x0 y0

Equation (6.9). however, may not determine the solution of (6.8)

uniquely; that is. (6.9) may have many solutions, of which only one will
satisfy the initial-value problem.

61 of 110
BRAC
University MAT 120

REDUCTION OF HOMOGENEOUS EQUATIONS

The homogeneous differential equation

dy
= f (x, y) (6.10)
dx
having the property that f (tx, ty) = f (x, y) (see Chapter 3) can be
transformed into a separable equation by making the substitution

y = xv (6.11)
along with its corresponding derivative

dy dv
=v+x (6.12)
dx dx
The resulting equation in the variables v and x is solved as a separable
differential equation; the required solution to Eq.(6.10) is obtained by
back substitution.

Alternatively, the solution to (6.10) can be obtained by rewriting the

differential equation as

dx 1
= (6.13)
dy f (x, y)
and then substituting

x = yu (6.14)
and the corresponding derivative

dx du
=u+y (6.15)
dy dy
into Eq.(6.13). After simplifying, the resulting differential equation
will be one with variables (this time, u and y) separable.

Ordinarily, it is immaterial which method of solution is used. Occa-

sionally, however, one of the substitutions (6.11) or (6.14) is definitely

62 of 110
BRAC
University MAT 120

superior to the other one. In such cases, the better substitution is usu-
ally apparent from the form of the differential equation itself.
x+1
Example 6.3.1. Solve y 0 =
y4 + 1
This equation, in differential form, is (x + 1)dx + (−y 4 − 1)dy = 0
which is separable. Its solution is

Z Z
(x + 1)dx + (−y 4 − 1)dy = c

or, by evaluating,

x2 y5
+x− −y+c
2 5
Since it is impossible algebraically to solve this equation explicitly
for y, the solution must be left in its present implicit form.
2xy
Example 6.3.2. Solve y 0 =
x2 − y 2
Phis differential equation is not separable. Instead it has the form
0
y = f (x, y), with

2xy
f (x, y) =
x2 − y 2
where

2(tx)(ty) t2 (2xy) 2xy

f (tx, ty) = = = = f (x, y)
(tx)2 − (ty)2 t2 (x2 − y 2 ) x2 − y 2
so it is homogenous. Substituting Eqs.(6.11) and (6.12) into the
differential equation as originally given, we obtain

dv 2x(xy)
v+x = 2
dx x − (xy)2
which can be algebraically simplified to

63 of 110
BRAC
University MAT 120

dv v(v 2 + 1)
x =− 2
dx v −1
or

1 v2 − 1
dx + dv = 0 (6.16)
x v(v 2 + 1)
Using partial fractions, we can expand (6.16) to

 
1 1 2v
dx + − + 2 dv = 0 (6.17)
x v v +1
The solution to this separable equation is found by integrating both
sides of (6.17). Doing so, we obtain ln |x| − ln |v| + ln(v 2 + 1) = c, which
can be simplified to

x(v 2 + 1) = kv (c = ln |k|) (6.18)

Substituting v = y/x into (6.18), we find the solution of the given

differential equation is x2 + y 2 = ky.

64 of 110
BRAC
University MAT 120

6.4 Exact First Order Differential Equations

DEFINING PROPERTIES

A differential equation

M (x, y)dx + N (n, y)dy = 0 (6.19)

is exact if there exists a function g(x, y) such that

dg(x, y) = M (x, y)dx + N (n, y)dy (6.20)

Test for exactness: If M(x, y) and N(x, y) are continuous functions

and have continuous first partial derivatives on some rectangle of the
xy-plane. then (6.19) is exact if and only if

∂M (x, y) ∂N (x, y)
= (6.21)
∂y ∂x

METHOD OF SOLUTION

To solve Eq.(6.19). assuming that it is exact, first solve the equations

∂g(x, y)
= M (x, y) (6.22)
∂x

∂g(x, y)
= N (x, y) (6.23)
∂y
for g(x, y). The solution to (6.19) is then given implicitly In

g(x, y) = c (6.24)

where c represents an arbitrary constant.

Equation (6.24) is immediate from Eqs. (6.19) and (6.20). If (6.20)

is substituted into If (6.19) we obtain dg(x, y(x)) = 0. Integrating this

65 of 110
BRAC
University MAT 120

R
equation
R (note that we can write 0 as 0 dx), we have dg(x, y(x)) =
0 dx, which, in turn, implies (6.24).

INTEGRATING FACTORS

In general, Eq. (6.19) is not exact. Occasionally, it is possible to

transform (6.19) into an exact differential equation by a judicious mul-
tiplication. A function I(x, y) is an integrating factor for (6.19) if the
equation

I(x, y)[M (x, y) + N (x, y)] = 0 (6.25)

is exact. A solution to (6.19) is obtained by solving the exact differ-

ential equation defined by (6.25). Some of the more common integrating
factors are displayed in Table 6.4 and the conditions that follow:
 
1 ∂M ∂N
if − ≡ g(x), a function of x alone, then
N ∂y ∂x
R
g(x)dx
I(x, y) = e (6.26)
 
1 ∂M ∂N
if − ≡ g(x), a function of y alone, then
M ∂y ∂x
R
− h(y)dy
I(x, y) = e (6.27)

66 of 110
BRAC
University MAT 120

Group of terms Integrating factor I(x, y) Exact differential dg(x, y)

x dy−y dx
y dx − x dy − x12 x2
= d( xy )

1 y dx−x dy
y dx − x dy y y2
= d( xy )

x dy−y dx
y dx − x dy 1
− xy xy
= d(ln xy )

x dy−y dx
y dx − x dy 1
− x2 +y 2 x2 +y 2
= d(arctan xy )

1 y dx+x dy
y dx + x dy xy xy
= d(ln xy)
h i
1 y dx+x dy −1
y dx + x dy (xy)n
,n >1 (xy)n
= d (n−1)(xy)n−1

1 y dy+x dx
= d 21 ln x2 + y 2
 
y dx + x dy x2 +y 2 x2 +y 2
h i
1 y dy+x dx −1
y dx + x dy (x2 +y 2 )n
,n >1 (x2 +y 2 )n
= d 2(n−1)(x2 +y2 )n−1

ay dx + bx dy xa−1 y b−1 xa−1 y b−1 (ay dx + bx dy) = d(xa y b )

Table 6.1: Common integrating factors

if M = yf (xy) and n = xg(xy), then

1
I(x, y) = (6.28)
xM − yN
In general, integrating factors are difficult to uncover. If a differential
equation does not have one of the forms given above, then a search for
an integrating factor likely will not be successful, and other methods of
solution are recommended
Example 6.4.1. Determine whether the differential equation 2xydx +
(1 + x2 )dy = 0 is exact. and solve it
This equation has the form of Eq. (6.19) with M (x, y) = 2xy and
N (x, y) = 1 + x2 . Since ∂M/∂y = ∂N/∂x = 2x, the differential
equation is exact.

