Calculus

Calculus

for scientists and engineers

with matlab

Jishan Hu, Weiping Li, and Yueping Wu

ii
Contents

1 Limit and Continuity 1

1.1 Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Exponential Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Trigonometric Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2 Limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Limit Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
One-Sided Limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Limits at Infinities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Infinite Limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.3 Continuous Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
Composite Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Intermediate Value Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
1.4 Some Transcendental Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
Exponential Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
Inverse Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
Logarithmic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
Inverse Trigonometric Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

2 Differentiation 55
2.1 Derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
Relationship Between Differentiability and Continuity . . . . . . . . . . . . . . . . . 61
2.2 Rules of Differentiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
Differentiation of Trigonometric Functions . . . . . . . . . . . . . . . . . . . . . . . . 65
Chain Rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
Differentiation of Inverse Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
2.3 Differentiation of Exponential and Logarithmic Functions . . . . . . . . . . . . . . . 75
Logarithmic Differentiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
2.4 Implicit Differentiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
2.5 Related Rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
iv CONTENTS

2.6 Linear Approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

Differential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
2.7 Higher Derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
Leibniz Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

3 Applications of Differentiation 97
3.1 Maximum and Minimum Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
3.2 Monotone Functions and Mean Value Theorem . . . . . . . . . . . . . . . . . . . . . 107
3.3 Tests for Local Extrema and Concavity . . . . . . . . . . . . . . . . . . . . . . . . . 114
Local Extrema . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
Concavity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
3.4 L’Hospital’s Rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
3.5 Curve Sketching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
3.6 Parametric Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
Polar Coordinates and Polar Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
3.7 Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
3.8 Newton’s Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180

4 Integration 185
4.1 Areas under Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
Properties of Definite Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
Average Value of a Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
4.2 Fundamental Theorem of Calculus . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
4.3 Indefinite Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
4.4 Substitution Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
Trigonometric Substitutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
4.5 Integration by Parts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
4.6 Some Common Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
Trigonometric Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
Rational Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
4.7 Numerical Approximation of Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . 236
4.8 Improper Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259

5 Applications of Integration 265

5.1 Areas of Regions Between Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
Area Under Parametric Curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
5.2 Volumes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
Method of Cylindrical Shells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
Solid Generated by Parametric Curve . . . . . . . . . . . . . . . . . . . . . . . . . . 285
5.3 Arc Lengths . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
Arc Length of Parametric Curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
5.4 Surface Area of Revolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 296
Surface Area of Revolution by Parametric Curve . . . . . . . . . . . . . . . . . . . . 302
5.5 Lengths and Areas in Polar Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . 307
Arc Length of Polar Curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307
CONTENTS v

Area of Polar Region . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309

5.6 Moments, Centers of Mass, and Centroids . . . . . . . . . . . . . . . . . . . . . . . . 313
Discrete Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
Center of Mass Along a Curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315
Centroid for Planar Region . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322
Pappus’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 326
5.7 Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 332

6 Infinite Sequences and Series 337

6.1 Infinite Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
6.2 Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350
Convergence of Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350
Divergence Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
6.3 Integral Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359
6.4 Convergence Tests for Nonnegative Series . . . . . . . . . . . . . . . . . . . . . . . . 369
Comparison Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369
Ratio Test and Root Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 372
6.5 Convergence Tests for General Series . . . . . . . . . . . . . . . . . . . . . . . . . . . 378
Alternating Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 378
Absolute Convergence and Conditional Convergence . . . . . . . . . . . . . . . . . . 382
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384

7 Series of Functions 391

7.1 Approximating Functions with Polynomials . . . . . . . . . . . . . . . . . . . . . . . 391
Taylor Polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 391
7.2 Represent Functions by Power Series . . . . . . . . . . . . . . . . . . . . . . . . . . . 396
Remainder Formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399
7.3 Properties of Power Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403
Radius of Convergence and Interval of Convergence . . . . . . . . . . . . . . . . . . . 403
Operations on Power Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 409
7.4 Applications of Taylor Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 420
Limits by Taylor Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 420
Approximation of Definite Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . 421
Solving Differential Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 422
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 424

Appendix: Introduction to MATLAB 429

Index 441
vi CONTENTS
 CHAPTER

1
Limit and Continuity

1.1 Functions

1.1 Definition: function

In the calculus of one variable, a function y = f (x) is a rule assigning a unique real
number y to a real number x in a subset D ⊂ R. Here x is called the independent variable
and y the dependent variable. The set D is called the domain of the function and the set R
of all y values is called the range of the function. The function is denoted by f , while f (x)
represents the value y at x, i.e., y = f (x).

1.2 Example: domain of function

1. The function f (x) = x2 assigns the square y

x2 for any number x. The domain D is the
whole real line, i.e., D = (−∞, +∞). The
2
range R is the interval [0, +∞).
y = x2 1

x
−1 O 1
2 Limit and Continuity

2. The function g(x) = x2 is defined on the y

interval [−1, 1]. The function g has the
same rule as the function f in (i) except y = g(x)
1
the domains are different. The function
f has the whole real line as the domain,
while g has the closed interval [−1, 1] as
its domain. x
−1 O 1

1 y
3. The function h(x) = is defined every-
x
where except at x = 0. So the domain
D is the whole real line with 0 deleted, 1
y=
i.e., D = (−∞, 0) ∪ (0, +∞). The range R x
is the whole real line with 0 deleted, i.e., 1
−1
R = (−∞, 0) ∪ (0, +∞). O 1
x
−1

1.3 Example: natural domain of function

The domain of a function sometimes is not given explicitly. In this case, the domain can be
found from the expression of the function or from the context. It is called the natural domain.

1. The natural domain of the polynomial y

f (x) = x3 + 1 is the whole real line.

y = x3 + 1
1

x
−1 O

2. The natural domain of the rational func- y

tion
x
g(x) =
(x − 1)(x + 2) y=
x
(x − 1)(x + 2)
is the whole real line with −2 and 1
deleted, i.e., −2 O 1
x

D = (−∞, −2) ∪ (−2, 1) ∪ (1, +∞).

1.1 Functions 3

√
3. The function h(x) = x is defined for any y
x ≥ 0, so the natural domain is [0, +∞).
√
y= x

x
O

4. The function
y
k(x) = (x − 1)(x + 2)
p

is defined for all x such that (x−1)(x+2) ≥

0. The inequality implies that x ≥ 1
or x ≤ −2. So the natural domain is
x
D = (−∞, −2] ∪ [1, +∞). −2 O 1

x2 − 1
5. The natural domain of p(x) = is (−∞, 1) ∪ (1, +∞). Note that x = 1 is not in
x−1
the domain of p since p(x) cannot be evaluated at x = 1. If we let q(x) = x + 1 whose
natural domain is (−∞, +∞), then p(x) = q(x) when x 6= 1. Be alert that p and q are
not the same functions, since they have different domains.

y y

x2 − 1
y= 2 y =x+1 2
x−1
1 1

x x
−1 O 1 −1 O 1

1.4 Example: application - natural domain

Consider a rectangle with one side with

length x and the other side with length y.
Suppose the sum of lengths of four sides is L. y
Then the area
L − 2x
A(x) = x · y = x · .
2 x

Since x is the length of one side, we have 0 ≤ x ≤ 21 L. So the natural domain of the
function A is the closed interval [0, 12 L]. When x = 0 or 12 L, the rectangle is degenerate, i.e.,
the area of the rectangle is zero.
4 Limit and Continuity

1.5 Example: transcendental functions

Typical transcendental functions are exponential functions, sine function, cosine function,
logarithmic functions. The following are examples of transcendental functions:

xπ , 2x , log2 x, xx , x1/x , sin x.

A general transcendental function is a function that does not satisfy a polynomial equation.

Exponential Functions

1.6 Definition: exponentiation and exponential function

For a positive number b (called the base) and a real number β (called the exponent), the
exponentiation bβ defines a real number.
The exponential function is f (x) = ax , where a is a positive number. It has domain R and
range (0, +∞).

1.7 Theorem: Laws of Exponents

Let a, b be positive numbers and α, β be real numbers. The following laws of exponents
hold:
α
1. aα · aβ = aα+β . 4. (a · b) = aα · bα .
aα 1
2. = aα−β . 5. a−α = .
aβ aα
β
3. (aα ) = aα·β . 6. a1 = a, a0 = 1.

1.8 Example: application - exponential function

A hot object tends to cool down to the surrounding temperature. Newton expressed this
as the following relationship (known as Newton’s Law of Cooling):

T (t) = Te + (T0 − Te )a−t ,

where T (t) is the temperature at a time t, T0 is the temperature at t = 0, Te is the temperature

of the environment around the cooling object, and a is some constant.
1.1 Functions 5

Trigonometric Functions
1.9 Definition: trigonometric functions

Trigonometric functions of an acute angle

are ratios of different pairs of sides of a right-
angled triangle, as shown in the figure:
opposite adjacent
sin θ = , cos θ = , en
us
e

hypotenuse hypotenuse hy
p ot opposite
opposite adjacent
tan θ = , cot θ = ,
adjacent opposite θ
hypotenuse hypotenuse adjacent
csc θ = , sec θ = .
opposite adjacent
In general, for an angle θ, with θ ∈ [0, 2π), y
if we let P (x, y) be the point on the terminal
P (x, y)
side of θ and if let r be the distance |OP |, the
values of the sine and the cosine are defined
r
as θ
y x
sin θ = , cos θ = . x
r r O

For θ outside [0, 2π), the values of the sine and the cosine are defined in terms of the
following periodic property:
sin(θ + 2π) = sin θ, cos(θ + 2π) = cos θ.
sin θ
The tangent function is defined by tan θ = . The cosecant, secant, and cotangent
cos θ
function are the reciprocals of the sine, cosine, and tangent functions.

y y

sin θ cos θ
1 1

O π π 3π 2π
θ O π π 3π 2π
θ
2 2 2 2
−1 −1
domain: (−∞, ∞) domain: (−∞, ∞)
range: [−1, 1] range: [−1, 1]

y y

tan θ cot θ

− 3π
2
− π2

O π 3π
θ −π O π θ
2 2

domain: (−∞, ∞) − {± π2 , ± 3π
2
,...} domain: (−∞, ∞) − {±π, ±3π, . . . }
range: (−∞, ∞) range: (−∞, ∞)
6 Limit and Continuity

y y

csc θ sec θ

sin θ − 3π
2
− π2 cos θ
−π O π θ O π 3π
θ
2 2

domain: (−∞, ∞) − {±π, ±3π, . . . } domain: (−∞, ∞) − {± π2 , ± 3π

2
,...}
range: (−∞, −1] ∪ [1, ∞) range: (−∞, −1] ∪ [1, ∞)

1.10 Theorem: exact values on special angles

For some special angles between 0 and 2,

π
we have the following table of exact values:

degree radian sine cosine

0 ◦
0 0 1
√
30◦ π
6
1
2 2
3

√ √
45◦ π
4 2
2
2
2

√
60◦ π
3 2
3 1
2

90◦ π
2 1 0

For other trigonometric functions, the exact values at these special angles can be obtained
by using
sin θ cos θ 1 1
tan θ = , cot θ = , sec θ = , csc θ = .
cos θ sin θ cos θ sin θ
For angles in other ranges, one may use the following equalities:

sin(x + π) = − sin x, cos(x + π) = − cos x, tan(x + π) = tan x,

csc(x + π) = − csc x, cos(x + π) = − sec x, cot(x + π) = cot x.

1.11 Theorem: Two Key Identities of Trigonometric Functions

• The basic relationship between the sine

and the cosine is given by the Pythagorean
r
identity: y
sin2 θ + cos2 θ = 1. θ
x
1.1 Functions 7

• Euler’s formula, the fundamental rela- Im

tionship between the trigonometric functions i
and the complex exponential function: for eiθ = cos θ + i sin θ

any real number θ: sin θ

iθ
e = cos θ + i sin θ, θ
O cos θ 1
Re

where i is the imaginary unit satisfying i2 =

−1.

Remark: complex functions

• Any complex number can be expressed in the form a + bi, where a and b are real
numbers. The number a is called the real part of a + bi, and b is its imaginary part.
• The complex exponential function eiθ can be treated similarly as other real expo-
nential functions.
• Many trigonometric identities can be proved by using the Pythagorean identity
sin2 θ + cos2 θ = 1 and Euler’s formula eiθ = cos θ + i sin θ. For instance, from the
following:

cos(α + β) + i sin(α + β)
= ei(α+β) = eiα · eiβ
= (cos α + i sin α) · (cos β + i sin β)
= (cos α cos β − sin α sin β) + i(sin α cos β + cos α sin β),

by identifying the real and imaginary parts on both sides, we have the angle sum
identities:

cos(α + β) = cos α cos β − sin α sin β,

sin(α + β) = sin α cos β + cos α sin β.

• The proof of Euler’s formula requires advanced knowledge of mathematics, so is

omitted here.

1.12 Example: piecewise function

There are various ways to express functions, numerically, graphically, or using mathemat-
ical expressions.
Some functions may have different mathematical expressions over different parts of its
domain. Here are some examples.
8 Limit and Continuity

1. y
if x ≤ 0;

x,
f (x) = y = f (x)
sin x, if x > 0.
1

x
−1 O 1
2π
π

−1

2. y
cos x, if x 6= 0;

g(x) =
0, if x = 0. y = g(x) 1

x
−π − 12 π O 1
2π
π

−1

3. y

 cos x, if x ≤ −1;

y = h(x) 2
h(x) = sin x, if −1 < x ≤ 1;
1
2, if x > 1.

x
−2π − 32 π −π − 12 π O 1
−1

1.13 MATLAB: symbolic function

MATLAB supports creating the symbolic function with the help of the command syms.
There are two ways to creat a symbolic function in MATLAB. The first way creates the
symbolic function by first creating the symbolic variable. The following example shows how
to create the function y = f (x) = x2 through creating the symbolic variable x.

>> syms x ;
>> y = x ^2;

where syms x creates the symbolic variable x.

The second way is to create the symbolic function and the symbolic variable simultaneously,
which is shown in the following:

>> syms f ( x ) ;
>> f ( x ) = x ^2;

where syms f(x) creates the symbolic function f(x) as well as the symbolic variable x.
Moreover, we can plot the figure of y = f (x) = x2 using the plot command in MATLAB.
The MATLAB code below shows how to plot the function y = f (x) = x2 in the domain
D = [−1, 1].
1.2 Limit 9

>> x = -1:0.01:1;
>> figure
>> plot (x , x .^2)
>> xlabel ( 'x ')
>> ylabel ( 'y = x ^2 ')

With the above MATLAB code, we can generate the figure of y = f (x) = x2 in the domain
D = [−1, 1], shown in the following:

0.9

0.8

0.7

0.6
y = x2

0.5

0.4

0.3

0.2

0.1

0
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
x

1.14 Engineering Example: signal

The signal is a basic concept in signal processing, communication systems and electrical
engineering. Signals in nature are mathematical functions. For example, the independent
variable can be the time; while the dependent variable can be electric current, flow rate,
sound, motion of an object, and energy, etc.

1.2 Limit
1.15 Example: behaviors near a point

Consider two types of behaviors of functions near a point.

1. Consider f (x) = x2 . When x approaches y

0, f (x) approaches 0.
y = x2
1

x
−1 O 1

1 1
2. Consider g(x) = and h(x) = sin . When x approaches 0, both g(x) and h(x) do
x x
not approach any finite number.
10 Limit and Continuity

y y

1
y=
x 1
1
1 y = sin
x
−1
x x
O 1 −1 O 1
−1
−1

1.16 Definition: limit of function

We say a function f with domain D has a finite limit L at a if f (x) approaches L when
x ∈ D approaches a. In this case, we write lim f (x) = L.
x→a

1.17 Example: tangent line

Consider the function f (x) = 12 x2 and the y

tangent line of the graph of y = f (x) at x = 1. secant line
We first consider the secant line of the graph y = 12 x2

near x = 1. 1
2
(1 + h)2
P
tangent line
The secant line passes through two points
P0 = (1, 12 ) and P = (1 + h, 12 (1 + h)2 ). The
slope of the secant line is 1
2
P0

2 (1+ h)2 − 12 h + 12 h2
1
x
S(h) = = = 1 + 12 h. O 1 1+h
(1 + h) − 1 h

As the point P approaches P0 , i.e., as h approaches 0, the secant line approaches the
tangent line. In the language of limits, the tangent line is the limit of the secant line. Hence
the limit of the slope of the secant line is the slope S of the tangent line, i.e.,

S = lim S(h) = lim (1 + 21 h) = 1.

h→0 h→0

1.18 Example: eliminating the zero factor

x−1 y
Consider the function f (x) = . The
x2 − 1
function is not defined at 1. In fact, if we sub- y=
x−1
x−1 0 x2 − 1
stitute x by 1 in 2 , we will get , which 1
x −1 0
is meaningless. x
−1 O 1
However, the limit of the function, −1
x−1
lim 2 , only involves values of the func-
x→1 x − 1
tion near x = 1, but not the value at x = 1.
1.2 Limit 11

Indeed, the value of the limit is 12 , that can be obtained by eliminating the same “zero”
factor x − 1 in the numerator and the denominator:
x−1 x−1
lim = lim factoring
x→1 x2 − 1 x→1 (x + 1)(x − 1)
1
= lim cancelling 0-factor
x→1 x + 1
= 12 ,

This example shows that the limit of of a function at x = 1 is irrelevant to the actual value
of the function at x = 1.
Remark:
Note that in the definition of limit, we do not require a in the domain D. In
particular, the actual value of f (a) even when a ∈ D plays no role in the definition of
limit as shown in the current example.

1.19 MATLAB: limit

MATLAB can calculate the limit of a function with command limit. The following ex-
ample shows how to calculate the limit of f (x) = x2 when x approaches 2, i.e., lim x2 , in
x→2
MATLAB.

>> syms x ;
>> limit ( x ^2 ,x ,2)

ans =

MATLAB returns 4 as the result of the limit.

x−1
The following example shows the calculation of lim 2 in MATLAB.
x→1 x − 1

>> syms x ;
>> f = ( x - 1) /( x ^2 - 1) ;
>> limit (f ,x ,1)

ans =

1/2

MATLAB returns 1
2 as the result of the limit.
12 Limit and Continuity

1.20 Engineering Example: unit-impulse signal

Consider the function p (t)

2 , if − ≤ t ≤ ;
( 1
p (t) = 1

0, otherwise.
2

The graph of p (t) is shown.

−O  t

Clearly, as  gets smaller, the graph of p (t) δ(t)

gets narrower and taller. In particular, as
 → 0, the function p (t) approaches the
Dirac delta function δ(t), and can be mathe-
matically written as

δ(t) = lim p (t). O

t
→0

Note that the Dirac delta function equals to 0 everywhere except for 0, and the “area”
under the function equals 1. The Dirac delta function δ(t) is depicted as shown.
In engineering fields, especially signal processing, the Dirac delta function is also known
as the unit-impulse signal. The output of a system with the unit-impulse signal input is
called the unit-impulse response. The unit-impulse response can completely describe a linear
time-invariant (LTI) system, such as communication channels.

Limit Rules
1.2 Limit 13

1.21 Theorem: arithmetic rules of limits

Suppose lim f (x) and lim g(x) exist. Then the following rules hold.
x→a x→a

1. Sum Rule:
lim [f (x) + g(x)] = lim f (x) + lim g(x).
x→a x→a x→a

2. Difference Rule:
lim [f (x) − g(x)] = lim f (x) − lim g(x).
x→a x→a x→a

3. Product Rule:
lim [f (x) g(x)] = lim f (x) · lim g(x).
x→a x→a x→a
In particular, lim [cf (x)] = c lim f (x), where c is any constant.
x→a x→a

4. Quotient Rule: If lim g(x) 6= 0, then

x→a

f (x) lim f (x)

lim = x→a .
x→a g(x) lim g(x)
x→a

1.22 Example: arithmetic rules

Using the product rule of limits, we get lim x2 = ( lim x)2 = a2 . By induction, it is easy
x→a x→a
to have lim xn = an for any positive integer n.
x→a
Then using the sum rule of limits, we get

lim (cn xn + cn−1 xn−1 + · · · + c1 x + c0 )

x→a
= cn an + cn−1 an−1 + · · · + c1 a + c0 .

Thus for any polynomial f (x), we have lim f (x) = f (a).

x→a
f (x)
If f (x) and g(x) are polynomials with g(a) 6= 0, then for the rational function r(x) = ,
g(x)
by using the quotient rule, we have

f (x) lim f (x) f (a)

lim r(x) = lim = x→a = = r(a).
x→a x→a g(x) lim g(x) g(a)
x→a

1.23 Pitfall: arithmetic rules - conditions not met

For every arithmetic rule, there are associated conditions under which the rule holds. If
these conditions are not all satisfied, then the corresponding rule may fail.
x2 − 1
For example, if we hope to calculate lim 4 , we cannot directly apply the quotient
x→1 x − 1
rule since lim (x − 1) = 0.
4
x→1
14 Limit and Continuity

The correct way to calculate the limit is y

first to simplify the expression by eliminating
x2 − 1
the “zero” factor: y=
x4 − 1
1
x2 − 1 x2 − 1 1
= = 2 . 1
x −1
4 (x − 1)(x + 1)
2 2 x +1 2

x
−1 O 1

Then we can apply the quotient rule to get

x2 − 1 1
lim = lim 2 = 12 .
x→1 x4 − 1 x→1 x + 1

1.24 Theorem: Squeeze Theorem

Suppose

1. f (x) ≤ g(x) ≤ h(x) near a (not necessarily at a).

2. lim f (x) = lim h(x) = L.
x→a x→a

Then lim g(x) = L.

x→a

1.25 Example: Squeeze Theorem

1
Calculate lim x sin .
x→0 x
We cannot use the product rule since y
1
lim sin does not exist. However, we can 1 y=x
x→0 x
apply the Squeeze Theorem. In fact, since
1
1

y = x sin
0 ≤ x sin  ≤ |x|, we have
 
x
x x
−1 O 1
1
−|x| ≤ x sin ≤ |x|. y = −x
x
1
Because lim |x| = lim (−|x|) = 0, by the Squeeze Theorem, we get lim x sin = 0.
x→0 x→0 x→0 x

1.26 Example: Squeeze Theorem and arithmetic rules

1 1
 
Calculate lim (x − 1)2 + sin .
x→1 x2 − 1 x−1
1 1
Since the values of can be made arbitrarily large when x approaches 1 and sin
x2 −1 x−1
1.2 Limit 15

1 1
 
is bounded by 1, + sin does not have limit (or approaches ∞) and is not
x2 − 1 x−1
bounded either. Thus we can apply neither the product rule nor the Squeeze Theorem.
However, we can do the following changes before we take the limit:

1 1
 
(x − 1) 2
+ sin
x2 − 1 x−1
1 1
= (x − 1)2 · + (x − 1)2 sin expanding
(x − 1)(x + 1) x−1
x−1 1
= + (x − 1)2 sin . cancelling 0-factor
x+1 x−1
For the first term above, we have
x−1
lim = 0. applying the quotient rule
x→1 x+1
For the second term, since
1
−(x − 1)2 ≤ (x − 1)2 sin ≤ (x − 1)2
x−1
and both (x − 1)2 and −(x − 1)2 approaches zero as x → 1, we apply the Squeeze Theorem
and get
1
lim (x − 1)2 sin = 0.
x→1 x−1
Thus, y

1 1
 
lim (x − 1)2 + sin
x→1 x2 − 1 x−1
x−1 1 1
= lim + lim (x − 1)2 sin x
x→1 x + 1 x→1 x−1 O 1

= 0 + 0 = 0.

One-Sided Limits

1.27 Definition: one-sided limits

We write
lim f (x) = L
x→a−

and say the left-hand limit of f (x) as x approaches a [or the limit of f (x) as x ap-
proaches a from the left] is equal to L, if f (x) is arbitrarily close to L for all x sufficiently
close to a with x < a.
16 Limit and Continuity

Similarly, we write
lim f (x) = L
x→a+

and say the right-hand limit of f (x) as x approaches a [or the limit of f (x) as x ap-
proaches a from the right] is equal to L, if f (x) is arbitrarily close to L for all x sufficiently
close to a with x > a.

1.28 Theorem: relationship - the limit and the one-sided limits

lim f (x) = L if and only if lim f (x) = lim+ f (x) = L.

x→a x→a− x→a

1.29 Example: one-sided limits

√
1. When we calculate the limit lim x, since y
x→0
x must be non-negative, we in fact are
√ cal-
culating the right-hand limit lim+ x. 1 y=
√
x
x→0

x
O 1

2. Consider the function

x + 1, if x ≤ 0;

f (x) = y
sin x, if x > 0.
• When x < 0, f (x) = x + 1. So, y = f (x) 1

lim f (x) = lim (x + 1) = 1. x

x→0− x→0− −1 O 1 π
2π

• When x > 0, f (x) = sin x. So, −1

lim f (x) = lim+ sin x = 0.

x→0+ x→0

Because lim− f (x) 6= lim+ f (x), we conclude that the limit lim f (x) does not exist.
x→0 x→0 x→0

3. Consider the function

if x ≤ 0;

x,
g(x) = y
sin x, if x > 0.
y = g(x)
• When x < 0, g(x) = x. So, 1

lim g(x) = lim− x = 0. x

x→0− x→0 −1 O 1 π
2π

• When x > 0, g(x) = sin x. So, −1

lim g(x) = lim+ sin x = 0.

x→0+ x→0
1.2 Limit 17

Because lim g(x) = 0 = lim+ g(x), we conclude that lim g(x) = 0.

x→0− x→0 x→0

1.30 MATLAB: one-sided limits

The MATLAB command limit can also calculate the one-sided limit of a function. The
following example shows how to calculate the left-hand limit of f (x) = x2 as x approaches
2, i.e., lim− x2 , in MATLAB.
x→2

>> syms x ;
>> limit ( x ^2 ,x ,2 , ' left ')

ans =

MATLAB returns 4 as the result of the left-hand limit.

The right-hand limit of f (x) = x2 as x approaches 2, i.e., lim+ x2 , can also be calculated
x→2
using similar MATLAB code, shown as follows:

>> syms x ;
>> limit ( x ^2 ,x ,2 , ' right ')

ans =

MATLAB returns 4 as the result of the right-hand limit.

1.31 Example: lim sin x and lim cos x

x→a x→a

Firstly, let us calculate lim sin x. y

x→0
From the diagram, we see that the arc- 1
length from the point (1, 0) to the point
(cos x, sin x) is equal to x which is bigger than
the blue side of the triangle with length equal
(cos x, sin x)
to sin x. Thus when 0 < x < π/2, we have
0 < sin x < x. x
x
O 1

By the Squeeze Theorem, we get lim+ sin x = 0. Because sin x is an odd function, i.e.,
x→0
sin(−x) = − sin x, we further have lim sin x = lim [− sin(−x)] = 0. Thus we get
x→0− −x→0+
lim sin x = 0.
x→0
18 Limit and Continuity

Furthermore, since

x + a 
   
x−a x − a 
| sin x − sin a| = 2 sin cos ≤ 2 sin
 
  ,
2 2  2 

by the Squeeze Theorem, we get lim sin x = sin a.

x→a
Since cos x = sin π2 − x = − sin x − π2 , we have

lim cos x = lim − sin x − π2



x→a x→a
= − lim π sin x = − sin a − π
= cos a.


x→a− 2 2

sin θ
1.32 Example: lim =1
θ→0 θ

sin θ
We cannot apply the quotient rule to calculate the limit lim , since the limit of the
θ→0 θ
denominator is zero.

We use geometry and the Squeeze Theo- y

rem to calculate the limit. In fact, from the T

diagram shown, for 0 < θ < π2 , we have 1

P
Area 4OSP < Area sector OSP
1 tan θ
< Area 4OST , sin θ

or θ
x
O cos θ Q S
1
· sin θ· < π1 1 1
· 1 · tan θ.
2


2 θ· 2π < 2

These inequalities give

sin θ
cos θ < < 1.
θ
sin θ
Since lim+ cos θ = cos 0 = 1, by the Squeeze Theorem, we obtain lim+ = 1. Because
θ→0 θ→0 θ
sin θ is an odd function, we further have

sin θ sin(−θ) sin θ

lim− = lim− = lim+ = 1.
θ→0 θ θ→0 −θ θ→0 θ
sin θ
Hence, lim = 1.
θ→0 θ
1.2 Limit 19

Limits at Infinities

1.33 Example: limits at infinities

1 y
For the function y = , we see that when
x
x takes arbitrarily large values, f (x) will ap-
proach 0. The line y = 0 is called the 1
asymptote of the function. This is one of the y=
x
key components in curve sketching. x
O

1.34 Definition: limits at infinities

• If the values of f (x) can be made arbitrarily close to L by taking x sufficiently large
positive values, then we write
lim f (x) = L.
x→+∞

• If the values of f (x) can be made arbitrarily close to L by taking x sufficiently large
negative values, then we write
lim f (x) = L.
x→−∞

• If the values of f (x) can be made arbitrarily close to L by taking |x| sufficiently large
values, then we write
lim f (x) = L.
x→∞

1.35 Definition: horizontal asymptote

The line y = b is called a horizontal asymptote of the curve y = f (x) if either

lim f (x) = b
x→+∞

or
lim f (x) = b.
x→−∞

1.36 Example: horizontal asymptote

The arithmetic rules and the Squeeze Theorem still hold for limits at infinities.
x2 + 1
For the rational function , we can apply some arithmetic rules of limits to
2x + x + 3
2
20 Limit and Continuity

calculate the limit at infinity:

x2 + 1
lim comparing the leading powers
x→∞ 2x2 + x + 3
1
1+ 2
= lim x dividing each term by x2
x→∞ 1 3
2+ + 2
x x
1
 
lim 1 + 2
x→∞ x
=  = 12 . using the arithmetic rules of limits
1 3

lim 2 + + 2
x→∞ x x

Thus, this function has a horizontal asymptote x = 12 .

x2 + 1
y=
2x2 + x + 3
y= 1
2

x
O

1.37 Example: limit at infinity - difference of two radicals

Rationalizing radicals is a usual approach to deal with limits of this type before applying
arithmetic rules:
√ √

lim x+1− x
x→+∞
√ √
√ √
√ √
x+1− x x+1+ x x+1+ x
= lim √ √ multiplying √ √
x→+∞ x+1+ x x+1+ x
√
2 √ 2
x + 1 − ( x)
= lim √ √ simplifying
x→+∞ x+1+ x
1
= lim √ √ simplifying
x→+∞ x+1+ x
= 0.
1.2 Limit 21

This shows that y = 0 is a horizontal y

asymptote of the function.
√ √
1 y= x+1− x

x
O 1

1.38 MATLAB: limits at infinities

With the help of positive infinity inf, MATLAB can calculate the limit of√a function
√ at
infinities. The MATLAB code in the following shows how to calculate lim x+1− x .


x→+∞

>> syms x ;
>> f = sqrt ( x +1) - sqrt ( x ) ;
>> limit (f ,x , inf )

ans =

From the MATLAB code above, we see that

√ √

lim x + 1 − x = 0.
x→+∞

1.39 Example: limit at infinity - polynomials

Let f (x) = xn + an−1 xn−1 + · · · + a1 x + a0 be a polynomial. We can rewrite

 an−1 a0 
f (x) = xn 1 + + ··· + n .
x x
an−1 a0
When x approaches ±∞, the term 1 + + · · · + n approaches 1, and the term xn
x n x
approaches ∞. In fact, if n is an even integer, x approaches +∞; if n is an odd integer, xn
approaches ±∞. Thus we get the following useful result:
• When n is an even positive integer, a degree-n polynomial f (x) = xn + · · · + a1 x + a0
approaches positive infinity when x approaches infinites.
• When n is an odd positive integer, f (x) = xn +an−1 xn−1 +· · ·+a1 x+a0 approaches posi-
tive infinity (respectively, negative infinity) when x approaches positive infinity (respectively,
negative infinity).
22 Limit and Continuity

1.40 Example: limits at infinity - slant asymptote

x3 − 2x2 + 1
Let us study the behaviors of the function f (x) = near the infinities.
2x2 + 1
We factor out the highest power in numerator and denominator and have

2 1 2 1
   
x 3
1 − + 1 − +
x3 − 2x2 + 1 x x3 x x3
= = x ·  .
2x2 + 1 1 1
  
x2 2 + 2 2+ 2
x x
By applying the arithmetic rules, we have
2 1
 
lim 1− + 3 y
x→+∞ x x
 = 12 ,
1

lim 2+ 2 x3 − 2x2 + 1
x→+∞ x y=
2x2 + 1

so that
f (x) x
lim = 12 . O
x→+∞ x
Furthermore, we have 1
1 x−

lim f (x) − 12 x
  2
y=
x→+∞
x − 2x2 + 1 1
 3 
= lim − x
x→+∞ 2x2 + 1 2

4x2 + x − 2
 
= lim − = −1.
x→+∞ 4x2 + 2
We see that as x → +∞, the difference between f (x) and 12 x − 1 approaches zero. This
means that, near the positive infinity, the function y = f (x) behaves like the function y =
2 x − 1. Similarly, the function also behaves like the function y = 2 x − 1 near the negative
1 1

infinity. The line y = 2 x − 1 is called a slant asymptote of the function f (x).

1

Remark:
An alternating approach of finding a slant asymptote for a rational function is to
use polynomial long division. For the above example, the long division gives

x3 − 2x2 + 1 = 12 x − 1 (2x2 + 1) + 12 (−x + 4)

so that
x3 − 2x2 + 1 −x + 4
= 21 x − 1 + 2 ,
2x + 1
2 4x + 2
The quotient 21 x − 1 gives the slant asymptote of the function f .
1.2 Limit 23

1.41 Engineering Example: initial value theorem

In electronic engineering and control system, the time-domain behaviour of a system can be
characterized by the step response, which is the output of a system in the time domain when
the input is a unit-step function. The unit-step input means the input changes from 0 to 1
in a very short time. In other words, there is a sudden input. Note that a sudden input may
cause a fast and possibly large system performance deviation from its long-term steady state,
which may have severe effects on a certain system component or even the overall system.
Moreover, if a sudden input occurs, it may take time for the system output to settle down to
some vicinity of its final state before the system can act properly. Hence, it is important to
understand how the system responds to a sudden input, i.e., understand the step response.
The initial slope of the step response is one of the important quantities to describe the
step response. One way to calculate the initial slope of the step response is to use the Initial
Value Theorem. Let the initial slope of the step response of a system be denoted by
lim f (t), the Initial Value Theorem says that
t→0

lim f (t) = lim sF (s),

t→0 s→+∞

where F (s) is the (one-sided) Laplace transform of the step response f (t) and s is a complex
number frequency parameter. Note that the Laplace transform provides an alternative way
of analyzing the system in the frequency domain.

Now, we use the Initial Value Theorem to

calculate the initial slope of the step response
for a resistor-capacitor (RC) circuit.
Given that the step response
 of the RC cir-
R
t
cuit is vC (t) = 1 − e− RC u(t), where R is −
−
vI (t) vC (t)
the resistance of the resistor, C is the capaci- + C
+
tance of the capacitor and u(t) is the unit-step
function, then the Laplace transform of this
1
step response is VC (s) = .
s (RC · s + 1)

According to the property of Laplace transform, the Laplace transform of the slope of the
step response can be calculated as
1
F (s) = sVC (s) = .
RC · s + 1
Then, we use the Initial Value Theorem to calculate the initial slope of the step response and
have
s 1
lim f (t) = lim sF (s) = lim = .
t→0 s→+∞ s→+∞ RC · s + 1 RC
1
This shows that the initial slope of the step response is .
RC
24 Limit and Continuity

1.42 Engineering Example: final value theorem

A system or process is in a steady state when the behaviour of this system or process
is unchanging in time. The steady state is an important concept in many fields, e.g., eco-
nomics, electronics, chemical engineering, etc. For example, many design specifications of
the electronic systems are given in terms of the steady-state characteristics. Given that the
step response of a system is f (t), then the steady state of a system is defined as lim f (t).
t→+∞
According to the definition, the steady state can be obtained by calculating the limit of the
step response when time t approaches infinity. Alternatively, we can use the Final Value
Theorem to calculate the steady state. The Final Value Theorem is provided in the follow-
ing.

lim f (t) = lim sF (s),

t→+∞ s→0

where F (s) is the (one-sided) Laplace transform of f (t) and s is a complex number frequency
parameter. Note that the Laplace transform provides an alternative way of analyzing the
system in the frequency domain.

Now, we use the Final Value Theorem

to calculate the steady state of a resistor- R

capacitor (RC) circuit. −

Given that the Laplace transform of the vI (t)

−
+ C vC (t)
1 +
step response is VC (s) = ,
s (RC · s + 1)
where R is the resistance of the resistor and
C is the capacitance of the capacitor.

By applying the Final Value Theorem, we have

s
lim vC (t) = lim sVC (s) = lim = 1.
t→+∞ s→0 s→0 s (RC · s + 1)

This shows that the steady state of the RC circuit is 1.

Infinite Limits

1.43 Example: infinite limits

1 1
We know both functions f (x) = sin and g(x) = do not have limit at x = 0. However,
x x
we do know the behavior of g(x) near 0. In fact, when x approaches 0 from the right (x > 0),
the values of g(x) can be made arbitrarily large positive, and when x approaches 0 from the
left (x < 0), the values of g(x) can be made arbitrarily large negative. The infinite limits are
related to vertical asymptotes of the graph of a function.
1.2 Limit 25

y y

1
y=
1 x
1
y = sin 1
x
−1
x x
−1 O 1 O 1
−1
−1

1.44 Definition: infinite limits

• If the values of f (x) can be made arbitrarily large by taking x sufficiently close to a (but
not equal to a), then we write
lim f (x) = ∞.
x→a

• If the values of f (x) can be made arbitrarily large positive by taking x sufficiently close
to a (but not equal to a), then we write

lim f (x) = +∞.

x→a

• If the values of f (x) can be made arbitrarily large negative by taking x sufficiently close
to a (but not equal to a), then we write

lim f (x) = −∞.

x→a

Similarly, if we take x sufficiently close to from the left-hand side or the right-hand side of
a, then we can define

lim f (x) = ∞, lim f (x) = +∞, lim f (x) = −∞,

x→a− x→a− x→a−

and
lim f (x) = ∞, lim f (x) = +∞, lim f (x) = −∞.
x→a+ x→a+ x→a+

1.45 Definition: vertical asymptote

The line x = a is called a vertical asymptote of the curve y = f (x) if at least one of the
following is true:

lim f (x) = +∞, lim f (x) = +∞, lim f (x) = +∞,

x→a x→a− x→a+
lim f (x) = −∞, lim f (x) = −∞, lim f (x) = −∞.
x→a x→a− x→a+
26 Limit and Continuity

1.46 Example: vertical asymptotes

x
Consider the function f (x) = .
(x − 1)(x + 2)
When x approaches 1 from the right, x − 1 approaches 0 and the values of f (x) can be
made arbitrarily large. To determine whether it is the positive infinity or negative infinity,
let us check the sign of f (x). In fact, when x approaches 1 from the right, (x − 1) > 0,
x + 2 approaches 3 and hence, for x near 1 and x > 1, f (x) has the positive sign. So
lim+ f (x) = +∞. Hence x = 1 is a vertical asymptote for the graph of the function. One
x→1
can calculate similarly to get

lim f (x) = −∞, y

x→1−
lim f (x) = +∞,
x→−2+
x
lim f (x) = −∞. y=
(x − 1)(x + 2)
x→−2−

It also has a horizontal asymptote line y = x

−2 O 1
0 since when x → ∞, if we ignore −1 in
(x − 1) and 2 in (x + 2), we get lim f (x) =
x→∞ x = −2 x=1
x
lim = 0.
x→∞ x2

1.47 Pitfall: zero of the denominator ⇐⇒

6 vertical asymptote

x2 − 1
Consider the function f (x) = .
(x + 2)(x − 1)
The denominator equals (x + 2)(x − 1) which have two roots x = −2 and x = 1. It seems
f (x) should have vertical asymptotes x = −2 and x = 1. However,
x2 − 1
lim
x→1 (x + 2)(x − 1)

(x − 1)(x + 1)
= lim factoring the numerator
x→1 (x + 2)(x − 1)
x+1
= lim cancelling 0-factor
x→1 x + 2
= 23 .

Thus x = 1 is not a vertical asymptote of y

the function. This example shows that a zero
of the denominator is not necessarily a verti- y=
x2 − 1
cal asymptote. (x + 2)(x − 1)
1

x
−2 O 1
1.3 Continuous Functions 27

1.3 Continuous Functions

1.48 Example: continuous function - motivation

|x| x
Consider the following functions: f (x) = , g(x) = , and h(x) = x. The graphs are
x x
given below.

y
y y

y = g(x)
y = f (x) 1 y = h(x)
1

x x x
O 1 O 1 O
−1

We can see that the graph of h(x) moves continuously as x moves along the x-axis while
the graphs of both f (x) and g(x) have a break at x = 0. In order to distinguish these two
different behaviors of functions, we introduce the concept of continuous function. Let us
check the difference between these three functions in order to get a correct definition. For
the function f (x), the limit lim f (x) does not exist. For the function g(x), the limit lim g(x)
x→0 x→0
exists but g(x) is not defined at x = 0. For the function h(x), the limit lim h(x) exists and
x→0
equals the function value at x = 0, i.e., lim h(x) = h(0).
x→0

1.49 Definition: continuous function

• A function f is continuous at x = a if

1. lim f (x) exists;

x→a

2. f (x) is defined at x = a;
3. lim f (x) = f (a).
x→a

• A function f is continuous in an interval (α, β) if it is continuous at every number in

the interval.
• A function f is continuous on an closed interval [α, β] if it is continuous at every
number in the interval (α, β), and also continuous at the endpoints. The latter means that

lim f (x) = f (α) and lim f (x) = f (β).

x→α+ x→β −

• If f is not continuous at a point a in its domain, we say that f is discontinuous at a.

28 Limit and Continuity

1.50 Theorem: continuity with arithmetic operations

If f and g are continuous at a and c is a constant, then the following functions are also
continuous at a:

1. f + g; 4. f · g;
2. f − g;
f
3. cf ; 5. , provided that g(a) 6= 0.
g

1.51 Example: discontinuous functions

1. Consider the function y

 |x| ,

if x 6= 0; y = f (x) 1
f (x) = x
0, if x = 0. x

O 1

−1
It is easy to calculate the one-sided limits:

|x| −x
lim f (x) = lim− = lim− = −1,
x→0− x→0 x x→0 x
|x| x
lim f (x) = lim+ = lim+ = 1.
x→0+ x→0 x x→0 x

Since the left and right-hand limits are not equal, the limit lim f (x) does not exists.
x→0
Hence, the function f is not continuous at x = 0.

2. Consider the function y

 x
 , if x 6= 0; y = g(x)
1
g(x) = x
0, if x = 0.

x
O 1

It is clear that g(x) = 1 if x 6= 0, so that

lim g(x) = 1, lim g(x) = 1.

x→0− x→0+

Since the left and right-hand limits are both equal to 1, the limit lim g(x) exists and
x→0
equals to 1. However, since lim g(x) = 1 6= g(0) = 0, the function is not continuous at
x→0
x = 0.

1.52 Example: continuous elementary functions

Besides the facts on continuity under arithmetic operations, we can prove the following
elementary functions are continuous in their domains.
1.3 Continuous Functions 29

1. Polynomials are continuous in R.

2. The functions sin x and cos x are continuous in R.

√
3. If n is an odd positive integer, the function f (x)
√ = n x is continuous for all x. If n is
an even positive integer, the function f (x) = n x is continuous for all x ≥ 0.

1.53 Engineering Example: amplitude modulation

Communication technology plays an important role in modern society by connecting peo-

ple together. For example, radio broadcast program has been very popular for providing
information and entertainment to people. There are two popular transmission technologies
used for radio broadcasting, namely, AM (amplitude modulation) and FM (frequency mod-
ulation). For both schemes, the programs (music, news, stories, etc.) are brought to people
by radio waves, which serve as the carrier signal. In particular, in AM, the carrier signal is a
sinusoidal function, and the sinusoidal waveform is modulated by an information-bearing
audio waveform before transmission. In this way, the amplitude (signal strength) of the
carrier signal is varied in proportion to the information-bearing signal to be transmitted.

1.54 Engineering Example: sinusoidal function - signal processing

An engineering system usually can be described by its response to a certain input function.
The sinusoidal function is one of such input functions. The sinusoidal function, denoted as
sin x, is one of the trigonometric functions, which is thus a periodical function. The sinusoidal
function sin x (i.e., with period 2π) is plotted in the following.

y = sin x 1

x
−8 −2π − 32 π −π − 12 π O 1 π 3 2π 8
2π 2π
−1

Many systems, e.g., audio system, can be described by their responses to the sinusoidal
function. For linear time-invariant (LTI) systems, the sinusoidal input is especially important,
because the output is also a sinusoidal function with the same frequency as the input, but
possibly different amplitudes and phase angels. Then, the frequency response, which is an
important way of charactering a system, will be utilized to measure the amplitude and the
phase angel of the sinusoidal output.

1.55 Engineering Example: sinusoidal function - aeronautics and astronautics

As one of the four “earth-like” terrestrial planets in the solar system, Mars has been
explored by multiple organizations and space agencies. For example, the National Aeronautics
and Space Administration (NASA) has been exploring Mars for more than 40 years. In
January 2004, NASA landed on the surface of Mars with the Mars exploration rovers, spirit
30 Limit and Continuity

and Opportunity. However, humans have not landed on Mars yet. To successfully complete
the “human landing on Mars” mission, the communication between Mars and Earth is vital.
But, there is normally a significant communication delay due to the long distance between
these two planets (on average, over 100 million kilometers). Since Mars and Earth orbit Sun
at different rates, the distance between them is constantly varying. A natural question thus
arises as to when is Mars closest to Earth (so that the communication delay is the smallest)?
According to some data collected by NASA scientists and equipments, it turns out that the
distance between Mars and Earth can be modelled as a sinusoidal function (with shift),
which is plotted in the following.

y (×108 km)


y = 1.72 sin π
386
t × 108 + 2.27 × 108
4

2.27

0.557
t (×102 days)
O 2 4 6 7.72 10

We can see from the above figure that the distance between Mars and Earth is in a 772-day
(i.e., around 26 months) cycle and the closest distance between these two planets is around
5.57 × 107 kilometers. Given that the radio waves which can carry information travel in space
at the speed of light, the one-way communication time is around 3.1 minutes when Mars is
closest to Earth.

1.56 Engineering Example: unit-step function

In mathematics, the unit-step function is y

defined as follows:
y = H(x) 1
1, if x ≥ 0;

H(x) =
0, if x < 0. −1 O 1
x

It is a discontinuous function and not contin- −1

uous at x = 0.

The unit-step function is a fundamental function in engineering and can be widely applied
to many fields, e.g., circuit analysis and communication systems. In particular, for circuit
analysis, the unit-step function can model the behaviour of a switch (on/off). Moreover, it
can also describe sudden changes of the voltage or the current in a circuit. In communication
systems, the sum of unit-step functions can be used to represent waveforms.
1.3 Continuous Functions 31

Composite Functions
1.57 Definition: composite function

Let f and g be two functions such that the range of g is contained in the domain of f , we
can get a new function, called the composite function h, defined by h(x) = f (g(x)).

1.58 Example: composite functions

√
Let f (x) = x and g(x) = x2 + √ 1. Then
the composite function f (g(x)) = x2 + 1. y
Note that the composition of two functions y = g(f (x))
depends on the order√ of the composition. For y = g(x)
example, g(f (x)) = ( x)2 + 1 = x + 1. Thus
f (g(x)) 6= g(f (x)).
y = f (g(x))
In fact, f (g(x)) and g(f (x)) have not only y = f (x)
different expressions, but also different do- 1

mains: the function f (g(x)) is defined for all x

O 1
real number x; the function g(f (x)) is defined
for x ≥ 0.

1.59 MATLAB: composite function

MATLAB can show the composite function with command compose. The following exam-
1
ple shows the composite function h(x) = f (g(x)), where f (x) = and g(x) = 1 + x2 .
x

>> syms x ;
>> f = 1/ x ;
>> g = 1+ x ^2;
>> h = compose (f , g )

h =

1/( x ^2 + 1)

1.60 Theorem: continuity of composition of continuous functions

Given two continuous functions f and g such that the composition f (g(x)) exists. Then
the composite function f (g(x)) is continuous.

1.61 Example: continuous composite functions

Let f (x) = sin x and g(x) = x2 + 1. Then the composite functions f (g(x)) = sin(x2 + 1)
and g(f (x)) = (sin x)2 + 1 are continuous.
32 Limit and Continuity

Intermediate Value Theorem

1.62 Theorem: Intermediate Value Theorem

Let f be a continuous function over a closed interval [a, b]. For any number ξ between f (a)
and f (b), there exists at least one c in the interval [a, b] such that f (c) = ξ.

y y

f (b) f (b)

ξ
ξ

f (a) f (a)

x x
O a c b O a c1 c2 c3 b

1.63 Remark: key hypothesis in the Intermediate Value Theorem

Continuity of f is the key hypothesis in the Intermediate Value Theorem. If f is not

continuous, then the conclusion of the theorem may not hold.

For example, consider the function f de- y

fined on [−1, 3]:
y = f (x)
2
sin x, if 0 ≤ x ≤ 1;

f (x) = ξ = 1.5
2, if 1 < x ≤ 3.

Clearly, f (0) = 0 and f (3) = 2. For the in- x

termediate value ξ = 1.5, there is no values c O 1 3

in [0, 3] such that f (c) = 1.5.

1.64 Example: Intermediate Value Theorem - existence of roots y = f (x)

1. Consider the function f (x) = x3 . It is y

clear that f is continuous on [−1, 1], with
f (−1) = −1 and f (1) = 1. To see whether 1
there exists c in the interval [−1, 1] such
that f (c) = 0, we examine the equation
f (x) = 0. Such a solution is a root of f . x
−1 O 1
It is easy to see that f indeed has a root
c = 0, for which we do not need to apply −1
the Intermediate Value Theorem.
1.4 Some Transcendental Functions 33

2. Consider another function g(x) = x3 −5x+ y

2. It is not obvious that g has a root in
[0, 1]. However, it is clear g is continu- 2
y = x3 − 5x + 2
ous, with g(0) = 2 and g(1) = −2. Since
−2 < 0 < 2 (i.e., 0 is an intermediate x
value), from the Intermediate Value Theo- O c 1

rem, there exists c in the interval [0, 1] such

that g(c) = 0, i.e., g(x) = 0 has a root in
−2

the interval [0, 1].

1.65 Example: Intermediate Value Theorem - third-order polynomial

The Intermediate Value Theorem can be used to show that f (x) = x3 + ax2 + bx + c, which
is continuous everywhere, always has a real root for any numbers a, b and c.
In fact, since lim f (x) = +∞, there exists b > 0 such that f (b) > 0. Similarly since
x→+∞
lim f (x) = −∞, there exists a < 0 such that f (a) < 0. Because 0 is an intermediate value
x→−∞
of f (a) and f (b), from the Intermediate Value Theorem, on the closed interval [a, b], there
exists c in [a, b] such that f (c) = 0.

1.4 Some Transcendental Functions

Exponential Functions

1.66 Theorem: properties of exponential functions

The exponential function is f (x) = ax , where a is a positive number.

1. The exponential function f (x) = ax has domain R and range (0, +∞).
2. The exponential function f (x) = ax is a continuous function.

3. The graph of the exponential function f (x) = ax depends the value of a.

4. If a > 1, then lim ax = +∞ and lim ax = 0; if 0 < a < 1, then lim ax = 0 and
x→+∞ x→−∞ x→+∞
lim ax = +∞.
x→−∞

5. The following are the graphs of ax for various bases.

34 Limit and Continuity

( 12 )x 10x 2x
( 14 )x 1.5x

0.8x
1.2x

1x

x
O 1

1.67 Definition: the number e

Among all exponential functions ax with different base a, there is a base which will play
an important role in calculus. This number is denoted by e, defined by
x
1

e = lim 1+ .
x→+∞ x

This limit exists and e is an irrational number, approximately 2.71828.

1.68 MATLAB: exponential functions

In MATLAB, the exponential functions can be simply written in the following form.

>> syms x ;
>> f = 3^ x

f =

3^ x

where 3 is the base.

Specifically, the command exp can be used for the exponential functions with base e. The
following examples show the exponential functions e and e2 , respectively.

>> exp (1)

ans =

2.7183

>> exp (2)

1.4 Some Transcendental Functions 35

ans =

7.3891

1.69 Example: exponential function - population growth

Suppose a bacteria culture with 100 bac-

teria triples in every hour. To write the y
function describing the bacteria population
2500
growth, let x be the variable representing
number of hours, and y = f (x) be the func- 2000
tion of bacteria population by x hours. We y = 100 · 3x
1500
observe
1000
f (1) = 100 × 3,
500
f (2) = 100 × 32 ,
100
f (3) = 100 × 33 . O 1 2 3
x

So, in general, we get y = 100 · 3x .

1.70 Example: exponential function - radioactive decay

Carbon-14 is a radioactive substance with half-life 5730 years. Half-life means that 1-pound
of Carbon-14 will be half-pound after 5730 years.
Suppose the initial amount of Carbon-14 is A0 and f (t) represents the amount after t
years. Then

f (0) = A0 ,
f (5730) = A0 /2,
f (2 × 5730) = A0 /22 .

So, f (k × 5730) = A0 /2k . Let k × 5730 = t. y

t
Then we get k = , so that the amount
5730
of Carbon-14 after t-years is A0
y = A0 · 2−t/5730
A0
f (t) = t = A0 · 2−t/5730 . 0.5 × A0
2 5730
0.25 × A0
1
If we further write a = 2− 5730 , then we get
t
y = f (t) = A0 at , which is an exponential O 5730 × 1 5730 × 2 5730 × 3

function.
36 Limit and Continuity

1.71 Example: exponential function - bank interest

A bank has a savings account with the annual interest rate r. Let P be the initial deposit.
If the bank provides interest m times per year, then after t years, the amount A is given by
 r mt
A=P 1+ .
m
If the interest is calculated continuously, i.e. let m go to infinity, the amount A is given by
 r mt
A= lim P 1 +
m→+∞ m
rt
r  mr

= P lim 1+
m→+∞ m
rt
r  mr
 
=P lim 1+
m→+∞ m
x rt
1
 
= P lim 1+ = P ert .
x→+∞ x

Here the number e appears naturally.

The graph compares different models for y

fixed P and r. It shows that the yield in- m = +∞
m=4
creases as m increases. In other words, the y = P ert m=1
bank will pay you more money if the inter-
est is compounded more often. The amount
r mt
will reach its maximum if the bank calculates

y =P 1+
m
the compound interest (for which they will P

avoid). O 1 2 3 4 5 6 7
t (year)

Inverse Function

1.72 Definition: one-to-one

We say f (x) is one-to-one if it never takes on the same value twice, i.e., if x1 6= x2 , then
f (x1 ) 6= f (x2 ), or equivalently, if f (x1 ) = f (x2 ), then x1 = x2 .

1.73 Theorem: horizontal line test

A function is one-to-one if and only if no horizontal line intersects its graph more than
once.
1.4 Some Transcendental Functions 37

1.74 Example: one-to-one functions

1. The function f (x) = x3 is a one-to-one y

function. It is easy to check that if f (x1 ) =
f (x2 ), then x1 = x2 . It is also easy to ver-
ify that the function f passes the horizon- y = x3 + 1
tal line test. 1

x
−1 O

2. The function g(x) = x2 defined over y

(−∞, +∞) is not one-to-one because
g(1) = 1 = g(−1). It is easy to see that
2
the function fails the horizontal line test.
y = x2 1

x
−1 O 1

3. The function h(x) = x2 with the domain y

[0, +∞) is one-to-one. In fact, if x21 = x22
and both x1 and x2 are non-negative, we
2
must have x1 = x2 . It is also easy to verify
that the function h passes the horizontal 1 y = x2
line test.
x
−1 O 1

1.75 Definition: inverse function

One can regard a function f with domain D and range R as a map f from D to R. If f
is one-to-one, then we can construct a map from R to D whose composition with f is the
identity map. The corresponding function of the new map is called the inverse function.
Let f be a one-to-one function with domain D and range R. Then its inverse function f −1
has domain R and range D and is defined by

f −1 (y) = x ⇐⇒ f (x) = y

for any y in R. Thus,

domain of f −1 = range of f ,
range of f −1 = domain of f .
38 Limit and Continuity

Theorem: cancellation equations

The function f and its inverse function f −1 satisfy the following cancellation
equations:
f −1 (f (x))
= x for every x in D;
f f −1 (x) = x for every x in R.

1.76 Method: find the inverse function

1. Write y = f (x).
2. Solve this equation for x in terms of y (if possible).
3. To express f −1 as a function of x, interchange x and y. The resulting equation is
y = f −1 (x).

1.77 Example: inverse functions

1. The function f (x) = x3 is a one-to-one y

function. √The inverse function √ of f is
f −1 (x) = 3 x since f (f −1 (x)) = ( 3 x)3 =
x
y = x3 =
y
x.
√
y= 3
x
The inverse function f −1 can be found by x
O
the following steps: solving y = x3 to get
√
x = 3 y, √then interchanging x and y to
have y = 3 x.

√
One can see the graph of the inverse function f −1 (x) = 3 x is the mirror image of that
of the function f (x) = x3 with respect to the diagonal line y = x as the mirror.

2. The function h(x) = x2 defined on the y

domain [0, +∞) is one-to-one.
√ Its inverse
function is h−1 (x) = x. y = x2

One also sees the

√graph of the inverse func- √
y= x
tion h−1 (x) = x is the mirror image of x
O
that of the function h(x) = x2 with respect
to the diagonal line y = x as the mirror.
x
=
y
1.4 Some Transcendental Functions 39

3. The function p(x) = x2 defined on the y

domain [−∞, 0) is one-to-one.
√ Its inverse
function is p−1 (x) = − x.

x
=
y = x2

y
One again sees the √ graph of the inverse
function p−1 (x) = − x is the mirror im- √ x
O y=− x
age of that of the function h(x) = x2 with
respect to the diagonal line y = x as the
mirror.

1.78 Method: obtain the graph of f −1 from the graph of f

If f is a one-to-one function, then f (a) = b y

if and only if f −1 (b) = a. So, the point (a, b)
is on the graph of f if and only if the point

x
=
y
(b, a) is on the graph of f −1 . Hence, the graph y = f (x)
of f −1 is the reflection of the graph of f about
the line y = x. x
O

y = f −1 (x)

1.79 Theorem: continuity of the inverse function

The inverse function of any continuous one-to-one function is also continuous.

1.80 MATLAB: inverse function

In MATLAB, the inverse function can be calculated using command finverse. The fol-
lowing example shows how to calculate the inverse function of f (x) = x3 + 1 in MATLAB.

>> syms x ;
>> f = x ^3 + 1;
>> finverse ( f )

ans =

( x - 1) ^(1/3)

√ The following figure plots both the function f (x) = x3 +1 and its inverse function f −1 (x) =
3
x − 1.

>> x = -3:0.02:3;
40 Limit and Continuity

>> y = x .^3 + 1;
>> y_i = nthroot ( x - 1 ,3) ;
>> figure
>> plot (x ,y , 'r ')
>> hold on
>> plot (x , y_i , 'b ')
>> axis ([ -3 3 -3 3]) % Change the axis limits so that both ...
the x - axis and the y - axis range from -3 to 3
>> xlabel ( 'x ')
>> ylabel ( 'y ')
>> grid
>> legend ({ '$ f ( x ) = x ^3 + 1$ ' , '$ f ^{ -1}( x ) = ...
\ sqrt [3]{ x -1}$ '} , ' Interpreter ' , ' latex ')

0
y

-1

-2

-3
-3 -2 -1 0 1 2 3
x

Logarithmic Functions

1.81 Definition: logarithmic function

Since the exponential function ax is one-to-one, it has an inverse. The inverse function of
f (x) = ax is called logarithmic function with base a, denoted by loga x. We use ln x for
loge x.

1.82 Theorem: Laws of Logarithms

Let a, b be positive numbers not equal to 1, A, B be positive numbers, and α be a real

numbers.

1. loga (A · B) = loga A + loga B. 4. aloga A = A.

logb A
 
A
2. loga = loga A − loga B. 5. loga A = .
B logb a
3. loga Aα = α loga A. 6. loga a = 1, loga 1 = 0.
1.4 Some Transcendental Functions 41

1.83 Theorem: properties of logarithmic functions

Assume a is positive number not equal to 1.

1. The logarithmic function f (x) = loga x has domain (0, +∞) and range R.
2. The logarithmic function f (x) = loga x is a continuous function.

3. The graph of the logarithmic function f (x) = loga x depends on the value of a.
4. If a > 1, then lim loga x = +∞ and lim+ loga x = −∞; if 0 < a < 1, then
x→+∞ x→0
lim loga x = −∞ and lim+ loga x = +∞.
x→+∞ x→0

5. The following are the graphs of loga x for various bases.

log1.5 x
log2 x
ln x
log4 x
log10 x

x
O 1

log1/4 x
log1/2 x

log0.8 x

1.84 Example: logarithmic functions

Let us look at some examples using the fact that exponential functions and logarithmic
functions are inverse to each other.

1. To solve the equation ln(x − 1) = 0, we take exponential on both sides of the equation
simultaneously to have eln(x−1) = e0 . Using the identity eln a = a, we get x − 1 = 1, so
that x = 2.

2. To solve the equation 2x−1 = 5, we take logarithm on both sides of the equation
simultaneously to have log2 2x−1 = log2 5. Using the identity loga aα = α, we get
x − 1 = log2 5, so that

x = 1 + log2 5 = log2 2 + log2 5 = log2 10.

1.85 MATLAB: logarithmic functions

MATLAB provides commands for logarithmic functions with different bases, e.g., log2
for base 2, log for base e, and log10 for base 10. The following examples show functions
42 Limit and Continuity

f (x) = log2 (x), g(x) = ln(x) and h(x) = log10 (x) in MATLAB.

>> syms x ;
>> f = log2 ( x )

f =

log ( x ) / log (2)

>> g = log ( x )

g =

log ( x )

>> h = log10 ( x )

h =

log ( x ) / log (10)

1.86 Engineering Example: decibel

The decibel (dB) is used to measure the ratio of two quantities with the same unit. For
example, in terms of decibel, the ratio of two quantities A and B, which have the same unit,
is defined as
 
A
10 log10 (dB).
B

The decibel has been widely used in a variety of science and engineering fields, such as
electronics and control theory. For example, in electronics, the gains of amplifiers, the signal-
to-noise ratio (SNR), and the attenuations of signals are often represented in decibel.
Note that the decibel can be used to express the ratio of A with respect to any absolute
value B or a fixed reference value B. In the latter case, a suffix which indicates the reference
value is usually appended to the decibel symbol. For example, if the reference value B is 1
volt, then the suffix is “V”, e.g., 10 dBV; if the reference value B is 1 milliwatt, then the
suffix is “m”, e.g., 10 dBm.
The use of the decibel has several advantages. First, since the decibel is on a logarithmic
scale, it means that a very big ratio can be represented in a convenient way. For example, if
A is a trillion times of B, then in terms of decibel, we have
 
A
10 log10 = 10 log10 1012 = 120 (dB),
B

which is a more compact representation than 1, 000, 000, 000, 000.

1.4 Some Transcendental Functions 43

Second, due to the logarithmic scale property, the decibel can be used for simplifying the
calculations of multiplied effects of a system. For example, a multi-component system has
three components with gains g1,dB (dB), g2,dB (dB) and g3,dB (dB), respectively. Then the
overall gain go,dB (dB) of the system can be calculated as

go,dB = 10 log10 (go )

= 10 log10 (g1 g2 g3 )
 g1,dB g2,dB g3,dB 
= 10 log10 10 10 10 10 10 10
 g1,dB   g2,dB   g3,dB 
= 10 log10 10 10 + 10 log10 10 10 + 10 log10 10 10
= g1,dB + g2,dB + g3,dB ,

where go , g1 , g2 and g3 are the overall gain, the gain of the 1st component, the gain of the 2nd
component and the gain of the 3rd component in the linear scale, respectively. The above
additive operation is more convenient than the multiplication of multiple gains.

1.87 Engineering Example: RC charging circuit

The capacitor is an electronic component which can store electrical energy. If a resistor
is connected in series with the capacitor to form a resistor-capacitor (RC) circuit, then the
capacitor will charge up gradually through the resistor until the voltage across this capacitor
reaches the value of the supply voltage.

An example of the RC circuit is shown, R

where R represents the resistance of the re- i(t)

sistor, C represents the capacitance of the −

−

Vs C vC (t)
capacitor, Vs represents the supply voltage, +
+
and vC (t) represents the voltage across the
capacitor.

Assume that Vs = 10 V, R = 4.7 × 104 Ω, C = 10−3 F, and the initial voltage (i.e.,
the voltage at t = 0) across the capacitor is 0, it can be shown that the voltage across the
capacitor is given by
t
vC (t) = Vs (1 − e− RC ) (t ≥ 0).

We can find the time taken for the capacitor to achieve vC (t) = 9.8 V.
In fact, to achieve vC (t) = 9.8 V, we need to solve the following equation:
t
Vs (1 − e− RC ) = 9.8.

Substituting the values of Vs , R and C into the equation above, we have

t
−
10(1 − e 4.7×104 ×10−3 ) = 9.8.
44 Limit and Continuity

After some algebraic manipulations, we find that the time taken can be represented in terms
of a logarithmic function, and it is shown as follows:

t = −47 ln(1 − 0.98) ≈ 183.86 (s).

Note that the maximum voltage the capacitor can achieve is the supply voltage Vs . The
result above thus shows that it takes 183.86 seconds which is around 3 minutes for the
capacitor to virtually fully charged (i.e., achieve 98% of its maximum value).
Moreover, in RC circuit, the term τ = RC is defined as the time constant. In this example,
τ = RC = 47 (s). We notice that 183.86
47 ≈ 4, which means the time taken for the capacitor to
achieve 98% of its maximum voltage is 4 times of the time constant. The time period taken
for the capacitor to reach this 4τ point (i.e., achieve 98% of the maximum voltage) is known
as the transient period.

Inverse Trigonometric Functions

1.88 Definition: inverse trigonometric functions

• The function sin x restricted on [− π2 , π2 ] y

has unique inverse function, denoted by π

sin−1 x or arcsin x, that has domain [−1, 1] 2

sin−1 x
and range [− π2 , π2 ]. That is, 1
sin x

sin −1
x = y (|x| ≤ 1) − π2 −1 O 1 π
x
2

⇐⇒ sin y = x, − π2 ≤y≤ π
2.
−1

− π2

• The function of cos x restricted on [0, π] y

has unique inverse function, denoted by π
cos−1 x or arccos x, that has domain [−1, 1]
and range [0, π]. That is,
− π2
cos−1 x
cos−1 x = y (|x| ≤ 1) 1
cos x
⇐⇒ cos y = x, 0 ≤ y ≤ π. x
−1 O 1 π π
2

−1
1.4 Some Transcendental Functions 45

• The function of tan x restricted on y

(− π2 , π2 ) has unique inverse function, denoted
by tan−1 x or arctan x, that has domain tan x
(−∞, +∞) and range (− π2 , π2 ). That is, π
2

tan−1 x
tan−1 x = y (x ∈ R) x
− π2 O π
2

⇐⇒ tan y = x, − π2 <y< π
2. − π2

• The function of cot x restricted on (0, π) y

has unique inverse function, denoted by
cot−1 x or arccot x, that has domain (−∞, ∞)
π

cot x
and range (0, π). That is, π
2
cot−1 x

cot−1
x = y (x ∈ R) π
x
O π

⇐⇒ cot y = x, y ∈ (0, π).

2

• The function of csc x restricted on y

[− π2 , 0) ∪ (0, π2 ], has unique inverse function,
denoted by csc−1 x or arccsc x, that has do- csc x
main |x| ≥ 1 and range [− π2 , 0) ∪ (0, π2 ]. That π
2
csc−1 x
is, − π2
1

−1 O 1 π
x
csc−1 x = y (|x| ≥ 1) −1
2

− π2
⇐⇒ csc y = x, y ∈ [− π2 , 0) ∪ (0, π2 ].

• The function of sec x restricted on [0, π2 )∪ y

( π2 , π], has
unique inverse function, denoted
by sec x or arcsec x, that has domain |x| ≥
−1 π
sec x
1 and range [0, π2 ) ∪ ( π2 , π]. That is, π
2
sec−1 x
1

sec−1 x = y (|x| ≥ 1) −1 O 1 π π
x
2

⇐⇒ sec y = x, y ∈ [0, 2)
π
∪ ( π2 , π]. −1
46 Limit and Continuity

1.89 Example: inverse trigonometric functions

√ 
1. Calculate sin−1 2 .
2

√ 
Solution Let θ = sin−1 2
2
. Then, by the definition of sin−1 , θ is an angle in [− π2 , π2 ]
√
and sin θ = 2 .
So θ = π4 .
2

 √ 
2. Calculate cos−1 − 22 .
 √ 
Solution Let θ = cos−1 − 22 . Then, by the definition of cos−1 , θ is an angle in [0, π]
√
and cos θ = − 2 .
2
So θ = 4 .
3π

3. Find cos(sin−1 c) for c in [0, π/2].

Solution Let θ = sin−1 c, then sin θ = c. 1
We can draw a triangle with the opposite c

side of length c and hypotenuse of length

√ 1. θ
√
Then the adjacent side has length 1 − c2 . 1 − c2
Hence,
p p
cos(sin−1 c) = cos θ = 1 − c2 /1 = 1 − c2 .

1.90 MATLAB: trigonometric and inverse trigonometric functions

In MATLAB, there are several commands for calculating the trigonometric and inverse
trigonometric functions, which are provided in the following table.
Command Function
sin Sine function
cos Cosine function
tan Tangent function
cot Cotangent function
sec Secant function
csc Cosecant function
asin Inverse sine function
acos Inverse cosine function
atan Inverse tangent function
acot
Inverse cotangent function
asec Inverse secant function
acscInverse cosecant function
The following examples show the calculations of sin π2 and tan−1 (1), respectively, in

MATLAB.

>> sin ( pi /2)

ans =
1.4 Some Transcendental Functions 47

>> atan (1)

ans =

0.7854

Besides calculating the values of the trigonometric and inverse trigonometric functions,
graphs of the trigonometric and inverse trigonometric functions can also be easily plotted by
MATLAB. The following code shows the plot of cos−1 (x) within the interval [−1, 1]:

>> x = -1:0.01:1;
>> y = acos ( x ) ;
>> figure
>> plot (x ,y , 'r ')
>> xlabel ( 'x ')
>> ylabel ( 'y = cos ^{ -1}( x ) ')
>> grid

The plotted figure is shown below.

3.5

2.5
y = cos -1 (x)

1.5

0.5

0
-1 -0.5 0 0.5 1
x

1.91 Theorem: continuity of basic elementary functions

The following types of basic elementary functions are continuous at every point in their
domains:

polynomials, rational functions, root functions,

trigonometric functions, inverse trigonometric functions,
exponential functions, logarithmic functions.
48 Limit and Continuity

1.92 Engineering Example: inverse tangent - aeronautics and astronautics

It is said that the National Aeronautics and Space Administration (NASA) astronauts will
explore the surface of Moon again by 2020. This time, the astronauts are going to stay, build
outposts, and pave a way for the journeys to Mars and beyond. It is thus critical to have a
reliable and robust surface-to-surface communication between astronauts on Moon, and also
keep constant communications with Earth. This will likely be accomplished by a combined
use of communication satellites in orbit around Moon and communication equipments on
the lunar surface. In the case of no satellite coverage over an outpost, the astronauts there
will purely rely on the communication equipments on the lunar surface, e.g., communication
towers. Since Moon is a sphere, the surface-to-surface communication is limited by the height
of the communication tower. In particular, the communication signals cannot be received
beyond the point of tangency P in the following figure.

Parameter Description
r radius of Moon (1738.14 km)
h tower height
d line of sight distance
P point of tangency
a arc length
θ central angle measure

In the case of no satellite coverage over a certain area on the lunar surface, if there is a
communication tower with height 60 meters, then calculate the maximum distance that the
astronaut can explore in this area.
First, we note that the area which the astronaut can explore should be within the com-
munication range of the communication tower, i.e., not beyond the point of tangency P.
Moreover, the distance that the astronaut can explore refers to the one on the lunar surface.
Hence, the maximum distance is the arc length a in the above figure.
In order to calculate the arc length a, we first need to calculate the central angel measure
θ which can be obtained by using the inverse tangent function. Specifically, we have
 
d
θ = tan −1
.
r

We have known the Moon radius r = 1738.14 km. Now, we need to calculate the line of
2
sight distance d. We note that r2 + d2 = (r + h) , so we have
q
2
d = (r + h) − r2
q
2
= (1738.14 + 0.06) − 1738.142 ≈ 14.44 (km).
1.4 Some Transcendental Functions 49

Substituting the value of d to the first equation, we obtain

14.44
 
θ = tan −1
≈ 0.0083.
1738.14

Finally, we can calculate the arc length a, i.e., the maximum distance that the astronaut
can explore, as follows:
θ
a= × 2πr = 0.0083 × 1738.14 ≈ 14.43 (km).
2π
50 Summary

Summary
Definition: basic concepts

• function : f: D ⊂R→R
(D = domain, R = range)

• function continuous at a : lim f (x) = f (a)

x→a

• function continuous in I : lim f (x) = f (a) for every a ∈ I

x→a

• composite function : f (g(x))

• one-to-one function : if x1 6= x2 , then f (x1 ) 6= f (x2 )

⇐⇒ f passes the horizontal line test.

• inverse function : f −1 (y) = x ⇐⇒ f (x) = y

(f has to be one-to-one)

Definition: limit

limit : lim f (x) = L

x→a
⇐⇒ f (x) approaches L when x ∈ D approaches a

Definition: asymptotes

• horizontal asymptote y = b if at least one of the following is true:

lim f (x) = b, lim f (x) = b.

x→+∞ x→−∞

• vertical asymptote x = a if at least one of the following is true:

lim f (x) = +∞, lim f (x) = +∞, lim f (x) = +∞,

x→a x→a− x→a+
lim f (x) = −∞, lim f (x) = −∞, lim f (x) = −∞.
x→a x→a− x→a+

Theorem: limit rules and theorems

Suppose lim f (x) and lim g(x) exist. Then the following rules hold.
x→a x→a

1. Sum Rule:
lim [f (x) + g(x)] = lim f (x) + lim g(x).
x→a x→a x→a

2. Difference Rule:

lim [f (x) − g(x)] = lim f (x) − lim g(x).

x→a x→a x→a
 51

3. Product Rule:
lim [f (x) g(x)] = lim f (x) · lim g(x).
x→a x→a x→a

In particular, lim [cf (x)] = c lim f (x), where c is any constant.

x→a x→a

4. Quotient Rule: If lim g(x) 6= 0, then

x→a

f (x) lim f (x)

lim = x→a .
x→a g(x) lim g(x)
x→a

5. Squeeze Theorem: If f (x) ≤ g(x) ≤ h(x) and lim f (x) = lim h(x) = L, then
x→a x→a
lim g(x) = L.
x→a

6. Limit and One-Sided Limit: lim f (x) = L if and only if lim− f (x) = lim+ f (x) =
x→a x→a x→a
L.

Theorem: continuity of basic elementary functions

The following types of basic elementary functions are continuous at every point in their
domains:

polynomials, rational functions, root functions,

trigonometric functions, inverse trigonometric functions,
exponential functions, logarithmic functions.

Theorem: composite function and inverse function

The composite function of two continuous functions is a continuous function.

The inverse function of a one-to-one continuous function is a continuous function.

Theorem: continuity of arithmetic operations

If f and g are continuous at a and c is a constant, then the following functions are also
continuous at a:

1. f + g; 4. f · g;
2. f − g;
f
3. cf ; 5. , provided that g(a) 6= 0.
g
52 Summary

Theorem: two important limits

sin x
x
1

1. lim = 1. 2. lim 1+ = e.
x→0 x x→+∞ x

Method: find and graph the inverse function

If f is a one-to-one function, then f (a) = b y

if and only if f −1 (b) = a. So, the point (a, b)
is on the graph of f if and only if the point

x
=
y
(b, a) is on the graph of f −1 . Hence, the graph y = f (x)
of f −1 is the reflection of the graph of f about
the line y = x. x
O

y = f −1 (x)

Theorem: Laws of Exponents

Let a, b be positive numbers and α, β be real numbers.

α
1. aα · aβ = aα+β . 4. (a · b) = aα · bα .
aα 1
2. = aα−β . 5. a−α = .
aβ aα
β
3. (aα ) = aα·β . 6. a1 = a, a0 = 1.

Theorem: Laws of Logarithms

Let a, b be positive numbers not equal to 1, A, B be positive numbers, and α be a real

numbers.

1. loga (A · B) = loga A + loga B. 4. aloga A = A.

logb A
 
A
2. loga = loga A − loga B. 5. loga A = .
B logb a
3. loga Aα = α loga A. 6. loga a = 1, loga 1 = 0.

Theorem: properties of exponential functions

The exponential function is f (x) = ax , where a is a positive number.

1. The exponential function f (x) = ax has domain R and range (0, +∞).
2. The exponential function f (x) = ax is a continuous function.
3. The graph of the exponential function f (x) = ax depends the value of a.
 53

4. If a > 1, then lim ax = +∞ and lim ax = 0; if 0 < a < 1, then lim ax = 0 and
x→+∞ x→−∞ x→+∞
lim ax = +∞.
x→−∞

5. The following are the graphs of ax for various bases.

( 12 )x 10x 2x
( 14 )x 1.5x

0.8x
1.2x

1x

x
O 1

Theorem: properties of logarithmic functions

Assume a is positive number not equal to 1.

1. The logarithmic function f (x) = loga x has domain (0, +∞) and range R.

2. The logarithmic function f (x) = loga x is a continuous function.

3. The graph of the logarithmic function f (x) = loga x depends the value of a.
4. If a > 1, then lim loga x = +∞ and lim+ loga x = −∞; if 0 < a < 1, then
x→+∞ x→0
lim loga x = −∞ and lim+ loga x = +∞.
x→+∞ x→0

5. The following are the graphs of loga x for various bases.

log1.5 x
log2 x
ln x
log4 x
log10 x

x
O 1

log1/4 x
log1/2 x

log0.8 x
54 Summary

Definition: some inverse functions

• y = loga x ⇐⇒ x = ay ;
• y = sin−1 x (|x| ≤ 1) ⇐⇒ sin y = x, − π2 ≤ y ≤ 2;
π

• y = cos−1 x (|x| ≤ 1) ⇐⇒ cos y = x, 0 ≤ y ≤ π;

• y = tan−1 x (x ∈ R) ⇐⇒ tan y = x, − π2 < y < 2;
π

• y = cot−1 x (x ∈ R) ⇐⇒ cot y = x, y ∈ (0, π);

• y = csc−1 x (|x| ≥ 1) ⇐⇒ csc y = x, y ∈ [− π2 , 0) ∪ (0, π2 ];

• y = sec−1 x (|x| ≥ 1) ⇐⇒ sec y = x, y ∈ [0, π2 ) ∪ ( π2 , π].

Theorem: Intermediate Value Theorem

Let f be a continuous function over a closed interval [a, b], for any number ξ between f (a)
and f (b), there exists at least one c in the interval [a, b] such that f (c) = ξ.

y y

f (b) f (b)

ξ
ξ

f (a) f (a)

x x
O a c b O a c1 c2 c3 b
 CHAPTER

2
Differentiation

2.1 Derivatives

2.1 Example: instantaneous velocity

Let y = f (t) = t2 be the distant function describing a movement of a particle on the x-axis
and t is the time variable. The velocity of the particle is not a constant and depends uopn the
time t. So we call the velocity at time t the instantaneous velocity v(t). How to calculate
the instantaneous velocity? We know how to find the average velocity. So let’s calculate the
average velocity from time t = 1 to time t. The distance traveled by the particle within the
time is f (t) − f (1) = t2 − 12 . The time spent is t − 1. Thus the average velocity is

f (t) − f (1) t2 − 1
= = t + 1.
t−1 t−1
Clearly this average velocity doesn’t equal the instantaneous velocity at t = 1. However, if
we use the average velocity to approximate the instantaneous velocity at t = 1, the approx-
imation is better if t is closer to 1. Using the idea of limits, it is reasonable to say that the
instantaneous velocity v(1) is the limit of the average velocity as t approaches 1, i.e.,

f (t) − f (1)
v(1) = lim = lim (t + 1) = 2.
t→1 t−1 t→1

2.2 Example: slope of tangent line

Consider the function f (x) = 12 x2 and the tangent line of the graph of y = f (x) at x = 1.
We first consider the secant line of the graph at x = 1.
56 Differentiation

secant line
y = 12 x2

1
2
(1 + h)2 tangent line

1
2

x
O 1 1+h

The secant line passes through two points P0 = (1, 12 ) and P = (1 + h, 12 (1 + h)2 ). The
slope of the secant line is

2 (1 + h)2 − 21 h + 21 h2
1
S(h) = = = 1 + 21 h.
(1 + h) − 1 h

As the point P approaches P0 , i.e., as h approaches 0, the secant line approaches the
tangent line. In the language of limits, the tangent line is the limit of the secant line. Hence
the limit of the slope of the secant line is the slope S of the tangent line, i.e.

f (1 + h) − f (1)
S = lim S(h) = lim = lim (1 + 12 h) = 1.
h→0 h→0 h h→0

2.3 Remark: derivative - motivation

Given a function y = f (x), take a point a in the domain of the function. The change in
f (x) − f (a)
y from a to x is f (x) − f (a). The average rate of change is . The instantaneous
x−a
rate of change at a is the limit of average rate of change when x approaches a, i.e.,

f (x) − f (a)
lim .
x→a x−a
Since many important concepts such as velocity and slope of the tangent line of a function
can be described in this way, we will call such kind of limit the derivative of the function at
x = a.