This equation was shown to be exact. We now determine a function

g(x, y) that satisfies Eqs. (6.22) and (6.23). Substituting M (x, y) = 2xy

67 of 110
BRAC
University MAT 120

into (6.22), we obtain ∂g/∂x = 2xy. Integrating both sides of this

equation with respect to x, we find
Z Z
∂g
dx = 2xy dx
∂x
or,

g(x, y) = x2 y + h(y) (6.29)

Note that when integrating with respect to x, the constant (with
respect to x) of integration can depend on y.
We now determine h(y). Differentiating (6.29) with respect to y,
we obtain ∂g/∂y = x2 + h0 (y). Substituting this equation along with
N (x, y) = 1 + x2 into (6.23), we have

x2 + h0 (y) = 1 + x2 or h0 (y) = 1
Integrating this last equation with respect to y, we obtain h(y) =
y + c1 (c1 = constant). Substituting this expression into (6.29) yields

g(x, y) = x2 y + y + c1
The solution to the differential equation, which is given implicitly by
(6.24) as g(x, y) = c, is

x2 y + y = c2 (c2 = c − c1 )
Solving for y explicitly, we obtain the solution as y = c2 /(x2 + 1).
0 2 + yexy
Example 6.4.2. Solve y =
2y − xexy
Rewriting this equation in differential form, we obtain

(2 + yexy )dx + (xexy − 2y)dy = 0

Here, M (x, y) = 2 + yexy and N (x, y) = xexy − 2y and, since
∂M/∂y = ∂N/∂x = exy + xyexy the differential equation is exact. Sub-
stituting M (x, y) into (6.22), we find ∂g/∂x = 2+yexy ; then integrating
with respect to x, we obtain

68 of 110
BRAC
University MAT 120

Z Z
∂g
dx = [2 + yexy ]dx
∂x
or

g(x, y) = 2x + exy + h(y) (6.30)

To find h(y), first differentiate (6.30) with respect to y, obtaining

∂g/∂y = xexy + h0 (y); then substitute this result along with N (x, y)
into (6.23) to obtain

xexy + h0 (y) = xexy − 2y or h0 (y) = −2y

It follows that h(y) = −y 2 + c1 . Substituting this h(y) into (6.30),

we obtain

g(x, y) = 2x + exy − y 2 + c1

The solution to the differential equation is given implicitly by (6.24)

as

2x + exy − y 2 = c2 (c2 = c − c1 )

69 of 110
BRAC
University MAT 120

6.5 Linear First Order Differential Equations

METHOD OF SOLUTION

A first-order linear differential equation has the form (see Chapter 3)

y 0 + p(x)y = q(x) (6.31)

An integrating factor for Eq. (6.31) is

R
p(x)dx
I(x) = e (6.32)

which depends only on y and is independent of v. When both sides

of (6.31) are multiplied by I(x), the resulting equation

I(x)y 0 + p(x)I(x)y = I(x)q(x) (6.33)

is exact. This equation can he solved by the method described in

Chapter 5. A simpler procedure is to rewrite (6.33) as

d(yI)
= Iq(x) (6.34)
dx
integrate both sides of this last equation with respect to x, and then
solve the resulting equation for Y .

REDUCTION OF BERNOULLI EQUATIONS

A Bernoulli differential equation has the form

y 0 + p(x)y = q(x)y n (6.35)

where n is a real number. The substitution

z = y 1−n (6.36)

transforms (6.35) into a linear differential equation in ihe unknown

function z(x).

70 of 110
BRAC
University MAT 120

Example 6.5.1. Find an integrating factor for y 0 − 3y = 6.

The differential equation has the form of Eq. 6.31), with p(x) = −3
and q(x) = 6, and is linear. Here

Z Z
p(x)dx = −3dx = −3x

so (6.32) becomes

R
I(x) = e p(x)dx
= e−3x
dz
Example 6.5.2. Solve − xz = −x
dx
This is a linear differential equation for the unknown function z(x). It
has the form of Eq. (6.31) with y replaced by z and p(x) = q(x) = −x.
The integrating factor is

2
R
I(x) = e (−x)dx
= e−x /2

Multiplying the differential equation by I(x), we obtain

2
/2 dz 2 2
e−x − xe−x /2
z = −xe−x /2
dx
or

dz −x2 /2 2
(ze ) = −xe−x /2
dx
Upon integrating both sides of this last equation, we have

2 2
ze−x /2
= −e−x /2
+c

whereupon

2
z(x) = cex /2
+1

71 of 110
BRAC
University MAT 120

6.6 Applications of First Order Differential Equa-

tions
GROWTH AND DECAY PROBLEMS

Let N (t) denote the amount of substance (or population) that is

either growing or decaying. It’ we assume that dN/dt. the lime rate
of change of this amount of substance, is proportional to the amount of
substance present. Then dN/dt = kN . or

dN
− kN = 0 (6.37)
dt
where k is the constant of proportionality.

We are assuming that N (t) is a differentiable, hence continuous, func-

tion of time. For population problems, where N (t) is actually discrete
and integer-valued, this assumption is incorrect. Nonetheless, (6.37)
still provides a good approximation to the physical laws governing such
a system.

TEMPERATURE PROBLEMS

Newton’s law of cooling, which is equally applicable lo healing, states

that the time rate of change of the temperature of a body is propor-
tional to the temperature difference between the body and its surround-
ing medium. Let T denote the temperature of the body and let Tm
denote the temperature of the surrounding medium. Then the time rate
of change of Ihe temperature of the body is dT /dt, and Newton’s law of
cooling can be formulated as dT /dt = −k(T − Tm ). or as

dT
+ kT = kTm (6.38)
dt
where k is a positive constant of proportionality Once k is chosen
positive, the minus sign is required in Newton’s law to make dT /dt
negative in a cooling process, when T is greater than Tm . and positive
in a heating process, when TMs less than Tm

72 of 110
BRAC
University MAT 120

FALLING BODY PROBLEMS

Consider a vertically falling body of mass m that is being influenced

only by gravity g and an air resistance that is proportional to the velocity
of the body. Assume that both gravity and mass remain constant and,
for convenience, choose the downward direction as the positive direction.
Newton’s second law of motion: The net force acting on a body
is equal to the time rate of change of the momentum of the body; or, for
constant mass,

dv
F =m (6.39)
dt
where F is the net force on the body and v is the velocity of the body,
both at time t.

For the problem at hand, there are two forces acting on the body: (1)
the force due to gravity given by the weight w of the body, which equals
mg, and (2) the force due to air resistance given by −kv, where k > 0
is a constant of proportionality. The minus sign is required because this
force opposes the velocity; that is, it acts in the upward, or negative,
direction. The net force F on the body is, therefore, F = mg − kv.
Substituting this result into (6.39), we obtain

dv
mg − kv = m
dt
or

dv k
+ v=g (6.40)
dt m
as the equation of motion for the body.

If air resistance is negligible or nonexistent, then k = 0 and (6.40)

simplifies to

dv
=g (6.41)
dt
When k > 0, the limiting velocity vl is defined by

73 of 110
BRAC
University MAT 120

mg
vl = (6.42)
k
Caution: Equations (6.40), (6.41), and (6.42), are valid only if the
given conditions are satisfied. These equations are not valid if, for ex-
ample, air resistance is not proportional to velocity but to the velocity
squared, or if the upward direction is taken to be the positive direction.
Consider a tank which initially holds V0 gal of brine that contains
a lb of salt. Another brine solution, containing b lb of salt per gallon,
is poured into the tank at the rate of e gal/min while, simultaneously,
the well-stirred solution leaves the tank at the rate of f gal/min . The
problem is to find the amount of salt in the tank at any time t.

Let Q denote the amount (in pounds) of salt in the tank at any time
t. The time rate of change of Q, dQ/dt, equals the rate at which salt
enters the tank minus the rate at which salt leaves the tank. Salt enters
the tank at the rate of be lb/min. To determine the rate at which salt
leaves the tank, we first calculate the volume of brine in the tank at any
time t, which is the initial volume V0 plus the volume of brine added et
minus the volume of brine removed f t . Thus, the volume of brine at
any time is

V0 + et − f t (6.43)
The concentration of salt in the tank at any time is Q/(V0 + et − f t),
from which it follows that salt leaves the tank at the rate of
 
Q
f lb/min
V0 + et − f t
Thus,
 
dQ Q
= be − f
dt V0 + et − f t
or
 
dQ Q
+f = be (6.44)
dt V0 + et − f t

74 of 110
BRAC
University MAT 120

ELECTRICAL CIRCUITS

The basic equation governing the amount of current I (in amperes) in

a simple RL circuit consisting of a resistance R (in ohms), an inductor L
(in henries), and an electromotive force (abbreviated emf) E (in volts)
is

dI R E
+ I= (6.45)
dt L L
For an RC circuit consisting of a resistance, a capacitance C (in
farads), an emf, and no inductance, the equation governing the amount
of electrical charge q (in coulombs) on the capacitor is

dq 1 E
+ q= (6.46)
dt RC r
The relationship between q and I is

dq
i= (6.47)
dt

ORTHOGONAL TRAJECTORIES

Consider a one-parameter family of curves in the xy-plane defined by

F (x, y, c) = 0 (6.48)
where c denotes the parameter. The problem is to find another one-
parameter family of curves, called the orthogonal trajectories of the fam-
ily (6.48) and given analytically by

G(x, y, k) = 0 (6.49)
such that every curve in this new family (6.49) intersects at right
angles every curve in the original family (6.48).