2.4 Definition: derivative

Given a function f and a point a in the domain of f , the derivative of f at x = a is

f (x) − f (a)
lim ,
x→a x−a
provided the limit exists. In this case, we say f has a derivative at x = a.
For single variable functions, the term “being differentiable” is used as synonymous with
“having a derivative”.
2.1 Derivatives 57

2.5 Definition: notations of derivative

There are several widely used notations for derivatives.
• The derivative of f (x) at x = a is denoted by f 0 (a).
f (x) − f (a)
• The limit lim can be written as
x→a x−a
f (a + h) − f (a)
lim ,
h→0 h
where we replace x by a + h. When x varies, h varies so that x approaches a if and only if h
goes to zero.
f (x) − f (a)
• The limit lim can be written as
x→a x−a
f (a + ∆x) − f (a)
lim ,
∆x→0 ∆x
where ∆x = x − a is the another symbol representing the change in x.
• We can also calculate the derivative of f at any point of the domain. Thus we define the
derivative of the function at x:
f (x + ∆x) − f (x)
f 0 (x) = lim .
∆x→0 ∆x
• If we write the change in y as ∆y = f (x + ∆x) − f (x), then the derivative can also be
expressed as
∆y
f 0 (x) = lim .
∆x→0 ∆x

∆y
Hence the change in y is ∆y, the change in x is ∆x, the average rate of change is , and
∆x
the (instantaneous) rate of change of the function at x is the limit of the average rate of
change as ∆x approaches zero.
df (x)
• Leibniz notation: f 0 (x) can also be expressed as . If we write y = f (x), we can
dx
dy
also write for f 0 (x).
dx
df (x) 

• We sometimes use f 0 (x)|x=a , or to denote f 0 (a).
dx  x=a

2.6 Example: calculate derivatives by the definition

1. Find f 0 (x) for f (x) = x2 .

Solution By the definition,

f (x + h) − f (x) (x + h)2 − x2
f 0 (x) = lim = lim
h→0 h h→0 h
2hx + h2
= lim = lim (2x + h) = 2x.
h→0 h h→0
58 Differentiation

In the above calculation, the changing variable is h, while x should be treated as a

constant.
√
2. Find f 0 (x) for f (x) = x where x > 0.
Solution By the definition,

f (x + h) − f (x)
f 0 (x) = lim
h→0 h
√ √
x+h− x
= lim
h→0 h
√ √ √ √
( x + h − x)( x + h + x)
= lim √ √
h→0 h( x + h + x)
h
= lim √ √
h→0 h( x + h + x)
1 1
= lim √ √ = √ .
h→0 ( x + h + x) 2 x

In the above calculation, again the changing

√ variable
√ is h, while x should be treated as
a constant. When we calculate lim h + x = x, we used the fact that the function
√ h→0
h + x is a continuous function as a function of h.
Also from the calculation above, we see that f 0 (0) does not exist.

2.7 MATLAB: derivative

In MATLAB, the derivative of a function can be calculated using command diff. More-
over, using the symbolic substitution command subs, we can obtain the value of the deriva-
tive at a certain point. The MATLAB code below shows the calculations of f 0 (x), where
f (x) = x2 , and the value of f 0 (3).

>> syms x ;
>> f = x ^2;
>> d = diff ( f )

d =

2* x

>> subs (d ,x ,3)

ans =

6
2.1 Derivatives 59

2.8 Example: tangent line

Find the equation of the tangent line of the graph of y = x2 at x = 1.

Solution By using the definition, we can y

calculate the derivative f 0 (1) = 2, which is
y = x2
the slope of the tangent line. Since the tan-
gent lien passes through the point (1, f (1)) = y − 1 = 2(x − 1)
(1, 1), the equation of the tangent line is 1

y − 1 = 2 · (x − 1).
x
O 1

2.9 Example: velocity

If we toss a ball straightly up into the air with initial velocity 98 m/s, the distance (mea-
sured in meters) of the ball to the ground is given by the function f (t) = −4.9t2 + 98t where
t is the time measured in seconds. Find the velocity of the ball at time t and find the time
and the distance travelled when the ball reaches the highest height.

Solution Since the velocity v(t) is the derivative of the distance function, we obtain

[−4.9(h + t)2 + 98(h + t)] − (−4.9t2 + 98t)

v(t) = f 0 (t) = lim
h→0 h
−4.9(2ht + h2 ) + 98h
= lim
h→0 h
= lim [−4.9(2t + h) + 98] = −9.8t + 98.
h→0

The ball reaches the highest height when the velocity is zero. So let v(t) = 0, we get
−9.8t + 98 = 0 and t = 10. So in ten seconds, the ball reaches the highest height. The
distance travelled during the ten seconds is f (10) = −490 + 980 = 490 meters.

2.10 Engineering Example: electric current

An electric current is the flow of electric charge. Electric currents can cause Joule heating
which creates light in the incandescent light bulbs. Electric currents can also cause magnetic
fields which are used in motor, inductors and generators.
In mathematics, the electric current can be defined as the rate at which the electric charge
flows through a given surface, which is given as follows:
dQ
I= ,
dt
where I represents the electric current, Q represents the electric charge, and t represents the
time.
60 Differentiation

2.11 Engineering Example: angular velocity

The angular velocity is a basic concept in physics. It is the rate at which a particle rotates
around a center point. In other words, the angular velocity represents how fast an object
goes around something. For example, the angular velocity of Earth represents how fast the
Earth obits the Sun.

Mathematically, the angular velocity is de-

fined as the rate of a particle’s angular dis- ω=
dφ
dt
placement change relative to the origin and φ
can be expressed as follows:
O

dφ
ω= ,
dt

where ω represents the angular velocity, φ represents the angular displacement, and t repre-
sents the time. The angular velocity can be depicted in the figure.
Suppose a person is jogging on a circular track with angular displacement φ(t) = 20 1
t
(radians). Calculate this person’s angular velocity.
According to the definition of angular velocity, we have

dφ(t) d( 1 t)
ω= = 20 = 1
(radians/second).
dt dt 20

2.12 Engineering Example: acceleration

Acceleration widely exists in our life. For example, when a car speeds up, the passengers
might experience a force of pushing them back to seats. This feeling occurs because they are
accelerating. In layman’s words, acceleration means any process where the velocity (either
the magnitude or the direction) changes. In particular, the magnitude of the velocity is called
speed. Mathematically, the magnitude of the acceleration is defined as the rate of the speed
change of an object with respect to time, and given as follows:
dv
a= ,
dt
where a represents the acceleration, v represents the speed, and t represents the time.
Suppose a person is running on the track. At the starting stage, the speed of this person
is v(t) = t (meter/second). Calculate the acceleration of this person.
According to the definition of acceleration, we have

dv(t) dt 2
a= = = 1 (meter/second ).
dt dt
2.2 Rules of Differentiation 61

Relationship Between Differentiability and Continuity

2.13 Theorem: differentiability and continuity

If f is differentiable at a, then f is continuous at a. The converse is false; that is, there are
functions that are continuous but not differentiable.
Pitfall: continuous but not differentiable

Indeed, consider the function f (x) = |x|. It is continuous everywhere. However,

since
f (h) − f (0) |h| − 0 h−0
lim = lim+ = lim+ = 1,
h→0+ h h→0 h h→0 h
f (h) − f (0) |h| − 0 −h − 0
lim− = lim− = lim− = −1,
h→0 h h→0 h h→0 h
f (h) − f (0)
we know that lim does not exist. So, f is not differentiable at x = 0.
h→0 h

Justification:

f (x) − f (a)
If f is differentiable at a, then lim = f 0 (a) exists. Thus,
x→a x−a
f (x) − f (a)
lim [f (x) − f (a)] = lim · (x − a)
x→a x→a x−a
f (x) − f (a)
= lim · lim (x − a) = f 0 (a) · 0 = 0.
x→a x−a x→a

Hence lim f (x) = f (a), so that f is continuous at a.

x→a

2.2 Rules of Differentiation

2.14 Theorem: rules of differentiation

Suppose f 0 (x) and g 0 (x) exist. Then the following rules of differentiation hold.

1. Sum Rule:
0
[f (x) + g(x)] = f 0 (x) + g 0 (x).

2. Difference Rule:
0
[f (x) − g(x)] = f 0 (x) − g 0 (x).

3. Product Rule:
0
[f (x) g(x)] = f (x) g 0 (x) + g(x) f 0 (x).
62 Differentiation

4. Quotient Rule: If g(a) 6= 0, then

0
f (x) g(x) f 0 (x) − f (x) g 0 (x)

= 2 .
g(x) [g(x)]

Justification:
By the definition, for the sum and difference rules, we have
0
[f (x) ± g(x)]
[f (x + h) ± g(x + h)] − [f (x) ± g(x)]
= lim
h→0 h
[f (x + h) − f (x)] ± [g(x + h) − g(x)]
= lim
h→0 h
f (x + h) − f (x) g(x + h) − g(x)
 
= lim ±
h→0 h h
f (x + h) − f (x) g(x + h) − g(x)
= lim ± lim
h→0 h h→0 h
= f 0 (x) ± g 0 (x).

For the product rule, we may re-write the derivative as follows:

0
[f (x)g(x)]
f (x + h)g(x + h) − f (x)g(x)
= lim
h→0 h
f (x + h)g(x + h) − f (x + h)g(x) f (x + h)g(x) − f (x)g(x)
 
= lim +
h→0 h h
g(x + h) − g(x) f (x + h) − f (x)
 
= lim f (x + h) · + · g(x) .
h→0 h h

If f is differentiable, f is continuous, so that

lim f (x + h) = f (x).
h→0

Hence,
0
[f (x)g(x)]
g(x + h) − g(x) f (x + h) − f (x)
= lim f (x + h) · lim + lim · g(x)
h→0 h→0 h h→0 h
= f (x) g 0 (x) + g(x) f 0 (x).

The quotient rule can be shown similarly.

2.2 Rules of Differentiation 63

2.15 Example: derivatives of polynomials

Show that

0
an xn + an−1 xn−1 + · · · + a1 x + a0 = nan xn−1 + · · · + a1 .

Solution We prove the equality in following steps.

1. If f (x) ≡ c, then f 0 (x) = 0.

In fact, by the definition,
c−c
f 0 (x) = lim = 0.
h→0 h

2. If f (x) = x, then f 0 (x) = 1.

In fact, by the definition,

(x + h) − x h
f 0 (x) = lim = lim = 1.
h→0 h h→0 h

3. If f (x) = xn for an positive integer n, then f 0 (x) = nxn−1 .

This can be proved by induction on n.
0
First, we know that the formula (xn ) = nxn−1 holds for n = 1.

0
Suppose the formula holds for n = k, i.e., xk = kxk−1 . Then, for n = k + 1, from
the product rule,

0
0
f 0 (x) = xk+1 = x · xk

0
= x0 · xk + x · xk
= xk + kxk = (k + 1)xk .

0
Hence, for any positive integer n, (xn ) = nxn−1 .
4. If f (x) = cg(x), then f 0 (x) = cg 0 (x).
In fact, applying the product rule gives
0
(cg) = c0 · g + c · g 0 = 0 + c · g 0 = cg 0 .

5. Let f (x) = an xn + · · · + a1 x + a0 be a polynomial. Then

f 0 (x) = nan xn−1 + · · · + a1 .

In fact,
0 0 0
f 0 (x) = (an xn ) + · · · + (a1 x) + (a0 ) sum rule
= nan x n−1
+ · · · + a1 + 0 items 1, 3, 4
= nan x n−1
+ · · · + a1 .
64 Differentiation

2.16 Example: reciprocal rule

Suppose f is differentiable. By applying the quotient rule, we get the following reciprocal
rule: 0
1 (1)0 · f (x) − 1 · f 0 (x) f 0 (x)

= = − .
f (x) f (x)2 f (x)2

2.17 Example: horizontal tangent line

Find points where the tangent lines of y = f (x) = 2x3 − 3x2 − 12x + 1 are horizontal.

Solution A line is horizontal if and only if the slope of the line is zero. Since the slope of
the tangent line is given by f 0 (x), we just need to find points where f 0 (x) = 0.

Since y

0
2x3 − 3x2 − 12x + 1
= 2 · 3x2 − 3 · 2x − 12 8

x
= 6(x2 − x − 2) −1 O 2

= 6(x − 2)(x + 1), y = 2x3 − 3x2 − 12x + 1

we know that at x = 2 and x = −1, the tan- −19

gent lines are horizontal.

2.18 MATLAB: rules of differentiation

With diff, we can easily calculate the derivatives of f (x) + g(x), f (x) − g(x), f (x) g(x)
f (x) f (x)
and . Following examples show the derivatives of f (x) g(x) and in MATLAB.
g(x) g(x)

>> syms f ( x ) g ( x ) ;
>> diff ( f ( x ) * g ( x ) ,x )

ans =

f ( x ) * diff ( g ( x ) , x ) + g ( x ) * diff ( f ( x ) , x )

>> diff ( f ( x ) / g ( x ) ,x )

ans =

diff ( f ( x ) , x ) / g ( x ) - ( f ( x ) * diff ( g ( x ) , x ) ) / g ( x ) ^2
2.2 Rules of Differentiation 65

Differentiation of Trigonometric Functions

2.19 Theorem: derivatives of trigonometric functions

1. (sin x)0 = cos x; 4. (csc x)0 = − csc x cot x;

2. (cos x)0 = − sin x; 5. (sec x)0 = sec x tan x;

3. (tan x)0 = sec2 x; 6. (cot x)0 = − csc2 x.

Justification:
The derivatives of all six trigonometric functions can be found as follows.

1. By using some trigonometric identities (addition and double angle identities), continuity
sin h
of sin x, and the limit lim = 1, we have
h→0 h

sin(x + h) − sin x
(sin x)0 = lim
h→0 h
sin x cos h + cos x sin h − sin x
= lim
h→0 h
sin x(cos h − 1) + sin h cos x
= lim
h→0 h
−2 sin2 (h/2) sin h
 
= lim sin x · + cos x ·
h→0 h h
sin(h/2) sin h
 
= sin x · lim − sin(h/2) · + cos x · lim
h→0 h/2 h→0 h

= 0 + cos x = cos x.

2. By using some trigonometric identities (addition and double angle identities), continuity
sin h
of sin x, and the limit lim = 1, we have
h→0 h

cos(x + h) − cos x
(cos x)0 = lim
h→0 h
cos x cos h − sin x sin h − cos x
= lim
h→0 h
cos x(cos h − 1) − sin h sin x
= lim
h→0 h
−2 sin2 (h/2) sin h
 
= lim cos x · − sin x ·
h→0 h h
sin(h/2) sin h
 
= cos x · lim − sin(h/2) · − sin x · lim
h→0 h/2 h→0 h

= 0 − sin x = − sin x.
66 Differentiation

3. By applying the quotient rule, we have

0
sin x (sin x)0 cos x − sin x(cos x)0

(tan x) =
0
=
cos x cos2 x
cos x cos x − sin x(− sin x) cos2 x + sin2 x
= = = sec2 x.
cos2 x cos2 x

4. By applying the quotient rule, we have

0
1 (1)0 sin x − (sin x)0

(csc x)0 = =
sin x sin2 x
− cos x
= = − cot x csc x.
sin2 x

5. By applying the quotient rule, we have

0
1 (1)0 cos x − (cos x)0

(sec x)0 = =
cos x cos2 x
sin x
= = tan x sec x.
cos2 x

6. By applying the quotient rule, we have

 cos x 0 (cos x)0 sin x − cos x(sin x)0
(cot x)0 = =
sin x sin2 x
− sin x sin x − cos x cos x
=
sin2 x
− cos x − sin2 x
2
= = − csc2 x.
sin2 x

2.20 MATLAB: derivatives of trigonometric functions

We can calculate the derivatives of trigonometric functions easily using command diff.
Moreover, with the help of the variable precision arithmetic command vpa and the symbolic
substitution command subs, we can calculate the value of the derivative of a trigonometric
function at a certain point. The following example shows the calculations of f 0 (x), where
f (x) = sin(x), and f 0 (1).

>> syms x ;
>> f = sin ( x ) ;
>> d = diff ( f )

d =

cos ( x )
2.2 Rules of Differentiation 67

>> v1 = subs (d ,x ,1)

v1 =

cos (1)

>> vpa ( v1 )

ans =

0.54030230586813971740093660744298

2.21 Example: tangent line - trigonometric functions

Find the equation of horizontal tangent line(s) of y = f (x) = sin x + cos x in (0, π).

Solution Horizontal tangent lines have zero slope. The slope of the tangent line is given
by the derivative of f . Thus we calculate

f 0 (x) = (sin x + cos x)0 = cos x − sin x = 0 =⇒ tan x = 1.

Solving the equation gives √ x = π/4. Since

y
f ( π4 ) = sin π4 + cos π4 = 2, we get the equa- √
tion of the horizontal tangent line y= 2
1
√ y = sin x + cos x
y = 2.
x
O 1 3 π
4π 4π

−1

2.22 Engineering Example: electric current - capacitor

Consider a circuit with a capacitor below, where C is the capacitance of the capacitor, i(t)
is the current through the capacitor, and vC (t) is the voltage across the capacitor.
vC (t)
− +

i(t)
C

Given that vC (t) = 4 cos t, where t ≥ 0, calculate the electric current i(t) where i(t) =
C dvdt
C (t)
.
First, we calculate dvdt
C (t)
, and we have

dvC (t)
= −4 sin t.
dt
68 Differentiation

Therefore, the current i(t) is given by

i(t) = −4C sin t

where “-” specifies the direction of the current.

Chain Rule
2.23 Theorem: chain rule

If g is differentiable at a and f is differentiable at g(a), then the composite function F ,

defined by F (x) = f (g(x)), is differentiable at a and the derivative F 0 (a) is given by the
product:
F 0 (a) = f 0 (g(a)) · g 0 (a).
Or
dy dy du
= · .
dx du dx
This is the so-called chain rule.

Justification:

Suppose u = g(x) is differentiable at a and y = f (u) is differentiable at b = g(a). If ∆x is

an increment in x and ∆u and ∆y are the corresponding increments in u and y, then, from
the hypotheses,
∆u ∆y
lim = g 0 (a), lim = f 0 (b).
∆x→0 ∆x ∆u→0 ∆u

Thus,

∆u = g 0 (a)∆x + ε1 ∆x, where ε1 → 0 as ∆x → 0;

∆y = f (b)∆u + ε2 ∆u,
0
where ε2 → 0 as ∆u → 0.

Hence,
∆y = [f 0 (b) + ε2 ]∆u = [f 0 (b) + ε2 ][g 0 (a) + ε1 ]∆x,
so that
∆y
= [f 0 (b) + ε2 ][g 0 (a) + ε1 ].
∆x
Since g is differentiable at a, it must be continuous at a and thus lim g(a + ∆x) = g(a) so
∆x→0
that ∆u goes to zero as ∆x goes to zero. This gives that both ε1 → 0 and ε2 → 0 as ∆x → 0.
Therefore,
∆y
F 0 (a) = lim = lim [f 0 (b) + ε2 ][g 0 (a) + ε1 ]
∆x→0 ∆x ∆x→0
= f 0 (b)g 0 (a) = f 0 (g(a))g 0 (a).

This proves the chain rule.

2.2 Rules of Differentiation 69

2.24 Example: chain rule

1. Compute the derivative of h(x) = sin2 x.

Solution Write y = f (u) = u2 and u = g(x) = sin x. Then h(x) = f (g(x)). Thus, by
the chain rule,

dy 

h (x) =
0
· g 0 (x) = 2u|u=g(x) · cos x = 2 sin x cos x.
du u=g(x)

2. Compute the derivative of h(x) = sin(x2 ).

Solution Write y = f (u) = sin u and u = g(x) = x2 . Then h(x) = f (g(x)). Thus, by
the chain rule,

dy 

h0 (x) = · g 0 (x) = cos u|u=g(x) · 2x = 2x cos(x2 ).
du u=g(x)

2.25 MATLAB: chain rule

The following example shows the calculation of F 0 (x), where F (x) = f (g(x)), in MATLAB.
It verifies that the Chain Rule holds in general.

>> syms f ( x ) g ( x ) ;
>> F ( x ) = compose ( f ( x ) ,g ( x ) )

F(x) =

f(g(x))

>> diff ( F ( x ) )

ans =

D ( f ) ( g ( x ) ) * diff ( g ( x ) , x )

2.26 Engineering Example: maximum wind speed

Consider the maximum wind speed at the site of the building. Denote the maximum
wind speed as V (the wind direction is ignored for simplicity) and the probability that this
maximum wind speed is no larger than v as FV (v). Empirical data indicate that, in many
locations, FV (v) has the following form:

FV (v) = e−( u )
v −k

where u and k are positive parameters.

70 Differentiation

Specifically, in Boston, plausible values of u and k are 49.4 miles per hour (mph) and 6.5,
respectively. Hence, in Boston, FV (v) can be written as

FV (v) = e−( 49.4 )

v −6.5

where the unit of the maximum wind speed V is mph.

Calculate F0V (v) by the chain rule.

−6.5
Let f (v) = ev and g(v) = − 49.4
v
. Then, FV (v) can be rewritten as FV (v) = f (g(v)).
By the chain rule, we have

F0V (v) = f 0 (g(v)) g 0 (v)

0
= eg(v) (−49.46.5 ) v −6.5
= ae−( 49.4 )
v −6.5
v −7.5 ,

where a = 6.5 × 49.46.5 ≈ 6.6395 × 1011 . Note that in probability theory and statistics, FV (v)
is called the cumulative distribution function of V and F0V (v) is called the probability density
function of V , which are both important functions to determine the statistical characteristics
of V .

2.27 Engineering Example: electric current in RC circuit

Consider a resistor-capacitor (RC) circuit below, where R is the resistance of the resistor,
C is the capacitance of the capacitor and vC (t) is the voltage across the capacitor.

R
i(t)
−
−
vI (t) C vC (t)
+
+

t dvC (t)
Given that vC (t) = 1−e− RC (t ≥ 0), calculate the electric current i(t) where i(t) = C .
dt
dvC (t) dvC (t)
First, we calculate . Let y = 1 − eu and u = − RC
t
. Then, can be written as
dt dt
dy
. By the chain rule, we have
dt
dy dy du 1 1 − t
 
= · = −e −u
= e RC .
dt du dt RC RC

dvC (t)
Substituting the above result into i(t) = C , we can obtain the electric current i(t), as
dt
shown in the following:
1 − t
i(t) = e RC .
R
2.2 Rules of Differentiation 71

2.28 Engineering Example: reliability function

In reliability engineering, which is an interdisciplinary field of engineering and engineering

management, the reliability function gives the probability that a component will survive
beyond any given specified time. Suppose the reliability function of a component is given by
√
R(t) = e− t
(t ≥ 0).

d (1 − R(t))
Calculate .
dt
d (1 − R(t))
First, given the expression of R(t), can be written as
dt
 √ 
d (1 − R(t)) d 1 − e− t
=
dt dt
√ d (1 − R(t))
Let f (t) = 1 − et and g(t) = − t. Then, can be rewritten as
dt
d (1 − R(t)) df (g(t))
= .
dt dt
By the chain rule, we have

df (g(t)) 1 1 √
 
= f 0 (g(t)) g 0 (t) = −eg(t) − √ = √ e− t .
dt 2 t 2 t

d (1 − R(t))
Note that R(t) and are the complementary cumulative distribution function
dt
and the probability density function of the lifetime of the component, respectively. Besides
the engineering field, the reliability function can also be used for the study in other fields,
e.g., the duration of marriage, the death of a patient, etc. In these cases, the reliability
function is commonly known as the survival function.

Differentiation of Inverse Functions

2.29 Theorem: differentiation of inverse function

If f is differentiable and has an inverse function f −1 over an interval I, then f −1 is also

differentiable on I whose derivative equals

df −1 (x) 1 1
= 0 −1 = .
dx f (f (x)) df (u) 


du u=f −1 (x)
72 Differentiation

For short, the formula can be re-written as

dx 1
= .
dy dy
dx

Justification:
We can apply the chain rule to prove the formula for differentiation of inverse functions.
If applying the differentiation with respect to x to the identity

f (f −1 (x)) = x,

we obtain
df (f −1 (x)) df (u)  df −1 (x)

= 1 =⇒ · =1
dx du u=f −1 (x) dx
df −1 (x) 1
=⇒ = .
dx df (u) 


du u=f −1 (x)

√
2.30 Example: derivative of n
x
√
Let f (x) = xn with n a positive integer, then f −1 (x) = n x. We write f (u) = un . Thus
df (u) df (u) 

√
= nun−1 . Thus we get = n( n x)n−1 , and
du du u=f −1 (x)
 1
0 √ 1 1 1
xn = ( n x)0 = √ = x n −1 .
n( n x)n−1 n

2.31 Example: derivative of xa

Consider the function xa where a is a rational number. If a is positive, we can write

a = n/m, where n and m are positive integers, so that
√
n
xa = m x .
√
Write g(x) = m x and f (u) = un . Then xa = f (g(x)). Applying the chain rule, we get

dun  √
0 √
n−1 1 1 −1

0
(xa ) = · m
x = n m
x · xm
du u= m√x
 m
n n−1 + 1 −1 n n −1
= x m m = x m = axa−1 .
m m
If a is negative, we can write xa = x−b , where b = −a is a positive rational number. Then
2.2 Rules of Differentiation 73

1
xa = . Applying the quotient rule gives
xb

0
a 0 xb bxb−1
(x ) = − 2 = − = −bx−b−1 = axa−1 .
(xb ) x2b

In summary, for any rational number a, we have

0
(xa ) = axa−1 .

2.32 Theorem: derivatives of inverse trigonometric functions

1 1
1. (sin−1 x)0 = √ ; 4. (csc−1 x)0 = − √ ;
1 − x2 |x| x2 − 1
1 1
2. (cos−1 x)0 = − √ ; 5. (sec−1 x)0 = √ ;
1 − x2 |x| x2 − 1
1 1
3. (tan−1 x)0 = ; 6. (cot−1 x)0 = − .
1 + x2 1 + x2

Justification:
We can use the formula of derivative of inverse functions to calculate the derivatives of the
inverse trigonometric functions. For example,

0 1 1
sin−1 x = 0 = .
(sin u) |u=sin−1 (x) cos(sin−1 x)
The last expression can be simplified as fol-
lows. Let θ = sin−1 x, then sin θ = x. Draw
a right triangle with one angle |θ| whose op-
posite side has length |x| and hypotenuse has 1
|x|
length 1. Then the adjacent
√ √ side has length
1 − x2 . Thus, cos θ = 1 − x2 . |θ|
√
1 − x2

Hence,

0 1 1
sin−1 x = =√ .
cos θ 1 − x2
The other five derivatives can be calculated similarly.

2.33 Engineering Example: roof truss design

There is a symmetrical roof truss shown below, where the span AB is 16 m.

74 Differentiation

θ
A B

Now, we need to design the rise r. Find the rate at which the angle θ between the rafter AC
and the span AB changes with the rise r, i.e., find dθ
dr .
First, let the length of the span AB be lAB . We note that tan θ = 1 lrAB = 8r . So, we have
2

r
θ = tan−1 .
8
Let u = 8r , f (x) = tan x, then θ = f −1 (u). According to the chain rule, we have

dθ df −1 (u)
=
dr dr
df −1 (u) du
= .
du dr
Moreover, according to the theorem for the differentiation of inverse function, we can calculate

df −1 (u) 1
=
du

df (x) 
dx 
x=f −1 (u)
1
=
sec2 tan−1 (u)

cos2 tan−1 (u)

=
sin2 tan−1 (u) + cos2 tan−1 (u)

1
=
1 + tan tan−1 (u)
2

1
= .
1 + u2
Therefore, we can get
dθ 1 du
=
dr 1 + u2 dr
1 1
=
r 2 8
1+


8
8
= 2 .
r + 64
2.3 Differentiation of Exponential and Logarithmic Functions 75

2.3 Differentiation of Exponential and Logarithmic Functions

2.34 Theorem: derivatives of exponential functions

(ex )0 = ex .
In general, for any a > 0,
(ax )0 = ax ln a.

Justification:
x
1

Since e = lim 1+ , by making a change of variable, we get
x→+∞ x

e = lim+ (1 + y)1/y .
y→0

From the continuity of the logarithmic function, we have

ln(1 + y)
lim = lim+ ln(1 + y)1/y = ln e = 1.
y→0+ y y→0

Denote y = eh − 1. It is easy to see that y → 0+ is equivalent to h → 0+ . Thus, we get

h ln(1 + y)
lim = lim = 1.
h→0+ eh − 1 y→0+ y

We further get
h −h heh
lim− = lim = lim = 1.
h→0 eh − 1 h→0+ e−h − 1 h→0+ eh − 1
h
So, we know that lim = 1. Hence,
h→0 eh − 1

ex+h − ex eh − 1
(ex )0 = lim = ex · lim = ex .
h→0 h h→0 h
In general,

ax+h − ax ah − 1
(ax )0 = lim = ax · lim
h→0 h h→0 h
e h ln a
−1
= ax ln a · lim = ax ln a.
h→0 h ln a

2.35 Engineering Example: atmospheric pressure

The atmospheric pressure on earth changes with the altitude of the surface. The weather
conditions, such as the temperature and the humidity, also affect the atmospheric pressure,
so it is necessary to know these to calculate an accurate atmospheric pressure at a certain lo-
76 Differentiation

cation. However, for general conditions, the National Aeronautics and Space Administration
(NASA) has averaged the conditions for all parts of the earth year-round, and it is shown
that the atmospheric pressure decreases as the altitude increases.
Specifically, at low altitudes above sea level (e.g., the first 1000 meters above sea level),
the atmospheric pressure decreases by approximately 11.3 Pa per meter. For higher altitudes
where the standard temperature lapse rate (i.e., the rate at which the temperature changes
with the altitude) is zero, the atmospheric pressure is shown in the following equation:
gM
− RT (h−hb )
P (h) = Pb e b

where
Parameter Explanation
Pb static pressure of layer b (Pa) 
2
g gravitational acceleration : 9.80665 m/s
M molar mass of Earth’s air : 0.0289644 (kg/mol)
R universal gas constant : 8.3144598 (J/mol/K)
Tb standard temperature of layer b (K)
hb height at bottom of layer b (m)
with b ∈ [0, 6] (b ∈ Z) in accordance with each of seven successive layers of the atmosphere.
dP (h)
Calculate .
dh
In P (h), we treat h as the independent variable and all the other parameters as constants.
Then, by the chain rule, we have
0
dP (h)

gM
− RT (h−hb ) gM
= Pb e b − (h − hb )
dh RTb
gM − gM (h−hb )
=− Pb e RTb .
RTb
Note that the result above represents the rate at which the atmospheric pressure changes
with the altitude.
For example, in the stratosphere (i.e., b = 1), the standard temperature lapse rate is 0. We
dP (h)
thus can use the above expressions of P (h) and to calculate the atmospheric pressure
dh
and its changing rate, respectively. In particular, when b = 1, we have

Parameter Value
P1 22632.10 (Pa)
T1 216.65 (K)
h1 11000 (m)
2.3 Differentiation of Exponential and Logarithmic Functions 77

dP (h)
Therefore, P (h) and in the stratosphere can be calculated as follows:
dh
9.80665×0.0289644
P (h) = 22632.10e− 8.3144598×216.65 (h−11000)
−4
≈ 2.2 × 104 e−1.6×10 (h−11000)
(Pa),
dP (h) 9.80665 × 0.0289644 9.80665×0.0289644
=− × 22632.10 × e− 8.3144598×216.65 (h−11000)
dh 8.3144598 × 216.65
−4
≈ −3.6e−1.6×10 (h−11000)
(Pa/m).

dP (h)
Note that the sign “−” in means that the atmospheric pressure decreases with the
dh
altitude.

2.36 Theorem: derivatives of logarithmic functions

1
(ln x)0 = .
x
In general, for any a > 0, a 6= 1,
1
(loga x)0 = .
x ln a

Justification:

Let f (u) = au and u = loga x. Then the equality aloga x = x can be re-written as au = x.
Applying the chain rule, we get
du
au ln a · = 1.
dx
Hence,
du 1 1
(loga x)0 = = u = .
dx a ln a x ln a

2.37 Example: derivative of ln |x|

1
Show that (ln |x|)0 = for any x 6= 0.
x
78 Differentiation

Solution In fact, we know that for x > 0, y

1
(ln |x|)0 = (ln x)0 = .
x 1
y = ln |x| y=
If x < 0, then, by the chain rule, 1
x

1 1 x
−1 O 1 2
(ln |x|)0 = [ln(−x)]0 = · (−1) = .
−x x

2.38 Example: derivative of composite function

Calculate the derivative of ln(ax + 1).

Solution By applying the chain rule, we have

0 1 ax ln a
[ln(ax + 1)] = · (ax + 1)0 = x .
ax +1 a +1

2.39 Engineering Example: entropy - information theory

First introduced by Claude Shannon in 1948, in information theory, the entropy is used
to quantitively measure the information which is produced by a stochastic source of data.
Specifically, when a low-probability event occurs at the source, then the event carries more
information than when a high-probability event occurs. The amount of information conveyed
by each event becomes a random variable whose expected value is the entropy. Hence, the
low-probability event has more entropy. In general, the more uncertain a event is, the more
information the event carries and the more the entropy of this event is. In other words, the
entropy refers to uncertainty or disorder.
For a communication system, if it has two inputs 0 and 1 and they occur with probability
p and 1 − p (0 < p < 1), respectively, then the entropy of the source is defined as
H(p) = −p log2 p − (1 − p) log2 (1 − p).

Calculate dH(p)
dp in this communication system.
By applying the product rule and the chain rule, we have
dH(p) 0 0
= − log2 p − p (log2 p) − (−1) log2 (1 − p) − (1 − p) (log2 (1 − p))
dp
1 −1
= − log2 p − p + log2 (1 − p) − (1 − p)
p ln 2 (1 − p) ln 2
= log2 (1 − p) − log2 p.

Note that dH(p)

dp shows the changing rate of the entropy H(p) with respect to the probability
of the input p. Now, we investigate how H(p) changes with p.
2.3 Differentiation of Exponential and Logarithmic Functions 79

In particular, since log2 (1 − p) is a decreasing function of p and log2 p is an increasing

function of p, dH(p)
dp is thus a decreasing function of p. Moreover, we notice that when p is
dH(p) dH(p)
small, dp > 0; while when p gets bigger and approaches 1, dp < 0. We thus know that
dH(p)
there exists a point p ∈ (0, 1) such that H(p ) = 0. Furthermore, since
∗ ∗
dp > 0 means
dH(p)
H(p) increases with p and < 0 means H(p) decreases with p, the entropy H(p) is thus
dp
maximized at p∗ . This shows that when p is neither too small nor too big, i.e., the input is
more uncertain, more entropy can be achieved.

Actually, we can plot the figure of H(p) as

H(p)
follows:
Indeed, we see from the figure above that y = log2 (1 − p) − log2 p

H(p) increases with p up to a certain point, 1

denoted as p∗ , then it decreases with p. In

other words, more entropy is obtained when
p is neither too big nor too small, i.e., when p
the input is more uncertain. O p∗ 1

Logarithmic Differentiation
2.40 Method: logarithmic differentiation

1. Take natural logarithms of both sides of an equation y = f (x) and use the Laws of
Logarithms to simplify.
2. Differentiate implicitly with respect to x.

3. Solve the resulting equation for y 0 .

2.41 Example: logarithmic differentiation

The “logarithmic differentiation” method can often be used to calculate the derivatives for
exponential functions in the form [u(x)]v(x) , or products u1 (x)u2 (x) · · · uk (x) with relatively
larger k.

1. Calculate the derivative of xµ where x > 0 and µ is a real constant.

Solution Let f (x) = xµ . Taking logarithm of both sides gives

ln f (x) = ln xµ = µ ln x.
Then we can take derivative of both sides of the last equation and apply the chain rule
to obtain
1 0 1
f (x) = µ · = µx−1 .
f (x) x
80 Differentiation

Thus,
f 0 (x) = f (x) · µx−1 = µxµ−1 .

2. Calculate the derivative of xx where x > 0.

Solution Let f (x) = xx . Taking logarithm of both sides gives

ln f (x) = ln xx = x ln x.

Then we can take derivative of both sides of the last equation and apply the chain rule
to obtain
1 0 1
f (x) = ln x + x · = ln x + 1.
f (x) x
Thus,
f 0 (x) = f (x) · (ln x + 1) = xx (ln x + 1).

√
ex 1 − x2
3. Calculate the derivative of f (x) = .
1 + x + x2

Solution We can calculate this derivative directly using the product rule and the
quotient rule. Since there are many terms, it is quite involved. In the following, we
make use of logarithm for a simpler calculation. In fact, we have
√
ex 1 − x2
ln f (x) = ln
1 + x + x2
= x + 21 ln(1 − x2 ) − ln(1 + x + x2 ).

Taking derivative of both sides of the equation and applying the chain rule, we obtain
1 0 x 1 + 2x
f (x) = 1 − − ,
f (x) 1 − x2 1 + x + x2

so that √
ex 1 − x2 1 + 2x
 
x
f (x) =
0
1− − .
1 + x + x2 1 − x2 1 + x + x2
2.4 Implicit Differentiation 81

2.4 Implicit Differentiation

2.42 Example: implicit differentiation - circle

For the function y = f (x) defined by y 2 + x2 = 1, we can take the derivative of y with
respect to x of both sides of the equation. By treating y as a function of x and applying the
chain rule, we get
d x2 + y 2 dy


= 0 =⇒ 2x + 2y · = 0,
dx dx
so that
dy x
=− .
dx y
This function y = f (x) is defined implicitly. We keep in mind that y is a function of x.
dy
We can find the derivative by “implicit differentiation”.
dx

2.43 Method: implicit differentiation

1. Replace y by y(x) in the given equation which defines the function y = y(x) implicitly.

2. Differentiate both sides of the equation with respect to x, by applying the rules of
differentiation.
dy
3. Solve for from the resulting equation.
dx

2.44 Example: implicit differentiation

The equation 2y 4 − x2 − sin y + 1 = 0 (x > 0) gives a curve on the plane. Use implicit
dy
differentiation to find y 0 = and then determine the tangent line of the curve at the point
dx
P = (1, 0).
Solution By treating y as a function of x, we differentiate both sides of the equation
2y 4 − 2x2 − sin y + 1 = 0 with respect to x and get

8y 3 y 0 − 2x − y 0 cos y = 0.

Solving the equation for y 0 gives

2x
y0 = .
8y 3 − cos y
82 Differentiation

The point P = (1, 0) satisfies the equation y

and thus lies on the curve. The tangent line
2y 4 − x2 − sin y + 1 = 0
of the curve at the point has slope
1
2(1)
y 0 |P = = −2.
8(03 ) − cos 0
x
O 1 2
Hence the tangent line at P is given by the
equation y = −2(x − 1)
y = −2(x − 1).

2.45 MATLAB: implicit differentiation

MATLAB supports the implicit differentiation if the symbolic function y(x) is appropriately
created. The following example shows the calculation of y 0 in MATLAB, where x and y satisfy
the equation 2y 4 − x2 − sin y + 1 = 0.

>> syms x y ( x ) t
>> f = 2* y ( x ) ^4 - x ^2 - sin ( y ( x ) ) + 1;
>> d = diff (f , x )

d =

8* y ( x ) ^3* diff ( y ( x ) , x ) - 2* x - cos ( y ( x ) ) * diff ( y ( x ) , x )

>> d0 = subs (d , diff ( y ( x ) ,x ) ,t )

d0 =

8* t * y ( x ) ^3 - 2* x - t * cos ( y ( x ) )

>> s = solve ( d0 , t )

s =

-(2* x ) /( cos ( y ( x ) ) - 8* y ( x ) ^3)

where d0 = subs(d,diff(y(x),x),t) is to replace diff(y(x),x) with t, so that the com-

mand solve can work to obtain the expression of t (i.e., the expression of diff(y(x),x) or
y 0 ). Note that if we solve d = 0 with respect to diff(y(x),x) directly, i.e., using command
solve(d,diff(y(x),x)), MATLAB is unable to find the explicit solution.
Moreover, based on the expression of diff(y(x),x) or y 0 , which was calculated above, we
can easily determine the slope of the tangent line at point (1, 0).
2.5 Related Rates 83

>> s_x = subs (s ,x ,1) % substitute x = 1

s_x =

-2/( cos ( y (1) ) - 8* y (1) ^3)

>> s_xy = subs ( s_x ,y ,0) % substitute y = 0

s_xy =

-2

MATLAB shows that the slope of the tangent line at point (1, 0) is −2.

2.5 Related Rates

2.46 Example: application - related rates

John observes a rising helium balloon. He I ♥ HK

stands 100 meters away from the launch site.
The balloon rises vertically at the constant
speed 2 meter/second. How fast is the angle
of the elevation of the balloon increasing 50 θ(t)
seconds after the launch? (The angle of ele- 100
vation is the angle between the ground and
John’s line of sight to the balloon.)

Solution Let y(t) be the distance traveled by the balloon from the launch site at time t.
Let θ(t) be the angle of elevation at time t.
The speed of the balloon is y 0 (t), which equals to 2 m/s. “How fast is the angle of the
elevation” means that we need to find the rate of change of θ(t), i.e., the derivative θ0 (t).
The relation between y(t) and θ(t) is given by
y(t) = 100 · tan θ(t).
Taking the differentiation of both sides of the equation above, we obtain
100 · θ0 (t)
y 0 (t) = .
cos2 θ(t)
Since we need to find θ0 (t), we obtain
y 0 (t) 2
θ0 (t) = cos2 θ(t) = cos2 θ(t).
100 100
84 Differentiation

We need to find cos θ(t) at t = 50. In fact,

y(50) = 2 × 50 = 100.

From the relation between y(t) and θ(t), we have 100 = 100 · tan θ(50), so that θ(50) = 4.
π

Thus we obtain √
2
cos θ(50) = cos( 4 ) =
π
.
2
Therefore, we have
 √ 2
2 2
θ0 (50) = · = 0.01.
100 2
Since two rates of change, i.e., derivatives θ0 (t) and y 0 (t), are related by the equation

y 0 (t)
θ0 (t) = cos2 θ(t).
100
This is a typical problem of related rates. In general, so-called related rates problems
involve finding a rate at which a quantity changes by relating that quantity to other quantities
whose rates of change are known.

2.47 Example: related rates - area

An oil tank springs an oil leak on the ground, and the oil spreads in a circular patch around
the tank. If the radius of the oil patch increases at a rate of 10 meters/hour, how fast is the
area of the patch increasing when the patch has a radius of 20 meters?

Solution
Step 1: Label all variables:
Let r(t) be the radius of the oil patch at time t and A(t) be the area of the oil patch at
time t.
The rate of the area change is the quantity A0 (t).
Step 2: Write the relation between concerned variables:
The area of the patch is
A(t) = πr(t)2 .
Step 3: Write the relations between rates:
We take the differentiation with respect to t on both sides of the equation above,

A0 (t) = 2πr(t) · r0 (t).

Step 4: List all known quantities:

The radius of the oil patch increases at a rate of 10 meters/hour, that is,

r0 (t) = 10.

At the time t0 when the patch has a radius of 20 meters, we have

r(t0 ) = 20, r0 (t0 ) = 10.

2.5 Related Rates 85

Step 5: Find the desired rate:

Combining all the results above gives

A0 (t0 ) = 2πr(t0 ) · r0 (t0 ) = 2π · 20 · 10 = 400π ≈ 1256.64 (m2 /h).

2.48 Example: related rates - distance

Two sailing boats sail towards a docking P 30

port. Boat A sails due west at the con- A
stant speed 8 miles/hour and Boat B sails due
north at the constant speed 10 miles/hour.
40
How fast is the distance between the boats
changing when the westbound boat is 30
miles from the port and the northbound boat
is 40 miles from the port? B

Solution
Step 1: Label all variables:
Let x(t) be the distance of the westbound boat to the port at time t. Let y(t) be the
distance of the northbound boat to the port at time t. Let z(t) be the distance between two
boats at time t.
The rate of the distance change is the quantity z 0 (t).
Step 2: Write the relation between concerned variables:
From the Pythagorean theorem,
z(t)2 = x(t)2 + y(t)2 .
Step 3: Write the relations between the rates:
We take the differentiation with respect to t on both sides of the equation above,
x(t) · x0 (t) + y(t) · y 0 (t)
2z(t) · z 0 (t) = 2x(t) · x0 (t) + 2y(t) · y 0 (t) =⇒ z 0 (t) = .
z(t)
Step 4: List all given quantities:
At the time t0 when westbound boat is 30 miles from the port, we have
x(t0 ) = 30, y(t0 ) = 40, x0 (t0 ) = −8, y 0 (t0 ) = −10.
Notice that the derivatives x0 (t) and y 0 (t) are negative, since the distances from both boats
to the port decrease in time.
Step 5: Determine other quantities from the given quantities:
We can determine the quantity z(t0 ) by using the Pythagorean identity:
p
z(t0 ) = x(t0 )2 + y(t0 )2 = 302 + 402 = 50.
p

Step 6: Find the desired rate:

Combining all the results above gives
x(t0 ) · x0 (t0 ) + y(t0 ) · y 0 (t0 ) −8 · 30 − 10 · 40
z 0 (t0 ) = = = −12.8 (mi/h).
z(t0 ) 50
86 Differentiation
y
2.6 Linear Approximation
y = 21 x2
2.49 Example: linear approximation
tangent line
y

y = 12 x2

1 tangent line
2

1
2

x
O 1O 1
x

Geometrically, the tangent line of a graph y = f (x) at x = a is close to the graph when x
is very close to a. Algebraically, the value f (a) + f 0 (a)(x − a) is very close to f (x) when x is
close to a, that is, when x ≈ a,

f (x) ≈ f (a) + f 0 (a)(x − a).

2.50 Theorem: linear approximation

Let f be differentiable at a. Then the linear function y = f (a) + f 0 (a)(x − a) approximates

y = f (x) near x = a, that is, when x ≈ a,

f (x) ≈ f (a) + f 0 (a)(x − a).

This linear function y = f (a) + f 0 (a)(x − a) is called the linear approximation (or
linearization) of f at x = a.

2.51 Example: linear approximation - numerical approximation

√ 1
Consider the function f (x) = 1 + x. Its derivative is f 0 (x) =
√ . The tangent
2 1+x
line at x = 0 is y = f (0) + f (0)x = 1 + 2 x. We know that, for small x, y = 1 + 12 x
0 1
√
approximates 1 + x. Numerically, we can verify this. For different values of small x, we
make the following table
2.6 Linear Approximation 87

√
x 1+x 1 + 21 x

0.1 1.04880884817 1.05

0.01 1.00498756211 1.005
0.001 1.00049987506 1.0005

We see that the approximation is getting better when x is closer to 0.

2.52 Example: linear approximation - numerical approximation

√
Find approximate value of 3
1.02.

√ 1
Solution Consider the function f (x) = 3
1 + x. Since f 0 (x) = , the linear
3 3 (1 + x)2
p
approximation of f is
y = 1 + 13 x.
√
Thus, we have an approximate value of 3 1.02:
√
1.02 ≈ 1 + 31 (0.02) ≈ 1.006667.
3

√
The real value of 3
1.02 is 1.0066227 · · · .

2.53 MATLAB: linear approximation

In MATLAB, the linear approximation of a function can be calculated using the command
taylor.
√ The following examples show the calculations of the linear approximations of f (x) =
3
1 + x at x = 0 and f (x) = cos x at x = a, respectively.

>> syms x ;
>> f = (1+ x ) ^(1/3) ;
>> taylor (f ,x ,0 , ' Order ' ,2)

ans =

x /3 + 1
>> syms x a ;
>> f = cos ( x ) ;
>> taylor (f ,x ,a , ' Order ' ,2)

ans =

cos ( a ) + sin ( a ) *( a - x )

Because the linear approximation f (x) ≈ f (a) + f 0 (a)(x − a) has two terms, we specify
88 Differentiation

‘Order’ and 2 in the command taylor in the MATLAB code above.

2.54 Engineering Example: small signal analysis

In electronic circuits, many of the components that make up of the circuits are nonlinear,
e.g., diodes, transistors and vacuum tubes. For these nonlinear components, the current
through them is not proportional to the voltage, which makes calculations and analysis
complicated. Moreover, there are many applications where nonlinear devices only operate
in a very restricted range of voltage or current, such as many sensor applications and audio
amplifiers. For easy calculations and analysis of the voltage and current, a method called
small signal analysis (or incremental analysis) is proposed to approximate circuits containing
nonlinear elements with linear equations over a restricted region. Specifically, if the time-
varying alternating current (AC) signal of an electronic current is small compared to the
deterministic direct current (DC) bias, then using linear approximation, the nonlinear
circuit elements are replaced by linear elements over a restricted range in the resultant AC-
equivalent circuit. The small signal analysis method widely applies to electronic circuits
where the AC signals are small compared to the DC ones, e.g., radio receivers, sensors,
telecommunications and signal processing circuits.
A nonlinear circuit example is shown in the following, where the diode is a nonlinear
component.
Parameter Description
iD
VD DC voltage
ID DC current
−
− vI vD vd AC voltage
+
+ id AC current
vD = VD + vd total voltage
iD = ID + id total current
vI input voltage

For the total voltage vD and the total current iD of the above circuit, we have iD = f (vD ).
Note that iD is a nonlinear function of vD due to the nonlinear diode. We will then use the
small signal analysis method to approximate the behaviour of this circuit near the DC bias
point.
In particular, at vD = VD , using linear approximation, the function iD = f (vD ) can be
approximated as
df (vD ) 

iD ≈ f (VD ) + · (vD − VD )
dvD vD =VD
df (vD ) 

= f (VD ) + · vd .
dvD  vD =VD

Note that  iD = ID + id and ID can be written as ID = f (VD ). Hence, the term

df (vD ) 
· vd can be treated as the AC current id , and we have
dvD  vD =VD

df (vD ) 

id = · vd .
dvD vD =VD
2.6 Linear Approximation 89

df (vD ) 

Further, we note that is independent of vd , so id is in proportion with vd
dvD vD =VD
after linear approximation, i.e.,

id ∝ vd .

A graphical interpretation of the small signal analysis is shown below.

iD

iD = f (vD )

df (vD ) 

A iD = f (VD ) + · vd
dvD vD =VD
id B

ID
vd

vD
O VD

The point (VD , ID ) representing the steady DC voltage and current is called the operating
point. Note that the small signal analysis is dependent on the operating point, i.e., the DC
bias voltage and current. From the figure above, we see that over a restricted range around
vD = VD , the time-varying AC current id becomes a linear function of the AC voltage vd (or
graphically, we approximate A with B in the restricted region), which facilitates the analysis
of the nonlinear circuit over this range.

Differential

2.55 Definition: differential

The differential of y = f (x) at x, denoted y

as dy or df (x), is
R
dy = f 0 (x) ∆x.
dy
Q ∆y
We often call the change in x, ∆x, as the P
S
differential of the independent variable, de- dx = ∆x
noted as dx, that is dx = ∆x. Hence,
x
dy = f 0 (x) dx. O x x + ∆x
90 Differentiation

2.56 Example: differential - numerical approximation

Compute the differential of the function f (x) = cos x. Use it to find approximate value of
cos 94◦ .

Solution For the function y = cos x, we have y 0 = − sin x, so that dy = − sin x dx. Thus,

cos(a + ∆x) − cos a = ∆y ≈ dy = − sin a dx = − sin a ∆x.

Notice that 94◦ = 90◦ + 4◦ = 21 π + 4

180 π = 12 π + 1
45 π. By taking a = 12 π and ∆x = 1
45 π, we
have

cos 94◦ = cos(a + ∆x) ≈ cos a − sin a · ∆x

= cos 21 π − sin 12 π · ( 45
1
π)
=0−1· 1
45 π ≈ −0.0698132.

The real value of cos 94◦ is −0.0697565 · · · .

2.7 Higher Derivatives

2.57 Definition: higher derivatives

The second derivative of f is the derivative of the derivative of f :

0
f 00 (x) = (f 0 ) (x).

The third derivative of f is the derivative of the second derivative of f :

0
f 000 (x) = (f 00 ) (x).

In general, the n-th derivative of f is the derivative of the (n − 1)th derivative of f :

 0
f (n) (x) = f (n−1) (x).

2.58 Example: higher derivatives

Calculate the second derivative of f (x) = sin x2 .

Solution We apply the chain rule to calculate the first derivative:

f 0 (x) = (cos x2 ) · 2x = 2x cos x2 .
Then we need to use product rule and the chain rule to calculate the second derivative:
0
f 00 (x) = (f 0 (x)) = (2x cos x2 )0
= (2x)0 · cos x2 + 2x · (cos x2 )0 = 2 cos x2 − 4x2 sin x2 .
2.7 Higher Derivatives 91

2.59 Example: higher derivatives - polynomials

dk
For any polynomial Pn (x) of degree n, show that Pn (x) = 0 for k > n.
dxk
dm m
Solution Since (xm )0 = mxm−1 for integer m, we know that (x ) = m! and hence
dxm
dm+1 m dk m
(x ) = (m!)0
= 0. Thus we obtain (x ) = 0 if k > m. If k > n, then
dxm+1 dxk
dk dk
Pn (x) = (a0 + a1 x + · · · + an xn )
dx k dxk
dk dk dk
= a0 k (1) + a1 k (x) + · · · + an k (xn )
dx dx dx
= 0 + 0 + · · · + 0 = 0.

2.60 Example: higher derivatives - sin x and cos x

Prove that
dn dn
sin x = sin(x + 12 nπ), cos x = cos(x + 21 nπ).
dxn dxn
Solution Both equalities can be proved by induction. Here we only demonstrate the proof
of the first equality.
Clearly, since (sin x)0 = cos x = sin(x + 21 π), the equality holds for n = 1.
dk
Suppose the equality holds for n = k, that is, sin x = sin(x + 12 kπ). Thus,
dxk
0
dk+1 d
 k
sin x = sin x
dxk+1 dxk
0
= sin(x + 21 kπ)


= sin(x + 12 kπ + 12 π)
= sin x + 12 (k + 1)π .
 

So, the equality holds for n = k + 1.

Hence, the first equality holds for all n ≥ 1.

2.61 Example: higher derivatives - implicit differentiation

For the function y = f (x) defined implicitly by y 2 + x2 = 1, by implicit differentiation, we

get
dy dy x
2y + 2x = 0 =⇒ =− .
dx dx y
For finding the second derivative, we can adopt one of the following two approaches.
92 Differentiation

dy x
(1) By differentiating = − once more, we have
dx y
dy
d2 y d

x
 1·y−x·
= − =− dx
dx2 dx y y2
 
x
1·y−x· −
y x2 + y 2 1
=− = − = − 3.
y2 y3 y
dy
(2) By implicitly differentiating 2y + 2x = 0, we have
dx
 2
dy d2 y
2 + 2y 2 + 2 = 0,
dx dx

so that
2 2
dy
 
x
1+ 1 + −
d2 y dx y
=− =−
dx2 y y
x2 + y 2 1
=− = − 3.
y3 y

2.62 Example: acceleration

g 2
Let f (t) = t be the distance function of a free-fall object and t be the time. Its derivative
2
is the velocity v(t) = f 0 (t) = gt. The rate of change of the velocity is the acceleration a(t):
a(t) = v 0 (t) = f 00 (t). Thus the acceleration is the second derivative of f (t).
In general, if y = f (t) is a distance function, then the first derivative of f is the velocity
v(t) = f 0 (t) and the second derivative of f is the acceleration a(t) = v 0 (t) = f 00 (t).

2.63 Engineering Example: Newton’s second law of motion

Newton’s second law of motion presents the relationship between the net force acting on
an object and the rate of the object’s momentum change. This is one of the most important
laws in physics. It was first proposed by Isaac Newton in his book “Philosophiae Naturalis
Principia Mathematica (Mathematical Principles of Natural Philosophy)”, which was first
published in 1687. In terms of magnitude, Newton’s second law is given in the following:
dp dv
F = =m ,
dt dt
where F is the net force acting on an object, p is the momentum of the object, v is the speed,
and t is the time.
Note that the speed v equals to the rate of an object’s displacement change with respect to
2.7 Higher Derivatives 93

dx
time, i.e., v = , where x is the displacement of an object. Then, we can rewrite Newton’s
dt
second law as
d2 x
F =m .
dt2
Suppose that a ball of mass 0.25 kg is rolling on the ground, and it starts to decelerate at
t = 0. The displacement of this ball from its original position at t = 0 is x(t) = 2t − 0.2t2
(0 ≤ t ≤ 5). What is the resistive force acting on the ball?
According to Newton’s second law, we have

d2 x(t) d2 2t − 0.2t2 d (2 − 0.4t)

F =m = 0.25 × = 0.25 × = −0.1 (N).
dt 2 dt 2 dt
The result above shows that the magnitude of the resistive force acting on the ball is 0.1 N.
The sign “−” means the direction of the resistive force is opposite to the rolling direction of
the ball.

Leibniz Formula

2.64 Theorem: Leibniz Formula

If f (n) and g (n) exits, so is (f · g)(n) and the Leibniz Formula holds:
n  
(n)
X n
[f (x) · g(x)] = f (i) (x) · g (n−i) (x).
i=0
i

This is also called the Leibniz Rule.

Justification:
The formula can be proved by induction. In fact, it is clear that the formula holds for
n = 1:
1  
0
X 1 (i)
[f (x) · g(x)] = f 0 (x) g(x) + f (x) g 0 (x) = f (x) · g (1−i) (x).
i=0
i
Suppose the formula holds for n = k, that is,
k  
(k)
X k
[f (x) · g(x)] = f (i) (x) · g (k−i) (x).
i=0
i
94 Differentiation

Thus,
(k+1)
[f (x) · g(x)]
" k   #0
X k
= f (x) · g
(i) (k−i)
(x)
i=0
i
k  h
X k i
= f (i+1) (x) · g (k−i) (x) + f (i) (x) · g (k+1−i) (x)
i=0
i
k  
X k
=f (k+1)
(x) g(x) + f (i) (x) · g (k+1−i) (x)
i=1
i − 1
k  
X k (i)
+ f (x) · g (k+1−i) (x) + f (x) g (k+1) (x)
i=1
i
= f (k+1) (x) g(x)
k    
X k k
+ + f (i) (x) · g (k+1−i) (x) + f (x) g (k+1)
i=1
i − 1 i
k+1
X k + 1 
= f (i) (x) · g (k+1−i) (x)
i=0
i

So, the Leibniz Formula holds for all n ≥ 1.

2.65 Example: Leibniz Formula

dn
Compute x2 ex for n ≥ 1.


dxn
Solution We know that

(x2 )0 = 2x, (x2 )00 = 2, (x2 )(i) = 0 for all i > 2.

Thus, by the Leibniz Formula, we have

dn n(n − 1) 2 00 x (n−2)
x2 ex = x2 (ex )(n) + n(x2 )0 (ex )(n−1) + (x ) (e )


dx n 2
= x + 2nx + n(n − 1) e .
 2  x
 95

Summary

Definition: basic concepts

f (x + h) − f (x)
derivative : f 0 (x) = lim
h→0 h
(n−1) 0
higher derivative : f (x) = f (x), n ≥ 2
(n)

differential : dy = f 0 (x) dx

linear approximation : f (x) ≈ f (a) + f 0 (a)(x − a)

Theorem: rules of differentiation

Suppose f 0 (x) and g 0 (x) exist. Then the following rules of differentiation hold.

1. Sum Rule:
0
[f (x) + g(x)] = f 0 (x) + g 0 (x).

2. Difference Rule:
0
[f (x) − g(x)] = f 0 (x) − g 0 (x).

3. Product Rule:
0
[f (x) g(x)] = f (x) g 0 (x) + g(x) f 0 (x).

4. Quotient Rule: If g(a) 6= 0, then

0
f (x) g(x) f 0 (x) − f (x) g 0 (x)

= 2 .
g(x) [g(x)]

5. Chain Rule: Suppose f 0 and g 0 exist. Then

0
[f (g(x))] = f 0 (g(x)) · g 0 (x),

or in Leibniz notation,
dy dy du
= · .
dx du dx
6. Inverse Function: Suppose f is one-to-one and differentiable with non-zero derivative.
Then
 −1 0 1
f (x) = 0 −1 .
f (f (x))

7. Leibniz Formula: Suppose f (n) and g (n) exist. Then

n  
(n)
X n (i)
[f (x) · g(x)] = f (x) · g (n−i) (x).
i=0
i
96 Summary

Theorem: derivatives of some elementary functions

0
1. (c) = 0 10. (cot x)0 = − csc2 x.
0
2. (xµ ) = µxµ−1 1
11. (sin−1 x)0 = √
0
3. (ex ) = ex , 1 − x2
0
(ax ) = ax ln a 1
12. (cos−1 x)0 = − √
1 1 − x2
4. (ln x)0 = ,
x 1
1 13. (tan−1 x)0 =
(loga x)0 = 1 + x2
x ln a
5. (sin x)0 = cos x 1
14. (csc−1 x)0 = − √
|x| x2 − 1
6. (cos x)0 = − sin x
1
7. (tan x)0 = sec2 x 15. (sec−1 x)0 = √
|x| x2 − 1
8. (csc x)0 = − csc x cot x
1
9. (sec x)0 = sec x tan x 16. (cot−1 x)0 = −
1 + x2

Method: logarithmic differentiation

1. Take natural logarithms of both sides of an equation y = f (x) and use the Laws of
Logarithms to simplify.
2. Differentiate implicitly with respect to x.
3. Solve the resulting equation for y 0 .

Method: implicit differentiation

1. Replace y by y(x) in the given equation which defines the function y = y(x) implicitly.
2. Differentiate both sides of the equation with respect to x, by applying the rules of
differentiation.
dy
3. Solve for from the resulting equation.
dx
 CHAPTER

3
Applications of Differentiation

3.1 Maximum and Minimum Values

3.1 Definition: extreme values

Suppose f is a function defined over a domain D and c is a number in D.
Definition: absolute extreme values

The number f (c) is the

• absolute maximum value of f if f (c) ≥ f (x) for all x in D;
• absolute minimum value of f if f (c) ≤ f (x) for all x in D.

Definition: local extreme values

The number f (c) is a

• local maximum value of f if f (c) ≥ f (x) when x is near c;

• local minimum value of f if f (c) ≤ f (x) when x is near c.

When f attains its extremum at a point x = c, we call x = c an extreme point of f .

3.2 Example: absolute maximum and absolute minimum

1. Consider f (x) = x2 defined over the interval [−1, 1]. From the graph of the function,
we see that
98 Applications of Differentiation

• f has the absolute maximum value

f (±1) = 1; y

• f has the absolute minimum value y = f (x)

f (0) = 0; 1

• f has two local maximum values

f (±1) = 1;
x
O 1
• f has one local minimum value −1

f (0) = 0.

2. Consider g(x) = x2 defined over the interval (−1, 1). From the graph of the function,
we see that

• g has no absolute maximum value;

y
• g has the absolute minimum value
g(0) = 0; y = g(x)
1
• g has no local maximum value;
• g has one local minimum value g(0) =
0. x
−1 O 1

3. Consider h(x) = x2 defined over (−∞, +∞). From the graph of the function, we see
that

• h has no absolute maximum value;

y
• h has the absolute minimum value
h(0) = 0; y = h(x)
1
• h has no local maximum value;
• h has one local minimum value
h(0) = 0. x
−1 O 1

1
4. Consider k(x) = defined over (0, +∞). From the graph of the function, we see that
x
• k has no absolute maximum value; y

• k has no absolute minimum value;

• k has no local maximum value; y = k(x)
1
• k has no local minimum value. x
O 1
3.1 Maximum and Minimum Values 99

5. Consider r(x) = cos x defined over (−∞, +∞). From the graph, we see that, for
n = 0, ±1, ±2, . . . ,

• r has the absolute maximum value

r(2nπ) = 1;
y
• r has the absolute minimum value
r((2n + 1)π) = −1;
y = r(x) 1
• r has local maximum values r(2nπ) = 8
x
1; −π − 21 π O
−1
1
2π
π

• r has local minimum values r((2n +

1)π) = −1.

6. Consider the function p(x) = xe−x with the graph given below. We see that

• p has the absolute maximum value

p(1) = 1/e; y
• p has no absolute minimum value; 1 y = p(x)
• p has a local maximum values p(1) =
x
1/e; O 1

• p has no local minimum value.

7. For the function q(x) = |x|, we see that

• q has no absolute maximum value;

y
• q has the absolute minimum value
q(0) = 0; y = q(x)
1
• q has no local maximum value;
• q has one local minimum value q(0) =
0. x
−1 O 1

3.3 Theorem: Fermat’s Theorem

If f has a local maximum or minimum value at an interior point c in the domain D of f ,
and if f 0 (c) exists, then f 0 (c) = 0.
A point c is an interior point of D if there is an open interval (a, b) such that c ∈ (a, b) ⊂
D.

Justification:
Without loss of generality, suppose f has a local maximum at c, that is,

f (x) ≤ f (c) for x sufficiently close to c.

100 Applications of Differentiation

This implies that

f (c + h) − f (c) ≤ 0,
for all h sufficiently close to 0, since c is an interior point. Thus, if h > 0 and h is sufficiently
small, we have
f (c + h) − f (c)
≤ 0.
h
Since f is differentiable at c, we have

f (c + h) − f (c)
f 0 (c) = lim+ ≤ 0.
h→0 h
Similarly, if h < 0 and h is sufficiently small, we have

f (c + h) − f (c)
≥ 0.
h
Since f is differentiable at c, we have

f (c + h) − f (c)
f 0 (c) = lim− ≥ 0.
h→0 h
Since both f 0 (c) ≤ 0 and f 0 (c) ≥ 0 hold, we get f 0 (c) = 0.
The case of a local minimum can be proved in a similar manner.

Remark: interior point ⇐⇒ not a boundary point

That a point is an interior point means that it is not a boundary point. For example, the
point 0 is an interior point of the interval [−1, 1] since 0 ∈ (−1, 1) ⊂ [−1, 1]. The point −1 is
not an interior point of [−1, 1], but a boundary point.
The point 0 is not an interior point of the interval (0, 1) since it is even not a point in
(0, 1). In fact, 0 is a boundary point of (0, 1). The number 10−100 is an interior point of
(0, 1) since 10−100 ∈ (0.5 × 10−100 , 1.5 × 10−100 ) ⊂ (0, 1).

Remark: locations of local extrema

Fermat’s Theorem concludes that a local extreme value f (c), where c is in the domain of
f , only occurs at one of the following three types of locations:

• an interior point where f 0 (c) = 0;

• an interior point where f 0 (c) does not exist;
• a boundary point.

3.4 Definition: critical point

Let f be defined over a domain D. A number c in D is called a critical point of f if

either f 0 (c) = 0 or f 0 (c) does not exist.
3.1 Maximum and Minimum Values 101

3.5 Theorem: Fermat’s Theorem - rephrase

If f has a local maximum or minimum value at an interior point c in the domain D of f ,

then c is a critical point of f .

3.6 Example: local extreme occur at critical points

1. Consider the function p(x) = 2x3 − 3x2 − 12x + 1 with the graph given below. The
derivative of p is

p0 (x) = 6x2 − 6x − 12 = 6(x2 − x − 2) = 6(x − 2)(x + 1).

Solving p0 (x) = 0 gives x = 2 or x = −1. y

Thus these two points are all the critical
points of p. From the graph, we see that
8
the function p attains local extreme values
at these points. x
−1 O 2

y = 2x3 − 3x2 − 12x + 1

−19

2. Consider the function q(x) = |x|. Since q(x) does not have derivative at x = 0, so that
x = 0 is a critical point. The function also attains a local minimum value as well as an
absolute minimum value at x = 0.

3.7 Pitfall: critical points do not necessarily yield local extreme values

Fermat’s Theorem states that a local extremed value only occurs at an interior critical
point or at a boundary point. However, not all critical points will yield local extreme values.

1. Consider g(x) = x3 . Since g 0 (x) = 3x2 , y

x = 0 is a critical point of g. From the
graph of function, we see that x = 0 is not y = g(x)
1
a local extreme point.
x
−1 O 1

−1
102 Applications of Differentiation

2. Consider the function

y
 cos x, if x ≤ −1;


h(x) = sin x, if −1 < x ≤ 1;

y = h(x) 2
2, if x > 1.

1
It is not differentiable at points x = −1
x
and 1, so that they are critical points. −2π − 32 π −π − 12 π O 1

From the graph, we see that x = −1 and −1

1 are not local extreme points.

3.8 Theorem: Extreme Value Theorem

Let f be a continuous function defined over a finite closed interval [a, b]. Then f (x) has at
least one absolute maximum point and one absolute minimum point in the interval [a, b].

3.9 Example: continuous function on a finite closed interval

The function f (x) = x2 is continuous on y

the finite closed interval [−1, 1]. So the Ex-
y = f (x)
treme Value Theorem applies. In fact, the 1
function has the absolute maximum value
f (−1) = f (1) = 1 and the absolute minimum
value f (0) = 0. x
−1 O 1

3.10 Pitfall: continuous function may not have absolute maximum

The function g(x) = x2 is continuous on y

the finite interval (−1, 1). However, the Ex-
y = g(x)
treme Value Theorem does not apply to the 1
function g, since the interval (−1, 1) is not
closed. In fact, it is easy to see that the func-
tion g does not have absolute maximum value. x
−1 O 1

3.11 Pitfall: discontinuous function may have absolute extrema

Consider the function h:

y
 cos x, if x ≤ −1;


h(x) = sin x, if −1 < x ≤ 1;

y = h(x) 2
2, if x > 1.

1
The function is not continuous on the domain x
(−∞, +∞), and the domain is not closed nor −2π − 32 π −π − 12 π O 1
−1
finite. Nevertheless, it has an absolute maxi-
mum and absolute minimum.
3.1 Maximum and Minimum Values 103

3.12 Method: finding absolute extrema

Given a continuous function f over a finite closed interval [a, b].

1. Find the critical points and evaluate the values of f at critical points.
2. Evaluate the values of f at the boundary points x = a and x = b.

3. The largest value and smallest value from steps 1 and 2 are the absolute maximum
value and the absolute minimum value respectively.

3.13 Example: absolute maximum and absolute minimum

Consider f (x) = x2 defined over the in- y

terval [−2, 1]. The function f is continuous,
and the domain [−2, 1] is a finite closed in- 4

terval. Since the function f satisfies all the

y = f (x)
hypotheses in the Extreme Value Theorem,
the function f must have the absolute maxi-
1
mum value and the absolute minimum value.
x
−1 O 1

To find the absolute extrema, we precede as follows:

(1) First try to determine the critical points. Indeed, by solving the equation f 0 (x) = 2x =
0, we see that the critical point of f is x = 0. Thus, we get the value f (0) = 0.
(2) Next we evaluate the values of f at two boundary points: f (−2) = (−2)2 = 4 and
f (1) = 1.
(3) Among these three values, the largest one is f (−2) = 4 and the smallest one is f (0) = 0.
So the absolute maximum value occurs at x = −2 and the absolute minimum value occurs
at x = 0.

3.14 MATLAB: maximum and minimum values

1. To obtain the local maximum or minimum value of a function, we can first plot the
function in MATLAB to see the rough location of the local maximum or minimum value.
The following code shows how to plot the figure of the function p(x) = 2x3 −3x2 −12x+1.

>> syms x ;
>> p ( x ) = 2* x ^3 - 3* x ^2 - 12* x + 1;
>> x_array = -3:0.05:3;
>> y_array = p ( x_array ) ;
>> figure
>> plot ( x_array , y_array , 'r ')
104 Applications of Differentiation

>> grid
>> xlabel ( 'x ')
>> ylabel ( 'p ( x ) ')

After running the MATLAB code above, 10

we have the figure for the function p. 0

From this figure, we see that, for the -10

function p, the local maximum value is

p(x)
-20

around x = −1 and the local minimum

value is around x = 2. According to Fer-
-30

mat’s Theorem, we know that p0 (xmax ) = -40

p0 (xmin ) = 0, where xmax and xmin are the -50

-3 -2 -1 0 1 2 3

points obtaining the local maximum value x

and the local minimum value, respectively.

To obtain the exact locations of the local maximum and minimum values of p, we need
the following MATLAB code.

>> d = diff ( p ) ;
>> d_fun = matlabFunction ( d )

d_fun =

function_handle with value :

@( x ) x .* -6.0+ x .^2.*6.0 -1.2 e1

>> max_p = fzero ( d_fun , -1)

max_p =

-1

>> min_p = fzero ( d_fun ,2)

min_p =

where matlabFunction(d) is to convert the symbolic function d to a function handle

which can be analyzed by the command fzero. Note that the command solve is
not used here, because MATLAB may not be able to obtain explicit solutions for some
equations; while fzero can numerically find the zero of a function near the initial guess.
MATLAB returns that the local maximum value can be achieved at x = −1, and the
local minimum value can be achieved at x = 2.
3.1 Maximum and Minimum Values 105

2. The following example shows how to find the absolute minimum value of the function
f (x) = x2 − 2.1x + 1.1025. First, we plot the curve of f . The MATLAB code is shown
below.

>> syms x ;
>> f ( x ) = x ^2 - 2.1* x + 1.1025;
>> x_array = -1:0.2:3;
>> y_array = f ( x_array ) ;
>> figure
>> plot ( x_array , y_array , 'r ')
>> xlabel ( 'x ')
>> ylabel ( 'f ( x ) ')
>> grid

Then, we have the figure showing the 4.5

graph of f . 4

3.5

From this figure, we see that the absolute 3

minimum value of f (x) is around x = 1. f(x)

2.5

To find the exact location of the abso- 2

lute minimum value, we need the following

1.5

1
MATLAB code. 0.5

0
-1 -0.5 0 0.5 1 1.5 2 2.5 3
x

>> d = diff ( f ) ;
>> d_fun = matlabFunction ( d )

d_fun =

function_handle with value :

@( x ) x .*2.0 -2.1 e1 ./1.0 e1

>> min_p = fzero ( d_fun ,1)

min_p =

1.0500

MATLAB returns that the absolute minimum value of f (x) can be achieved at x = 1.05.
106 Applications of Differentiation

3.15 Engineering Example: fugacity coefficient

In chemical thermodynamics, the fugacity is an effective partial pressure of a real gas. It

equals the pressure of an ideal gas which has the same temperature and molar Gibbs free
energy as this real gas. For example, the Nitrogen gas (N2 ) at 0 ℃ and a pressure of P = 100
atmospheres (atm) has a fugacity of f = 97.03 atm. This means that the molar Gibbs free
energy of this real Nitrogen gas is the same with the ideal gas at 0 ℃ and a pressure of 97.03
atm. The pressure and the fugacity of a gas is related through a parameter called fugacity
f
coefficient, which is defined as ϕ = , where f is the fugacity and P is the pressure. For
P
example, the fugacity coefficient of the Nitrogen gas at 0 ℃ and a pressure of P = 100 atm
is 0.9703. Note that for the ideal gas, the fugacity coefficient is always 1.
The fugacity coefficient can be rewritten in the form of the compressibility factor which
is a correction factor describing the deviation of a real gas from ideal gas behaviour. In
particular, under certain conditions, the fugacity coefficient in terms of the compressibility
factor is provided as follows:
1
 
ϕ = f (x) = x − 1 − ln (x − 1) − ln 1 +
x
where x (x > 1) is the compressibility factor.
Use MATLAB to calculate the minimum value of the fugacity coefficient function and the
corresponding compressibility factor.
First, we plot the function f (x) = x − 1 − ln (x − 1) − ln 1 + x1 in MATLAB to see the

rough location of the minimum value. The code is shown in the following.

>> syms x
>> f ( x ) = x - 1 - log ( x - 1) - log (1 + 1/ x ) ;
>> x_array = 1.1:0.1:10;
>> y_array = f ( x_array ) ;
>> figure
>> plot ( x_array , y_array , 'r ')
>> grid
>> xlabel ( 'x ')
>> ylabel ( 'f ( x ) ')

From the graph of f , we see that the min- 7

imum value is around x = 2. According to 6

Fermat’s Theorem, to obtain the minimum 5
value of f (x), we need to calculate the value
of x which satisfies f 0 (x) = 0.
4
f(x)

0
0 2 4 6 8 10
x

Next, we run the following MATLAB code.

3.2 Monotone Functions and Mean Value Theorem 107

>> df = diff ( f )

df ( x ) =

1/( x ^2*(1/ x + 1) ) - 1/( x - 1) + 1

>> df_fun = matlabFunction ( df )

df_fun =

function_handle with value :

@( x ) -1.0./( x -1.0) +1.0./ x .^2./(1.0./ x +1.0) +1.0

>> x_min = fzero ( df_fun ,2)

x_min =

1.8393

It shows that f 0 (1.8393) = 0, i.e., the minimum value of f (x) is achieved when the compress-
ibility factor x = 1.8393.
Substituting x = 1.8393 into the fugacity coefficient function f (x), we have the minimum
value of the fugacity coefficient shown as follows.

>> f_min = vpa ( f ( x_min ) )

f_min =

0.58031458931953300345263227407065

The result shows that the minimum value of the fugacity coefficient is around 0.5803.

3.2 Monotone Functions and Mean Value Theorem

3.16 Definition: monotone functions
A function f is said to be
• increasing on an interval I if
f (x1 ) < f (x2 ) whenever x1 < x2 in I;

• decreasing on an interval I if
f (x1 ) > f (x2 ) whenever x1 < x2 in I.
108 Applications of Differentiation

Such functions are also said to be monotone (or monotonic) functions.

3.17 Example: monotone functions

1. The function f (x) = x2 is decreasing on y

the interval (−∞, 0) and increasing on the
y = x2
interval (0, +∞). 2

x
−1 O 1

2. The function g(x) = x3 is increasing on y

the whole real line (−∞, +∞).
1
y = x3
x
O 1

1 y
3. The function h(x) = is deceasing on the
x
interval (−∞, 0) as well as on the interval
(0, +∞). y=
1
x
1
−1
x
O 1
−1

4. The function k(x) = sin x is increasing on y

the interval (− 12 π, 21 π) and decreasing on
y = sin x
the interval ( 12 π, 32 π). 1

x
− 12 π O 1
2π
3
2π
−1

3.18 Theorem: Mean Value Theorem

Suppose f is a function that satisfies the following:

1. f is continuous on the closed interval [a, b].

2. f is differentiable in the open interval (a, b).
3.2 Monotone Functions and Mean Value Theorem 109

Then there exists at least one number c in (a, b) such that

f (b) − f (a)
f 0 (c) =
b−a
or, equivalently,
f (b) − f (a) = f 0 (c)(b − a).

y y

f (b) f (b)

f (a) f (a)

x x
O a c b O a c1 c2 b

Justification:

An equation of the line segment connecting (a, f (a)) and (b, f (b)) is
f (b) − f (a)
y = f (a) − (x − a).
b−a
As shown in the figure, at x with a ≤ x ≤ b, consider the auxiliary function
f (b) − f (a)
h(x) = f (x) − f (a) − (x − a).
b−a

f (b)

h(x)

f (x)
f (a) f (b) − f (a)
f (a) + (x − a)
b−a

x
O a x b

Since f is continuous on [a, b], so is h. By the Extreme Value Theorem, there exist absolute
maximum point M and absolute minimum point m.
If one of m and M is attained at an interior point of (a, b), since h is differentiable in (a, b),
by Fermat’s Theorem, there exists c ∈ (a, b) such that h0 (c) = 0. Since
f (b) − f (a)
h0 (x) = f 0 (x) − ,
b−a
110 Applications of Differentiation

we get
f (b) − f (a)
f 0 (c) = .
b−a
If both m and M are attained at the boundary points of [a, b], since

f (b) − f (a)
h(a) = f (a) − f (a) − (a − a) = 0,
b−a
f (b) − f (a)
h(b) = f (b) − f (a) − (b − a)
b−a
= f (b) − f (a) − [f (b) − f (a)] = 0,

we must have m = M = 0. This means that h must be a constant function. Thus, for any
c ∈ (a, b), we have h0 (c) = 0. Similarly as above, we again have

f (b) − f (a)
f 0 (c) = .
b−a

3.19 Pitfall: when the hypotheses in the Mean Value Theorem fail

The hypotheses of the Mean Value Theorem are that the function f is continuous on the
closed interval [a, b] and differentiable in the open interval (a, b). We will see that if any of
these fails, then the conclusion of the Mean Value Theorem does not necessarily hold.
1. Consider the function f defined on [0, 1]:
1, if 0 < x ≤ 1;

f (x) =
0, if x = 0.
Clearly, f is discontinuous at x = 1, so that f is not continuous on [0, 1]. Obviously,
for any x ∈ (0, 1), f 0 (x) = 0 and
f (1) − f (0) 1−0
= = 1.
1−0 1−0
So there exists no c in (0, 1) such that
f (1) − f (0)
f 0 (c) = .
1−0
2. Consider the function g(x) = |x| defined over [−1, 1]. We know that g 0 (0) does not
exist, so g is not differentiable in (−1, 1). It is easy to see that
g(1) − g(−1) 1−1
= = 0,
1 − (−1) 2
and
−1, if −1 < x < 0;

g (x) =
0
1, if 0 < x < 1.
Clearly, there exists no c inside (−1, 1) such that
g(1) − g(−1)
g 0 (c) = .
1 − (−1)
3.2 Monotone Functions and Mean Value Theorem 111

3.20 Theorem: Monotone Test

Suppose f is differentiable on an interval I.

1. If f 0 (x) > 0 on I, then f is increasing on I.

2. If f 0 (x) < 0 on I, then f is decreasing on I.

Justification:

Suppose f 0 (x) > 0 on (a, b). We take any two numbers x1 , x2 ∈ (a, b) such that with
x1 < x2 . By the Mean Value Theorem, there is c ∈ (x1 , x2 ) such that

f (x1 ) − f (x2 ) = f 0 (c)(x1 − x2 ) < 0,

which implies that f (x1 ) < f (x2 ). Hence, f is increasing on (a, b).
The other case can be proved similarly.

3.21 Example: determine intervals of monotonicity

Consider the function f (x) = 2x3 − 3x2 − 12x + 1. To find the intervals where f is
monotone, by the Monotone Test, we need to determine the intervals in which f 0 keeps its
sign unchanged.
In fact,
f 0 (x) = 6x2 − 6x − 12 = 6(x2 − x − 2) = 6(x − 2)(x + 1),
so that it has two real roots x = −1 and x = 2. These two numbers divide the interval
(−∞, +∞) into three open intervals (−∞, −1), (−1, 2), and (2, +∞). Since the function f 0
is continuous and nonzero over each of these intervals, by the Intermediate Value Theorem,
it must be either positive or negative on each interval. In (−∞, −1), since f 0 is positive for
large negative x, f 0 (x) is positive on (−∞, −1). The function f 0 changes its sign on two sides
of x = −1 and x = 2, so that f 0 (x) is negative on (−1, 2) and positive on (2, +∞).

y y

x x
−1 O 2 −1 O 2

y = f (x)
y = f 0 (x) −19

Thus the function f (x) is increasing over the intervals (−∞, −1) and (2, +∞), and de-
creasing over the interval (−1, 2).
112 Applications of Differentiation

3.22 Method: solving inequalities by using the number line

There is a method that works for solving inequalities of the form f (x) > 0, or inequalities
involving ≥, <, or ≤, where f is a polynomial. It consists of the following major steps:
(1) Solve the equation f (x) = 0 to find all real roots.
(2) Specify the roots on the number line.
(3) Sketch the graph of y = f (x) by starting from x near −∞. If the order of f is odd, the
graph of f is below the number line; if the order of f is even, the graph is above the number
line.
(4) From the left to the right, the curve y = f (x) intersects the number line at the roots,
one after another.
(5) The curve y = f (x) crosses the number line at roots when the multiplicities are odd
and does not cross at roots when the multiplicities are even.
(6) The inequality f (x) > 0 holds on the interval over which the graph of y = f (x) is above
the number line.
The method can be modified for inequalities with f being rational functions or even tran-
scendental functions.

3.23 Theorem: zero derivative implies constant function

Suppose f is differentiable in (a, b). If f 0 (x) = 0 for all x in (a, b), then f is a constant
function.

Justification:

For any two numbers x1 , x2 in (a, b), by the Mean Value Theorem, there is a number
x∗ ∈ (x1 , x2 ) ⊂ (a, b) such that

f (x1 ) − f (x2 ) = f 0 (x∗ )(x1 − x2 ).

Since f 0 (x) = 0 for all x in an interval (a, b), we have f 0 (x∗ ) = 0. Thus, f (x1 ) = f (x2 ). This
shows that at any two numbers in (a, b), the values of f are the same. So, f is constant on
(a, b).

3.24 Example: identity proving

Show that when |x| ≤ 12 , the following identity holds:

3 arccos x − arccos(3x − 4x3 ) = π.

Solution When |x| < 21 ,

0
3 arccos x − arccos(3x − 4x3 )


3 1
= −√ +p (3 − 12x2 )
1 − x2 1 − (3x − 4x3 )2
3 1
= −√ +√ (3 − 12x2 ) = 0,
1 − x2 1 − x2 · (1 − 4x2 )
3.2 Monotone Functions and Mean Value Theorem 113

so we have
3 arccos x − arccos(3x − 4x3 ) = C, − 21 < x < 12 ,
where C is a constant. Setting x = 0 gives C = π. Hence,

3 arccos x − arccos(3x − 4x3 ) = π, − 12 < x < 21 .

It is easy to verify that the identity also holds when x = ± 21 . Therefore,

3 arccos x − arccos(3x − 4x3 ) = π, |x| ≤ 12 .

3.25 MATLAB: monotone function

The following example shows how to determine the monotone intervals of the function
f (x) = 2x3 − 3x2 − 12x + 1 in MATLAB.

>> syms x ;
>> f ( x ) = 2* x ^3 - 3* x ^2 - 12* x + 1;
>> d ( x ) = diff ( f ( x ) ,x ) ;
>> rts_d = solve ( d ( x ) == 0 , x )

rts_d =

-1
2

>> x_array = -3:0.05:3;

>> figure
>> plot ( x_array , d ( x_array ) , 'b ')
>> xlabel ( 'x ')
>> ylabel ( 'f ' '( x ) ')
>> grid

Note that for a string (e.g., x inside ‘x’ and f”(x) inside ‘f”(x)’ ), we need to use double
single quotes to obtain a single quote (simply using one single quote represents conjugate
transpose in MATLAB), e.g., we use ‘f”(x)’ to represent f 0 (x).

From the MATLAB code above, we have 60

the graph of f 0 and see that f 0 (x) = 0 has 50

two roots: x = −1 and x = 2. We also see 40

that f 0 is negative on (−1, 2), and positive 30

on (−∞, −1) and (2, +∞). According to the
f'(x)

20
Monotone Test, we know that f is decreasing 10
on the interval (−1, 2), and f is increasing on 0
the intervals (−∞, −1) and (2, +∞). -10

-20
-3 -2 -1 0 1 2 3
x
114 Applications of Differentiation

Actually, we can plot f to check the above results.

>> f ( x ) = 2* x ^3 - 3* x ^2 - 12* x + 1;
>> figure
>> plot ( x_array , f ( x_array ) , 'r ')
>> xlabel ( 'x ')
>> ylabel ( 'f ( x ) ')
>> grid

We see from the graph of f that indeed 10

f is decreasing on (−1, 2), and increasing on

(−∞, −1) and (2, +∞).
0

-10

f(x)
-20

-30

-40

-50
-3 -2 -1 0 1 2 3
x

3.3 Tests for Local Extrema and Concavity

Local Extrema

3.26 Theorem: First Derivative Test

Suppose that c is a critical point of a continuous function f .

1. If f 0 changes from positive to negative at c, then f has a local maximum value at c.

2. If f 0 changes from negative to positive at c, then f has a local minimum value at c.

3. If f 0 does not change sign at c, then f has no local extreme value at c.

3.3 Tests for Local Extrema and Concavity 115

y y

f 0 (x) < 0 f 0 (x) > 0 f 0 (x) > 0 f 0 (x) < 0

c x c x
O O

f 0 changes from positive to negative at c f 0 changes from negative to positive at c

y y

f 0 (x) > 0 f 0 (x) < 0

f 0 (x) > 0 f 0 (x) < 0

c x c x
O O

No maximum or minimum No maximum or minimum

Justification:

Near c, suppose f 0 (x) is positive for a < x < c and negative for c < x < b. Then f (x)
is continuous on [a, c] and differentiable on (a, c). Thus f (x) is increasing over [a, c], i.e. for
any x with a < x < c, f (x) < f (c). By the similar argument, f (x) is decreasing over [c, b],
i.e. for any x with c < x < b, f (c) > f (x). Therefore c is a local maximum point.
The same argument applies to the other two cases.

3.27 Example: local extrema using the First Derivative Test

1. Consider the function f (x) = 2x3 − 3x2 − 12x + 1. Its derivative is

f 0 (x) = 6x2 − 6x − 12 = 6(x − 2)(x + 1).

116 Applications of Differentiation

By solving the equation f 0 (x) = 0, we see y

that f has critical points x = −1, 2. It
is easy to see that f 0 (x) is positive for
8
x < −1, negative for −1 < x < 2, and
positive for x > 2. Thus, by the First x
−1 O 2
Derivative Test, we know that x = −1 is a
y = 2x3 − 3x2 − 12x + 1
local maximum point and x = 2 is a local
−19
minimum point.

2. Consider the function g(x) = x3 . Its y

derivative is g 0 (x) = 3x2 . By solving the
equation g 0 (x) = 0, we see that g has one
y = x3
critical point x = 0. It is easy to see 1

that near x = 0, g 0 (x) does not change

x
signs. Thus, by the First Derivative Test, −1 O 1

we know that x = 0 is not a local extreme −1

point.

3. Consider the function

h(x) = |x|. y

It is not differentiable at x = 0, so that y = |x|

1
h has a critical point at x = 0. When
x < 0, h0 (x) = −1 < 0; when x > 0,
h0 (x) = 1 > 0. Thus, by the First Deriva-
x
tive Test, x = 0 is a local minimum point. −1 O 1

4. Consider r(x) = 1
2 sin x − 12 x cos x over (−5, 5). Its derivative is

r0 (x) = 1
2 cos x − 21 (cos x − x sin x) = 12 x sin x.

Solving r0 (x) = 12 x sin x = 0 in the interval (−5, 5) gives three critical points −π, 0, π.

In fact, the derivative r0 (x) = 21 x sin x has y

simple root at x = −π and π, and a double y = r0 (x)
1
root at x = 0. The later is because
x
−π O π 5
r0 (x) sin x −5
−1
lim = 1
lim
2 x→0 = 1
2 6= 0.
x→0 x2 x

Since sin x is positive for −5 < x < −π, we know that r0 (x) is negative near x = −5.
Thus, we can sketch the graph of r0 , as shown.
3.3 Tests for Local Extrema and Concavity 117

From the graph of r0 , we see that r0 (x) changes from negative to positive at the critical
point x = −π, does not change sign at the critical point x = 0, and changes from
positive to negative at the critical point x = π.

Hence, by the First Derivative Test, f has y

a local minimum value at x = −π, not a
local extreme at x = 0, and a local maxi- y = r(x) 1
mum value at x = π. −π
x
−5 O π 5
−1

3.28 Theorem: Second Derivative Test

Suppose f 00 is continuous near c.

1. If f 0 (c) = 0 and f 00 (c) > 0, then f has a local minimum value at c.

2. If f 0 (c) = 0 and f 00 (c) < 0, then f has a local maximum value at c.

3. If f 0 (c) = 0 and f 00 (c) = 0, it is inconclusive about local extreme of f at c.

Justification:

1. Assume f 0 (c) = 0 and f 00 (c) > 0. Since f 00 is continuous near c, we know that f 00 (x) > 0
near c. Thus f 0 is an increasing function near c. Hence for x near c, f 0 (x) < f 0 (c) = 0 for
x < c, and f 0 (x) > f 0 (c) = 0 for x > c. By the First Derivative Test, c is a local minimum
point.
2. The proof for this case is similar as case 1.
3. We use two functions to demonstrate this. They both satisfy the hypotheses, but the
first function has no local extreme value at c while the second has a local maximum value at
c.
(1) Consider f (x) = x3 . It is easy to see that f 0 (0) = 0, f 00 (0) = 0, and f 00 (x) = 6x is
continuous near x = 0. Since f (x) < 0 = f (0) when x < 0 and f (x) > 0 = f (0), f (0) is not
a local extreme value.
(2) Consider g(x) = x4 . It is easy to see that g 0 (0) = 0, g 00 (0) = 0, and g 00 (x) = 12x2 is
continuous near x = 0. Since g(x) ≥ 0 = g(0) for all real x, g(0) is a local minimum value.

3.29 Example: local extrema using the Second Derivative Test

1. Consider f (x) = 2x3 − 3x2 − 12x + 1. Its derivatives are

118 Applications of Differentiation

f 0 (x) = 6x2 − 6x − 12 y
= 6(x − 2)(x + 1),
f 00 (x) = 12x − 6. 8

x
Solving f 0 (x) = 0 gives x = −1, 2. Since −1 O 2

f 00 (−1) = −18 < 0, f (−1) = 8 is a lo- y = 2x3 − 3x2 − 12x + 1

cal maximum value; f 00 (2) = 18 > 0, thus −19
f (2) = −19 is a local minimum value.

2. Consider r(x) = 1
2 sin x − 12 x cos x over (−5, 5). Its derivatives are

r0 (x) = 2 cos x − 2 (cos x −

1 1
x sin x) = 21 x sin x,
r00 (x) = 2 (sin x + x cos x).
1

Solving r0 (x) = 12 x sin x = 0 in the interval y

(−5, 5) gives three critical points −π, 0, π.
Since r00 (−π) = 21 π > 0, r has a local min- y = r(x) 1
imax value at x = −π; r00 (π) = − 12 π < 0, −π
x
−5 O π 5
r has a local minimum value at x = π. −1

However, since r00 (0) = 0, the Second Derivative Test is inconclusive in this case. In
fact, we can apply the First Derivative Test to conclude that r has not a local extreme
at x = 0, since r0 does not change sign at the critical point x = 0.

3.30 Pitfall: the Second Derivative Test may not be applicable

Consider the function f (x) = x|x|. When x 6= 0,

−2x, if x < 0;

f 0 (x) =
2x, if x > 0.

To calculate the derivative f 0 (0), we use the definition of derivative.

f (x) − f (0) −2x2

lim− = lim− = lim− (−2x) = 0,
x→0 x−0 h→0 x h→0
f (x) − f (0) 2x2
lim = lim+ = lim+ (2x) = 0.
x→0+ x−0 h→0 x h→0

Since the left limit and right limit are equal, f 0 (0) = 0. Thus f 0 (x) = 2|x| for all x ∈ R.
3.3 Tests for Local Extrema and Concavity 119

Clearly f 0 (x) has no derivative at x = 0. y

Hence f does not have the second derivative
at x = 0, so that the Second Derivative Test
cannot be applied to the function f . y = x|x|
However, we can still apply the First 1

Derivative Test to show that f (0) is not a x

O 1
local extreme value, since f 0 (0) = 0 and f 0 −1

does not change sign at x = 0. −1

3.31 MATLAB: tests for local extrema

The following example shows the tests for local extrema of the function f (x) = 2x3 − 3x2 −
12x + 1.
We first test the local extrema of f using First Derivative Test.

>> syms x ;
>> f ( x ) = 2* x ^3 - 3* x ^2 - 12* x + 1;
>> d ( x ) = diff ( f ( x ) ,x ) ;
>> rts_d = solve ( d ( x ) == 0 , x )

rts_d =

-1
2

>> x_array = -3:0.05:3;

>> figure
>> plot ( x_array , d ( x_array ) , 'b ')
>> xlabel ( 'x ')
>> ylabel ( 'f ' '( x ) ')
>> grid

From the MATLAB code above, we have 60

the graph of f 0 and see that f 0 (x) = 0 has 50

two roots: x = −1 and x = 2. We also see 40

that f 0 is negative on (−1, 2), and positive on 30

(−∞, −1) and (2, +∞). According to First
f'(x)

20
Derivative Test, we know that f has a local 10
maximum value at x = −1 and a local mini- 0
mum value at x = 2. -10

-20
-3 -2 -1 0 1 2 3
x
120 Applications of Differentiation

Next, we test the local extrema of f using Second Derivative Test.

>> d2 ( x ) = diff ( f ( x ) ,x ,2)

d2 ( x ) =

12* x - 6

>> d2 ( -1)

ans =

-18

>> d2 (2)

ans =

18

We see from the MATLAB code above that f 00 (−1) = −18 < 0 and f 00 (2) = 18 > 0. Hence,
according to Second Derivative Test, f has a local maximum value at x = −1 and a local
minimum value at x = 2.
Actually, we can plot f to check the above results.

>> figure
>> plot ( x_array , f ( x_array ) , 'r ')
>> xlabel ( 'x ')
>> ylabel ( 'f ( x ) ')
>> grid

We see from the figure that f indeed has a 10

local maximum value at x = −1 and a local

minimum value at x = 2.
0

-10
f(x)

-20

-30

-40

-50
-3 -2 -1 0 1 2 3
x
3.3 Tests for Local Extrema and Concavity 121

3.32 Engineering Example: genetics

According to genetics, if a parent provides a gene A with probability p (0 ≤ p ≤ 1) and a

gene a with probability 1−p, then the offspring is of the genotype AA with probability p2 , the
genotype Aa with probability 2p(1−p) and the genotype aa with probability (1−p)2 . Suppose
that there are 100 people. Among them, 25 of them is of the genotype AA, 47 of them is of the
genotype Aa, and 28 of them is of the genotype aa. Let f (p) be the probability of observing
this specific sequence among these 100 people. Then, we can calculate the probability f (p)
as follows:

25 h i28
47 2
f (p) = p2 [2p (1 − p)] (1 − p)
= 247 p97 (1 − p)103 .

Calculate the probability p such that f (p) is maximized.

First, using the rules of differentiation, we calculate the first derivative of f in the following:

f 0 (p) = 247 97p96 (1 − p)103 − 103p9 (1 − p)102

 

= 247 p96 (1 − p)102 (97 − 200p).

Next, we calculate the value of p in the open interval (0, 1) at which f can achieve the
local maximum value. Note that when p ∈ (0, 1), we have p96 (1 − p)102 > 0. Hence, from the
expression above, we know that f 0 (p) = 0 when p = 20097
= 0.485 and

f 0 (p) > 0, if p < 0.485;

(

f 0 (p) < 0, if p > 0.485.

According to First Derivative Test, f can achieve the local maximum value at p = 0.485.
To verify whether f can achieve its absolute maximum value at p = 0.485, we need to
calculate the values of f (p) at the boundary points p = 0 and p = 1, respectively. We can
easily calculate that f (0) = f (1) = 0 < p(0.485). Therefore, we conclude that f can achieve
the absolute maximum value at p = 0.485.
MATLAB solution:
Alternatively, this problem can also be solved using MATLAB.
First, we calculate f 0 and the root of f 0 (p) = 0 using the following MATLAB code.

>> syms p
>> f ( p ) = 2^47* p ^97*(1 - p ) ^103;
>> df ( p ) = simplify ( diff ( f ( p ) ) )

df ( p ) =

-140737488355328* p ^96*(200* p - 97) *( p - 1) ^102

>> df_root = solve ( df ( p ) == 0 ,p >0 ,p <1) % specify the range ...

of p
122 Applications of Differentiation

df_root =

97/200

In order to verify whether p = 200 97

is a local maximum point or a local minimum point, we
plot the figure of f 0 using the following MATLAB code.

>> p_array = 0:0.01:1;

>> df_array = df ( p_array ) ;
>> figure
>> plot ( p_array , df_array , 'r ')
>> grid
>> xlabel ( 'p ')
>> ylabel ( 'f ' '( p ) ')

The figure is plotted as shown. We see 2

10-45

from the figure that f 0 (p) > 0 if p < 200

97
1.5

and f (p) < 0 if p > 200 . According to

0 97
1

First Derivative Test, we thus know that at 0.5

p = 200
97
, f achieves a local maximum value.
f'(p)

-0.5

-1

-1.5

-2
0 0.2 0.4 0.6 0.8 1
p

At last, to verify whether p = 200 97

is an absolute maximum point, we need to calculate
the value of f (0) and f (1), respectively. It can be easily shown that f (0) = f (1) = 0 and
97
> 0. Therefore, p = 20097
is an absolute maximum point and f achieves the absolute


f 200
maximum value at p = 200 97
.

3.33 Engineering Example: beam testing problem

Quality control is a process of reviewing the quality of all factors to confirm the fulfillment
of the quality requirements. A quality control engineer is testing the failure rate of a certain
type of beams under stress. To conduct the testing, beams are put under steady stress for
some time. Suppose that the beam can pass the test with probability p, then it cannot
pass the test with probability 1 − p. If the quality control engineer conducts trials for 100
beams and 5 of them do not pass the test, then the probability of observing this sequence of
successes and failures is

f (p) = p95 (1 − p)5 .

Calculate the probability p such that f (p) is maximized.

3.3 Tests for Local Extrema and Concavity 123

First, using the rules of differentiation, we calculate the first derivative of f (p) in the
following:

f 0 (p) = 95p94 (1 − p)5 − 5p95 (1 − p)4

= p94 (1 − p)4 (95 − 100p).

Next, we calculate the value of the local maximum point p in the open interval (0, 1). Note
that when p ∈ (0, 1), we have p94 (1 − p)4 > 0. Hence, from the expression above, we know
that f 0 (p) = 0 when p = 100
95
= 0.95 and

f 0 (p) > 0, if p < 0.95;

(

f 0 (p) < 0, if p > 0.95.

According to First Derivative Test, we know that f (p) can achieve a local maximum value
at p = 0.95.
To verify whether p = 0.95 is an absolute maximum point, we now calculate the values of
f (p) at p = 0 and p = 1, respectively. It can be easily calculated that f (0) = f (1) = 0 and
f (0.95) > 0. Therefore, we conclude that f (p) is maximized at p = 0.95.
MATLAB solution:
Alternatively, this problem can also be solved using MATLAB.
First, we calculate f 0 (p) and the root of f 0 (p) = 0 using the following MATLAB code.

>> syms p
>> f ( p ) = p ^95*(1 - p ) ^5;
>> df ( p ) = simplify ( diff ( f ( p ) ) )

df ( p ) =

-5* p ^94*(20* p - 19) *( p - 1) ^4

>> df_root = solve ( df ( p ) == 0 ,p >0 ,p <1) % specify the range ...

of p

df_root =

19/20

In order to check whether p = 19 20 is a local maximum point or a local minimum point, we

plot the figure of f 0 (p) using the following MATLAB code:

>> p_array = 0:0.01:1;

>> df_array = df ( p_array ) ;
>> figure
>> plot ( p_array , df_array , 'b ')
>> grid
124 Applications of Differentiation

>> xlabel ( 'p ')

>> ylabel ( 'f ' '( p ) ')

The figure is plotted as shown. 6

10
-8

From the figure, we see that f 0 (p) > 0 if 4

p < 2019
and f 0 (p) < 0 if p > 1920 . Hence, 2

according to First Derivative Test, we know 0

20 is a local maximum point.

that p = 19

f'(p)
-2

-4

-6

-8

-10
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
p

At last, to check whether p = 1920 is an absolute maximum point, we need to calculate the
values of f
(p) at p = 0 and p = 1, respectively. It can be easily shown that f (0) = f (1) = 0
and f 19 20 > 0. Therefore, p = 20 is an absolute maximum point and f is maximized at
19

p = 20 .
19

3.34 Engineering Example: profit maximization

In economics, profit maximization is a short run or long run process by which a firm can
determine the price, input and output levels to make the greatest profit. Given price p ∈ R,
the demand of a product is d(p). If the unit cost is c, then to maximize the profit, the seller’s
objective is

max d(p)(p − c)
p

Calculate the value of p which satisfies the seller’s objective in the following cases:
1. If d(p) is a linear demand, i.e., d(p) = a − bp.
2. If d(p) is an exponential demand, i.e., d(p) = λ exp − pθ .

3. In the constant elasticity of substitution case, i.e., d(p) = ap−b (b > 1).
Let the objective function be f (p) = d(p)(p − c). First, by the Product Rule, we can easily
calculate f 0 (p) as follows:

f 0 (p) = d0 (p)(p − c) + d(p).

Next, we calculate the value of p which maximizes f (p) in different cases, respectively.

1. If d(p) = a − bp, we have

f 0 (p) = (−b)(p − c) + a − bp
= −2bp + bc + a.
3.3 Tests for Local Extrema and Concavity 125

Then, the root of f 0 (p) = 0 is p = 1

c+ a
.


2 b
It can be easily verified that

f 0 (p) > 0, if p < 1

c+ a
;
(

2 b

f 0 (p) < 0, if p > 1

c+ a


2 b .

Hence, according to First Derivative Test, we know that p = 12 c + ab maximizes f (p),

i.e., it satisfies the seller’s objective. Note that the optimal price p = 2 c + b increases
1 a

with the unit cost c.

 p
2. If d(p) = λ exp − , we have
θ
λ  p  p
f 0 (p) = − exp − (p − c) + λ exp −
θ θ  θ
 p  λ  c
= exp − − p+ 1+ λ .
θ θ θ

Then, we can calculate that the root of f 0 (p) = 0 is p = θ + c.

It can be easily verified that

f 0 (p) > 0, if p < θ + c;


f 0 (p) < 0, if p > θ + c.

Hence, according to First Derivative Test, we know that p = θ + c maximizes f (p), i.e.,
it satisfies the seller’s objective. Note that p = θ + c > c, which means the optimal
price is greater than and increasing with the unit cost c.
3. If d(p) = ap−b , we have

f 0 (p) = (−ab)p−(b+1) (p − c) + ap−b

= p−b (a − ab) + abcp−1 .
 

Then, the root of f 0 (p) = 0 is p = b−1 .

bc

It can be easily verified that

f 0 (p) > 0, if p < b−1 ;

bc


f 0 (p) < 0, if p > bc

b−1 .

Hence, according to First Derivative Test, we know that p = b−1 bc

maximizes f (p), i.e.,
it satisfies the seller’s objective. Note that p = b−1 > c, which means the optimal price
bc

is greater than and increasing with the unit cost c.

126 Applications of Differentiation

3.35 Engineering Example: maximum power transfer

RS

−
VS + RL

Consider the circuit in the following figure. A load of resistance RL is connected over a
power source of voltage VS and internal resistance RS .
According to Ohm’s Law, we have
VS
Electric Current : I = ;
RS + RL
VS2 RL
Output Power : P = 2.
(RS + RL )

Calculate the the value of RL such that the output power P is maximized.
V 2R
Let f (RL ) = P = (R S+RL )2 , then by Quotient Rule, we have
S L

2
VS2 (RS + RL ) − 2VS2 RL (RS + RL )
f 0 (RL ) = 4
(RS + RL )
V (RS − RL )
2
= S 3 .
(RS + RL )

Then, the root of f 0 (RL ) = 0 is RL = RS . Moreover, we can easily verify that

f (RL ) > 0, if 0 < RL < RS ;

 0
f 0 (RL ) < 0, if RL > RS .

Hence, according to First Derivative Test, we know that RL = RS maximizes f (RL ), i.e.,
the maximum power transfer occurs in the load when the load resistance (RL ) equals to the
source resistance (RS ). This is called a “matched condition”, and it illustrates the theorem
that the maximum power would be transferred to the output when the impedance of the
external device matches that of the source. Note that improper impedance matching would
lead to excessive power use and dissipation.

3.36 Engineering Example: projectile motion

Consider the case where a ball is thrown with initial velocity u and an angle θ
(0◦ ≤ θ ≤ 90◦ ) against the ground.

1. Find the relationship between the horizontal distance that the ball can reach sx , and
the variables u and θ.
3.3 Tests for Local Extrema and Concavity 127

2. Prove that in order to throw a ball as far as possible, we need to start the motion with
an angle of 45◦ .

y y

θ = 60◦

θ = 45◦
θ = 30◦
θ
x x
O O

1. First, we note that for the motion, the only force that affects the system is the gravi-
tational force in the vertical direction. For the vertical movement, we have
uy = u sin θ
where uy is the velocity in the y-axis direction.
With the velocity in the y-axis direction uy , we can calculate the displacement of the
ball in the y-direction at time t, which is shown in the following:
sy = uy t − 21 gt2
= (u sin θ)t − 12 gt2 .

When the ball hits the ground again, i.e., sy = 0, we have

(u sin θ)t − 12 gt2 = 0.
=⇒ t 21 gt − u sin θ = 0.

2u sin θ
Solving the above equation, we can obtain two roots: t = 0 and t = . Note
g
that at t = 0, the ball is at its initial position. Hence, the ball hits the ground again
2u sin θ 2u sin θ
at t = . In other words, the projectile motion ends at t = and the ball
g g
hits the ground again.
Next, we consider the horizontal movement of the ball. The velocity in the x-axis
direction can be calculated as follows:
ux = u cos θ.

With the velocity in the x-axis direction ux , we can calculate the displacement of the
ball in the x-direction at time t, which is shown in the following:
sx = ux t = (u cos θ)t.

2u sin θ
At t = , the displacement of the ball in the x-direction is
g
2u sin θ 2u2
sx = u cos θ = sin θ cos θ.
g g
128 Applications of Differentiation

2u2
2. Let f (θ) = sx = sin θ cos θ. To calculate the maximum horizontal movement of
g
the ball, i.e., to maximize sx , we first need to calculate f 0 (θ). According to Product
Rule, the calculation is shown in the following:

2u2
f 0 (θ) = cos2 θ − sin2 θ .


g

In the range of 0, π2 , f 0 (θ) = 0 has a solution at θ = π4 . Moreover, it can be easily

verified that
f (θ) > 0, if 0 ≤ θ < π4 ;
 0
f 0 (θ) < 0, if π4 < θ ≤ π.

Hence, according to First Derivative Test, we know that θ = π4 is a local maximum

point. In order to prove that θ = π4 is also an absolute maximum point, we need to
calculate
the values of f (0) and f (π). It can be easily calculated that f (0) = f (π) =
0 < f π4 . Therefore, θ = π4 is also an absolute maximum point, i.e., in order to throw
a ball as far as possible, we need to start the motion with an angle of 45◦ .

MATLAB solution:
2u2
We now turn to the use of MATLAB. Let f (θ) = sx = sin θ cos θ. Specifically, we
g
calculate f (θ) and the roots of f (θ) = 0 using MATLAB.
0 0

>> syms a x % a :2 u ^2/ g , x : theta

>> f ( x ) = a * sin ( x ) * cos ( x ) ;
>> df ( x ) = diff ( f ( x ) ,x )

df ( x ) =

a * cos ( x ) ^2 - a * sin ( x ) ^2

>> root_df = solve ( df ( x ) == 0 , x )

root_df =

pi /4

From the MATLAB
code above, we see that f 0 π4 = 0, i.e., f 0 (45◦ ) = 0. In order to

check whether f 0 π4 = 0 is a local maximum point or a local minimum point, we need

to know the sign of f 0 (θ) (i.e., positive or negative) when 0 ≤ θ < π4 and π4 < θ ≤ π,
2u2
respectively. Since a = > 0, we plot the figures of g (θ) = cos2 θ − sin2 θ for
g
0 ≤ θ < π4 and π4 < θ ≤ π, respectively, using the following MATLAB code.
3.3 Tests for Local Extrema and Concavity 129

>> syms x % x : theta

>> g ( x ) = cos ( x ) ^2 - sin ( x ) ^2;
>> x_array1 = linspace (0 , pi /4 ,100) ;
>> g_array1 = g ( x_array1 ) ;
>> subplot (1 ,2 ,1)
>> plot ( x_array1 , g_array1 , 'r ')
>> xlabel ( '\ theta ')
>> ylabel ( 'g (\ theta ) ')
>> grid
>> x_array2 = linspace ( pi /4 , pi /2 ,100) ;
>> g_array2 = g ( x_array2 ) ;
>> subplot (1 ,2 ,2)
>> plot ( x_array2 , g_array2 , 'b ')
>> xlabel ( '\ theta ')
>> ylabel ( 'g (\ theta ) ')
>> grid

The figure plotted is shown as follows:

1 0

0.9 -0.1

0.8 -0.2

0.7 -0.3

0.6 -0.4
g( )

g( )

0.5 -0.5

0.4 -0.6

0.3 -0.7

0.2 -0.8

0.1 -0.9

0 -1
0 0.2 0.4 0.6 0.8 0.8 1 1.2 1.4

Note that π
4≈ 0.7854. We thus see from the figure above that

g (θ) > 0, if 0 ≤ θ < π4 ; f (θ) > 0, if 0 ≤ θ < π4 ;

  0
=⇒
g (θ) < 0, if 4 < θ ≤ π
π
f 0 (θ) < 0, if π4 < θ ≤ π.

According to First Derivative Test, we know that θ = π4 is a local maximum point. In

order to prove that θ = π4 is also an absolute maximum point, we need to calculate the
values of f (0) and f (π). It can be easily obtained that f (0) = f (π) = 0 < f π4 .

Therefore, θ = π4 is also an absolute maximum point, i.e., in order to throw a ball as

far as possible, we need to start the motion with an angle of 45◦ .
130 Applications of Differentiation

Concavity

3.37 Definition: concave up and concave down

Suppose f is defined over an interval I. The function f is said to be

• concave up if its graph is above all its tangent lines over the interval;
• concave down if its graph is below all its tangent lines over the interval.

A point P is an inflection point of f if f is continuous at P and changes concavity at P .

3.38 Example: concave up and concave down

y y

2 y

1 1 y = x3 1
y = −x2
O −1
x x x
−1 O 1 −1 1 O 1
y=x 2
−1 −1

1. The function y = x2 is concave up on R.

2. The function y = −x2 is concave down on R.

3. The function y = x3 is concave down over the interval (−∞, 0) and concave up over
the interval (0, ∞). The point 0 is an inflection point of f .

3.39 Theorem: Concavity Test

Suppose f 00 exists on an interval I.

1. If f 00 (x) > 0 on I, then f is concave up on I.

2. If f 00 (x) < 0 on I, then f is concave down on I.

3.40 Example: Concavity Test

1. For the function y1 = x2 , we have y100 = 2 > 0. Thus, by the Concavity Test, this
function is concave up on R. For the function y2 = −x2 , we have y200 = −2 < 0 and the
function is concave down on R.
3.3 Tests for Local Extrema and Concavity 131

y y

2 1
y = −x2
1
O x
−1 1

x −1
−1 O 1
y=x 2

2. For the function f (x) = x3 , f 00 (x) = 6x. y

Since f 00 (x) < 0 when x < 0, by the Con-
cavity Test, f is concave down on (−∞, 0).
y = x3 1
Similarly, since f 00 (x) > 0 when x > 0, f
is concave up on (0, ∞). Thus, 0 is an in- −1
x
flection point. O 1

−1

3. Consider g(x) = 2x3 − 3x2 − 12x + 1. y

The second derivative g 00 (x) = 12x − 6 =
6(2x − 1). Since g 00 (x) < 0 when x < 21 , by
8
the Concavity Test, g is concave down on
(−∞, 12 ). Similarly, since g 00 (x) > 0 when x
−1 O 2
( 12 , − 11
2 )
x > 12 , g is concave up on ( 12 , +∞). Thus,
x = 21 is an inflection point.
−19
y = 2x3 − 3x2 − 12x + 1

3.41 MATLAB: concavity

The example below shows the concavity test of the function f (x) = 2x3 − 3x2 − 12x + 1.

>> syms x ;
>> f ( x ) = 2* x ^3 - 3* x ^2 - 12* x + 1;
>> d2 ( x ) = diff ( f ( x ) ,x ,2)

d2 ( x ) =

12* x - 6

>> rts_d2 = solve ( d2 ( x ) == 0 , x )

rts_d2 =
132 Applications of Differentiation

1/2

>> x_array = -3:0.05:3;

>> figure
>> plot ( x_array , d2 ( x_array ) , 'r ')
>> grid

From the MATLAB code, we know that f 00 30

is a linear function and f 00 (x) = 0 has one 20

root: x = 12 . Furthermore, from the figure of 10

f 00 above, we see that f 00 (x) < 0 if x < 21 and 0

f 00 (x) > 0 if f 00 (x) > 12 . Hence, according

f''(x)
-10
to the Concavity Test, f is concave down on -20
−∞, 12 and concave up on 21 , +∞ , while



-30
x = 12 is thus an inflection point.
-40

-50
-3 -2 -1 0 1 2 3
x

Actually, we can plot f to check the above results.

>> figure
>> plot ( x_array , f ( x_array ) , 'r ')
>> xlabel ( 'x ')
>> ylabel ( 'f ( x ) ')
>> grid

We can see from the graph of
f that f is in- 10

deed concave down on −∞, 12 and concave

0
up on 2 , +∞ .
1


-10
f(x)

-20

-30

-40

-50
-3 -2 -1 0 1 2 3
x
3.4 L’Hospital’s Rule 133

3.4 L’Hospital’s Rule

3.42 Example: indeterminate forms

sin x
1. In calculating limits, we often encounter indeterminate forms such as lim where it
x→0 x
0
becomes the indeterminate form if we naively plug x = 0 into the expression. In order
0
sin x
to calculate the limit, we used some geometry of the sine function to get lim = 1.
x→0 x

x2 − 1
2. For the limit lim , if we directly plug x = 1 into the expression, we will get the
x→1 x − 1
0
indeterminate form . To calculate the actual limit, we can perform the cancellation
0
x2 − 1 x2 − 1
= x + 1 and lim = lim (x + 1) = 2.
x−1 x→1 x − 1 x→1

ln x 0
3. For the limit lim , it is an indeterminate form if we plug x = 1 into the
x→1 x − 1 0
expression. We will develop a general approach to find such a limit.

3.43 Definition: indeterminate forms

The following are typical indeterminate forms:

indeterminate form of type 00

)
two standard indeterminate forms
indeterminate form of type ∞∞
indeterminate form of type 0 · ∞ indeterminate product

indeterminate form of type ∞ − ∞ indeterminate difference

indeterminate form of type 00 



indeterminate form of type ∞0 indeterminate powers

indeterminate form of type 1 ∞ 

Here the form 0

0 means a limit of the form

f (x)
lim ,
x→a g(x)

where f (x) → 0 and g(x) → 0 as x → a. The other indeterminate forms arise similarly in
the corresponding symbolic expressions.

3.44 Example: indeterminate forms

The following are some examples of indeterminate forms:

134 Applications of Differentiation

0 ln x • 00 : lim xx ;
• : lim ;
0 x→1 x − 1 x→0+

∞ ex
• : lim ; • ∞0 : lim x1/x ;
∞ x→∞ x x→+∞
• 0 · ∞: lim+ (sin x) ln x;
x→0 x
1

√ √
• ∞ − ∞: lim ( x + 1 − x); • 1∞ : lim 1− .
x→+∞ x→∞ x

3.45 Theorem: L’Hospital’s Rule

Suppose

1. f and g are differentiable functions;

2. g 0 (x) is not zero on an open interval I containing a except possibly at a;

3. lim f (x) = 0 = lim g(x) or lim f (x) = ∞ = lim g(x);
x→a x→a x→a x→a

f 0 (x)
4. lim exists or is ∞.
x→a g 0 (x)

Then
f (x) f 0 (x)
lim = lim 0 .
x→a g(x) x→a g (x)

The rule also holds when limits are one-sided limits or when x → a is replaced by x → ∞.

0
3.46 Example: form
0
ln x
Find the limit lim .
x→1 x−1

Solution For the given limit, we can verify the hypotheses of l’Hospital’s Rule:
1. Both functions ln x and x − 1 are differentiable;
2. (x − 1)0 = 1 is nowhere zero near x = 1;
3. lim ln x = 0 and lim (x − 1) = 0;
x→1 x→1
(ln x)0 1
4. lim = lim = 1.
x→1 (x − 1)0 x→1 x
Thus, legitimately we can apply l’Hospital’s Rule and get

ln x H (ln x)0 1
lim = lim = lim = 1.
x→1 x − 1 x→1 (x − 1)0 x→1 x
3.4 L’Hospital’s Rule 135

∞
3.47 Example: form
∞
ex
Find the limit lim .
x→+∞ x

Solution For the given limit, we can verify the hypotheses of l’Hospital’s Rule:
1. Both functions ex and x are differentiable;
2. x0 = 1 is nowhere zero;
3. lim ex = +∞ and lim x = +∞;
x→+∞ x→+∞
(ex )0
4. lim = lim ex = +∞.
x→+∞ (x)0 x→+∞
Thus, legitimately we can apply l’Hospital’s Rule and get

ex H (ex )0
lim = lim = lim ex = +∞.
x→+∞ x x→+∞ (x)0 x→+∞

3.48 Example: using l’Hospital’s Rule repeatedly

ex
Find the limit lim .
x→+∞ x2
Analysis: If we apply l’Hospital’s Rule, we get

ex H (ex )0 ex
lim = lim = lim .
x→+∞ x2 x→+∞ (x2 )0 x→+∞ 2x

For the last limit, we again apply l’Hospital’s Rule and get

ex H (ex )0 ex
lim = lim = lim = +∞.
x→+∞ 2x x→+∞ (2x) 0 x→+∞ 2

These two steps need to be justified by checking the hypotheses of l’Hospital’s Rule.

ex
Solution For the limit lim , we can verify the hypotheses of l’Hospital’s Rule:
x→+∞ 2x
1. Both functions e and 2x are differentiable;
x

2. (2x)0 = 2 is nowhere zero when x is large;

3. lim ex = +∞ and lim 2x = +∞;
x→+∞ x→+∞
(ex )0 ex
4. lim = lim = +∞.
x→+∞ (2x)0 x→+∞ 2
Thus, legitimately we can apply l’Hospital’s Rule and get

ex H (ex )0 ex
lim = lim = lim = +∞.
x→+∞ 2x x→+∞ (2x)0 x→+∞ 2

ex
For the original limit lim , we can verify the hypotheses of l’Hospital’s Rule:
x→+∞ x2
1. Both functions ex and x2 are differentiable;
2. (x2 )0 = 2x is nowhere zero when x is large;
136 Applications of Differentiation

3. lim ex = +∞ and lim x2 = +∞;

x→+∞ x→+∞
(e ) x 0
ex
4. As shown above, lim = lim = +∞.
x→+∞ (x2 )0 x→+∞ 2x
Thus, legitimately we can apply l’Hospital’s Rule and get

ex H (ex )0
lim = lim = +∞.
x→+∞ x2 x→+∞ (x2 )0

Remark:
Since the verification of the hypotheses of l’Hospital’s Rule is straightforward, often
ex
we omit this step. Thus, for finding the limit lim 2 , typically, we simply write:
x→+∞ x

ex H (ex )0 ex
lim = lim = lim
x→+∞ x2 x→+∞ (x2 )0 x→+∞ 2x

H (e )
x 0
ex
= lim = lim = +∞.
x→+∞ (2x)0 x→+∞ 2

3.49 Pitfall: l’Hospital’s Rule

1 − sin x
1. Consider lim .
x→0 1 + sin x
If one calculates the limit by using l’Hospital’s Rule without verifying the hypotheses,
one would get
1 − sin x H (1 − sin x)0 − cos x
lim = lim = lim = −1.
x→0 1 + sin x x→0 (1 + sin x)0 x→0 cos x

However, this is a wrong value! The correct value can be obtained by the continuity:
1 − sin x 1 − sin 0
lim = = 1.
x→0 1 + sin x 1 + sin 0
The reason for the failure is that one of the hypotheses of l’Hospital’s Rule is not
satisfied. In fact, we see that lim (1 − sin x) = 1, which is neither zero nor ∞.
x→0

x + sin x
2. Consider lim .
x→+∞ x
If one applies l’Hospital’s Rule, one would get
x + sin x H (x + sin x)0
lim = lim = lim (1 + cos x).
x→+∞ x x→+∞ (x)0 x→+∞

Clearly, the last limit does not exist. However, we have

x + sin x sin x
 
lim = lim 1+ = 1.
x→+∞ x x→+∞ x
3.4 L’Hospital’s Rule 137

This contradiction shows that not all indeterminate form can apply l’Hospital’s Rule.
The failure in this example is due to the fact that the limit lim (1 + cos x) does not
x→+∞
exist. This is one of the hypotheses of l’Hospital’s Rule.

3.50 Example: 0 · ∞ form

Find the limit lim+ (sin x) ln x.

x→0

Solution The given limit is indeterminate form of 0 · ∞. We may first convert it to a form
∞
of then apply l’Hospital’s Rule, as shown in the following:
∞
ln x
lim (sin x) ln x = lim+ −1
x→0+ x→0 (sin x)
0
H (ln x)
= lim+ h i0
x→0 −1
(sin x)
x−1
= lim+ −2
− (sin x) · cos x
x→0

sin2 x
= lim+ = 0.
x→0 −x cos x

Remark:
0
We may first convert it to a form of then apply l’Hospital’s Rule, as shown in the
0
following:
sin x
lim+ (sin x) ln x = lim+ −1
x→0 x→0 (ln x)
0
H (sin x)
= lim+ h i0
x→0 −1
(ln x)
cos x
= lim+ −2
x→0− (ln x) · x−1
= lim+ −x ln2 x · cos x .
 
x→0

We see that the last limit seems more complicated than the original one. This shows
0 ∞
that sometimes it is an art to convert indeterminate form to or in order to
0 ∞
rightfully apply l’Hospital’s Rule.
138 Applications of Differentiation

3.51 Example: ∞ − ∞ form

√ √
Find the limit lim ( x + 1 − x).
x→+∞

Solution The given limit is indeterminate form of ∞ − ∞. We may first convert it to a

0
form of then apply l’Hospital’s Rule, as shown in the following:
0
√
√ √ √ x+1

lim ( x + 1 − x) = lim x √ −1
x→+∞ x→+∞ x
√
1+x −1−1
= lim
x→+∞ x−1/2
√
0
H 1 + x−1 − 1
= lim
0
x→+∞ x−1/2
2 (1
+ x−1 )−1/2 · (−x−2 )
1
= lim
x→+∞ − 21 x−3/2
1
= lim √ √ = 0.
x→+∞ x · 1 + x−1

3.52 Example: 00 form

Find the limit lim+ xx .

x→0

Solution Let y = xx . If taking logarithm of both sides, we get ln y = ln xx = x ln x.

∞ ln x
Clearly, x ln x is 0 · ∞ form. We can change it into form −1 and apply l’Hospital’s Rule
∞ x
to get
ln x
lim+ x ln x = lim+
x→0 x→0 x−1
H (ln x)0
= lim+ −1 0
x→0 (x )
−1
x
= lim+
x→0 −x−2
= lim+ (−x) = 0.
x→0

Since y = eln y , by the continuity of the exponential function, we get

lim xx = lim+ eln y = e0 = 1.

x→0+ x→0
3.4 L’Hospital’s Rule 139

3.53 Example: ∞0 form

Find the limit lim x1/x .

x→+∞

ln x
Solution Let y = x1/x . If taking logarithm of both sides, we get ln y = ln x1/x = .
x
ln x ∞
Clearly, is form. Applying l’Hospital’s Rule gives
x ∞
ln x H (ln x)0
lim = lim
x→+∞ x x→+∞ (x)0

x−1
= lim = 0.
x→+∞ 1

Since y = eln y , by the continuity of the exponential function, we get

lim x1/x = lim eln y = e0 = 1.

x→+∞ x→+∞

3.54 Example: 1∞ form

x
1

Find the limit lim 1− .
x→∞ x
x
1 1
  
Solution Let y = 1 − . If taking logarithm of both sides, we get ln y = x ln 1 − .
x x
0
Clearly, it is 0 · ∞ form. We can change it into form
0
1
 
ln 1 −
x
−1
.
x
If we let z = x−1 , then x → ∞ ⇐⇒ z → 0. Thus, by applying l’Hospital’s Rule, we get

1
 
ln 1 −
x ln(1 − z)
lim = lim
x→∞ x−1 z→0 z
0
H [ln(1 − z)]
= lim
z→0 (z) 0

−1
= lim = −1.
z→0 1 − z

Since y = eln y , by the continuity of the exponential function, we get

x
1

lim 1 − = lim eln y = e−1 .
x→∞ x x→∞
140 Applications of Differentiation

3.55 MATLAB: indeterminate forms

We use MATLAB to calculate the following limits:

ln x • lim+ xx ;
• lim ;
x→1 x − 1 x→0

ex
• lim ; • lim x1/x ;
x→∞ x x→+∞
• lim+ (sin x) ln x;
x→0 x
1

√ √
• lim ( x + 1 − x); • lim 1 − .
x→+∞ x→∞ x

>> syms x ;
>> f = log ( x ) /( x -1) ;
>> limit (f ,x ,1)

ans =

>> f = exp ( x ) / x ;
>> limit (f ,x , inf )

ans =

Inf

>> f = sin ( x ) * log ( x ) ;

>> limit (f ,x ,0 , ' right ')

ans =

>> f = sqrt ( x + 1) - sqrt ( x ) ;

>> limit (f ,x , inf )

ans =

>> f = x ^ x ;
>> limit (f ,x ,0 , ' right ')

ans =
3.5 Curve Sketching 141

>> f = x ^(1/ x ) ;
>> limit (f ,x , inf )

ans =

>> f = (1 - 1/ x ) ^ x ;
>> limit (f ,x , inf )

ans =

exp ( -1)

3.5 Curve Sketching

3.56 Definition: symmetry

(A) Suppose the domain D of f is symmetric about the origin.

• f is even if f (−x) = f (x) for all x in D.
• f is odd if f (−x) = −f (x) for all x in D.
(B) A function f is said to be periodic if f (x+p) = f (x) for every number x in its domain,
where p is a positive constant. The smallest such number p is called the period.

3.57 Example: even and odd functions

1. f (x) = x2 , g(x) = cos x, and h(x) = 3x4 − 2x2 − 1

2 are even functions.

2 y y

1 y = cos x 1 1

x x x
−1 O 1 − 21 π O 1
2
π −1 O 1
y=x 2 −1
−1
y = 3x4 − 2x2 − 1
2
142 Applications of Differentiation

Suppose (a, b) is a point on the graph of an even function y = F (x). Since

F (−a) = F (a) = b,

we know that (−a, b) is also on the graph of F . This means that the graph of every
even function is symmetric with respect to the y-axis.
2. f (x) = x3 , g(x) = cos x, and h(x) = −x5 + 2x3 − x are odd functions.

y y
y
1 1
y = x3 y = sin x 1
x x
−1 O 1 x −1 O 1
− 21 π O 1
2
π
−1 −1
−1
y = −x5 + 2x3 − x

Suppose (a, b) is a point on the graph of an odd function y = F (x). Since

F (−a) = −F (a) = −b,

we know that (−a, −b) is also on the graph of F . This means that the graph of every
odd function is symmetric with respect to the origin.

3.58 Method: guidelines for curve sketching

By following the guidelines, you are able to make a sketch that displays the most important
aspects of a given function y = f (x).
A. Domain: Domain D is the set of values of x for which f is defined.
B. Intercepts:
The y-intercept is f (0). The x-intercepts are the real solutions of the equation f (x) = 0.
C. Symmetry:
(i) If f (−x) = f (x) for all x ∈ D, then f is an even function and the curve is symmetric
about the y-axis.
(ii) If f (−x) = −f (x) for all x ∈ D, then f is an odd function and the curve is symmetric
about the origin.
(iii) If f (x + p) = f (x) for all x ∈ D, where p is a positive constant, then f is a
periodic function and the smallest such number p is the period.

D. Asymptotes:
(i) The line y = L is a horizontal asymptote of the curve y = f (x) if either

lim f (x) = L or lim f (x) = L.

x→+∞ x→−∞
3.5 Curve Sketching 143

(ii) The line x = a is a vertical asymptote of the curve y = f (x) if at least one of the
following statements is true:

lim f (x) = +∞, lim f (x) = +∞, lim f (x) = +∞,

x→a x→a− x→a+
lim f (x) = −∞, lim f (x) = −∞, lim f (x) = −∞.
x→a x→a− x→a+

E. Intervals of Increase and Decrease:

Apply the Increasing/Decreasing Test:
(a) If f 0 (x) > 0 on an interval I, then f is increasing on I.
(b) If f 0 (x) < 0 on an interval I, then f is decreasing on I.

F. Local Maximum and Minimum Values:

First find the critical points, the numbers c where f 0 (c) = 0 or f 0 (c) does not exist. Then
apply the First Derivative Test or the Second Derivative Test to determine whether f (c) is a
local maximum or minimum value.
(i) First Derivative Test: Suppose that c is a critical number of a continuous function f .
(a) If f 0 changes from positive to negative at c, then f has a local maximum at c.
(b) If f 0 changes from negative to positive at c, then f has a local minimum at c.
(c) If f 0 does not change sign at c, then f has no local maximum or minimum at c.
(ii) Second Derivative Test: Suppose f 00 is continuous near c.
(a) If f 0 (c) = 0 and f 00 (c) > 0, then f has a local minimum at c.
(b) If f 0 (c) = 0 and f 00 (c) < 0, then f has a local maximum at c.

G. Concavity and Points of Inflection:

(i) Apply the Concavity Test:
(a) If f 00 (x) > 0 on an interval, then the graph of f is concave upward on that interval.
(b) If f 00 (x) < 0 on an interval, then the graph of f is concave downward on that
interval.
(ii) Inflection Points: At these points, the curve changes the direction of concavity.

H. Sketch the Curve: Sketch the curve y = f (x) by using the information in items A-G.

3.59 Example: curve sketching

Sketch the function f (x) = 2x3 − 3x2 − 12x + 1.

A. Domain: The function f is defined at every point, so that the domain is the whole
real line (−∞, +∞).
B. Intercepts: The y-intercept is f (0) = 1. The x-intercepts are the real solutions of the
equation f (x) = 0, that is,
2x3 − 3x2 − 12x + 1 = 0.
144 Applications of Differentiation

The exact values of solutions can be approximated by using the Intermediate Value Theorem.
In fact, we can obtain the following bounds for the three roots of f :

 f (−2) = −3, f (−1.5) = 5.5 =⇒ −2 < x1 < −1.5;



f (0) = 1, f (0.5) = −5.5 =⇒ 0 < x2 < 0.5;

f (3) = −8, f (3.5) = 8 =⇒ 3 < x3 < 3.5.


C. Symmetry: The function f is neither even nor odd, and it is not a periodic function
either.
D. Asymptotes: The function f is defined everywhere and thus has no vertical asymptote.
Since
3 12 1
 
lim (2x3 − 3x2 − 12x + 1) = lim x3 2 − − 2 + 3 = ±∞,
x→±∞ x→±∞ x x x
the function f does not have any horizontal asymptote either.
E. Intervals of Increase or Decrease: Since

f 0 (x) = 6x2 − 6x − 12 = 6(x2 − x − 2) = 6(x + 1)(x − 2),

we see that the derivative f 0 has two simple roots x = −1 and x = 2. These two points
divide the interval (−∞, +∞) into three open intervals (−∞, −1), (−1, 2), and (2, +∞). We
solve the inequalities f 0 (x) ≷ 0 by using the number line and know that f 0 is positive on
(−∞, −1), negative in (−1, 2), and positive in (2, +∞). Thus, by the Monotone Test, the
function f is increasing over the intervals (−∞, −1) and (2, +∞), and decreasing over the
interval (−1, 2).
F. Local Maximum and Minimum Values: The roots of f 0 , x = −1 and x = 2, are
critical points of f . The second derivative

f 00 (x) = 12x − 6

gives f 00 (−1) < 0 and f 00 (2) > 0. By the Second Derivative Test, f has a local maximum
value f (−1) = 8 and a local minimum value f (−1) = −19.
G. Concavity and Inflection Points: Since f 00 (x) < 0 when x < 21 , by the Concavity
Test, f is concave down on (−∞, 12 ). Similarly, since f 00 (x) > 0 when x > 21 , f is concave up
on ( 12 , +∞). Thus, x = 12 is an inflection point, with f ( 12 ) = − 11
2 .
H. Sketch the Curve: Use the results obtained in items A-G, we can sketch the graph
of f (x) = 2x3 − 3x2 − 12x + 1.

x
−1 O 2

y = 2x3 − 3x2 − 12x + 1

−19
3.5 Curve Sketching 145

3.60 Example: curve sketching

x
Sketch the function f (x) = .
x2 + 1
A. Domain: The function f is defined at every point, so that the domain is the whole
real line (−∞, +∞).
B. Intercepts: The y-intercept is f (0) = 0. The x-intercept(s) is the real solutions of the
equation f (x) = 0, that is,
x
= 0,
x2 + 1
which gives the unique solution x = 0. Thus, the curve passes through the x-axis at the
origin.
C. Symmetry: Since
−x x
f (−x) = =− 2 = −f (x),
(−x) + 1
2 x +1
the function f is odd, so the graph is symmetric about the origin.
The function f is not a periodic function.
D. Asymptotes: The function f is defined everywhere and thus has no vertical asymptote.
Since
x 1
lim = lim = 0,
x→±∞ x2 + 1 x→±∞ 1
x+
x
the function f has the x-axis as its horizontal asymptote.
E. Intervals of Increase or Decrease, and
F. Local Maximum and Minimum Values: Since
(x + 1)(x − 1)
f 0 (x) = − ,
(x2 + 1)2
we see that the derivative f 0 has two simple roots x = −1 and x = 1. These two points
divide the interval (−∞, +∞) into three open intervals (−∞, −1), (−1, 1), and (1, +∞). We
solve the inequalities f 0 (x) ≷ 0 by using the number line, we know that f 0 is negative on
(−∞, −1), positive in (−1, 1), and negative in (1, +∞). Thus, by the Monotone Test and the
First Derivative Test, we can have the following table:
interval sign of f 0 monotonicity
(−∞, −1) − &
−1 0 local min
(−1, 1) + %
1 0 local max
(1, +∞) − &
This table gives the intervals in which the function is monotone and the locations of the local
extrema. In fact, the local minimum value is f (−1) = − 12 and the local maximum value is
f (1) = 12 .
G. Concavity and Inflection Points: The second derivative f 00 is
2x(x2 − 3)
f 00 (x) = .
(x2 + 1)3
146 Applications of Differentiation

√ √
It has three simple roots x = − 3, x = 0, and x √= 3. √These three √ points divide
√ the
interval (−∞, +∞) into four open intervals (−∞, − 3), (− 3, 0), (0, 3), and ( 3, +∞).
We solve the
√ inequalities f (x)
00
√ ≷ 0 by using the number
√ line and know √that f 00 is negative
in (−∞, − 3), positive in (− 3, 0), negative in (0, 3), and positive in ( 3, +∞). Thus, by
the Concavity Test, we can have the following table:

interval sign of f 00 concavity

√
(−∞,√− 3) − _
−√ 3 0 inflection
(− 3, 0] + ^
0√ 0 inflection
(0,√ 3) − _
√ 3 0 inflection
( 3, +∞) + ^

This table gives the intervals in which the function is concave up/down and √ the locations of
the inflection points. In fact, the inflection points are at x = 0 and x = ± 3.
H. Sketch the Curve: Use the results obtained in items A-G, we can sketch the graph
x
of f (x) = 2 .
x +1

1
x
y=
x2 + 1
x
−1 O 1

−1

3.61 Example: curve sketching

x3
Sketch the function f (x) = .
x2 − 1
x3
A. Domain: The function f (x) = is defined everywhere except ±1, so that the
x2 − 1
domain of f is (−∞, −1) ∪ (−1, 1) ∪ (1, +∞).
B. Intercepts: The y-intercept is f (0) = 0. The x-intercept(s) is the real solutions of the
equation f (x) = 0, that is,
x3
= 0,
x −1
2

which gives the unique solution x = 0. Thus, the curve passes through the x-axis at the
origin.
3.5 Curve Sketching 147

C. Symmetry: Since
(−x)3 x3
f (−x) = = − = −f (x),
(−x)2 − 1 x2 − 1
the function f is odd, so the graph is symmetric about the origin.
The function f is not a periodic function.
D. Asymptotes: Since
x3 x3
lim = lim = −∞,
x→−1− x2 − 1 x→1− x2 − 1
x3 x3
lim = lim = +∞,
x→−1+ x2 − 1 x→1+ x2 − 1
Moreover, by long division, we have
x3 x
=x+ 2 ,
x2 −1 x −1
so that
x3
 
x
lim − x = lim 2 .=0
x→∞ x2 − 1 x→∞ x − 1

This shows that y = x is a slant asymptote along both infinity directions, i.e., the graph of
f (x) gets closer to the line y = x when x → ∞. We also see that the graph of f is above the
asymptote for x > 0 and below the asymptote for x < 0.
E. Intervals of Increase or Decrease, and
F. Local Maximum and Minimum Values: The derivative f 0 is
x2 (x2 − 3)
f 0 (x) = .
(x2 − 1)2
Its denominator
√ has two
√ double roots at x = ±1, while the numerator has two simple roots
x = − 3 and x = 3, and a double root √ x = 0. √These five points divide the √ inter-
val
√ (−∞, +∞) into six open intervals (−∞, − 3), (− 3, −1), (−1, 0), (0, 1), (1, 3), and
( 3, +∞). We solve the inequalities
√ f 0
(x) ≷ 0 by using the number line and know that f 0
changes its signs only at x = ± 3. By the Monotone Test and the First Derivative Test, we
can have the following table:
interval sign of f 0 monotonicity
√
(−∞,√ − 3) + %
− 3
√ 0 local max
(− 3, −1) − &
−1 not exist
(−1, 0) − &
0 0 no extreme
(0, 1) − &
1√ not exist
(1,√ 3) − &
√ 3 0 local min
( 3, +∞) + %
148 Applications of Differentiation

This table gives the intervals in which the function is monotone and the locations of the local
√ √
extrema. In fact, the local maximum value is f (− 3) = − 3 2 3 and the local minimum value
√ √
is f ( 3) = 3 2 3 .
G. Concavity and Inflection Points: The second derivative f 00 is

2x(x2 + 3)
f 00 (x) = .
(x2 − 1)3

Its denominator has two triple roots at x = ±1, while the numerator has one simple root
x = 0. These three points divide the interval (−∞, +∞) into four open intervals (−∞, −1),
(−1, 0), (0, 1), and (1, +∞). We solve the inequalities f 00 (x) ≷ 0 by using the number line
and know that f 00 changes its signs at x = ±1 and x = 0. By the Concavity Test, we can
have the following table:

interval sign of f 00 concavity

(−∞, −1) − _
−1 not exist
(−1, 0) + ^
0 0 inflection
(0, 1) − _
1 not exist
(1, +∞) + ^

The function has one inflection point at x = 0, with f (0) = 0. The other two points x = ±1
are not inflection points, since they are not in the domain of f .
H. Sketch the Curve: Use the results obtained in items A-G, we can sketch the graph
x3
of f (x) = 2 .
x −1

x3
y=
x2 − 1 1

x
−1 O 1
−1

3.62 MATLAB: curve sketching

There are several commands to plot figures in MATLAB.

x
1. The following example shows the plots of f (x) = 2x3 − 3x2 − 12x + 1 and g(x) =
x2 +1
3.5 Curve Sketching 149

in the same figure.

syms x ;
f ( x ) = 2* x ^3 - 3* x ^2 - 12* x + 1;
g ( x ) = x /( x ^2 + 1) ;

x_array = -3:0.05:3; % x - axis ranges from -3 to 3 with ...

step size 0.05
subplot (2 ,1 ,1) % create 2 rows and 1 column of graphs ...
and place the current figure in the 1 st position
plot ( x_array , f ( x_array ) , 'r ')
xlabel ( 'x ')
ylabel ( 'f ( x ) ')
grid

subplot (2 ,1 ,2)
x_array2 = linspace ( -3 ,3 ,121) ; % a vector with 121 ...
evenly spaced points from -3 to 3
plot ( x_array2 , g ( x_array2 ) , 'b ')
xlabel ( 'x ')
ylabel ( 'g ( x ) ')
grid

Note that we use different commands to generate vectors for x-axis, but the vectors
generated, i.e., x_array and x_array2, are the same.
The figure plotting f and g is shown below.

0
f(x)

-20

-40
-3 -2 -1 0 1 2 3
x
0.5
g(x)

-0.5
-3 -2 -1 0 1 2 3
x

x3
2. The following example shows the plot of f (x) = using the command fplot.
x2 − 1

syms x ;
f ( x ) = x ^3/( x ^2 - 1) ;

fplot ( f ( x ) )
150 Applications of Differentiation

xlabel ( 'x ')

ylabel ( 'f ( x ) ')
grid

where fplot(f(x)) plots the figure of f on the default interval [−5, 5] along the x-axis.
The figure of f on [−5, 5] is shown below.

f(x) 0

-2

-4

-6
-5 -4 -3 -2 -1 0 1 2 3 4 5
x

We can see from the figure above that x = −1 and x = 1 are vertical asymptotes of the
curve y = f (x).
Note that we can input the expression of f into fplot directly without the need of
defining a symbolic function beforehand. In this way, f should be input in the form of
an anonymous function. Moreover, we can also specify the interval along the x-axis for
the figure to be plotted. The following example shows this alternative way of using the
command fplot.

fplot (@( x ) x .^3./( x .^2 - 1) ,[ -4 ,4]) % 1 st input : ...

anonymous function ; 2 nd input : x - axis ranges from -4 ...
to 4
xlabel ( 'x ')
ylabel ( 'f ( x ) ')
grid

2
f(x)

-2

-4

-6

-4 -3 -2 -1 0 1 2 3 4
x

The figure above shows f on the interval [−4, 4] (except at ±1).

3.6 Parametric Curves 151

3.6 Parametric Curves

3.63 Definition: parametric equations and parametric curve

Suppose that x and y are both given as functions of a third variable t (called a parameter)
by the so-called parametric equations:

x = f (t), y = g(t), a ≤ t ≤ b.

As t varies, the point (x, y) = (f (t), g(t)) varies and traces out a curve, which we call a
parametric curve. The points (f (a), g(a)) and (f (b), g(b)) are called the initial point and
terminal point of the curve.

3.64 Example: parametric equations and parametric curve

The parametric equations

x = cos t, y = sin t, 0 ≤ t ≤ 2π

describe the unit circle. In fact, for each value of t, the corresponding ordered pair (x, y) is
a point on the circle centered at (0, 0) with radius 1. To be more explicit, for some special
values of t, we can generate the following table and the corresponding points (x, y) which are
on the circle.

t (x, y)

0 (1, 0) t= 2 , (0, 1)
π

√ √ √ √
√ √ t= 4 , (− 2 , 2 )
3π 2 2
t= 4,( 2 , 2 )
π 2 2
π
4 ( 2 , 2 )
2 2

π
2 (0, 1) t = π, (−1, 0) t = 0, (1, 0)
√ √ t = 2π, (1, 0)
x
O
3π
4 (− 2 , 2 )
2 2

π (−1, 0)
3π
2 (0, −1) t= 2 , (0, −1)
3π

2π (1, 0)

From the figure, we see that as t increases from the initial value t = 0 to the final value
of t = 2π, the curve is generated from the initial point (1, 0) to the final point (1, 0). The
parametric equations trace out the unit circle in the counterclockwise direction. The initial
point and the final point coincide at the same location.
152 Applications of Differentiation

3.65 Example: parametric equations of a line

The parametric equations

x = x0 + at, y = y0 + bt, −∞ < t < +∞,

b
where x0 , y0 , a, and b are constants, describ a line with slope , if a 6= 0. If a = 0 and b 6= 0,
a
the line is vertical.

3.66 Example: parametric equations of cycloid

Suppose a point P starting at the origin on the circumference of a circle of radius r traces
out a curve as the circle rolls along the x-axis. Show that the parametric equations of the
curve are
x = r(t − sin t), y = r(1 − cos t),
where the parameter t is the angle of the rotation of the circle with t = 0 when P is at the
origin.

Solution We take t, the angle of the rotation of the circle, as the parameter. Suppose the
circle has rotated through t radians. From the figure, we see that the distance it has rolled
from the origin is
|OT | = arc P T = rt,
so that the coordinates of the center of the circle is C(rt, r). Let the coordinates of P be
(x, y). Then

x = |OT | − |P Q| = rt − r sin t = r(t − sin t),

y = |T C| − |QC| = r − r cos t = r(1 − cos t).

These give the parametric equations of the cycloid.

r C(rt, r)
P Q

y
x
x
O T 2πr
rt

3.67 Remark: non-uniqueness of parametrization

A curve can have more than one parametrizations. For example, it is easy to verify that
3.6 Parametric Curves 153

the unit circle x2 + y 2 = 1 admits the following parametrizations:

(1) x = cos t, y = sin t, 0 ≤ t < 2π;

(2) x = sin t, y = cos t, π
2 <t≤ 5π
2 .

3.68 Theorem: tangent to parametric curve

Suppose a parametric curve is given by the parametric equations

x = f (t), y = g(t).

If f and g are differentiable functions, and if f 0 (t) 6= 0, then the chain rule gives the slope of
the tangent line at (x, y) = (f (t), g(t)):

dy dy/dt g 0 (t)
= = 0 .
dx dx/dt f (t)

Theorem: horizontal tangent

The parametric curve has a horizontal tangent at (x, y) = (f (t), g(t)) if

dy dx
= g 0 (t) = 0, = f 0 (t) 6= 0.
dt dt

Theorem: vertical tangent

The parametric curve has a vertical tangent at (x, y) = (f (t), g(t)) if

dx dy
= f 0 (t) = 0, = g 0 (t) 6= 0.
dt dt

3.69 MATLAB: parametric curves

In MATLAB, parametric curves can be plotted using the command plot. The following
examples show the plot of two parametric curves.
1. The parametric equations are
x = cos t, y = sin t, 0 ≤ t ≤ 2π
The parametric curve is plotted using the following code:

>> t = linspace (0 ,2* pi ,200) ;

>> x = cos ( t ) ;
>> y = sin ( t ) ;
>> figure
154 Applications of Differentiation

>> plot (x ,y , 'r ')

>> grid
>> xlabel ( 'x ')
>> ylabel ( 'y ')

The parametric curve is shown below.

0.8

0.6

0.4

0.2

0
y

-0.2

-0.4

-0.6

-0.8

-1
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
x

2. The parametric equations are

x = t − sin t, y = 1 − cos t, 0 ≤ t ≤ 4π

The parametric curve is plotted using the following code:

>> t = linspace (0 ,4* pi ,200) ;

>> x = t - sin ( t ) ;
>> y = 1 - cos ( t ) ;
>> figure
>> plot (x ,y , 'b ')
>> grid
>> xlabel ( 'x ')
>> ylabel ( 'y ')

The parametric curve is shown below.

1.8

1.6

1.4

1.2

1
y

0.8

0.6

0.4

0.2

0
0 2 4 6 8 10 12 14
x
3.6 Parametric Curves 155

Polar Coordinates and Polar Curves

3.70 Definition: Cartesian coordinates and polar coordinates

A point (x, y) in Cartesian coordinates can y

be expressed in polar coordinates by P (r, θ) = P (x, y)

x = r cos θ, y = r sin θ.
y
Conversely, we can express r and θ in terms
of x and y: θ
x x
O
y
r =x +y ,
2 2 2
tan θ = .
x

3.71 Definition: polar curve

The graph of a polar equation r = f (θ), or more generally F (r, θ) = 0, consists of all points
P that have at least one polar representation (r, θ) whose coordinates satisfy the equation.

3.72 Example: polar curve - Archimedean spiral

The general Archimedean spiral is a polar curve described by the equation

r = a + bθ, θ ≥ 0,

where a and b are real numbers. Sketch the curve with a = 2 and b = 1.
Solution For θ ≥ 0, in Cartesian coordinates, the equation r = 2 + θ gives a line of slope
1 in the first quadrant, starting from the point (0, 2), as in the following left figure. As θ
increases from 0, it shows that r increases monotonically from 2. It gives a spiral curve, as
shown in the following right figure.

105◦ 90◦ 75◦

120◦ 60◦

135◦ 45◦
r
150◦ 30◦

◦
165 15◦

180◦ 0◦
2 4 6 8
r =2+θ 195◦ 345◦

◦
210 330◦
2
◦ ◦
225 315

O 1
θ 240◦ 300◦
255◦ 270 ◦ 285◦

3.73 Example: polar curve - three-leaved rose

Sketch the curve with the given polar equation

r = 2 sin 3θ.
156 Applications of Differentiation

Solution The function is periodic of period 23 π, and a full revolution is 2π, so the least
common multiple is 2π. Thus, we sketch r = 2 sin 3θ, 0 ≤ θ ≤ 2π, in Cartesian coordinates
in the following left figure.
(1) As θ increases from 0 to π6 , r increases from 0 to 2; as θ increases from π6 to π3 , r
decreases from 2 to 0. Since r is positive, the curve is in the first quadrant.
(2) As θ increases from π3 to π2 , r decreases further from 0 to −2; as θ increases from π2 to
3 , r increases from −2 to 0. Since r is negative, the curve is in the lower-half plane.
2π

(3) As θ increases from 2π 3 to 6 , r increases from 0 to 2; as θ increases from 6 to π, r

5π 5π

decreases from 2 to 0. Since r is positive, the curve is in the second quadrant.

Thus, as θ varies from 0 to π, the curve traces a three-leaved rose.
(4) As θ increases from π to 2π, since r = −2 sin(3θ − π), we know that the curve traces
the same three-leaved rose once more.
105◦ 90◦ 75◦
r 120◦ 60◦

135◦ 45◦
r = 2 sin 3θ ◦
2 150 30◦
Ä Á
165◦ 15◦
À Á Ä Å
1
3π
2
3π
π 2π Å À
θ 180◦ 0◦
O 1 2

 à 195 ◦
345◦
 Ã
−2 210◦ 330◦

225◦ 315◦

240◦ 300◦
255◦ 270◦ 285◦

3.74 Example: polar curve sketching

Sketch the curve with the given polar equation

r = a + b cos θ,

where (A) a = b = 4; (B) a = 3 and b = 5.

Solution For any given a and b, the function r = a + b cos θ is periodic of period 2π. Since
a full revolution is 2π, we only need to sketch the polar curve for 0 ≤ θ ≤ 2π.
(A) For 0 ≤ θ ≤ 2π, we sketch r = 4 + 4 cos θ in Cartesian coordinates in the following left
figure. As θ increases from 0 to π, it shows that r decreases from 4 to 0. Since r is positive,
the curve is in the upper-half plane. As θ increases from π to 2π, r increases from 0 to 4.
Again since r is positive, the curve is in the lower-half plane. Thus, as θ varies from 0 to 2π,
the curve completes one loop, as shown in the following right figure. This is a cardioid.
3.6 Parametric Curves 157

105◦ 90◦ 75◦

r 120◦ 60◦

135◦ 45◦

150◦ 30◦
r = 4 + 4 cos θ À
◦
165 15◦
8
180◦ 0◦
2 4 6 8
À Á 195◦ 345◦
4
Á
◦
210 330◦

◦ ◦
225 315

O π 2π
θ 240◦ 300◦
◦ ◦
255 270◦ 285

(B) For 0 ≤ θ ≤ 2π, we sketch r = 3 + 5 cos θ in Cartesian coordinates in the following left
figure.
(1) As θ increases from 0 to cos−1 (− 35 ) ≈ 2.2143 (rad) ≈ 126.87◦ , r decreases from 8 to 0.
Since r is positive, the curve is in the upper-half plane.
(2) As θ increases from cos−1 (− 53 ) to π, r decreases further from 0 to −3. Since r is
negative, the curve is in the fourth quadrant.
(3) As θ increases from π to 2π − cos−1 (− 53 ) ≈ 4.0689 (rad) ≈ 233.13◦ , r increases from
−2 to 0. Since r is negative, the curve is in the first quadrant.
In fact, when θ varies from cos−1 (− 53 ) to 2π − cos−1 (− 53 ), the curve completes the small
loop.
(4) As θ increases from 2π − cos−1 (− 53 ) to 2π, r increases from 0 to 8, and the curve is in
the lower-half plane.
The curve also completes a large loop, as shown in the following right figure. We get a
limaçon.

r
105◦ 90◦ 75◦
120◦ 60◦

135◦ 45◦
r = 3 + 5 cos θ
150◦ −1
(− 35 )
À 30◦
8 θ = cos

◦
165 15◦
Â
À Ã 180◦ 0◦
4 2 4 6 8
Á
195◦ 345◦

◦
θ = 2π − cos−1 (− 35 )
210 Ã 330◦
O π 2π
θ
Á Â 225◦ 315◦
−2
240◦ 300◦
255◦ 270◦ 285◦

3.75 Method: determine symmetry of a polar curve

(a) If a polar equation is unchanged when θ

is replaced by −θ, the curve is symmetric (r, θ)
about the polar axis. θ=0
O
(r, −θ)
158 Applications of Differentiation

(b) If the equation is unchanged when r is re-

placed by −r, or when θ is replaced by (r, θ)
θ + π, the curve is symmetric about the θ=0
pole. (−r, θ)
O

(c) If the equation is unchanged when θ is re-

(r, π − θ) (r, θ)
placed by π − θ, the curve is symmetric
about the vertical line θ = π2 . θ=0
O

3.76 Theorem: tangent to polar curve

Suppose a polar curve is given by the equation r = f (θ). Then we can write its parametric
equations as
x = r cos θ = f (θ) cos θ, y = r sin θ = f (θ) sin θ.
dx
If f is differentiable, and if = f 0 (θ) cos θ − f (θ) sin θ 6= 0, then the chain rule gives the
dθ
slope of the tangent line at (x, y) = (f (θ) cos θ, f (θ) sin θ):

dy dy/dθ f 0 (θ) sin θ + f (θ) cos θ

= = 0 .
dx dx/dθ f (θ) cos θ − f (θ) sin θ

Theorem: horizontal tangent

The parametric curve has a horizontal tangent at (x, y) = (f (θ) cos θ, f (θ) sin θ) if

dy dx
= f 0 (θ) sin θ + f (θ) cos θ = 0, = f 0 (θ) cos θ − f (θ) sin θ 6= 0.
dθ dθ

Theorem: vertical tangent

The parametric curve has a vertical tangent at (x, y) = (f (θ) cos θ, f (θ) sin θ) if

dx dy
= f 0 (θ) cos θ − f (θ) sin θ = 0, = f 0 (θ) sin θ + f (θ) cos θ 6= 0.
dθ dθ

3.77 Example: highest point of limaçon

Find the highest point of limaçon r = 3 + 5 cos θ.

3.6 Parametric Curves 159

Solution The function r = 3 + 5 cos θ is periodic of period 2π. Since a full revolution is
2π, we only need to determine the highest point for 0 ≤ θ ≤ 2π.
The parametric equations for the limaçon are

x = (3 + 5 cos θ) cos θ, y = (3 + 5 cos θ) sin θ, 0 ≤ θ ≤ 2π.

dy dy dx dx
At the highest point, we must have = · =0· = 0. It is easy to have
dθ dx dθ dθ
dy
= (−5 sin θ) sin θ + (3 + 5 cos θ) cos θ
dθ
= −5(1 − cos2 θ) + (3 + 5 cos θ) cos θ
= 10 cos2 θ + 3 cos θ − 5.

dy
Solving = 0 gives
dθ
√ √
cos θ = 20 (−3
1
+ 209) ≈ 0.5728 or 20 (−3
1
− 209) ≈ −0.8728.

Thus, for 0 < θ < 2π, there are four critical points:
√
θ1 = cos−1 201
(−3 + 209) ≈ 0.9608,


√
θ2 = cos−1 201
(−3 − 209) ≈ 2.6318,


√
θ3 = 2π − cos−1 20 1
(−3 + 209) ≈ 3.6514,


√
θ4 = 2π − cos−1 20 1
(−3 − 209) ≈ 5.3224,

at which the corresponding Cartesian coordinates (x, y) are

(x1 , y1 ) = (3.35926, 4.8067),

(x2 , y2 ) = (1.19074, −0.6657),
(x3 , y3 ) = (1.19074, 0.6657),
(x4 , y4 ) = (3.35926, −4.8067).

The Cartesian coordinates of two endpoints when θ = 0 and θ = 2π coincide at (8, 0). By
comparing the values of the y-coordinates at these five points, we see that the highest point
of the curve attains at (x1 , y1 ), approximately at (3.35926, 4.8067).

105◦ 90◦ 75◦

120◦ 60◦

135◦ 45◦

150◦ 30◦
θ= cos(− 53 )

165◦ 15◦

180◦ 0◦
2 4 6 8
◦
195 345◦

θ = 2π − cos(− 35 )
210◦ 330◦

225◦ 315◦

240◦ 300◦
255◦ 270◦ 285◦
160 Applications of Differentiation

3.78 MATLAB: polar curves

In MATLAB, the command polarplot facilitates the plot of polar curves. Two examples
are provided below to show how to plot polar curves in MATLAB.
1. The polar curve is described by the following equation:
r = 2 sin 3θ.
This polar curve can be plotted using the following code:

>> theta = linspace (0 ,2* pi ,200) ;

>> r = 2* sin (3* theta ) ;
>> figure
>> polarplot ( theta , r )

The polar curve plotted is shown below.

90
120 2 60

1.5
150 30
1

0.5

180 0 0

210 330

240 300
270

2. The polar curve is described by the following equation:

r = 4 + 4 cos θ,
This polar curve can be plotted using the following code:

>> theta = linspace (0 ,2* pi ,200) ;

>> r = 4 + 4* cos ( theta ) ;
>> figure
>> polarplot ( theta ,r , 'r ')

The polar curve plotted is shown below.

90
120 8 60

6
150 30
4

180 0 0

210 330

240 300
270
3.7 Optimization 161

3.7 Optimization
3.79 Example: optimization - largest area

John has 120 meters of fencing to build a rectangular fence along a straight river. He needs
to build three sides, and the side along the river does not need fencing. What are the lengths
of sides of the field to have the largest area?

x x

Solution This is a typical example of optimization problems. Its solution involves a few
steps, shown below.
Step 1: Introduce variables.
Let x be the length of the side of the rectangle perpendicular to the river and y be the
length of the other side.
Step 2: Write down the function to be optimized and the constraints on variables. Simplify
the function under the constraints.
(1) We need to maximize the area of the rectangle, which is A = xy.
(2) There are some constraints for variables x and y. Clearly, we must have 2x + y = 120,
or y = 120 − 2x. We also have 0 ≤ x ≤ 60 and 0 ≤ y ≤ 120.
(3) Substitute y = 120 − 2x into the function we need to optimize, we get
A(x) = x(120 − 2x) = 120x − 2x2 ,
where 0 ≤ x ≤ 60.
Step 3: Solve the optimization problem.
This is a problem of finding maximum for the function A(x) = 120x − 2x2 defined in a
finite closed interval [0, 120].
In fact, the function A is continuous on [0, 120] and differentiable in (0, 120). The derivative
of A is
A0 (x) = 120 − 4x.
Solving A0 (x) = 0 gives only one critical point x = 30 in [0, 120]. We can use one of the
following two approaches to determine the absolute maximum value of A in the interval
[0, 120].
(a) We evaluate the function A at two boundary points x = 0 and x = 120 and the critical
point at x = 30:
A(0) = 0 = A(60), A(30) = 30 × 60 = 1800.
By choosing the largest value, we get the absolute maximum value A(30) = 1800.
(b) It is easy to see that
A (x) > 0, if 0 < x < 30;
 0
A0 (x) < 0, if 30 < x < 120.
162 Applications of Differentiation

Thus, by the First Derivative Test, the function A has a local maximum value at x = 30.
Since A is increasing on (0, 30), and decreasing on (30, 120), the local maximum value is also
the absolute maximum.
Hence, the function A attains the absolute maximum value A(30) = 1800.

3.80 Example: optimization - Snell’s law of refraction

Fermat’s principle says that light travels along the path of shortest traveling time. In the
figure below, we consider a light ray traveling from point A located in one medium to point
B in another medium. The media are separated by the x-axis. Assume that the speed of
light is respectively c1 and c2 in the media. Denote by θ1 the angle of incidence and θ2 the
angle of refraction.