We first implicitly differentiate (6.48) with respect to x, then elimi-

nate c between this derived equation and (6.48). This gives an equation

75 of 110
BRAC
University MAT 120

connecting x, y, and y 0 , which we solve for y 0 to obtain a differential

equation of the form

dy
= f (x, y) (6.50)
dx
The orthogonal trajectories of (6.48) are the solutions of

dx 1
x= (6.51)
dy f (x, y)
For many families of curves, one cannot explicitly solve for dy/dx and
obtain a differential equation of the form (6.50). We do not consider such
curves in this book.
Example 6.6.1. What constant interest rate is required if an initial
deposit placed into an account that accrues interest compounded contin-
uously is to double its value in six years?
The balance N (t) in the account at any time t is governed by (6.37)

dN
− kN = 0
dt
which has as its solution

N (t) = cekt (6.52)

We are not given an amount for the initial deposit, so we denote it

as N0 . At t = 0, N (0) = N0 , which when substituted into (6.52) yields

N0 = cek(0) = c

and (6.52) becomes

N (t) = N0 ekt (6.53)

We seek the value of k for which N = 2N0 when t = 6. Substituting

these values into (6.53) and solving for k, we find

76 of 110
BRAC
University MAT 120

2N0 = N0 ek∗6
e6k = 2
6k = ln |2|
1
k= ln |2| = 0.1155
6
Example 6.6.2. An RL circuit has an emf given (in volts) by 3 sin 2t, a
resistance of 10 ohms, an inductance of 0.5 henry, and an initial current
of 6 amperes. Find the current in the circuit at any time t.
Here, E = 3 sin 2t, R = 10, and L = 0.5; hence (6.45) becomes

dI
+ 20I = 6 sin 2t
dt
This equation is linear, with solution (see section 6.5)
Z Z
20t
d(Ie ) + 6e20t sin 2t dt

Carrying out the integrations (the second integral requires two inte-
grations by parts), we obtain

30 3
I = ce−20t + sin 2t − cos 2t
101 101
At t = 0, 1 = 6; hence,

30 3 3
6 = ce−20∗0 + sin 2 ∗ 0 − cos 2 ∗ 0 or 6 = c −
101 101 101
whence c = 609/101. The current at any time t is

609 −20t 30 3
I= e + sin 2t − cos 2t
101 101 101
the current is the sum of a transient current, here (609/101)e−20t ,
and a steady-state current,

30 3
sin 2t − cos 2t
101 101

77 of 110
BRAC
University MAT 120

6.7 Graphical & Numerical Methods of Solving First

Order Differential Equations
QUALITATIVE METHODS

In Chapter 2, we touched upon the concept of qualitative methods

regarding differential equations; that is, techniques which are used when
analytical solutions are difficult or virtually impossible to obtain. In
this chapter, and in the two succeeding chapters, we introduce several
qualitative approaches in dealing with differential equations.

DIRECTION FIELDS

Graphical methods produce plots of solutions to first-order differen-

tial equations of the form

y 0 (x, y) (6.54)
where the derivative appears only on the left side of the equation.
Example 6.7.1.
(a) For the problem y 0 = −y + x + 2, we have f (x, y) = −y + x + 2.
(b) For the problemy 0 = y 2 + 1, we have f (x, y) = y 2 + 1.
(c) For the problem y 0 = 3, we have f (x, y) = 3. Observe that in a
particular problem, f (x, y) may be independent of x, of y, or of x
and y.
Equation (6.54) defines the slope of the solution curve y(x) at any
point (x, y) in the plane. A line element is a short line segment that
begins at the point (x, y) and has a slope specified by (6.54); it represents
an approximation to the solution curve through that point. A collection
of line elements is a direction field. The graphs of solutions to (6.54) are
generated from direction fields by drawing curves that pass through the
points at which line elements are drawn and also are tangent to those
line elements.

If the left side of Eq. (6.54) is set equal to a constant, the graph
of the resulting equation is called an isocline. Different constants de-
fine different isoclines, and each isocline has the property that all line

78 of 110
BRAC
University MAT 120

elements emanating from points on that isocline have the same slope,
a slope equal to the constant that generated the isocline. When they
are simple to draw, isoclines yield many line elements at once which is
useful for constructing direction fields.

EULER’S METHOD

If an initial condition of the form

y(x0 ) = y0 (6.55)

is also specified, then the only solution curve of Eq. (6.54) of interest
is the one that passes through the initial point (X0 , y0 ).

To obtain a graphical approximation to the solution curve of Eqs.

(6.54) and (6.55), begin by constructing a line element at the initial
point (x0 , y0 ) and then continuing it for a short distance. Denote the
terminal point of this line element as (x1 , y1 ). Then construct a second
line element at (x1 , y1 ) and continue it a short distance. Denote the
terminal point of this second line element as (x2 , y2 ). Follow with a third
line element constructed at (x2 , y2 ) and continue it a short distance. The
process proceeds iteratively and concludes when enough of the solution
curve has been drawn to meet the needs of those concerned with the
problem.

If the difference between successive x values are equal, that is, if for
a specified constant h, h = x1 − x0 = x2 − x1 = x3 − x2 = ..., then the
graphical method given above for a first-order initial-value problem is
known as Euler’s method. It satisfies the formula

yn+1 = yn + hf (xn , yn ) (6.56)

for n = 1, 2, 3, .... This formula is often written as

yn+1 = yn + hyn0 (6.57)

where

79 of 110
BRAC
University MAT 120

yn0 = f (xn , yn ) (6.58)

as required by Eq. (6.54).

STABILITY

The constant h in Eqs. (6.56) and (6.57) is called the step-size, and its
value is arbitrary. In general, the smaller the step-size, the more accurate
the approximate solution becomes at the price of more work to obtain
that solution. Thus, the final choice of h may be a compromise between
accuracy and effort. If h is chosen too large, then the approximate
solution may not resemble the real solution at all, a condition known
as numerical instability. To avoid numerical instability, Euler’s method
is repeated, each time with a step-size one half its previous value, until
two successive approximations are close enough to each other to satisfy
the needs of the solver.

80 of 110
Chapter 7

Higher Order Differential Equations

7.1 Linear Differential Equations

LINEAR DIFFERENTIAL EQUATIONS

An nth-order linear differential equation has the form

bn (x)y (n) + bn−1 (x)y (n−1) + ... + b2 (x)y 00 + b1 (x)y 0 + b0 (x)y = g(x)
(7.1)
where g(x) and the coefficients bj (x) (j = 0, 1, 2, ..., n) depend solely
on the variable x. In other words, they do not depend on y or on any
derivative of y,

If g(x) ≡ 0, then Eq (7.1) is homogeneous; if not, (7.1) is nonho-

mogeneous. A linear differential equation has constant coefficients if
all the coefficients bj (x) in (7.1) are constants; if one or more of these
coefficients is not constant, (7.1) has variable coefficients.
Theorem 7.1.1. Consider the initial-value problem given by the linear
differential equation (7.1) and the n initial conditions

y(x0 ) = c0 , y 0 (x0 ) = c1 , y 00 (x0 ) = c2 , ..., y (n−1) (x0 ) = cn−1 (7.2)

If g(x) and bj (x) (j = 0, 1, 2, ..., n) are continuous in some interval I

containing x0 and if bn (x) 6= 0 in I, then the initial-value problem given
by (7.1) and (7.2) has a unique (only one) solution defined throughout
I.