A
θ1
a

d−x
x C
b
θ2
B
d

The total time for light to travel along the path ACB is
T (x) = (time from A to C) + (time from C to B)
√
a2 + x2 b2 + (d − x)2
p
= + .
c1 c2
We shall determine the value of x by applying Fermat’s Principle to minimize the function
T on [0, d].
In fact, the derivative of T is
x d−x
T 0 (x) = √ − p .
c1 a +x
2 2 c2 b + (d − x)2
2

We claim that the equation T 0 (x) = 0 has a unique solution on [0, d]. Indeed, the equation
has a solution by the Intermediate Value Theorem, and
d d
T 0 (0) = − √ < 0, T 0 (d) = √ > 0.
c2 b2 + d2 c1 a2 + d2
The solution is unique since T 0 is increasing on (0, d):
a2 2 b2
T 00 (x) = (a + x2 )−3/2 + (b2 + (d − x)2 )−3/2 > 0.
c1 c2
Let x = x0 be the unique solution of the equation T 0 (x) = 0 on (0, d). The above argument
shows that
T (x) < 0, if 0 < x < x0 ;
 0
T 0 (x) > 0, if x0 < x < d.
3.7 Optimization 163

Thus, by the First Derivative Test, the function T has a local minimum at x = x0 . Since
T is increasing on (x0 , d), and decreasing on (0, x0 ), the local minimum is also the absolute
minimum. Hence, along the path ACB when x = x0 , light travels from A to B in the shortest
time.
When x = x0 , the equation T 0 (x0 ) = 0 gives

x0 d − x0
= .
+ b2 + (d − x0 )2
p p
c1 a2 x20 c2

In terms of the angles of incidence and refraction, we have

sin θ1 sin θ2
=
c1 c2
which gives Snell’s law
sin θ1 c1
= ,
sin θ2 c2

3.81 Example: optimization - shortest distance

Find points on the parabola y = 12 x2 that is closest to the point (4, 1).

Solution The distance from a point (x, y) on the parabola to the point (4, 1) is
q
d = (x − 4)2 + ( 12 x2 − 1)2 .

We need to minimize the function d. It is equivalent to minimize the function f (x) = d2 (x),
which is computationally easier.
In fact, the derivative f 0 is
0
f 0 (x) = (x − 4)2 + ( 12 x2 − 1)2 = 2(x − 4) + 2( 21 x2 − 1)x = x3 − 8.


By solving f 0 (x) = 0, we see that f has only one critical point x = 2. It is easy to see that
f (x) < 0, if x < 2;
 0
f 0 (x) > 0, if x > 2.
Thus, by the First Derivative Test, the function f has a local minimum value at x = 2. Since
f is decreasing on (−∞, 2), and increasing on (2, +∞), the local minimum value is also the
absolute minimum.

Hence, the shortest distance is attained y

when x = 2, with
√
d = (2 − 4)2 + (22 /2 − 1)2 = 5.
p
y = 12 x2

2 (2, 2)

1 (4, 1)
x
O 1 2 4
164 Applications of Differentiation

3.82 MATLAB: optimization

John has 120 meters of fencing to build a rectangular fence along a straight river. He needs
to build three sides, and the side along the river does not need fencing. What are the lengths
of sides of the field to have the largest area?
The following MATLAB code solves the above optimization problem.

>> syms x y S ( x ) ;
>> y = 120 - 2* x ; % the side parallel to the river
>> S ( x ) = x * y ; % area
>> d ( x ) = diff ( S ( x ) ,x )

d(x) =

120 - 4* x

>> rts_d = solve ( d ( x ) == 0 , x )

rts_d =

30

>> x_array = 0:0.5:60; % x < 60

>> figure
>> plot ( x_array , d ( x_array ) , 'r ')
>> xlabel ( 'x ')
>> ylabel ( 'S ' '( x ) ')
>> grid

150

100

50
S'(x)

-50

-100

-150
0 10 20 30 40 50 60
x

We see from the above MATLAB code that S 0 (x) = 0 has one root: x = 30. Moreover,
we note from the above figure of S 0 (x) that S 0 (x) > 0 on (0, 30) and S 0 (x) < 0 on (30, 60).
According to the First Derivative Test, S(x) has a local maximum value at x = 30. Moreover,
since S 0 (x) > 0 on (0, 30), i.e., S(x) is increasing on (0, 30), and S 0 (x) < 0 on (30, 60), i.e.,
S(x) is decreasing on (30, 60), we know that S(x) also has an absolute maximum value at
3.7 Optimization 165

x = 30. Therefore, to have the largest area, the length perpendicular to the river should be
30 meters, and the side parallel to the river should be 120 − 2 × 30 = 60 meters.

3.83 Engineering Example: bending moment

When an external force is applied to a structural element, e.g., beam, and it causes this el-
ement to bend, we say the bending moment is induced. The bending moment is an important
concept in physics and engineering, especially civil and mechanical engineering.
Suppose there is a beam of length L with uniform distributed loading of q per unit length.
At the distance x of the beam, the bending moment M is given by
q qL
M = − x2 + x.
2 2
This is illustrated in the figure below.

x
L

Find the value of x which gives the maximum bending moment and calculate this maximum
bending moment.
First, let f (x) = M = − 2q x2 + qL
2 x (0 ≤ x ≤ L). Then, we have

qL
f 0 (x) = −qx + .
2
It is easy to see that
f 0 (x) > 0, if 0 ≤ x < L2 ;


f 0 (x) < 0, if L2 < x ≤ L.

Thus, by the First Derivative Test, we know that f (x) has a local maximum value at x = L2 .
Moreover, since f 0 (x) > 0 on 0, L2 , i.e., f (x) is increasing on 0, L2 , and f 0 (x) < 0 on L2 , L ,




i.e., S(x) is decreasing on L2 , L , we know that f (x) also has an absolute maximum value at


x = L2 . Hence, at x = L2 , the bending moment is the maximum.

To calculate the corresponding maximum bending moment, we substitute x = L2 into f (x)
and have
 2
qL2
 
L q L qL L
f =− + = .
2 2 2 2 2 8
qL2
The result above shows that the maximum bending moment is 8 .
166 Applications of Differentiation

3.84 Engineering Example: tank volume optimization

One piece of metal is provided to build a water tank. For this purpose, four corners of the
metal should be cut, as illustrated in the figure below.
x

3m

6m

Find the value of x such that the volume of the water tank is maximized and calculate the
maximum volume of the water tank.
First, the volume of the water tank can be calculated as follows:

V (x) = (6 − 2x)(3 − 2x)x

= 4x3 − 18x2 + 18x.

The derivative of V (x) is

V 0 (x) = 12x2 − 36x + 18.

As we know, the roots of an equation ax2 + bx + c = 0 are given by

√
−b ± b2 − 4ac
x= .
2a
We thus can calculate the roots of V 0 (x) = 12x2 − 36x + 18 = 0 as
√
36 ± 362 − 4 × 12 × 18
x=
24
≈ 0.634 or 2.366.

Since 2x < 3, the value of x should satisfy 0 < x < 1.5, which shows that x cannot be 2.336.
We will then check whether x = 0.634 is a maximum point.
The second derivative of V (x) can be easily calculated and shown below:

V 00 (x) = 24x − 36.

Substituting x = 0.634 into V 00 (x), we have

V 00 (0.634) = 15.216 − 36 < 0.

According to the Second Derivative Test, we know that V (x) has a local maximum value at
x = 0.634. To check whether x = 0.634 is also an absolute maximum point, we rewrite V 0 (x)
as

V 0 (x) = 2(x − 0.634)(x − 2.366).

3.7 Optimization 167

Obviously, we know that

V 0 (x) > 0, if 0 < x < 0.634;


V 0 (x) < 0, if 0.634 < x < 1.5,

which means V (x) is increasing on (0, 0.634) and decreasing on (0.634, 1.5). Therefore, x =
0.634 is also an absolute maximum point, i.e., the volume of the water tank is maximized at
x = 0.634 (m). Finally, substituting x = 0.634 into V (x), we have the maximum volume of
the water tank as

V (0.634) = 4 × 0.6343 − 18 × 0.6342 + 18 × 0.634 ≈ 5.196 m3 .

3.85 Engineering Example: entropy optimization

In information theory, the entropy refers to the uncertainty or the disorder of an event. If
an event has more uncertainty, then it carries more information and has more entropy.
For a communication system, if it has two inputs 0 and 1 and they occur with probability
p and 1 − p (0 < p < 1), respectively, then the entropy of the source is given by

H(p) = −p log2 p − (1 − p) log2 (1 − p).

Find the value of p such that the entropy H(p) is maximized and calculate the maximum
entropy.
By applying the product rule and the chain rule, we have

dH(p) 0 0
= − log2 p − p (log2 p) − (−1) log2 (1 − p) − (1 − p) (log2 (1 − p))
dp
1 −1
= − log2 p − p + log2 (1 − p) − (1 − p)
p ln 2 (1 − p) ln 2
= log2 (1 − p) − log2 p.

According to Fermat’s Theorem, we first calculate the value of p such that dH(p)
dp = 0. It

dH(p) 
can be calculated that dp  = 0. Moreover, since log2 x is an increasing function, we
p=0.5
notice that when p < 0.5, dH(p)
dp > 0; while when p > 0.5, dH(p)
dp < 0. Therefore, according to
the first derivative test, the entropy H(p) is maximized at p = 0.5.
Substituting p = 0.5 into the expression of H(p), we have

H(p) = −0.5 × log2 0.5 − (1 − 0.5) × log2 (1 − 0.5)

= − log2 0.5 = 1 (bit).

The maximum entropy is thus 1 bit.

168 Applications of Differentiation

3.8 Newton’s Method

3.86 Method: Newton’s iteration

To find approximations to a solution r of the equation f (x) = 0, we start with an initial

guessing value x0 , which is not equal to r in general. To improve the approximation, consider
the tangent line to the curve y = f (x) at the point (x0 , f (x0 )). An equation of the line is

y − f (x0 ) = f 0 (x0 )(x − x0 ),

whose x-intercept gives the first approximation x1 , obtained by putting y = 0 in the last
equation:

f (x0 )
0 − f (x0 ) = f 0 (x0 )(x1 − x0 ) =⇒ x1 − x0 = −
f 0 (x0 )
f (x0 )
=⇒ x1 = x0 − 0 .
f (x0 )

The formula for x1 is meaningful if f 0 (x0 ) 6= 0.

We repeat this procedure with x0 replaced by the first approximation x1 , using the tangent
line at (x1 , f (x1 )). This gives a second approximation:

f (x1 )
x2 = x1 − .
f 0 (x1 )

If we keep repeating this process, we generate a sequence of approximations

x1 , x2 , x3 , x4 , . . . . In general, if xn is the nth approximation and f 0 (xn ) 6= 0, then the
next approximation is given by

f (xn )
xn+1 = xn − .
f 0 (xn )

This recursive formula is called Newton’s iteration and the sequence {xn } generated is
called Newton’s approximations.

y
(x0 , f (x0 ))

(x1 , f (x1 ))

(x2 , f (x2 ))
r
x
O x3 x2 x1 x0
3.8 Newton’s Method 169

Theorem: convergence of Newton’s approximations

If x0 is chosen to be sufficiently close to a solution r of f (x) = 0, and if f 0 (xn ) 6= 0

for n = 1, 2, 3, . . . , then the sequence {xn } generated by Newton’s iteration converges
to r:
lim xn = r.
n→∞

3.87 Example: Newton’s method

Given f (x) = x3 − 5x + 1. Use six steps of Newton’s method to approximate all three real
roots of f .

1. Choose an initial guessing value for each root.

The function f is a polynomial of degree 3, so it has at most three real roots. Since

 f (−3) < 0, f (−2) > 0;



f (0) > 0, f (1) < 0;

f (2) < 0, f (3) > 0,


by the Intermediate Value Theorem, f must have one root in each of the three intervals
(−3, −2), (0, 1), and (2, 3).
Hence, it is plausible to choose the initial guessing values −2.5, 0.5, 2.5 for Newton’s
iterations for approximating the three roots of f .
2. Write the recursive formula of Newton’s method.
For f (x) = x3 − 5x + 1, we have

f 0 (x) = (x3 − 5x + 1)0 = 3x2 − 5.

Thus, the recursive formula of Newton’s method is

f (xn ) x3n − 5xn + 1 2x3n − 1

xn+1 = xn − = x n − = .
f 0 (xn ) 3x2n − 5 3x2n − 5

3. Perform the iteration.

For the initial guessing value x0 = −2.5,

2x30 − 1 2(−2.5)3 − 1
x1 = = = −2.345455,
3x20 − 5 3(−2.5)2 − 5
2x3 − 1 2(−2.345455)3 − 1
x2 = 12 = = −2.330203,
3x1 − 5 3(−2.345455)2 − 5
..
.
2x34 − 1
x5 = = −2.330059.
3x24 − 5
170 Applications of Differentiation

We list Newton’s approximations of all three roots in the following table.

k xk xk xk
0 −2.500000 0.500000 2.500000
1 −2.345455 0.176471 2.200000
2 −2.330203 0.201568 2.131933
3 −2.330059 0.201640 2.128428
4 −2.330059 0.201640 2.128419
5 −2.330059 0.201640 2.128419

Remark: when to stop for Newton’s iteration

There is a need for a good estimation of the error of the approximation. However,
since the actual value of the root is not know, it is not easy to get the estimation
of the error. The rule of thumb is that if we want n decimal accuracy, we should
compute until two successive approximations agree up to n decimal places.

Remark: how many steps needed

For a given accuracy, the number of steps of Newton’s iteration can vary. In the
example, if the initial guessing value is x0 = 1.3, then Newton’s approximations are

k xk
0 1.300000
1 48.485718
2 32.346600
.. ..
. .
12 2.128441
13 2.128419

It takes 13 steps of Newton’s iterations to achieve 6 decimal accuracy.

3.88 Pitfall: convergence of Newton’s method

In Newtons’s method, f 0 (xn ) appears in the denominator of the formula. When f 0 (xn ) is
zero, the method breaks down. When f 0 (xn ) is close to zero, the method is slow to converge
to the actual root or may fail to converge.
8x2
Consider f (x) = 2 . Then the recursive formula of Newton’s method is
3x + 1
f (xn ) xn (3x2n + 1) xn
xn+1 = xn − = x n − = (1 − 3x2n ).
f 0 (xn ) 2 2

Clearly the function f has 0 as the only real root. For the following three initial guessing
3.8 Newton’s Method 171

values x0 = 1, 0.15, and 1.1, we list the corresponding Newton’s approximations:

k xk xk xk
0 1.000000 0.150000 1.200000
1 −1.000000 0.069938 −1.992000
2 1.000000 0.034456 10.860575
3 −1.000000 0.017166 −1.92 × 103
4 1.000000 0.008576 1.06 × 1010
5 −1.000000 0.004287 −1.76 × 1030
6 1.000000 0.002143 8.21 × 1090

From the computations, we see that the approximations with the initial guessing values
x0 = 1 and x0 = 1.2 fail to converge.

3.89 MATLAB: Newton’s method

The following example shows how to calculate the roots of f (x) = x3 − 5x + 1 in MATLAB.
We note that f is a polynomial of degree 3, so it has at most three real roots. To roughly
estimate these roots, we first plot the figure of f as follows.

>> syms x ;
>> f ( x ) = x ^3 - 5* x + 1;
>> x_array = -3:0.05:3;
>> figure
>> plot ( x_array , f ( x_array ) , 'r ')
>> xlabel ( 'x ')
>> ylabel ( 'f ( x ) ')
>> grid

15

10

5
f(x)

-5

-10

-15
-3 -2 -1 0 1 2 3
x

From the figure above, we see that there are 3 roots of f , and these roots are on intervals
(−3, −2), (0, 1) and (2, 3), respectively. Hence, we choose the initial guessing values −2.5,
0.5 and 2.5 for Newton’s iterations for approximating the three roots of f (x).
Next, we use Newton’s method to approximate these three roots of f . The MATLAB code
is provided below.
172 Applications of Differentiation

syms x ;
f ( x ) = x ^3 - 5* x + 1; % function
d ( x ) = diff ( f ( x ) ,x ) ;

n = 5;
N_max = 10; % maximum no . of iterations
x = zeros (1 , N_max ) ;
x0 = - 2.5; % initial guess
x (1) = x0 ;

for num = 1: N_max

x ( num + 1) = x ( num ) - f ( x ( num ) ) / d ( x ( num ) ) ;
if abs ( x ( num + 1) - x ( num ) ) <10^( -( n + 1) ) % to ensure n ...
decimal accuracy
break ; % accuracy achieved , so exit !
end
end

x_iter = x (1:( num + 1) ) % show iteration results

By inputting different initial guessing values, we can obtain approximations of all three
roots. Moreover, we can have 15-digit approximations by using the command format long.
The Newton’s approximations of all three roots with initial guessing values −2.5, 0.5 and 2.5,
respectively, are listed in the following table.

k xk xk xk
0 −2.500000000000000 0.500000000000000 2.500000000000000
1 −2.345454545454545 0.176470588235294 2.200000000000000
2 −2.330203407719357 0.201568074338339 2.131932773109244
3 −2.330058752527104 0.201639675087802 2.128428202971012
4 −2.330058739567982 0.201639675723405 2.128419063906659
5 2.128419063844577

Note that to ensure the 5 decimal accuracy, it requires the absolute difference of two
successive approximations to be less than 10−6 in the MATLAB code. In this case, the
initial guessing values of −2.5 and 0.5 result in 4 iterations, and the initial guessing value of
2.5 results in 5 iterations.

3.90 Engineering Example: projectile - Newton’s method

Consider the case where a ball is thrown with initial velocity u and an angle θ
(0◦ ≤ θ ≤ 90◦ ) against the ground, as shown in the following figure. Then, at time t, the
1
displacement of the ball in the y-direction is sy = u sin θt − gt2 , where g ≈ 9.8 m/s2 is
2
the standard gravity. Suppose that the initial velocity is 40 m/s and the angle against the
ground is 45◦ . Use Newton’s method to calculate the time t at which the ball is 10 meters
3.8 Newton’s Method 173

above the ground during the falling down stage.

θ
x
O

When u = 40 m/s and θ = 45◦ , the displacement of the ball in the y-direction is
9.8 2 √
f (t) = 40 sin(45◦ )t − t = 20 2t − 4.9t2 .
2
√
Note that 2 ≈ 1.414, we thus have
f (t) ≈ 28.28t − 4.9t2 .
First, we need to find an initial guessing value x0 . We notice that
f (5) ≈ 18.9 > 10;


f (6) ≈ −6.7 < 10,

so, by the Intermediate Value Theorem, one of the roots of f (t) = 10 is in the interval (5, 6).
Moreover, since f (5) > f (6), the root of f (t) = 10 in the interval (5, 6) is the time during
the falling down stage of the ball. Hence, we set the initial guessing value as x0 = 5.5.√
Before performing the iterations, we first define a function g(t) = f (t) − 10 = 20 2t −
4.9t2 − 10. To write the recursive formula of Newton’s method, we need to calculate g 0 (t),
which is shown in the following
√
g 0 (t) = 20 2 − 9.8t.
Then, the recursive formula of Newton’s method can be written as
√
g(tn ) 20 2tn − 4.9t2n − 10
tn+1 = tn − 0 = tn − √ .
g (tn ) 20 2 − 9.8tn
From the initial guessing value t0 = 5.5, we have
g(t0 ) g(5.5)
t1 = t0 − = 5.5 − 0
g 0 (t0 ) g (5.5)
−2.661508
= 5.5 − = 5.396099,
−25.615729
g(t1 ) g(5.396099)
t2 = t1 − 0 = 5.396099 − 0
g (t1 ) g (5.396099)
−0.052906
= 5.396099 − = 5.393948,
−24.597499
g(t2 ) g(5.393948)
t3 = t2 − 0 = 5.393948 − 0
g (t2 ) g (5.393948)
−0.000019
= 5.393948 − = 5.393947.
−24.576419
174 Applications of Differentiation

We see that three steps are required to achieve 5 decimal accuracy. Moreover, since t3 =
5.393947, we can conclude that at t = 5.393947, the ball is in the falling down stage and is
10 meters above the ground.
MATLAB solution: √
We now turn to the use of MATLAB. Let g(t) = f (t) − 10 = 20 2t − 4.9t2 − 10. We will
run the following MATLAB code of the Newton’s method to calculate the result.

syms t ;
g ( t ) = 20* sqrt (2) * t - 4.9* t ^2 - 10;
d ( t ) = diff ( g ( t ) ,t ) ;

n = 5; % n decimal accuracy
N_max = 10; % maximum no . of iterations
x = zeros (1 , N_max ) ;
x0 = 5.5; % initial guess
x (1) = x0 ;

for num = 1: N_max

x ( num + 1) = x ( num ) - g ( x ( num ) ) / d ( x ( num ) ) ;
if abs ( x ( num + 1) - x ( num ) ) <10^( -( n + 1) ) % to ensure n ...
decimal accuracy
break ; % accuracy achieved , so exit !
end
end

x_iter = x (1:( num + 1) ) % show iteration results

Starting from the initial guessing value t0 = 5.5, we can obtain approximations of the root
near 5.5. Note that by using the command format long, we can have 15-digit approxima-
tions. The Newton’s approximations are listed in the following table.

k tk
0 5.500000000000000
1 5.396098675752262
2 5.393948136613312
3 5.393947214525773

We see that 3 iterations are needed to ensure 5 decimal accuracy. At t = 5.393947214525773,

the ball is in the falling down stage and is 10 meters above the ground.

3.91 Engineering Example: water tank design

One piece of metal is provided to build a water tank. For this purpose, four corners of the
metal should be cut, as illustrated in the figure below.
3.8 Newton’s Method 175

3m

6m

By Newton’s method, find the value of x such that the volume of the water tank is 3.5 m3
and the water tank covers the least area on the ground.
First, the volume of the water tank can be calculated as follows:

V (x) = (6 − 2x)(3 − 2x)x

= 4x3 − 18x2 + 18x.

To find the value of x such that the volume of the water tank V (x) is 3.5 m3 , we actually
need to find x which satisfies 4x3 − 18x2 + 18x = 3.5. Let g(x) = 4x3 − 18x2 + 18x − 3.5.
The derivative of g(x) is

g 0 (x) = 12x2 − 36x + 18.

Then, the recursive formula of Newton’s method can be written as

g(xn )
xn+1 = xn −
g 0 (xn )
4x3 − 18x2n + 18xn − 3.5
= xn − n .
12x2n − 36xn + 18

The function g(x) is a polynomial of degree 3, so it has at most three real roots. Since

 g(0) = −3.5 < 0, g(1) = 0.5 > 0;



g(1) = 0.5 > 0, g(1.5) = −3.5 < 0;

g(3) = −3.5 < 0, g(4) = 36.5 > 0,


by the Intermediate Value Theorem, g(x) must have one root in each of the three intervals
(0, 1), (1, 1.5), and (3, 4). Note that 2x < 3, so we have 0 < x < 1.5. Moreover, the area that
the water tank covers on the ground can be written as (6 − 2x)(3 − 2x) which is a decreasing
function of x. Hence, if we are looking for the value of x such that the water tank covers the
least area on the ground, the value of x should be as large as possible, i.e., it is within the
interval (1, 1.5). Therefore, we choose the initial value x0 as 1.25.
Now, we use the recursive formula of Newton’s method to calculate the root of g(x) = 0
176 Applications of Differentiation

within (1, 1.5). From the initial guessing value x0 = 1.25, we have

g(x0 ) g(1.25)
x1 = x0 − = 1.25 − 0
g 0 (x0 ) g (1.25)
−1.3125
= 1.25 − = 1.090909,
−8.25
g(x1 ) g(1.090909)
x2 = x1 − 0 = 1.090909 − 0
g (x1 ) g (1.090909)
−0.092035
= 1.090909 − = 1.077746,
−6.991735
g(x2 ) g(1.077746)
x3 = x2 − 0 = 1.077746 − 0
g (x2 ) g (1.077746)
−0.000863
= 1.077746 − = 1.077620.
−6.860419

We see that three steps are required to achieve 3 decimal accuracy. Moreover, since x3 =
1.077620, we can conclude that the volume of the water tank is 3.5 m3 and the water tank
covers the least area on the ground if x = 1.078 m.

MATLAB solution:
We now turn to the use of MATLAB. Let g(x) = 4x3 − 18x2 + 18x − 3.5. We will run the
following MATLAB code of the Newton’s method to calculate the result.

syms x ;
g ( x ) = 4* x ^3 - 18* x ^2 + 18* x - 3.5;
d ( x ) = diff ( g ( x ) ,x ) ;

n = 3; % n decimal accuracy
N_max = 10; % maximum no . of iterations
x_array = zeros (1 , N_max ) ;
x0 = 1.25; % initial guess
x_array (1) = x0 ;

for num = 1: N_max

x_array ( num + 1) = x_array ( num ) - ...
g ( x_array ( num ) ) / d ( x_array ( num ) ) ;
if abs ( x_array ( num + 1) - x_array ( num ) ) <10^( -( n + 1) ) % ...
to ensure n decimal accuracy
break ; % accuracy achieved , so exit !
end
end

x_iter = x_array (1:( num + 1) ) % show iteration results

3.8 Newton’s Method 177

Starting from the initial guessing value x0 = 1.25, we can obtain approximations of the
root near 1.25. The Newton’s approximations are listed in the following table.

k tk
0 1.250000000000000
1 1.090909090909091
2 1.077745540511498
3 1.077620217678657
4 1.077620206075159

We see that in order to ensure 3 decimal accuracy, 4 iterations are needed if running the
MATLAB code above. At x = 1.077620206075159, i.e., x = 1.078 m, the volume of the water
tank is 3.5 m3 and the water tank covers the least area on the ground.

3.92 Engineering Example: tank optimization - Newton’s method

One piece of metal is provided to build a water tank. For this purpose, four corners of the
metal should be cut, as illustrated in the figure below.
x

3m

6m

By Newton’s method, find the value of x such that the volume of the water tank is maximized.
First, the volume of the water tank can be calculated as follows:
V (x) = (6 − 2x)(3 − 2x)x
= 4x3 − 18x2 + 18x.
The derivative of V (x) is
V 0 (x) = 12x2 − 36x + 18.
To find the value of x such that the volume of the water tank V (x) is maximized, we need
to find x which satisfies V 0 (x) = 12x2 − 36x + 18 = 0, i.e., find the root of V 0 (x) = 12x2 −
36x + 18 = 0. As such, we let f (x) = V 0 (x) = 12x2 − 36x + 18, and the derivative of f (x) is
f 0 (x) = 24x − 36.
Then, the recursive formula of Newton’s method can be written as
f (xn )
xn+1 = xn −
f 0 (xn )
12x2n − 36xn + 18
= xn − .
24xn − 36
178 Applications of Differentiation

Before using Newton’s method to find the value of x, we need to find an initial guessing
value x0 . Hence, we plot the figure of f (x) = V 0 (x), as shown below.

f (x) = V 0 (x)

x
O 1.5

Note that 2x < 3, so we have 0 < x < 1.5. To find the value of x which maximizes V (x),
we need to find the root of V 0 (x) = 0 within the interval (0, 1.5). From the figure of V 0 (x)
above, we notice that there is one root within the interval (0, 1.5) and it is roughly in the
middle of this interval. Thus, we choose the initial guessing value x0 around this root and
set it as 0.75.
Now, we use the recursive formula of Newton’s method to calculate the root of f (x) =
V 0 (x) = 0 within (0, 1.5). From the initial guessing value x0 = 0.75, we have
f (x0 ) f (0.75)
x1 = x0 − = 0.75 − 0
f (x0 )
0 f (0.75)
−2.25
= 0.75 − = 0.625,
−18
f (x1 ) f (0.625)
x2 = x1 − 0 = 0.625 − 0
f (x1 ) f (0.625)
0.1875
= 0.625 − = 0.633928,
−21.0
f (x2 ) f (0.633928)
x3 = x2 − 0 = 0.633928 − 0
f (x2 ) f (0.633928)
0.000968
= 0.633928 − = 0.633974.
−20.785728
We see that three steps are required to achieve 3 decimal accuracy. Moreover, from the
figure of V 0 (x) above, we see that if x is smaller than the root of V 0 (x) within (0, 1.5), then
V 0 (x) > 0, i.e., V 0 (x) is an increasing function; if x is larger than this root, then V 0 (x) < 0,
i.e., V 0 (x) is a decreasing function. So this root is an absolute maximum point. Therefore,
we can conclude that the volume of the water tank is maximized when x = 0.634 (m).

MATLAB solution:
We now turn to the use of MATLAB. Let f (x) = V 0 (x) = 12x2 − 36x + 18. We will run
the following MATLAB code of the Newton’s method to calculate the result.

syms x ;
f ( x ) = 12* x ^2 - 36* x + 18;
d ( x ) = diff ( f ( x ) ,x ) ;
3.8 Newton’s Method 179

n = 3; % n decimal accuracy
N_max = 10; % maximum no . of iterations
x_array = zeros (1 , N_max ) ;
x0 = 0.75; % initial guess
x_array (1) = x0 ;

for num = 1: N_max

x_array ( num + 1) = x_array ( num ) - ...
f ( x_array ( num ) ) / d ( x_array ( num ) ) ;
if abs ( x_array ( num + 1) - x_array ( num ) ) <10^( -( n + 1) ) % ...
to ensure n decimal accuracy
break ; % accuracy achieved , so exit !
end
end

x_iter = x_array (1:( num + 1) ) % show iteration results

Starting from the initial guessing value x0 = 0.75, we can obtain approximations of the
root near 0.75. The Newton’s approximations are listed in the following table.

k tk
0 0.750000000000000
1 0.625000000000000
2 0.633928571428571
3 0.633974594992636

We see that 3 iterations are needed to ensure 3 decimal accuracy. Moreover, as explained in
the first part, the root of V 0 (x) within (0, 1.5) is an absolute maximum point. Therefore, we
can conclude that when x = 0.634 (m), the volume of the water tank is maximized.
Remark:

The Newton’s method is utilized to find the root of f (x) = 0 by iteratively approx-
imating the root from an initial guessing value. Hence, any function whose root is
required to be found can be set to be f (x), e.g., V 0 (x) instead of V (x) is set to be
f (x) in this example.
180 Summary

Summary
Definition: basic concepts

Extrema :
absolute maximum : if f (c) ≥ f (x) for all x in D
absolute minimum : if f (c) ≤ f (x) for all x in D
local maximum : if f (c) ≥ f (x) when x is near c
local minimum : if f (c) ≤ f (x) when x is near c

SpecialPoints :
critical point c : either f 0 (c) = 0 or f 0 (c) does not exist
interior point c of D : if there exists (a, b) such that c ∈ (a, b) ⊂ D
inflection point P : at P , f is continuous and changes concavity

Monotonicity :
increasing : if f (x1 ) < f (x2 ) whenever x1 < x2 in (a, b)
decreasing : if f (x1 ) > f (x2 ) whenever x1 < x2 in (a, b)

Concavity :
concave up : if the graph of f is above all its
tangent lines over (a, b)
concave down : if the graph of f is below all its
tangent lines over (a, b)

Theorem: Fermat’s Theorem

If f has a local maximum or minimum value at an interior point c in the domain D of f ,
then c is a critical point of f .

Theorem: Extreme Value Theorem

Let f be a continuous function defined over a finite closed interval [a, b]. Then f (x) has at
least one absolute maximum point and one absolute minimum point in the interval [a, b].

Method: find absolute extrema

Given a continuous function f over a finite closed interval [a, b].

1. Find the critical points and evaluate the values of f at critical points.
2. Evaluate the values of f at the boundary points x = a and x = b.
3. The largest value and smallest value from steps 1 and 2 are the absolute maximum
 181

value and the absolute minimum value respectively.

Theorem: Mean Value Theorem

Suppose f is a function that satisfies the following:

1. f is continuous on the closed interval [a, b].

2. f is differentiable in the open interval (a, b).

Then there exists at least one number c in (a, b) such that

f (b) − f (a)
f 0 (c) =
b−a
or, equivalently,
f (b) − f (a) = f 0 (c)(b − a).

y y

f (b) f (b)

f (a) f (a)

x x
O a c b O a c1 c2 b

Theorem: zero derivative implies constant function

Suppose f is differentiable in (a, b). If f 0 (x) = 0 for all x in (a, b), then f is a constant
function.

Theorem: Monotone Test

Suppose f is differentiable on an interval I.

1. If f 0 (x) > 0 on I, then f is increasing on I.

2. If f 0 (x) < 0 on I, then f is decreasing on I.

Theorem: First Derivative Test

Suppose that c is a critical point of a continuous function f .

1. If f 0 changes from positive to negative at c, then f has a local maximum value at c.

182 Summary

2. If f 0 changes from negative to positive at c, then f has a local minimum value at c.

3. If f 0 does not change sign at c, then f has no local extreme value at c.

y y

f 0 (x) < 0 f 0 (x) > 0 f 0 (x) > 0 f 0 (x) < 0

c x c x
O O

f 0 changes from positive to negative at c f 0 changes from negative to positive at c

y y

f 0 (x) > 0 f 0 (x) < 0

f 0 (x) > 0 f 0 (x) < 0

c x c x
O O

No maximum or minimum No maximum or minimum

Theorem: Second Derivative Test

Suppose f 00 is continuous near c.

1. If f 0 (c) = 0 and f 00 (c) > 0, then f has a local minimum value at c.

2. If f 0 (c) = 0 and f 00 (c) < 0, then f has a local maximum value at c.

3. If f 0 (c) = 0 and f 00 (c) = 0, it is inconclusive about local extreme of f at c.

Theorem: Concavity Test

Suppose f 00 exists on an interval I.

1. If f 00 (x) > 0 on I, then f is concave up on I.

2. If f 00 (x) < 0 on I, then f is concave down on I.
 183

Theorem: l’Hospital’s Rule

Suppose

1. f and g are differentiable functions;

2. g 0 (x) is not zero on an open interval I containing a except possibly at a;

3. lim f (x) = 0 = lim g(x) or lim f (x) = ∞ = lim g(x);

x→a x→a x→a x→a

f (x)
0
4. lim exists or is ∞.
x→a g 0 (x)

Then
f (x) f 0 (x)
lim = lim 0 .
x→a g(x) x→a g (x)

The rule also holds when limits are one-sided limits or when x → a is replaced by x → ∞.

Method: guidelines for curve sketching

By following the guidelines, you are able to make a sketch that displays the most important
aspects of a given function y = f (x).
A. Domain: Domain D is the set of values of x for which f is defined.
B. Intercepts:
The y-intercept is f (0). The x-intercepts are the real solutions of the equation f (x) = 0.
C. Symmetry:
(i) If f (−x) = f (x) for all x ∈ D, then f is an even function and the curve is symmetric
about the y-axis.
(ii) If f (−x) = −f (x) for all x ∈ D, then f is an odd function and the curve is symmetric
about the origin.
(iii) If f (x + p) = f (x) for all x ∈ D, where p is a positive constant, then f is a
periodic function and the smallest such number p is the period.

D. Asymptotes:
(i) The line y = L is a horizontal asymptote of the curve y = f (x) if either

lim f (x) = L or lim f (x) = L.

x→∞ x→−∞

(ii) The line x = a is a vertical asymptote of the curve y = f (x) if at least one of the
following statements is true:

lim f (x) = ∞, lim f (x) = ∞, lim f (x) = ∞,

x→a x→a− x→a+
lim f (x) = −∞, lim f (x) = −∞, lim f (x) = −∞.
x→a x→a− x→a+

E. Intervals of Increase or Decrease:

Apply the Increasing/Decreasing Test:
184 Summary

(a) If f 0 (x) > 0 on an interval I, then f is increasing on I.

(b) If f 0 (x) < 0 on an interval I, then f is decreasing on I.

F. Local Maximum and Minimum Values:

First find the critical points, the numbers c where f 0 (c) = 0 or f 0 (c) does not exist. Then
apply the First Derivative Test or the Second Derivative Test to determine whether f (c) is a
local maximum or minimum value.
(i) First Derivative Test: Suppose that c is a critical number of a continuous function f .
(a) If f 0 changes from positive to negative at c, then f has a local maximum at c.
(b) If f 0 changes from negative to positive at c, then f has a local minimum at c.
(c) If f 0 does not change sign at c, then f has no local maximum or minimum at c.
(ii) Second Derivative Test: Suppose f 00 is continuous near c.
(a) If f 0 (c) = 0 and f 00 (c) > 0, then f has a local minimum at c.
(b) If f 0 (c) = 0 and f 00 (c) < 0, then f has a local maximum at c.

G. Concavity and Points of Inflection:

(i) Apply the Concavity Test:
(a) If f 00 (x) > 0 on an interval, then the graph of f is concave upward on that interval.
(b) If f 00 (x) < 0 on an interval, then the graph of f is concave downward on that
interval.
(ii) Inflection Points: At these points, the curve changes the direction of concavity.

H. Sketch the Curve: Sketch the curve y = f (x) by using the information in items A-G.

Theorem: convergence of Newton’s approximations

If x0 is chosen to be sufficiently close to a solution r of f (x) = 0, and if f 0 (xn ) 6= 0 for

n = 1, 2, 3, . . . , then the sequence {xn } generated by Newton’s iteration

f (xn )
xn+1 = xn − , n = 0, 1, 2, . . .
f 0 (xn )
converges to r:
lim xn = r.
n→∞
 CHAPTER

4
Integration

4.1 Areas under Curves

4.1 Example: area under a curve

To find the area A of the region bounded by the curve y = f (x) = x2 and the x-axis over
1
[0, 1], we divide the interval [0, 1] into n subintervals of equal length . Let
n
1 2 n−1 n
x0 = 0, x1 = , x2 = , ..., xn−1 = , xn = =1
n n n n
be the endpoints of these subintervals. For each i (1 ≤ i ≤ n), over the i-th subinterval
[xi−1 , xi ], we use the the area of the rectangle with height f (xi ) to approximate the area of
the region between the curve and the x-axis over the interval [xi−1 , xi ].

y y

RSn RS2n

x1 x2 xi−1 xi xn = 1
x 1
x
O O
186 Integration

The sum of n rectangles is

RSn = f ( n1 ) · n1 + f ( n2 ) · n1 + · · · + f ( nn ) · 1
n
 2 2

= n12 + n2 2 + · · · + nn2 · n1

1 n(n + 1)(2n + 1)
= 1 + 2 2 + . . . + n2 · 3 = .
n 6n3
Clearly RSn approximates the area A, and when n get larger, the approximation gets better.
n(n + 1)(2n + 1)
Since lim RSn = lim = 13 , we get A = 13 .
n→∞ n→∞ 6n3
We can similarly choose the left endpoint of each subinterval instead of the right endpoint
to make a new approximation to A. Denote it as LSn . By the same manner, we also get
A = lim LSn = 13 .
n→∞

y y

LSn Sn

x1 x2 xi−1 xi xn = 1
x x1 x2 xi−1 ∗ xi xn = 1
x
O O xi

For any fixed n, if we choose x∗i in the interval [xi−1 , xi ] arbitrarily and use the rectangle
of height f (x∗i ) over the base [xi−1 , xi ] to make approximation, we will get the sum
1
Sn = [f (x∗1 ) + . . . + f (x∗n )] · .
n
Clearly LSn ≤ Sn ≤ RSn . As n → ∞, since both LSn and RSn converge to the same limit
3 , by using a similar argument as the Squeeze Theorem, we get
1

lim Sn = 13 .
n→∞

4.2 MATLAB: Riemann sums

In MATLAB, the command rsums is used to interactively approximate the integral
ˆ 1 of a
function by Riemann sum. The following example shows the approximation of x dx by
2
0
using the command.
4.1 Areas under Curves 187

>> syms x ;
>> f = x ^2;
>> rsums (f ,[0 ,1])

After running the MATLAB code above, the figure shown in the following is generated.
0.332500

0.8

0.6

0.4

0.2

0
0 0.2 0.4 0.6 0.8 1
10

The figure above displays a graph of f (x) = x2 using 10 rectangles. In other words,
the interval [0, 1] is divided into 10 subintervals. Note that the height of each rectangle
is determined by the value of f (x) = x2 in the middle of each subinterval. In this case,
ˆ 1
x2 dx ≈ 0.332500, which we can see from the top of the figure. Furthermore, we can
0
adjust the number of the rectangles (from 2 to 128) by using the slider below the figure. For
example, if we choose the number of the rectangles to be 80, the graph of f (x) = x2 is shown
below.

ˆ 1
From the figure above, we see that x2 dx ≈ 0.333320 which is closer to the real value
ˆ 1 0

of x2 dx (i.e., 13 ).
0
188 Integration

4.3 Remark: sigma notation

We often encounter sums with multiple terms, for example,

1 + 2 2 + · · · + n2 .

For convenience, we use a sigma notation to write

n
X
1 + 2 2 + · · · + n2 = i2 .
i=1

Here i2 is the general term of the summation. The symbol Σ is the Greek capital letter
Sigma corresponding to our letter ‘S’, and refers to the initial letter of the word ‘Sum’. So
this expression means the sum of all the terms i2 , where i takes the values from 1 to n.
In general, for fixed numbers k and n with k ≤ n, let ai be a number depending on i.
Then the sum ak + ak+1 + · · · + an can be represented in terms of the following sigma sum
notation:
Xn
ai = ak + ak+1 + · · · + an .
i=k

4.4 Definition: definite integral

Let f be a function defined over [a, b]. Divide the interval into n subintervals of equal
b−a
length :
n
[x0 , x1 ], [x1 , x2 ], . . . , [xn−1 , xn ].
b−a
Let ∆x = xi − xi−1 = . Choose x∗i in [xi−1 , xi ] for each subinterval. The Riemann sum
n
is
n
X
f (x∗i )∆x.
i=1

The function f is said to be integrable if the limit of the Riemann sum exists and has a
unique limit, i.e.,
Xn
lim f (x∗i )∆x = L
∆x→0
i=1

for all choices of

x∗i .The limit is called the definite integral of f (or Riemann integral)
from a to b, denoted by
ˆ b
f (x) dx,
a
where the function f is called the integrand, a is the lower limit, and b is the upper limit.
Theorem: necessary condition

If f is integrable on [a, b], then f must be bounded on [a, b].

4.1 Areas under Curves 189

4.5 Theorem: integrable functions

If f is continuous over [a, b] or bounded on [a, b] with a finite number of discontinuous

points, then f is integrable on [a, b].

4.6 Example: integrable functions

The function y

 cos x, if x ≤ −1;

y = h(x) 2
h(x) = sin x, if −1 < x ≤ 1;
2, if x > 1 1

x
is integrable on [−2, 2] since it is discontin- −2π − 32 π −π − 21 π O
−1
1

uous only at the points x = −1 and x = 1.

The function is bounded and continuous away
from these two discontinuities.

4.7 Pitfall: nonexistence of definite integrals

1 y
The integrand of the definite integral,
x
ˆ 2
1 1
dx, y=
x
−1 x 1
−1
is continuous everywhere except at x = 0. x
O 2
However, the integral does not exist since
f (x) is not bounded.
1
The function is not integrable on [−1, 2].
x

4.8 Definition: area and net area

• Let R be a region. The net area of R is the area of the part of R above the x-axis
minus the area of the part of R below the x-axis. If R is bounded by the graph y = f (x)
and the x-axis over the interval [a, b], then the net area of the region R is
ˆ b
f (x) dx.
a

y = f (x)
+

a +
x
− − b
190 Integration

• Let R be a region. The area of R is the area of the part of R above the x-axis plus
the area of the part of R below the x-axis. If R is bounded by the graph y = f (x) and the
x-axis over the interval [a, b], then the area of the region R is
ˆ b
|f (x)| dx.
a

y y

y = f (x) y = |f (x)|

a
x a
x
b b

4.9 Example: area and net area

ˆ 1 y
1. The integral x dx calculates the area of
0
the region between the curve y = x and the y=x
x-axis over the interval [0, 1]. This gives 1
a triangular region above the x-axis, with
the area 12 . Thus, the area equals the net
x
area in this case. O 1

ˆ 1 y
2. The integral x dx calculates the net
−1
area of the region between the curve y = x 1

and the x-axis over the interval [−1, 1]. y=x

The net area is the area of the region in −1

x
the first quadrant, which is 12 , minus the O 1

area of the region in the third quadrant,

which is also 12 . Hence the net area is 0. −1

The area of the region is 12 + 12 = 1.

4.1 Areas under Curves 191

ˆ 1 y
3. The integral x dx calculates the net
−2
area of the region between the curve y = x 1
y=x
and the x-axis over the interval [−2, 1]. −2 −1
The net area is the area of the region in the O 1
x
first quadrant, which is 12 , minus the area
of the region in the third quadrant, which −1

is 2. Hence the net area is − 23 . However,

the area of the region is 12 + 2 = 25 . −2

4.10 MATLAB: definite integrals

To calculate definite integrals, we can use the commands int and integral. Specifically,
int performs symbolic integration and integral performs numerical integration. The fol-
ˆ 1
lowing example shows the calculation of (1−x2 ) dx using int and integral, respectively.
−1

>> syms x ;
>> f = 1 - x ^2;
>> int (f , -1 ,1)

ans =

4/3

>> integral (@( x ) 1 - x .^2 , -1 ,1)

ans =

1.3333

We see from the MATLAB code above that the command int returns a symbolic number
3,while the command integral returns a floating-point number 1.3333. Note that the first
4

input of the command int is a symbolic expression, but the first input of the command
integral should be a function handle.
Besides int and integral, the command quadgk can also be used for numerical integration.

4.11 MATLAB: definite integrals - polynomials

If the integrand is a polynomial, besides int and integral, we can also use the command
ˆ 1
polyint to calculate the integral. The following example shows the calculation of (1 −
−1
x2 ) dx using polyint.
First, we calculate the coefficients of the new polynomial obtained after integration, and
192 Integration

we have the following:

>> p = [ -1 0 1]; % -x ^2 + 1
>> q = polyint ( p ) % coefficients of the polynomial after ...
integration

q =

-0.3333 0 1.0000 0

We see from the MATLAB code above that after integration, the new polynomial is
−0.3333x3 + x. Next, we calculate the values of −0.3333x3 + x at the lower limit −1 and the
upper limit 1, respectively, and obtain the result of the integral using the following code.

>> a = -1; % lower limit

>> b = 1; % upper limit
>> q_val = polyval (q ,[ a b ]) % evaluate the values of q at a ...
and b , respectively

q_val =

-0.6667 0.6667

>> polyint_ans = diff ( q_val ) % q_val (2) - q_val (1)

polyint_ans =

1.3333
ˆ 1
The result shows that (1 − x2 )dx = 1.3333. Note that for diff(X),
−1
if the input X is a vector, the command diff(X) returns the result of
[X(2) − X(1), X(3) − X(2), . . . , X(n) − X(n − 1)].

4.12 Engineering Example: energy stored in a capacitor

The capacitor is an electrical component that can store electrical energy in an electric field.
Capacitors have been widely used as important parts of electrical circuits in many common
electrical devices. The following figures show a real capacitor and a symbol for the capacitor
in circuits, respectively.
4.1 Areas under Curves 193

V
− +

The energy stored in a capacitor is equal to the work done in establishing the electric
field from an uncharged state by moving charge for the negative to the positive plate of the
capacitor. Mathematically, the energy W stored in a capacitor can be written as a definite
integral, as shown in the following:
ˆ Q
W = V (q) dq,
0

where Q is the total charge stored in the capacitor, q is a variable representing the charge,
and V is the voltage across the capacitor (which is a function of q).

4.13 Engineering Example: PID controller

The Proportional-Integral-Derivative (PID) controller is a closed-loop feedback component

widely used in industrial control systems and different types of other applications, e.g., the
cruise control on a car. The PID controller first calculates an error value as the difference
between a desired setpoint and a measured output. Then, based on the error value, a cor-
rection term based on the proportional (P), integral (I) and derivative (D) terms of the PID
controller is applied at the input. This process is executed continuously to achieve the desired
setpoint in an optimal way as well as to keep the system in a steady state.
Mathematically, the overall control function of the PID controller can be written as
ˆ t
de(t)
u(t) = Kp e(t) + Ki e(t0 ) dt0 + Kd
| {z }
| 0 {z | {zdt }
Proportional (P) term
}
Integral (I) term Derivative (D) term

where Kp ≥ 0, Ki ≥ 0 and Kd ≥ 0 are the coefficients for the proportional (P), integral (I),
and derivative (D) terms, respectively, and e(t) is the error value.

4.14 Engineering Example: electrical energy

Any object that carries charges will experience electrical force created by electrical potential
difference (voltage) in electrical field. The force will drive charges to move orderly in one
direction. Such orderly movement of charges will form current. All electrical devices work
based on the transformation of electrical energy to other types of energy, for example, heat,
light, mechanical energy, etc. The power generated/consumed by a component (battery,
resistor, etc.) is calculated by the product between the voltage across the component and
the current going through the component. Consider a circuit as shown in the following:
194 Integration

i(t) R

−
−
vs (t) C vC (t)
+
+

The power generated by the power source at time t can be obtained by vs (t)i(t), where
vs (t) is the voltage of the source and i(t) is the electric current. As a result, the electrical
energy E provided by the source during T can be calculated as
ˆ T
E= vs (t)i(t) dt.
0

Properties of Definite Integrals

4.15 Theorem: properties of definite integrals

Let f and g be integrable functions over an interval I and let a < b < c be points in I.
1. If f and g are integrable over [a, b], then f + g is integrable over [a, b], and we have
ˆ b ˆ b ˆ b
[f (x) + g(x)] dx = f (x) dx + g(x) dx.
a a a

2. If f is integrable over [a, b], then kf (x) is integrable over [a, b], and we have
ˆ b ˆ b
kf (x) dx = k f (x) dx.
a a

3. If f is integrable over [a, b] and [b, c], then f is integrable over [a, c], and we have
ˆ b ˆ c ˆ c
f (x) dx + f (x) dx = f (x) dx.
a b a
ˆ a
4. f (x) dx = 0.
a

5. If f and g are both integrable over [a, b], and f (x) ≤ g(x), then
ˆ b ˆ b
f (x) dx ≤ g(x) dx.
a a

6. ˆ  ˆ
 b  b
f (x) dx ≤ |f (x)| dx.
 

 a  a
4.1 Areas under Curves 195

7. (Mean Value Theorem for Integrals) If f is continuous on [a, b], then there is c ∈
[a, b] such that
ˆ b
f (x) dx = f (c)(b − a).
a

Justification: Part 1
The proof of these properties follows directly from the definition of the definite integral.
Let’s prove the last two statements.
Consider the Riemann sum of the function f for any partition:

a = x0 < x1 < · · · < xn−1 < xn = b,

which divides the interval [a, b] into n subintervals. For each index i, choose x∗i inside the
subinterval [xi−1 , xi ] and make the Riemann sum
n
X
Rn (f ) = f (x∗i )∆xi ,
i=1

where ∆xi = xi −xi−1 > 0 is the length of the subinterval [xi−1 , xi ]. Using the same partition
and same choice of x∗i , we can also have the Riemann sum of the function |f (x)| ,
n
X
Rn (|f |) = |f (x∗i )|∆xi .
i=1

Recall the triangle inequality

|α + β| ≤ |α| + |β|.
Apply the inequality to the Riemann sum, we get
Xn  X n
f (x∗i )∆xi  ≤ f (x∗i )∆xi , i.e., |Rn (f )| ≤ Rn (|f |).
   

i=1 i=1

Take limits of both sides as n → ∞, we get

lim |Rn (f )| ≤ lim Rn (|f |), order rule of limits

n→∞ n→∞
 
 lim Rn (f ) ≤ lim Rn (|f |), continuity of the absolute value function
 
n→∞ n→∞
ˆ  ˆ
 b  b
f (x) dx ≤ |f (x)| dx.
 

 a  a
196 Integration

Justification: Part 2

If f is continuous on [a, b], by the Extreme Value Theorem, the function f can attain its
minimum m and maximum M :

f (c1 ) = m, f (c2 ) = M,

where c1 , c2 ∈ [a, b]. Since m ≤ f (x) ≤ M , we have

ˆ b
m(b − a) ≤ f (x) dx ≤ M (b − a).
a

Thus,
ˆ
1 b
f (c1 ) ≤ f (x) dx ≤ f (c2 ).
b−a a
By the Intermediate Value Theorem, there is a number c between c1 and c2 , such that
ˆ
1 b
f (c) = f (x) dx.
b−a a

Obviously, c ∈ [a, b] and the equality holds:

ˆ b
f (x) dx = f (c)(b − a).
a

4.16 Example: simple definite integrals

ˆ 1 y
1. The integral x dx represents the area
0
of the triangular region bounded by the y=x
graph y = x and the x-axis over [0, 1], 1
which equals 21 . So,
ˆ 1 x
O 1
x dx = 12 .
0
4.1 Areas under Curves 197

2. From the properties of definite integrals, y

we have the net area of the region bounded
by the graph y = 2x+1 and the x-axis over 3

[0, 1]:
ˆ 1 ˆ 1 ˆ 1 y = 2x + 1

(2x + 1) dx = (2x) dx + 1 dx
0 0 0
ˆ 1 ˆ 1 1

=2 x dx + 1 dx
0 0
x
=2· 1
2 + 1 = 2. O 1

3. For the piecewise function

if −2 ≤ x ≤ −1;

 x, y
f (x) = 1, if −1 < x ≤ 1;
−x, if 1 < x ≤ 3, y = f (x)

1
by using the different expressions over dif-
ferent intervals, we have −2 −1 O 1 2 3
x
ˆ 3
−1
f (x) dx
−2 −2
ˆ −1 ˆ 1 ˆ 3
= f (x) dx + f (x) dx + f (x) dx −3
−2 −1 1
ˆ −1 ˆ 1 ˆ 3
= x dx + 1 dx + (−x) dx
−2 −1 1
= − 32 + 2 − 4 = − 72 .

We note that the contributions from the integrals with negative integrands are negative.

Average Value of a Function

4.17 Example: average temperature

Since temperatures fluctuate throughout the day, as well as varying substantially depending
on the locations, the average temperature gives you a more accurate picture of the tempera-
ture in a specific location than a single measurement ever could. For example, knowing the
average temperature of a given week in Hong Kong will certainly help a traveler to prepare
his trip to the city.
The average temperature can be calculated from several temperature measurements using
the formula:
sum of measured temperatures
average temperature = .
number of measurements
198 Integration

For instance, for calculating the average temperature on January 1, 2019, one may first
measure temperature every hour in Central, Hong Kong to get a dataset of 24 temperatures.
Then using the formula above, one may obtain an approximation of the average temperature.
Apparently, increasing the number of measurements should get a better result in reflecting
Hong Kong’s temperature situation on the New Year Day.

4.18 Definition: average value of a function

The average value of f on the interval [a, b] is

ˆ
1 b
fave = f (x) dx,
b−a a

provided that the integral exists.

Justification:

To compute the average value of a function y = f (x) on [a, b], we start by dividing
the interval [a, b] into n equal subintervals, each with length ∆x = (b − a)/n. Then we
choose points x∗1 , . . . , x∗n in successive subintervals and calculate the average of the numbers
f (x∗1 ), . . . , f (x∗n ):
f (x∗1 ) + · · · + f (x∗n )
.
n
Since n = (b − a)/∆x, the average value becomes

f (x∗1 ) + · · · + f (x∗n ) 1
= [f (x∗1 )∆x + · · · + f (x∗n )∆x]
(b − a)/∆x b−a
n
1 X
= f (x∗i )∆x,
b − a i=1

so that ˆ
n
1 X 1 b
lim f (x∗i )∆x = f (x) dx.
n→∞ b − a b−a a
i=1

Therefore, we define the average value of f on [a, b] as

ˆ
1 b
fave = f (x) dx.
b−a a

4.19 Example: average distance from the Pluto to the Sun

The perihelion is the shortest distance from a planet to the Sun, and aphelion is the longest
distance. The Sun is at one of the two foci. For the Pluto, the perihelion is 4.44 billion km
and the aphelion is 7.38 billion km.
4.1 Areas under Curves 199

As shown in the figure, assume that the or- y

bit of the Pluto is b

2 2
x y
2
+ 2 = 1, Pluto
a b
d
where 0 < b < a.√ It is known that the foci
x
are located at (± a2 − b2 , 0). √
Let us assume Sun a

the coordinates of the Sun is ( a2 − b2 , 0). aphelion perihelion

Then the distance from the Pluto, located at (x, y), to the Sun is
q p
d = (x − a2 − b2 )2 + y 2
s
x2
p  
= (x − a − b ) + b 1 − 2
2 2 2 2
a
r 2
p
= a−1 a2 − a2 − b2 x
 p  p
= a−1 a2 − a2 − b2 x . a2 > a2 − b2 x for −a ≤ x ≤ a

Thus, the average distance from the Pluto to the Sun is

ˆ a
1  p 
D= · a−1 a2 − a2 − b2 x dx
2a −a
ˆ a ˆ a
1 1 p
= · a dx − a−1 a2 − b2 x dx
2a −a 2a −a
1
= · a · 2a − 0 = a.
2a
In other words, the average distance equals to the average of the perihelion and aphelion:
perihelion + aphelion
D= .
2

4.20 Theorem: geometric interpretation on average value

If f is continuous on [a, b], then there exists y

a number c in [a, b] such that
ˆ b
1
f (c) = fave = f (x) dx, f (c)
b−a a
that is
ˆ b x
O a c b
f (x) dx = f (c)(b − a).
a
200 Integration

4.21 Engineering Example: electricity voltage in China

The signal we get from the wall socket is from the power plant. This signal fluctuates in
terms of both the magnitude and the direction, we thus call it “Alternating Current (AC)
signal”. Specifically, the voltage provided by the AC signal can be formulated as

v(t) = v0 sin (2πf t)

where f is the electricity frequency. Note that f = 50 Hz in China.

It is well known that China generally uses a 220 volts voltage. We may wonder whether
v0 = 220 or not. Actually, v0 represents the peak value of the voltage and v0 = 311. Then,
why do we say “China uses a 220 volts voltage”?
Let’s consider the following two circuits.

v −
v(t) R R
+

As we said, in practice, we use the AC signal, as shown in the left figure. Assume that
the AC signal provides power PAC for the resistance R. Can we find an equivalent Direct
Current (DC) signal, as shown in the right figure, which provides power PDC for R such that
PDC = PAC ?
In order to answer this question, we first need to calculate the power PAC . Let the period
of the AC signal be denoted by T . Based on Ohm’s law, the calculation is shown below:
ˆ ˆ
1 T
v 2 (t) 1 T
PAC = dt = v02 sin2 (ωt) dt
T 0 R TR 0
ˆ T ˆ ˆ !
1 2 1 − cos(2ωt) 1 v02 T v02 T
= v dt = dt − cos(2ωt)dt
TR 0 0 2 TR 2 0 2 0
1 v02 1 v02 1 v2
 T 
= − sin(2ωt) = 0.

R 2 T 2 2ω 0 2R

For the DC circuit in the right figure, based on Ohm’s law, the power provided for R can
be calculated as
v2
PDC = .
R
If PDC = PAC , we have

v2 v2 v0
= 0 ⇒ v = √ ≈ 220 (V).
R 2R 2
Therefore, the “220 volts” is a DC equivalent voltage such that the DC signal provides the
same power as the AC signal v(t) = 311 sin (100πt).
4.2 Fundamental Theorem of Calculus 201

4.2 Fundamental Theorem of Calculus

4.22 Theorem: Fundamental Theorem of Calculus

Let f be a continuous function on [a, b].

1. Then the function ˆ x

A(x) = f (t) dt
a

is continuous on [a, b], differentiable in (a, b), and A0 (x) = f (x).

2. If F is a differentiable function [a, b] such that F 0 (x) = f (x), then
ˆ b
f (x) dx = F (b) − F (a).
a

Justification: Part 1

Since f is continuous over [a, b], it has an absolute maximum and an absolute minimum
over [a, b], and thus it is bounded, i.e., there exists B such that |f (x)| ≤ B for all x in [a, b].
Let c be a point in [a, b]. Let x be an arbitrary point and assume x > c. By the properties
of integration, we have
ˆ x ˆ c  ˆ x 
|A(x) − A(c)| =  f (t) dt − f (t) dt =  f (t) dt
   

ˆ ax ˆa x c

≤ |f (t)| dt ≤ B dt = B(x − c).

c c

Thus when x → c+ , by the Squeeze Theorem, lim+ |A(x) − A(c)| = 0, i.e., lim+ A(x) = A(c).
x→c x→c
By the same argument, we get lim A(x) = A(c). Therefore we conclude that lim A(x) =
x→c− x→c
A(c), i,e., A is continuous at c.
To show that f is differentiable in (a, b), we assume c is any point in (a, b). Without loss
of generality, assume x > c. We write:
ˆ x ˆ c ˆ x
A(x) − A(c) − f (c)(x − c) = f (t) dt − f (t) dt − f (c) dt
ˆax ˆax c

= f (t) dt − f (c) dt
ˆc x c

= [f (t) − f (c)] dt.

c

Let M (x) be the maximum value of the function |f (t) − f (c)| over the interval between c and
x, i.e.,
|f (t) − f (c)| ≤ M (x) for all t between c and x.
202 Integration

It depends on x. Since f (x) is continuous, M (x) goes to zero as x goes to 0. Thus we get
 ˆ x
f (t) − f (c) 
 
 A(x) − A(c)
− f (c) =  dt
 
  
 x−c c x−c
ˆ x
 f (t) − f (c) 

≤  x − c  dt properties of integrals
 
c
ˆ x
M (x)
≤ dt the order rule
c x−c
M (x)
= (x − c) = M (x),
x−c
which goes to zero as x goes to 0. Thus
 
A(x) − A(c) A(x) − A(c)
lim − f (c) = 0, i.e. lim = f (c).
x→c x−c x→c x−c

Thus we obtain A0 (c) = f (c).

Justification: Part 2
b−a
Partition the interval [a, b] with n subintervals of equal length . By the Mean Value
n
Theorem, there exists x∗i in [xi−1 , xi ] such that

b−a
F (xi ) − F (xi−1 ) = F 0 (x∗i )(xi − xi−1 ) = f (x∗i ) · .
n
Then, we take the Riemann sum of the function f on [a, b]:
n n
X b−a X
f (x∗i ) · = [F (xi ) − F (xi−1 )] = F (xn ) − F (x0 ) = F (b) − F (a).
i=1
n i=1

Thus by the definition of definite integral, we have

ˆ b n
X b−a
f (x) dx = lim f (x∗i ) · = F (b) − F (a).
a n→∞
i=1
n

4.23 Example: evaluate definite integrals

Evaluate ˆ ˆ ˆ
1 π 1
x2 dx, sin x dx, and ex dx.
0 0 0
Solution
4.2 Fundamental Theorem of Calculus 203

1 3 0
1. Since = x2 , by the Fundamental Theorem of Calculus, we have


3x
ˆ 1
1 3 1
x2 dx = = 13 .

3x 0
0

2. Since (− cos x)0 = sin x, by the Fundamental Theorem of Calculus, we have

ˆ π
π
sin x dx = − cos x|0 = − cos π − (− cos 0) = 1 + 1 = 2.
0

3. Since (ex )0 = ex , by the Fundamental Theorem of Calculus, we have

ˆ 1
1
ex dx = ex |0 = e1 − e0 = e − 1.
0

4.24 Example: area and net area

1. Find the net area of the region between the graph of f (x) = x2 − 1 and the x-axis over
[−2, 2].

Solution The net area of the region is the y

integral of f over [−2, 2]. It is easy to see
3
that
0
3 x − x = x − 1.
1 3 2 y = x2 − 1

Thus, by the Fundamental Theorem of 1

Calculus, the net area is O
ˆ 2 −2 −1 1 2
x

2
(x2 − 1) dx = 13 x3 − x −2 −1
−2
= 13 (23 ) − 2 − 13 (−2)3 − (−2) = 34 .
   

2. Find the area of the region between the graph of f (x) = x2 − 1 and the x-axis over
[−2, 2].
Solution The area of the region equals the sum of the area of the region above the
x-axis and the area of the region below the x-axis.

We need to find the intervals over which f does not change its sign. Solve

x2 − 1 = 0 =⇒ x = 1 or x = −1.

Thus f (x) is positive over [−2, −1], negative over [−1, 1], and positive over [1, 2]. Thus,
204 Integration

by the Fundamental Theorem of Calculus, the total area equals

ˆ 2
x − 1 dx
 2 
−2
ˆ −1 ˆ 1 ˆ 2
= (x2 − 1) dx − (x2 − 1) dx + (x2 − 1) dx
−2 −1 1

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


− 31 (13 ) − 1 + 13 (−1)3 − (−1)

   

+ 31 (23 ) − 2 − 13 (13 ) − 1
   

= 44
3 .

y y

3 3

y = x2 − 1 y = |x2 − 1|

1 1

O x x
−2 −1 1 2 −2 −1 O 1 2

−1 −1

4.25 Example: Derivatives of Integrals

Find the derivatives

ˆ ˆ x2
d x
d
sin t dt
2
and sin t2 dt.
dx 0 dx 1
Solution

1. We can use directly the Fundamental Theorem of Calculus to compute the first integral:
ˆ x
d
sin t2 dt = sin x2 .
dx 0
ˆ u
2. Here we need to apply the chain rule of differentiation. If we let f (u) = sin t2 dt
1
and u = g(x) = x2 , we get
ˆ x2
sin t2 dt = f g(x) .


1
4.3 Indefinite Integrals 205

So we apply the chain rule to get

ˆ x2 ˆ
d d u
d(x2 )
 
sin t2 dt = sin t2 dt 

·
dx 1 du 1 u=x2 dx
= sin u 2
· 2x

u=x2
= sin x · 2x = 2x sin x4 .
4

4.26 Pitfall: nonexistence of definite integrals

ˆ 2
1
Consider the integral 2
dx.
−1 x
0
1 1

Since − = 2 , if one had used the Fundamental Theorem of Calculus for this integral,
x x
one would get the integral equal to
ˆ 2
2
1 1 

dx = − incorrect equality
−1 x2 x −1
= − 12 + (−1) = − 23 .

1
The computation is incorrect because the integrand is not defined at x = 0 which is a
x2
point in [−1, 2]. In fact, if one uses Riemann sum to approximate this integral, the limit of
the Riemann sum diverges.

4.3 Indefinite Integrals

4.27 Remark:
From the Fundamental Theorem of Calculus, in order to calculate a definite integral
ˆ b
f (x) dx,
a

it is important to know a function F (x) whose derivative is f (x), i.e., F 0 (x) = f (x). This
leads to the concept of indefinite integral, equivalently, antiderivative.

4.28 Definition: indefinite integral

The indefinite integral of f , denoted by

ˆ
f (x) dx
206 Integration

is the collection of all functions F whose derivative equals f , i.e.,

F 0 (x) = f (x).

Any function F satisfying F 0 (x) = f (x) is called an antiderivative of f .

4.29 Example: finding indefinite integral

ˆ
Find 2x dx.

Solution We know that (x2 )0 = 2x and (x2 + 1)0 = 2x. In fact, for any constant C,
(x + C)0 = 2x.
2

We know from the Mean Value Theorem that zero derivative implies constant function.
Thus, any differentiable function F with F 0 (x) = 2x must differ from x2 at most by a
constant. Hence, by the definition of indefinite integral, we have
ˆ
2x dx = x2 + C.

4.30 Theorem: linearity of indefinite integral

ˆ ˆ ˆ
1. [f (x) + g(x)] dx = f (x) dx + g(x) dx.
ˆ ˆ
2. kf (x) dx = k f (x) dx, where k is a constant.

Justification:
The equalities follow from the linearity of differentiation.

4.31 Theorem: indefinite integrals of some elementary functions

ˆ ˆ
1. k dx = kx + C 6. cos x dx = sin x + C
ˆ
xµ+1 ˆ
2. xµ dx = (µ 6= −1)
µ+1 7. sin x dx = − cos x + C
ˆ
dx
3. = ln |x| + C
x ˆ
ˆ 8. ex dx = ex + C
dx
4. = arctan x + C
x2 + 1
ˆ ˆ
dx ax
5. √ = arcsin x + C 9. ax dx = +C
1 − x2 ln a
4.3 Indefinite Integrals 207

ˆ ˆ
dx
10. = sec2 x dx = tan x + C
cos2 x
ˆ ˆ
dx
11. = csc2 x dx = − cot x + C
sin2 x
ˆ
12. sec x tan x dx = sec x + C
ˆ
13. csc x cot x dx = − csc x + C

Justification:

For formula 1, it is easy to see that (kx + C)0 = k for any constant C. We know from the
Mean Value Theorem that zero derivative implies constant function. Thus, any differentiable
function F with F 0 (x) = k must differ from kx at most by a constant. Hence, by the definition
of indefinite integral, we have ˆ
k dx = kx + C.

Similar argument applies to other formulae, except formula 3.

For formula 3, if x > 0, we have ln |x| = ln x and
1
(ln x)0 = .
x
Since, from the Mean Value Theorem, zero derivative implies constant function, we know that
1
any differentiable function F with F 0 (x) = must differ from ln x at most by a constant.
x ˆ
Hence, by the definition of indefinite integral, we know that for x > 0, x−1 dx = ln x + C.
If x < 0, we have ln |x| = ln(−x) and, applying the chain rule,
1
0 1
ln(−x) · (−x)0 = .
=
−x x
ˆ
By a similar argument as above, we know that for x < 0, x−1 dx = ln(−x) + C.
Combining two cases into one format, we obtain formula 3:
ˆ
x−1 dx = ln |x| + C.

4.32 Remark: constants in indefinite integrals

It is easy to see that

ˆ
1 dx = x + C1 ,
ˆ
2x dx = x2 + C2 ,
ˆ
(1 + 2x) dx = x + x2 + C,
208 Integration

where C1 , C2 , and C are all arbitrary constants.

On the other hand, if we use the linearity of indefinite integrals, we have
ˆ ˆ ˆ
(1 + 2x) dx = 1 dx + 2x dx

= x + C1 + x2 + C2 = x + x2 + C1 + C2 .

It seems that we get different results from different ways of computations.

However, the constant C in indefinite integrals means that it is an arbitrary constant. For
our example, since C1 and C2 are arbitrary constants, so is C1 + C2 . Therefore, it makes
sense to write x + x2 + C1 + C2 as x + x2 + C.

4.33 Example: definite integral - Fundamental Theorem of Calculus

ˆ 1
1
 
Evaluate x + 2 cos x − 2
5
dx.
0 x +1

Solution We can first evaluate the indefinite integral:

ˆ 
1

x + 2 cos x − 2
5
dx
x +1
ˆ ˆ ˆ
1
= x5 dx + 2 cos x dx − dx applying linearity
x2 + 1
= 61 x6 + 2 sin x − tan−1 x + C using the known integrals

Then, by applying the Fundamental Theorem of Calculus, we get

ˆ 1
1


1
x5 + 2 cos x − 2 dx = 16 x6 + 2 sin x − tan−1 x 0
0 x +1
= 1
6 + 2 sin 1 − tan−1 1
= 1
6 + 2 sin 1 − π4 .

4.34 Remark: displacement function and velocity function

If s(t) is the displacement function, then v(t) = s0 (t) is the velocity function from the
definition of derivatives.
Conversely, if we are given the velocity function v(t), then the indefinite integral
ˆ
v(t) dt

is a collection of displacement functions having v(t) as its velocity.

4.3 Indefinite Integrals 209

4.35 Example: velocity - application

A particle moves on a line with the velocity at time t, v(t) = t2 − t − 6 (measured in meters
per second).

1. Find the displacement function of the particle as a function of the time t. Assume the
particle is at the position of the origin when time t = 0.

Solution Let f be the displacement function of the particle satisfying f (0) = 0. Since
f 0 (t) = v(t), in order to find f , we just need to calculate the indefinite integral
ˆ ˆ
f (t) = v(t) dt = (t2 − t − 6) dt = 31 t3 − 12 t2 − 6t + C.

The initial condition f (0) = 0 will determine the constant C. In fact, C = 0.

2. Find the displacement traveled during the time period 1 ≤ t ≤ 4.

Solution It requires us to calculate the following definite integral.

ˆ 4

4
v(t) dt = 31 t3 − 12 t2 − 6t 1 = − 29 .
1

3. Find the distance traveled during the time period 1 ≤ t ≤ 4.

ˆ 4
Solution This amounts to calculate the definite integral |v(t)| dt.
1
To do so, we need to find intervals where v has the same sign. In fact, let v(t) = 0, we
get
t2 − t − 6 = 0 =⇒ (t − 3)(t + 2) = 0 =⇒ t = 3 or t = −2.

It is easy to see that v(t) is positive on (−∞, −2), negative on (−2, 3), and positive on
(3, ∞). Therefore the distance traveled during the time period 1 ≤ t ≤ 4 is
ˆ 4 ˆ 3 ˆ 4
|v(t)| dt = − v(t) dt + v(t) dt
1 1 3

3
4
=− 1 3
3t − 12 t2 − 6t 1 + 1 3
3t − 21 t2 − 6t 3 = 61
6 .

4.36 MATLAB: indefinite integrals

In MATLAB, the command intˆcan be used to

ˆ calculate indefinite integrals. The following
examples show the calculation of x dx and
n
t2 − t − 6 dt.

>> syms x ;
210 Integration

>> f = x ^ n ;
>> int (f , x )

ans =

piecewise ( n == -1 , log ( x ) , n 6= -1 , x ^( n + 1) /( n + 1) )

>> syms t ;
>> g = t ^2 - t - 6;
>> m = int ( g )

m =

-( t *( - 2* t ^2 + 3* t + 36) ) /6

>> m_cl = collect ( m ) % rewrite m in terms of the powers of t

m_cl =

t ^3/3 - t ^2/2 - 6* t

>> pretty ( m_cl ) % rewrite m_cl in a format that resembles ...

type - set mathematics
3 2
t t
-- - -- - 6 t
3 2
ˆ
We see from the MATLAB code above that xn dx is a piecewise function. With the help
of collect and pretty, we can rewrite the results calculated by int. Note that for all the
results calculated by int, the constant term C is not shown.

4.4 Substitution Method

4.37 Example: substitution method
2
For the function ex , we apply the chain rule and get
d x2 2
e = 2xex .
dx
We can decompose the computation into several steps. First, let u = x2 . Then
d x2 d u  du

e = e · .
dx du u=x2 dx
4.4 Substitution Method 211

By the definition of indefinite integral, we get

ˆ
2 2
2xex dx = ex + C.

Or, if we let u = x2 , then we get du = (x2 )0 dx = 2x dx, and the integral can be rewritten as
ˆ ˆ ˆ
2 2 2
2xex dx = ex 2x dx
| {z } = eu du = eu + C = ex + C.
du

This process is a typical case of the substitution method.

4.38 Theorem: substitution method for indefinite integral

Let u = g(x) be a differentiable function and f be a continuous function. Suppose the

range of g is contained in the domain of f . Then
ˆ ˆ
f g(x) · g (x) dx = f (u) du.

0

Justification:

It follows directly from the chain rule for differentiation. Let F (x) be a function whose
derivative is f (x), i.e., F 0 (x) = f (x). Equivalently, we have
ˆ
f (u) du = F (u) + C.

Then by the chain rule of differentiation, we have

d
dF (u) 0
F g(x) = · g (x) = f (u) · g 0 (x) = f g(x) · g 0 (x),


dx du
or equivalently
ˆ ˆ
d
f g(x) · g 0 (x) dx = F g(x) dx



dx
= F g(x) + C = F (u) + C u = g(x)


ˆ
= f (u) du.

4.39 Example: substitution method for indefinite integrals

Using the substitution method to evaluate the integral

ˆ
x
√ dx.
x+1
212 Integration

1. Let u = x + 1. Then du = dx. Thus, we have

ˆ
x
√ dx
x+1
ˆ
u−1
= √ du u = x + 1, du = dx
u
ˆ  
1 1
= u 2 − u− 2 du simplifying
3 1
= 23 u 2 − 2u 2 + C evaluating
3 1
= 3 (x +
2
1) − 2(x + 1) + C
2 2 replacing u by x + 1
√
= (x + 1)3 − 2 x + 1 + C.
2
p
3

√
2. We can also make the following substitution u = x + 1. Then x = u2 − 1 and
dx = 2u du. Thus, we have
ˆ
x
√ dx
x+1
ˆ 2
u −1 √
= 2u du u = x + 1, dx = 2u du
u
ˆ
= (2u2 − 2) du simplifying

= 23 u3 − 2u + C evaluating
√ √
= 23 (x + 1)3 − 2 x + 1 + C. replacing u by x+1
p

From this example, we see that there might be more than one way to make substitution.

4.40 Theorem: substitution method for definite integral

Let u = g(x), where g 0 (x) is continuous on [a, b], and let f be continuous on an interval
containing the range of g. Then
ˆ b ˆ g(b)
f g(x) · g 0 (x) dx = f (u) du.


a g(a)

4.41 Example: substitution method for definite integrals

Evaluate ˆ π
2
sin3 x cos x dx.
0
Solution If we let u = sin x, then

du = (sin x)0 dx = − cos x dx.

4.4 Substitution Method 213

When x = 0, u = sin 0 = 0; when x = 2,

π
u = sin π2 = 1. Thus the integral can be evaluated
as follows:
ˆ π2
sin3 x |cos{z
x dx}
0
−du
ˆ 1
substituting u = sin x, du = − cos x dx
= u3 (−du)
0 and changing the upper and lower limits
1
= − 41 u4 0
=− 1
− 0 = − 14 .


4

4.42 Engineering Example: electrical power in a circuit

The electrical power can be supplied to business and homes by the electrical power industry
through electrical power grids, and it can be converted into other forms of energy, such as
heat, light and motion. The amount of power produced in a circuit is related to the electric
current, the voltage, etc.
In particular, in the case of an alternating current (AC) circuit with Ohmic resistive loads,
the Ohm’s law can be combined, and thus the electrical power during one time period (0, T )
is given by
ˆ T
R
P = i2 (t)dt,
T 0

where i(t) is the alternating current.

Suppose that i(t) = A sin (ωt), where A is the amplitude and ω is the angular frequency.
Calculate P .
First, if i(t) = A sin (ωt), the electrical power P can be calculated by solving the following
integral:
ˆ T
A2 R
P = sin2 (ωt) dt.
T 0

Let u = ωt, then

du = ωdt.

When t = 0, u = ω × 0 = 0; when t = T , u = ωT . Thus the integral of P can be rewritten

as follows:
ˆ
A2 R ωT
P = sin2 udu.
Tω 0
214 Integration

Moreover, we note that sin2 u = 1

2 − 12 cos(2u), the integral above thus can be rewritten as
ˆ
A2 R 1 1ωT  
P = − cos(2u) du
Tω 0 2 2
ˆ ˆ
A2 R ωT A2 R ωT
= du − cos(2u)du
2T ω 0 2T ω 0
ˆ
1 A2 R ωT
= A2 R − cos(2u)du.
2 2T ω 0

Furthermore, we let v = 2u, then

dv = 2du.

When u = 0, v = 2 × 0 = 0; when u = ωT , v = 2 × ωT = 2ωT . We thus have

ˆ ˆ
ωT
1 2ωT
cos(2u)du = cos vdv
0 2 0
1 2ωT 1
= [sin v]0 = sin (2ωT ) .
2 2
Therefore, we can obtain
1 2 1
P = A R− A2 R sin (2ωT ) .
2 4T ω
Note that T = ω ,
2π
so P can be further simplified as

1 2 1 2 1
P = A R− A R sin (4π) = A2 R.
2 8π 2

4.43 Engineering Example: hazard function

The hazard function provides the instantaneous failure rate of a component at time t.
Mathematically, the hazard function h(t) can be defined as the ratio of the probability density
function f (t) and the complementary cumulative distribution function R(t) of the lifetime of
a component, i.e.,
f (t)
h(t) = .
R(t)
Note that R(t) can be calculated in terms of f (t) as follows:
ˆ t
R(t) = 1 − f (u)du,
0

so hazard function can be rewritten as

f (t)
h(t) = ´t .
1 − 0 f (u)du
4.4 Substitution Method 215

√
Suppose that the probability density function of a component is f (t) = 2√ 1
t
e− t
. Find
h(t).
In order to calculate h(t), we first calculate the integral in the denominator:
ˆ ˆ
t
1 t
1 √
f (u)du = √ e− u du.
0 2 0 u
√
Let v = u, then
1
dv = √ du.
2 u
√ √
When u = 0, v = 0 = 0; when u = t, v = t. We thus have
ˆ t ˆ √
t
f (u)du = e−v dv
0 0
√t √
= − e−v 0 = 1 − e− t .


Substituting the above result into the expression of h(t) and after some algebraic manipu-
lations, we have
1
h(t) = √ .
2 t

Trigonometric Substitutions
√
4.44 Example: integrals involving a2 − x2

Find the area of a circle of radius a.

Solution The area of the circle equals four y

√
times the area of the part of the circle in the y= a2 − x2
a
first quadrant, which can be √ calculated via
the area below the curve y = a2 − x2 over x2 + y 2 = a2
the interval [0, a], i.e., we need to calculate
the following integral a
x
O
ˆ ap
4 a2 − x2 dx.
0

In order to remove the radical in the integrals, we use the following trigonometric substi-
tution for integrals involving a2 − x2 . Let x = a sin θ. Then dx = a cos θ dθ, and when x = 0,
216 Integration

θ = 0; when x = a, θ = π2 . Hence,
ˆ ap
4 a2 − x2 dx
0
ˆ π
2 p
=4 a2 − a2 sin2 θ · a cos θ dθ making substitution x = a sin θ
0
ˆ π
2
=4 a cos θ · a cos θ dθ using 1 − sin2 θ = cos2 θ
0
ˆ π
2
= 4a 2
2 (1
1
+ cos 2θ) dθ using the double-angle formula
0
π
= 2a2 θ + sin 2θ 0
1
evaluating

 2
2
= 2a π
+ 0 = πa2 . simplifying
2


2

4.45 Method: trigonometric substitutions

Expression Corresponding Substitution Useful Identity

√
a2 − x2 x = a sin θ, − π2 ≤ θ ≤ π
2 1 − sin2 θ = cos2 θ
√
a 2 + x2 x = a tan θ, − π2 < θ < π
2 1 + tan2 θ = sec2 θ
√ 0 ≤ θ < π2 or
x2 − a2 x = a sec θ, sec2 θ − 1 = tan2 θ
π ≤ θ < 3π2

√
4.46 Example: integrals involving x2 − a2

Evaluate the integral ˆ

dx
.
(x2 − 1)3
p

Solution In order to remove the radical in the integrals, we use the trigonometric substi-
tution x = sec θ. Then we have

x2 − 1 = sec2 θ − 1 = tan2 θ, dx = sec θ tan θ dθ.