81
BRAC
University MAT 120

When the conditions on bn (x) in Theorem 7.1.1 hold, we can divide

Eq (7.1) by bn (x) to get

y (n) + an−1 (x)y (n−1) + ... + a2 (x)y 00 + a1 (x)y 0 + a0 (x)y = φ(x) (7.3)

where aj (x) = bj (x)/bn (x) (j = 0, 1, 2, ..., n−1) and φ(x) = g(x)/bn (x)

Let us define the differential operator L(y) by

L(y) ≡ y (n) + an−1 (x)y (n−1) + ... + a2 (x)y 00 + a1 (x)y 0 + a0 (x)y (7.4)

where ai (x) (i = 0, 1, 2, ..., n − 1) is continuous on some interval of

interest. Then (7.3) can be rewritten as

L(y) = φ(x) (7.5)

and, in particular, a linear homogeneous differential equation can he

expressed as

L(y) = 0 (7.6)

LINEARLY INDEPENDENT SOLUTIONS

A set of functions {y1 (x), y2 (x), ..., yn (x)} is linearly dependent on

a ≤ x ≤ b if there exist constants c1 , c2 , ... , cn , not all zero, such that

c1 y1 (x) + c2 y2 (x) + ... + cn yn (x) ≡ 0 (7.7)

on a ≤ x ≤ b
Example 7.1.1. The set {x, 5x, 1, sin x} is linearly dependent on [−1, 1]
since there exist constants c1 = −5, c2 = 1, c3 = 0, and c4 = 0, not all
zero, such that (7.7) is satisfied. In particular,

−5 · x + 1 · 5x + 0 · 1 + 0 · sin x ≡ 0

82 of 110
BRAC
University MAT 120

Note that c1 = c2 = ... = cn = 0 is a set of constants that always

satisfies (7.7). A set of functions is linearly dependent if there exists
another set of constants, not all zero, that also satisfies (7.7). If the
only solution to (7.7) is c1 = c2 = ... = cn = 0, then the set of functions
{y1 (x), y2 (x), ..., yn (x)} is linearly independent on a ≤ x ≤ b.

Theorem 7.1.2. The nth-order linear homogeneous differential equa-

tion L(y) = 0 always has n linearly independent solutions. If y1 (x),
y2 (x), ..., yn (x) represent these solutions, then the general solution of
L(y) = 0 is

y(x) = c1 y1 (x) + c2 y2 (x) + ... + cn yn (x) (7.8)

where c1 , c2 , ... , cn denote arbitrary constants.

THE WRONSKIAN

The Wronskian of a set of functions {z1 (x), z2 (x), ..., zn (x)} on the
interval a ≤ x ≤ b, having the property that each function possesses
n − 1 derivatives on this interval, is the determinant

z1 z2 ... zn
z10 z20 ... zn0
W (z1 , z2 , ..., zn ) = z100 z200 ... zn00 (7.9)
.. .. ..
. . .
(n−1) (n−1) (n−1)
z1 z2 ... zn

Theorem 7.1.3. If the Wronskian of a set of n functions defined on the

interval a ≤ x ≤ b is nonzero for at least one point in this interval, then
the set of functions is linearly independent there. If the Wronskian is
identically zero on this interval and if each of the functions is a solution
to the same linear differential equation, then the set of functions is
linearly dependent.

Caution: Theorem 7.1.3 is silent when the Wronskian is identically

zero and the functions are not known to be solutions of the same linear

83 of 110
BRAC
University MAT 120

differential equation. In this case, one must test directiy whether Eq.
(7.7) is satisfied.

NON-HOMOGENEOUS EQUATIONS

Let yp denote any particular solution of Eq. (7.5) and let yh (hence-
forth called the homogeneous or complementary solution) represent the
general solution of the associated homogeneous equation L(y) = 0.
Theorem 7.1.4. The general solution to L(y) = φ(x) is

y = yh + yp (7.10)

Example 7.1.2. Two solutions of y 00 − 2y 0 + y = 0 are ex and xex . Is

the general solution y = c1 ex + c2 xex ?
We have

ex xex
W (ex , xex ) = = e2x 6≡ 0
ex ex + xex
It follows, first from Theorem 7.1.3 that the two particular solutions
are linearly independent and then from 7.1.2, that the general solution
is

y = c1 ex + c2 xex

Example 7.1.3. Use the results of Example 7.1.2 to find the general
solution of

y 00 − 2y 0 + y = x2

if it is known that x2 + 4x + 6 is a particular solution.

We have from Example 7.1.2 that the general solution to the associ-
ated homogeneous differential equation is

yh = c1 ex + c2 xex

84 of 110
BRAC
University MAT 120

Since we are given that yp = x2 + 4x + 6, it follows from Theorem

7.1.4 that

y = yh + yp = c1 ex + c2 xex + x2 + 4x + 6

7.2 Second Order Linear Homogeneous Differential

Equations
INTRODUCTORY REMARK

Thus far we have concentrated on first-order differential equations.

We will now turn our attention to the second-order case. After inves-
tigating solution techniques, we will discuss applications of these differ-
ential equations.

THE CHARACTERISTIC EQUATION

Corresponding to the differential equation

y 00 + a1 y 0 + a0 y = 0 (7.11)

in which a1 and a0 are constants, is the algebraic equation

λ2 + a1 λ + a0 = 0 (7.12)

which is obtained from Eq (7.11) by replacing y 00 , y 0 and y by λ2 ,

λ1 and λ0 = 1, respectively. Equation (7.12) is called the characteristic
equation of (7.11).

Example 7.2.1. The characteristic equation of y 00 + 3y 0 − 4y = 0 is

λ2 + 3λ − 4 = 0; the characteristic equation of y 00 − 2y 0 + y = 0 is
λ2 − 2λ + 1 = 0.

Characteristic equations for differential equations having dependent

variables other than y are obtained analogously, by replacing the jth
derivative of the dependent variable by λj (j = 0, 1, 2).

85 of 110
BRAC
University MAT 120

The characteristic equation can be factored into

(λ − λ1 )(λ − λ2 ) = 0 (7.13)

THE GENERAL SOLUTION

The general solution of (7.11) is obtained directly from the roots of

(7.13). There are three cases to consider.

Case 1. λ1 and λ2 both real and distinct. Two linearly inde-

pendent solutions are eλ1 x and eλ2 x , and the general solution is (Theorem
7.1.2)

y = c1 eλ1 x + c2 eλ2 x (7.14)

In the special case λ2 = −λ1 , the solution (7.14) can be rewritten

as y = k1 cosh λ1 x + k2 sinh λ1 x

Case 2. λ1 = a + ib, a complex number. Since a1 and a0

in (7.11) and (7.12) are assumed real, the roots of (7.12) must appear
in conjugate pairs; thus, the other root is λ2 = a − ib. Two linearly
independent solutions are e(a+ib)x and e(a−ib)x , and the general complex
solution is

y = d1 e(a+ib)x + d2 e(a−ib)x (7.15)

which is algebraically equivalent to

y = c1 eax cos bx + c2 eax sin bx (7.16)

Case 3. λ1 = λ2 . Two linearly independent solutions are eλ1 x

and xeλ1 x , and the general solution is

y = c1 eλ1 x + c2 xeλ1 x (7.17)

86 of 110
BRAC
University MAT 120

Warning: The above solutions are not valid if the differential equa-
tion is not linear or does not have constant coefficients. Consider, for
example, the equation y 00 − x2 y = 0. The roots of the characteristic
equation are λ1 = x and λ2 = −x, but the solution is not

2 2
y = c1 e(x)x + c2 e(−x)x = c1 ex + c2 e−x

Example 7.2.2. Solve y 00 − y 0 − 2y = 0.

The characteristic equation is λ2 − λ − 2 = 0, which can be factored
into (λ + 1)(λ − 2) = 0. Since the roots λ1 = −1 and λ2 = 2 are real
and distinct, the solution given by (7.14) as

y = c1 e−x + c2 e2x

Example 7.2.3. Solve y 00 − 8y 0 + 16y = 0.

The characteristic equation is
λ2 − 8λ + 16 = 0

which can be factored into

(λ − 4)2 = 0

The roots λ1 = λ2 = 4 are real and equal, so the general solution is

given by (7.17) as
y = c1 e4x + c2 xe4x

87 of 110
BRAC
University MAT 120

7.3 nth order Linear Homogeneous Differential Equa-

tions
THE CHARACTERISTIC EQUATION

The characteristic equation of the differential equation

y (n) + an−1 y (n−1) + ... + a1 y 0 + a0 y = 0 (7.18)

with constant coefficients aj (j = 0, 1, ..., n − 1) is

λn + an−1 λn−1 + ... + a1 λ + a0 = 0 (7.19)

The characteristic equation (7.19) is obtained from (7.18) by replac-
ing y (j) by λj (j = 0, 1, ..., n − 1). Characteristic equations for differ-
ential equations having dependent variables other than y are obtained
analogously, by replacing the jth derivative of the dependent variable
by λj (j = 0, 1, ..., n − 1).
Example 7.3.1. The characteristic equation of y (4) −3y 000 +2y 00 −y = 0
is λ4 − 3λ3 + 2λ2 − λ = 0. The characteristic equation of

d5 x d3 x dx
− 3 + 5 − 7x = 0
dt5 dt3 dt
is

λ5 − 3λ3 + 5λ − 7 = 0

Caution: Characteristic equations are only defined for linear homo-

geneous differential equations with constant coefficients.

THE GENERAL SOLUTION

The roots of the characteristic equation determine the solution of

(7.18). If the roots λ1 , λ2 , ..., λn are all real and distinct the solution is

y = c1 eλ1 x + c2 eλ2 x + ... + cn eλn x (7.20)

88 of 110
BRAC
University MAT 120

If the roots are distinct, but some are complex, then the solution is
again given by (7.20). Those terms involving complex exponentials can
be combined to yield terms involving sines and cosines. If λk is a root of
multiplicity p [that is, if (λ − λk )p is a factor of the characteristic equa-
tion, but (λ − λk )p+1 is not] then there will be p linearly independent
solutions associated with λk given by eλk x , xeλk x , x2 eλk x , ... , xp−1 eλk x .
These solutions are combined in the usual way with the solutions asso-
ciated with the other roots to obtain the complete solution.

In theory it is always possible to factor the characteristic equation,

but in practice this can be extremely difficult, especially for differential
equations of high order. In such cases, one must often use numerical
techniques to approximate the solutions.
Example 7.3.2. Solve y 000 − 6y 00 + 11y 0 − 6y = 0
The characteristic equation is λ3 − 6λ2 + 11λ − 6 = 0, which can be
factored into

(λ − 1)(λ − 2)(λ − 3) = 0

The roots are λ1 = 1, λ2 = 2 and λ3 = 3; hence the solution is

y = c1 ex + c2 e2x + c3 e3x

Example 7.3.3. Solve y (4) + 8y 000 + 24y 00 + 32y 0 + 16y = 0

The characteristic equation, λ4 + 8λ3 + 24λ2 + 32λ + 16 = 0, can be
factored into (λ + 2)4 = 0. Here λ1 = −2 is a root of multiplicity four;
hence the solution is

y = c1 e−2x + c2 xe−2x + c3 x2 e−2x + c4 x3 e−2x

89 of 110
BRAC
University MAT 120

7.4 Method of Undetermined Coefficients

The general solution to the linear differential equation L(y) = φ(x) is
given by Theorem 7.1.4 as y = yh + yp where yp denotes one solution to
the differential equation and yh is Ihe general solution to the associated
homogeneous equation, L(y) = 0. Methods for obtaining yh when the
differential equation has constant coefficients are given in the previous
sections. In this chapter and the next, we give methods for obtaining a
particular solution yp once yh is known.

SIMPLE FORM OF THE METHOD

The method of undetermined coefficients is applicable only if φ(x)

and all of its derivatives can be written in terms of the same finite set of
linearly independent functions. which we denote by {y1 (x), y2 (x), ..., yn (x)}.
The method is initiated by assuming a particular solution of the form

yp (x) = A1 y1 (x) + A2 y2 (x) + ... + An yn (x)

where A1 , A2 , ... , An denote arbitrary multiplicative constants.

These arbitrary constants are then evaluated by substituting the pro-
posed solution into the given differential equation and equating the co-
efficients of like terms.

Case 1. φ(x) = pn (x), an nth degree polynomial in x.

Assume a solution of the form

yp = An xn + An−1 xn−1 + ... + A1 x + A0 (7.21)

where Aj (j = 0, 1, 2, ..., n) is a constant to be determined.

Case 2. φ(x) = keαx where k and α are known constants.

Assume a solution of the form

yp = Aeαx (7.22)

where A is a constant to be determined.

90 of 110
BRAC
University MAT 120

Case 3. φ(x) = k1 sin βx + k2 cos βx where k1 , k2 and β

are known constants. Assume a solution of the form

yp = A sin βx + B cos βx (7.23)

where A and B are constants to be determined.

Note: (7.23) in its entirety is assumed even when k1 or k2 is zero,

because the derivatives of sines or cosines involve both sines and cosines.

GENERALIZATIONS

If φ(x) is the product of terms considered in Cases 1 through 3,

take yp to be the product of the corresponding assumed solutions and
algebraically combine arbitrary constants where possible. In particular,
if φ(x) = eαx pn (x) is the product of a polynomial with an exponential,
assume

yp = eαx (An xn + An−1 xn−1 + ... + A1 x + A0 ) (7.24)

where Aj is as in Case 1. If, instead, φ(x) = eαx pn (x) sin βx is
the product of a polynomial, exponential, and sine term, or if φ(x) =
eαx pn (x) cos βx is the product of a polynomial, exponential, and cosine
term, then assume

yp = eαx sin βx(An xn + ... + A1 x + A0 ) + eαx cos βx(Bn xn + ... + B1 x + B0 )

(7.25)
where Aj and Bj (j = 0, 1, ..., n) are constants which still must be
determined.

if φ(x) is the sum (or difference) of terms already considered, then

we take yp to be the sum (or difference) of the corresponding assumed
solutions and algebraically combine arbitrary constants where possible.

MODIFICATIONS

If any term of the assumed solution, disregarding multiplicative con-

stants, is also a term of yh (the homogeneous solution), then the as-
sumed solution must be modified by multiplying it by xm , where m

91 of 110
BRAC
University MAT 120

is the smallest positive integer such that the product of xm with the
assumed solution has no terms in common with yh .

LIMITATIONS OF THE METHOD

In general, if φ(x) is not one of the types of functions considered

above, or if the differential equation does not have constant coefficients,
then other methods apply.
Example 7.4.1. Solve y 00 − y 0 − 2y = 4x2
It can be shown that yh = c1 e−x + c2 e2x . Here φ(x) = 4x2 , a second
degree polynomial. Using (7.21), we assume that
yp = A2 x2 + A1 x + A0

Thus, yp0 = 2A2 x + A1 and yp00 = 2A2 . Substituting these results into
the differential equation, we have

2A2 − (2A2 x + A1 ) − 2(A2 x2 + A1 x + A0 ) = 4x2

or, equivalently,

(−2A2 )x2 + (−2A2 − 2A1 )x + (2A2 − A1 − 2A0 ) = 4x2 + (0)x + 0

Equating the coefficients of like powers of x, we obtain

−2A2 = 4 − 2A2 − 2A 1 = 0 2A2 − A1 − 2A0 = 0

Solving this system, we find that A2 = −2, A1 = 2 and A0 = −3.