4.4 Substitution Method 217

Thus
ˆ ˆ
dx sec θ tan θ
= dθ substituting x = sec θ
(x2 1)3 tan3 θ
p
−
ˆ
cos θ
= dθ simplifying
sin2 θ
ˆ
1
= du substituting u = sin θ
u2
1
=− +C evaluating
u
1
=− +C changing the variable from u to θ
sin θ
x
= −√ + C. changing the variable from θ to x
x2 − 1
Here we use the following diagram to convert
the variable from θ to x: x √
x2 − 1
√
x2 − 1
sin θ = . θ
x 1

√
4.47 Example: integrals involving a2 + x2
ˆ
dx
Evaluate the integral .
(1 + x2 )2

Solution Although the integrand contains no radicals, a trigonometric substitution can

still apply to simplify the term (1 + x2 )k . Indeed, using the trigonometric substitution
x = tan θ, we have
1 + x2 = 1 + tan2 θ = sec2 θ, d(tan θ) = sec2 θ dθ.
The integral is
ˆ ˆ ˆ
dx sec2 θ
= dθ = cos2 θ dθ
(1 + x2 )2 sec4 θ
ˆ
= 2 (1
1
+ cos 2θ) dθ = 12 θ + 1
4 sin 2θ + C.

To convert θ back to x, we have θ = tan−1 x. From the triangle below, we have

x 1
sin θ = √ , cos θ = √ , √
1 + x2 1 + x2 1 + x2
x
2x
sin 2θ = 2 sin θ cos θ = .
1 + x2 θ
1
Thus the integral is ˆ
dx x
= 1
tan−1 x + + C.
(1 + x2 )2 2 2(1 + x2 )
218 Integration

4.5 Integration by Parts

4.48 Example: product rule and integration by parts

Let u(x) and v(x) be differentiable functions of x. The product rule of differentiation gives
that
(uv)0 = u0 v + uv 0 .
Rewrite the expression as
uv 0 = (uv)0 − u0 v.
ˆ
As an example, consider the integral xex dx. If we let

u(x) = x, v 0 (x) = ex ,

then, for the latter, we may choose v = ex . From the equality uv 0 = (uv)0 − u0 v, we get
ˆ
xex dx
ˆ ˆ
= x(e ) dx = u(x)v 0 (x) dx
x 0
u = x and v = ex
ˆ h i

0
= u(x)v(x) − u0 (x)v(x) dx applying the product rule
ˆ
= u(x)v(x) − u0 (x)v(x) dx definition of indefinite integral
ˆ
= xex − ex dx

= xex − ex + C.

This process is a typical case of so-called integration by parts.

4.49 Theorem: integration by parts

Suppose u and v are differentiable functions of x. Then

ˆ ˆ
u(x) · v 0 (x) dx = u(x) · v(x) − v(x) · u0 (x) dx,
| {z } | {z }
dv du

or ˆ ˆ
u dv = uv − v du.

Furthermore, for definite integral,

ˆ b ˆ b
b
u(x) · v (x) dx = [u(x) ·
0
v(x)]a − v(x) · u0 (x) dx.
a a
4.5 Integration by Parts 219

Justification:
By the product rule of differentiation, we get

(uv)0 = u0 v + uv 0 =⇒ uv 0 = (uv)0 − u0 v.

Integrating the both sides of the latter equality gives

ˆ ˆ ˆ ˆ
uv dx = (uv) dx − u v dx = uv − u0 v dx,
0 0 0

or equivalently ˆ ˆ
u dv = uv − v du.

By further applying the Fundamental Theorem of Calculus, we get the formula of integra-
tion by parts for definite integral.

4.50 Example: repeatedly integrating by parts

Evaluate the integral ˆ

e2x sin x dx.

Solution We make the following choice of u and v:

Functions in original integral u = e2x dv = sin x dx

Functions in new integral du = 2e2x dx v = − cos x

We obtain
ˆ
e2x sin x dx
ˆ ˆ
= e2x (− cos x)0 dx = u(x)v 0 (x) dx u = e2x and v = − cos x
ˆ h i

0
= u(x)v(x) − u0 (x)v(x) dx applying the product rule
ˆ
= u(x)v(x) − u0 (x)v(x) dx definition of indefinite integral
ˆ
= e (− cos x) − (− cos x)2e2x dx
2x

ˆ
= −e2x cos x + 2 e2x cos x dx.
ˆ ˆ
By comparing the last integral e2x cos x dx with the original integral e2x sin x dx,
we see that integration by parts converts the sine function to the cosine function in the
220 Integration

integrands. This motivates us to apply integration by part once more to the integral
ˆ
e2x cos x dx.

To do so, we make the following choice of u and v:

Functions in original integral u = e2x dv = cos x dx

Functions in new integral du = 2e2x dx v = sin x

We obtain
ˆ
e2x cos x dx
ˆ ˆ
= e (sin x) dx =
2x 0
u(x)v 0 (x) dx u = e2x and v = sin x
ˆ h i

0
= u(x)v(x) − u0 (x)v(x) dx applying the product rule
ˆ
= u(x)v(x) − u0 (x)v(x) dx definition of indefinite integral
ˆ
= e2x sin x − sin x · 2e2x dx
ˆ
= e sin x − 2 e2x sin x dx.
2x

Combining two relations obtained above gives

ˆ ˆ
e sin x dx = −e cos x + 2e sin x − 4 e2x sin x dx.
2x 2x 2x

Therefore, by treating the desired integral as an unknown, we get

ˆ
e2x sin x dx = − 15 e2x cos x + 52 e2x sin x + C.

4.51 Pitfall: choices of u and v

If we choose u = ex , dv = x dx and hence v = 21 x2 to evaluate the integral

ˆ
xex dx.
4.5 Integration by Parts 221

Integrating by parts, we get

ˆ ˆ
1 2 0
xex dx = ex dx


2x
ˆ
= 12 x2 · ex − 1 2
2x · (ex )0 dx
ˆ
= 21 x2 · ex − 1 2
2x · ex dx.

Now we see the resulting integration on the right hand side is more complicated than the
original one. Thus we made a wrong choice of u and v.
Instead, if we let u = x and v = ex , we get
ˆ ˆ
xe dx = x(ex )0 dx
x

ˆ
= xex − ex dx = xex − ex + C.

4.52 Example: recursive formula

In many cases, integration by parts can be used to yield recursive formula for integrals, so
that some complicated integrals can be evaluated through simpler integrals.
ˆ π2
Consider the integral Ip = cosp x dx.
0
For p > 1, if taking u(x) = cosp−1 x and v(x) = sin x, we integrate by parts to have
ˆ π
2
Ip = cosp−1 x · (sin x)0 dx
0
ˆ π
π 2
= cosp−1 x · sin x 02 − sin x · (p − 1) cosp−2 x · (− sin x) dx

0
ˆ π
2
= (p − 1) (1 − cos2 x) cosp−2 x dx
0
= (p − 1)Ip−2 − (p − 1)Ip .

Hence, we have the following recursive formula:

p−1
Ip = Ip−2 , p > 1.
p
It is easy to have that
ˆ π
2
I0 = 1 dx = 2;
π
0
ˆ π
2
I1 = cos x dx = 1.
0
222 Integration

Therefore, we have
ˆ π
2 2n − 1 2n − 1 2n − 3
cos2n x dx = I2n = I2n−2 = · I2n−4
0 2n 2n 2n − 2
2n − 1 2n − 3 1
= ··· = · · · · I0
2n 2n − 2 2
(2n − 1)(2n − 3) · · · 1 π
= ·
2n(2n − 2) · · · 2 2
(2n)! · π
= 2n+1 ;
2 (n!)2
ˆ π
2 2n 2n 2n − 2
cos2n+1 x dx = I2n+1 = I2n−1 = · I2n−3
0 2n + 1 2n + 1 2n − 1
2n 2n − 2 2
= ··· = · · · · I1
2n + 1 2n − 1 3
(2n)(2n − 2) · · · 2
= ·1
(2n + 1)(2n − 1) · · · 3
22n (n!)2
= .
(2n + 1)!

These hold for all integers n ≥ 0.

4.53 Example: use substitution or integration by parts?

Expressions of similar integrals sometimes need different integration techniques.

Consider the integral of the form
ˆ
xp ln x dx.

1. Assume p = −1. In this case, we use the substitution method. We take u = ln x. Then
du = x−1 dx. Thus,
ˆ ˆ
x−1 ln x dx = u du = 12 u2 + C = 12 ln2 x + C.
4.5 Integration by Parts 223

2. Assume p 6= −1. In this case, we use integration by parts. In fact,

ˆ ˆ 0
xp+1

x ln x dx =
p
ln x · dx
p+1
ˆ p+1
xp+1 x
= ln x − (ln x)0 dx
p+1 p+1
ˆ p+1
xp+1 x
= ln x − x−1 dx
p+1 p+1
ˆ
xp+1 xp
= ln x − dx
p+1 p+1
xp+1 xp+1
= ln x − + C.
p+1 (p + 1)2

4.54 Engineering Example: electric current - inductor

The inductor is a two-terminal electrical component that can store energy in a magnetic
field and is characterized by its inductance. The inductance is defined as the ratio of the
voltage to the rate of the change of the electric current. Mathematically, the inductance L is
defined as
v(t)
L = di(t)
dt
where v(t) is the voltage across the inductor and i(t) is the electric current through the
inductor.
Alternatively, according to the fundamental theorem of calculus, the electric current
through the inductor i(t) can be calculated as follows:
ˆ
1 t
i(t) = v(u)du.
L 0
Suppose that L = 5 × 10−3 (H) and v(t) = 3te−t (V), find i(t).
First, we know that
ˆ t
3
i(t) = × 103 ue−u du.
5 0
Integrating by parts, the integral above can be solved as
ˆ t ˆ t

0
ue dt = −
−u
u e−u du
0 0
ˆ t
 −u t
= − ue 0
+ e−u du
0
t
= −te−t − e−u 0


= 1 − e−t − te−t .
Therefore, the electric current through the inductor is
i(t) = 600 1 − e−t − te−t (A).


224 Integration

4.6 Some Common Integrals

Trigonometric Integrals

4.55 Example: integrating products of sin mx and cos nx

1. Calculate ˆ
sin 2x cos 3x dx.

Solution By using the identity

sin α cos β = 1
2 [sin(α − β) + sin(α + β)] ,

we have
ˆ ˆ
sin 2x cos 3x dx = 1
2 (− sin x + sin 5x) dx

= 1
2 cos x − 1
10 cos 5x + C.

2. Calculate ˆ
cos 4x cos x dx.

Solution By using the identity

cos α cos β = 1
2 [cos(α − β) + cos(α + β)] ,

we have
ˆ ˆ
cos 4x cos x dx = 1
2 (cos 3x + cos 5x) dx

= 1
6 sin 3x + 1
10 sin 5x + C.

4.56 Method: integrating products of sin mx and cos nx

For evaluating
ˆ ˆ ˆ
sin mx cos nx dx, sin mx sin nx dx, cos mx cos nx dx,

use the identities:

sin α cos β = 1
2 [sin(α − β) + sin(α + β)]
sin α sin β = 1
2 [cos(α − β) − cos(α + β)]
cos α cos β = 1
2 [cos(α − β) + cos(α + β)]
4.6 Some Common Integrals 225

4.57 Engineering Example: energy in a capacitor - calculation

The capacitor is a two-terminal electronic component that can store electrical energy in an
electric field. The electric power is the rate (per unit time) at which the electrical energy is
transferred by an electric circuit. Hence, to calculate the energy that is stored in a capacitor,
we can calculate the electric power of the capacitor first. In particular, the instantaneous
electric power of the capacitor Pc (t) can be calculated by

Pc (t) = ic (t)vc (t)

where ic (t) is the electric current through the capacitor at time t and vc (t) is the voltage
across the capacitor at time t. Then, the energy stored in a capacitor during the time interval
(0, Tc ) is given by
ˆ Tc ˆ Tc
Wc = Pc (t)dt = ic (t)vc (t)dt.
0 0

Suppose that the voltage across a capacitor with capacitance C and zero energy storage
at time 0 is vc (t) = A sin (ωt), where A is the amplitude and ω is the angular frequency.
Calculate Wc which is the energy stored during the time interval (0, Tc ).
First, the electric current through the capacitor can be calculated as

dvc (t) A · d[sin (ωt)]

ic (t) = C =C = CAω cos (ωt) .
dt dt
Hence, the energy stored in the capacitor during the time interval (0, Tc ) can be calculated
by solving the following integral:
ˆ Tc
Wc = CAω cos (ωt) sin (ωt) dt.
0

Let u = cos (ωt), then we have du = −ω sin (ωt) dt. Moreover, when t = 0, u = 1; when
t = Tc , u = cos (ωTc ). So we can get
ˆ cos(ωTc ) ˆ 1
Wc = −CA u du = CA udu
1 cos(ωTc )
2 1
= CA = 12 CA 1 − cos2 (ωTc )
1 

2 u cos(ωTc )

= 2 CA sin (ωTc ) .
1 2

4.58 Example: integrating products of powers of sin x and cos x

1. Calculate ˆ
sin2 x cos x dx.

Solution We apply the substitution method and let u = sin x. Then we have du =
226 Integration

cos x dx and get

ˆ ˆ
sin2 x cos x dx = u2 du

= 13 u3 + C = 1
3 sin3 x + C.

2. Calculate ˆ
sin3 x cos2 x dx.

Solution If we let u = cos x, then du = − sin x dx and

sin2 x cos2 x = (1 − cos2 x) cos2 x = (1 − u2 )u2 .

Applying the substitution method gives

ˆ ˆ
sin x cos x dx = (1 − u2 )u2 (−du)
3 2

= 15 u5 − 13 u3 + C = 1
5 cos5 x − 1
3 cos3 x + C.

3. Calculate ˆ
sin2 x cos2 x dx.

Solution By the double-angle formula, we write the integrand as

sin2 x cos2 x = 12 (1 − cos 2x) · 21 (1 + cos 2x)

= 14 (1 − cos2 2x)
= 41 1 − 21 (1 + cos 4x)
 

= 18 (1 − cos 4x).

Thus,
ˆ ˆ
sin2 x cos2 x dx = 8 (1
1
− cos 4x) dx

= 18 x − 1
32 sin 4x + C.

4.59 Method: integrating products of powers of sin x and cos x

For integrals of the form ˆ

sin xm x cosn x dx,

one can integrate by the following approach:

1. When m odd, n real: split off sin x, rewrite even power of sin x in terms of cos x, then
substitute u = cos x.
2. When n odd, m real: split off cos x, rewrite even power of cos x in terms of sin x, then
4.6 Some Common Integrals 227

substitute u = sin x.

3. When m and n both even nonnegative integers: use half-angle identities to transform
the integrand into a polynomial in terms of cos 2x, and apply previous strategies.

4.60 Theorem: the universal trigonometric substitution

Consider any integral of the form

ˆ
R(sin x, cos x) dx,

where R is rational in sin x and cos x. Then the universal trigonometric substitution

t = tan( 21 x) or x = 2 tan−1 t

converts the integral into some integral of a rational function of t.

Justification:

For the given substitution t = tan( 21 x),

from the triangle below, we have
√
t 1 + t2
sin( 12 x) = √ , t
1 + t2
1 1
x
cos( 12 x) = √
2
, 1
1 + t2
so that
t 1 1
sin x = 2 sin( 21 x) cos( 12 x) = 2 √ ·√ = ,
1+t 2 1+t 2 1 + t2
1 1 − t2
cos x = 2 cos2 ( 12 x) − 1 = 2 · −1= .
1+t 2 1 + t2
Thus, under the substitution, the function R(sin x, cos x) becomes
1 1 − t2
 
R , ,
1 + t2 1 + t2
which is a rational function in t. Furthermore, since
2
dt = 1
sec2 ( 12 x) dx or dx = dt,
2 1 + t2
we get
ˆ ˆ
1 1 − t2 2
 
R(sin x, cos x) dx = R , · dt.
1 + t 1 + t2
2 1 + t2
This shows that the universal trigonometric substitution converts the given integral into some
integral of a rational function of t.
228 Integration

4.61 Example: integrating the secant function

ˆ
Calculate sec x dx.

Solution We notice that if we make the

substitution t = tan( 12 x), then, from the tri-
angle below, we have √
1 + t2
t
1
cos( 12 x) = √ ,
1 + t2
1
2
x
1

so that
1 1
sec x = =
cos x 2 cos2 ( 12 x) − 1
1 1 + t2
= = .
1 1 − t2
2· − 1
1 + t2
It is easy to have
2
dt = 1
sec2 ( 12 x) dx or dx = dt.
2 1 + t2
Thus, by applying the substitution rule, we get
ˆ ˆ
1 + t2 2
sec x dx = · dt
1 − t2 1 + t2
ˆ
2
= dt. rational integrand
1 − t2
For the last integral, we write the integrand as its form of the partial fraction decomposition:
2 1 1
= + .
1 − t2 1−t 1+t
Hence,
ˆ ˆ 
1 1

sec x dx = + dt
1−t 1+t
= − ln |1 − t| + ln |1 + t| + C
 1 + tan( 21 x) 
 
= ln 
  + C.
1 − tan( 1 x) 
2
4.6 Some Common Integrals 229

Rational Functions

4.62 Example: partial fractions for rational functions

ˆ 
1 1

1. Calculate the integral + dx.
x−2 x+1
We can use substitution method.
ˆ  ˆ ˆ
1 1 dx dx

+ dx = + splitting the integral
x−2 x+1 x−2 x+1
ˆ ˆ
du dv
= + letting u = x − 2 and v = x + 1
u v
= ln |u| + ln |v| + C evaluating integrals
= ln |x − 2| + ln |x + 1| + C. converting variables back to x
ˆ
2x − 1
2. Calculate the integral dx.
x2 −x−2
We notice that
2x − 1 1 1
= + ,
x2 − x − 2 x−2 x+1
so that the integral equals the previous one.

It seems a mystery to have the identity. However, there is an algorithm to derive this,
as demonstrated below.
Step 1 Factorize the denominator:

x2 − x − 2 = (x − 2)(x + 1).

Step 2 Write the partial fraction decomposition with coefficients to be determined:

2x − 1 A B
= + .
x2−x−2 x−2 x+1

Step 3 Clear the denominators:

2x − 1 A B
= +
x2 −x−2 x−2 x+1
A(x + 1) + B(x − 2) (A + B)x + A − 2B
= = .
x2 − x − 2 x2 − x − 2

Step 4 Solve for A and B by letting numerators of both sides equal:

2x − 1 = (A + B)x + A − 2B =⇒ A + B = 2, A − 2B = −1
=⇒ A = 1, B = 1.

The above process is a special case of the method of partial fractions.

230 Integration

4.63 Remark: integrals of rational functions

There are two facts worth knowing:

1. Any real rational function can be expressed as partial fractions;

2. Any partial fraction can be integrated.

Both can be effectively implemented on computers symbolically.

4.64 Method: integrating rational functions by partial fractions

For an improper rational function P (x)/Q(x), where P and Q are polynomials with deg P ≥
deg Q, one can use long division to reduce it to the form
P (x) R(x)
= S(x) + ,
Q(x) Q(x)
where S(x) and R(x) are polynomials and R(x)/Q(x) is a proper rational function with
deg R < deg Q. Hence, for integrating rational functions, it is essential to integrate proper
rational functions.
For any proper rational function, we can integrate it by using partial fractions in one of
the four cases:
1. The denominator Q has only distinct real roots.
In this case, we write
Q(x) = A(x − λ1 )(x − λ2 ) · · · (x − λk ).
In this case, a proper rational function R(x)/Q(x) can be expressed as
R(x) A1 A2 Ak
= + + ··· + .
Q(x) x − λ1 x − λ2 x − λk
For each fraction, we can make a substitution t = x − λ to integrate:
ˆ ˆ
1 1
dx = dt = ln |t| + C.
x−λ t
2. The denominator Q is a product of linear factors and some of them are repeated.
A1
Suppose the function (x−λ) is repeated r times. Then instead of the single term ,
x−λ
one needs to use the form
A1 A2 Ar
+ + ··· + .
x − λ (x − λ) 2 (x − λ)r
1
For each fraction of the form , we can make a substitution to integrate. In
(x − λ)m
fact, if m = 1, we can integrate the fraction as in Case 1. If m > 1, by making the
substitution t = x − λ, we get
ˆ ˆ
1 1 t−m+1
dx = dt = + C.
(x − λ)m tm −m + 1
4.6 Some Common Integrals 231

3. The denominator Q contains irreducible quadratic factors and none of them is repeated.
Suppose x2 + ax + b is an irreducible quadratic factor, with a2 − 4b < 0. Then the
partial fraction for the proper rational function R(x)/Q(x) will have a term of the form

Ax + B
x2 + ax + b
whose integration can be re-written as
ˆ ˆ
Ax + B 1
+ a) + (B − 12 aA)
2 A(2x
dx = dx
x + ax + b
2 x2 + ax + b
ˆ 1
2 A(x + ax + b) + (B − 2 aA)
2 0 1
= dx
x2 + ax + b
ˆ 1 ˆ
2 A(x + ax + b) (B − 21 aA)
2 0
= dx + dx.
x2 + ax + b x2 + ax + b
For the last two integrals on the right-hand side, they can be compute by using different
substitution. In fact, for the first integral, by using the substitution t = x2 + ax + b, it
can be reduced to ˆ
dt
= ln |t| + C;
t
for the second integral, by using the substitution t = x + 21 a, we can reduce the last
integral to the form ˆ
1 1
 
t
dt = tan −1
+ C.
t2 + α 2 α α

4. The denominator Q contains a repeated irreducible quadratic factor.

Suppose Q contains the factor (x2 + ax + b)r , where a2 − 4b < 0. Then instead of the
A1 x + B 1
single factor 2 , one needs to use the form
x + ax + b
A1 x + B 1 A2 x + B 2 Ar x + B r
+ + ··· + 2 .
x2 + ax + b (x2 + ax + b)2 (x + ax + b)r

As in Case 3, we re-write the integral of a typical term as

ˆ
Ax + B
dx
(ax + bx + c)m
2
ˆ 1 ˆ
2 A(ax + bx + c) (B − 21 aA)
2 0
= dx + dx.
(ax + bx + c)
2 m (ax2 + bx + c)m

In fact, if m = 1, we can integrate the fraction as in Case 3. If m > 1, the first integral
on the right-hand side, by using the substitution t = x2 + ax + b, it can be reduced to
ˆ
dt 1
=− + C;
t m (m − 1)tm−1
232 Integration

for the second integral, by using the substitution t = x + 12 a, we can reduce it to the
form ˆ
1
dt.
(t2 + α2 )m
By integrating by parts, we have
ˆ
1
dt
(t + α2 )m−1
2
ˆ
t
= 2 − t · (−m + 1)(t2 + α2 )−m · 2t dt
(t + α2 )m−1
ˆ
t
= 2 + 2(m − 1) (t2 + α2 )−m (t2 + α2 − α2 ) dt
(t + α2 )m−1
ˆ 
1 α2

t
= 2 + 2(m − 1) − 2 dt.
(t + α2 )m−1 (t2 + α2 )m−1 (t + α2 )m

So, we have the recursive formula

ˆ
1
dt
(t + α2 )m
2
ˆ
1 t 2m − 3 1
= + dt.
2(m − 1)α2 (t2 + α2 )m−1 2(m − 1)α2 (t2 + α2 )m−1

Using this formula recursively reduces the integral to the integral

ˆ
1 1
 
t
dt = tan −1
+ C.
t2 + α 2 α α

4.65 Example: partial fractions for various rational functions

ˆ
x6
1. Calculate the integral dx.
+1
x2
x6
Solution The integrand 2 is an improper rational function. By long division, we
x +1
have
x6 −1
= (x2 − x + 1) + 2 .
x +1
2 x +1
−1
The denominator x2 + 1 is an irreducible quadratic polynomial, so the term 2 is
x +1
in the form of partial fraction already. Hence,
ˆ ˆ 
x6

−1
dx = (x 2
− x + 1) + dx
x2 + 1 x2 + 1
ˆ ˆ
1
= (x2 − x + 1) dx − dx
x2 + 1
= 31 x3 − 12 x2 + x − tan−1 x = C.
4.6 Some Common Integrals 233

ˆ
2x2 + 1
2. Calculate the integral dx.
(x + 1)(x − 1)3
Solution We express the proper rational integrand as its partial fraction decomposition:

2x2 + 1 A1 B1 B2 B3
= + + + .
(x + 1)(x − 1) 3 x + 1 x − 1 (x − 1) 2 (x − 1)3

By clearing the denominators, we get

2x2 + 1 = A1 (x − 1)3 + B1 (x + 1)(x − 1)2 + B2 (x + 1)(x − 1) + B3 (x + 1).

So,
x = −1 : 3 = A1 · (−2)3 =⇒ A1 = − 83 ;
x=1: 3 = B3 · 2 =⇒ B3 = 32 ;
differentiating then x = 1 : 4 = B2 · 2 + B3 =⇒ B2 = 45 ;
coefficients of x3 : 0 = A1 + B 1 =⇒ B1 = 83 .
Thus, we have the partial fraction decomposition:

2x2 + 1 1 1 1 1
= − 38 · + 3
· + 5
· + 3
· .
(x + 1)(x − 1) 3 x+1 8 x−1 4 (x − 1)2 2 (x − 1)3

Hence,
ˆ
2x2 + 1
dx
(x + 1)(x − 1)3
ˆ 
1 1 1 1

= − 38 · +3· + 5
· +2·
3
dx
x+1 8 x−1 4 (x − 1)2 (x − 1)3
1 1
= − 83 ln |x + 1| + 3
ln |x − 1| − 5
· −3· + C.
8 4 x − 1 4 (x − 1)2
ˆ
x
3. Calculate the integral dx.
x3 + 1
Solution We express the proper rational integrand as its partial fraction decomposition:
x x A1 B1 x + C1
= = + .
x3 + 1 (x + 1)(x2 − x + 1) x + 1 x2 − x + 1

By clearing the denominators, we get

x = A1 (x2 − x + 1) + (B1 x + C1 )(x + 1).

So,
x = −1 : −1 = A1 · 3 =⇒ A1 = − 31 ;
x=0: 0 = A1 + C1 =⇒ C1 = 13 ;
coefficients of x2 : 0 = A1 + B 1 =⇒ B1 = 31 .
234 Integration

Thus, we have the partial fraction decomposition:

x 1 1
x + 13
= − 13 · + 23 .
x3 +1 x+1 x −x+1
Note that (x2 − x + 1)0 = 2x − 1, so we further write the integrand as

1 6 (2x − 1) + 2
1 1
x
= − 1
· + .
x3 + 1 3 x+1 x2 − x + 1
Hence,
ˆ
x
dx
x3 + 1
ˆ 
1 6 (2x − 1) + 2
1 1
= 1
−3 · + dx
x+1 x2 − x + 1
ˆ
dx
= − 31 ln |x + 1| + 16 ln(x2 − x + 1) + 12 .
x2 − x + 1
The last integral can be integrated as
ˆ ˆ
dx dx
= completing square
x −x+1
2 (x − 12 )2 + 43
ˆ
q
4 sec t dt
3 2 q
= x − 1 = 4 tan t
3
3
sec 2t
ˆ q 4

= 3 dt
4

q q q
= 43 t + C = 43 tan−1 43 (x − 1) + C.

Therefore,
ˆ
x 2x − 1
dx = − 13 ln |x + 1| + 1
ln(x2 − x + 1) + √1 tan−1 √ + C.
x3 +1 6 3 3
ˆ
x2 + 1
4. Calculate the integral dx.
x4 + 1
Solution The denominator x4 + 1 can be factored as a product of irreducible quadratic
factors:

x4 + 1 = (x4 + 2x2 ) + 1 − 2x2

√
= (x2 + 1)2 − ( 2x)2
√ √
= (x2 − 2x + 1)(x2 + 2x + 1).

Thus, we can express the proper rational integrand as its partial fraction decomposition:
x2 + 1 A1 x + B 1 A2 x + B 2
= √ + √ .
x4 + 1 x − 2x + 1 x + 2x + 1
2 2
4.6 Some Common Integrals 235

By clearing the denominators, we get

√ √
x2 + 1 = (A1 x + B1 )(x2 + 2x + 1) + (A2 x + B2 )(x2 − 2x + 1).

So,
x=0: 1 = B1 + B2 ;
coefficients of x3 : 0 = A1 + A2 ;
√ √
coefficients of x : 0 = A1 + 2B1 + A2 − 2B2 ;
√ √
coefficients of x2 : 1 = B1 + 2A1 + B2 − 2A2 .

Solving the system gives

A1 = 0, A2 = 0, B1 = 12 , B2 = 21 .

Thus, we have the partial fraction decomposition:

x2 + 1 1 1
= √2 + √2 .
x +1
4
x2 − 2x + 1 x2 + 2x + 1
Since
ˆ ˆ
dx dx
√ = completing square
x2 − 2x + 1 (x − √1 )2
2
+ 1
2
ˆ √1 sec t dt2
2
= x− √1 = √1 tan t
2 sec t
1 2 2 2
ˆ
√
= 2 dt
√ √ √
= 2t + C = 2 tan−1 ( 2x − 1) + C,

similarly, ˆ
dx √ √
√ = 2 tan−1 ( 2x + 1) + C,
x2 − 2x + 1
Therefore,
ˆ
x2 + 1 √ √
dx = √1 tan−1 ( 2x − 1) + √1 tan−1 ( 2x + 1) + C.
x4 + 1 2 2

4.66 MATLAB: partial fractions for rational functions

For integrating rational functions, one important step is to perform the partial fraction
decomposition for the rational functions. In MATLAB, the partial fraction decomposition
can be executed by using the command partfrac. The following MATLAB code shows how
2x2 + 1
to perform the partial fraction decomposition for the rational functions and
(x + 1)(x − 1)3
236 Integration

x
, respectively.
x3 + 1

>> syms x
>> f ( x ) = (2* x ^2 + 1) /(( x + 1) *( x - 1) ^3) ;
>> partfrac ( f ( x ) ,x )

ans =

3/(8*( x - 1) ) - 3/(8*( x + 1) ) + 5/(4*( x - 1) ^2) + 3/(2*( x - ...

1) ^3)

>> g ( x ) = x /( x ^3 + 1) ;
>> partfrac ( g ( x ) ,x )

ans =

( x /3 + 1/3) /( x ^2 - x + 1) - 1/(3*( x + 1) )

´ Based
2x2 +1
on the above
´ xresults of the partial fraction decomposition, the integrals of
(x+1)(x−1) 3 dx and x3 +1 dx can be solved by utilizing some integration methods such
as the substitution method.

4.7 Numerical Approximation of Integrals

4.67 Example: numerical integration

ˆ 1
2
Consider the integral ex dx.
0
2
We know that there is a function F (x) whose derivative is ex and the integral above equals
F (1) − F (0). However, there is no explicit expression of F (x) in terms of known functions
such as polynomials, trigonometric functions and their inverses, and transcendental functions
and their inverses. We can use the definition of definite integrals to approximate the integral.
2
Recall the definition of definite integrals. Let f (x) = ex . Divide the interval [0, 1] into n
equal subintervals:
1 n−1 n
x0 = 0, x1 = , . . . , xn−1 = , xn = = 1.
n n n
To approximate the integral numerically, one can use the Riemann sum of left end points:
ˆ 1
2 1
ex dx ≈ Ln = [f (x0 ) + f (x1 ) + · · · + f (xn−1 )] · ,
0 n
4.7 Numerical Approximation of Integrals 237

or the Riemann sum of right end points:

ˆ 1
2 1
ex dx ≈ Rn = [f (x1 ) + f (x2 ) + · · · + f (xn )] · .
0 n

Or, one can use the Riemann sum of middle points. In fact, if we let
x0 + x1 1 x1 + x2 3 xn−1 + xn 2n − 1
x∗1 = = , x∗2 = = , . . . , x∗n = =
2 2n 2 2n 2 2n
be the middle points of each subinterval, then the Riemann sum of middle points is
1
Mn = [f (x∗1 ) + f (x∗2 ) + · · · + f (x∗n )] · .
n
For instance, if one takes n = 10, the approximation gives:
ˆ 1
2
ex dx ≈ M10 = f ( 20 ) + f ( 20 ) + · · · + f ( 19
20 ) ·
 1 3
 1
10
0
= e0.0025 + e0.0225 + · · · + e0.9025 · 0.1

≈ 1.460393.

Or one can approximate the region under the graph of the function f on [xi−1 , xi ] by a
trapezoidal region, equivalently, the average of Riemann sums of left end points and right
end points:
Ln + Rn 1
Tn = = [f (x0 ) + 2f (x1 ) + · · · + 2f (xn−1 ) + f (xn )] .
2 2n

4.68 Method: numerical integration

Let f be integrable function over [a, b]. Divide the interval [a, b] into n equal subintervals

b−a (n − 1)(b − a)
x0 = a, x1 = a + , . . . , xn−1 = a + , xn = b.
n n
• The left point rule is the Riemann sum y
of left end points of f on [a, b]. The re-
gion under the graph of the function f on
[xi−1 , xi ] is approximated by a rectangle.
ˆ b
f (x) dx ≈ Ln
a
x
b−a x0 x1 xi−1 xi xn
= [f (x0 ) + f (x1 ) + · · · + f (xn−1 )] · .
n left point rule
238 Integration

• The right point rule is the Riemann sum y

of right end points of f on [a, b]. The re-
gion under the graph of the function f on
[xi−1 , xi ] is approximated by a rectangle.
ˆ b
f (x) dx ≈ Rn
a
x
b−a x0 x1 xi−1 xi xn
= [f (x1 ) + f (x2 ) + · · · + f (xn )] · .
n right point rule

• The trapezoidal rule is the average of y

the Riemann sums of left end points and
right end points. The region under the
graph of the function f on [xi−1 , xi ] is ap-
proximated by a trapezoid.
ˆ b
f (x) dx ≈ Tn x
a x0 x1 xi−1 xi xn

b−a
= [f (x0 ) + 2f (x1 ) + · · · + 2f (xn−1 ) + f (xn )] · . trapezoidal rule
2n

• The midpoint rule is the Riemann sum y

of middle points of f on [a, b]. The re-
gion under the graph of the function f on
[xi−1 , xi ] is approximated by a rectangle.
ˆ b
f (x) dx ≈ Mn
a
b−a x0 x1 xi−1 xi xn x
= [f (x∗1 ) + f (x∗2 ) + · · · + f (x∗n )] · , x̄i
n midpoint rule
xi−1 + xi
where x∗i = .
2

4.69 Example: various numerical integrations and their errors

We use various rules to estimate the integral

ˆ 1
dx
I= .
0 1+x
2

Take n = 4. The partition gives four subintervals

[0, 0.25], [0.25, 0.5], [0.5, 0.75], [0.75, 1].

4.7 Numerical Approximation of Integrals 239

Each subinterval has length 0.25. The values at the partition points are

y0 = 1.000000, y1 = 0.941176, y2 = 0.800000,

y3 = 0.640000, y4 = 0.500000.

Then we get the approximate values of the integral I according to various rules

R4 = 0.25 × (0.941176 + 0.800000 + 0.640000 + 0.500000) ≈ 0.720294,

L4 = 0.25 × (1.000000 + 0.941176 + 0.800000 + 0.640000) ≈ 0.845294,
L4 + R4 0.25
T4 = = × (1.000000 + 2 × 0.941176 + 2 × 0.800000
2 2
+2 × 0.640000 + 0.500000)

≈ 0.782794.

The middle points are

0.125 < 0.375 < 0.625 < 0.875,
and the values at the middle points are

ȳ1 = 0.984615, ȳ2 = 0.876712, ȳ3 = 0.719101, ȳ4 = 0.566372.

Then the midpoint rule gives

M4 = 0.25 × (0.984615 + 0.876712 + 0.719101 + 0.566372) ≈ 0.786700.

If we double the number of partition points by taking n = 8, then the length of each
subinterval is 0.125, and we have

i xi yi x̄i ȳi
0 0 1.000000
1 0.125 0.984615 0.0625 0.996109
2 0.250 0.941176 0.1875 0.966038
3 0.325 0.876712 0.3125 0.911032
4 0.500 0.800000 0.4375 0.839344
5 0.625 0.719101 0.5625 0.759644
6 0.750 0.640000 0.6875 0.679045
7 0.875 0.566372 0.8125 0.602353
8 1.000 0.500000 0.9375 0.532225

Then we get the approximations

L8 ≈ 0.815997, R8 ≈ 0.753497, T8 ≈ 0.784747, M8 ≈ 0.785721.

We actually know the exact value of the integral

ˆ 1
dx
I= = π4 ≈ 0.785398.
0 1 + x2
240 Integration

We may compare the approximate values with the actual value in the following table.

error n=4 n=8

I − Ln −0.059896 −0.030599
I − Rn 0.065104 0.031901
I − Tn 0.002604 0.000651
I − Mn −0.001302 −0.000323

4.70 Theorem: error bounds of midpoint and trapezoidal rules

Suppose f 00 is continuous and |f 00 (x)| ≤ K on [a, b]. Then

ˆ  ˆ 
 b  K(b − a)3  b  K(b − a)3
f (x) dx − Tn  ≤ , f (x) dx − M ≤ .
   
n
12n 2 24n2
 
 a   a 

4.71 Remark: error bounds

We can make the following observations from the theorem.

1. Both midpoint rule and trapezoidal rule produce better approximation as n increases.
2. Generally, the error of the midpoint rule is about half of the error of the trapezoidal
rule.

4.72 Example: error bound of trapezoidal rule

Consider the integral ˆ 1

2 cos x dx.
0
How large n should we take so that trapezoidal rule is accurate within 0.0001?
Since |(2 cos x)00 | = | − 2 cos x| ≤ 2, from the error bound, we know that the number n
should satisfy
2
< 0.0001,
12n2
1
so that n > √ ≈ 40.8.
0.0006
Hence, if we take an integer n ≥ 41, then the trapezoidal rule will yield a numerical value
that approximates the given integral with error smaller than 0.0001.

4.73 Method: Simpson’s Rule

Let f be integrable function over [a, b]. For an even integer n, divide the interval [a, b] into
n equal subintervals

b−a (n − 1)(b − a)
x0 = a, x1 = a + , . . . , xn−1 = a + , xn = b.
n n
4.7 Numerical Approximation of Integrals 241

ˆ b
Simpson’s rule is the approximation to the integral f (x) dx:
a
ˆ b
b−a
f (x) dx ≈ Sn = · [f (x0 ) + 4f (x1 ) + 2f (x2 ) + 4f (x3 ) + · · ·
a 3n
+2f (xn−2 ) + 4f (xn−1 ) + f (xn )] .

Justification:

(1) Let us first derive a formula for the area under a parabola of equation y = ax2 + bx + c
passing through the three points (−h, y0 ), (0, y1 ), and (h, y2 ).

y The area is
ˆ h
y = ax2 + bx + c
(0, y1 ) (h, y2 ) A= (ax2 + bx + c) dx
−h
h
= + 12 bx2 + cx
1 3
(−h, y0 ) 3 ax x=−h
= 2
3 ah
3
+ 2ch = 13 h(2ah2 + 6c).
x
−h O h

(2) Since the parabola passes through (−h, y0 ), (0, y1 ), and (h, y2 ), we have

y0 = a(−h)2 + b(−h) + c,
y1 = c,
y2 = ah2 + bh + c.

It is easy to see that

y0 + 4y1 + y2 = a(−h)2 + b(−h) + c + 4c + (ah2 + bh + c) = 2ah2 + 6c,

 

which is the same the area under the parabola, so

A = 31 h(y0 + 4y1 + y2 ).

This formula shows that the area under the parabola depends only on the values of y0 , y1 , y2 ,
and h. It is clear that shifting this parabola horizontally does not change the area under it.
ˆ b
(3) For an integral f (x) dx, we divide [a, b] into an even number n of subintervals of
a
b−a
equal length ∆x = by using n + 1 points
n
x0 = a, x1 = a + ∆x, x2 = a + 2∆x, ..., xn = a + n∆x = b.

For each consecutive pair of subintervals, we apply the area under the parabola. Adding all
242 Integration

the areas gives Simpson’s rule:

ˆ b
f (x) dx ≈ 13 h(y0 + 4y1 + y2 ) + 31 h(y2 + 4y3 + y4 )
a
+ · · · + 13 h(yn−2 + 4yn−1 + yn )
= 13 h(y0 + 4y1 + 2y2 + 4y3 + 2y4 + · · · + 2yn−2 + 4yn−1 + yn )
= Sn .

4.74 Theorem: error bound of Simpson’s rule

Suppose f (4) is continuous and f (4) (x) ≤ K on [a, b]. Then

 

ˆ 
 b  K(b − a)5
f (x) dx − Sn  ≤ .
 
180n4

 a 

4.75 Remark: Simpson’s rule

1. In Simpson’s rule, the integrand f is approximated by a piecewise quadratic function.

In comparison, in the left point rule, the right point rule, the trapezoidal rule, or the
midpoint rule, the function f is approximated by a piecewise linear function.
2. Simpson’s rule is related to other rules by

S2n = 13 Tn + 32 Mn .

3. The error of Simpson’s rule is smaller than the error in the trapezoidal rule. In other
words, Simpson’s rule generally yields a better result in evaluating integrals.

4.76 Example: error bound of Simpson’s rule

Consider the integral ˆ 1

2 cos x dx.
0
How large n should we take so that Simpson’s rule is accurate to within 0.0001?
Thus |(2 cos x)(4) | = |2 cos x| ≤ 2. The number n should satisfy
2
< 0.0001,
180n4
1
so that n > √ ≈ 3.25.
4
0.009
Hence, if we take an integer n ≥ 4, then the Simpson’s rule will yield a numerical value
that approximates the given integral with error smaller than 0.0001.
4.7 Numerical Approximation of Integrals 243

4.77 MATLAB: left point rule

The left-point rule of approximating an integral can be executed in MATLAB

ˆ 1 via program-
dx
ming. In particular, the following MATLAB code shows the calculation of using
0 1 + x2
the left-point rule.

Ln = 0;
a = 0; % interval start
b = 1; % interval end
n = 4; % number of subintervals
itv = ( b - a ) / n ; % subinterval length

x = linspace (a ,b , n + 1) ; % n subintervals , n + 1 points

y = 1./(1 + x .^2) ;

for k = 1: n % y value index

Ln = Ln + itv * y ( k ) ;
end

In the MATLAB code above, the interval [0, 1] is partitioned into 4 subintervals. After
executing the above MATLAB code, the result can be seen by typing Ln in the command
window, as shown in the following.

>> Ln

Ln =

0.8453

4.78 MATLAB: right point rule

The right-point rule of approximating an integral can be executed in MATLAB

ˆ 1via pro-
dx
gramming. In particular, the following MATLAB code shows the calculation of
0 1 + x2
using the right-point rule.

Rn = 0;
a = 0; % interval start
b = 1; % interval end
n = 4; % number of subintervals
itv = ( b - a ) / n ; % subinterval length

x = linspace (a ,b , n + 1) ; % n subintervals , n + 1 points

244 Integration

y = 1./(1 + x .^2) ;

for k = 1: n % y value index

Rn = Rn + itv * y ( k +1) ;
end

In the MATLAB code above, the interval [0, 1] is partitioned into 4 subintervals. After
executing the above MATLAB code, the result can be seen by typing Rn in the command
window, as shown in the following.

>> Rn

Rn =

0.7203

4.79 MATLAB: midpoint rule

The midpoint rule of approximating an integral can be executed in MATLAB

ˆ 1 via program-
dx
ming. In particular, the following MATLAB code shows the calculation of using
0 1 + x2
the midpoint rule.

Mn = 0;
a = 0; % interval start
b = 1; % interval end
n = 4; % number of subintervals
itv = ( b - a ) / n ; % subinterval length

x = linspace (a ,b , n + 1) ; % n subintervals , n + 1 points

for k = 1: n % y value index

x_m = ( x ( k ) + x ( k + 1) ) /2; % midpoint
y_m = 1/(1 + x_m ^2) ; % y value at midpoint
Mn = Mn + itv * y_m ;
end

In the MATLAB code above, the interval [0, 1] is partitioned into 4 subintervals. After
executing the above MATLAB code, the result can be seen by typing Mn in the command
window, as shown in the following.

>> Mn

Mn =
4.7 Numerical Approximation of Integrals 245

0.7867

4.80 MATLAB: trapezoidal rule

In MATLAB, the command trapz can be used to compute the approximation of an integral
using the trapezoidal rule.
ˆ 1 The following MATLAB code shows an example of how to use
dx
trapz to approximate .
0 1 + x2

>> x = 0:0.25:1; % 4 subintervals with length 0.25 each

>> y = 1./(1 + x .^2) ;
>> T4 = trapz (x , y )

T4 =

0.7828

In the MATLAB code above, the interval [0, 1] is partitioned into 4 subintervals of equal
length 0.25. Note that if there is only one input in trapz, i.e., the command is written as
trapz(y), then the length of the subinterval (or the gap between two neighbouring points)
will be set to the default value 1.

4.81 MATLAB: Simpson’s rule

ˆ 1
dx
The following MATLAB code shows the calculation of using Simpson’s rule.
0 1 + x2

Sn = 0;
a = 0; % interval start
b = 1; % interval end
n = 4; % even number of subintervals
itv = ( b - a ) / n ; % subinterval length

x = linspace (a ,b , n + 1) ; % n subintervals , n + 1 points

y = 1./(1 + x .^2) ;

for k = 1:2:( n -1) % y value index

sum_term = y ( k ) + 4* y ( k + 1) + y ( k + 2) ;
Sn = Sn + itv * sum_term /3;
end

In the MATLAB code above, the interval [0, 1] is partitioned into 4 subintervals. After
executing the above MATLAB code, the result can be seen by typing Sn in the command
window, as shown in the following.
246 Integration

>> Sn

Sn =

0.7854

4.82 Engineering Example: error function

The error function is an important function in statistics, applied mathematics, physics, etc.
For example, it is useful in determining the bit error rate (BER) of a digital communication
system. The error function is defined as
ˆ x
2 2
erf(x) = √ e−t dt.
π 0

Use the left point rule, the right point rule, the midpoint rule, the trapezoidal rule and the
Simpson’s rule, respectively, to approximate erf(3).
First, erf(3) can be written as
ˆ
2 3
2
erf(3) = √ e−t dt.
π 0

Take n = 10. The partition gives ten subintervals as follows:

[0, 0.3], [0.3, 0.6], [0.6, 0.9], [0.9, 1.2], [1.2, 1.5],
[1.5, 1.8], [1.8, 2.1], [2.1, 2.4], [2.4, 2.7], [2.7, 3].
2
Each subinterval has length 0.3. Let y = f (t) = √2 e−t .
π
The values of y = f (t) at the
partition points are

y0 = 1.128379167095513, y1 = 1.031260909618963,
y2 = 0.787243431714287, y3 = 0.501968574240363,
y4 = 0.267344347003539, y5 = 0.118930289223629,
y6 = 0.044191723332011, y7 = 0.013715649999807,
y8 = 0.003555648680878, y9 = 0.000769924759855,
y10 = 0.000139253051947,

where yi denotes the value of y = f (t) at the ith partition point. Then we get the approximate
values of erf(3) according to various rules. In particular, by the left point rule, the right point
4.7 Numerical Approximation of Integrals 247

rule, the trapezoidal rule and the Simpson’s rule, respectively, we have
9
X
L10 = 0.3 yi = 1.169207899700653,
i=0
10
X
R10 = 0.3 yi = 0.830735925487584,
i=1
9
0.3 X
T10 = (yi + yi+1 ) = 0.999971912594118,
2 i=0
0.3
S10 = (y0 + 4y1 + 2y2 + 4y3 + · · · + 2y8 + 4y9 + y10 )
3
= 0.999977011297936.

Before using the midpoint rule to approximate erf(3), we first need to get the middle points
of each interval:

x∗1 = 0.15, x∗2 = 0.45, x∗3 = 0.75, x∗4 = 1.05, x∗5 = 1.35,
x∗6 = 1.65, x∗7 = 1.95, x∗8 = 2.25, x∗9 = 2.55, x∗10 = 2.85.

The corresponding values of y = f (t) at each middle point are

f (x∗1 ) = 1.103274126650824, f (x∗2 ) = 0.921532013012220,

f (x∗3 ) = 0.642931069195207, f (x∗4 ) = 0.374666956970448,
f (x∗5 ) = 0.182369986539023, f (x∗6 ) = 0.074146103374068,
f (x∗7 ) = 0.025179684949841, f (x∗8 ) = 0.007142319022018,
f (x∗9 ) = 0.001692213637679, f (x∗10 ) = 0.000334886877468.

The approximation of erf(3) by the midpoint rule is thus

10
X
M10 = 0.3 f (x∗i ) = 0.999980808068639.
i=1

Actually, the exact value of erf(3) can be calculated by MATLAB using command erf, as
shown below:

>> y = erf (3)

y =

0.9 999779 095030 01

248 Integration

We may compare the approximate values with the actual value in the following table.

error n = 10
y − Ln −0.169229990197652
y − Rn 0.169241984015417
y − Tn 5.996908883365570 × 10−6
y − Sn 8.982050653250084 × 10−7
y − Mn −2.898565637621253 × 10−6

4.8 Improper Integrals

4.83 Example: improper integrals over unbounded intervals

1
Consider the area of the graph of the function f (x) = above the x-axis over the interval
x2
[1, +∞). Since the interval [1, +∞) is unbounded, we can consider the area over a finite
interval [1, T ], and let T get larger and larger to approximate the required area. We need to
calculate ˆ T T
1 1  1

2
dx = −  =1− T.
1 x x 1
Thus the required area is
ˆ T
1 1
 
lim dx = lim 1− = 1.
T →+∞ 1 x2 T →+∞ T

We can use the following notation to denote this area, called an improper integral
ˆ ˆ
+∞
1 T
1
dx = lim dx.
1 x2 T →+∞ 1 x2

1
If we want to know the area of the graph of the function f (x) = above the x-axis over
x
the interval [1, +∞), we need to calculate the improper integral
ˆ ˆ
+∞
1 T
1
dx = lim dx = lim (ln T − 1) = +∞.
1 x T →+∞ 1 x T →+∞

Thus the required area is infinite.

Comparing these two improper integrals, the former is a finite number and the latter is
infinite. Since the improper integrals are in fact the limits of some integrals, we call the
former convergent and the latter divergent.
4.8 Improper Integrals 249

4.84 Definition: improper integrals over infinite intervals

Suppose f is a function defined over an in- y

terval [a, +∞) such that f is integrable over
[a, b] for any b > a. The improper integral
is defined as a limit of a definite integral as
follows:
ˆ +∞ ˆ b
f (x) dx = lim f (x) dx. a x
a b→+∞ a b

We say the improper integral is convergent if the limit exists, otherwise divergent.

Similarly, we define ˆ ˆ
a a
f (x) dx = lim f (x) dx.
−∞ b→−∞ b
ˆ ∞ ˆ a
If f (x) dx and f (x) dx are convergent, then we define
a −∞
ˆ ∞ ˆ a ˆ ∞
f (x) dx = f (x) dx + f (x) dx,
−∞ −∞ a

where a is any fixed real number.

4.85 Example: improper integral over unbounded interval

ˆ 0
Consider the improper integral xex dx.
−∞

b −1
x
O

y = xex
−1

ˆ 0
By definition, we first calculate the definite integral xex dx on a finite interval. From
b
integration by parts, we have
ˆ 0 ˆ 0 ˆ 0
0 0
xex dx = x d(ex ) = xex |b − ex dx = −beb − ex |b = −beb − 1 + eb .
b b b

Thus, ˆ 0
xex dx = lim (−beb − 1 + eb ) = −1.
−∞ b→−∞
250 Integration

Hence this improper integral is convergent to −1.

4.86 Example: divergent improper integral over unbounded interval

ˆ 0
Consider the improper integral cos x dx.
−∞

y
y = cos x
1
b
x
O
−1

ˆ 0
By definition, we first calculate the definite integral cos x dx on a finite interval. From
b
integration by parts, we have
ˆ 0
0
cos x dx = sin x|b = − sin b.
b

Thus, ˆ 0
cos x dx = − lim sin b,
−∞ b→−∞

that is, the given improper integral is divergent.

4.87 Example: convergence/divergence of improper integral

ˆ +∞
1
Consider the improper integral dx (p > 0, p 6= 1).
1 xp

y y

1 x 1
y y== cos
(p > 1) y= (0 < p < 1)
xp xp
1 1

b
x x
O 1 b O 1 b

ˆ b
1
By definition, we first calculate the definite integral dx on a finite interval:
1 xp
ˆ b
b
1 x1−p  b1−p 1
dx = = − .
1 xp 1−p 1  1−p 1−p
4.8 Improper Integrals 251

Thus,
ˆ +∞
1 b1−p 1 1
 
dx = lim − = , when p > 1;
1 xp b→+∞ 1−p 1−p p−1
ˆ +∞
1 b1−p 1
 
dx = lim − = ∞, when p < 1.
1 xp b→+∞ 1−p 1−p

So, when p > 1, the given improper integral is convergent; when p < 1, it is divergent.

4.88 Definition: improper integrals with unbounded integrands

Suppose f is a function defined over an in- y

terval [a, b) such that f is integrable over [a, c]
for any c with a < c < b, and suppose f is
unbounded in [c, b). We define the improper
integral as a limit of a definite integral as fol-
lows:
ˆ b ˆ c x
a c b
f (x) dx = lim f (x) dx.
a c→b a

We say the improper integral is convergent if the limit exists, otherwise divergent.

Similarly, if f is unbounded near a, we define

ˆ b ˆ b
f (x) dx = lim f (x) dx.
a c→a c
ˆ c ˆ b
If f is unbounded near a and b, and if f (x) dx and f (x) dx are convergent, then we
a c
define ˆ ˆ ˆ
b c b
f (x) dx = f (x) dx + f (x) dx,
a a c
where c is any fixed real number with a < c < b.

4.89 Example: improper integral with unbounded integrand

ˆ 1
Consider the improper integral ln x dx. The integrand ln x becomes unbounded near
0
x = 0.
252 Integration

c 1
x
O

y = xex
−1

ˆ 1
By definition, we first calculate the definite integral ln x dx on [c, 1]. From integration
c
by parts, we have
ˆ 1 ˆ 1
1 1
ln x dx = x ln x|c − 1 dx = −c ln c − x|c = −c ln c − 1 + c.
c c

Thus, ˆ 1
ln x dx = lim (−c ln c − 1 + c) = −1.
0 c→0

Here we used l’Hospital’s Rule and had lim (−c ln c) = 0. Hence, the given improper integral
c→0
is convergent to −1.

4.90 Example: convergence/divergence of improper integrals

ˆ 1
Consider the improper integral xp dx (p < 0, p 6= −1). The integrand xp becomes
0
unbounded near x = 0.

y y

y = xp (p < −1) y = xp (−1 < p < 0)

1 1

c x x c x x
O 1 O 1
4.8 Improper Integrals 253

ˆ 1
By definition, we first calculate the definite integral xp dx on [c, 1]:
c

ˆ 1
1
xp+1  1 cp+1
x dx =
p
= − .
c p+1 c  p+1 p+1

Thus,
ˆ 1
1 cp+1 1
 
xp dx = lim − = , when −1 < p < 0;
0 c→0 p+1 p+1 p+1
ˆ 1
1 cp+1
 
x dx = lim
p
− = +∞, when p < −1.
0 c→0 p+1 p+1

So, when −1 < p < 0, the given improper integral is convergent; when p < −1, it is
divergent.

4.91 Pitfall: improper integrals

Consider the integral ˆ 1

1
dx.
−1 x
If one mistakenly thought this was a definite integral, he would have calculated as follows
ˆ 1
1 1
dx = ln |x|−1 = ln 1 − ln 1 = 0.
−1 x

1
This is incorrect since the integral is an improper integral, for being unbounded near
x
0 ∈ [−1, 1].
We could write a sum of two improper integrals
ˆ 1 ˆ 0 ˆ 1
1 1 1
dx = dx + dx.
−1 x −1 x 0 x

One may naively calculate the improper integral as limits of the following definite integrals,
for b > 0,
ˆ −b ˆ
1 1
1 −b 1
dx + dx = ln |x|−1 + ln |x|b = ln b − ln 1 + ln 1 − ln b = 0.
−1 x b x

So, one might conclude that the improper integral is zero. In fact, this is a wrong answer
again.
The correct way is to apply the definition. For each of the two improper integrals,
ˆ 0 ˆ 1
1 1
dx and dx,
−1 x 0 x
254 Integration

one should first evaluate the corresponding definite integral. Indeed, for −1 < a < 0 and
0 < b < 1, one gets
ˆ a ˆ 1
1 1 a 1
dx + dx = ln |x| + ln |x| = ln |a| − ln 1 + ln 1 − ln b.
 
−1 x b x −1 b

Clearly when a →ˆ 0 and b → 0, the expression above diverges. Therefore, the original
1
1
improper integral dx is divergent.
−1 x
Similar results also hold for other types of improper integrals.

4.92 Theorem: Comparison Test for improper integrals

Let f and g be continuous over [a, +∞) and 0 ≤ f (x) ≤ g(x). Then
ˆ +∞ ˆ +∞
1. If g(x) dx converges, then f (x) dx converges.
a a
ˆ +∞ ˆ +∞
2. If f (x) dx diverges, then g(x) dx diverges.
a a

Similar results hold for other types of improper integrals.

4.93 Example: Comparison Test for convergence

ˆ +∞
1 + sin x
Consider the improper integral dx.
1 x2
Since we have the inequality y

1 + sin x 2
2
≤ 2, 2
x x
and
ˆ +∞ we know the improper integral 2
1 y=
2 x2

2
dx converges, the improper integral
x 1 + sin x
ˆ1 +∞ y=
1 + sin x x2
x x
dx converges. O 1
1 x2

4.94 MATLAB: improper integrals over infinite intervals

In MATLAB, the commands integral, quadgk and int support calculating improper
integrals over infinite intervals, i.e., the input which specifies theˆ interval of the integral can
0
be -Inf or Inf. The following example shows the calculation of xex dx using integral,
−∞
quadgk and int, respectively.
4.8 Improper Integrals 255

>> integral_ans = integral (@( x ) x .* exp ( x ) ,- Inf ,0)

integral_ans =

-1

>> quadgk_ans = quadgk (@( x ) x .* exp ( x ) ,- Inf ,0)

quadgk_ans =

-1.0000

>> syms t
>> f ( t ) = t * exp ( t ) ;
>> int_ans = int ( f ( t ) ,- Inf ,0)

int_ans =

-1

Note that the first input of the commands integral and quadgk should be a function handle,
and the first input of the command int is a symbolic expression.

4.95 MATLAB: improper integral with unbounded integrand

For convergent improper integrals with unbounded integrands, we can use the commands
integral, quadgk and int to calculate the results. The following example shows the calcu-
ˆ 1
lation of x−0.5 dx in MATLAB.
0

>> integral_ans = integral (@( x ) x .^( -0.5) ,0 ,1)

integral_ans =

2.0000

>> quadgk_ans = quadgk (@( x ) x .^( -0.5) ,0 ,1)

quadgk_ans =

2.0000

>> syms t
>> f ( t ) = t ^( -0.5) ;
>> int_ans = int ( f ( t ) ,0 ,1)
256 Integration

int_ans =

However, for divergent improper integrals with unbounded integrands,

ˆ 1 we need ˆto1 use the
1
command int. The following example shows the calculations of dx and x−2 dx
−1 x 0
using int, respectively.

>> syms t
>> f ( t ) = 1/ t ;
>> int_ans = int ( f ( t ) , -1 ,1)

int_ans =

NaN

>> f ( t ) = t ^( -2) ;
>> int_ans = int ( f ( t ) ,0 ,1)

int_ans =

Inf
ˆ 1 ˆ 1
1 1
It shows that the result of dx is NaN, i.e., Not-a-Number, which means dx is
x x
ˆ 1 −1
ˆ 1 −1

divergent. The result of x−2 dx is Inf, i.e., ∞, which also means x−2 dx is divergent.
0 0

4.96 Engineering Example: Fourier transform

Mathematically, the Fourier transform fˆ(ξ) (ξ ∈ R) of an integrable function f (x) is defined

as
ˆ ∞
fˆ(ξ) = f (x)e−2πixξ dx,
−∞

where f (x) is in the time domain, and fˆ(ξ) is in the frequency domain.
The Fourier transform decomposes a function in the time domain into the frequency-
domain components which make it up. It allows us to analyze a signal in the time domain
from a different perspective, i.e., in the frequency domain. The Fourier transform has wide
applications in communication engineering, differentiation equation analysis area, quantum
mechanics, etc. For example, the Fourier transform is used to help with the deduction of
noise in a communication system. Moreover, the Fourier transform can facilitate the analysis
of partial differential equations.
4.8 Improper Integrals 257

4.97 Engineering Example: Laplace transform

The Laplace transform takes a function of a real variable t (e.g., time) to a function of
a complex variable s (e.g., frequency). Mathematically, the Laplace transform f (s) of an
integrable function f (t) (ξ ∈ R+ ) is defined as follows
ˆ ∞
ˆ
f (s) = f (t)e−st dt,
0

where f (t) is in the time domain, and f (s) is in the frequency domain.
Similar to the Fourier transform, the Laplace transform provides an alternative way of
analyzing a signal in the time domain, i.e., in the frequency domain. The Laplace transform
has been widely used in engineering and physics. For example, the Laplace transform has
been used to analyze the linear time-invariant system in control theory by transforming the
convolution to a multiplication which is easier to solve due to the algebraic form. Another
important application of the Laplace transform is to solve partial differential equations, which
has been extensively used in mechanical engineering and electrical engineering. In particular,
the Laplace transform can turn a linear differential equation to an algebraic equation which
can be easily solved by the formal rules of algebra, and the original linear differential equation
can then be solved using the inverse Laplace transform.

4.98 Engineering Example: unit-step function - integral form

In mathematics, the unit-step function can y

defined in the following:
y = H(x) 1
1, if x ≥ 0;

H(x) =
0, if x < 0. −1 O 1
x
−1

Alternatively, H(x) can also be defined as an integral of the Dirac delta function δ(s),
which is provided as follows:
ˆ x
H(x) = δ(s) ds
−∞

where δ(s) is the Dirac delta function defined as

if x ≤ 0;

x,
δ(s) =
sin x, if x > 0,
ˆ ∞
and δ(s) ds = 1.
−∞
In communication theory, sum of the unit-step functions can be used for representations
of waveforms. Moreover, the unit-step response, which is the time behaviour of the output
of a general system when its input is a unit-step function, can be used to describe the
characteristics of a system and is thus widely used in electronic engineering and control
theory.
258 Integration

4.99 Engineering Example: continuous-time convolution

In mathematics, the continuous-time convolution is defined as an improper function as

follows:
ˆ ∞
(f ? g)(t) = f (τ )g(t − τ ) dτ.
−∞

The function produced by convolution, i.e., (f ? g)(t), shows how the shape of one function
(e.g., f (t)) is modified by the other one (e.g., g(t)). The continuous-time convolution has
wide applications in probability, statistics, signal processing and many other engineering
areas. For example, the convolution is an important computation tool to characterize the
linear time invariant (LTI) systems.

4.100 Engineering Example: integrator - signal processing

The integrator is a device whose output signal is the time integral of its input signal. It is
an important part in many engineering and scientific applications, such as signal processing
circuits and mechanical engineering devices.
In signal processing area, block diagrams are commonly seen to illustrate the signal flow
paths. One example is shown below, where the quadrate block represents an integrator.
ˆ
x(t) y(t)

Given that the input signal is x(t) and the output signal is y(t), then the integrator
represents the following integral:
ˆ t
y(t) = x(τ ) dτ.
−∞
 259

Summary

Definition: basic concepts

ˆ b n
X
definite integral : f (x) dx = lim f (x∗i )∆x
∆x→0
ˆa i=1

indefinite integral : f (x) dx = F (x) + C, where F 0 (x) = f (x)

Theorem: properties of definite integrals

Let f and g be integrable functions over an interval I and let a < b < c be points in I.

1. If f and g are integrable over [a, b], then f + g is integrable over [a, b], and we have
ˆ b ˆ b ˆ b
[f (x) + g(x)] dx = f (x) dx + g(x) dx.
a a a

2. If f is integrable over [a, b], then kf (x) is integrable over [a, b], and we have
ˆ b ˆ b
kf (x) dx = k f (x) dx.
a a

3. If f is integrable over [a, b] and [b, c], then f is integrable over [a, c], and we have
ˆ b ˆ c ˆ c
f (x) dx + f (x) dx = f (x) dx.
a b a

ˆ a
4. f (x) dx = 0.
a

5. If f and g are both integrable over [a, b], and f (x) ≤ g(x), then
ˆ b ˆ b
f (x) dx ≤ g(x) dx.
a a

6. ˆ  ˆ
 b  b
f (x) dx ≤ |f (x)| dx.
 

 a  a

7. (Mean Value Theorem for Integrals) If f is continuous on [a, b], then there is c ∈
[a, b] such that
ˆ b
f (x) dx = f (c)(b − a).
a
260 Summary

Theorem: linearity of indefinite integral

ˆ ˆ ˆ
1. [f (x) + g(x)] dx = f (x) dx + g(x) dx.
ˆ ˆ
2. kf (x) dx = k f (x) dx, where k is a constant.

Theorem: Fundamental Theorem of Calculus

Let f be a continuous function on [a, b].

1. Then the function ˆ x

A(x) = f (t) dt
a

is continuous on [a, b], differentiable in (a, b), and A0 (x) = f (x).

2. If F is a differentiable function [a, b] such that F 0 (x) = f (x), then
ˆ b
f (x) dx = F (b) − F (a).
a

Theorem: Indefinite Integrals of Some Elementary Functions

ˆ ˆ
1. k dx = kx + C 6. cos x dx = sin x + C
ˆ
xµ+1 ˆ
2. xµ dx = (µ 6= −1)
µ+1 7. sin x dx = − cos x + C
ˆ
dx
3. = ln |x| + C
x ˆ
ˆ 8. ex dx = ex + C
dx
4. = arctan x + C
x2 + 1
ˆ ˆ
dx ax
5. √ = arcsin x + C 9. ax dx = +C
1 − x2 ln a
ˆ ˆ
dx
10. = sec2 x dx = tan x + C
cos2 x
ˆ ˆ
dx
11. = csc2 x dx = − cot x + C
sin2 x
ˆ
12. sec x tan x dx = sec x + C
ˆ
13. csc x cot x dx = − csc x + C
 261

Theorem: substitution method

Theorem: for indefinite integral

Let u = g(x) be a differentiable function and f be a continuous function. Suppose

the range of g is contained in the domain of f . Then
ˆ ˆ
f g(x) · g 0 (x) dx = f (u) du.

Theorem: for definite integral

Let u = g(x), where g 0 (x) is continuous on [a, b], and let f be continuous on an
interval containing the range of g. Then
ˆ b ˆ g(b)
f g(x) · g 0 (x) dx = f (u) du.


a g(a)

Theorem: integration by parts

Suppose u and v are differentiable functions of x. Then

ˆ ˆ
u(x) · v 0 (x) dx = u(x) · v(x) − v(x) · u0 (x) dx,
| {z } | {z }
dv du

or ˆ ˆ
u dv = uv − v du.

Furthermore, for definite integral,

ˆ b ˆ b
b
u(x) · v (x) dx = [u(x) ·
0
v(x)]a − v(x) · u0 (x) dx.
a a

Method: numerical integration

Let f be integrable function over [a, b]. Divide the interval [a, b] into n equal subintervals

b−a (n − 1)(b − a)
x0 = a, x1 = a + , . . . , xn−1 = a + , xn = b.
n n
• The left point rule is the Riemann sum of left end points of f on [a, b]:
ˆ b
b−a
f (x) dx ≈ Ln = [f (x0 ) + f (x1 ) + · · · + f (xn−1 )] · .
a n
262 Summary

• The right point rule is the Riemann sum of right end points of f on [a, b]:
ˆ b
b−a
f (x) dx ≈ Rn = [f (x1 ) + f (x2 ) + · · · + f (xn )] · .
a n

• The trapezoidal rule is the average of Riemann sum of left end points and right end
points:
ˆ b
b−a
f (x) dx ≈ Tn = [f (x0 ) + 2f (x1 ) + · · · + 2f (xn−1 ) + f (xn )] · .
a 2n

The region under the graph of the function f on [xi−1 , xi ] is approximated by a trape-
zoid.

• The midpoint rule is the Riemann sum of middle points of f on [a, b]:
ˆ b
b−a
f (x) dx ≈ Mn = [f (x∗1 ) + f (x∗2 ) + · · · + f (x∗n )] · ,
a n
xi−1 + xi
where x∗i = .
2
• Simpson’s rule is the following approximation
ˆ b
b−a
f (x) dx ≈ Sn = · [f (x0 ) + 4f (x1 ) + 2f (x2 ) + 4f (x3 )
a 3n
+ · · · + 2f (xn−2 ) + 4f (xn−1 ) + f (xn )] ,

where n is an even integer.

Theorem: error bounds of numerical integration

Theorem: error bounds of the midpoint and trapezoidal rules

Suppose f 00 is continuous and |f 00 (x)| ≤ K on [a, b]. Then

ˆ 
 b  K(b − a)3
f (x) dx − Tn  ≤ ,
 
12n2

 a 
ˆ 
 b  K(b − a)3
f (x) dx − Mn  ≤ .
 
24n2

 a 
 263

Theorem: error bounds of Simpson’s rule

Suppose f (4) is continuous and f (4) (x) ≤ K on [a, b]. Then

 

ˆ 
 b  K(b − a)5
f (x) dx − Sn  ≤ .
 
180n4

 a 

Definition: improper integrals

ˆ +∞ ˆ b
• f (x) dx = lim f (x) dx.
a b→+∞ a
ˆ a ˆ a
• f (x) dx = lim f (x) dx.
−∞ b→−∞ b
ˆ ∞ ˆ a ˆ ∞
• f (x) dx = f (x) dx + f (x) dx.
−∞ −∞ a
ˆ b ˆ c
• f (x) dx = lim f (x) dx, if f is unbounded near b.
a c→b a
ˆ b ˆ b
• f (x) dx = lim f (x) dx, if f is unbounded near a.
a c→a c
ˆ b ˆ c ˆ b
• f (x) dx = f (x) dx + f (x) dx, if f is unbounded near a and b (c is any fixed
a a c
real number between a and b).

Theorem: Comparison Test for improper integrals

Let f and g be continuous over [a, +∞) and 0 ≤ f (x) ≤ g(x). Then
ˆ +∞ ˆ +∞
1. If g(x) dx converges, then f (x) dx converges.
a a
ˆ +∞ ˆ +∞
2. If f (x) dx diverges, then g(x) dx diverges.
a a

Similar results hold for other types of improper integrals.

264 Summary
 CHAPTER

5
Applications of Integration

5.1 Areas of Regions Between Curves

5.1 Theorem: area of a region bounded by two curves

Let f and g be two continuous functions y

with f (x) ≥ g(x) over the interval [a, b]. The
y = f (x)
area of the region between the graphs of f
and g over [a, b] is x=b

ˆ b
x=a
x
[f (x) − g(x)] dx.
a y = g(x)

5.2 Example: area bounded by two curves

1
Find the area of the closed region bounded by the graphs of the functions f (x) =
x2 + 1
and g(x) = 21 .

Solution We sketch the graphs of the given functions.

266 Applications of Integration

We need first to find the intersections of the y

two curves so that we know the interval over
1
which the two curves bound a closed region. y=
x2 + 1
In fact, by solving the equation
1 y= 1
= 12 , 2
x2 +1 x
−1 O 1
we have x = ±1.
Over the interval [−1, 1], f (x) ≥ g(x). Thus, the area of the region bounded by two curves
is
ˆ 1
1
 
1  1
1
dx = tan−1 x x=−1 − 12 x x=−1

−
−1 x +1
2 2

= tan−1 1 − tan−1 (−1) − 1 −1

2 − 2
= π
4 − −π
4 −1= π
2 − 1.

5.3 Example: multiple regions

Find the area of the closed region bounded by the graphs of the functions f (x) = x + 2
and g(x) = |x| over the interval [−2, 1].

Solution Even though the interval [−2, 1] is given, the integral

ˆ 1
[f (x) − g(x)] dx
−2

is not the area of the region bounded by the curves over [−2, 1]. This is because that neither
f (x) ≥ g(x) hold over the whole interval [−2, 1], nor g(x) ≥ f (x) over the whole interval.
Hence, we need to determine the interval(s) over which f (x) ≥ g(x) and f (x) ≤ g(x),
respectively.

In fact, by solving the equation f (x) = y

g(x), or x + 2 = |x|, we get a unique solu-
tion x = −1. We see that g(x) ≥ f (x) when
y =x+2
x ≤ −1; f (x) ≥ g(x) when x ≥ −1, as shown
in the figure.
y = |x|
x
−2 −1 O 1 2
5.1 Areas of Regions Between Curves 267

Thus, the area of the region bounded by the two curves is

ˆ −1 ˆ 1
[g(x) − f (x)] dx + [f (x) − g(x)] dx
−2 −1
ˆ−1 ˆ 1
= [|x| − (x + 2)] dx + [x + 2 − |x|] dx
−2 −1
ˆ −1 ˆ 0 ˆ 1
= [−x − (x + 2)] dx + [x + 2 − (−x)] dx + [x + 2 − x] dx
−2 −1 0
−1 0 1
= (−x2 − 2x)−2 + (x2 + 2x)−1 + 2x|0
= (−1 + 2) − (−4 + 4) + 0 − (1 − 2) + 2 − 0 = 4.

Geometrically, we see that the area is the same as

ˆ 1
|f (x) − g(x)| dx.
−2

5.4 Theorem: area between two curves

In general, the area between the curve y = f (x) and y = g(x) over [a, b] is
ˆ b
|f (x) − g(x)| dx.
a

5.5 Example: areas between two curves

Find the area between y = cos x and y = sin x over [0, π2 ].

Solution We need to calculate the integral

ˆ π
2
| sin x − cos x| dx.
0

Firstly we find intervals where sin x ≥ y

cos x. To do so, in the interval [0, π2 ], we solve
the equation y = cos x

sin x = cos x
y = sin x
and get x = π4 . We know that sin x ≤ cos x x
when 0 ≤ x ≤ π4 ; sin x ≥ cos x when π4 ≤ x ≤ O π
4
π
2

2 , as shown in the figure.

π
268 Applications of Integration

Hence, the area between y = cos x and y = sin x over [0, π2 ] is

ˆ π
2
| sin x − cos x| dx
0
ˆ π
4
ˆ π
2
= (cos x − sin x) dx + (sin x − cos x) dx
π
0 4
π π
= (sin x + cos x)|04 + (− cos x − sin x)| π2
4

= (sin π4 + cos π4 ) − (sin 0 + cos 0) + (− cos π2 − sin π2 ) − (− cos π4 − sin π4 )

√ √ √
= 2 − 1 − 1 + 2 = 2 2 − 2.

5.6 Theorem: area bounded by two curves along the y-axis

Let A be the area of the region that lies be- y

tween two curves x = f (y) and x = g(y) and
y=d
between the horizontal lines y = c and y = d,
where f and g are continuous functions and
f (y) ≥ g(y) for all x in [c, d]. Then
ˆ d x = g(y) x = f (y)
A= [f (y) − g(y)] dy.
c x
y=c

5.7 Example: areas between two curves

Find the common area of two disks x2 + y 2 ≤ r2 and (x − r)2 + y 2 ≤ r2 .

Solution Solving two simultaneous equa-

tions y

x2 + y 2 = r 2 , (x − r)2 + y 2 = r2
x=r− x=
p p
√ r2 − y2 r2 − y2
gives y = ± 23 r. It is easy to get the two arcs
x
that enclosed
p the region are x = r − y
p
2 2 O

and x = r − r − y , as shown in the figure.

2 2

Thus, the common region is

ˆ √
3
2 r hp  p i ˆ √
3
2 r  p 
√ r2 − y 2 − r − r2 − y 2 dy = √ r−2 r2 − y 2 dy.
3 3
− 2 r − 2 r

To evaluate the last integral, we make a trigonometric substitution y = r sin t. Then the area
5.1 Areas of Regions Between Curves 269

is
ˆ √
3
2 r  p  ˆ π
3

√ 2 r2 − y 2 − r dy = (2r cos t − r) · r cos t dt

− 23 r −π
3
ˆ π
3
= 2r 2
2 cos2 t − cos t dt


0
ˆ π
3
= 2r2 [(1 + cos 2t) − cos t] dt
0
 π
= 2r2 · t + 21 sin 2t − sin t 03

 √ √  √
= 2r2 · π3 + 12 · 23 − 23 = ( 2π 3 − 2 )r .
3 2

5.8 MATLAB: area bounded by two curves along the x-axis

Find the area between y = cos x and y = sin x over [0, π2 ].

Solution First, we plot the curves of given functions.

>> syms x
>> y1 ( x ) = cos ( x ) ;
>> y2 ( x ) = sin ( x ) ;
>> x_array = linspace (0 , pi /2 ,200) ;
>> y1_array = y1 ( x_array ) ;
>> y2_array = y2 ( x_array ) ;
>> figure
>> plot ( x_array , y1_array , 'b ' , x_array , y2_array , 'r ') % plot ...
two curves on the same figure
>> legend ( 'y = cos x ' , 'y = sin x ')
>> xlabel ( 'x ')
>> ylabel ( 'y ')
>> grid

Note that to plot two curves on the same figure, we can write specifications of the two curves
using one plot command instead of two plot commands. The figure plotted is shown below.
270 Applications of Integration

1
y = cos x
0.9 y = sin x

0.8

0.7

0.6

0.5
y

0.4

0.3

0.2

0.1

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
x

We see that y = cos x and y = sin x intersect at a point around x = 0.8. So, we try to find a
point around x = 0.8 such that cos x = sin x. This is shown in the following MATLAB code.

>> Delta_y = y1 - y2

Delta_y ( x ) =

cos ( x ) - sin ( x )

>> Delta_y_hd = matlabFunction ( Delta_y ) % convert symbolic ...

expression to function handle

Delta_y_hd =

function_handle with value :

@( x ) cos ( x ) - sin ( x )

>> x0 = fzero ( Delta_y_hd ,0.8)

x0 =

0.7854

We see from the result above that y = cos x and y = sin x intersect at x0 = 0.7854. Hence,
we can determine the two intervals over which the two curves bound a closed region, namely,
[0, x0] and x0, π2 . According to these two intervals, we calculate the area bounded by the


two curves as follows:

5.1 Areas of Regions Between Curves 271

>> area1 = int ( Delta_y ,x ,0 , x0 ) ;

>> area2 = int ( - Delta_y ,x , x0 , pi /2) ;
>> area = area1 + area2

area =

2*2^(1/2) - 2
√
The result shows that the area bounded by y = cos x and y = sin x over [0, π2 ] is 2 2 − 2.

5.9 MATLAB: area bounded by two curves along the y-axis

Find the common area of two disks x2 + y 2 ≤ 32 and (x − 3)2 + y 2 ≤ 32 .

Solution First, we notice that x2 + y 2 = 32 and (x − 3)2 + y 2 = 32 are two circles.
Specifically, x2 +y 2 = 32 is a circle with the center at (0, 0) and radius 3, and (x−3)2 +y 2 = 32
is a circle with the center at (3, 0) and radius 3. We thus can plot a figure with these two
circles using the following MATLAB code:

>> cla % clear the axes

>> c1 = viscircles ([0 0] ,3 , ' Color ' , 'b ') ; % center : [0 0] , ...
radius : 3
>> c2 = viscircles ([3 0] ,3 , ' Color ' , 'r ') ; % center : [3 0] , ...
radius : 3
>> grid
>> legend ([ c1 , c2 ] ,{ 'x ^2 + y ^2 = 3^2 ' , '( x - 3) ^2 + y ^2 = 3^2 ' }) ;
>> xlabel ( 'x ')
>> ylabel ( 'y ')

Note that the command viscircles can plot a circle with given center and radius. Moreover,
to specify the colour of the circle, besides the parameter specifying the colour, e.g., ‘b’ and
‘r’, we also need to type in ‘Color’ in the command viscircles. Furthermore, since the
command viscircles does not clear the target axes before plotting circles, we need to first
remove any graphs that have been previously plotted in the axes using the cla command.
The figure plotted is shown below.
272 Applications of Integration

y
-1 x 2 + y2 = 32
(x - 3)2 + y 2 = 3 2

-2

-3
-3 -2 -1 0 1 2 3 4 5 6
x

From the figure above,

p we see that the intersections of the two circles are the ones of
the two functions x = 32 − y 2 and x = 3 − 32 − y 2 . In particular, these two functions
p

intersect at points around y = −2.5 and y = 2.5, respectively. p

Therefore, we try
pto find the
points around y = −2.5 and y = 2.5, respectively, such that 32 − y 2 = 3 − 32 − y 2 by
using the following MATLAB code.

>> syms y
>> x1 = sqrt (3^2 - y ^2) ;
>> x2 = 3 - sqrt (3^2 - y ^2) ;
>> Delta_x = x1 - x2

Delta_x =

2*(9 - y ^2) ^(1/2) - 3

>> Delta_x_hd = matlabFunction ( Delta_x ) % convert symbolic ...

expression to function handle

Delta_x_hd =

function_handle with value :

@( y ) sqrt ( - y .^2+9.0) .*2.0 -3.0

>> y1 = fzero ( Delta_x_hd , -2.5)

y1 =

-2.5981

>> y2 = fzero ( Delta_x_hd ,2.5)

5.1 Areas of Regions Between Curves 273

y2 =

2.5981

The results above show that x = 32 − y 2 and x = 3 − 32 − y 2 intersect at y1 = −2.5981

p p

and y2 = 2.5981. Hence, we can determine the interval over which these two functions bound
a closed region, i.e., [y1, y2]. Then, we calculate the area bounded by the two curves as
follows:

>> area = int ( Delta_x , y1 , y2 )

area =

6* pi - (9*3^(1/2) ) /2

We see from√the result above that the common area of two disks x2 +y 2 ≤ 32 and (x−3)2 +y 2 ≤
32 is 6π − 9 2 3 .

Area Under Parametric Curve

5.10 Theorem: area under a parametric curve

If a curve is traced out once by the parametric equations x = f (t) and y = g(t), α ≤ t ≤ β,
then the area under the curve is ˆ β
g(t) f 0 (t) dt.
α

Justification:
ˆ b
The area under a curve y = F (x) over [a, b] is A = F (x) dx. By applying the substitu-
a
tion rule, we get
ˆ β
A= F (f (t)) f 0 (t) dt,
α

where F (a) = α and F (b) = β. Since the curve y = F (x), a ≤ x ≤ b, is traced out by the
given parametric equations, we have

y = F (x) = F (f (t)),

and y = g(t), so that F (f (t)) = g(t). Hence, we obtain the area formula in parametric
equations.
274 Applications of Integration

5.11 Example: area under one arch of the cycloid

Find the area under one arch of the cycloid

x = r(t − sin t), y = r(1 − cos t), 0 ≤ t ≤ 2π.

2r

x
O 2πr

Solution It is known that, for a given curve above the x-axis, the area under the curve
ˆ b
over an interval [a, b] is A = y(x) dx. Thus, if the curve is given by parametric equations
a
x = x(t), y = y(t), then these equations can be served as a substitution.
Hence, the area under one arch of the cycloid is
ˆ 2πr
A= y dx
0
ˆ 2π
x = r(t − sin t)

0
= r(1 − cos t) · [r(t − sin t)] dt
0 y = r(1 − cos t)
ˆ 2π
= r2 (1 − 2 cos t + cos2 t) dt
0
ˆ 2π
= r2 1 − 2 cos t + 12 (1 + cos 2t) dt double-angle formula
 
0
2π
= r2 − 2 sin t + sin 2t = 3πr2 .
3 1
2t 4 t=0

5.12 MATLAB: area under one arch of the cycloid

Find the area under one arch of the cycloid

x = 3(t − sin t), y = 3(1 − cos t), 0 ≤ t ≤ 2π.

Solution First, we plot the graph of the cycloid when 0 ≤ t ≤ 2π.

>> t_array = linspace (0 ,2* pi ,200) ;

>> x_array = 3*( t_array - sin ( t_array ) ) ;
>> y_array = 3*(1 - cos ( t_array ) ) ;
>> figure
>> plot ( x_array , y_array , 'r ')
>> grid
>> xlabel ( 'x ')
5.2 Volumes 275

>> ylabel ( 'y ')

The figure plotted is shown below.

3
y
2

0
0 2 4 6 8 10 12 14 16 18 20
x

Then, the area under one arch of the cycloid can be calculated as follows:

>> syms t
>> x ( t ) = 3*( t - sin ( t ) ) ;
>> y ( t ) = 3*(1 - cos ( t ) ) ;
>> dx ( t ) = diff ( x ( t ) ,t ) ;
>> area = int ( y ( t ) * dx ( t ) ,0 ,2* pi )

area =

27* pi

The result above shows that the area is 27π.

5.2 Volumes
5.13 Definition: volume

Let S be a solid that lies between x = a

and x = b. If the cross-sectional area of S in
the plane Px , through x and perpendicular to
the x-axis, is A(x), where A is a continuous
function, then the volume of S is x=a A(x) x=b
Xn ˆ b x
V = lim A(x∗i )∆x = A(x) dx.
n→∞ a
i=1

Similarly, if S is a solid that lies between y = c and y = d and if the cross-sectional area
276 Applications of Integration

of S through y perpendicular to the y-axis is A(y), where A is a continuous function, then

the volume of S is ˆ d
Xn
V = lim A(yi∗ )∆y = A(y) dy.
n→∞ c
i=1

We normally say the above method as finding volume by using slices.

5.14 Example: volume of a pyramid - using slices

Find the volume of a pyramid whose base is a square with side L and whose height is h.

Solution We place the origin O of the coordinates at the vertex of the pyramid and the
x-axis along its central axis. Any plane Px that passes through x and is perpendicular to
the x-axis intersects the pyramid in a square with side of length s, say. We can express s in
terms of x by observing from the similar triangles:
x h
=
s L
xL L2
and so s = . Thus, the cross-sectional area is A(x) = s2 = 2 x2 .
h h

y
z

s L
x
x
O
x x
h
h

The pyramid lies between x = 0 and x = h, so its volume is

ˆ h ˆ h
L2 2 L2 1 3 h
V = A(x) dx = x dx = ( x = 31 L2 h.
0 0 h2 h2 3 0

5.15 Example: volume of a solid - using slices

Consider a solid whose base is a circular disk with radius r. Find the volume of the solid
if parallel cross-sections perpendicular to the base are isosceles triangles with height h and
unequal side in the base.

Solution We sketch the diagram of the situation as in the figure.

5.2 Volumes 277

√
y= r 2 − x2

y
x
O

√ x
y = − r 2 − x2

Solving the equation x2 + y 2 = r2 for x, we have the expressions of the top-half and the
bottom-half of the circle, p
x = ± r2 − y2 .
It is clear that the circle intersects the x-axis at x = ±r.
By symmetry, without loss of generality, we can assume that cross-sections are perpendic-
ular to the x-axis. For −r ≤ x ≤ r, the area of the cross-section is
√ √
A(x) = 21 · 2 r2 − x2 · h = h r2 − x2 .

Hence, the volume of the solid is

ˆ r ˆ r p ˆ r p
V = A(x) dx = h r2 − x2 dx = h r2 − x2 dx.
−r −r −r

We observe that the last integral is the area of semi-circular disk of radius r, so

V = h · 21 πr2 = 12 πr2 h.

5.16 Example: volume of a ball - using slices

It is known that the area of circular disk of radius r is πr2 . Find the volume of the ball of
radius r.
√
Solution A ball of radius r can be generated by rotating the upper semicircle y = r2 − x2
about the x-axis. For −r ≤ x ≤ r, the cross-section is a circular disk, with area π(r2 − x2 ),
as shown in the figure.
278 Applications of Integration

y
√
y= r 2 − x2

x
−r O r

Thus, the volume of the ball is

ˆ r
V = π(r2 − x2 ) dx
−r
ˆ r
r
=2 π(r2 − x2 ) dx = 2π (r2 x − 31 x3 )0 = 34 πr3 .
0

5.17 Example: volume of a solid torus - using slices

Consider the region bounded by the curves

(x − R)2 + y 2 = r2 , 0 < r < R.

The solid obtained by rotating the region about the y-axis is the solid torus, as shown in the
figure.

R + r2 − y2
p
p
R − r − y2
2

y
R
R+r
x
O R−r

(x − R)2 + y 2 = r2

We shall use the method of slices to compute the volume of the solid.
The region and a typical slice are shown in the figure.
Wepsee that the slice is a washer, with outer radius R + r2 − y 2 , and inner radius
p

R − r2 − y 2 . So the cross-sectional area is

 p 2  p 2
π R + r2 − y2 − π R − r2 − y2 .
5.2 Volumes 279

Thus, the volume of the solid torus is

ˆ r   p 2  p 2 
V = π R+ r −y 2 2 2
−π R− r −y 2 dx
−r
ˆ r p
=π 2R · 2 r2 − y 2 dy factoring
−r
ˆ r p
= 4πR r2 − y 2 dy.
−r

Since the last integral represents the area of the semicircular region of radius r, we have

V = 4πR · 12 πr2 = 2π 2 Rr2 .

5.18 MATLAB: volume of a ball - using slices

Find the volume of a ball with radius 3.

Solution First, to well visualize the ball and the cross-section, we plot the ball and a
cross-section for a fixed x value in MATLAB by the following code:

% plot the sphere with radius r

N = 40;
r = 3; % sphere radius

[ XX , YY , ZZ ] = sphere ( N ) ; % unit sphere generated by three ...

( N +1) -by -( N +1) matrices XX , YY and ZZ
X = r * XX ; % expand the sphere radius
Y = r * YY ;
Z = r * ZZ ;

figure
surf (X ,Y ,Z , ' FaceAlpha ' ,0.2 , ' FaceColor ' , 'b ') % ' FaceAlpha ': ...
face transparency - ranges from 0 ( fully transparent ) to 1 ...
( fully opaque )
xlabel ( 'x ')
ylabel ( 'y ')
zlabel ( 'z ')

% plot the cross - section

x0 = 2; % cross - section is at x = 2
r0 = sqrt ( r ^2 - x0 ^2) ; % radius of the cross - sectional circle

y1 = linspace (0 , r0 ,200) ;
z1 = zeros (1 , length ( y1 ) ) ;

for ind = 1: length ( y1 ) % first quadrant

280 Applications of Integration

y = y1 ( ind ) ;
z1 ( ind ) = sqrt ( r0 ^2 - y ^2) ;
end

y2 = - y1 (1: length ( y1 ) ) ; % second quadrant

z2 = z1 (1: length ( z1 ) ) ;

y3 = - y1 (1:( length ( y1 ) ) ) ; % third quadrant

z3 = - z1 (1:( length ( z1 ) ) ) ;

y4 = y1 (1:( length ( y1 ) ) ) ; % fourth quadrant

z4 = - z1 (1:( length ( z1 ) ) ) ;

y_array = [ fliplr ( y1 ) ,y2 , fliplr ( y3 ) , y4 ]; % plot the circle ...

anticlockwise
z_array = [ fliplr ( z1 ) ,z2 , fliplr ( z3 ) , z4 ];
x_array = ones (1 , length ( y_array ) ) * x0 ;
x4 = ones (1 , length ( y4 ) ) * x0 ;

hold on
plot3 ( x_array , y_array , z_array , 'r ' , ' LineWidth ' ,4)

Note that the command sphere can be used to generate a ball with unit radius, the
command surf can be used to plot the surface of the ball, and the command plot3 can
be used to plot curves in three-dimensional space. Moreover, the face transparency and the
face colour of the ball can be specified when using surf, and the width of the curve can be
specified when using plot3.
The figure plotted is shown below.
5.2 Volumes 281

Since the cross-section is a circular disk with area π(32 − x2 ), as shown in the figure above,
we thus can calculate the volume of the ball by solving the following integral:
ˆ 3
V = π(32 − x2 )dx.
−3

It can be implemented in MATLAB as follows:

>> syms x
>> f ( x ) = pi *(3^2 - x ^2) ;
>> V = int ( f ( x ) , -3 ,3)

V =

36* pi

The volume of the ball is thus 36π.

Method of Cylindrical Shells

5.19 Theorem: volume by cylindrical shells

The volume of the solid, obtained by ro- y

tating about the y-axis the region under the
curve y = f (x) from a to b (0 ≤ a < b), is y = f (x)
ˆ b
x
V = 2πxf (x) dx. a b
a

We normally say the above method as finding volume by using shells.

Justification:

We divide the interval [a, b] into n subintervals [xi−1 , xi ] of equal width ∆x and let x∗i be a
point in the ith subinterval. If the rectangle with base [xi−1 , xi ] and height f (x∗i ) is rotated
about the y-axis, then the result is a cylindrical shell with average radius x∗i , height f (x∗i ),
and thickness ∆x, so its volume is

Vi = (circumference) × (height) × (thickness) = (2πx∗i ) · f (x∗i ) · ∆x.

282 Applications of Integration

f (x∗i )
a b
x
∆x
x∗i

a b
x
x∗i

f (x∗i )
∆x
2πx∗i

Hence an approximation to the volume V of S is given by the sum of the volumes of these
shells:
Xn Xn
V ≈ Vi = 2πx∗i f (x∗i )∆x.
i=1 i=1

This gives the volume:

n
X ˆ b
V = lim 2πx∗i f (x∗i )∆x = 2πxf (x) dx,
n→∞ a
i=1

where the terms in the formula can be interpreted as

x: radius of a typical cylinder shell

2πx : circumference
f (x) : height
dx : thickness of the shell
ˆ b
(2πx) [f (x)] (dx)
a | {z } | {z } |{z}
circumference height thickness

5.20 Example: volume of a ball - using shells

Find the volume of the ball of radius r.

5.2 Volumes 283

x2 + y 2 = r 2
r

x h
x
O

−r

The region and a typical shell √

are shown in the figure. We see that the shell has radius x,
circumference 2πx, and height 2 r2 − x2 . So the volume is
ˆ r p
V = 2πx · 2 r2 − x2 dx
0
ˆ rp
= (−2π) r2 − x2 · (r2 − x2 )0 dx
0
r
3
= (−2π) · 32 (r2 − x2 ) 2  = 43 πr3 .
0

5.21 Example: volume of a solid torus - using shells

Consider the region bounded by the curves

(x − R)2 + y 2 = r2 , 0 < r < R.

The solid obtained by rotating the region about the y-axis is the solid torus, as shown in the
figure.

(x − R)2 + y 2 = r2

R h
x
O x

We shall use the method of cylindrical shells to compute the volume of the solid.
The region and a typical shell are shown in the figure.
We see that the shell has radius x, circumference 2πx, and height 2 r2 − (x − R)2 . So
p
284 Applications of Integration

the volume is
ˆ R+r
V = 2πx · 2 r2 − (x − R)2 dx
p
R−r
ˆr p
= 4π(u + R) r2 − u2 du u = x − R, du = dx
−r
ˆ r  p p 
= 4π u r2 − u2 + R r2 − u2 du
−r
ˆ r p
= 4π R r2 − u2 du odd function
−r
ˆ r p
= 4πR r2 − u2 du.
−r

Since the last integral represents the area of the semicircular region of radius r, we have

V = 4πR · 12 πr2 = 2π 2 Rr2 .

5.22 Example: rotating about a horizontal line

Find the volume of the solid obtained by rotating the region bounded by the curves
√
y = x and y = x2 ,

about the horizontal line y = −k (k > 0).

√
Solution By solving the simultaneous equations y = x and y = x2 , we see that there are
two intersections (0, 0) and (1, 1). So the region bounded by the given curves is 0 ≤ x ≤ 1 as
well as 0 ≤ y ≤ 1.
We compute the volume of the solid by slices and shells.

1. The region and a typical slice are shown in the following figures.

y y
y = x2 y = x2
√ √
y= x y= x
1 1

√
x+k

x x
O x 1 x2 + k O 1

y = −k y = −k

√
We see that the slice is a washer, with inner radius x2 + k and outer radius x + k. So
5.2 Volumes 285

the volume is
ˆ 1  √
V = π( x + k)2 − π(x2 + k)2 dx

0
ˆ 1 √
=π x + 2k x − x4 − 2kx2 dx


0
= π
30 (9 + 20k).

2. The region and a typical shell are shown in the following figures.

y y
√
y y = x2 y = x2
√ √
y 2
y= x y= x
1 1

y
y+k
x x
O 1 O 1

y = −k y = −k

√
We see that the shell has radius y + k, circumference 2π(y + k), and height y − y2 .
So the volume is
ˆ 1
√
V = 2π(y + k) · ( y − y 2 ) dy
0
ˆ 1  
3 1
= 2π y 2 − y 3 + ky 2 − ky 2 dx
0
= π
30 (9 + 20k).

Solid Generated by Parametric Curve

5.23 Example: solid generated by a parametric curve

Rotate the area under one arch of the cycloid

x = r(t − sin t), y = r(1 − cos t), 0 ≤ t ≤ 2π,

about the x-axis and the y-axis, respectively. Find the volumes of the two solids.
286 Applications of Integration

2r

x
O 2πr

Solution
1. When rotating the region about the x-axis, we compute the volume of the solid by
slices.
A typical slice is shown in the following figure.

2r

x
O 2πr

ˆ 2π
The volume of the solid is V = π[y(x)]2 dx. We can use the parametric equations
0
for the cycloid as a substitution for the definite integral. Thus, the volume of the solid
is
ˆ 2πr
V = π[y(x)]2 dx
0
ˆ 2π
x = r(t − sin t)

2 0
= π [r(1 − cos t)] · [r(t − sin t)] dt
0 y = r(1 − cos t)
ˆ 2π
= πr3 (1 − cos t)3 dt
0
ˆ 2π
= πr 3
1 − 3 cos t + 3 cos2 t − cos3 t dt expanding


0
ˆ 2π
= πr 3
1 − 3 cos t + 23 (1 + cos 2t)

0

−(1 − sin2 x) cos t dt



2π
= πr3 − 3 sin t + sin 2t − sin t + sin3 t
5 3 1
2t 4 3 t=0
= 5π r .
2 3

2. When rotating the region about the y-axis, we compute the volume of the solid by
shells.
5.2 Volumes 287

A typical shell is shown in the following figure.

2r

x
O 2πr

ˆ 2π
The volume of the solid is V = 2πx · y(x) dx. We use the parametric equations for
0
the cycloid as a substitution for the definite integral. Thus, the volume of the solid is
ˆ 2πr
V = 2πx · y(x) dx
0
ˆ 2π
0
= 2πr(t − sin t) · [r(1 − cos t)] · [r(t − sin t)] dt
0
ˆ 2π
= 2πr3 (t − sin t)(1 − cos t)2 dt.
0

The function sin t · (1 − cos t)2 is a 2π periodic odd function, so

ˆ 2π ˆ π
sin t · (1 − cos t)2 dt = sin t · (1 − cos t)2 dt = 0.
0 −π

Hence,
ˆ 2π
V = 2πr3 t(1 − cos t)2 dt
0
ˆ 2π
= 2πr3 t(1 − 2 cos t + cos2 t) dt
0
ˆ 2π
= 2πr 3
t 1 − 2 cos t + 21 (1 + cos 2t) dt
 
0
2π
= 2πr · − 2t sin t − 2 cos t + 14 t sin 2t + cos 2t
3
3 2 1
4t 8 t=0
= 6π 3 r3 .
288 Applications of Integration

5.3 Arc Lengths

5.24 Definition: arc length

Suppose a curve is defined by y = f (x), where f is continuous and a ≤ x ≤ b. We approxi-

mate C by a collection of line segments, by dividing the interval [a, b] into n subintervals with
end points x0 , x1 , . . . , xn and equal width ∆x. Denote the endpoints of the line segments
to be P0 , P1 , . . . , Pn , where Pi lines on the curve C with coordinates (xi , f (xi )). Then, the
length L of the curve C is defined as
n
X
L = lim |Pi−1 Pi |.
n→∞
i=1

y
Pn

P1
Pi−1 Pi
P0

x0 x1 xi−1 xi xn
x

5.25 Theorem: formulas for arc length of a smooth curve

If f 0 is continuous on [a, b], then the length of the curve y = f (x), a ≤ x ≤ b, is

ˆ b ˆ bq
s  2
dy 2
L= 1+ dx = 1 + [f 0 (x)] dx.
a dx a

Similarly, If g 0 is continuous on [c, d], then the length of the curve x = g(y), c ≤
y ≤ d, is
ˆ d ˆ dq
s  2
dx 2
L= 1+ dy = 1 + [g 0 (y)] dy.
c dy c

Definition: arc length function

If f 0 is continuous on [a, b], the distance along the curve from the initial point
(a, f (a)) to the point (x, f (x)) defines the arc length function:
ˆ xq
2
s(x) = 1 + [f 0 (t)] dt.
a
5.3 Arc Lengths 289

Justification:

Let ∆yi = yi − yi−1 = f (xi ) − f (xi−1 ). Then

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

p p

By the Mean Value Theorem,

∆yi = f (xi ) − f (xi−1 ) = f 0 (x∗i )(xi − xi−1 ) = f 0 (x∗i )∆x,

where x∗i is between xi−1 and xi . Thus,

q
2
|Pi−1 Pi | = (∆x)2 + (∆yi )2 = (∆x)2 + [f 0 (x∗i )∆x]
p
q
2
= 1 + [f 0 (x∗i )] ∆x.

Hence,
n
X n q
X 2
L = lim |Pi−1 Pi | = lim 1 + [f 0 (x∗i )] ∆x
n→∞ n→∞
i=1 i=1
ˆ b q
2
= 1 + [f 0 (x)] dx.
a

5.26 Example: arc length

3
Find the arc length of the curve y = x 2 y
from (0, 0) to (1, 1).
1
3
Solution The equation y = x 2 gives y 0 = 3

3 12 y = x2
2 x . Thus, the arc length of the curve from
(0, 0) to (1, 1) is x
O 1

ˆ 1 p ˆ 1 q
L= 1 + y 02 dx = 1 + 94 x dx
0 0
ˆ 13
4 1 9
= t · 2 4
dt t=1+ x
1
9 4
h i 13
4
√
2 32
= 4
9 3t = 13 13−8
27 .
t=1
290 Applications of Integration

Remark:

2 dx 1
The same curve can be expressed as x = y 3 , which gives = 32 y − 3 . Thus, the
dy
arc length from (0, 0) to (1, 1) is
ˆ 1 ˆ 1q
s  2
dx 2
L= 1+ dy = 1 + 94 y − 3 dy.
0 dy 0

Evaluation of the last integral (an improper integral!) is technically more complicated.
So, when there are more than one choices in finding a quantity (like area, volume,
arc length, etc), one may need to choose a wise setup.

5.27 Example: arc length of astroid

2
Find the circumference of the astroid x 3 +
2 2
y = a3 .
3

Solution The astroid is symmetric about y

the x-axis and the y-axis. So, its circumfer-
ence is four time the arc length in the first a
quadrant. In the first quadrant, y ≥ 0. In 2 2 2
x3 + y3 = a3
this region, solving the equation of the as-
troid gives x
O a

 2 2
 32
y = a3 − x3 ,

where 0 ≤ x ≤ a, so that

dy  2 2
 21 1 1
 2 2
 21
= 23 · a 3 − x 3 · (−1) 32 x− 3 = −x− 3 a 3 − x 3 .
dx
Thus, the circumference of the astroid is
ˆ ar h  i
2 2 2
L=4 1 + x− 3 a 3 − x 3 dx
ˆ
0
a
1 1
=4 a 3 x− 3 dx an improper integral
0
ˆ a
1 1
= 4a 3 lim+ x− 3 dx
→0
1
h  2 ia
= 4a 3 lim+ 32 x 3 = 6a.
→0 x=
5.3 Arc Lengths 291

5.28 MATLAB: arc length

3
Find the arc length of the curve y = x 2 from (0, 0) to (1, 1).
3
Solution First, we plot the curve of y = x 2 within the interval [0, 1].

>> syms x
>> f ( x ) = x ^(3/2) ;
>> x_array = linspace (0 ,1 ,200) ;
>> figure
>> plot ( x_array , f ( x_array ) , 'b ')
>> xlabel ( 'x ')
>> ylabel ( 'y ')
>> grid

The figure plotted is shown below.

To calculate the arc length of the curve in MATLAB, we provide two methods.
1. We calculate the arc length of the curve by using the arc length function.

>> syms x
>> f ( x ) = x ^(3/2) ;
>> df ( x ) = diff ( f ( x ) ,x )

df ( x ) =

(3* x ^(1/2) ) /2

>> arc_len = int ( sqrt (1 + df ( x ) ^2) ,0 ,1)

arc_len =

(13*13^(1/2) ) /27 - 8/27

√
The result above shows that the arc length is 13 13
27 − 27 .
8
292 Applications of Integration

2. We calculate an approximation of the arc length based on the arc length definition.

N = 1000; % number of line segments

x_array = linspace (0 ,1 , N ) ;
y_array = x_array .^(3/2) ;
arc_len = 0;

for ind = 2: length ( x_array )

x2 = x_array ( ind ) ;
x1 = x_array ( ind - 1) ;
y2 = y_array ( ind ) ;
y1 = y_array ( ind - 1) ;
seg_len = sqrt (( y2 - y1 ) ^2 + ( x2 - x1 ) ^2) ;
arc_len = arc_len + seg_len ;

end

Note that the bigger the number of line segments N is, the more accurate the approxi-
mation is.
The result of the arc length can be seen when we type in arc_len in the command
window, as shown below.

>> format long

>> arc_len

arc_len =

1.43 97096 857556 22

It is shown that when N = 1000, √

the approximation is 1.439709685755622, which is
very close to the exact result 13 27
13−8
≈ 1.439709873371550.

Arc Length of Parametric Curve

5.29 Theorem: arc length for parametric curve

Suppose a parametric curve is given by the parametric equations

x = f (t), y = g(t), α ≤ t ≤ β,

where f 0 and g 0 are continuous and the parametric curve is traversed exactly once as t
5.3 Arc Lengths 293

increases from α to β. Then the arc length is

ˆ β  2  2 ˆ βq
s
dx dy 2 2
L= + dt = [f 0 (t)] + [g 0 (t)] dt.
α dt dt α

Justification: Part 1

We divide the parameter interval [α, β] y

into n subintervals of equal width ∆t. If
Pi−1 P
t0 , t1 , t2 , . . . , tn are the endpoints of these i

subintervals, then xi = f (ti ) and yi = C

g(ti ) are the coordinates of points Pi (xi , yi )
that lie on C and the polygon with vertices
P2
P0 , P1 , . . . , Pn approximates C. x
P1
P0
α t
ti−1 ti β

The length L of C is defined as the limit of the lengths of these approximating polygons
as n → ∞:
Xn
L = lim |Pi−1 Pi |.
n→∞
i=1

Justification: Part 2

By the Mean Value Theorem, there are t∗i and t∗∗

i in (ti−1 , ti ) such that

∆xi = f (ti ) − f (ti−1 ) = f 0 (t∗i )∆t,

∆yi = g(ti ) − g(ti−1 ) = g 0 (t∗∗
i )∆t.

Thus,
q
|Pi−1 Pi | =(∆xi )2 + (∆yi )2 = [f 0 (t∗i )]∆t)2 + [g 0 (t∗∗
i )]∆t)
p
2
q
= [f 0 (t∗i )]2 + [g 0 (t∗∗
i )] ∆t,
2

and so
n q
X
L = lim [f 0 (t∗i )]2 + [g 0 (t∗∗
i )] ∆t.
2
n→∞
i=1

The last sum resembles a Riemann sum for the function [f 0 (t)]2 + [g 0 (t)]2 but it is not
p

exactly a Riemann sum because t∗i 6= t∗∗

i in general. Nevertheless, if f and g are continuous,
0 0

it can be shown that the limit is the same as if ti and ti were equal, namely,
∗ ∗∗

ˆ β q
2 2
L= [f 0 (t)] + [g 0 (t)] dt.
α
294 Applications of Integration

5.30 Example: arc length of a parametric curve

Find the circumference of the astroid, given

by the parametric equations y

x = a cos3 t, y = a sin3 t,
a x = a cos3 t,
y = a sin3 t
where 0 ≤ t ≤ 2π. (0 ≤ t ≤ 2π)

x
Solution The astroid is symmetric about O a

the x-axis and the y-axis. So, its circumfer-

ence is four time the arc length in the first
quadrant. Thus,

ˆ π
2
q
L=4 [3a cos2 t · (− sin t)]2 + [3a sin2 t · (cos t)]2 dt
0
ˆ π
2
= 12a sin t cos t dt
0
 π2
= 12a sin2 t = 6a.
1
2 t=0

5.31 MATLAB: arc length of a parametric curve

Find the circumference of the astroid, given by the parametric equations

x = cos3 t, y = sin3 t,

where 0 ≤ t ≤ 2π.
Solution First, we plot the curve of x = cos3 t, y = sin3 t where 0 ≤ t ≤ 2π.

>> syms t
>> x ( t ) = cos ( t ) ^3;
>> y ( t ) = sin ( t ) ^3;
>> t_array = linspace (0 ,2* pi ,200) ;
>> figure
>> plot ( x ( t_array ) ,y ( t_array ) , 'r ')
>> xlabel ( 'x ')
>> ylabel ( 'y ')
>> grid

The figure plotted is shown below.

5.3 Arc Lengths 295

0.8

0.6

0.4

0.2

y
-0.2

-0.4

-0.6

-0.8

-1
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
x

We see from the figure above that the plotted astroid is symmetric about the x-axis and the
y-axis. So, its circumference is four times the arc length in the first quadrant. We thus first
calculate the arc length of the curve in the first quadrant, and then multiply it by 4.
To calculate the arc length of the astroid in MATLAB, we provide two methods.
1. We calculate the arc length of the astroid by using the arc length function.

>> syms t
>> x ( t ) = ( cos ( t ) ) ^3;
>> y ( t ) = ( sin ( t ) ) ^3;
>> dx ( t ) = diff ( x ( t ) ,t )

dx ( t ) =

-3* cos ( t ) ^2* sin ( t )

>> dy ( t ) = diff ( y ( t ) ,t )

dy ( t ) =

3* cos ( t ) * sin ( t ) ^2

>> arc_len = 4* int ( sqrt (( dx ( t ) ) ^2 + ( dy ( t ) ) ^2) ,0 , pi /2)

arc_len =

The result above shows that the arc length of the astroid is 6.
2. We calculate an approximation of the arc length based on the arc length definition.

N = 1000; % number of line segments

296 Applications of Integration

t_array = linspace (0 , pi /2 , N ) ;
x_array = cos ( t_array ) .^3;
y_array = sin ( t_array ) .^3;
arc_len1 = 0; % arc length in the 1 st quadrant

for ind = 2: length ( t_array )

x2 = x_array ( ind ) ;
x1 = x_array ( ind - 1) ;
y2 = y_array ( ind ) ;
y1 = y_array ( ind - 1) ;
seg_len = sqrt (( x2 - x1 ) ^2 + ( y2 - y1 ) ^2) ;
arc_len1 = arc_len1 + seg_len ;

end

arc_len = 4* arc_len1 ;

Note that the bigger the number of line segments N is, the more accurate the approxi-
mation is.
The result of the arc length can be seen when we type in arc_len in the command
window, as shown below.

>> format long

>> arc_len

arc_len =

5.99 99993 819181 76

It is shown that when N = 1000, the approximation is 5.999999381918176, which is

very close to the exact result 6.

5.4 Surface Area of Revolution

5.32 Definition: surface area of revolution

Consider the surface is obtained by rotating the curve defined by y = f (x), where f is
positive and a ≤ x ≤ b. In order to define its surface area, we divide the interval [a, b] into
n subintervals with endpoints x0 , x1 , . . . , xn and equal width ∆x. The part of the surface
between xi−1 and xi is approximated by taking the line segment Pi−1 Pi and rotating it
about the x-axis, where Pi (xi , f (xi )) lies on the curve. The result is a band with slant height
5.4 Surface Area of Revolution 297

|Pi−1 Pi |. Therefore, in the case where f is positive we define the surface area obtained by
rotating the curve y = f (x), a ≤ x ≤ b, about the x-axis as

y
n
X
S = lim Si , Pi−1 Pn
n→∞ Pi
i=1
P0
where Si is the area of the band with slant
height |Pi−1 Pi |. x

5.33 Theorem: formula for surface area of revolution

If f 0 is continuous on [a, b], then the surface area of the surface obtained by rotating
the curve y = f (x), a ≤ x ≤ b, about the x-axis, is
ˆ b ˆ b
s  2
dy
q
2
S= 2π y(x) 1 + dx = 2π f (x) 1 + [f 0 (x)] dx.
a dx a

Similarly, if g 0 is continuous on [c, d], then the surface area of the surface obtained
by rotating the curve x = g(y), c ≤ y ≤ d, about the y-axis, is
ˆ d ˆ d
s  2
dx
q
2
S= 2π x(y) 1 + dy = 2π g(y) 1 + [g 0 (y)] dy.
c dy c

Justification: Part 1
Let us first derive a formula for the area of a circular cone with base radius r and slant
height l.

l θ
2πr

In fact, we can flat it to form a sector of a circle with radius l and central angle θ = 2πr/l.
Since it is a part of a disk of radius l, we have the area of the sector to be

A = 21 l2 θ = 21 l2 · (2πr/l) = πrl.
298 Applications of Integration

Justification: Part 2
Thus, the area of the band, or frustum of a cone, with slant height l and upper and lower
radii r1 and r2 is
A = πr2 (l1 + l) − πr1 l1 = π [(r2 − r1 )l1 + r2 l] .
From similar triangles we have
l1 l1 + l
=
r1 r2
which gives
r2 l1 = r1 l1 + r1 l or (r2 − r1 )l1 = r1 l.
Hence,
A = π [r1 l + r2 l] = 2πrl = (average circumference) × (slant height),
where r = 2 (r1
1
+ r2 ) is the average radius of the band.

l1

r1

r2

Justification: Part 3

Suppose f 0 is continuous on [a, b]. For a band with slant height l = |Pi−1 Pi | and average
radius r = 12 (f (xi−1 ) + f (xi )), its surface area is
Si = 2π · 12 (f (xi−1 ) + f (xi )) · |Pi−1 Pi |.
By the Pythagorean Theorem and the Mean Value Theorem,
q
2
|Pi−1 Pi | = (xi − xi−1 )2 + [f (xi ) − f (xi−1 )]
q q
2 2
= (∆x)2 + [f 0 (x∗i )∆x] = 1 + [f 0 (x∗i )] ∆x.
Thus, by approximating 21 (f (xi−1 ) + f (xi )) using f (x∗i ), we get
q
2
Si ≈ 2π · f (x∗i ) 1 + [f 0 (x∗i )] ∆x.
Therefore,
n
X n
X q
2
S = lim Si = lim 2π · f (x∗i ) 1 + [f 0 (x∗i )] ∆x
n→∞ n→∞
i=1 i=1
ˆ b q
2
= 2π f (x) 1 + [f 0 (x)] dx.
a
5.4 Surface Area of Revolution 299

5.34 Theorem: general formula for surface area of revolution

For rotations about the x-axis and the y-axis, respectively, the surface area formula becomes
ˆ ˆ
S = 2πy ds or S = 2πx ds,

where s s
2 2
dy dx
 
ds = 1+ dx or 1+ dy.
dx dy
In general, the surface area of rotation is
ˆ
(2πr) (ds)
| {z } |{z}
circumference arc length

y
y

5.35 Example: surface area of a sphere

A sphere can be generated by rotating the semicircle

p
y = r 2 − x2 , −r ≤ x ≤ r,

about the x-axis, as shown in the figure. Find its surface area.

y
√
y= r 2 − x2

x
−r O r
300 Applications of Integration

Solution It is easy to see that

dy x
= −√ .
dx r 2 − x2
Thus, the surface area of the sphere is
ˆ
s 2
r 
p x
S= 2π r 2 − x2 1 + −√ dx
−r r 2 − x2
ˆ r
= 2π r dx = 4πr2 .
−r

5.36 Example: surface area of a torus

A torus can be generated by rotating the circle

(x − R)2 + y 2 = r2 , 0 < r < R,

about the y-axis, as shown in the figure.

(x − R)2 + y 2 = r2

x
O R

The torus
p consists of two portions: the outer portion generated by the rightmost semicircle
x =pR + r2 − y 2 (−r ≤ y ≤ r), and the inner portion by the leftmost semicircle x =
R − r2 − y 2 (−r ≤ y ≤ r). It is easy to see that

dx y
along the rightmost semicircle : = −p ;
dy r − y2
2
dx y
along the leftmost semicircle : =p .
dy r2 − y2
5.4 Surface Area of Revolution 301

Thus, the surface area of the torus is

ˆ
v !2
u
r  p u y
S= 2π R + r − y t1 +
2 2 −p dy
−r r2 − y2
ˆ
v !2
u
r p  u y
+ 2π R − r − y t1 +
2 2 p dy
−r r2 − y2
ˆ r
r
= 4πR p dy
−r r2 − y2
ˆ r
r
= 8πR p dy even function
0 r2 − y2
ˆ π
2 r cos θ dθ
= 8πRr y = r sin θ
0 r cos θ
π
= 8πRr · = 4π 2 Rr.
2

5.37 MATLAB: surface area of a sphere

A sphere can be generated by rotating the semicircle

p
y = 32 − x2 , −3 ≤ x ≤ 3,

about the x-axis. Find its surface area.

Solution We calculate the surface area of the sphere by two methods.
1. The surface area of the sphere is calculated by using the formula for surface area of
revolution.

>> syms x
>> y ( x ) = sqrt (3^2 - x ^2) ;
>> dy ( x ) = diff ( y ( x ) ,x )

dy ( x ) =

-x /(9 - x ^2) ^(1/2)

>> int_fun = 2* pi * y ( x ) * sqrt (1 + dy ( x ) ^2) ; % integrand

>> S = int ( int_fun , -3 ,3)

S =

36* pi

The result above shows that the surface area of the sphere is 36π.
302 Applications of Integration

2. An approximation of the surface area is calculated based on the definition.

R = 3; % radius of the sphere

N = 1000; % number of line segments

x_array = linspace ( -R ,R , N ) ;
y_array = sqrt (3^2 - x_array .^2) ;
S_sum = 0;

for ind = 2: length ( x_array )

x2 = x_array ( ind ) ;
x1 = x_array ( ind - 1) ;
y2 = y_array ( ind ) ;
y1 = y_array ( ind - 1) ;
ell = sqrt (( x2 - x1 ) ^2 + ( y2 - y1 ) ^2) ;
S = 2* pi *0.5*( y1 + y2 ) * ell ;
S_sum = S_sum + S ;

end

Note that the bigger the number of line segments N is, the more accurate the approxi-
mation is.
The result of the surface area can be seen when we type in S_sum in the command
window, as shown below.

>> format long

>> S_sum

S_sum =

1.1 309702 679604 28 e +02

It is shown that when N = 1000, the approximation is 113.0970267960428, which is

very close to the exact result 36π ≈ 113.0973355292326.

Surface Area of Revolution by Parametric Curve

5.38 Theorem: surface area of revolution by parametric curve

Suppose a parametric curve is given by the parametric equations

x = f (t), y = g(t), α ≤ t ≤ β,
where f and g are continuous and g(t) ≥ 0. Then the surface area of revolution about the
0 0
5.4 Surface Area of Revolution 303

x-axis is
ˆ ˆ
s 2 2
β
dx dy β
 q
2 2
S= 2πy + dt = 2πg(t) [f 0 (t)] + [g 0 (t)] dt.
α dt dt α

Similarly, the surface area of revolution about the y-axis is

ˆ β ˆ β
s 
2  2
dx dy
q
2 2
S= 2πx + dt = 2πf (t) [f 0 (t)] + [g 0 (t)] dt.
α dt dt α

Justification:
The general formula for the surface area of revolution about the x-axis is
ˆ
S = 2πy ds.

For the parametric curve we have

ds = [f 0 (t)]2 + [g 0 (t)]2 dt.

p

Thus, by the substitution rule, we get

ˆ β q
2 2
S= 2πg(t) [f 0 (t)] + [g 0 (t)] dt.
α

5.39 Example: solid generated by a parametric curve

Rotate the area under one arch of the cycloid

x = r(t − sin t), y = r(1 − cos t), 0 ≤ t ≤ 2π,

about the x-axis and the y-axis, respectively. Find the surface areas of the two solids.

2r

x
O 2πr

Solution Generally, the surface area of revolution is

ˆ
(2πr) (ds) .
| {z } |{z}
circumference arc length
304 Applications of Integration

1. When the area under one arch of the cycloid rotates about the x-axis, a typical circle
has radius y = y(t), while the arc length is given by

ds = x0 (t)2 + y 0 (t)2 dt.

p

2r

x
O 2πr

Thus, the surface area of the solid is

ˆ 2π
S= 2πy(t) · x0 (t)2 + y 0 (t)2 dt
p
0
ˆ 2π q
02 02
= 2π · r(1 − cos t) · [r(t − sin t)] + [r(1 − cos t)] dt
0
ˆ 2π q
= 2πr 2
(1 − cos t) (1 − cos t)2 + sin2 t dt
0
ˆ 2π √
= 2πr2 (1 − cos t) · 2 − 2 cos t dt
0
ˆ 2π
= 2πr2 2 sin2 ( 12 t) · 2 sin( 12 t) dt half-angle formula
ˆ0
π
= 8πr2 (1 − cos2 u) sin u · 2 du t = 2u
0
π
= 16πr2 − cos u + 13 cos3 u u=0 = 16πr2 · 4
= 64
 2
3 3 πr .

2. When the area under one arch of the cycloid rotates about the y-axis, a typical circle
has radius x = x(t), while the arc length is given by

ds = x0 (t)2 + y 0 (t)2 dt.

p

x(t)
2r

x
O 2πr
5.4 Surface Area of Revolution 305

Thus, the surface area of the solid is

ˆ 2π
S= 2πx(t) · x0 (t)2 + y 0 (t)2 dt
p
0
ˆ 2π q
02 02
= 2π · r(t − sin t) · [r(t − sin t)] + [r(1 − cos t)] dt
0
ˆ 2π q
= 2πr2 (t − sin t) (1 − cos t)2 + sin2 t dt
0
ˆ 2π √
= 2πr2 (t − sin t) · 2 − 2 cos t dt
0
ˆ 2π
= 2πr2 (t − sin t) · 2 sin( 12 t) dt half-angle formula
ˆ0
π
= 4πr2 (2u − sin 2u) sin u · 2 du t = 2u
ˆ0
π
= 8πr2 (2u sin u − sin 2u sin u) du
0
π
= 8πr2 −2u cos u + 3
sin u + 1
sin 3u u=0 = 8πr2 · 2π = 16π 2 r2 .

2 6

5.40 MATLAB: solid generated by a parametric curve

Rotate the area under one arch of the cycloid

x = 3(t − sin t), y = 3(1 − cos t), 0 ≤ t ≤ 2π,

about the x-axis. Find the surface area of the generated solid.
Solution We calculate the surface area of the solid by two methods.
1. The surface area of the solid is calculated by using the formula for surface area of
revolution.

>> syms t
>> x ( t ) = 3*( t - sin ( t ) ) ;
>> y ( t ) = 3*(1 - cos ( t ) ) ;
>> dx ( t ) = diff ( x ( t ) ,t ) ;
>> dy ( t ) = diff ( y ( t ) ,t ) ;
>> int_fun = 2* pi * y ( t ) * sqrt ( dx ( t ) ^2 + dy ( t ) ^2)

int_fun =

-6* pi *(3* cos ( t ) - 3) *( sin ( t ) ^2 + (3* cos ( t ) - 3) ^2/9) ^(1/2)

>> int_fun2 = matlabFunction ( int_fun ) ; % convert to ...

function handle
>> S = integral ( int_fun2 ,0 ,2* pi )
306 Applications of Integration

S =

603.1858

The result above shows that the surface area of the solid is 603.1858. Note that the com-
mand integral can only deal with a function handle, so we need to convert the symbolic
expression int_fun to the function handle int_fun2 using command matlabFunction.

2. An approximation of the surface area is calculated based on the definition.

N = 1000; % number of line segments

t_array = linspace (0 ,2* pi , N ) ;

x_array = 3*( t_array - sin ( t_array ) ) ;
y_array = 3*(1 - cos ( t_array ) ) ;
S_sum = 0;

for ind = 2: length ( x_array )

x2 = x_array ( ind ) ;
x1 = x_array ( ind - 1) ;
y2 = y_array ( ind ) ;
y1 = y_array ( ind - 1) ;
ell = sqrt (( x2 - x1 ) ^2 + ( y2 - y1 ) ^2) ;
S = 2* pi *0.5*( y1 + y2 ) * ell ;
S_sum = S_sum + S ;

end

Note that the bigger the number of line segments N is, the more accurate the approxi-
mation is.
The result of the surface area can be seen when we type in S_sum in the command
window, as shown below.

>> S_sum

S_sum =

603.1845

It is shown that when N = 1000, the approximation is 603.1845.

5.5 Lengths and Areas in Polar Coordinates 307

5.5 Lengths and Areas in Polar Coordinates

Arc Length of Polar Curve

5.41 Theorem: arc length of polar curve

Suppose a polar curve is given by the equa-

tion r = f (θ). If f 0 is continuous, then the
arc length of the polar curve from θ = α to r = f (θ)
θ = β is
ˆ βq θ=β
2 2
L= [f (θ)] + [f 0 (θ)] dθ
α θ=α
ˆ β θ=0
O
= r2 + (dr/dθ)2 dθ.
p
α

Justification:
We write its parametric equations as

x = r cos θ = f (θ) cos θ, y = r sin θ = f (θ) sin θ.

Since
2 2
dx dy
 
2 2
+ = [f 0 (θ) cos θ − f (θ) sin θ] + [f 0 (θ) sin θ + f (θ) cos θ]
dθ dθ
2 2
= [f (θ)] + [f 0 (θ)] ,

the length is
ˆ ˆ
s 2 2
β
dx dy β
 q
2 2
L= + dθ = [f (θ)] + [f 0 (θ)] dθ.
α dθ dθ α

5.42 Example: arc length of polar curve

Find the arc length of the cardioid

r = a + a cos θ,

where a > 0.
308 Applications of Integration

105◦ 90◦ 75◦

120◦ 60◦

135◦ 45◦

150◦ 30◦

◦
165 15◦

180◦ 0◦
a

195◦ 345◦

◦
210 330◦

◦ ◦
225 315

240◦ 300◦
◦ ◦
255 270◦ 285

Solution The given polar curve r = a+a cos θ is periodic of period 2π. Since r0 = −a sin θ,
we get the arc length
ˆ 2π p
L= (a + a cos θ)2 + (−a sin θ)2 dθ
0
ˆ 2π √
= a 2 + 2 cos θ dθ
0
ˆ 2π
= 2a cos( 1 θ) dθ
 
2
0
ˆ π ˆ 2π
= 2a cos( 21 θ) dθ + 2a (−1) cos( 21 θ) dθ
0 π
π 2π
= 2a 2 sin( 12 θ) θ=0 − 2a 2 sin( 12 θ) θ=π = 8a.
 

5.43 MATLAB: arc length of polar curve

Find the arc length of the cardioid

r = 1 + cos θ.

The arc length of the cardioid can be calculated in MATLAB using the following code:

>> syms theta

>> r ( theta ) = 1 + cos ( theta ) ;
>> dr ( theta ) = diff ( r ( theta ) , theta )

dr ( theta ) =

- sin ( theta )

>> int_fun = sqrt ( r ( theta ) ^2 + dr ( theta ) ^2) ;

>> arc_len = int ( int_fun , theta ,0 ,2* pi )

arc_len =
5.5 Lengths and Areas in Polar Coordinates 309

The result above shows that the arc length of the cardioid is 8.

Area of Polar Region

5.44 Theorem: area of polar region

Let R be a polar region bounded by the po-

lar curve r = f (θ) and by the rays θ = α to
θ = β, where f is a positive continuous func- r = f (θ)
tion and the angles satisfy 0 < β − α ≤ 2π. R
Then the area of R is θ=β

ˆ β ˆ β
θ=α
A= 1 2
2 r dθ = 2 [f (θ)] dθ.
1 2
θ=0
α α O

Theorem: area between two polar curves

Let R be a polar region bounded by

the polar curves r = f (θ), r = g(θ),
and by the rays θ = α to θ = β, where
f (θ) ≥ g(θ) ≥ 0 and the angles satisfy R r = f (θ)

0 < β − α ≤ 2π. Then the area A of the θ=β

polar region R is r = g(θ)

ˆ β   θ=α
2 2
A= 1
2 [f (θ)] − [g(θ)] dθ. O
θ=0
α

Justification:

For the polar curve r = f (θ), we write its parametric equations as

x = r cos θ = f (θ) cos θ, y = r sin θ = f (θ) sin θ.

We divide the interval [α, β] into subintervals of equal width ∆θ with endpoints
θ0 , θ1 , θ2 , . . . , θn . If we choose θi∗ in the ith subinterval [θi−1 , θi ], then the area ∆Ai of
the ith region is approximated by the area of the sector of a circle with central angle ∆θ and
310 Applications of Integration

radius f (θi∗ ). The area can be approximated by

(central angle)
∆Ai ≈ π(radius)2 · = 21 (radius)2 × (central angle)
2π
2
= 12 [f (θi∗ )] ∆θ.

So, an approximation to the total area A of R is

n
X 2
A≈ 1
2 [f (θi∗ )] ∆θ.
i=1

Hence, we obtain the formula for the area A of the polar region R
n
X ˆ β
2 2
A = lim 1
2 [f (θi∗ )] ∆θ = 1
2 [f (θ)] dθ.
n→∞ α
i=1

θ = θi

θ = θi−1
f (θi∗ )

r = f (θ)

r
θ=β
θ

θ=α

area = 12 r2 θ O
θ=0

5.45 Example: area between two polar curves

Find the area of the region that lies inside the circle r = 9 cos θ outside of the cardioid
r = 3 + 3 cos θ.

105◦ 90◦ 75◦

120◦ 60◦

135◦ 45◦
π
θ= 3
150◦ 30◦

◦
165 15◦

180◦ 0◦
2 4 6 8

195◦ 345◦

◦
210 330◦
θ = − π3
◦ ◦
225 315

240◦ 300◦
255◦ 270 ◦ 285◦

Solution The circle r = 9 cos θ makes a complete rotation when θ varies from − π2 to 2;
π
5.5 Lengths and Areas in Polar Coordinates 311

the cardioid r = 3 + 3 cos θ make a complete loop when θ varies from −π to π. By solving the
simultaneous equations r = 9 cos θ and r = 3 + 3 cos θ, we find the intersections at θ = − π3
and θ = π3 . Since r(−θ) = r(θ) for both polar equations, these curves are symmetric about
the polar axis.
Thus, the area of the region that lies inside the circle outside of the cardioid is
ˆ π
3
A=2 1
(9 cos θ)2 − (3 + 3 cos θ)2 dθ
 
2
0
ˆ π
3
= (−9 − 18 cos θ + 72 cos2 θ) dθ
0
ˆ π
3
= [−9 − 18 cos θ + 36(1 + cos 2θ)] dθ
0
π
= [27θ − 18 sin θ + 18 sin 2θ]θ=0
3

= 9π.

5.46 Example: area between two polar curves

The limaçon r = 3 + 5 cos θ consists of an inner loop and an outer loop. Find the area
between these loops.
105◦ 90◦ 75◦
120◦ 60◦

135◦ 45◦

◦
150 30◦
θ = cos−1 (− 53 )

165◦ 15◦

◦
180 0◦
2 4 6 8

195◦ 345◦

◦
θ = 2π − cos−1 (− 35 )
210 330◦

◦ ◦
225 315

240◦ 300◦
◦ ◦
255 270◦ 285

Solution The given polar curve r = 3+5 cos θ is periodic of period 2π. Since r(−θ) = r(θ),
the polar graph is symmetric to the polar axis. As θ varies from 0 to arccos(− 35 ), the polar
equation describes the upper-half outer loop; as θ varies from π to 2π − arccos(− 35 ), it
describes the upper-half inner loop. Thus, the area between these loops is
ˆ arccos(− 35 ) ˆ 2π−arccos(− 35 )
A=2 1
2 (3 + 5 cos θ)2
dθ − 2 2 (3 + 5 cos θ) dθ.
1 2
0 π
We can compute the following indefinite integral:
ˆ ˆ
(3 + 5 cos θ)2 dθ = (9 + 30 cos θ + 25 cos2 θ) dθ
ˆ
= 9 + 30 cos θ + 2 (1
25
+ cos 2θ) dθ
 

= 43
2 θ + 30 sin θ + 25
4 sin 2θ + C.
312 Applications of Integration

Hence,
arccos(− 35 )
A= θ + 30 sin θ + sin 2θ
 43 25
2 4 θ=0
2π−arccos(− 35 )
− 2 θ + 30 sin θ + sin 2θ θ=π
 43 25
4
= 2 · arccos(− 5 ) + 30 · ( 5 ) + 4 · 2( 5 )(− 5 )
43 3 4 25 4 3

2 · 2π − arccos(− 5 ) − 30 · (− 5 ) − 4
− 43 · 2(− 45 )( 35 ) +
3 4 25 43
 
2 ·π
= 36 − 432 π + 43 · arccos(− 5 )
3

≈ 63.6705.

5.47 MATLAB: area between two polar curves

Find the area of the region that lies inside the circle r = 9 cos θ outside of the cardioid
r = 3 + 3 cos θ.
First, we plot these two polar curves in MATLAB.

>> syms theta

>> r1 ( theta ) = 9* cos ( theta ) ;
>> r2 ( theta ) = 3 + 3* cos ( theta ) ;
>> theta_array = linspace ( - pi , pi ,200) ;
>> figure
>> polarplot ( theta_array , r1 ( theta_array ) , 'r ')
>> hold on
>> polarplot ( theta_array , r2 ( theta_array ) , 'b ')

The figure plotted is shown below.

90

120 60
8

6
150 30
4

180 0 0

210 330

240 300

270

In order to calculate the area of the specified region, we need to first find the intersections
of these two polar curves, which is shown in the following:
5.6 Moments, Centers of Mass, and Centroids 313

>> solve ( r1 ( theta ) == r2 ( theta ) )

ans =

- pi /3
pi /3

We see that the circle r = 9 cos θ intersects with the cardioid r = 3 + 3 cos θ at θ = − π3 and
θ = π3 . Then, we can calculate the area of the specified region below.

>> int_fun = 0.5*( r1 ( theta ) ^2 - r2 ( theta ) ^2) ;

>> S = int ( int_fun , - pi /3 , pi /3)

S =

9* pi

The area of the specified region is thus 9π.

5.6 Moments, Centers of Mass, and Centroids

Discrete Systems

5.48 Theorem: moment and center of mass for discrete system

Theorem: moment and center of mass for one-dimensional discrete system

For a system of n particles with masses m1 , m2 , . . . , mn located at the points

x1 , x2 , . . . , xn on the x-axis, the center of mass of the system is located at
n
P
mi xi
M i=1
x̄ = = Pn ,
m
mi
i=1

n
where M = mi xi is the moment of the system about the origin.
P
i=1
314 Applications of Integration

Theorem: moments and center of mass for two-dimensional discrete system

For a system of n particles with masses m1 , m2 , . . . , mn located at the points

(x1 , y1 ), (x2 , y2 ) . . . , (xn , yn ) in the xy-plane, the center of mass of the system is lo-
cated at (x̄, ȳ), where
n
P n
P
m i xi mi yi
My Mx
x̄ = = i=1
n , ȳ = = i=1
n .
m P m P
mi mi
i=1 i=1

n
Here My = mi xi is the moment of the system about the y-axis and Mx =
P
i=1
n
mi yi is the moment of the system about the x-axis.
P
i=1

Justification:
Consider a situation with two particles with masses m1 and m2 on a rod that lies along
x-axis. Suppose the distances from the masses to fulcrum are d1 and d2 respectively. The
rod will balance if
m1 d1 = m2 d2 .

d1 d2
m1 m1

Denote the coordinates of m1 and m2 to be x1 and x2 , and the center of mass at x̄. We
see that d1 = x̄ − x1 and d2 = x2 − x̄. Then we have
m1 x1 + m2 x2
m1 (x̄ − x1 ) = m2 (x2 − x̄) =⇒ x̄ = .
m1 + m2
If we have a system of n particles with masses m1 , m2 , . . . , mn located at the points
x1 , x2 , . . . , xn on the x-axis, and if the rod balances, then the center of mass x̄ satisfies:
n
P
n
X mi xi
mi (xi − x̄) = 0 =⇒ x̄ = i=1
Pn .
i=1 mi
i=1

For a system of n particles with masses m1 , m2 , . . . , mn located at the points

(x1 , y1 ), (x2 , y2 ) . . . , (xn , yn ) in the xy-plane, we assume that the center of mass of the system
is located at (x̄, ȳ). Then, similarly, we have
5.6 Moments, Centers of Mass, and Centroids 315

 n y
X
mi (xi − x̄) = 0,





i=1 m3
 n
X mn
mi (yi − ȳ) = 0,

 m2


m5
i=1 m8
x
so that (x̄, ȳ)
m4
n n
P P m1
mi xi mi yi m7
My Mx
x̄ = = i=1
n , ȳ = = i=1
n .
m P m P m6
mi mi
i=1 i=1

Center of Mass Along a Curve

5.49 Theorem: center of mass along a curve

Consider a continuous system of mass distributed along a thin curve C : y = f (x), a ≤ x ≤

b, with density ρ(x). Then the center of mass is (x̄, ȳ), where
ˆ
1 b
x̄ = 1 + [f 0 (x)]2 dx,
p
ρ(x) · x
m a
ˆ
1 b
ȳ = ρ(x) · f (x) 1 + [f 0 (x)]2 dx,
p
m a
ˆ b
and m = 1 + [f 0 (x)]2 dx is the total mass along the curve C .
p
ρ(x) ·
a

Theorem: centroid of a curve

In particular, if ρ(x) = ρ is a constant, then the center of mass is the same as the
centroid of the curve, with
ˆ
1 b
x̄ = 1 + [f 0 (x)]2 dx,
p
x
L a
ˆ
1 b
ȳ = f (x) 1 + [f 0 (x)]2 dx,
p
L a
ˆ b
where L = 1 + [f 0 (x)]2 dx is the length of the curve C .
p
a
316 Applications of Integration

Justification:

We divide the interval [a, b] into n subintervals with endpoints x0 , x1 , . . . , xn and equal
width ∆x. Then the ith arc can be approximated by a line segment from (xi−1 , f (xi−1 )) to
(xi , f (xi )), whose approximate length is, by the Mean value Theorem,
s 2
f (xi ) − f (xi−1

(∆x) + (f (xi ) − f (xi−1 )) = 1 + ∆x
p
2 2
xi − xi−1
q
2
= 1 + [f 0 (x∗i )] ∆x.

The mass along the ith piece of arc is approximately

q
2
ρ(x∗i ) 1 + [f 0 (x∗i )] ∆x.

so the moment about the y-axis is

q
2
ρ(x∗i ) 1 + [f 0 (x∗i )] ∆x · x∗i .

Adding these moments, we obtain the approximate moment of C . By taking the limit as
n → ∞ we obtain the moment of C about the y-axis:
n
X q
2
My = lim ρ(x∗i )x∗i 1 + [f 0 (x∗i )] ∆x
n→∞
i=1
ˆ b q
2
= ρ(x) · x 1 + [f 0 (x)] dx.
a

In a similar fashion, the moment along the ith piece of arc about the x-axis is
q
2
ρ(x∗i ) 1 + [f 0 (x∗i )] ∆x · f (x∗i )

so we get the moment about the x-axis:

n
X q
2
Mx = lim ρ(x∗i )f (x∗i ) 1 + [f 0 (x∗i )] ∆x
n→∞
i=1
ˆ b q
2
= ρ(x) · f (x) 1 + [f 0 (x)] dx.
a

The mass along the ith piece of arc is

q
2
ρ(x∗i ) 1 + [f 0 (x∗i )] ∆x,

so we get the total mass of the curve C :

ˆ b q
2
m= ρ(x) · 1 + [f 0 (x)] dx.
a
5.6 Moments, Centers of Mass, and Centroids 317

 
My Mx
Hence, the center of mass of the curve is located at the point (x̄, ȳ) = , , that is,
m m
ˆ b q ˆ b q
2 2
ρ(x) · x 1 + [f 0 (x)]
dx ρ(x) · f (x) 1 + [f 0 (x)] dx
x̄ = ˆa b q , ȳ = aˆ b q .
2 2
ρ(x) · 1 + [f (x)] dx
0 ρ(x) · 1 + [f (x)] dx
0
a a

It is easy to deduce the centroid from the center of mass when the density is a constant.

5.50 Example: centroid of a semicircle

√ Find the centroid of the semicircle y =

y
r2 − x2 (−r ≤ x ≤ r). y=
√
r 2 − x2

Solution By symmetry, the centroid (x̄, ȳ)

must lie on the y-axis, so that x̄ = 0. The
length of the semicircle is L = πr. Since −r O r
x

1 x
y 0 = 12 (r2 − x2 )− 2 · (−2x) = − √ .
r2 − x2
we have
ˆ
s 2
1 r 
p x
ȳ = 1 + −√
r 2 − x2 · dx
L −r r 2 − x2
ˆ r
1 1 2
= r dx = · 2r2 = r.
πr −r πr π
Hence, the centroid of the semicircle is at y
(x̄, ȳ) = (0, π2 r). y=
√
r 2 − x2

0, π2 r

x
−r O r

5.51 MATLAB: centroid of a semicircle

√
Find the centroid of the semicircle y = 32 − x2 (−3 ≤ x ≤ 3).
Solution First, we notice that the length of the semicircle is L = 3π. Then, by symmetry,
we know that the centroid (x̄, ȳ) must lie on the y-axis, so that x̄ = 0. We will calculate ȳ
below.

>> syms x
318 Applications of Integration

>> f ( x ) = sqrt (3^2 - x ^2) ;

>> df ( x ) = diff ( f ( x ) ,x )

df ( x ) =

-x /(9 - x ^2) ^(1/2)

>> L = 3* pi ;
>> int_fun_y = f ( x ) * sqrt (1 + df ( x ) ^2) ;
>> cen_y = int ( int_fun_y , -3 ,3) / L

cen_y =

6/ pi

The result above shows that ȳ = π.

6

5.52 Theorem: center of mass along a parametric curve

Consider a continuous system of mass distributed along a parametric curve C given by the
parametric equations
x = f (t), y = g(t), α ≤ t ≤ β,
with density ρ(t). Suppose f 0 and g 0 are continuous. Then the center of mass of C is (x̄, ȳ),
where
ˆ
1 β
x̄ = ρ(t) · f (t) [f 0 (t)]2 + [g 0 (t)]2 dt,
p
L α
ˆ
1 β
ȳ = ρ(t) · g(t) [f 0 (t)]2 + [g 0 (t)]2 dt,
p
L α
ˆ β
and m = [f 0 (t)]2 + [g 0 (t)]2 dt is the total mass of C .
p
ρ(t) ·
α

Theorem: centroid of a parametric curve

In particular, if ρ(t) = ρ is a constant, then the center of mass is the same as the
centroid of the curve, with
ˆ
1 β
x̄ = f (t) [f 0 (t)]2 + [g 0 (t)]2 dt,
p
L α
ˆ
1 β
ȳ = [f 0 (t)]2 + [g 0 (t)]2 dt,
p
g(t)
L α
ˆ β
where L = [f 0 (t)]2 + [g 0 (t)]2 dt is the length of the curve C .
p
α
5.6 Moments, Centers of Mass, and Centroids 319

Justification:
The moment of C about the y-axis is
ˆ
My = ρx ds.

For the parametric curve we have

ds = [f 0 (t)]2 + [g 0 (t)]2 dt.

p

Thus, by the substitution rule, we get

ˆ β q
2 2
My = ρ(t) · f (t) [f 0 (t)] + [g 0 (t)] dt.
α

Hence, we get
qˆ β
2 2
ρ(t) · f (t) [f 0 (t)] + [g 0 (t)] dt
My
x̄ = = ˆ β
α
.
m
q
2 2
ρ(t) · [f (t)] + [g (t)] dt
0 0
α
Similarly, we have
q ˆ β
2 2
ρ(t) · g(t) [f 0 (t)] + [g 0 (t)] dt
Mx
ȳ = = ˆ β
α
.
m
q
2 2
ρ(t) · [f (t)] + [g (t)] dt
0 0
α

5.53 Example: centroid of a parametric curve

Find the centroid of one arch of the cycloid y

x = r(t − sin t), y = r(1 − cos t), 2r

where 0 ≤ t ≤ 2π. x
O 2πr

Solution The length of one arch of the cycloid is

ˆ 2π q
2
L= [r(1 − cos t)] + (−r sin t)2 dt
0
ˆ 2π √
=r 2 − 2 cos t dt
0
ˆ 2π
= 2r sin( 12 t) dt half-angle formula
0
2π
= 2r −2 cos( 21 t) t=0 = 8r.

320 Applications of Integration

Without loss of generality, assume the density ρ = 1. Then the moment of the curve about
the y-axis is
ˆ 2π q
02 02
My = r(t − sin t) · [r(t − sin t)] + [r(1 − cos t)] dt
0
ˆ 2π q
= r2 (t − sin t) (1 − cos t)2 + sin2 t dt
0
ˆ 2π √
= r2 (t − sin t) · 2 − 2 cos t dt
0
ˆ 2π
= r2 (t − sin t) · 2 sin( 12 t) dt half-angle formula
ˆ
0
π
= 2r2 (2u − sin 2u) sin u · 2 du t = 2u
ˆ0 π
= 4r2 (2u sin u − sin 2u sin u) du
0
π
= 4r2 −2u cos u + 23 sin u + 16 sin 3u u=0


= 4r2 · 2π = 8πr2 ;

the moment of the curve about the x-axis is

ˆ 2π q
2
Mx = r(1 − cos t) · [r(1 − cos t)] + (−r sin t)2 dt
0
ˆ 2π q
=r 2
(1 − cos t) (1 − cos t)2 + sin2 t dt
0
ˆ 2π √
= r2 (1 − cos t) · 2 − 2 cos t dt
0
ˆ 2π
= r2 2 sin2 ( 12 t) · 2 sin( 21 t) dt half-angle formula
ˆ
0
π
= 4r2 (1 − cos2 u) sin u · 2 du t = 2u
0
π
= 8r2 − cos u + 13 cos3 u u=0


= 8r2 · 4
3 = 32 2
3 r .

Hence the centroid of the curve is (x̄, ȳ), y

where
My 8πr2 2r
x̄ = = = πr,
8r πr, 43 r


L
32 2 x
Mx 3 r
O 2πr
ȳ = = = 43 r.
L 8r
5.6 Moments, Centers of Mass, and Centroids 321

5.54 MATLAB: centroid of a parametric curve

Find the centroid of one arch of the cycloid

x = 3(t − sin t), y = 3(1 − cos t),

where 0 ≤ t ≤ 2π.
Solution For one arch of the cycloid, the centroid is calculated in MATLAB using the
following code:

>> syms t
>> x ( t ) = 3*( t - sin ( t ) ) ;
>> y ( t ) = 3*(1 - cos ( t ) ) ;
>> dx ( t ) = diff ( x ( t ) ,t ) ;
>> dy ( t ) = diff ( y ( t ) ,t ) ;
>> int_fun_L = sqrt ( dx ( t ) ^2 + dy ( t ) ^2) ;
>> L = int ( int_fun_L ,0 ,2* pi )

L =

24

>> int_fun_x = x ( t ) * sqrt ( dx ( t ) ^2 + dy ( t ) ^2) ;

>> cen_x = int ( int_fun_x ,0 ,2* pi ) / L

cen_x =

3* pi

>> int_fun_y = y ( t ) * sqrt ( dx ( t ) ^2 + dy ( t ) ^2) ;

>> int_fun_y2 = matlabFunction ( int_fun_y ) ; % convert to ...
function handle
>> cen_y = integral ( int_fun_y2 ,0 ,2* pi ) / L

cen_y =

The result above shows that the centroid is (3π, 4). Note that the command int cannot
ˆ 2π q
0 0
solve the integral 3 (1 − cos t) (3(t − sin t)) + (3(1 − cos t)) dt, so we use the command
0
integral to numerically evaluate the integral.
322 Applications of Integration

Centroid for Planar Region

5.55 Theorem: centroid for planar region between two curves

Consider a flat plate (called a lamina) with uniform density ρ that occupies a region R. If
R lies between two curves y = f (x) and y = g(x), where f (x) ≥ g(x) for a ≤ x ≤ b, then the
centroid of R is (x̄, ȳ), where
ˆ ˆ
1 b
1 b n
2 2
o
x̄ = x [f (x) − g(x)] dx, ȳ = 1
2 [f (x)] − [g(x)] dx,
A a A a
ˆ b
where A = [f (x) − g(x)] dx is the area of R.
a

Justification: Part 1

First we consider the case that g(x) ≡ 0.

We divide the interval [a, b] into n subintervals with endpoints x0 , x1 , . . . , xn and equal
width ∆x. Denote x̄i = 21 (xi−1 + xi ) to be the midpoint of the ith subinterval. Then the
centroid of the ith approximating rectangle is (x̄i , 21 f (x̄i )). Its area is f (x̄i )∆x, so its mass
is ρf (x̄i )∆x. Thus, for the ith approximating rectangle, the moment about the y-axis is
ρf (x̄i )∆x · x̄i
Adding these moments, we obtain the moment of the polygonal approximation to R. By
taking the limit as n → ∞ we obtain the moment of R about the y-axis:
X n ˆ b
My = lim ρx̄i f (x̄i )∆x = ρ xf (x) dx.
n→∞ a
i=1

In a similar fashion, the moment of the ith approximating rectangle about the x-axis is
ρf (x̄i )∆x · 21 f (x̄i )
so we get the moment of R about the x-axis:
n
X ˆ b
2 2
Mx = lim ρ· 1
2 [f (x̄i )] ∆x = ρ 1
2 [f (x)] dx.
n→∞ a
i=1

2 f (x̄i )
1

x0 x1 xi−1 xi xn
x
x̄i
5.6 Moments, Centers of Mass, and Centroids 323

The mass of the lamina is ˆ b

m = ρA = ρ f (x) dx.
a

So, the center of mass of the plate (or the centroid of R since the density is uniform) is
located at the point (x̄, ȳ), where
ˆ b ˆ b
2
xf (x) dx 1
2 [f (x)] dx
My Mx
x̄ = = ˆa b , ȳ = = a
ˆ b
.
m m
f (x) dx f (x) dx
a a

Justification: Part 2

Next, for g(x) 6≡ 0, the mass of the lamina is

ˆ b
m = ρA = ρ [f (x) − g(x)] dx,
a

where A is the area bounded by the functions y = f (x) and y = g(x) over the interval [a, b].
The moment of R about the y-axis and about the x-axis are, respectively,
ˆ b
My = ρ x [f (x) − g(x)] dx,
a
ˆ b n o
2 2
Mx = ρ 1
2 [f (x)] − [g(x)] dx.
a

Hence,
ˆ ˆ
1 b
1 b n
2 2
o
x̄ = x [f (x) − g(x)] dx, ȳ = 1
2 [f (x)] − [g(x)] dx.
A a A a

5.56 Example: centroid of a semicircular plate

y
As shown in the figure, we place the plate y=
√
r 2 − x2
so
√ that the circular arc is y = f (x) =
r2 − x2 (−r ≤ x ≤ r). Find the centroid
of the plate.
x
−r O r

Solution By symmetry, the centroid (x̄, ȳ) must lie on the y-axis, so that x̄ = 0. The area
324 Applications of Integration

of the semicircular plate is A = 12 πr2 . Thus,

ˆ
1 r 1 2
ȳ = [f (x)] dx
A −r 2
ˆ r p
1 2
= 1 2· 1
2 r 2 − x2 dx
2 πr −r
ˆ r
2
= 2 r2 − x2 dx even function


πr 0 y
2  r
= 2 r2 x − 13 x3 x=0 = 3π 4
r. y=
√
r 2 − x2
πr
Hence, the centroid of the semicircular
0, 3π
4


r
plate is at (x̄, ȳ) = (0, 3π
4
r).
x
−r O r

5.57 Example: centroid of a cycloid plate

Find the centroid of the plate between one y

arch of the cycloid
2r
x = r(t − sin t), y = r(1 − cos t),
x
2πr
where 0 ≤ t ≤ 2π, and the x-axis. O

Solution By applying the substitution rule using the parametric equations, the area of
one arch of the cycloid is
ˆ 2πr
A= y dx
0
ˆ 2π
x = r(t − sin t)

0
= r(1 − cos t) · [r(t − sin t)] dt
0 y = r(1 − cos t)
ˆ 2π
= r2 (1 − 2 cos t + cos2 t) dt
0
ˆ 2π
= r2 1 − 2 cos t + 12 (1 + cos 2t) dt double-angle formula
 
0
2π
= r2 − 2 sin t + sin 2t = 3πr2 .
3 1
2t 4 t=0

Without loss of generality, assume the density ρ = 1. Then the moment of the plate about
5.6 Moments, Centers of Mass, and Centroids 325

the y-axis is
ˆ 2πr
My = xy dx
0
ˆ 2π
x = r(t − sin t)

0
= r(t − sin t) · r(1 − cos t) · [r(t − sin t)] dt
0 y = r(1 − cos t)
ˆ 2π
= r3 (t − sin t)(1 − cos t)2 dt = 3π 2 r3 ;
0

the moment of the plate about the x-axis is

ˆ 2πr
Mx = 2 y dx
1 2
0
ˆ 2π
x = r(t − sin t)

= 2 [r(1
1
− cos t)]2 · [r(t − sin t)]0 dt
0 y = r(1 − cos t)
ˆ 2π
= 1 3
2r (1 − cos t)3 dt = 52 πr3 .
0

Hence the centroid of the plate is (x̄, ȳ), y

where
My 3π 2 r3 2r
x̄ = = = πr,
3πr2 πr, 56 r


L
5 x
Mx πr3 O 2πr
ȳ = = 2 2 = 56 r.
L 3πr

5.58 MATLAB: centroid of a semicircular plate

√
Find the centroid of the semicircular plate whose circular arc is y = f (x) = 32 − x2
(−3 ≤ x ≤ 3).
2 . Moreover, by
Solution First, we notice that the area of the semicircular plate is A = 9π
symmetry, the centroid (x̄, ȳ) must lie on the y-axis, so that x̄ = 0. Then, we calculate ȳ
below.

>> syms x
>> y ( x ) = sqrt (3^2 - x ^2) ;
>> A = 3^2* pi /2;
>> int_fun = 0.5* f ( x ) ^2;
>> cen_y = int ( int_fun , -3 ,3) / A

cen_y =

4/ pi

The result above shows that ȳ = π.

4
326 Applications of Integration

5.59 MATLAB: centroid of a cycloid plate

Find the centroid of the plate between one arch of the cycloid

x = 3(t − sin t), y = 3(1 − cos t),

where 0 ≤ t ≤ 2π, and the x-axis.

Solution The centroid of the plate is calculated in MATLAB using the following code:

>> syms t
>> x ( t ) = 3*( t - sin ( t ) ) ;
>> y ( t ) = 3*(1 - cos ( t ) ) ;
>> dx ( t ) = diff ( x ( t ) ,t ) ;
>> int_fun_A = y ( t ) * dx ( t ) ;
>> A = int ( int_fun_A ,0 ,2* pi )

A =

27* pi

>> int_fun_x = x ( t ) * y ( t ) * dx ( t ) ;
>> cen_x = int ( int_fun_x ,0 ,2* pi ) / A

cen_x =

3* pi

>> int_fun_y = 0.5* y ( t ) ^2* dx ( t ) ;

>> cen_y = int ( int_fun_y ,0 ,2* pi ) / A

cen_y =

5/2

The result above shows that the centroid of the plate is 3π, 52 .

Pappus’s Theorem
5.60 Example: relationship - arc length and surface area of revolution

We have calculated several geometric quantities associated with the cycloid

x = r(t − sin t), y = r(1 − cos t), 0 ≤ t ≤ 2π.

We know that the arc length is L = 8r and the centroid of the arc is (x̄, ȳ) = (πr, 34 r).

1. For the solid obtained by rotating the region under the curve about the x-axis, we know
5.7 Work 327

that the surface area is S = 3 πr .

64 2
It is easy to verify that

S= 64
3 πr
2
= 2π( 43 r) × (8r) = 2π ȳ × L.

2. For the solid obtained by rotating the region under the curve about the y-axis, we know
that the surface area is S = 16π 2 r2 . It is easy to verify that

S = 16π 2 r2 = 2π(πr) × (8r) = 2πx̄ × L.

5.61 Example: relationship - area and volume of revolution

We have calculated several geometric quantities associated with the area under the cycloid

x = r(t − sin t), y = r(1 − cos t), 0 ≤ t ≤ 2π.

We know that the area under the arc is A = 3πr2 and the centroid of region under the arc is
(x̄, ȳ) = (πr, 65 r).

1. For the solid obtained by rotating the region under the curve about the x-axis, we know
that the volume is V = 5π 2 r3 . It is easy to verify that

V = 5π 2 r3 = 2π( 56 r) · (3πr2 ) = 2π ȳ · A.

2. For the solid obtained by rotating the region under the curve about the y-axis, we know
that the volume is V = 6π 3 r3 . It is easy to verify that

V = 6π 3 r3 = 2π(πr) · (3πr2 ) = 2πx̄ · A.

5.62 Theorem: Pappus’s Theorem

1. The surface area S of a surface of revolution generated by rotating a plane curve C

about an axis external to C and on the same plane is equal to the product of the arc
length L of C and the distance d traveled by the centroid of C .

2. The volume V of a solid of revolution generated by rotating a plane region R about an

external axis is equal to the product of the area A of R and the distance d traveled by
the centroid of R.

5.7 Work

5.63 Theorem: work done in moving an object along a line

Suppose that the object moves along the x-axis in the positive direction, from x = a to
x = b, and at each point x between a and b a force f (x) acts on the object, where f is a
328 Applications of Integration

continuous function. Then the total work by the force is

ˆ b
W = f (x) dx.
a

Justification:
Suppose that the object moves along the x-axis in the positive direction, from x = a
to x = b, and at each point x between a and b a force f (x) acts on the object, where f
is a continuous function. We divide the interval [a, b] into n subintervals with endpoints
x0 , x1 , . . . , xn and equal width ∆x. We choose a sample point x∗i in the ith subinterval
[xi−1 , xi ]. Then the force at that point is f (x∗i ). If n is large, then ∆x is small, and since f is
continuous, the values of f don’t change very much over the interval [xi−1 , xi ]. In other words,
f is almost constant on the interval and so the work Wi that is done in moving the particle
from xi−1 to xi is approximately given by Wi ≈ f (x∗i ) × ∆x. Thus we can approximate the
total work by
X n
W ≈ f (x∗i )∆x.
i=1

Therefore, the work done in moving the object from a to b as the limit of this quantity as
n → ∞. Since the right side is a Riemann sum, we recognize its limit as being a definite
integral:
Xn ˆ b
W = lim f (x∗i )∆x = f (x) dx.
n→∞ a
i=1

5.64 Example: spring stretching

A spring exerts a force of 10 N when stretched 10 cm beyond its natural length. How much
work is required to stretch the spring 20 cm beyond its natural length?

Solution According to Hooke’s law, the force needed to extend or compress a spring by
some distance x meters away from its natural length scales linearly with respect to that
distance, that is, f (x) = kx. When the spring is stretched 10 cm beyond its natural length,
the force is 0.1k. By the hypothesis, 10 = 0.1k, so that k = 100. Thus, f (x) = 100x.
Hence the work done in stretching the spring 20 cm beyond its natural length is
ˆ 0.2
0.2
W = 100x dx = 50x2 x=0 = 2 J.

0

5.65 Example: cable lifting

A cable 60 meters long whose mass is 30 kilograms is hanging over the edge of a tall
building and does not touch the ground. How much work is required to lift the top 12 meters
of the cable to the top of the building?

Solution To lift the top 12 meters of the cable to the top of the building, the total work
5.7 Work 329

done consists of two parts: (1) W1 for moving the first top 12 meters of the cable to the top
the building; (2) W2 for moving the bottom (60 − 12) = 48 meters of the cable 12 meters
higher.
(1) Let us place the origin at the top of the building and the x-axis pointing downward.
To compute the work W1 , we divide the cable into small segments with length ∆x. If x∗i is a
point in the ith such segment, then all points in the segment are lifted by approximately the
same amount, namely x∗i . Since the cable weights 6030
= 12 kilograms per meter, so the weight
of the ith segment is 2 ∆x. Thus, the work done on the ith segment is
1

( 12 ∆x) · x∗i · g = 21 gx∗i ∆x,

By adding all these approximations and letting the number of segments becomes large (equiv-
alently ∆x → 0), we get the work done for moving the first top 12 meters of the cable to the
top the building:
n ˆ 12
2 12
X
W1 = lim 1 ∗
∆x = 2 gx dx
1
= = 36g.
1 
n→∞ 2 gxi 4 gx x=0
i=1 0

(2) Similar as above, we divide the second portion of cable into small segments with length
∆x. This gives the weight of the ith segment to be 12 ∆x. Thus, the work done on the ith
segment is
( 21 ∆x) · 12 · g = 6g∆x,
By adding all these approximations and letting the number of segments becomes large (equiv-
alently ∆x → 0), we get the work done for moving the bottom 48 meters of the cable to the
top the building:
n
X ˆ 60
W2 = lim 6g∆x = 6g dx = 6g × (60 − 12) = 288g.
n→∞ 12
i=1

In fact, we can compute W2 without using integration. Since this portion of cable exerts a
force of 21 × 48g along the direction of motion, the work done is W2 = 12 × 48g × 12 = 288g.
Hence the total work done is

W = W1 + W2 = 36g + 288g = 324g = 3175.2 J.

5.66 Example: pumping liquid

A frustum of a cone of height H, top radius R and base radius r contains some liquid of
density ρ. Suppose that the depth of the liquid in the frustum is h. How much work does it
take to pump the liquid to the top of the container?

Solution Let x be the distance to the top of the container. Then the liquid extends from
x = H − h to x = H. We sketch a diagram of the situation.
330 Applications of Integration

R
O R

x∗i
H
ri
ri ∆x h
rH
r
r R−r

By similar triangles, the radius r(x) of the disk at depth x satisfies

r(x) − r H −x
= ,
R−r H
which gives
H −x
r(x) = (R − r) + r.
H
We divide the interval [H − h, H] into subintervals. Then the thin layer corresponding to
[xi−1 , xi ] has approximate volume

∆Vi ≈ π[r(x∗i )]2 ∆x,

and the work needed to pump the layer to the top of the container is approximately

∆Wi ≈ (gρ∆Vi )x∗i = πgρx∗i [r(x∗i )]2 ∆x,

where g is the gravitational constant. Therefore the total work needed to pump all the liquid
to the top of container is approximately
n
X n
X
∆Wi = πgρ x∗i [r(x∗i )]2 ∆x.
i=1 i=1

Taking the limit as ∆x → 0, we get the total work

ˆ H ˆ
πgρ H
W = πgρ x[r(x)]2 dx = 2 x[(R − r)(H − x) + rH]2 dx
H−h H H−h
= πgρH 2 R2 a2 b + 21 a(2a − 3b)b2 + 13 (1 − a)(1 − 3a)b3 − 14 (1 − a)2 b4 ,
 

r h
where a = and b = .
R H
5.7 Work 331

5.67 MATLAB: spring stretching

A spring exerts a force of 10 N when stretched 10 cm beyond its natural length. How much
work is required to stretch the spring 20 cm beyond its natural length?
Solution The problem can be solved in MATLAB as follows:

>> syms k
>> g ( k ) = 0.1* k ; % Hooke ' s law
>> k0 = solve ( g ( k ) == 10) % solve the constant k

k0 =

100

>> syms x
>> int_fun = k0 * x ;
>> W = int ( int_fun ,0 ,0.2)

W =

We see that the work required is 2 (J).