Hence

yp = −2x2 + 2x − 3

and the general solution is

y = yh + yp = c1 e−x + c2 e2x − 2x2 + 2x − 3

92 of 110
BRAC
University MAT 120

Example 7.4.2. Solve y 000 − 6y 00 + 11y 0 − 6y = 2xex

It can be shown that yh = c1 ex +c2 e2x +c3 e3x . Here φ(x) = eαx pn (x),
where α = −1 and pn (x) = 2x, a first degree polynomial. Using (7.24),
we assume that yp = e−x (A1 x + A0 ), or
yp = A1 xe−x + A0 e−x

Thus,

yp0 = −A1 xe−x + A1 e−x − A0 e−x

yp00 = A1 xe−x − 2A1 e−x + A0 e−x
yp000 = −A1 xe−x + 3A1 e−x − A0 e−x

Substituting these results into the differential equation and simplify-

ing, we obtain

−24A1 xe−x + (26A1 − 24A0 )e−x = 2xe−x + (0)e−x

Equating coefficients of like terms, we have

−24A1 = 2 26A1 − 24A0 = 0

from which A1 = −1/12 and A0 = −13/144.

So,

1 −x 13 −x
yp = − xe − e
12 144
and the general solution is

1 −x 13 −x
y = c1 ex + c2 e2x + c3 e3x − xe − e
12 144

93 of 110
BRAC
University MAT 120

7.5 Initial Value Problems for Linear Differential

Equations
Initial-value problems are solved by applying the initial conditions to
the general solution of the differential equation. It must be emphasized
that the initial conditions are applied only to the general solution and
not to the homogeneous solution yh , even though it is yh that possesses
all arbitrary constants that must be evaluated. The one exception is
when the general solution is the homogeneous solution; that is, when
the differential equation under consideration is itself homogeneous.
Example 7.5.1. Solve y 00 − y 0 − 2y = 4x2 ; y(0) = 1, y 0 (0) = 4.
The general solution of the differential equation is

y = c1 e−x + c2 e2x − 2x2 + 2x − 3

Therefore,

y 0 = −c1 e−x + 2c2 e2x − 4x + 2

Applying the first initial condition to the equation for y, we obtain

y(0) = c1 e−(0) + c2 e2(0) − 2(0)2 + 2(0) − 3 = 1 or c1 + c2 = 4

Applying the second initial condition to the equation for y 0 , we obtain

y 0 (0) = −c1 e−(0) + 2c2 e2(0) − 4(0) + 2 = 0 or − c1 + 2c2 = 2

Solving the two equations above simultaneously, we find that c1 = 2

and c2 = 2. Substituting these values into the general solution, we
obtain the solution of the initial-value problem as

y = 2e−x + 2e2x − 2x2 + 2x − 3

Example 7.5.2. Solve y 00 + 4y 0 + 8y = sin x; y(0) = 1, y 0 (0) = 0

Here yh = e−2x (c1 cos 2x + c2 sin 2x), and, by the method of undeter-
mined coefficients,

94 of 110
BRAC
University MAT 120

7 4
yp = sin x − cos x
65 65
Thus, the general solution to the differential equation is

7 4
y = e−2x (c1 cos 2x + c2 sin 2x) + sin x − cos x
65 65
Therefore,

y 0 = −2e−2x (c1 cos 2x + c2 sin 2x) + e−2x (−2c1 sin 2x + 2c2 cos 2x)
7 4
+ cos x + sin x
65 65
Applying the first initial condition to the general solution, we obtain

69
c1 =
65
Applying the second initial condition to the general solution, we ob-
tain

7
−2c1 + c2 = −
65
Solve the two equations above simultaneously, we find that c1 =
69/65 and c2 = 131/130. Substituting these values into the general
solution, we obtain the solution of the initial-value problem as
 
69 131 7 4
y = e−2x cos 2x + sin 2x + sin x − cos x
65 130 65 65

95 of 110
BRAC
University MAT 120

7.6 Application of Second Order Linear Differential

Equations
SPRING PROBLEMS

A simple spring system consists of a mass m attached to the lower

end of a spring that is itself suspended vertically from a mounting. The
system is in its equilibrium position when it is at rest. The mass is set
in motion by one or more of the following means: displacing the mass
from its equilibrium position, providing it with an initial velocity, or
subjecting it to an external force F (t).

Hooke’s law: The restoring force F of a spring is equal and

opposite to the forces applied to the spring and is proportional to the ex-
tension (contraction) l of the spring as a result of the applied force; that
is, F = −kl, where k denotes the constant of proportionality, generally
called the spring constant.

Example 7.6.1. A steel ball weighing 128 lb is suspended from a spring,

whereupon the spring is stretched 2 ft from its natural length. The ap-
plied force responsible for the 2-ft displacement is the weight of the ball,
128 lb. Thus, F = -128 lb. Hooke’s law then gives -128 = -k(2), or k =
64 lb/ft.

For convenience, we choose the downward direction as the positive

direction and take the origin to be the center of gravity of the mass in
the equilibrium position. We let x denote the displacement of the mass
of the spring. We assume that the mass of the spring is negligible and
can be neglected and that air resistance, when present, is proportional
to the velocity of the mass. Thus, at any time t, there are three forces
acting on the system: (1) F (t), measured in the positive direction; (2) a
restoring force given by Hooke’s law as Fs = −kx, k > 0; and (3) a force
due to air resistance given by Fa = −ax, a > 0, where a is the constant
of proportionality. Note that the restoring force Fs always acts in a
direction that will tend to return the system to the equilibrium position:
if the mass is below the equilibrium position, then x is positive and −kx
is negative; whereas if the mass is above the equilibrium position, then x

96 of 110
BRAC
University MAT 120

is negative and −kx is positive. Also note that because a > 0 the force
Fa due to air resistance acts in the opposite direction of the velocity and
thus tends to retard, or damp, the motion of the mass.

It now follows from Newton’s second law that mẍ = −kx−aẋ+F (t),
or

a k F (t)
ẍ + ẋ + x = (7.26)
m m m
If the system starts at t = 0 with an initial velocity v0 and from an
initial position x0 , we also have the initial conditions

x(0) = x0 ẋ(0) = v0 (7.27)

The force of gravity does not explicitly appear in the first equation,
but it is present nonetheless. We automatically compensated for this
force by measuring distance from the equilibrium position of the spring.
If one wishes to exhibit gravity explicitly, then distance must be mea-
sured from the bottom end of the natural length of the spring. That is,
the motion of a vibrating spring can be given by

a k F (t)
ẍ + ẋ + x = g +
m m m

ELECTRICAL CIRCUIT PROBLEMS

We have a simple electrical circuit consisting of a resistor R in ohms;

a capacitor C in farads, an inductor L in henries; and an electromotive
force (emf) E(t) in volts, usually a battery or generator, all connected
in a series. The current I flowing through the circuit is measured in
amperes and the charge q on the capacitor is measured in coulombs.

Kirchoff ’s loop law: The algebraic sum of the voltage drops in a

simple closed electric circuit is zero.

It is known that the voltage drops across a resistor, a capacitor, and

an inductor are respectively RI, (1/C)q, and L(dI/dt) where q is the

97 of 110
BRAC
University MAT 120

charge on the capacitor. The voltage drop across an emf is −E(t). Thus,
from Kirchhoff s loop law, we have

dI 1
RI + L + q − E(t) = 0 (7.28)
dt C
The relationship between q and I is

dq dI d2 q
I= = 2 (7.29)
dt dt dt
Substituting these values into the initial equation, we obtain

d2 q R dq 1 1
+ + q = E(t) (7.30)
dt2 L dt LC L
The initial conditions for q are

dq
q(0) = q0 = I(0) = I0 (7.31)
dt t=0

To obtain a differential equation for the current, we differentiate the

initial equation with respect to t and then substitute Eq. (7.29) into the
resulting equation to obtain

d2 I R dI 1 1 dE(t)
+ + I = (7.32)
dt2 L dt LC L dt
The first initial condition is I(0) = I0 . The second initial condition
is obtained from Eq. (7.28) by solving for dI/dt and then setting t = 0.
Thus,

dI 1 R 1
= E(0) − I0 − q0 (7.33)
dt t=0 L L LC
An expression for the current can be gotten either by solving Eq.
(7.32) directly or by solving Eq. (7.30) for the charge and then differ-
entiating that expression.