332 Summary

Summary

Theorem: area

• If f (x) ≥ g(x),
y
ˆ b
A= [f (x) − g(x)] dx y = f (x)
a
x=b

• In general, x=a
x
ˆ b
A= |f (x) − g(x)| dx y = g(x)
a

• For a region under the parametric curve x = f (t), y = g(t) (α ≤ t ≤ β),

ˆ β
A= g(t) f 0 (t) dt
α

• For a polar region bounded by the polar

curves r = f (θ), r = g(θ), and by the rays
θ = α to θ = β, R r = f (θ)
θ=β
ˆ β
r = g(θ)
 
2 2
A= 1
2 [f (θ)] − [g(θ)] dθ
α θ=α
θ=0
O

Theorem: volume

• by slices:
ˆ b
V = A(x) dx x=a A(x) x=b
a x

• by shells: y

ˆ b
V = (2πx) [f (x)] (dx) y = f (x)
a | {z } | {z } |{z}
circumference height thickness x
a b
 333

Theorem: arc length

• For y = f (x) (a ≤ x ≤ b),

ˆ bq
L=
2
1 + [f 0 (x)] dx. y
a

• For parametric curve x = f (t), y = g(t)

(α ≤ t ≤ β),
ˆ βq
2 2
L= [f 0 (t)] + [g 0 (t)] dt.
α
• In general, x
ˆ
L= (ds) .
|{z}
arc length

• For the polar curve r = f (θ) from θ = α

to θ = β,
ˆ βq r = f (θ)
2 2
L= [f (θ)] + [f 0 (θ)] dθ
α θ=β
ˆ β
= r2 + (dr/dθ)2 dθ
p
θ=α
α θ=0
O

Theorem: surface area of revolution

• For a curve y = f (x) (a ≤ x ≤ b),

ˆ b q
2 y
S= 2π f (x) 1 + [f 0 (x)] dx;
a

• For a parametric curve x = f (t), y = g(t)

(α ≤ t ≤ β),
ˆ β q x
2 2
S= 2πg(t) [f 0 (t)] + [g 0 (t)] dt.
α

• In general,
ˆ
(2πr) (ds) .
| {z } |{z}
circumference arc length
334 Summary

Theorem: center of mass for discrete system

• For one-dimensional system:

n
P
mi xi
M i=1
x̄ = = Pn .
m
mi
i=1

• For two-dimensional system:

n
P n
P
m i xi mi yi
My Mx
x̄ = = i=1
n , ȳ = i=1
= Pn .
m P m
mi mi
i=1 i=1

Theorem: center of mass

• For mass distributed along a curve y = f (x) (a ≤ x ≤ b),

ˆ
1 b
x̄ = 1 + [f 0 (x)]2 dx,
p
ρ(x) · x
m a
ˆ
1 b
ȳ = ρ(x) · f (x) 1 + [f 0 (x)]2 dx,
p
m a
ˆ b
where m = 1 + [f 0 (x)]2 dx is the total mass of the curve.
p
ρ(x) ·
a
• For mass distributed along a parametric curve x = f (t), y = g(t) (α ≤ t ≤ β),
ˆ
1 β
x̄ = ρ(t) · f (t) [f 0 (t)]2 + [g 0 (t)]2 dt,
p
m α
ˆ
1 β
ȳ = ρ(t) · g(t) [f 0 (t)]2 + [g 0 (t)]2 dt,
p
m α
ˆ β
where m = [f 0 (t)]2 + [g 0 (t)]2 dt is the total mass of the curve.
p
ρ(t) ·
α
• For a plate occupying the region bounded by two curves y = f (x) and y = g(x) (a ≤
x ≤ b), with f (x) ≥ g(x), the centroid is at (x̄, ȳ):
ˆ ˆ
1 b
1 b n
2 2
o
x̄ = x [f (x) − g(x)] dx, ȳ = 1
2 [f (x)] − [g(x)] dx,
A a A a
ˆ b
where A = [f (x) − g(x)] dx is the area of the region.
a
 335

Theorem: Pappus’s Theorem

1. The surface area S of a surface of revolution generated by rotating a plane curve C

about an axis external to C and on the same plane is equal to the product of the arc
length L of C and the distance d traveled by the centroid of C .

2. The volume V of a solid of revolution generated by rotating a plane region R about an

external axis is equal to the product of the area A of R and the distance d traveled by
the centroid of R.

Theorem: work done by a force

ˆ b
W = f (x) dx.
a
336 Summary
 CHAPTER

6
Infinite Sequences and Series

6.1 Infinite Sequences

6.1 Definition: infinite sequence

An infinite sequence (or sequence for short) is an ordered list of numbers of the form

{a1 , a2 , a3 , . . . , an , . . . }.

For each n, the term an is called the nth term of the sequence.
For convenience, sometimes we write a sequence as {an }.

6.2 Example: infinite sequences

The following are some examples of sequences.

1. The sequence {1, 2, 3, . . . , n, . . . } consists of all the positive integers, with the nth term
of the sequence being an = n.
2. The nth term of the sequence {1, −1, 1, −1, 1, −1, . . . } is an = (−1)n−1 .
3. The terms in the sequence

{3, 3.1, 3.14, 3.141, 3.1415, . . . }

are from the truncations of the number π. The nth term of the sequence is the num-
ber obtained by truncating π to n − 1 decimal place. For this sequence, there is no
elementary formula available to express an in terms of n.
4. For a given expression of a function f , by evaluating the function at integers, we can
obtain a sequence
{a1 , a2 , a3 , . . . , an , . . . } ,
338 Infinite Sequences and Series

where an = f (n) (n = 1, 2, 3, . . . ). For instance, the infinite sequence

cos π1 , cos π2 , cos π3 , . . . , cos nπ , . . .



can be generated by the function f (x) = cos πx .

5. The infinite sequence
( r )
√ p √ √ √
q p q p
2, 2 + 2, 2 + 2 + 2, 2 + 2 + 2 + 2, . . .

can be defined as
√
a1 = 2; an = 2 + an−1 , for n ≥ 2.
p

By using the formula, we can generate the whole sequence recursively. In fact, for the
first few terms:
√
when n = 1 : a1 = 2;
√
when n = 2 : a2 = 2 + a1
p √
= 2 + 2;
√
when n = 3 : a3 = 2 + a2
√
q p
= 2 + 2 + 2;
√
when n = 4 : a4 = 2 + a3
r
√
q p
= 2 + 2 + 2 + 2.

6.3 Definition: limit of infinite sequence

• An infinite sequence {an } has the limit L, an

which is a finite number, and we write

lim an = L or an → L as n → ∞
n→∞ L

if we can make the terms an as close to L

as we like by taking n sufficiently large.
n
O 1 2 3 4 5 6 7 8 9 10

If lim an exists, we say the sequence converges (or is convergent). Otherwise, we

n→∞
say the sequence diverges (or is divergent).
• The infinite limit lim an = ∞ means that we can make the terms an as large as we like
n→∞
by taking n sufficiently large. In this case, we say the sequence diverges to infinity.
Similarly, we can definite lim an = −∞.
n→∞
6.1 Infinite Sequences 339

6.4 Example: convergent/divergent sequences

1. The sequence 1 1 1 1

1, 2, 3, . . . , n, . . .
1
converges to 0, since we can make the terms an = as close to 0 as we like by taking
n
n sufficiently large. In fact, for instance, to make the terms an close to 0 within 0.1,
that is,
|an − 0| < 0.1,
we just need to take n > 10, as shown in the figure.

n an n an
1 1.0 11 0.090909
2 0.5 12 0.083333 an
3 0.333333 13 0.076923
4 0.25 14 0.071429
5 0.2 15 0.066667
6 0.166667 20 0.05 0.5

7 0.142857 30 0.033333
8 0.125 40 0.025 0 + 0.1
n
9 0.111111 50 0.02 0 − 0.1 1 2 3 4 5 10

10 0.1 60 0.016667

2. The sequence
2, (2) , (2) , . . . , (2) , . . .
1 1 2 1 4 1 n

converges to 0, since we can make the terms an = ( 21 )n as close to 0 as we like by taking

n sufficiently large. In fact, for instance, to make the terms an close to 0 within 0.05,
that is,
|an − 0| < 0.05,
we just need to take n > 4, as shown in the figure.

n an
1 0.5
2 0.25 an
3 0.125
4 0.0625 0.5
5 0.03125
6 0.015625
7 0.0078125 0 + 0.05
8 0.00390625 0 − 0.05 1 2 3 4 5 10
n
9 0.00195313
10 0.000976563
340 Infinite Sequences and Series

3. The sequence
{0.9, 0.99, 0.999, 0.9999, . . . }
converges to 1, since the terms increase in magnitude and approach 1. For instance, to
make the terms an close to 0 within 0.07, that is,

|an − 0| < 0.07,

we just need to take n > 1.

an
n an
1 + 0.07
1
1 0.9 1 − 0.07

2 0.99
3 0.999 0.5

4 0.9999
5 0.99999
n
6 0.999999 1 2 3 10

4. The sequence
1 8 3 n2 + 2 · (−1)n n
 
− , , ,..., ,...
3 9 19 2n2 + 1
n2 + 2 · (−1)n
converges to 2,
1
since we can make the terms an = as close to 12 as we
2n2 + 1
like by taking n sufficiently large. In fact, for instance, to make the terms an close to
2 within 0.15, that is,
1
an − 1  < 0.15,
 
2

we just need to take n > 6, as shown in the figure.

n an n an an
1 −0.333333 11 0.407407
2 0.888889 12 0.581315
3 0.157895 13 0.421829 1
+ 0.15
4 0.727273 14 0.569975 2
1
2

5 0.294118 15 0.432373 1
2 − 0.15

6 0.657534 20 0.549313
7 0.353535 30 0.533037 1 2 3 4 5 6 7 10 20
n
8 0.620155 40 0.524836
9 0.386503 60 0.516595
10 0.597015 80 0.512460
6.1 Infinite Sequences 341

5. The sequence an

{sin 1, sin 2, sin 3, . . . , sin n, . . . } 1

diverges, since an = sin n does not appear

to approach any finite number as n → ∞. n
12 5 10 20 30 40 50

−1

6. The sequence
1 2 3
 
n an
, , ,..., ,...
ln 2 ln 3 ln 4 ln(n + 1)
diverges to ∞. This is because,
x H 1 5
lim = lim
x→∞ ln(x + 1) x→∞ 1
x+1 2
= lim (x + 1) = ∞, 1
x→∞ n
1 2 3 4 5 10 20
n
we have lim an = lim = ∞.
n→∞ n→∞ ln(n + 1)

6.5 Theorem: limit of a sequence from limit of a function

If lim f (x) = L and f (n) = an when n is an integer, then lim an = L. The same result
x→∞ n→∞
holds when L = ∞ or −∞.

6.6 Theorem: Limit Laws for sequences

Suppose {an } and {bn } are convergent sequences and c is a constant. Then the following
hold.

1. lim (an + bn ) = lim an + lim bn .

n→∞ n→∞ n→∞

2. lim (an − bn ) = lim an − lim bn .

n→∞ n→∞ n→∞

3. lim can = c lim an . In particular, lim c = c.

n→∞ n→∞ n→∞

4. lim (an bn ) = lim an · lim bn .

n→∞ n→∞ n→∞

an lim an
5. lim = n→∞ if lim bn 6= 0.
n→∞ bn lim bn n→∞
n→∞
342 Infinite Sequences and Series

h ip
6. lim apn = lim an if p > 0 and an > 0.
n→∞ n→∞

6.7 Example: limits of sequences

1
1. Show that lim = 0 (p > 0).
np
n→∞
1 1
Solution Because lim = 0, we know that lim = 0. Thus, for p > 0, by the
x→∞ x n→∞ n
Limit Laws, p
1 1

lim p = lim = 0p = 0.
n→∞ n n→∞ n

3
n 2 + 2n − 1
2. Compute lim 3 .
n→∞ n2 + 1
Solution Since
1 1 1 1
 
3
3 n2 1+2· 1 − 3 1+2· 1 − 3
n + 2n − 1
2
n 2 n 2 n n
= = 2 2
,
3
n2 + 1 1 1
 
n
3
2 1+ 3 1 + 3
n2 n2

from the Limit Laws, we have

1 1
3 1 + 2 · lim 1 − lim
n 2 + 2n − 1 n→∞ n 2
3
n→∞ n 2 1+2·0−0
lim = = = 1.
n→∞
3
n2 + 1 1 1+0
1 + lim 3
n→∞ n 2

6.8 Theorem: Squeeze Theorem

If an ≤ bn ≤ cn for n ≥ n0 and lim an = lim cn = L, then lim bn = L.

n→∞ n→∞ n→∞

6.9 Example: limits of sequence - Squeeze Theorem

cos n
Show that lim √ = 0.
n→∞ n
Solution Since −1 ≤ cos n ≤ 1 for all n, we have
1 cos n 1
−√ ≤ √ ≤ √ .
n n n

1 1 cos n
 
Because lim √ = lim − √ = 0, by the Squeeze Theorem, we get that lim √ = 0.
n→∞ n n→∞ n n→∞ n
6.1 Infinite Sequences 343

6.10 Example: zero limit and its absolute value

Show that lim |an | = 0 if and only if lim an = 0.

n→∞ n→∞
Solution
“=⇒”: Suppose lim |an | = 0. By the Limit Laws, lim (−|an |) = 0. Since
n→∞ n→∞

−|an | ≤ an ≤ |an |,

by the Squeeze Theorem, lim an = 0.

n→∞
“⇐=”: Suppose lim an = 0. By the Limit Laws,
n→∞
q
lim
a2n + n−1 = lim (a2n + n−1 )
p
n→∞ n→∞
q
= lim a2n + lim n−1
n→∞ n→∞
q p
= lim a2n + 0 = 02 + 0 = 0.
n→∞

Since
0 ≤ |an | ≤ a2n + n−1 ,
p

by the Squeeze Theorem, lim |an | = 0.

n→∞

6.11 Definition: two asymptotic relations

• The notation
an  bn , as n → ∞,
which is read “an is much smaller than bn as n tends to ∞,” means
an
lim = 0.
n→∞ bn

• The notation
an ∼ bn , as n → ∞,
which is read “an is asymptotic to bn as n tends to ∞,” means that

an − bn  bn , as n → ∞,

or, equivalently,
an
lim = 1.
n→∞ bn

6.12 Theorem: growth rates of sequences

For any real numbers p > 0, q > 0, r > 0, and b > 1, the following asymptotic relations
hold: as n → ∞,
lnq n  np  np lnr n  bn  n!  nn .
344 Infinite Sequences and Series

Remark: Stirling’s formula

A more precise formula for n! is given by Stirling’s formula:

√  n n
n! ∼ 2πn , as n → ∞.
e

Justification: lnq n  np

Since, as n → ∞,
1/q
lnq n lnq n ln n

ln n  n
q p
⇐⇒  1 ⇐⇒  1 ⇐⇒  1 ⇐⇒ ln n  np/q ,
np np np/q

we only need to show for any positive real number p,

ln n  np , as n → ∞.

In fact, by l’Hospital’s Rule,

1
ln x H x = lim 1 = 0,
lim = lim
x→∞ xp x→∞ pxp−1 x→∞ pxp

ln n
so that lim = 0. Thus, ln n  np as n → ∞.
n→∞ np

Justification: np  np lnr n

Since, as n → ∞,

np  np lnr n ⇐⇒ 1  lnr n ⇐⇒ 1  ln n,
1 1
we only need to show that lim = 0. The latter is obvious since lim = 0.
n→∞ ln n x→∞ ln x

Justification: np lnr n  bn

Since lnq n  np as n → ∞ for any positive p and q, to show that np lnr n  bn , we only
need to show that the asymptotic relation np  bn for any positive p and b > 1. Furthermore,
since
np  bn ⇐⇒ n  (b1/p )n ,
it is sufficient to show that n  bn for any b > 1.
In fact, we may write b = 1 + a, with a > 0. Then, by the binomial expansion, we have

bn = (1 + a)n = 1 + na + 21 n(n − 1)a2 + · · · + an > 21 n(n − 1)a2 .

We get
n 2 1
0< < 2· .
bn a n−1
6.1 Infinite Sequences 345

1 n
Because lim = 0, by the Squeeze Theorem, we have lim n = 0. That is, n  bn .
n→∞ n−1 n→∞ b

Justification: bn  n!

For the given number b, take a sufficiently large K such that b < K. Then, for n > K, we
have
n factors
z }| {
n
b b · b···b · b
0< =
n! n · (n − 1) · · · 2 · 1
b b b b b b
= · ··· · ··· ·
n n−1 K K −1 2 1
b b b b
< · ··· · .
n K −1 2 1
Since
1
   
b b b b b b b
lim · ··· · = b· ··· · · lim = 0,
n→∞ n K −1 2 1 K −1 2 1 n→∞ n
bn
by the Squeeze Theorem, we have lim = 0. That is, bn  n! as n → ∞.
n→∞ n!

Justification: n!  nn

Since
n! n · (n − 1) · · · 2 · 1 n n−1 2 1 1
0< = = · ··· · < .
nn n
| · n ·
{z· · n · n
} n n n n n
n factors

1 n!
It is known that lim = 0. By the Squeeze Theorem, we have lim n = 0. That is,
n→∞ n n→∞ n
n!  nn as n → ∞.

6.13 Theorem: geometric sequences

Let r be a real number. Then

 0, if |r| < 1;


lim rn = 1, if r = 1;
n→∞
diverges, if r ≤ −1 or r > 1.


Justification:

1. Consider 0 ≤ |r| < 1.

When r = 0, it is clear that lim rn = 0.
n→∞
346 Infinite Sequences and Series

1
When 0 < r < 1, we put b = and know that b > 1. Since 1  bn as n → ∞, we have
r
1
lim = 0 =⇒ lim rn = 0.
n→∞ bn n→∞

When −1 < r < 0, we have 0 < |r| < 1. Since

−|r|n ≤ rn ≤ |r|n ,

and lim |r|n = lim (−|r|n ) = 0, by the Squeeze Theorem, we know that lim rn = 0.
n→∞ n→∞ n→∞

2. If r = 1, then r ≡ 1 for all n ≥ 1. In this case, it is clear that lim r = 1.

n n
n→∞

3. Suppose r ≤ −1 or r > 1.
When r > 1, we know that 1  rn , so that {rn } diverges to ∞.
When r ≤ −1, {|r|n } diverges to ∞, so that {rn } oscillates and never approach any
finite number. Hence, {rn } diverges in this case.

6.14 Theorem: sequential limit for continuous function

If lim an = L and the function f is continuous at L, then

n→∞

lim f (an ) = f (L).

n→∞

6.15 Example: limit of sequence

√
1. Compute the limit lim n
a for a > 0.
n→∞
√ ln a √
Solution Let an = ln n a = . Then n a = ean . We know that 1  n as n → ∞,
n
so that lim an = 0. Because ex is a continuous function at x = 0, we have
n→∞
√
lim n
a = lim ean = e0 = 1.
n→∞ n→∞

√
2. Compute the limit lim n
n.
n→∞
√ ln n √
Solution Let an = ln n n = . Then n n = ean . We know that ln n  n as n → ∞,
n
so that lim an = 0. Because ex is a continuous function at x = 0, we have
n→∞
√
lim n
n = lim ean = e0 = 1.
n→∞ n→∞

n
2

3. Compute the limit lim 1 − .
n→∞ n
6.1 Infinite Sequences 347

n n
2 2 2
   
Solution Let an = ln 1 − = n · ln 1 − . Then 1 − = ean . Replacing
n n n
n by a continuous variable x−1 , we see that n → ∞ is a special case when x → 0+ .
Applying l’Hospital’s Rule gives
−2
ln (1 − 2x)
= lim+ 1 − 2x = −2,
H
lim x−1 · ln (1 − 2x) = lim+
x→0+ x→0 x x→0 1
so we have lim an = −2. Because ex is a continuous function at x = −2, we have
n→∞
n
2

lim 1− = lim ean = e−2 .
n→∞ n n→∞

6.16 Definition: terminology for sequences

• {an } is increasing if an+1 > an . For example, {1, 2, 3, . . . }.

• {an } is non-decreasing if an+1 ≥ an . For example, {1, 1, 2, 2, 3, 3, . . . }.

• {an } is decreasing if an+1 < an . For example, {0, −1, −2, −3, . . . }.
• {an } is non-increasing if an+1 ≤ an . For example, {0, 0, −1, −1, −2, −2, . . . }.
• {an } is monotonic if it is either non-increasing or non-decreasing.
• {an } is bounded if there is number M such that |an | ≤ M for all n.

6.17 Example: product of sequences

Suppose {an } is bounded and lim bn = 0. Show that lim an bn = 0.

n→∞ n→∞
Solution Since {an } is bounded, there is a number M such that |an | ≤ M for all n. Thus,
for every n,
−M |bn | ≤ an bn ≤ M |bn |.
Since lim bn = 0 implies lim |bn | = 0, by the Limit Laws, we know that lim M |bn | =
n→∞ n→∞ n→∞
lim (−M |bn |) = 0. Hence, by the Squeeze Theorem, we get lim an bn = 0.
n→∞ n→∞

6.18 Theorem: Monotonic Sequence Theorem

Every bounded, monotonic sequence is convergent.

348 Infinite Sequences and Series

Remark:
The proof of the theorem is based on a fundamental property of the real numbers
that if a sequence (or a nonempty set) is bounded above, then it has a least upper
bound. It is an expression of the fact that there is no gap or hole in the real number
line. We omit the details of the proof because it is beyond the scope of the text.

6.19 Example: sequence generated by recurrence relation

Show the sequence {an }, defined by the recurrence relation,

an = 12 an−1 + 1, (n ≥ 2); a1 = 1,

has a limit. What is the limit?

Analysis: By plotting the first few terms an

of the sequence, we see that the sequence ap-
2
pears increasing and bounded.
1

n
O 1 2 3 4 5 6 7 8 9 10

Solution We shall prove the following statement by induction: for n ≥ 1,


an+1 > an ,
Sn :
1 ≤ an < 2.

In fact, for n = 1, we have

a2 = 21 a1 + 1 = 1
2 ·1+1= 3
2 > 1 = a1 ,
1 = a1 < 2.

So S1 is true. Suppose the statement is true for some k:


ak+1 > ak ,
Sk :
1 ≤ ak < 2.

Then,
ak+2 = 21 ak+1 + 1 > 12 ak + 1 = ak+1 ,
and
1 ≤ ak < ak+1 = 21 ak + 1 < 1
2 · 2 + 1 = 2,
that is 
ak+2 > ak+1 ,
Sk+1 :
1 ≤ ak+1 < 2,
Sk+1 is also true. Hence, by the principle of induction, the statement Sn is true for all n ≥ 1.
6.1 Infinite Sequences 349

Because {an } is increasing and bounded, by the Monotonic Sequence Theorem, lim an
n→∞
exists, say L.
To find the limit L, we let n tend to ∞ on the both sides of the recurrence relation
an = 21 an−1 + 1. This yields L = 12 L + 1. Solving the equation gives L = 2.

6.20 MATLAB: limit of sequence

We will compute the limit of the following sequences in MATLAB.

√
1. lim n a for a > 0.
n→∞
Solution

>> syms a n
>> assume (a , ' positive ') % a >0
>> assume (n , ' positive ') % n >0
>> assumeAlso (n , ' integer ') % n is also an integer
>> limit ( a ^(1/ n ) ,n , Inf )

ans =

√
2. lim n
n.
n→∞
Solution

>> syms n
>> assume (n , ' positive ') % n >0
>> assumeAlso (n , ' integer ') % n is also an integer
>> limit ( n ^(1/ n ) ,n , Inf )

ans =

n
2

3. lim 1− .
n→∞ n
Solution

>> syms n
>> assume (n , ' positive ') % n >0
>> assumeAlso (n , ' integer ') % n is also an integer
>> limit ((1 - 2/ n ) ^n ,n , Inf )
350 Infinite Sequences and Series

ans =

exp ( -2)

6.2 Series
Convergence of Series
6.21 Definition: convergent series
∞
For a given series ak = a1 + a2 + a3 + · · · , let sn denote its nth partial sum:
P
k=1

n
X
sn = ak = a1 + a2 + · · · + an .
k=1

If the sequence {sn } is convergent and lim sn = s exists as a real number, then the series
n→∞
ak is called convergent and we write
P

∞
X
a1 + a2 + a3 + · · · = s or ak = s.
k=1

The number s is called the sum of the series. If the sequence {sn } is divergent, then the
series is called divergent.

6.22 Theorem: geometric series

Let a 6= 0 and r be a real number. Then

 a
∞  , if |r| < 1;
1−r
X
ark−1 =
diverges, if |r| ≥ 1.

k=1

Definition: geometric series

The series
∞
X
ark−1 = a + ar + ar2 + · · ·
k=1

is called a geometric series. Each term is obtained from the preceding one by
multiplying it by the common ratio r.
6.2 Series 351

Justification:
If r = 1, the nth partial sum sn = a + a + · · · + a = na diverges either to ∞ or −∞, as n
tends to ∞.
If r 6= 1, the nth partial sum sn can be found as follows: by multiplying r to both side of

sn = a + ar + ar2 + · · · + arn−1 ,

we get

rsn = ar + ar2 + · · · + arn−1 + arn .

Subtracting these equations, we get

sn − rsn = a − arn ,

so that
a − arn
sn = .
1−r
Thus, if −1 < r < 1, since rn → 0 as n → ∞, we get

a − arn a − a · lim rn a
n→∞
lim sn = lim = = .
n→∞ n→∞ 1 − r 1−r 1−r
If r ≤ −1 or r > 1, the sequence {rn } is divergent, so {sn } is also divergent. Thus, if
r ≤ −1 or r > 1, the geometric series diverges.

6.23 Example: repeating decimal

Express 3.0127 as a fraction.

Solution The given number can be written as a geometric series:

3.0127 = 3 + 0.0127 = 3 + 0.0127 + 0.0000127 + 0.0000000127 + · · ·

= 3 + 0.0127 + 0.0127 · 10−3 + 0.0127 · 10−6 + · · ·
| {z }
geometric series
0.0127
=3+
1 − 10−3
127 × 10−4
= 3 + −3
10 × (103 − 1)
127 30097
=3+ = .
9990 9990

Remark:
It is easy to generalize the above argument to show that any repeating decimal can
be expressed as a fraction. Hence, any repeating decimal must be a rational number.
352 Infinite Sequences and Series

6.24 Example: telescoping series

∞ 1 1
 
1. Evaluate the series − k+1 .
P
k=1 2k 2
Solution The partial sum for the given series is
n 
1 1
X 
sn = −
2k 2k+1
k=1
1 1
A1 − 1
+ A1 − 1
+ · · · + A1 − 1
       
= − +
21
22 2A2
23 2A3
24 2An 2n+1
1 1
= 1 − n+1 .
2 2
Thus,
1 1 1
 
lim sn = lim − n+1 = .
n→∞ n→∞ 21 2 2
Hence,
∞ 
1 1 1
X 
− = .
2k 2 k+1 2
k=1

This is called telescoping series since the partial sum collapses (like a pirate’s collapsing
telescope) into just two terms.
∞ 1
2. Evaluate the series .
P
k=1 k(k + 1)
Solution The partial sum for the given series can be evaluated by writing it as a
telescoping sum:
n n 
1 1 1
X X 
sn = = −
k(k + 1) k k+1
k=1 k=1
1 1 1 1 1 1 1 1
       
= − + C − + C − + ··· + C −
1 2 2C 3 3C 4 nC n+1
1 1
= − .
1 n+1
Thus,
1 1
 
lim sn = lim − = 1.
n→∞ n→∞ 1 n+1
Hence,
∞
X 1
= 1.
k(k + 1)
k=1
6.2 Series 353

6.25 Theorem: properties of convergent series

1. If ak converges and c is a real number, then the series cak converges and
P P

X X
cak = c ak .

2. If ak and bk converge, then the series (ak ± bk ) converges and

P P P

X X X
(ak ± bk ) = ak ± bk .

∞ ∞
3. If N is a positive integer, then ak and ak either both converge or both diverge.
P P
k=1 k=N
In general, changing a finite number of terms in a convergent series does not change its
convergence, although it does change the value of the series.

Justification:

n ∞
1. Let sn = ak and lim sn = s. Then the nth partial sum for the series cak is
P P
k=1 n→∞ k=1

n
X n
X
un = cak = c ak = csn ,
k=1 k=1

and

lim un = lim csn = c · lim sn = c · s.

n→∞ n→∞ n→∞

That is, cak is convergent and

P

∞
X ∞
X
cak = c ak .
k=1 k=1

2. Let
n
X n
X
sn = ak , tn = bk ,
k=1 k=1
∞
and lim sn = s, lim tn = t. Then the nth partial sum for the series (ak + bk ) is
P
n→∞ n→∞ k=1

n
X n
X n
X
un = (ak + bk ) = ak + bk = sn + tn ,
k=1 k=1 k=1

and

lim un = lim (sn + tn ) = lim sn + lim tn = s + t.

n→∞ n→∞ n→∞ n→∞
354 Infinite Sequences and Series

That is, (ak + bk ) is convergent and

P

∞
X ∞
X ∞
X
(ak + bk ) = ak + bk .
k=1 k=1 k=1

Similarly we can show that (ak − bk ) is convergent and

P

∞
X ∞
X ∞
X
(ak − bk ) = ak − bk .
k=1 k=1 k=1

3. For fixed integer N with 1 < N < n,

n
X n
X N
X −1
ak = ak − ak .
k=N k=1 k=1

∞ ∞
If ak converges, then the right side has a finite value as n tends to ∞, so
P P
ak
k=1 k=N
converges. Similarly, it is easy to see that, from the equality
n
X N
X −1 n
X
ak = ak + ak ,
k=1 k=1 k=N

∞ ∞
if ak converges, then ak converges.
P P
k=N k=1
If bk is generated by changing a finite number of terms in a convergent series ak ,
P P
then there exists a sufficiently large N , such that, when n > N ,
n
X n
X
bk = ak .
k=N k=N

Thus, when n > N , the equality holds:

n
X n
X
bk = ak + K,
k=1 k=1

∞
where K is a fixed constant. It is easy to see from the last equality that bk converges,
P
k=1
∞
while its value may differ from the value of ak .
P
k=1

6.26 Remark: finding the sums

Besides the geometric series and telescoping series, generally, it is not easy to find the sum
of a series. However, it is possible to study convergence of a series without knowing its sum.
6.2 Series 355

6.27 MATLAB: telescoping series

In MATLAB, we can use the command symsum to evaluate the sum of a series. In the
following, we will show two examples.
∞ 1 1
 
1. .
P
−
k=1 2 2k+1
k

Solution

>> syms k
>> f = 1/2^ k - 1/2^( k + 1) ;
>> symsum (f ,k ,1 , Inf )

ans =

1/2

∞ 1
2. .
P
k=1 k(k + 1)
Solution

>> syms k
>> f = 1/( k *( k + 1) ) ;
>> symsum (f ,k ,1 , Inf )

ans =

6.28 Engineering Example: discrete-time convolution

The convolution is an operation on two functions to produce a third function that shows
how the shape of one of the original functions is modified by the other one. For two functions
f [n] and g[n] defined on the set Z of integers, the discrete-time convolution of f [n] and g[n]
is given by
∞
X
f [n] ? g[n] = f [m]g[n − m].
m=−∞

The discrete-time convolution has wide applications in engineering fields, e.g., computer
vision, natural language processing and digital signal processing (DSP) systems.
For example, the block diagram for a linear time-invariant (LTI) system is shown below,
where x[n] is the input and y[n] is the output.
356 Infinite Sequences and Series

x[n] LTI system y[n]

The output of the system y[n] is the convolution of the input x[n] with the system’s
response to an impulse, i.e., the impulse response. Mathematically, it can be written as

y[n] = x[n] ? h[n]

X∞
= x[m]h[n − m]
m=−∞

where h[n] is the impulse response of the LTI system.

6.29 Engineering Example: discrete-time correlation

Correlation is a way to compare signals with each other and search for similarities between
signals. The signal correlation can be performed either between two different signals or with
one signal.
Specifically, the correlation between two different signals is called cross-correlation.
Given two signals x[n] and y[n], the cross-correlation between x[n] and y[n] is defined as
∞
X
Rxy [k] = x[m]y[m − k].
m=−∞

There are many applications of signal cross-correlation in signal processing systems. For
example, when a signal is corrupted by another undesirable signal, e.g., noise, during trans-
mission, the cross-correlation is then used to estimate the signal at the receiver side.
The correlation with one signal is called auto-correlation. For the signal x[n], its auto-
correlation is defined as
X∞
Rxx [k] = x[m]x[m − k].
m=−∞

The auto-correlation is widely applied in radar, sonar, satellite, wireless communication sys-
tems, etc.

6.30 Engineering Example: geometric distribution

Assume a programmer has 85% chance of finding a bug every time he compiles his code.
Then, the probability that he finds the bug at the kth (k ≥ 1) trial can be calculated as

p(k) = (1 − 0.85)k−1 0.85.

Note that the probability that the programmer finds the bug at the kth trial, i.e., p(k), obeys
the geometric distribution.
Suppose that the total number of chances that the programmer can try to find the bug
every workday is n. Then, the probability that the programmer can find the bug before he
6.2 Series 357

gets off work is

n
X n
X
sn = p(k) = (1 − 0.85)k−1 0.85.
k=1 k=1

In particular, when n → ∞, it can be proved that

∞
X
s= (1 − 0.85)k−1 0.85 = 1.
k=1

∞
X
The result above shows that the series (1 − 0.85)k−1 0.85 converges to 1. Note that this
k=1
∞
X
result holds for any bug finding probability q ∈ (0, 1], i.e., (1 − q)k−1 q = 1 for any
k=1
0 < q ≤ 1. It means if given enough times of trial, no matter how proficient the programmer
is, he can find the bug for sure.

6.31 Engineering Example: Poisson distribution

The Poisson distribution is a discrete probability distribution that describes the likelihood
of a given number of events occurring within a fixed interval of time (or space), if these events
occur 1) with a knowing constant average rate, and 2) independently of the time since the
last event occurred. The Poisson distribution can be used to model quite a large number of
processes, e.g., the number of births expected during the night in a hospital, the number of
arrivals in any service facility such as at an ATM, bank, bus station and petrol pump, and
the number of clicks on any website.
Let the probability of k arrivals at a bus station during a fixed time interval be denoted as
p(k). Since p(k) obeys the Poisson distribution, we have

λk
p(k) = e−λ
k!
where λ is the average number of arrivals at the bus station during the fixed time interval.
Furthermore, it can be proved that
∞ ∞
X X λk
s= p(k) = e−λ = 1.
k!
k=0 k=0

The convergent series above means there is either no arrival or at least one arrival at the bus
station during the fixed time interval, which is consistent with the intuition.
358 Infinite Sequences and Series

Divergence Test

6.32 Theorem: Divergence Test

If ak converges, then lim ak = 0. Hence, if lim ak 6= 0, then the series diverges.

P
k→∞ k→∞

Justification:
n
If ak is convergent, by the definition, lim sn = s exists, where sn = ak . Thus,
P P
n→∞ k=1

lim ak = lim (sk − sk−1 ) = lim sk − lim sk−1 = s − s = 0.

k→∞ k→∞ k→∞ k→∞

6.33 Example: divergent series

∞ (−1)k k 2
Show that the series diverges.
P
k=1 k + 1
2

Solution Since, as k → ∞,
 (−1)k k 2  k2 k2 1
 
 k2 + 1  = k2 + 1 = =
1
→ 1 6= 0,
 
1

k 1+ 2
2 1 +
k k2

(−1)k k 2
we know that lim 6= 0. Hence, by the Divergence Test, the given series is divergent.
k→∞ k 2 + 1

6.34 Remark: Divergence Test

1. If lim ak 6= 0, then either the limit does not exist, or the limit exists but does not
k→∞
equal to 0.
2. If lim ak = 0, the Divergence Test is inconclusive. In other words, the zero limit of
k→∞
the sequence {an } is not sufficient for the convergence of the series ak . An example
P
P1
of such kind of series is the so-called harmonic series .
k

6.35 Example: harmonic series

∞ 1
Show that the harmonic series diverges.
P
k=1 k
Solution We can show that the partial sums s2 , s4 , s8 , s16 , s32 , . . . become unbounded, so
6.3 Integral Test 359

the sequence {sn } does not have a finite limit. For the first few terms, we see that

s2 = 1 + 12 ,
s4 = 1 + 1
+ 1
+ 1
>1+ 1
+ 1
+ 1
= 1 + 22 ,



2 3 4 2 4 4
s8 = 1 + 1
+ 1
+ 1
+ 1
+ 1
+ 1
+ 1



2 3 4 5 6 7 8
>1+ 1
+ 1
+ 1
+ 1
+ 1
+ 1
+ 1
= 1 + 32 ,



2 4 4 8 8 8 8
s16 = 1 + 1
+ 1
+ 1
+ 1
+ 1
+ 1
+ 1
+ 9 + 10 + · · · + 16
1 1 1




2 3 4 5 6 7 8
>1+ 1
+ 1
+ 1
+ 1
+ 1
+ 1
+ 1
+ 16 + 16 + · · · + 16
1 1 1
= 1 + 24 .




2 4 4 8 8 8 8

In fact, we see generally that

n
. s2n > 1 +
2
Thus, {sn } does not have a finite limit. That is, the harmonic series diverges.

6.36 MATLAB: divergent series

∞ (−1)k k 2
Show the convergence/divergence of the series in MATLAB.
P
k=1 k + 1
2

Solution We will use the Divergence Test to verify the convergence/divergence of the
series, as shown in the following:

>> syms k
>> f = ( -1) ^ k * k ^2/( k ^2 + 1) ;
>> limit ( abs ( f ) ,k , Inf )

ans =

Note that in MATLAB, the command  abs(x)  can calculate the absolute value of x. From
 (−1)k k 2  (−1) k 2
k
the result above, we see that limk→∞  2  = 1 6= 0, which means limk→∞ 6= 0.
k +1  k2 + 1
∞ (−1) kk 2
Hence, according to the Divergence Test, we know that the series is divergent.
P
k=1 k + 1
2

6.3 Integral Test

6.37 Theorem: Integral Test

Suppose f is a continuous, positive, decreasing function on [1, ∞)

ˆ and let ak = f (k). Then
∞ ∞
the series ak is convergent if and only if the improper integral f (x) dx is convergent.
P
k=1 1
360 Infinite Sequences and Series

In other words:
ˆ ∞ ∞
1. If f (x) dx is convergent, then ak is convergent;
P
1 k=1
ˆ ∞ ∞
2. If f (x) dx is divergent, then ak is divergent.
P
1 k=1

y = f (x)

Justification:
ˆ ∞ ∞
1. Suppose f (x) dx converges. To show that f (k) converges, since f ≥ 0, we only
P
1 k=1
n
need to show that f (k) is bounded with respect to n.
P
k=1

For each positive integer k, put h(x) = y

f (k+1) for x ∈ [k, k+1). Then h is a func-
tion defined on [1, ∞). Since f is decreases
monotonically, we know that h ≤ f . Hence
n
X n
X
f (k) = f (1) + f (k)
k=1 k=2 a2
ˆ n a3 a
4 a5 a6 . . .
≤ f (1) + h(x) dx O 1 2 3 4 5 6
x
ˆ1
n ˆ ∞
≤ f (1) + f (x) dx ≤ f (1) + f (x) dx,
1 1
n
X
so that f (k) is bounded.
k=1
ˆ ∞ ∞
2. Suppose f (x) dx is divergent. We show that f (k) converges by contradiction.
P
1 k=1
 y = f (x)

6.3 Integral Test 361

For each positive integer k, put g(x) =

f (k) for x ∈ [k, k + 1). Then g is a func-
tion defined on [1, ∞). Since f is decreases y
monotonically, it is clear that g ≥ f . For
any b, we choose an integer n such that
n ≥ b. Then
ˆ b ˆ n ˆ n
f (x) dx ≤ f (x) dx ≤ g(x) dx a1
1 1 1
a2
n−1 ∞ a3
X X a4 a5 . . .
= f (k) ≤ f (k), O 1 2 3 4 5 6
x
k=1 k=1
ˆ b
so that f (x) dx is bounded, which im-
1
plies that the improper integral is conver-
gent.ˆ This contradicts to the hypothesis
∞
that f (x) dx is divergent.
1

6.38 Remark: Integral Test

y = f (x)
The hypotheses in the Integral Test say that f is a continuous, positive, decreasing function
on [1, ∞). However, it is not necessary to start the series or the integral at n = 1. In fact,
if f is decreasing for x larger than some number N , then the same conclusion holds. The
reason is that changing a finite number of terms in a convergent series does not change its
convergence.

As an example, consider the infinite se-

∞ 1
ries . As usual, if one takes
P
k=1 (2k − 5)2 y
1
f (x) = , then f is not decreasing
(2x − 5)2
on [1, ∞) (also not continuous in this inter-
val). However, we can consider the interval
[3, ∞) on which f is continuous, positive, and
decreasing. We can similarly have the con-
∞ 1
clusion that is convergent since
P
x
k=3 (2k − 5)
2 1 2 3 4 5 6
O
ˆ ∞
1
dx is convergent.
3 (2x − 5)2

Since
∞ 2 ∞
X 1 X 1 X 1
= + ,
(2k − 5) 2 (2k − 5) 2 (2k − 5)2
k=1 k=1 k=3
∞ 1
we know that is indeed convergent.
P
k=1 (2k − 5)2
362 Infinite Sequences and Series

6.39 Example: harmonic series - Integral Test

∞ 1
Apply the Integral Test to show that the harmonic series diverges.
P
k=1 k
1
Solution Consider the function f (x) = . It is clear that f is a continuous, positive,
x ˆ ∞
1
decreasing function on [1, ∞). The improper integral dx is divergent:
1 x
ˆ ∞ ˆ y = f (x)
1 b
1 b
dx = lim dx = lim ln x|x=1 = lim ln b = ∞.
1 x b→∞ 1 x b→∞ b→∞

∞ 1
Hence, by the Integral Test, the harmonic series is divergent.
P
k=1 k

6.40 Example: Integral Test

∞ k
1. Determine whether the series converges.
P
k=1 k2 +1
Solution Consider the function f (x) =
x y
. It is clear that f is a continuous,
x2 + 1
positive function on [1, ∞). It is decreas-
ing on (1, ∞), since 1
2

0
1 · (x2 + 1) − x · 2x

x
f 0 (x) = =
x +1
2 (x2 + 1)2
x
x2 − 1 O 1 2 3 4 5 6
=− 2 < 0.
(x + 1) 2
ˆ ∞
x
The improper integral 2+1
dx is divergent:
1 x
ˆ ∞ ˆ b
x x b
dx = lim dx = lim 21 ln(x2 + 1)x=1
1 x2 + 1 b→∞ 1 +1 x2 b→∞
1
= lim ln(b2 + 1) − ln 2 = ∞.

b→∞ 2

∞ k
Hence, by the Integral Test, the series is divergent.
P
k=1 k2 +1
1 ∞
2. Determine whether the series converges.
P
k=1 +2 k2
1
Solution Consider the function f (x) = 2 . It is clear that f is a continuous, posi-
x +2
6.3 Integral Test 363

ˆ ∞
1
tive, decreasing function on [1, ∞). The improper integral dx is convergent:
1 x2 + 2
ˆ ∞ ˆ
1 1 b
dx = lim dx
1 x2 +2 b→∞ 1 x 2+2
 b
= lim √12 tan−1 √12 x 

b→∞
h   x=1 i
= lim √2 tan
1 −1 √ b − tan−1
1
2
√1
2
b→∞
 
= √12 π2 − tan−1 √12 .

∞ 1
Hence, by the Integral Test, the series is convergent.
P
k 2+2
k=1

1 ∞
3. Determine whether the series converges.
P
√
k=1 2k −1
1
Solution Consider the function f (x) = √ . It is clear that f is a continuous,
2x − 1 ˆ ∞
y = f (x)
1
positive, decreasing function on [1, ∞). The improper integral √ dx is di-
1 2x −1
vergent:
ˆ ∞ ˆ
1 1b
√ dx = lim dx
√
1 2x − 1 b→∞ 1 2x − 1
√ b √ 
= lim 2x − 1x=1 = lim 2b − 1 − 1 = ∞.
b→∞ b→∞

∞ 1
Hence, by the Integral Test, the series is divergent.
P
√
k=1 2k − 1
∞ ln k
4. Determine whether the series converges.
P
k=1 k

Solution Consider the function f (x) =

ln x y
. It is clear that f is a continuous,
x
positive function on [1, ∞). It is decreas-
ing on (3, ∞), since 1
2

1

ln x
0 · x − ln x · 1
f (x) =
0
= x
x x2 x
O 1 2 3 4 5 6
ln x − 1
=− 2 < 0.
(x + 1)2
364 Infinite Sequences and Series

ˆ ∞
ln x
The improper integral dx is divergent:
3 x
ˆ ∞ ˆ
ln x ln x
b
dx = lim dx
3 x b→∞ 3 x
b
= lim 12 ln2 xx=3 = lim 1
ln2 b − ln2 3 = ∞.


b→∞ b→∞ 2

∞ ln k ∞ ln k
Thus, by the Integral Test, the series is divergent. Hence is divergent.
P P
k=3 k k=1 k

∞ 1
5. Show that the series √ converges.
P
k=1 k3
1
Solution Consider the function f (x) = √ . It is clear that f is a continuous, positive,
x3 ˆ ∞
1
decreasing function on [1, ∞). The improper integral √ dx is convergent:
1 x3
ˆ ∞ ˆ b
1 1
√ dx = lim √ dx
x 3 b→∞ x3
1 1
b
2  1
 
= lim − √  = lim 2 1 − √ = 2.
b→∞ x x=1 b→∞ b
∞ 1
Hence, by the Integral Test, the series √ is convergent. This is a special case of
P
k=1 k3
the p-series.

6.41 Theorem: convergence of the p-series

∞ 1
The p-series converges for p > 1 and diverges for p ≤ 1.
P
p
k=1 k

Justification:
1 1 1
If p < 0, then lim = ∞. If p = 0, then lim p = 1. In either case lim p 6= 0, so
k→∞ kp k→∞ k k→∞ k
the p-series diverges by the Divergence Test.
1
If p > 0, then the function f (x) = p is clearly continuous, positive, and decreasing on
x
6.3 Integral Test 365

ˆ ∞
1
[1, ∞). For the improper integral dx,
1 xp
ˆ ∞ ˆ b
1 1 b
x−p+1 
when p 6= 1 : dx = lim dx = lim
1 xp b→∞ 1 x
p b→∞ −p + 1 x=1

1
 −p+1 
b
= lim −
b→∞ −p + 1 −p + 1
1

, if p > 1;

= p−1
 ∞, if p < 1;
ˆ ∞ ˆ b
1 1 b
when p = 1 : dx = lim dx = lim ln x|x=1
1 xp b→∞ 1 x b→∞

= lim ln b = ∞,
b→∞

so that it converges if p > 1 and diverges if p ≤ 1. It follows from the Integral Test that the
∞ 1
converges if p > 1 and diverges if 0 < p ≤ 1.
P
p-series p
k=1 k

6.42 MATLAB: Integral Test

We will determine whether the following series are convergent by applying the Integral
Test.
∞ k
1. .
P
k 2+1
k=1
Solution

>> syms x
>> f ( x ) = x /( x ^2 + 1) ;
>> int ( f ( x ) ,1 , Inf )

ans =

Inf
ˆ ∞ ∞
x k
Since dx = we thus know that is divergent.
P
∞,
1 x +1
2
k=1 k + 1
2

∞ 1
2. .
P
k=1 k2 +2
Solution

>> syms x
366 Infinite Sequences and Series

>> f ( x ) = 1/( x ^2 + 2) ;
>> int ( f ( x ) ,1 , Inf )

ans =

( pi *2^(1/2) ) /4 - (2^(1/2) * atan (2^(1/2) /2) ) /2

ˆ ∞
1 ∞ 1
We see from the result above that dx < ∞, we thus know that
P
1 x2 +2 k=1 k + 2
2
is convergent.
∞ 1
3. .
P
√
k=1 2k −1
Solution

>> syms x
>> f ( x ) = 1/ sqrt (2* x - 1) ;
>> int ( f ( x ) ,1 , Inf )

ans =

Inf
ˆ ∞
1 ∞ 1
Since dx = ∞, the series is thus divergent.
P
√ √
1 2x − 1 k=1 2k −1
∞ ln k
4. .
P
k=1 k
Solution

>> syms x
>> f ( x ) = log ( x ) / x ;
>> int ( f ( x ) ,1 , Inf )

ans =

Inf
ˆ ∞
ln x ∞ ln k
Since dx = ∞, the series is thus divergent.
P
1 x k=1 k
∞ 1
5. √ .
P
k=1 k3

>> syms x
6.3 Integral Test 367

>> f ( x ) = 1/ sqrt ( x ^3) ;

>> int ( f ( x ) ,1 , Inf )

ans =

2
ˆ ∞
1 ∞ 1
We see from the result above that √ dx < ∞, the series √ is thus conver-
P
x 3 k=1 k3
1
gent.

6.43 Theorem: remainder estimate for the Integral Test

Suppose f is a continuous, positive, decreasing function on [1, ∞). Let ak = f (k) and
∞ ∞
suppose ak = s is convergent. Denote Rn = s − sn = ak . Then
P P
k=1 k=n+1
ˆ ∞ ˆ ∞
1. f (x) dx ≤ Rn ≤ f (x) dx;
n+1 n
ˆ ∞ ˆ ∞
2. sn + f (x) dx ≤ s ≤ sn + f (x) dx.
n+1 n

Justification:

1. We apply the same ideas as in the Integral Test, assuming that f is decreasing on
[n, ∞).y =
Comparing
f (x) the areas of the rectangles with
y =the
f (x)area under y = f (x) for x > n,
we see that ˆ ∞
Rn = an+1 + an+2 + · · · ≤ f (x) dx.
n
Similarly, we have ˆ ∞
Rn = an+1 + an+2 + · · · ≥ f (x) dx.
n+1

y y

an+1
an+1
an+2 an+2 an+3
an+3 a ... an+4
n+4 ...

O n
x O n+1
x
368 Infinite Sequences and Series

2. Adding sn to each side of the inequalities in part 1, we get the desired inequalities in
part 2.

6.44 Example:

In 1734, Euler derived the equality

∞
X 1 1 1 1 π2
= + + + · · · = .
k2 12 22 32 6
k=1

1. How many terms of the series must be used to obtain an approximation that is within
10−4 of the exact value of the series?
2. Find an approximation to the series using 50 terms of the series.
1
Solution The function associated with this series is f (x) = . Clearly it is continuous,
x2
positive, and decreasing on [1, ∞).

1. Using the bound on the remainder in the Integral Test, we have

ˆ ∞ ˆ ∞
dx 1
Rn < f (x) dx = 2
= .
n n x n
1
Thus, to ensure that Rn < 10−4 , we need to choose n so that < 10−4 , or n > 10, 000.
n
10,001 1
Hence, the finite series gives an approximation that is within 10−4 of the exact
P
k=1 k2
value of the series.
2. Using the bounds on the series in the Integral Test, we have
ˆ ∞ ˆ ∞
dx dx
s50 + 2
< s < s 50 + 2
,
51 x 50 x

so that
1 1
s50 + < s < s50 + ,
51 50
50 1
where s50 = . By using a calculator, we have s50 ≈ 1.62513273, which has 8th
P
2
k=1 k
decimal place accuracy. Thus,

1.64474057 < s < 1.64513273.

Taking the average of these two bounds as our approximation of s, we find that s ≈
1.64493665.
π2
For comparison, = 1.64493406 · · · . We see that our approximation matches first
6
five decimal places.
6.4 Convergence Tests for Nonnegative Series 369

6.4 Convergence Tests for Nonnegative Series

Comparison Test

6.45 Theorem: Comparison Test

Suppose that ak and bk are series with nonnegative terms.

P P

1. If bk is convergent and ak ≤ bk for all k, then ak is also convergent;

P P

2. If bk is divergent and ak ≥ bk for all k, then ak is also divergent.

P P

Justification:

1. Suppose bk is convergent. Let

P

n
X n
X ∞
X
sn = ak , tn = bk , t= bk
k=1 k=1 k=1

Since both series have positive terms, the sequences {sn } and {tn } are increasing.
Because tn → t, we have tn ≤ t for all n. Since ak ≤ bk , we have sn ≤ tn . Thus,
sn ≤ t for all n. This means that {sn } is increasing and
Pbounded above and therefore
converges by the Monotonic Sequence Theorem. Thus ak converges.
2. Suppose bk is divergent. We show that ak is divergent by contradiction.
P P

In fact, if ak is convergent, because bk ≤ ak , from part 1, we know that bk is

P P
convergent, a contradiction.

6.46 Example: Comparison Test

∞ k
1. Determine whether the series converges.
P
k=1 3k 3 − k2 + 1
Solution The given series has all positive term
terms. For any integer k ≥ 1,
1
k
k k 3k − k + 1 3 2

= 3 1

3k − 2k + 1
3 2k + (k − k ) + 1
3 2
0.5
k 2

k k
≤ 3 < 3
2k + 1 2k
k
1 1 2 3 4 5 10

< 2.
k
∞ 1
Since the p-series is convergent (p = 2), by the Comparison Test, the series
P
2
k=1 k
∞ k
is convergent.
P
k=1 3k 3 − k2 + 1
370 Infinite Sequences and Series

∞ ln k
2. Determine whether the series converges.
P
2
k=1 k

Solution The given series has nonnega-

tive terms.
√ For any integer k ≥ 1, since term
ln k < k, we know that
√ 1
ln k k 1 ln k

< 2 = 3. k2
1
k2 k k2 3
k2
0.5
P 1
∞
Since the p-series 3 is convergent
k=1 k 2
k
(p = 32 ), by the Comparison Test, the se- 1 2 3 4 5 10

∞ ln k
ries is convergent.
P
2
k=1 k

6.47 Example: Comparison Test - decimal series

Show that any decimal series of the form

0.a1 a2 a3 · · · ,

where ai is an integer satifying 0 ≤ ai ≤ 9, is convergent. In other words, it always represents

a real number.
Solution The decimal series can be rewritten as a series
∞
X
0.a1 a2 a3 · · · = a1 · 10−1 + a2 · 10−2 + a3 · 10−3 + · · · = ak · 10−k .
k=1

It is clear that it is a series with nonnegative terms. Since 0 ≤ ak ≤ 9, we know that

0 ≤ ak · 10−k ≤ 9 · 10−k for all k ≥ 1.

∞
The geometric series 9 · 10−k has common ratio 10−1 , so it converges. Hence, by the
P
k=1
Comparison Test, the decimal series always converges.

6.48 Theorem: Limit Comparison Test

Suppose that ak and bk are series with ak ≥ 0 and bk > 0. Assume

P P

ak
lim = c.
k→∞ bk
1. If 0 < c < ∞ (that is, c is a finite positive number), then ak and bk either both
P P
converge or both diverge.

2. If c = 0 and bk converges, then ak converges.

P P

3. If c = ∞ and bk diverges, then ak diverges.

P P
6.4 Convergence Tests for Nonnegative Series 371

Justification:

1. If 0 < c < ∞, let m and M be positive numbers such that m < c < M . Because
ak
→ c as k → ∞, there is an integer K such that
bk
ak
m< <M for k > K,
bk
or equivalently,
mbk < ak < M bk for k > K.
If bk converges, so does bk . Thus ak converges by part 1 of the Comparison
P P P
MP
Test. If diverges, so does and part 2 of the Comparison Test shows that
P
bk mbk
an diverges.
P

2. If c = 0, then, similarly as in Case 1, there is an integer K such that

ak < M bk for k > K.

It follows from part 1 of the Comparison Test that if bk converges, then ak con-
P P
verges.
3. If c = ∞, similarly as in Case 1, there is an integer K such that

mbk < ak for k > K.

It follows from part 2 of the Comparison Test that if bk diverges, then ak diverges.
P P

6.49 Example: Limit Comparison Test

∞ k
1. Determine whether the series converges.
P
k=1 3k 3 − k2 + 1
Solution The given series has all positive terms. We calculate the limit of the ratio of
∞ 1
the kth terms of the series and the p-series (p = 2):
P
2
k=1 k

k
3k 3 − 2k + 1 = lim k3 1 1
lim = lim = .
k→∞ 1 k→∞ 3k − 2k + 1
3 k→∞ 2 1 3
3− 2 + 3
k2 k k
Since the p-series (p = 2) is convergent, by the Limit Comparison Test, the series
∞ k
is convergent.
P
k=1 3k 3 − k2 + 1

∞ k2
2. Determine whether the series converges.
P
k=1 3k − k + 1
3 2

Solution The given series has all positive terms. We calculate the limit of the ratio of
372 Infinite Sequences and Series

∞ 1
the kth terms of the series and the harmonic series :
P
k=1 k

k2
3k 3 − 2k + 1 = lim k3 1 1
lim = lim = .
k→∞ 1 k→∞ 3k − 2k + 1
3 k→∞ 2 1 3
3− 2 + 3
k k k
Since the harmonic series is divergent, by the Limit Comparison Test, the series
∞ k2
is divergent.
P
k=1 3k − k + 1
3 2

∞ ln k
3. Determine whether the series converges.
P
2
k=1 k
3
Solution The given series has nonnegative terms. Since ln k  k 2 as k → ∞, we have
ln k
2 ln k
lim k = lim 1 = 0.
k→∞ 1 k→∞ k 2
3
k2
∞ 1
It is known that the p-series 3 is convergent (p = 2 ), by the Limit Comparison
3
P
k=1 k 2
∞ ln k
Test, the series is convergent.
P
2
k=1 k

∞ ln k
4. Determine whether the series converges.
P
k=1 k
Solution The given series has nonnegative terms. Since 1  ln k as k → ∞, we have
ln k
lim k = lim ln k = ∞.
k→∞ 1 k→∞
k
∞ 1
It is known that the harmonic series is divergent, by the Limit Comparison Test,
P
k=1 k
∞ ln k
the series is divergent.
P
k=1 k

Ratio Test and Root Test

6.50 Theorem: Ratio Test

Suppose that ak is a series with positive terms.

P
6.4 Convergence Tests for Nonnegative Series 373

ak+1
1. If lim = L < 1, then the series ak converges.
P
k→∞ ak
ak+1 ak+1
2. If lim = L > 1 or lim = ∞, then the series ak diverges.
P
k→∞ ak k→∞ ak

ak+1
3. If lim = 1, the test is inconclusive.
k→∞ ak

Justification:

ak+1
1. If lim = L < 1, we take any number r such that L < r < 1. Thus, there is K
k→∞ ak
such that when k ≥ K,
 
 ak+1
 < r − L =⇒ −(r − L) < ak+1 − L < r − L


 ak − L ak
ak+1
=⇒ <r
ak
=⇒ ak+1 < r · ak .
Putting k successively equal to K, K + 1, K + 2, . . . in the inequality, we obtain
aK+1 < r · aK ,
aK+2 < r · aK+1 < r2 · aK ,
aK+3 < r · aK+2 < r3 · aK ,
..
.
so that
aK+k < rk · aK , for all k ≥ 1.
∞
Since |r| = r < 1, the geometric series aK rk converges. By the Comparison Test,
P
k=1
∞ ∞
the series ak converges. Hence, ak converges.
P P
k=K+1 k=1

2. To show ak diverges, by the Divergence Test, it is sufficient to show that lim ak 6= 0.

P
k→∞
ak+1
Since lim = L > 1, there is K such that when k ≥ K,
k→∞ ak
 
 ak+1
 < (L − 1) =⇒ −(L − 1) < ak+1 − L


 ak − L  ak
ak+1
=⇒ >1
ak
=⇒ ak+1 > ak > · · · > aK .
Hence, ak ≥ aK for k ≥ K. This implies that lim ak 6= 0.
k→∞
ak+1
Similarly as above, if lim = ∞, we can also show that lim ak 6= 0.
k→∞ ak k→∞
374 Infinite Sequences and Series

∞ 1 ∞ 1
3. Consider the p-series with p = 1 and with p = 2 > 1. The former is
P P
2
k=1 k k=1 k
divergent and the latter is convergent. For both series, we have
ak+1
lim = 1.
k→∞ ak
ak+1
Hence, the Ratio Test is inconclusive if lim = 1.
k→∞ ak

6.51 Example: Ratio Test

∞ bk
1. Determine whether the series converges, where b > 0.
P
k=1 k!
Solution The given series has all positive terms. Since, for any b > 0,

bk+1
(k + 1)! b
lim = lim = 0,
k→∞ bk k→∞ k + 1

k!
∞ bk
by the Ratio Test, the series is convergent for any positive b.
P
k=1 k!

∞ k!
2. Determine whether the series converges.
P
k
k=1 k
Solution The given series has all positive terms. The ratio of two consecutive terms is

(k + 1)!
(k + 1)k+1 1 1
= = k .
k! (k + 1) k
1
kk 1+
kk k
x k
1 1
 
Since e = lim 1+ , we have e = lim 1 + , so that
x→∞ x k→∞ k

(k + 1)!
(k + 1)k+1 1 1
lim = = < 1.
k→∞ k! 
1
k e
kk lim 1 +
k→∞ k
∞ k!
By the Ratio Test, the series is convergent.
P
k
k=1 k
6.4 Convergence Tests for Nonnegative Series 375

6.52 MATLAB: Ratio Test

We will determine whether the following two series converge by applying the Ratio Test.
∞ k
X b
1. where b > 0.
k!
k=1
Solution

>> syms b k
>> assume (b , ' positive ') % b > 0
>> assume (k , ' positive ') % k > 0
>> assumeAlso (k , ' integer ') % k is also an integer
>> a ( k ) = b ^ k / factorial ( k ) ;
>> ratio_a = simplify ( a ( k + 1) / a ( k ) ) ;
>> limit ( ratio_a ,k , inf )

ans =

Note that we can set assumptions on symbolic objects using commands assume and
assumeAlso. To calculate factorial in MATLAB, the command factorial is used.
Moreover, the command simplify is used for symbolic simplification, so that the cal-
culation of the limit for the factorial ratio becomes feasible.
∞ k
ak+1 X b
The result above shows that lim = 0 < 1, then the series is convergent.
k→∞ ak k!
k=1
∞
X k!
2. .
kk
k=1
Solution

>> syms k
>> assume (k , ' positive ') % k > 0
>> assumeAlso (k , ' integer ') % k is also an integer
>> a ( k ) = factorial ( k ) / k ^ k ;
>> ratio_a = simplify ( a ( k + 1) / a ( k ) ) ;
>> limit ( ratio_a ,k , inf )

ans =

exp ( -1)

∞
ak+1 X k!
From the result above, we see that lim = e−1 < 1, so that the series is
k→∞ ak kk
k=1
convergent.
376 Infinite Sequences and Series

6.53 Theorem: Root Test

Suppose that ak is a series with nonnegative terms.

P

√
1. If lim k ak = L < 1, then the series ak converges.
P
k→∞
√ √
2. If lim ak = L > 1 or lim ak = ∞, then the series ak diverges.
P
k k
k→∞ k→∞
√
3. If lim k ak = 1, the test is inconclusive.
k→∞

Justification:
√
1. If lim k ak = L < 1, we take any number r such that L < r < 1. Thus, there is K
k→∞
such that when k ≥ K,
√ √
| k ak − L| < r − L =⇒ −(r − L) < k ak − L < r − L
√
=⇒ k ak < r
=⇒ ak < rk .

When |r| < 1, the geometric series r converges. By the Comparison Test, the series
P k
∞ ∞
ak converges. Hence, ak converges.
P P
k=K k=1

2. To show ak diverges, by the Divergence Test, it is sufficient to show that lim ak 6= 0.

P
k→∞
√
Since lim ak = L > 1, there is K such that when k ≥ K,
k
k→∞
√ √
| k ak − L| < (L − 1) =⇒ −(L − 1) < k ak − L
√
=⇒ k ak > 1
=⇒ ak > 1.

This implies that lim ak 6= 0.

k→∞
√
Similarly as above, if lim k ak = ∞, we can also show that lim ak 6= 0.
k→∞ k→∞

∞1 1 ∞
3. Consider the p-series with p = 1 and with p = 2 > 1. The former is
P P
k k 2
k=1 k=1
divergent and the latter is convergent. For both series, we have
√
lim k
ak = 1.
k→∞
√
Hence, the Root Test is inconclusive if lim k ak = 1.
k→∞
6.4 Convergence Tests for Nonnegative Series 377

6.54 Example: Root Test

k
∞ 3k − 1

1. Determine whether the series converges.
P
k=1 4k + 1
Solution The given series has all positive terms. It is easy to have
1
3−
s
k
3k − 1 3k − 1 k = 3 < 1.

lim = lim = lim
k

k→∞ 4k + 1 k→∞ 4k + 1 k→∞ 1 4

4+
k
k
∞ 3k − 1

By the Root Test, the series is convergent.
P
k=1 4k + 1

∞
 −k2
k
2. Determine whether the series converges.
P
k=1 k+1
x
1

Solution The given series has all positive terms. Since e = lim 1 + , we have
x→∞ x
k
1

e = lim 1 + , so that
k→∞ k
s
−k2 k k
k+1 1
  
k
lim = lim = lim 1 + = e > 1.
k

k→∞ k+1 k→∞ k k→∞ k

∞
 −k2
k
By the Root Test, the series is divergent.
P
k=1 k+1

6.55 MATLAB: Root Test

We will determine whether the following two series converge by applying the Root Test.
∞  k
X 3k − 1
1. .
4k + 1
k=1
Solution

>> syms k
>> assume (k , ' positive ') % k > 0
>> assumeAlso (k , ' integer ') % k is also an integer
>> a ( k ) = ((3* k - 1) /(4* k + 1) ) ^ k ;
>> root_a = ( a ( k ) ) ^(1/ k ) ;
>> limit ( root_a ,k , inf )

ans =
378 Infinite Sequences and Series

3/4

∞  k
ak+1 3 X 3k − 1
The result above shows that lim = < 1, then the series is
k→∞ ak 4 4k + 1
k=1
convergent.
2
∞  −k
X k
2. .
k+1
k=1
Solution

>> syms k
>> assume (k , ' positive ') % k > 0
>> assumeAlso (k , ' integer ') % k is also an integer
>> a ( k ) = ( k /( k + 1) ) ^( - k ^2) ;
>> root_a = ( a ( k ) ) ^(1/ k ) ;
>> limit ( root_a ,k , inf )

ans =

exp (1)

2
∞  −k
ak+1 X k
We see from the result above that lim = e > 1, so that the series
k→∞ ak k+1
k=1
is divergent.

6.5 Convergence Tests for General Series

Alternating Series

6.56 Example: alternating series

An alternating series is a series whose terms are alternately positive and negative. The
following are two examples of alternating series:
∞
1 1 1 1 1 X (−1)k−1
1− + − + − + ··· = ,
2 3 4 5 6 k
k=1
∞ √
√ √ √ √ √ X
−1 + 2− 3+ 4− 5+ 6 − ··· = (−1)k
2 3 4 5 6 k
k.
k=1
6.5 Convergence Tests for General Series 379

∞
In general, an alternating series ak has its kth term in the form
P
k=1

ak = (−1)k−1 bk or ak = (−1)k bk ,

where bk > 0 for all k.

6.57 Theorem: Alternating Series Test

The alternating series

∞
X
(−1)k−1 bk = b1 − b2 + b3 − b4 + b5 − b6 + · · · , bk > 0
k=1

converges if

1. if {bk } is a positive non-increasing sequence, bk ≥ bk+1 > 0, and

2. lim bk = 0.
k→∞

Justification:
First consider the even partial sums:

s2 = b1 − b2 ≥ 0, since b1 ≥ b2
s4 = s2 + (b3 − b4 ) ≥ s2 , since b3 ≥ b4
..
.
s2n = s2n−2 + (b2n−1 − b2n ) ≥ s2n−2 since b2n−1 ≥ b2n

Also,
s2n = b1 − (b2 − b3 ) − (b4 − b5 ) − · · · − (b2n−2 − b2n−1 ) − b2n ≤ b1 .
So, {s2n } is increasing and bounded above. Hence, it is convergent by the Monotonic Sequence
Theorem. Denote s = lim s2n . Now we compute the limit of the odd partial sums:
n→∞

lim s2n+1 = lim (s2n + b2n+1 ) = lim s2n + lim b2n+1 = s + 0 = s.

n→∞ n→∞ n→∞ n→∞

Since both the even and odd partial sums converge to s, we have lim sn = s and so the
n→∞
series is convergent.

6.58 Example: alternating harmonic series

The alternating harmonic series

∞ ∞
1 1 1 1 1 X (−1)k−1 X
1 − + − + − + ··· = = (−1)k−1 bk ,
2 3 4 5 6 k
k=1 k=1
380 Infinite Sequences and Series

1
where bk = satisfies
k
bk ≥ bk+1 > 0 and lim bk = 0.
k→∞

Hence, by the Alternating Series Test, the series converges.

6.59 MATLAB: Alternating Series Test

∞
X 1
Determine whether the series (−1)k−1 bk where bk = converges.
k
k=1
1
Solution Clearly, we know that bk = > 0 when k ≥ 1. Hence, we will check whether
k
lim bk = 0 and bk ≥ bk+1 in the following:

>> syms k
>> assume (k , ' positive ') % k > 0
>> assumeAlso (k , ' integer ') % k is also an integer
>> b ( k ) = 1/ k ;
>> limit ( b ( k ) ,k , inf )

ans =

>> Delta_b = b ( k ) - b ( k + 1) ;
>> isAlways ( Delta_b > 0)

ans =

logical

Note that the command isAlways can be used to test whether a mathematical statement is
true or not. If the mathematical statement is true, then the result will be 1; if it is false,
then the result will be 0.
From the above result, we see that lim bk = 0 and bk > bk+1 . Therefore, according to the
∞
X 1
Alternating Series Test, the series (−1)k−1 bk where bk = is convergent.
k
k=1

6.60 Theorem: remainder estimate for alternating series

∞
If s = (−1)k bk is the sum of an alternating series that satisfies
P
k=1
6.5 Convergence Tests for General Series 381

1. bk+1 ≤ bk for all k;

2. lim bk = 0
k→∞

then
|Rn | = |s − sn | ≤ bn+1 .

Justification:

The even partial sums {s2n } form an increasing sequence:

s2 = b1 − b2 ≥ 0, since b1 ≥ b2
s4 = s2 + (b3 − b4 ) ≥ s2 , since b3 ≥ b4
..
.
s2n = s2n−2 + (b2n−1 − b2n ) ≥ s2n−2 since b2n−1 ≥ b2n

Similarly, the odd partial sums {s2n+1 } form a decreasing sequence. Because we know that
both {s2n } and {s2n+1 } converge to s, we get

s2n ≤ s ≤ s2n+1 , for all n.

It follows that
|Rn | = |s − sn | ≤ |sn+1 − sn | = bn+1 .

6.61 Example:

Consider the series

∞
X (−1)k 1 1 1
=1− + − + ··· .
k! 2! 3! 4!
k=0

1. How many terms of the series must be used to obtain an approximation that is within
10−4 of the exact value of the series?
2. What is the approximation value?
Solution Write
∞ n
X (−1)k X (−1)k
= + Rn .
k! k!
k=0 k=0

1. Using the bound on the remainder in the Alternating Series Test, we have
1
|Rn | < .
(n + 1)!
1
Thus, to ensure that |Rn | < 10−4 , we need to choose n so that < 10−4 , or
(n + 1)!
7 (−1)k
n > 7. Hence, the finite series gives an approximation that is within 10−4 of
P
k=0 k!
the exact value of the series.
382 Infinite Sequences and Series

2. Using the bounds on the series in the Alternating Series Test, we have
1 1
s7 − < s < s7 + ,
8! 8!
7 (−1)k
where s7 = . By using a calculator, we have s7 ≈ 0.36786, which has 5th
P
k=0 k!
decimal place accuracy. Thus,

0.36784 < s < 0.36789.

Taking the average of these two bounds as our approximation of s, we find that s ≈
0.36787.
∞ (−1)k
For comparison, = e−1 = 0.3678794412 · · · . We see that our approximation
P
k=0 k!
indeed is within 10−4 of the exact value of the series.

Absolute Convergence and Conditional Convergence

6.62 Definition: absolute convergece and conditional convergence

• A series ak is called absolutely convergent if the series of absolute values

P P
|ak |
is convergent.
• A series ak is called conditionally convergent if it is convergent but not absolutely
P
convergent.

6.63 Theorem: convergence of absolutely convergent series

If a series ak is absolutely convergent, then it is convergent.

P

Justification:
It is clear that
0 ≤ ak + |ak | ≤ 2|ak |, for all k.
If
P ak is absolutely convergent, then 2|ak | is convergent. By the Comparison Test,
P P
(ak + |ak |) is convergent. Thus,
X X X
ak = (ak + |ak |) − |ak |

is convergent.
6.5 Convergence Tests for General Series 383

6.64 Example: absolutely/conditionally convergent series

∞ 1
1. The p-series , with p = 2, is convergent, so the alternating series
P
2
k=1 k

∞
X (−1)k−1 1 1 1
=1− + − + ···
k2 4 9 16
k=1

is absolutely convergent.
∞ (−1)k−1
2. The alternating harmonic series is convergent by the Alternating Series
P
k=1 k 
∞ 1 ∞  (−1)k−1 

Test, but the harmonic series =  is divergent, so the alternating
P P 
k=1 k k
 
k=1
harmonic series is conditionally convergent.

6.65 Remark: differences between absolutely and conditionally convergent series

There are some fundamental differences between absolutely convergent series and condi-
tionally convergent series.
Riemann discovered a striking result, known as the Riemann Rearrangement Theorem. It
says that for any given number α, a conditionally convergent series may be rearranged in its
terms so that the rearranged series converges to α.
384 Summary

Summary

Definition: limit of infinite sequence

• lim an = L ⇐⇒ we can make the terms an as close to L as we like by taking n

n→∞
sufficiently large.
– lim an exists ⇐⇒ {an } converges (or is convergent).
n→∞
– lim an does not exist ⇐⇒ {an } diverges (or is divergent).
n→∞

• lim an = ∞ ⇐⇒ we can make the terms an as large as we like by taking n sufficiently

n→∞
large.
– lim an = ∞ ⇐⇒ {an } diverges to infinity.
n→∞
– lim an = −∞ ⇐⇒ {an } diverges to negative infinity.
n→∞

Definition: terminology for sequences

• {an } is increasing if an+1 > an . For example, {1, 2, 3, . . . }.

• {an } is non-decreasing if an+1 ≥ an . For example, {1, 1, 2, 2, 3, 3, . . . }.

• {an } is decreasing if an+1 < an . For example, {0, −1, −2, −3, . . . }.
• {an } is non-increasing if an+1 ≤ an . For example, {0, 0, −1, −1, −2, −2, . . . }.
• {an } is monotonic if it is either non-increasing or non-decreasing.

• {an } is bounded if there is number M such that |an | ≤ M for all n.

6.66 Definition: two asymptotic relations

an
• an  bn (as n → ∞) ⇐⇒ lim = 0.
n→∞ bn
an
• an ∼ bn (as n → ∞) ⇐⇒ an − bn  bn (as n → ∞) ⇐⇒ lim = 1.
n→∞ bn

Theorem: limit of a sequence from limit of a function

If lim f (x) = L and f (n) = an when n is an integer, then lim an = L. The same result
x→∞ n→∞
holds when L = ∞ or −∞.

Theorem: Limit Laws for sequences

Suppose {an } and {bn } are convergent sequences and c is a constant. Then the following
hold.
 385

1. lim (an + bn ) = lim an + lim bn .

n→∞ n→∞ n→∞

2. lim (an − bn ) = lim an − lim bn .

n→∞ n→∞ n→∞

3. lim can = c lim an . In particular, lim c = c.

n→∞ n→∞ n→∞

4. lim (an bn ) = lim an · lim bn .

n→∞ n→∞ n→∞

an lim an
5. lim = n→∞ if lim bn 6= 0.
n→∞ bn lim bn n→∞
n→∞
h ip
6. lim apn = lim an if p > 0 and an > 0.
n→∞ n→∞

Theorem: Squeeze Theorem

If an ≤ bn ≤ cn for n ≥ n0 and lim an = lim cn = L, then lim bn = L.

n→∞ n→∞ n→∞

Theorem: growth rates of sequences

For any real numbers p > 0, q > 0, r > 0, and b > 1, the following asymptotic relations
hold: as n → ∞,
lnq n  np  np lnr n  bn  n!  nn .

Theorem: geometric sequences

Let r be a real number. Then

 0, if |r| < 1;


lim rn = 1, if r = 1;
n→∞
diverges, if r ≤ −1 or r > 1.


Theorem: sequential limit for continuous function

If lim an = L and the function f is continuous at L, then

n→∞

lim f (an ) = f (L).

n→∞

Theorem: Monotonic Sequence Theorem

Every bounded, monotonic sequence is convergent.

386 Summary

Definition: convergent series

∞ n
ak converges to s ⇐⇒ lim sn = s, where sn = ak .
P P
k=1 n→∞ k=1

Theorem: geometric series

Let a 6= 0 and r be a real number. Then

 a
∞  , if |r| < 1;
1−r
X
ark−1 =
diverges, if |r| ≥ 1.

k=1

Theorem: properties of convergent series

1. If ak converges and c is a real number, then the series cak converges and
P P

X X
cak = c ak .

2. If ak and bk converge, then the series (ak ± bk ) converges and

P P P

X X X
(ak ± bk ) = ak ± bk .

∞ ∞
3. If N is a positive integer, then ak and ak either both converge or both diverge.
P P
k=1 k=N
In general, changing a finite number of terms in a convergent series does not change its
convergence, although it does change the value of the series.

Theorem: Divergence Test

If ak converges, then lim ak = 0. Hence, if lim ak 6= 0, then the series diverges.

P
k→∞ k→∞

Theorem: Integral Test

Suppose f is a continuous, positive, decreasing function on [1, ∞)

ˆ and let ak = f (k). Then
∞ ∞
the series ak is convergent if and only if the improper integral f (x) dx is convergent.
P
k=1 1
In other words:
ˆ ∞ ∞
1. If f (x) dx is convergent, then ak is convergent;
P
1 k=1
ˆ ∞ ∞
2. If f (x) dx is divergent, then ak is divergent.
P
1 k=1
 387

Theorem: remainder estimate for the Integral Test

Suppose f is a continuous, positive, decreasing function on [1, ∞). Let ak = f (k) and
∞ ∞
suppose ak = s is convergent. Denote Rn = s − sn = ak . Then
P P
k=1 k=n+1
ˆ ∞ ˆ ∞
1. f (x) dx ≤ Rn ≤ f (x) dx;
n+1 n
ˆ ∞ ˆ ∞
2. sn + f (x) dx ≤ s ≤ sn + f (x) dx.
n+1 n

Theorem: convergence of the p-series

∞ 1
The p-series converges for p > 1 and diverges for p ≤ 1.
P
p
k=1 k

Theorem: Comparison Test

Suppose that ak and bk are series with nonnegative terms.

P P

1. If bk is convergent and ak ≤ bk for all k, then ak is also convergent;

P P

2. If bk is divergent and ak ≥ bk for all k, then ak is also divergent.

P P

Theorem: Limit Comparison Test

Suppose that ak and bk are series with ak ≥ 0 and bk > 0. Assume

P P

ak
lim = c.
k→∞ bk
1. If 0 < c < ∞ (that is, c is a finite positive number), then ak and bk either both
P P
converge or both diverge.

2. If c = 0 and bk converges, then ak converges.

P P

3. If c = ∞ and bk diverges, then ak diverges.

P P

Theorem: Ratio Test

Suppose that ak is a series with positive terms.

P

ak+1
1. If lim = L < 1, then the series ak converges.
P
ak
k→∞

ak+1 ak+1
2. If lim = L > 1 or lim = ∞, then the series ak diverges.
P
k→∞ ak k→∞ ak
388 Summary

ak+1
3. If lim = 1, the test is inconclusive.
k→∞ ak

Theorem: Root Test

Suppose that ak is a series with nonnegative terms.

P

√
1. If lim k ak = L < 1, then the series ak converges.
P
k→∞
√ √
2. If lim ak = L > 1 or lim ak = ∞, then the series ak diverges.
P
k k
k→∞ k→∞
√
3. If lim k ak = 1, the test is inconclusive.
k→∞

Theorem: Alternating Series Test

The alternating series

∞
X
(−1)k−1 bk = b1 − b2 + b3 − b4 + b5 − b6 + · · · , bk > 0
k=1

converges if

1. if {bk } is a positive non-increasing sequence, bk ≥ bk+1 > 0, and

2. lim bk = 0.
k→∞

Theorem: remainder estimate for alternating series

∞
If s = (−1)k bk is the sum of an alternating series that satisfies
P
k=1

1. bk+1 ≤ bk for all k;

2. lim bk = 0
k→∞

then
|Rn | = |s − sn | ≤ bn+1 .

Definition: absolute convergece and conditional convergence

• A series ak is called absolutely convergent if the series of absolute values

P P
|ak |
is convergent.
• A series ak is called conditionally convergent if it is convergent but not absolutely
P
convergent.
 389

Theorem: convergence of absolutely convergent series

If a series ak is absolutely convergent, then it is convergent.

P
390 Summary
 CHAPTER

7
Series of Functions

7.1 Approximating Functions with Polynomials

Taylor Polynomials

7.1 Example: linear approximation and quadratic approximation

1. We know that if a function f is differen-

tiable at a point a, then it can be ap-
proximated near a by its tangent line y = y
f (a) + f 0 (a)(x − a) (or “linearization”).
This is a linear approximation to f at the
point a. Since it is a first-degree polyno- y = f (x)
mial, it is convenient to write
y = p1 (x)
p1 (x) = f (a) + f 0 (a)(x − a).