98 of 110
BRAC
University MAT 120

BUOYANCY PROBLEMS

Consider a body of mass m submerged either partially or totally in

a liquid of weight density ρ. Such a body experiences two forces, a
downward force due to gravity and a counter force governed by:

Archimedes’ principle: A body in liquid experiences a buoyant

upward force equal to the weight of the liquid displaced by that body.

Equilibrium occurs when the buoyant force of the displaced liquid

equals the force of gravity on the body. The situation for a cylinder of
radius r and height H where h units of cylinder height are submerged at
equilibrium, is like this: at equilibrium, the volume of water displaced
by the cylinder is πr2 h, which provides a buoyant force of πr2 hρ that
must equal the weight of the cylinder mg. Thus,

πr2 hρ = mg (7.34)

Motion will occur when the cylinder is displaced from its equilibrium
position. We arbitrarily take the upward direction to be the positive
x-direction. If the cylinder is raised out of the water by x(t) units,
then it is no longer in equilibrium. The downward or negative force on
such a body remains mg but the buoyant or positive force is reduced to
πr2 [h − x(t)]ρ. It now follows from Newton’s second law that

mẍ = πr2 [h − x(t)]ρ − mg (7.35)

Substituting (7.34) into this last equation, we can simplify it to

mẍ = −πr2 x(t)ρ

or

πr2 ρ
ẍ + x=0 (7.36)
m

99 of 110
BRAC
University MAT 120

CLASSIFYING SOLUTIONS

Vibrating springs, simple electrical circuits, and floating bodies are

all governed by second-order linear differential equations with constant
coefficients of the form

ẍ + a1 ẋ + a0 = f (t) (7.37)

For vibrating spring problems defined by Eq. (7.26), a1 = a/m,

a0 = k/m, and f (t) = F (t)/m. For buoyancy problems defined by
Eq.(7.36), a1 = 0, a0 = πr2 ρ/m, and f (t) ≡ 0. For electrical circuit
problems, the independent variable x is replaced either by q in Eq. (7.30)
or I in (7.32).

The motion or current in all of these systems is classified as free and

undamped when f (t) ≡ 0 and a1 = 0. It is classified as free and damped
when f (t) is identically zero but a1 is not zero. For damped motion,
there are three separate cases to consider, depending on whether the
roots of the associated characteristic equation re (1) real and distinct, (2)
equal, or (3) complex conjugate. These cases are respectively classified
as (1) overdamped, (2) critically damped, and (3) oscillatory damped (or,
in electrical problems, underdamped ). If f (t) is not identically zero, the
motion or current is classified as forced.

A motion or current is transient if it “dies out“ (that is, goes to zero)

as t → ∞. A steady-state motion or current is one that is not transient
and does not become unbounded. Free damped systems always yield
transient motions, while forced damped systems (assuming the external
force to be sinusoidal) yield both transient and steady-state motions.

Free undamped motion defined by Eq. (7.37) with a1 = 0 and f (t) ≡

0 always has solutions of the form

x(t) = c1 cos ωt + c2 sin ωt (7.38)

which defines simple harmonic motion. Here c1 , c2 , and ω are con-

stants with ω often referred to as circular frequency. The natural fre-
quency f is

100 of 110
BRAC
University MAT 120

ω
f= (7.39)
2π
and it represents the number of complete oscillations per time unit
undertaken by the solution. The period of the system of the time re-
quired to complete one oscillation is

1
T = (7.40)
f
Equation (7.38) has the alternate form

x(t) = (−1)k A cos (ωt − φ) (7.41)

p
where the amplitude A = c21 + c22 , the phase angle φ = arctan (c2 /c1 ),
and k is zero when c1 is positive and unity when c1 is negative.

Example 7.6.2. A steel ball weighing 128 lb is suspended from a spring,

whereupon the spring is stretched 2 ft from its natural length. The ball is
started in motion with no initial velocity by displacing it 6 in above the
equilibrium position. Assuming no air resistance, find (a) an expression
for the position of the ball at any time t, and (b) the position of the ball
at t = π/12 sec.

(a) The equation of motion is governed by Eq. (7.26). There

is no externally applied force, so F (t) = 0, and no resistance from the
surrounding medium, so a = 0. The motion is free and undamped. Here
g = 32 ft/sec2 , m = 128/32 = 4 slugs, and it follows from Example 7.6.1
that k = 64 lb/ft. Equation (7.26) becomes ẍ + 16x = 0. The roots of
its characteristic equation are λ = ±4i, so its solution is

x(t) = c1 cos 4t + c2 sin 4t

At t = 0, the position of the ball is x0 = − 12 ft (the minus sign is
required because the ball is initially displaced above the equilibrium po-
sition, which is in the negative direction). Applying this initial condition
to the equation above, we find that

101 of 110
BRAC
University MAT 120

1
− = x(0) = c1 cos 0 + c2 sin 0 = c1
2
so the initial equation becomes

1
x(t) = − cos 4t + c2 sin 4t
2
The initial velocity is given as v0 = 0 ft/sec. Differentiating the
equation above, we obtain

v(t) = ẋ(t) = 2 sin 4t + 4c2 cos 4t

whereupon

0 = v(0) = 2 sin 0 + 4c2 cos 0 = 4c2

Thus, c2 = 0, and x(t)’s expression simplifies to

1
x(t) = − cos 4t
2
as the equation of motion of the steel ball at any time t.

(b) At t = π/12,

π 1 4π 1
x = − cos = − ft
12 2 12 4
Example 7.6.3. A mass of 1/4 slug is attached to a spring, where-
upon the spring is stretched 1.28 ft from its natural length. The mass is
started in motion from the equilibrium position with an initial velocity
of 4 ft/sec in the down-ward direction. Find the subsequent motion of
the mass if the force due to air resistance is −2ẋ lb.
Here m = 1/4, a = 2, F (t) ≡ 0 (there is no external force), and,
from Hooke’s law, k = mg/l = (1/4)(32)/1.28 = 6.25. Equation (7.26)
becomes

102 of 110
BRAC
University MAT 120

ẍ + 8ẋ + 25x = 0

The roots of the associated characteristic equation are λ1 = −4 + 3i

and λ2 = −4 − 3i, which are complex conjugates; hence this problem is
an example of oscillatory damped motion. The solution of the equation
above is

x = e−4t (c1 cos 3t + c2 sin 3t)

The initial conditions are x(0) = 0 and ẋ(0) = 4. Applying these

conditions, we find that c1 = 0 and c2 = 34 ; thus, x = 34 e−4t sin 3t. Since
x → 0 as t → ∞, the motion is transient.

7.7 Reduction of Linear Differential Equation to First

Order Equations
AN EXAMPLE

Consider the following second-order differential equation:

d2 x dx
t4 + (sin t) − 4x = ln t (7.42)
dt2 dt
We see that (7.42) implies

d2 x 4 sin t dx ln t
= x − + 4 (7.43)
dt2 t4 t4 dt t
Since that derivatives can be expressed in many ways — using primes
dx
or dots are but two of them — we let v = = x0 = ẋ and v 0 =
dt
2
dx
= x00 = ẍ. Then Eq. (7.42) can be written as the following matrix
dt2
equation:

      
ẋ 0 1 x 0
= 4 − sin t + ln t (7.44)
v̇ t4 t4 v t4

103 of 110
BRAC
University MAT 120

because ẋ = 0x + 1v̇ and v̇ = t44 x − sin t ln t

t4 v + t4 . We note, finally, that
Eq (7.42) can also be expressed as

dx(t)/dt = A(t)x(t) + f (t) (7.45)

Note that if x(0) = 5 and ẋ(0) = −12 in (7.42), then these initial
conditions are written as x(0) = 5, v(0) = −12.

REDUCTION OF AN nth-ORDER EQUATION

As in the case of the second-order differential equation, with associ-

ated initial conditions, we can recast higher order initial-value problems
into a first-order matrix system as illustrated below:

dn x dn−1 x
bn (t) n + bn−1 (t) n−1 + ... + b1 (t)ẋ + b0 (t)x = g(t); (7.46)
dt dt
˙
x(t0 ) = c0 , x(t0 ) = c1 , ..., x(n−1) (t0 ) = cn−1 (7.47)
with bn (t) 6= 0, can be reduced to the first-order matrix system.

ẋ(t) = A(t)x(t) + f (t)

x(t0 ) = c (7.48)
where A(t), f (t) c, and the initial time t0 are known. The method
of reduction is as follows.

Step 1. Rewrite (7.46) so that dn x/dtn appears by itself. Thus,

dn x dn−1 x
= an−1 (t) n−1 + ... + a1 (t)ẋ + a0 (t)x + f (t) (7.49)
dtn dt
where aj (t) = −bj (t)/bn (t) (j = 0, 1, ..., n−1) and f (t) = g(t)/bn (t).

Step 2. Define n new variables (the same number as the order

of the original differential equation); x1 (t), x2 (t), ... , xn (t), by the
equations

dx(t) d2 x(t) dn−1 x

x1 (t) = x(t), x2 (t) = , x3 (t) = , ..., xn (t) = n−1 (7.50)
dt dt2 dt

104 of 110
BRAC
University MAT 120

These new variables are interrelated by the equations

ẋ1 (t) = x2 (t)

ẋ2 (t) = x3 (t)
ẋ3 (t) = x4 (t)
......... (7.51)
ẋn−1 (t) = xn (t)

Step 3. Express dxn /dt in terms of the new variables. Proceed

by first differentiating the last equation of (7.50) to obtain

d dn−1 x(t) dn x(t)

 
ẋn (t) = =
dt dtn−1 dtn
Then, from Eqs. (7.49) and (7.50),

dn−1 x(t)
ẋn (t) = an−1 (t) + ... + a1 (t)ẋ(t) + a0 (t)x(t) + f (t)
dtn−1
= an−1 (t)xn (t) + ... + a1 (t)x2 (t) + a0 (t)x1 (t) + f (t)

For convenience, we rewrite this last equation so that x1 (t), appears

before x2 (t), etc. Thus,

ẋn (t) = a0 (t)x1 (t) + a1 (t)x2 (t) + ... + an−1 (t)xn (t) + f (t) (7.52)

Step 4. Equations (7.51) and (7.52) are a system of first-order

linear differential equations in x1 (t), x2 (t), ... , xn (t). This system is
equivalent to the single matrix equation ẋ(t) = A(t)x(t) + f (t) if we
define

 
x1 (t)
 x (t) 
 2 
x(t) ≡  ..  (7.53)
 . 
xn (t)

105 of 110
BRAC
University MAT 120

 
0
 0 
 
 .. 
f (t) ≡  .  (7.54)
 
 0 
f (t)

 
0 1 0 0 ... 0
 0 0 1 0 ... 0 
 
 0 0 0 1 ... 0 
A(t) ≡  .. (7.55)
 
 . .. .. .. ..
. . . .
 
 0 0 0 0 ... 1 
a0 (t) a1 (t) a2 (t) a3 (t) ... an−1 (t)

Step 5. Define

 
c0
 c1 
c≡
 
.. 
 . 
cn−1

Then the initial conditions (7.47) can be given by the matrix (vector)
equation x(t0 ) = c. This last equation is an immediate consequence of
Eqs. (7.53), (7.54), and (7.47), since

     
x1 (t0 ) x(t0 ) c0
 x (t )   ẋ(t )   c 
 2 0   0   1 
x(t0 ) =  ..  =  ..  =  ..  ≡ c
 .   .   . 
(n−1)
xn (t0 ) x (t0 ) cn−1

Observe that if no initial conditions are prescribed, Steps 1 through

4 by themselves reduce any linear differential Eqs. (7.46) to the matrix
equation ẋ(t) = A(t)x(t) + f (t).

106 of 110
BRAC
University MAT 120

REDUCTION OF A SYSTEM

A set of linear differential equations with initial conditions also can

be reduced to System (7.48). The procedure is nearly identical to the
method for reducing a single equation to matrix form; only Step 2
changes. With a system of equations, Step 2 is generalized so that
new variables are defined for each of the unknown functions in the set.
Example 7.7.1. Put the initial-value problem

ẍ + 2ẋ − 8x = et ; x(0) = 1, ẋ(0) = −4

into the form of System (7.48).
Following Step 1, we write ẍ = −2ẋ + 8x + et ; hence, a1 (t) = −2,
a0 (t) = 8, and f (t) = et . Then, defining x1 (t) = x and x2 (t) = ẋ (the
differential equation is second-order, so we need two new variables), we
obtain) ẋ1 = x2 . Following Step 3, we find

d2 x
ẋ2 = 2 = −2ẋ + 8x + et = −2x2 + 8x1 + et
dt
Thus,

ẋ1 = 0x1 + 1x2 + 0

ẋ2 = 8x1 − 2x2 + et
These equations are equivalent to the matrix equation ẋ(t) = A(t)x(t)+
f (t) if we define
     
x1 (t) 0 1 0
x(t) ≡ A(t) ≡ f (t) ≡ t
x2 (t) 8 −2 e
 
1
Furthermore, if we also define c ≡ , then the initial conditions
−4
can be given by x(t0 ) = c, where t0 = 0.
Example 7.7.2. Convert the differential equation ẍ − 6ẋ + 9x = t into
the matrix equation

ẋ(t) = A(t)x(t) + f (t)

107 of 110
BRAC
University MAT 120

Here we omit Step 5, because the differential equation has no pre-

scribed initial conditions. Following Step 1, we obtain

ẍ = 6ẋ − 9x + t

Hence a1 (t) = 6, a0 (t) = −9, and f (t) = t. If we define two new

variables, x1 (t) = x and x2 (t) = ẋ, we have

ẋ1 = x2 and ẋ2 = ẍ = 6ẋ − 9x + t = 6x2 − 9x1 + t

Thus,

ẋ1 = 0x1 + 1x2 + 0

ẋ2 = −9x1 + 6x2 + t

These equations are equivalent to the matrix equation ẋ(t) = A(t)x(t)+

f (t) if we define

     
x1 (t) 0 1 0
x(t) ≡ A(t) ≡ f (t) ≡
x2 (t) −9 6 t

108 of 110
Appendices

109
List of Figures

1.1 The integral as a sum . . . . . . . . . . . . . . . . . . . . . 4

1.2 Integral of an odd function . . . . . . . . . . . . . . . . . . . 8
1.3 The logarithmic integral . . . . . . . . . . . . . . . . . . . . 8
1.4 Integral comparison test . . . . . . . . . . . . . . . . . . . . 14
1.5 Graph of example 1.5.4 . . . . . . . . . . . . . . . . . . . . . 15
3.1 Trigonometric substitutions . . . . . . . . . . . . . . . . . . . 32
3.2 Half-angle substitution . . . . . . . . . . . . . . . . . . . . . 33
5.1 Arc length . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.2 Graph of x = t2 , y = 2t3 , −1 ≤ t ≤ 1 .
. . . . . . . . . . . . . . 44
5.3 Surface of revolution . . . . . . . . . . . . . . . . . . . . . . 46
5.4 Tetrahedron for Example 5.3.1 . . . . . . . . . . . . . . . . . 49
5.5 Volume of revolution; the disc method . . . . . . . . . . . . . . 51
5.6 Volume of revolution; the shell method . . . . . . . . . . . . . . 52
5.7 Diagram for Example 5.4.3 . . . . . . . . . . . . . . . . . . . 54

110