Clearly, this polynomial p1 matches f in

value and in slope at a. That is, a
x

p1 (a) = f (a) and p01 (a) = f 0 (a).

392 Series of Functions

2. If f has a fairly constant slope near a, the

tangent line provides a good approxima-
tion. However, if the graph of f near a y
has a larger curvature, then the linear ap-
proximation may not be good enough. To
improve accuracy, we consider a quadratic y = f (x)
approximating polynomial p2 by adding y = p2 (x)
one new term to the linear polynomial: y = p1 (x)

p2 (x) = f (a) + f (a)(x − a) + c2 (x − a) .

0 2

The constant c2 should be chosen so that a

x
p2 is a good approximation to f near a. To
do so, we require that p2 agree with f in
value, slope, and concavity at a:
p2 (a) = f (a), p02 (a) = f 0 (a), and p002 (a) = f 00 (a),
where we assume that f (a), f 0 (a), and f 00 (a) exist. In fact, we can easily verify that
indeed p2 satisfies the first and the second equations. The third equation holds if c2
satisfies 2c2 = f 00 (a), or c2 = 21 f 00 (a). Hence, the quadratic approximating polynomial
is
f 00 (a)
p2 (x) = f (a) + f 0 (a)(x − a) + (x − a)2 .
2

√
7.2 Example: linear approximation and quadratic approximation for x
√
Find the linear approximation and √the quadratic approximation for x at x = 1. Then
use them to approximate the value of√ 1.03.
Solution For the function f (x) = x, we have
1 3
f 0 (x) = 12 x− 2 , f 00 (x) = − 14 x− 2 .

Thus, the linear approximation and the quadratic approximation at x = 1 are, respectively,

p1 (x) = f (1) + f 0 (1)(x − 1)

= 1 + 12 (x − 1),
f 00 (1)
p2 (x) = f (1) + f 0 (1)(x − 1) + (x − 1)2
2
= 1 + 12 (x − 1) − 81 (x − 1)2 .
√
Thus, by using these formulas, we obtain the approximating values of 1.03:

p1 (1.03) = 1 + 12 (1.03 − 1) = 1.015,

p2 (1.03) = 1 + 12 (1.03 − 1) − 81 (x − 1)2 = 1.0148875.
√
For comparison, the value of 1.03 is 1.014889156509222 · · · . We see that the quadratic
approximation indeed yields better accuracy.
7.1 Approximating Functions with Polynomials 393

7.3 Definition: Taylor polynomials

Let f be a function with f 0 , f 00 , . . . , and f (n) defined at a. The nth-order Taylor poly-
nomial for f with its center at a is

f 00 (a) f (n) (a)

Tn (x) = f (a) + f 0 (a)(x − a) + (x − a)2 + · · · + (x − a)n .
2! n!

7.4 Theorem: uniqueness of Taylor polynomials

Suppose f is a function with f 0 , f 00 , . . . , and f (n) defined at a. If Pn is a polynomial of

degree n satisfying

Pn (a) = f (a), Pn0 (a) = f 0 (a), Pn00 (a) = f 00 (a), ..., Pn(n) (a) = f (n) (a),

Then
f 00 (a) f (n) (a)
Pn (x) ≡ Tn (x) = f (a) + f 0 (a)(x − a) + (x − a)2 + · · · + (x − a)n .
2! n!

Justification:
Write Pn as
Pn (x) = c0 + c1 (x − a) + · · · + cn (x − a)n .
The equation Pn (a) = f (a) gives c0 = f (a). Differentiating the expression of Pn and evalu-
ating at x = a gives Pn0 (a) = c1 , so that c1 = f 0 (a). In general, differentiating the expression
(k) f (k) (a)
of Pn for k times and evaluating at x = a gives Pn (a) = ck · k!, so that ck = . Hence,
k!
f 00 (a) f (n) (a)
Pn (x) = f (a) + f 0 (a)(x − a) + (x − a)2 + · · · + (x − a)n = Tn (x).
2! n!

7.5 Example: Taylor polynomials for ex

Find the Taylor polynomials of order n = 0, 1, 2, 3, and 4 for f (x) = ex centered at 0.

Graph f and the polynomials. Use T4 (x) to approximate the value of e0.1 . What is the
nth-order Taylor polynomial centered at 0 for general n?
Solution For any integer k ≥ 0, we have

f (k) (x) = ex =⇒ f (k) (0) = 1.

394 Series of Functions

Thus, the first few Taylor polynomials for f (x) = ex centered at 0 are:

T0 (x) = f (0) = 1,
T1 (x) = f (0) + f 0 (0)x = 1 + x,
f 00 (0) 2 x2
T2 (x) = f (0) + f 0 (0)x + x =1+x+ ,
2! 2!
f 00 (0) 2 f 000 (0) 3 x2 x3
T3 (x) = f (0) + f (0)x +
0
x + x =1+x+ + ,
2! 3! 2! 3!
f 00
(0) f 000
(0) f (4)
(0) x2 x3 x4
T4 (x) = f (0) + f 0 (0)x + x2 + x3 + x4 = 1 + x + + + .
2! 3! 3! 2! 3! 4!

y
y = T4 (x)

7
y = T3 (x)

x
6 y=e

5
y = T2 (x)
4

2 y = T1 (x)

y = T1 (x)
y = T0 (x)
x
−1 O 1

Using the Taylor polynomial T4 , we get the approximation to e0.1 (to 8th decimal places):

T4 (0.1) ≈ 1.10517083.

The actual value of e0.1 to 8th decimal places is 1.10517092.

In general, we have the nth-order Taylor polynomial centered at 0:
n
x2 x3 xn X xk
Tn (x) = 1 + x + + + ··· + = .
2! 3! n! k!
k=0

7.6 Theorem: asymptotic approximation of Taylor polynomials

f (x) − Tn (x)
If f (n) (a) exists, then lim = 0, or equivalently, by using little-o notation,
x→a (x − a)n
n
X f (k) (a)
f (x) = Tn (x) + o(|x − a|n ) = (x − a)k + o(|x − a|n ), as x → a.
k!
k=0
7.1 Approximating Functions with Polynomials 395

Justification:
By applying l’Hospital’s Rule, we have

f (n) (a)
f (x) − Tn (x) f (x) − Tn−1 (x) − (x − a)n
lim = lim n!
x→a (x − a)n x→a (x − a)n
f (x) − Tn−1 (x) f (n) (a)
= lim −
x→a (x − a)n n!
H f (x) − Tn−1 (x) f (a) H
0 0 (n)
= lim − = ···
x→a n(x − a)n−1 n!
(n−1)
H f (n−1) (x) − Tn−1 (x) f (n) (a)
= lim −
x→a n(n − 1) · · · 2(x − a) n!
f (n−1) (x) − f (n−1) (a) f (n) (a)
= lim −
x→a n(n − 1) · · · 2(x − a) n!
1 f (n−1) (x) − f (n−1) (a) f (n) (a)
= · lim −
n! x→a x−a n!
1 f (a)
(n)
= · f (n) (a) − = 0.
n! n!

7.7 Pitfall: Taylor approximation

From the conclusion that f (x) = Tn (x) + o(|x − a|n ) (as x → a), we know that Tn (x)
approximates f (x) as x → a and the accuracy improves as n increases. However, this
does not imply that the sequence {Tn (x)} approaches f (x) as n → ∞. In fact, generally
f (x) 6= lim Tn (x).
n→∞

As an example, consider the function

y
 −1
e x2 , if x 6= 0,
f (x) =
0, if x = 0.
y = f (x)
1

We can show that f (n) (0) = 0 (n ≥ 0) by

induction. So, the n-th order Taylor polyno-
x
mial for f at 0 is −1 1

Tn (x) = 0 + 0 · x + · · · + 0 · xn = 0.
By Applying l’Hospital’s Rule, we can see that as x → 0,

f (x) = Tn (x) + o(|x|n ) = o(|x|n ), for all n ≥ 0.

However, for any x 6= 0,

f (x) 6= 0 = 0 + 0 · x + · · · + 0 · xn + · · · = lim Tn (x).

n→∞
396 Series of Functions

7.8 MATLAB: Taylor polynomials

Find the Taylor polynomial of order n = 4 for f (x) = ex centered at 0.

Solution In MATLAB, we can use the command taylor to calculate the Taylor polynomial
approximation for a function. The following MATLAB code shows how to find the Taylor
polynomial of order 4 for f (x) = ex centered at 0.

>> syms x
>> f ( x ) = exp ( x ) ;
>> taylor ( f ( x ) ,x ,0 , ' Order ' ,4 + 1) % centered at 0 , order 4 ...
plus one constant term

ans =

x ^4/24 + x ^3/6 + x ^2/2 + x + 1

Since the first term of the Taylor polynomial is a constant, the Taylor polynomial of order
4 for f (x) = ex has 5 terms. In other words, the value of ‘Order’ in the command taylor
is the required order of the Taylor polynomial plus 1.

7.2 Represent Functions by Power Series

7.9 Definition: remainder of Taylor series

Let Tn be the Taylor polynomial of order n for f at a:

f 00 (a) f (n) (a)

Tn (x) = f (a) + f 0 (a)(x − a) + (x − a)2 + · · · + (x − a)n .
2! n!
The term Rn (x) = f (x) − Tn (x) is called the remainder of the Taylor series. The latter
refers the following “formal” power series of f :

f 00 (a) f (n) (a)

f (a) + f 0 (a)(x − a) + (x − a)2 + · · · + (x − a)n + · · · .
2! n!
This power series is also called Taylor expansion of f at a. A Taylor series centered at 0
is called a Maclaurin series.

7.10 Definition: power series

A power series is a series of the form

∞
X
ck xk = c0 + c1 x + c2 x2 + c3 x3 + · · · ,
k=0

where x is a variable and the ck ’s are constants called the coefficients of the series.
7.2 Represent Functions by Power Series 397

More generally, a series of the form

∞
X
ck (x − a)k = c0 + c1 (x − a) + c2 (x − a)2 + · · · ,
k=0

is called a power series in (x − a) or a power series centered at a or a power series

about a.

7.11 Pitfall: power series

For any given function f , if f (n) (a) (n ≥ 0) all exist, then the Taylor series for f at a is
defined as
∞
X f (k) (a)
(x − a)k .
k!
k=0

The series is defined formally, since the series may not converge for some x at which f is
defined.
1
As an example, consider the function f (x) = . It has the following power series
1+x
(centered at 0):
1 − x + x2 − x3 + x4 − x5 + · · · .
In fact, we can have by induction that

(−1)n n!
f (n) (x) = ,
(1 + x)n

so that f (n) (0) = (−1)n n!. Thus, the Taylor series for f at 0 is
∞ ∞ ∞
X f (k) (0) X (−1)k k! X
xk = xk = (−1)k xk .
k! k!
k=0 k=0 k=0

It is clear that the function f is defined everywhere except at x = −1, but the series only
converges for |x| < 1.

7.12 Theorem: convergence theorem of Taylor series

If lim Rn (x) = 0 in an open interval I that contains a, then

n→∞

∞
X f (k) (a)
f (x) = lim Tn (x) = (x − a)k , x ∈ I.
n→∞ k!
k=0

Justification:

If lim Rn (x) = 0, then

n→∞

lim Tn (x) = lim [f (x) − Rn (x)] = f (x) − lim Rn (x) = f (x).

n→∞ n→∞ n→∞
398 Series of Functions

It is clear that {Tn (x)} are the partial sums of the series
∞
X f (k) (a) f 00 (a) f (k) (a)
(x − a)k = f (a) + f 0 (a)(x − a) + (x − a)2 + · · · + (x − a)k + · · · .
k! 2! k!
k=0

Thus, we have
∞
X f (k) (a)
f (x) = (x − a)k .
k!
k=0

7.13 Engineering Example: Z-transform

The Z-transform converts a sequence of discrete-time signals into a complex frequency-

domain representation for easy interpretations. The Z-transform can be defined as either a
two-sided or one-sided transform.
Specifically, given a discrete-time signal x[n], the two-sided Z-transform is defined as
∞
X
X(z) = x[n]z −n .
n=−∞

where z is, in general, a complex number. Alternatively, if the signal x[n] is defined only for
n ≥ 0, the one-sided Z-transform is defined as
∞
X
X(z) = x[n]z −n .
n=0

The Z-transform has been applied widely in science and engineering fields, e.g., probability
theory, digital signal processing (DSP) and control systems.

7.14 Engineering Example: continuous-time Fourier series

The Fourier series can represent periodic signals as sums of complex exponentials (or, sine
functions and cosine functions). Given a continuous-time periodic signal x(t), its continuous-
time Fourier series is defined as
∞
2π
X
x(t) = ak ej T kt

k=−∞

where ak is the Fourier series coefficient determined by x(t), and T is the period of x(t). From
the above definition, we see that the Fourier series can represent signals by their harmonics
2π
ej T kt .
The Fourier series has many applications in electrical engineering, signal processing, image
processing, acoustics, optics, etc. For example, since the Fourier series can represent signals
by their harmonics, it is thus very useful for the signals which can be naturally described by
harmonics, e.g., music signals.
Moreover, for a linear time-invariant (LTI) system with frequency response H(jω), if the
7.2 Represent Functions by Power Series 399

∞
2π
X
input is given by x(t) = ak ej T kt
, then the output is
k=−∞

∞
2π
 
2π
X
y(t) = ak H j k ej T kt .
T
k=−∞

Since the amplitude of each component is scaled by H j 2π T k , as seen above, the output of

the LTI system is thus a “filtered” version of the input. In other words, the Fourier series
motives a new representation of a system as a filter. This is useful for many systems, such
as the speech synthesis system.

Remainder Formulas
7.15 Theorem: integral form of the remainder

If f (n+1) is continuous in an open interval I that contains a, then

ˆ
1 x
Rn (x) = (x − t)n f (n+1) (t) dt, x ∈ I.
n! a

Justification:
We prove the remainder formula by induction.
For n = 0, the formula gives
ˆ
1 x
(x − t)0 f 0 (t) dt = f (x) − f (a) = f (x) − T0 (x) = R0 (x).
0! a
So, the formula is true for n = 0.
Suppose the formula holds for n = k:
ˆ
1 x
Rk (x) = (x − t)k f (k+1) (t) dt, x ∈ I.
k! a
Then, by integration by parts, we get
ˆ x
1
(x − t)k+1 f (k+2) (t) dt
(k + 1)! a
x ˆ
1 1 x

= (x − t)k+1 f (k+1) (t) + (x − t)k f (k+1) (t) dt
(k + 1)! t=a k! a
f (k+1) (a)
= Rk (x) − (x − a)k+1
(k + 1)
f (k+1) (a)
= f (x) − Tk (x) − (x − a)k+1
(k + 1)
= f (x) − Tk+1 (x) = Rk+1 (x).
400 Series of Functions

So, the formula is also true for n = k + 1.

Hence, the remainder formula holds for all n ≥ 0.

7.16 Theorem: Lagrange remainder

If f (n+1) is continuous in an open interval I that contains a, then there is c between a and
x such that
f (n+1) (c)
Rn (x) = (x − a)n+1 .
(n + 1)!
This formula is called the Lagrange remainder of Taylor series.

Justification:
Let x ∈ I be a fixed number.
If x > a, on the interval [a, x], the function f (n+1) has its minimum and maximum at s1
and s2 :
m = f (n+1) (s1 ), M = f (n+1) (s2 ),
respectively. Since m ≤ f (n+1) (t) ≤ M on [a, x], we have
ˆ ˆ ˆ
m x 1 x M x
(x − t)n dt ≤ (x − t)n f (n+1) (t) dt ≤ (x − t)n dt,
n! a n! a n! a

so that ˆ x
m(x − a) n+1
≤ (n + 1) (x − t)n f (n+1) (t) dt ≤ M (x − a)n+1 ,
a
or equivalently, ˆ
(n + 1) x
m≤ (x − t)n f (n+1) (t) dt ≤ M.
(x − a)n+1 a

By the Intermediate Value Theorem, there is c ∈ [a, x], such that

ˆ x
(n + 1)
f (n+1) (c) = (x − t)n f (n+1) (t) dt.
(x − a)n+1 a

Hence, from the integral form of the remainder,

ˆ
1 x f (n+1) (c)
Rn (x) = (x − t)n f (n+1) (t) dt = (x − a)n+1 .
n! a (n + 1)!

If x < a, the argument is similar.

7.17 Theorem: remainder estimate

If f (n+1) (x) ≤ M in an open interval I that contains a, then

 

M
|Rn (x)| ≤ |x − a|n+1 , x ∈ I.
(n + 1)!
7.2 Represent Functions by Power Series 401

Hence, if f (n+1) (x) ≤ M for every n in I, then

 

∞
X f (k) (a)
f (x) = (x − a)k , x ∈ I.
k!
k=0

Justification:

If f (n+1) (x) ≤ M in I, by the Lagrange remainder,

 

(c)
 (n+1) 
f M
|Rn (x)| =  (x − a)n+1  n+1

 ≤ (n + 1)! |x − a| , x ∈ I.

(n + 1)!

We know that |x − a|n+1  (n + 1)! as n → ∞. Thus, if f (n+1) (x) ≤ M for every n in I,

 

then lim Rn (x) = 0. Hence, by the convergence theorem of the Taylor series,
n→∞

∞
X f (k) (a)
f (x) = (x − a)k , x ∈ I,
k!
k=0

7.18 Example: convergent Taylor series

1. Show that for all x ∈ R,

∞
X xk x2 x3 x4
ex = =1+x+ + + + ··· .
k! 2! 3! 4!
k=0

Solution Let f (x) = ex . For all n, since f (n) (x) = ex , we have f (n) (0) = e0 = 1. Thus,
n n
X f (k) (0) X xk
Tn (x) = xk =
k! k!
k=0 k=0

is well-defined for every n ≥ 0. Its Lagrange remainder is

f (n+1) (c) n+1
Rn (x) = f (x) − Tn (x) = x ,
(n + 1)!
where c is between 0 and x. For any fixed x,
 
(c) = ec ≤ e|x| ,
 (n+1) 
f
so that
e|x| e|x|
− |x|n+1 ≤ Rn (x) ≤ |x|n+1 .
(n + 1)! (n + 1)!
Since |x|n+1  (n + 1)! as n → ∞, by the Squeeze Theorem, lim Rn (x) = 0. Hence,
n→∞
by the convergence theorem of Taylor series, we have
∞
X xk
ex = , x ∈ R.
k!
k=0
402 Series of Functions

2. Show that for all x ∈ R,

∞
X (−1)k 2k+1 x3 x5 x7
sin x = x =x− + − 7 + ··· ,
(2k + 1)! 3! 5! x
k=0
∞
X (−1)k 2k x2 x4 x6
cos x = x =1− + − 6 + ··· .
(2k)! 2! 4! x
k=0

Solution Let f (x) = sin x. Directly computation gives, for k = 0, 1, 2, . . . ,

f (4k) (0) = 0, f (4k+1) (0) = 1, f (4k+2) (0) = 0, f (4k+3) (0) = −1.

Thus, for n ≥ 1,
n
X (−1)k 2k+1
T2n+1 (x) = T2n+2 (x) = x .
(2k + 1)!
k=0

The Lagrange remainder is

f (n+1) (c) n+1

Rn (x) = f (x) − Tn (x) = x ,
(n + 1)!

where c is between 0 and x. For any fixed x, it is easy to see that f (n) (x) is either
± sin x or ± cos x. Thus, we have
 
(c) ≤ 1,
 (n+1) 
f

so that
1 1
− |x|n+1 ≤ Rn (x) ≤ |x|n+1 .
(n + 1)! (n + 1)!
Since |x|n+1  (n + 1)! as n → ∞, by the Squeeze Theorem, lim Rn (x) = 0. Hence,
n→∞
by the convergence theorem of Taylor series, we have
∞
X (−1)k 2k+1 x3 x5 x7
sin x = x =x− + − 7 + ··· , x ∈ R.
(2k + 1)! 3! 5! x
k=0

In a similar manner, we can have

∞
X (−1)k x2 x4 x6
cos x = x2k = 1 − + − + ··· , x ∈ R.
(2k)! 2! 4! 6!
k=0
7.3 Properties of Power Series 403

7.3 Properties of Power Series

Radius of Convergence and Interval of Convergence

7.19 Example: power series

1. The polynomial Pn (x) = a0 + a1 x + a2 x2 + · · · + an xn is a power series centered at 0.

It has finite number of terms and is meaningful (convergent) for any x ∈ R.
∞
2. The power series k!xk converges only when x = 0 (it is a convention that 0! = 1).
P
k=0
k
1

In fact, for any x 6= 0, we know that, as k → ∞,  k!, so that 1  k!|x|k .
|x|
∞
Thus, lim k!xk 6= 0 for any x 6= 0. Hence, if x 6= 0, the series k!xk always diverges
P
k→∞ k=0
by the Divergence Test.
∞ xk
3. The power series √ converges for any x ∈ R.
P
k=0 k!
In fact, for any x ∈ R,
 
 xk+1 
 
 (k + 1)! 
p 
|x|
lim = lim √ = 0.
k+1
 k 
k→∞  x  k→∞
√ 
 k! 

∞ xk
By the Ratio Test, the series √ is absolutely convergent, so it is convergent.
P
k=0 k!
∞
4. The power series xk converges for x ∈ (−1, 1) and diverges for x ∈ (−∞, −1]∪[1, ∞).
P
k=0
In fact, for x ∈ (−1, 1), the power series is a convergent geometric series. For x ∈
(−∞, −1] ∪ [1, ∞), the limit of kth term xk is not zero: lim xk 6= 0. By the Divergence
k→∞
Test, the power series is divergent.
∞ (−1)k xk
5. The power series converges for x ∈ (−1, 1] and diverges for x ∈ (−∞, −1]∪
P
k=0 k + 1
(1, ∞).
In fact, for any fixed x ∈ R,
 (−1)k+1 xk+1 
 
 
 k+2  (k + 1)|x|
lim  = lim = |x|.
 (−1) x 
k k k+2

k→∞ k→∞
 
 k+1 
404 Series of Functions

By the Ratio Test, the series is absolutely convergent for |x| < 1 and divergent for
|x| > 1. Hence, the series is convergent for x ∈ (−1, 1) and divergent for x ∈ (−∞, −1)∪
(1, ∞).
∞ 1
For x = −1, the power series is the divergent harmonic series .
P
k=1 k
∞ (−1)k−1
For x = 1, the power series is the convergent alternating harmonic series .
P
k=1 k
∞
 k
k
6. The power series (x + 1)k converges for x ∈ (−2, 0) and diverges for
P
k=0 k+1
x ∈ (−∞, −2] ∪ [0, ∞).
In fact, for any fixed x ∈ R,
v 
k
k · |x + 1|
u
u k 
lim tk
(x + 1)k  = lim = |x + 1|.

 k+1  k→∞ k + 1

k→∞

By the Root Test, the series is absolutely convergent for |x + 1| < 1 and divergent
for |x + 1| > 1. Hence, the series is convergent for x ∈ (−2, 0) and divergent for
x ∈ (−∞, −2) ∪ (0, ∞).
For x = −2 or x = 0, since
 k 
 k  1 1
lim  (x + 1)k  = lim  k = 6= 0,
 
k→∞  k + 1  k→∞ 1 e
1+
k

by the Divergence Test, the series diverges.

7.20 Theorem: convergence of power series

∞
For a given power series ck (x − a)k there are only three possibilities:
P
k=0

1. The series converges only when x = a.

2. The series converges for all x.
3. There is a positive number R such that the series converges if |x − a| < R and diverges
if |x − a| > R.
7.3 Properties of Power Series 405

Definition: radius of convergence

The number R is called the radius of convergence of the power series.

By convention, the radius of convergence is R = 0 when the power series converges
only at the center a and R = ∞ when the power series converges for all x. The
interval of convergence of a power series is the interval that consists of all values
of x for which the series converges.
Remark: interval of convergence

When 0 < R < ∞, the theorem says nothing about the convergence of the
series at the endpoint points |x − a| = R (i.e., x = a ± R) of the interval.
Generally, the interval of convergence cannot be obtained from the radius of
convergence.

Justification:

The theorem is a consequence of the following claim: if ak rk converges, then ak xk

P P
absolutely converges for |x| < |r|.
In fact, for convergent series ak rk , we have lim ak rk = 0. Thus, there is K such that
P
k→∞
when k > K,
|ak xk | < 1.
For |x| < |r|, this implies that when k > K,
 x k  x k
|ak xk | = |ak rk | ·   ≤   .
   
r r
x P  x k
Since   < 1, the geometric series   converges. Then, by the comparison test, the
 
r r
∞ ∞
series ak x absolutely converges. Hence,
k
ak xk absolutely converges.
P P
k=K k=1

7.21 Method: determine the interval of convergence

∞
Let ck (x − a)k be a given power series.
P
k=0

1. Compute the radius of convergence R by using the Ratio Test (or sometimes the Root
Test) by
1 1
 
 c 
R=   = lim  k  or R = .
lim k |ck |
p
 ck+1  k→∞ ck+1
lim   k→∞
k→∞  ck 

2. The Ratio and Root Tests are inconclusive when x is an endpoint of the interval of
convergence. Apply some other test to determine whether the series is convergent or
406 Series of Functions

divergent at an endpoint.

7.22 Example: radius of convergence and interval of convergence

1. Determine the radius of convergence and the interval of convergence of the power series
∞ x2k
.
P
k=0 (k + 1)
k

Solution Since s
 x2k  1

lim  (k + 1)k  = |x| k→∞
k   2
lim = 0,
k→∞ k+1
by the Root Test, for all x, the power series absolutely converges, so it converges. Thus,
the radius of convergence is R = ∞ and the interval of convergence is (−∞, ∞).
2. Determine the radius of convergence and the interval of convergence of the power series
∞ k(x + 1)k
.
P
k=0 k+1
Solution

(a) Compute the radius of convergence by the Ratio Test.

Since
 (k + 1)(x + 1)k+1 
 
 
 k+2  (k + 1)2
lim = |x + 1| · lim = |x + 1|,
 k(x + 1) 
k k→∞ k(k + 2)
 
k→∞
 
 k+1 

by the Ratio Test, when |x + 1| < 1, the power series converges; when |x + 1| > 1,
it diverges. Thus, the radius of convergence is R = 1.
So, the series converges if −2 < x < 0; diverges if x < −2 or x > 0.
(b) Determine whether the power series is convergent at the endpoints.
∞ (−1)k k
At the left endpoint x = −2, the series is . Since
P
k=0 k + 1

 (−1)k k 
 
lim   = lim k = 1 6= 0,
k→∞ k + 1  k→∞ k + 1

(−1)k k
so that lim 6= 0. From the Divergence Test, the series is divergent .
k→∞ k + 1
∞ k
At the right endpoint x = 0, the series is . Again from the Divergence
P
k=0 k + 1
Test, it is divergent from the Divergence Test.
∞ k(x + 1)k
Hence, for the power series , the radius of convergence R = 1 and the
P
k=0 k+1
interval of convergence is (−2, 0).
7.3 Properties of Power Series 407

3. Determine the radius of convergence and the interval of convergence of the power series
∞ (x − 2)k
.
P
√
k=0 k+1
Solution

(a) Compute the radius of convergence by the Ratio Test.

Since
 (x − 2)k+1 
 
√  √
 k + 1 + 1 k+1
lim   = |x − 2| · lim √ = |x − 2|,
 (x − 2)  k+1+1
k→∞  k  k→∞
 √k + 1 

by the Ratio Test, when |x − 2| < 1, the power series converges; when |x − 2| > 1,
it diverges. Thus, the radius of convergence is R = 1.
So, the series converges if 1 < x < 3; diverges if x < 1 or x > 3.
(b) Determine whether the power series is convergent at the endpoints.
∞ (−1)k
At the left endpoint x = 1, the series is , which is an alternating series.
P
√
k=0 k+1
1
Let bk = √ . Then, {bk } is a positive decreasing sequence and lim bk = 0.
k+1 k→∞
By the Alternating Series Test, this series is convergent.
∞ 1
At the right endpoint x = 3, the series is . Since
P
√
k=0 k+1
1
√
lim k√+ 1 = 1,
k→∞ k
∞ 1
and the p-series √ is divergent (p = 12 ), by the Limit Comparison Test, the
P
k=0 k
∞ 1
series is divergent.
P
√
k=0 k+1
∞ (x − 2)k
Hence, for the power series , the radius of convergence R = 1 and the
P
√
k=0 k+1
interval of convergence is [1, 3).

7.23 MATLAB: radius of convergence and interval of convergence

Determine the radius of convergence and the interval of convergence of the power series
∞ k(x + 1)k
.
P
k=0 k+1
Solution
First, we compute the radius of convergence by the Ratio Test.
408 Series of Functions

>> syms k x
>> assume (k , ' positive ') % k >0
>> assumeAlso (k , ' integer ') % k is also an integer
>> a ( k ) = k *( x + 1) ^ k /( k + 1) ;
>> ratio_a = simplify ( abs ( a ( k + 1) / a ( k ) ) ) ;
>> L = limit ( ratio_a ,k , inf )

L =

abs ( x + 1)

According to the Ratio Test, when |x + 1| < 1, the power series is convergent; when |x + 1| >
1, it is divergent. Thus, the radius of convergence is R = 1.
Moreover, from the result above, we can know that the series converges if −2 < x < 0, and
diverges if x < −2 or x > 0. To determine the interval of convergence, we need to determine
whether the power series is convergent at the endpoints x = −2 and x = 0, which is shown
in the following:

>> a_left = subs ( a ( k ) ,x , -2) % when x = -2

a_left =

(( -1) ^ k * k ) /( k + 1)

>> lim_a_left = limit ( abs ( a_left ) ,k , inf )

lim_a_left =

>> a_right = subs ( a ( k ) ,x ,0) % when x = 0

a_right =

k /( k + 1)

>> lim_a_right = limit ( a_right ,k , inf )

lim_a_right =

k(−2 + 1)k k(0 + 1)k

We see from the results above that lim = 1 6= 0 and lim = 1 6= 0.
k→∞ k+1 k→∞ k+1
So, according to the Divergence Test, the power series is divergent at both endpoints x = −2
and x = 0. Therefore, the interval of convergence of this power series is (−2, 0).
7.3 Properties of Power Series 409

Operations on Power Series

7.24 Theorem: combining power series

Suppose ck xk and dk xk converge to f (x) and g(x), respectively, in an open interval

P P
I containing 0. Then

1. Sum and Difference: (ck ± dk )xk converges to f (x) ± g(x) in I, that is,
P

∞
X ∞
X ∞
X
(ck ± dk )xk = ck xk ± dk xk , x ∈ I.
k=0 k=0 k=0

(c0 dk + c1 dk−1 + · · · + ck−1 d1 + ck d0 ) xk converges to f (x) · g(x) in I,

P
2. Product:
that is,  
∞ ∞ ∞
k
! !
X X X X
 cj dk−j  xk = k
ck x dk x k
, x ∈ I.
k=0 j=0 k=0 k=0

3. Quotient: If d0 6= 0, then in a neighborhood of x = 0, the power series ak xk

P
f (x)
converges to , that is,
g(x)
X∞

∞
ck xk
X
ak xk = k=0
∞ ,
X
k=0 k
dk x
k=0

where  
k
c0 1  X
a0 = , ak = ck − dj ak−j  for k ≥ 1.
d0 d0 j=1

4. Composition: IfPh(x) = bxm , where m is a positive integer and b 6= 0 a real number,

the power series ck (h(x))k converges to f (h(x)) for x such that h(x) ∈ I.

7.25 Example: product and quotient of power series

1. The product can be found by a vertical set-up. For instance, for the power series
representations

ex = 1 + x + 21 x2 + 16 x3 + · · · , x ∈ R,
cos x = 1 − 1 2
2! x + 1 4
4! x − ··· , x ∈ R,
410 Series of Functions

to compute the Maclaurin series for ex cos x, we collect like terms just as for polynomials:

1 + x + 21 x2 + 16 x3 + · · ·

× 1 − 12 x2 + ···

1 + x + 21 x2 + 16 x3 + · · ·
− 12 x2 − 12 x3 − · · ·
..
+ .

1+x − 31 x3 − · · ·

Thus,
ex cos x = 1 + x − 13 x3 − · · · , x ∈ R.

2. The quotient can be found by a vertical set-up. For instance, for the power series
representations

sin x = x − 3! x +
1 3 1 5
5! x − · · · , x ∈ R,
cos x = 1 − 2! x +
1 2 1 4
4! x − · · · , x ∈ R,

sin x
to compute the Maclaurin series for tan x = , we use a procedure like long division:
cos x
x + 13 x3 + 2 5
15 x + ···
1 − 12 x2 + 1 4
24 x − ··· x − 16 x3 + 1
120 x
5
− ···
x− 1 3
2x + 1 5
24 x − ···
1 3
3x − 1 5
30 x + ···
1 3
3x − 1 5
6x + ···
2 5
15 x + ···
2 5
15 x + ···
···

Thus,
tan x = x + 13 x3 + 2 5
15 x + ··· , x ∈ (− π2 , π2 ).
The Maclaurin series for the tangent function tan x has a smaller interval of convergence
due to the fact that the denominator cos x becomes zero at x = ± π2 .

7.26 MATLAB: product and quotient of power series

∞ xk
1. Given two power series f (x) = (x ∈ (−∞, ∞)) and =
P
g(x)
k=0 k!
7.3 Properties of Power Series 411

∞ x2k
(−1)k (x ∈ (−∞, ∞)), find the power series for the product of these two power
P
k=0 (2k)!
series, i.e., f (x)g(x).
Solution

>> syms x k
>> assume (k , ' integer ') % k is an integer
>> a ( k ) = x ^ k / factorial ( k ) ;
>> b ( k ) = ( -1) ^ k * x ^(2* k ) / factorial (2* k ) ;
>> f ( x ) = symsum ( a ( k ) ,k ,0 , inf )

f(x) =

exp ( x )

>> g ( x ) = symsum ( b ( k ) ,k ,0 , inf )

g(x) =

cos ( x )

>> prod_term = f ( x ) * g ( x )

prod_term =

exp ( x ) * cos ( x )

>> prod_series = ...

taylor ( prod_term ,x , ' ExpansionPoint ' ,0 , ' Order ' ,6)

prod_series =

- x ^5/30 - x ^4/6 - x ^3/3 + x + 1

It is shown that, up to the order of 5, the product of these two series can be written as

f (x)g(x) = 1 + x − 13 x3 − 61 x4 − 1 5
30 x + ··· , x ∈ R.

∞ x2k+1
2. Given two power series f (x) = (−1)k (x ∈ (−∞, ∞)) and g(x) =
P
k=0 (2k + 1)!
∞ x2k
(−1)k (x ∈ (−∞, ∞)), find the power series for the quotient of these two
P
k=0 (2k)!
f (x)
power series, i.e., .
g(x)
Solution
412 Series of Functions

>> syms x k
>> assume (k , ' integer ') % k is an integer
>> a ( k ) = ( -1) ^ k * x ^(2* k + 1) / factorial (2* k + 1) ;
>> b ( k ) = ( -1) ^ k * x ^(2* k ) / factorial (2* k ) ;
>> f ( x ) = symsum ( a ( k ) ,k ,0 , inf )

f(x) =

sin ( x )

>> g ( x ) = symsum ( b ( k ) ,k ,0 , inf )

g(x) =

cos ( x )

>> quot_term = f ( x ) / g ( x )

quot_term =

sin ( x ) / cos ( x )

>> quot_series = ...

taylor ( quot_term ,x , ' ExpansionPoint ' ,0 , ' Order ' ,6)

quot_series =

(2* x ^5) /15 + x ^3/3 + x

It is shown that, up to the order of 5, the quotient of these two series can be written as

f (x)
= x + 13 x3 + 2 5
15 x + ··· .
g(x)

Note that the denominator g(x) = cos x becomes 0 when x = ± π2 . Hence, the power
series for the quotient is convergent when x ∈ − 2 , 2 .
π π

7.27 Theorem: differentiation and integration of power series

∞
X
If the power series ck (x − a)k has radius of convergence R > 0, then the function f
k=0
defined by
∞
X
f (x) = c0 + c1 (x − a) + c2 (x − a)2 + · · · = ck (x − a)k
k=0
is differentiable (and therefore continuous) on the interval (a − R, a + R) and
7.3 Properties of Power Series 413

∞
1. f 0 (x) = c1 + 2c2 (x − a) + 3c3 (x − a)2 + · · · = kck (x − a)k−1 .
P

ˆ k=1
(x − a)2 (x − a)3 ∞ (x − a)k+1
f (x) dx = C + c0 (x − a) + c1 + c2 + ··· = C + .
P
2. ck
2 3 k=0 k+1
The radii of convergence of the these power series are both still R.

7.28 Example: differentiation of power series

Find 1 + 3x2 + 5x4 + 7x6 + · · · .

Solution Finite geometric series gives
n
X 1 − xn+1
xk = 1 + x + x2 + x3 + · · · + xn = ,
1−x
k=0

so that
∞
1 X
= (−1)k xk = 1 + x + x2 + x3 + · · · , x ∈ (−1, 1).
1−x
k=0

Substituting x by −x (or making a composition) gives

∞
1 X
= (−1)k xk = 1 − x + x2 − x3 + · · · , x ∈ (−1, 1).
1+x
k=0

Thus, differentiating these two equalities, respectively, gives

0
1

= 1 + 2x + 3x2 + 4x3 + · · · , x ∈ (−1, 1);
1−x
0
1

= −1 + 2x − 3x2 + 4x3 − · · · , x ∈ (−1, 1).
1+x

The difference of the last two equalities is

0  0
1 1

2 1 + 3x2 + 5x4 + 7x6 + · · · =


−
1−x 1+x
1 1 2(1 + x2 )
= + = ,
(1 − x)2 (1 + x) 2 (1 − x2 )2

so that
1 + x2
1 + 3x2 + 5x4 + 7x6 + · · · = , x ∈ (−1, 1).
(1 − x2 )2

7.29 Example: integration of power series

1. Finite geometric series gives

n
X 1 − (−x)n+1
(−1)k xk = 1 − x + x2 − x3 + · · · + (−1)n xn = ,
1+x
k=0
414 Series of Functions

so that
∞
1 X
= (−1)k xk = 1 − x + x2 − x3 + · · · , for x ∈ (−1, 1).
1+x
k=0

Integrating the equality gives

x2 x3 x4
ln(1 + x) = C + x − + − + ··· , for x ∈ (−1, 1).
2 3 4
By putting x = 0 on both of the last equality, we get C = 0. Hence, for x ∈ (−1, 1),
we have
∞
X xk x2 x3 x4
ln(1 + x) = (−1)k−1 =x− + − + ··· .
k 2 3 4
k=1

At the endpoints x = −1, the Maclaurin series is the divergent harmonic series. At
the other endpoint x = 1, the Maclaurin series is the convergent alternating harmonic
series. We can get the Maclaurin series for the function ln(1 + x):
∞
X xk x2 x3 x4
ln(1 + x) = (−1)k−1 =x− + − + ··· , for x ∈ (−1, 1].
k 2 3 4
k=1

The following result can be applied to justify the equality at x = 1:

∞ ∞
If ck converges and f (x) = ck xk (−1 < x < 1), then lim f (x) = ck .
P P P
k=0 x→1− k=0

ˆ
dx
2. Notice that = tan−1 x + C. So, to find the Maclaurin series for the function
1 + x2
1
tan−1 x, we may start with the Maclaurin series for the function :
1+x
1
= 1 − x + x2 − x3 + · · · , x ∈ (−1, 1).
1+x
Substituting x by x2 gives
1
= 1 − x2 + x4 − x6 + · · · , x ∈ (−1, 1).
1 + x2
Integrating the equality gives

tan−1 x = C + x − 31 x3 + 15 x5 − 17 x7 + · · · , x ∈ (−1, 1).

By putting x = 0 on both of the last equality, we get C = 0. Hence, for x ∈ (−1, 1),
we have
∞
X x2k+1
tan−1 x = (−1)k = x − 13 x3 + 15 x5 − 71 x7 + · · · .
2k + 1
k=0
7.3 Properties of Power Series 415

At each of the endpoints x = ±1, the Maclaurin series is a convergent alternating series,
by Alternating Series Test. We can get the Maclaurin series for the function tan−1 x:
∞
X x2k+1
tan−1 x = (−1)k = x − 31 x3 + 51 x5 − 17 x7 + · · · , for x ∈ [−1, 1].
2k + 1
k=0

Again, the equality holds at both endpoints due to the same result stated above.

7.30 MATLAB: differentiation of power series

∞
!0 ∞
!0
X X
Find the power series for x k
− (−1) x
k k
, x ∈ (−1, 1).
k=0 k=0
Solution

>> syms x k
>> assume (k , ' integer ') % k is an integer
>> a(k) = x^k;
>> b ( k ) = ( -1) ^ k * x ^ k ;
>> term1 = symsum ( a ( k ) ,k ,0 , inf ) ;
>> term2 = symsum ( b ( k ) ,k ,0 , inf ) ;
>> term1_diff = diff ( term1 , x )

term1_diff =

piecewise (1 < x , 0 , abs ( x ) < 1 , 1/( x - 1) ^2)

>> term2_diff = diff ( term2 , x )

term2_diff =

piecewise ( x < -1 , 0 , abs ( x ) < 1 , -1/( x + 1) ^2)

>> Delta_term_diff = term1_diff - term2_diff ;

>> D e l t a _ t e r m _ d i f f _ s e r i e s = ...
taylor ( Delta_term_diff ,x , ' ExpansionPoint ' ,0 , ' Order ' ,7)

Delta_term_diff_series =

14* x ^6 + 10* x ^4 + 6* x ^2 + 2

∞
!0 ∞
!0
X X
It is shown that, up to the order of 6, the power series for x k
− (−1) x
k k
can
k=0 k=0
416 Series of Functions

be written as
∞
!0 ∞
!0
X X
x k
− (−1) x
k k
= 2 + 6x2 + 10x4 + 14x6 + · · · , x ∈ (−1, 1).
k=0 k=0

7.31 MATLAB: integration of power series

ˆ X
∞
Find the power series for (−1)k xk dx, x ∈ (−1, 1).
k=0
Solution

>> syms x k
>> assume (k , ' integer ') % k is an integer
>> a ( k ) = ( -1) ^ k * x ^ k ;
>> int_fun = symsum ( a ( k ) ,k ,0 , inf ) ;
>> int_term = int ( int_fun , x ) ;
>> int_series = taylor ( int_term ,x , ' ExpansionPoint ' ,0 , ' Order ' ,5)

int_series =

- x ^4/4 + x ^3/3 - x ^2/2 + x

ˆ X
∞
It is shown that, up to the order of 4, the power series for (−1)k xk dx, x ∈ (−1, 1), can
k=0
be written as ˆ X
∞
1 1 1
(−1)k xk dx = x − x2 + x3 − x4 + · · · .
2 3 4
k=0

7.32 Theorem: common Maclaurin series and their radii of convergence

1 ∞
= xk = 1 + x + x2 + x3 + · · · , x ∈ (−1, 1), R = 1.
P
1.
1 − x k=0
∞ xk x x2 x3
ex = =1+ + + + ··· , x ∈ (−∞, ∞), R = ∞.
P
2.
k=0 k! 1! 2! 3!
∞ x2k+1 x3 x5 x7
sin x = (−1)k =x− + + ··· , x ∈ (−∞, ∞), R = ∞.
P
3. −
k=0 (2k + 1)! 3! 5! 7!
∞ x2k x2 x4 x6
cos x = (−1)k =1− + + ··· , x ∈ (−∞, ∞), R = ∞.
P
4. −
k=0 (2k)! 2! 4! 6!
∞ x2k+1 x3 x5 x7
tan−1 x = (−1)k =x− + + ··· , x ∈ [−1, 1], R = 1.
P
5. −
k=0 2k + 1 3 5 7
7.3 Properties of Power Series 417

∞ xk x2 x3 x4
ln(1 + x) = (−1)k−1 =x− + + ··· , x ∈ (−1, 1], R = 1.
P
6. −
k=1 k 2 3 4
∞ α(α − 1) 2 α(α − 1)(α − 2) 3
 
α k
(1 + x)α = x = 1 + αx + x + x + · · · , at least for
P
7.
k=0 k 2! 3!
x ∈ (−1, 1), R = 1.

7.33 Example: binomial expansion

√
Expand the function y = a2 + x2 (a > 0) as its Maclaurin series.
Solution Write the function as
1
y = a 1 + (x/a)2 2 .


By the binomial expansion

∞  
X α
(1 + x)α = xk , x ∈ (−1, 1).
k
k=0

we get, for −a < x < a,

p ∞ 1
X
a2 + x2 = a 2 (x/a)2k
k
k=0

2(2 − 1)  x 4 2 ( 2 − 1)( 2 − 2)
 1  x 2 1 1 1 1 1  x 6 
=a 1+ 2
· + · + · + ···
1! a 2! a 3! a
2(2 − 1) −3 4 ( − 1)( 12 − 2) −5 6
1 1 1 1 1
=a+ 2
· a−1 x2 + ·a x + 2 2 · a x + ··· .
1! 2! 3!

7.34 Theorem: theorem of analytic function

Suppose f has a power series representation (expansion) at a:

∞
X
f (x) = ck (x − a)k , |x − a| < R,
k=0

where R > 0 or R = ∞. Then

1. The coefficients are given by the formula
f (k) (a)
ck = , k ≥ 0.
k!
f (x) − Tn (x)
2. lim = 0, or equivalently,
x→a (x − a)n
n
X f (k) (a)
f (x) = Tn (x) + o(|x − a|n ) = (x − a)k + o(|x − a|n ), as x → a.
k!
k=0
418 Series of Functions

3. f (x) = lim Tn (x) for |x − a| < R.

n→∞

Justification:

1. We know that it is legitimate to differentiate the infinite series term by term at any
number x satisfying |x − a| < R. Thus, for each k, k = 0, 1, 2, . . . , we differentiate and
substitute x = a to have

f (k) (a) = 2 · 3 · 4 · · · k · ck = k! · ck ,

so that
f (k) (a)
ck = , k ≥ 0.
k!
n
2. By part 1, we get Tn (x) = ck (x − a)k . Thus,
P
k=0

∞
(x − a)n+1 · cn+k (x − a)k
P
f (x) − Tn (x) k=0
lim = lim = 0.
x→a (x − a)n x→a (x − a)n

n
3. By part 1, we get Tn (x) = ck (x − a)k . Thus,
P
k=0

∞
X
f (x) = ck (x − a)k = lim Tn (x).
n→∞
k=0

7.35 Example: find higher derivative

For f (x) = tan−1 x, find f (2019) (0).

Solution We know that the function f (x) = tan−1 x has the following Maclaurin series:
∞
X x2k+1
tan−1 x = (−1)k , x ∈ [−1, 1].
2k + 1
k=0

By the uniqueness of Taylor expansion, we have

f (2k+1) (0) (−1)k

= , k ≥ 0.
(2k + 1)! 2k + 1

Thus,
f (2019) (0) = (−1)1009 · 2018! = −2018!.
7.3 Properties of Power Series 419

7.36 MATLAB: find higher derivative

For f (x) = tan−1 x, find f (2019) (0).

Solution First, the Maclaurin series of f (x) = tan−1 x can be written as
∞
X
f (x) = tan−1 x = ck x k , x < R.
k=0

According to the theorem of analytic function, the coefficient of the term with order k is
given by
f (k) (0)
ck = , k ≥ 0.
k!
Hence, we have f (2019) (0) = 2019!c2019 . To calculate f (2019) (0), we thus only need to calculate
c2019 , which is shown as follows:

>> syms x
>> f ( x ) = atan ( x ) ;
>> f_series = taylor ( f ( x ) ,x , ' ExpansionPoint ' ,0 , ' Order ' ,2019 ...
+ 1) ; % expand f ( x ) up to order 2019
>> c_array = coeffs ( f_series ) ; % find all the coefficients ...
of the series , the coefficient of the term with the ...
highest order is the last element of the array
>> ind = length ( c_array ) ; % the location of the coefficient ...
for x ^2019 in the array
>> c = c_array ( ind ) % the value of the last coefficient , ...
i . e . , the one for x ^2019

c =

-1/2019

In the MATLAB code above, the command coeffs can return all the coefficients of a poly-
nomial, and the command length can return the length of a vector.
We see from the result above that c2019 = − 2019
1
. Therefore, we have

f (2019) (0) = 2019!c2019 = −2018!.

420 Series of Functions

7.4 Applications of Taylor Series

Limits by Taylor Series

7.37 Example: limit by Taylor series

x3 − 6x cos(x2 ) + 6 sin x
Find lim .
x→0 x5
Solution The limit has the indeterminate form 00 . Although we may apply l’Hospital’s
Rule, alternatively we use the technique of Taylor series. In fact, near x = 0, we have

sin x = x − 3! x + 5! x + · · ·
1 3 1 5
,
cos(x ) = 1 −
2
2! (x ) + · · · ,
1 2 2

so that, as x → 0,

sin x = x − 61 x3 + 1
120 x
5
+ o(|x|5 ),
cos(x2 ) = 1 − 21 x4 + o(|x| ),
4

Hence,

x3 − 6x cos(x2 ) + 6 sin x
lim 5
 x 1 4
x→0
x − 6x 1 − 2 x + o(|x|4 ) + 6 x − 61 x3 + 1
+ o(|x|5 )
3
  5

120 x
= lim
x→0 x5
x + o(|x| )
61 5 5
= lim 20 = 61
20 .
x→0 x5

7.38 Example: limit by Taylor series

h 2
i
Find lim x4 sin x−4 − x2 e−1/x + x2 .


x→∞
Solution If making the substitution x = t−1 , we see that x → ∞ ⇐⇒ t → 0+ . Thus,
h 2
i h 2
i
lim x4 sin x−4 − x2 e−1/x + x2 = lim+ t−4 sin(t4 ) − t−2 e−t + t−2


x→∞ t→0
2
sin(t4 ) − t2 e−t + t2
= lim+ .
t→0 t4

The last limit has the indeterminate form 00 . We use the technique of Taylor series. In fact,
near t = 0, we have

sin(t4 ) = t4 + · · · ,
2
e−t = 1 + 1! (−t )
1 2
+ ··· ,
7.4 Applications of Taylor Series 421

so that, as t → 0+ ,

sin(t4 ) = t4 + o(|t|4 ),
2
e−t = 1 − t2 + o(|t|2 ),

Hence,
2
sin(t4 ) − t2 e−t + t2
lim+
t→0 t4
t + o(|t|4 ) − t2 1 − t2 + o(|t|2 ) + t2
4   
= lim+
t→0 t4
2t + o(|t| )
4 4
= lim+ = 2.
t→0 t4
Therefore, h i
2
lim x4 sin x−4 − x2 e−1/x + x2 = 2.


x→∞

Approximation of Definite Integrals

7.39 Example: approximation of a definite integral
ˆ 1
2
Approximate e−x dx within 10−4 of the exact value.
0
Solution We begin with the Maclaurin series
∞
X xk x x2 x3
ex = =1+ + + + ··· , x ∈ (−∞, ∞).
k! 1! 2! 3!
k=0

By replacing x with −x2 , we have

∞
2 X (−x2 )k x2 x4 x6
e−x = =1− + − + ··· , x ∈ (−∞, ∞).
k! 1! 2! 3!
k=0

Thus,
ˆ 1 1
x3 x5 x7

2
e−x dx = + + · · · 

x− −
0 3 · 1! 5 · 2! 7 · 3! x=0
∞
1 1 1 X (−1)k
=1− + − + ··· = .
3 · 1! 5 · 2! 7 · 3! (2k + 1) · k!
k=0

This shows that the definite integral can be expressed as a convergent alternating series, with
1
bk = satisfying:
(2k + 1) · k!
0 < bk+1 < bk , lim bk = 0.
k→∞
422 Series of Functions

Using the bound on the remainder in the Alternating Series Test, we have
ˆ 
n
 1
−x2
X (−1)k  1
|Rn | =  e dx − < .

 0 (2k + 1) · k!  (2n + 3) · (n + 1)!
k=0

1
Thus, to ensure that |Rn | < 10−4 , we need to choose n so that < 10−4 , or
(2n + 3) · (n + 1)!
6 (−1)k
n > 6. Hence, the finite series gives an approximation that is within 10−4
P
k=0 (2k + 1) · k!
of the exact value of the series.
Using the bounds on the series in the Alternating Series Test, we have
ˆ 1
1 2 1
s6 − < e−x dx < s6 + ,
15 · 7! 0 15 · 7!
6 (−1)k
where s6 = . By using a calculator, we have s6 ≈ 0.74684, which has 5th
P
k=0 (2k + 1) · k!
decimal place accuracy. Thus,

0.74683 < s < 0.74685.

ˆ 1
2
Taking the average of these two bounds as our approximation of e−x dx, we find that
ˆ 1 0
−x2
e dx ≈ 0.74684.
0 ˆ 1
2
For comparison, e−x dx = 0.7468241324 · · · . We see that our approximation indeed is
0
within 10−4 of the exact value of the series.

Solving Differential Equations

7.40 Example: solving differential equation by power series

Solve the differential equation y 0 (t) = y(t)−1, subject to the initial value condition y(0) = 3.
Solution We assume the solution has a power series solution of the form
∞
X
y(t) = ck tk = c0 + c1 t + c2 t2 + c3 t3 + · · · ,
k=0

where the coefficients {ck } will be determined.

Since y(0) = 3, we get c0 = 3.
The other coefficients can be determined one by one. In fact, substituting the series solution
into the differential equation gives

c1 + 2c2 t + 3c3 t2 + 4c4 t3 + · · · = (c0 + c1 t + c2 t2 + c3 t3 + · · · ) − 1.

7.4 Applications of Taylor Series 423

Putting t = 0 on both sides of the equation gives

c1 = c0 − 1 =⇒ c1 = 2. since c0 = 3

We differentiate both sides of the above equation to have

2c2 + 3 · 2c3 t + 4 · 3c4 t2 + · · · = (c1 + 2c2 t + 3c3 t2 + · · · ) − 0.

Again, putting t = 0 on both sides of the last equation gives

2c2 = c1 =⇒ c2 = 1. since c1 = 2

We repeat the above process by differentiating the equation once more then putting t = 0 on
the resulted equation. Recursively, we get:
1 1 1
c3 = , c4 = , c5 = , ....
3 4·3 5·4·3
Thus, we obtain an “formal” solution:

t3 t4 t5
y(t) = 3 + 2t + t2 + + + + ···
3 4·3 5·4·3
The formal solution can be re-written as (formally!)

t2 t3 t4 t5
 
y(t) = 1 + 2 1 + t + + + + + · · · .
2! 3! 4! 5!

We recognize that the right-hand side is actually the Maclaurin series of the function 1 + 2et .
Finally, we can verify that the formally obtained solution is indeed the true solution of the
original problem, since

y 0 (t) = (1 + 2et )0 = 2et = (1 + 2et ) − 1 = y(t) − 1;

y(0) = (1 + 2et ) = 3.

t=0
424 Summary

Summary

Definition: basic concepts

• Let f be a function with f 0 , f 00 , . . . , and f (n) defined at a. The nth-order Taylor

polynomial for f with its center at a is

f 00 (a) f (n) (a)

Tn (x) = f (a) + f 0 (a)(x − a) + (x − a)2 + · · · + (x − a)n .
2! n!

• Let Tn be the Taylor polynomial of order n for f at a. The term Rn (x) = f (x) − Tn (x)
is called the remainder of the Taylor series. The latter refers the following “formal”
power series of f :

f 00 (a) f (n) (a)

f (a) + f 0 (a)(x − a) + (x − a)2 + · · · + (x − a)n + · · · .
2! n!
This power series is also called Taylor expansion of f at a. A Taylor series centered
at 0 is called a Maclaurin series.

• A power series in (x − a) or a power series centered at a or a power series

about a is a series of the form
∞
X
ck (x − a)k = c0 + c1 (x − a) + c2 (x − a)2 + · · · .
k=0

Theorem: uniqueness of Taylor polynomials

Suppose f is a function with f 0 , f 00 , . . . , and f (n) defined at a. If Pn is a polynomial of

degree n satisfying

Pn (a) = f (a), Pn0 (a) = f 0 (a), Pn00 (a) = f 00 (a), ..., Pn(n) (a) = f (n) (a),

Then
f 00 (a) f (n) (a)
Pn (x) ≡ Tn (x) = f (a) + f 0 (a)(x − a) + (x − a)2 + · · · + (x − a)n .
2! n!

Theorem: asymptotic approximation of Taylor polynomials

f (x) − Tn (x)
If f (n) (a) exists, then lim = 0, or equivalently, by using little-o notation,
x→a (x − a)n
n
X f (k) (a)
f (x) = Tn (x) + o(|x − a|n ) = (x − a)k + o(|x − a|n ), as x → a.
k!
k=0
 425

Theorem: convergence theorem of Taylor series

If lim Rn (x) = 0 in an open interval I that contains a, then

n→∞

∞
X f (k) (a)
f (x) = lim Tn (x) = (x − a)k , x ∈ I.
n→∞ k!
k=0

Theorem: integral form of the remainder

If f (n+1) is continuous in an open interval I that contains a, then

ˆ
1 x
Rn (x) = (x − t)n f (n+1) (t) dt, x ∈ I.
n! a

Theorem: Lagrange remainder

If f (n+1) is continuous in an open interval I that contains a, then there is c between a and
x such that
f (n+1) (c)
Rn (x) = (x − a)n+1 .
(n + 1)!
This formula is called the Lagrange remainder of Taylor series.

Theorem: remainder estimate

If f (n+1) (x) ≤ M in an open interval I that contains a, then

 

M
|Rn (x)| ≤ |x − a|n+1 , x ∈ I.
(n + 1)!

Hence, if f (n+1) (x) ≤ M for every n in I, then

 

∞
X f (k) (a)
f (x) = (x − a)k , x ∈ I.
k!
k=0

Theorem: convergence of power series

∞
For a given power series ck (x − a)k there are only three possibilities:
P
k=0

1. The series converges only when x = a.

2. The series converges for all x.
3. There is a positive number R such that the series converges if |x − a| < R and diverges
if |x − a| > R.
426 Summary

Definition: radius of convergence

The number R is called the radius of convergence of the power series.

By convention, the radius of convergence is R = 0 when the power series converges
only at the center a and R = ∞ when the power series converges for all x. The
interval of convergence of a power series is the interval that consists of all values
of x for which the series converges.

Method: determine the interval of convergence

∞
Let ck (x − a)k be a given power series.
P
k=0

1. Compute the radius of convergence R by using the Ratio Test (or sometimes the Root
Test) by
1 1
 
 c 
R=   = lim  k  or R = .
lim
p
 ck+1  k→∞ ck+1 k
|ck |
lim   k→∞
k→∞  ck 

2. The Ratio and Root Tests are inconclusive when x is an endpoint of the interval of
convergence. Apply some other test to determine whether the series is convergent or
divergent at an endpoint.

Theorem: combining power series

Suppose ck xk and dk xk converge to f (x) and g(x), respectively, in an open interval

P P
I containing 0. Then

1. Sum and Difference: (ck ± dk )xk converges to f (x) ± g(x) in I, that is,
P

∞
X ∞
X ∞
X
(ck ± dk )xk = ck xk ± dk xk , x ∈ I.
k=0 k=0 k=0

(c0 dk + c1 dk−1 + · · · + ck−1 d1 + ck d0 ) xk converges to f (x) · g(x) in I,

P
2. Product:
that is,  
∞ ∞ ∞
k
! !
X X X X
 cj dk−j  xk = ck x k dk xk , x ∈ I.
k=0 j=0 k=0 k=0

3. Quotient: If d0 6= 0, then in a neighborhood of x = 0, the power series ak xk

P
 427

f (x)
converges to , that is,
g(x)
∞
X
∞
ck xk
X k=0
ak xk = ∞ ,
X
k=0 k
dk x
k=0

where  
k
c0 1  X
a0 = , ak = ck − dj ak−j  for k ≥ 1.
d0 d0 j=1

4. Composition: IfPh(x) = bxm , where m is a positive integer and b 6= 0 a real number,

the power series ck (h(x))k converges to f (h(x)) for x such that h(x) ∈ I.

Theorem: differentiation and integration of power series

∞
X
If the power series ck (x − a)k has radius of convergence R > 0, then the function f
k=0
defined by
∞
X
f (x) = c0 + c1 (x − a) + c2 (x − a)2 + · · · = ck (x − a)k
k=0

is differentiable (and therefore continuous) on the interval (a − R, a + R) and

∞
1. f 0 (x) = c1 + 2c2 (x − a) + 3c3 (x − a)2 + · · · = kck (x − a)k−1 .
P

ˆ k=1
(x − a)2 (x − a)3 ∞ (x − a)k+1
f (x) dx = C + c0 (x − a) + c1 + c2 + ··· = C + .
P
2. ck
2 3 k=0 k+1
The radii of convergence of the these power series are both still R.

Theorem: common Maclaurin series and their radii of convergence

1 ∞
= xk = 1 + x + x2 + x3 + · · · , x ∈ (−1, 1), R = 1.
P
1.
1 − x k=0
∞ xk x x2 x3
ex = =1+ + + + ··· , x ∈ (−∞, ∞), R = ∞.
P
2.
k=0 k! 1! 2! 3!
∞ x2k+1 x3 x5 x7
sin x = (−1)k =x− + + ··· , x ∈ (−∞, ∞), R = ∞.
P
3. −
k=0 (2k + 1)! 3! 5! 7!
∞ x2k x2 x4 x6
cos x = (−1)k =1− + + ··· , x ∈ (−∞, ∞), R = ∞.
P
4. −
k=0 (2k)! 2! 4! 6!
∞ x2k+1 x3 x5 x7
tan−1 x = (−1)k =x− + + ··· , x ∈ [−1, 1], R = 1.
P
5. −
k=0 2k + 1 3 5 7
428 Summary

∞ xk x2 x3 x4
ln(1 + x) = (−1)k−1 =x− + + ··· , x ∈ (−1, 1], R = 1.
P
6. −
k=1 k 2 3 4
∞ α(α − 1) 2 α(α − 1)(α − 2) 3
 
α k
(1 + x)α = x = 1 + αx + x + x + · · · , at least for
P
7.
k=0 k 2! 3!
x ∈ (−1, 1), R = 1.

Theorem: theorem of analytic function

Suppose f has a power series representation (expansion) at a:

∞
X
f (x) = ck (x − a)k , |x − a| < R,
k=0

where R > 0 or R = ∞. Then

1. The coefficients are given by the formula

f (k) (a)
ck = , k ≥ 0.
k!

f (x) − Tn (x)
2. lim = 0, or equivalently,
x→a (x − a)n
n
X f (k) (a)
f (x) = Tn (x) + o(|x − a|n ) = (x − a)k + o(|x − a|n ), as x → a.
k!
k=0

3. f (x) = lim Tn (x) for |x − a| < R.

n→∞
Appendix: Introduction to MATLAB

7.41 MATLAB: What is MATLAB?

MATLAB is short for “Matrix Laboratory”. It is a multi-paradigm numerical computing
environment and proprietary programming language developed by MathWorks. MATLAB
allows matrix manipulations, plotting of functions and data, implementation of algorithms,
creation of user interfaces, and interfacing with programs written in other languages, including
C, C++, C#, Java, Fortran and Python. Although MATLAB is intended primarily for
numerical computing, an optional toolbox uses the MuPAD symbolic engine, allowing access
to symbolic computing abilities [Wikipedia].

7.42 MATLAB: Command Window

The command window is where the commands of MATLAB are typed in and calculations
are carried out. In particular, in the command window, the MATLAB commands are typed
in after two arrows >>, as shown in the following figure.
430 Appendix: Introduction to MATLAB

After typing in the command and pressing “Enter” (in Windows) or “return” (in Mac), the
calculation will be carried out.

7.43 MATLAB: Current Folder and Workspace

Next to the command window, there are two other windows, namely, “Current Folder”
window and “Workspace” window, as shown in the following figure.

In particular, the “Current Folder” window on the left enables us to view, create, open,
move, and rename files and folders in the current folder. For the “Workspace” window on
the right, it enables us to view and interactively manage each variable or object used in
MATLAB. For example, we can edit the content a variable directly in the workspace by right
clicking the scalar and selecting the “Edit Value”, as shown in the following figure.
 431

Note that if we click the “Close” option under the triangle menu, the “Current Folder” or
“Workspace” window will be closed.

7.44 MATLAB: M-File

The MATLAB commands can be put into one file called m-file (or script file), and then
we can instruct MATLAB to execute these commands.
To create a m-file, we simply need to click the “New” option from the “HOME” menu and
select “Script”, as shown in the following figure.
432 Appendix: Introduction to MATLAB

To open an existing m-file, we simply need to click the “Open” option from the “HOME”
menu and select the file required from the directory, as shown in the following figure.

To save an m-file, we can click the “Save” option from the “EDITOR” menu, as shown in
the following figure, and choose a suitable destination from the directory for the m-file. We
can also save an existing m-file to another destination by clicking the “Save As" option, and
then choose a suitable destination from the directory. Note that the m-file should be saved
with a “.m” extension.
 433

In the m-file, unlike in the command window, there are no double arrows after which the
command should be typed in. Moreover, all the commands in the m-file will not be executed
before clicking the “Run” button in the “EDITOR” menu which is shown in the figure above.
After clicking the “Run” button, the result will be shown if there is no error; otherwise, the
error information will be displayed in the command window.

7.45 MATLAB: Assignment of A Value and Semicolon

To explain how to assign a value and when to use the semicolon in MATLAB, we first
provide some examples as follows. Specifically, the following examples show the calculations
of cos(1) + 1 in the command window.

>> cos (1) + 1

ans =

1.5403

>> r1 = cos (1) + 1

r1 =

1.5403

>> r2 = cos (1) + 1;

>> r2

r2 =
434 Appendix: Introduction to MATLAB

1.5403

We note that “=” means “assign a value” in MATLAB (The symbol “==” represents “equal”
in MATLAB). If no variable is specified, then the answer of cos(1)+1 is assigned to the default
variable ans, as can be seen from the first command of the MATLAB code above. Note that
in MATLAB, the number “1” in cos(1) is in radians (not degrees). If a certain variable is
specified, e.g., r1 in the second command, then the value of cos(1) + 1 is assigned to r1 after
executing this command.
Next, we explain the use of the semicolon (i.e., “;”). Specifically, if no semicolon is used
after a command, then the result of this command will be displayed directly after executing
this command; while if a semicolon is used after a command, then the result of this command
will not be displayed, as can be seen from the last two commands of the MATLAB code above.

7.46 MATLAB: Comments

In MATLAB, comments are written after %, and MATLAB does not attempt to compute
the commenst. For example, in the MATLAB code below, “Change the axis limits so
that both the x-axis and the y-axis range from -3 to 3” are the comments.

7.47 MATLAB: Help

In MATLAB, there is extensive documentation explaining the use of all its built-in func-
tions. To check the use of a function, we simply need to type in help function name after
>> in the command window. For example, in the following figure, two commands, namely,
help cos and help plot, are typed in to check the use of cos and plot, respectively.
 435

7.48 MATLAB: Vector

The most common variable type in MATLAB is matrix. In Calculus course, we will need
to frequently use the one-dimensional matrices, i.e., vectors, in MATLAB. In the following,
we will introduce two ways to define a vector in MATLAB.
1. initial:increment:terminator
The first way to define a vector is by specifying the initial value, the increment (or
step), and the terminator of the vector.

>> vec1 = 1:2:9

vec1 =

1 3 5 7 9

>> vec2 = 1:2:10

vec2 =

1 3 5 7 9

>> vec = 1:5

vec =

1 2 3 4 5
436 Appendix: Introduction to MATLAB

The MATLAB code above defines two vectors named vec1 and vec2, respectively.
There are 5 elements in both of the vectors, namely, 1, 3, 5, 7 and 9. In particular,
vec1 starts at 1 (i.e., the initial value is 1), the increment from the previous value (i.e.,
the step) is 2, and it stops at 9.
However, sometimes the vector does not stop at the terminator specified in the com-
mand. With the value of each element changing from the value of the previous ele-
ment by adding an increment, the vector stops at a value which is not exceeding the
terminator but closest to the terminator. For example, vec2 stops at 9 instead of 10.
Note that if the increment is not specified in the command, the default increment is
1, as can be seen from the third command above.
2. linspace

Using the command linspace is another way of defining a vector.

>> vec3 = linspace (1 ,9 ,5)

vec3 =

1 3 5 7 9

>> vec4 = linspace (1 ,10 ,5)

vec4 =

1.0000 3.2500 5.5000 7.7500 10.0000

From the MATLAB code above, we see that the command linspace(x1,x2,N) defines
an N -element vector starting at x1 and stopping at x2. In this way, the increment
x2 − x1
equals to .
N −1
Note that if we use the command linspace(x1,x2), then a 100-element vector with
initial value x1 and ending value x2 will be generated.
 437

7.49 MATLAB: Basic Computations

Some basic matrix operators are listed in the following.

Operator Meaning
+ Matrix addition
- Matrix subtraction
* Matrix multiplication
.* Element-by-element multiplication
./ Element-by-element division
ˆ Matrix exponentiation
.ˆ Element-by-element exponentiation

We see from the above table that some operators for matrices are the same with the ones
for scalars, but some are different and require an additional . before the scalar operator.

7.50 MATLAB: Plot

MATLAB has powerful graph-plotting features. For example, the command plot can be
used to generate a graph based on two vectors. The following MATLAB code generates the
figure of y = x2 by using the command plot.

>> x = -1:0.01:1;
>> y = x .^2;
>> figure % create a figure window
>> plot (x ,y , 'r ') % graph is in red
>> xlabel ( 'x ') % X - axis label
>> ylabel ( 'y = x ^2 ') % Y - axis label
>> grid % add grid lines

The figure generated is shown in the following:

438 Appendix: Introduction to MATLAB

Note that by clicking the third icon, we can save the figure. Moreover, after clicking the fifth
icon, i.e., the arrow, we can choose the “Property Editor” option from the “View” menu, as
highlighted in the following figure.

Then, a “Property Editor” window will appear as follows, and we can edit the figure, e.g.,
change the font, etc.
439
440 Appendix: Introduction to MATLAB
Index

e, 34 under a parametric curve, 273

p-series, 364 MATLAB, 274
arithmetic rules of limits, 13
absolute convergence, 382 asymptote
absolute extreme, 97 horizontal, 19
absolute maximum, 97 slant, 22, 147
absolute minimum vertical, 25
MATLAB, 105
absolution minimum, 97 cancellation equations, 38
alternating harmonic series, 379 center of mass, 323
MATLAB, 380 along a curve, 315
Alternating Series Test discrete system
MATLAB, 380 1d, 313
remainder estimate, 380 2d, 314
antiderivative, 206 parametric curve, 318
arc length, 288 centroid, 322
MATLAB, 291 of a curve
parametric curve MATLAB, 317
MATLAB, 294 parametric curve, 318
polar curve MATLAB, 321
MATLAB, 308 planar region between two curves, 322
arc length function, 288 MATLAB, 325
area chain rule, 68
between two curves Comparison Test, 369
general case, 267 for improper integrals, 254
bounded by two curves, 265 composite function, 31
along the y-axis, 268 MATLAB, 31
polar curve, 309 concave down, 130
MATLAB, 312 concave up, 130
polar region, 309 concavity, 130
442 Index

concavity test, 130 function, 1

conditional convergence, 382 higher derivatives, 90
constant function, 112 improper integral
continuity over infinite interval, 249
composite functions, 31 unbounded integrand, 251
elementary functions, 6, 47 indefinite integral, 205
inverse function, 39 infinite limit, 25
continuous function, 27 integral, 188
in an interval, 27 interior point, 99, 101
continuous functions inverse function, 37
extreme value theorem, 102 inverse trigonometric functions, 44
convergence limit, 10
improper integral limit at infinities, 19
over infinite interval, 249 local extreme, 97
unbounded integrand, 251 local maximum, 97
series local minimum, 97
geometric series, 350 logarithmic function, 40
the p-series, 364 monotone, 107
convergence test odd function, 141
sequence one-sided limits, 15
Comparison Test, 369 one-to-one, 36
Integral Test, 359 periodic function, 141
Limit Comparison Test, 370 polar coordinates, 155
Ratio Test, 372 polar curve, 155
Root Test, 376 series
convergent series, 350 absolute convergence, 382
critical point, 100 conditional convergence, 382
curve kketching, 142 convergence, 350
cycloid, 152 divergence, 350
the number e, 34
definite integral, 188 trigonometric functions, 5
MATLAB, 191 vector
polynomial MATLAB, 435
MATLAB, 191 defnition
definition exponentiation, 4
absolute extreme, 97 delta function, 12
absolute maximum, 97 derivative, 56
absolute minimum, 97 arithmetic rules, 61
composite function, 31 chain rule, 68
concave down, 130 exponential functions, 75
concave up, 130 Fermat’s theorem, 99, 101
continuous function, 27 higher, 90
critical point, 100 inverse function, 71
definite integral, 188 inverse trigonometric functions, 73
derivative, 56 logarithmic functions, 77
differential, 89 MATLAB, 58
even function, 141 monotone test, 111
exponential function, 4 notations, 57
 443

trigonometric functions, 65 parametric curve, 302

derivative of inverse function, 71 volume by cylindrical shells, 281
derivatives of exponential functions, 75 function, 1
derivatives of inverse trigonometric functions, absolute extreme, 97
73 absolute maximum, 97
derivatives of logarithmic functions, 77 absolute minimum, 97
derivatives of trigonometric functions, 65 arc length, 288
determine the interval of convergence composite
MATLAB, 407 MATLAB, 31
determine the radius of convergence composition, 31
MATLAB, 407 concave down, 130
differentiable, 56 concave up, 130
differential, 89 continuous, 27
differentiation of power series critical point, 100
MATLAB, 415 decreasing, 107
Dirac delta function, 12 differentiable, 56
discontinuous, 27 Dirac delta, 12
divergent series, 350 discontinuous, 27
domain, 1 domain, 1
even, 141
error bound exponential
midpoint rule, 240 MATLAB, 34
Simpson’s rule, 242 increasing, 107
trapezoidal rule, 240 inverse, 37
Euler’s formula, 7 MATLAB, 39
even function, 141 inverse trigonometric, 44
exponential function local extreme, 97
MATLAB, 34 local maximum, 97
extreme point, 97 local minimum, 97
extreme value theorem, 102 logarithmic, 40
MATLAB, 41
Fermat’s Theorem, 99, 101 monotone, 107
first derivative test, 114 natural domain, 2
MATLAB, 119 odd, 141
formula periodic, 141
arc length, 288 range, 1
parametric curve, 292 transcendental, 4
polar curve, 309 trigonometric, 5
area MATLAB, 46
bounded by two curves, 265 Fundamental Theorem of Calculus, 201
bounded by two polar curves, 309
center of mass geometric sequences, 345
along a curve, 315 geometric series, 350
parametric curve, 318 graph of the inverse function, 39
centroid growth rates of sequences, 343
parametric curve, 318 Guidelines for Curve Sketching, 142
planar region between two curves, 322
surface area of revolution, 297 harmonic series
444 Index

alternating, 379 Intermediate Value Theorem, 32

MATLAB, 380 inverse function, 37
divergence, 358 cancellation equations, 38
by Integral Test, 362 derivative, 71
higher derivatives, 90 find, 38
horizontal asymptote, 19 graph, 39
horizontal line test, 36 MATLAB, 39
trigonometric
implicit differentiation, 81 MATLAB, 46
MATLAB, 82 inverse trigonometric functions, 44
improper integral MATLAB, 46
over infinite interval, 249
convergence, 249 l’Hospital’s Rule, 134
MATLAB, 254 Lagrange remainder, 400
unbounded integrand, 251 laws of exponents, 4
convergence, 251 laws of logarithms, 40
MATLAB, 255 left point rule, 237
increasing/decreasing test, 111 MATLAB, 243
indefinite integral, 205 left-hand limit, 15
linearity, 206 MATLAB, 17
MATLAB, 209 Leibniz Formula, 93
some elementary functions, 206 Leibniz Rule, 93
infinite limit, 25 limit, 10 
integrable, 188
x
1
integral lim 1+ = e, 34
x→+∞ x
lower limit, 188 sin θ
numerical approximation lim = 1, 18
θ→0 θ
left point rule, 237 arithmetic rules, 13
midpoint rule, 238 at infinity, 19
right point rule, 238 at negative infinity, 19
Simpson’s rule, 241 at positive infinity, 19
trapezoidal rule, 238 finite, 10
upper limit, 188 infinite, 25
integral form of the remainder, 399 one-sided, 15
Integral Test, 359 sequence, 338
MATLAB, 365 convergence, 338
remainder estimate, 367 divergence, 338
integrand, 188 from a function, 341
integrating products of sin mx and cos nx, MATLAB, 349
224 Limit Comparison Test, 370
integrating products of powers of sin x and limit laws for sequences, 341
cos x, 226 limit of sequence, 338
integrating rational functions by partial linearity
fractions, 230 indefinite integral, 206
integration by parts, 218, 227 linearization, 86
integration of power series local extreme, 97
MATLAB, 416 local maximum, 97
interior point, 99, 101 MATLAB, 103
 445

local minimum, 97 taylor, 396

MATLAB, 103 trapz, 245
logarithmic differentiation, 79 viscircles, 271
logarithmic function, 40 vpa, 66
MATLAB, 41 absolute minimum, 105
lower limit, 188 alternating harmonic series, 380
Alternating Series Test, 380
Maclaurin series, 396 arc length, 291
common functions, 416 parametric curve, 294
MATLAB polar curve, 308
Inf, 256 area
NaN, 256 polar curve, 312
abs, 359 under a parametric curve, 274
assumeAlso, 375 centroid
assume, 375 of a curve, 317
cla, 271 parametric curve, 321
coeffs, 419 planar region between two curves, 325
collect, 210 chain rule, 69
compose, 31, 69 command help, 434
diff, 58, 192 command window, 429
exp, 34 comment, 434
factorial, 375 composite function, 31
finverse, 39 concavity test, 131
format long, 172 current folder, 430
fplot, 149 definite integral, 191
fzero, 104 polynomial, 191
help, 434 derivative, 58
inf, 21 determine the interval of convergence,
integral, 191, 254, 255, 306 407
int, 191, 209, 254, 255 determine the radius of convergence, 407
isAlways, 380 differentiation of power series, 415
length, 419 exponential function, 34
limit, 11 figure plot, 437
linspace, 149, 436 harmonic series
log2, log10, log, 42 alternating, 380
matlabFunction, 104, 306 implicit differentiation, 82
partfrac, 236 improper integral
plot, 9, 153, 269, 437 over infinite interval, 254
polarplot, 160 unbounded integrand, 255
polyint, 191 increasing/decreasing test, 113
pretty, 210 indefinite integral, 209
quadgk, 254, 255 Integral Test, 365
rsums, 186 integration of power series, 416
simplify, 375 inverse function, 39
solve, 82 inverse trigonometric functions, 46
subplot, 149 left point rule, 243
subs, 58 left-hand limit, 17
symsum, 355 limit
446 Index

sequence, 349 graph of the inverse function, 39

limits at infinities, 21 implicit differentiation, 81
local maximum, 103 integrating products of sin mx and
local minimum, 103 cos nx, 224
logarithmic function, 41 integrating products of powers of sin x
m-file, 431 and cos x, 226
method integrating rational functions by partial
Simpson’s rule, 245 fractions, 230
midpoint rule, 244 logarithmic differentiation, 79
monotone test, 113 numerical integration, 237
Newton’s iteration, 171 Simpson’s approximation, 240
Newton’s method, 171 Simpson’s rule
one-sided limit, 17 MATLAB, 245
parametric curves, 153 solving inequality, 112
partial fractions for rational functions, trigonometric substitutions, 216
235 midpoint rule, 238
polar curves, 160 error bound, 240
product of power series, 410 MATLAB, 244
quotient of power series, 410 moment, 313, 314
Ratio Test, 375 about the x-axis, 314
Riemann sum, 186 about the y-axis, 314
right point rule, 243 about the origin, 313
right-hand limit, 17 monotone, 107
Root Test, 377 monotone test, 111
semicolon, 433 Monotonic Sequence Theorem, 347
series
divergence, 359 natural domain, 2
Newton’s iteration, 168
single quote, 113
MATLAB, 171
substitution, 58
Newton’s Law of Cooling, 4
surface area of revolution, 301
Newton’s method, 168
parametric curve, 305
MATLAB, 171
symbolic function, 8
notations of derivative, 57
Taylor polynomials, 396
numerical integration
telescoping series, 355
left point rule, 237
trapezoidal rule, 245
midpoint rule, 238
trigonometric functions, 46
right point rule, 238
value assignment, 433
Simpson’s rule, 241
vector, 149
trapezoidal rule, 238
vector definition, 435
volume, 279 odd function, 141
work, 331 one-sided limit
workspace, 430 MATLAB, 17
Mean Value Theorem, 108 one-sided limits, 15
method one-to-one, 36
determine the interval of convergence,
405 parametric curve, 151
find the absolute extrema, 103 parametric curves
find the inverse function, 38 MATLAB, 153
 447

parametric equations, 151 Riemann sum, 188

partial fractions for rational functions MATLAB, 186
MATLAB, 235 right point rule, 238
partial sum, 350 MATLAB, 243
period, 141 right-hand limit, 16
periodic function, 141 MATLAB, 17
pitfall Root Test, 376
arithmetic rules fail, 13 MATLAB, 377
continuous but not differentiable, 61 rules of differentiation, 61
critical points, 101
improper integral, 253 second derivative test, 117
zero of the denominator ⇐⇒ 6 vertical MATLAB, 120
asymptote, 26 sequential limit, 346
polar coordinates, 155 series
polar curve, 155 absolute convergence, 382
slope of the tangent, 158 conditional convergence, 382
polar curves convergence, 350
MATLAB, 160 divergence, 350
power series, 396 MATLAB, 359
centered at a, 396 partial sum, 350
product of power series sigma notation, 188
MATLAB, 410 Simpson’s rule, 241
properties error bound, 242
exponential functions, 33 slant asymptote, 22, 147
logarithmic functions, 41 Squeeze Theorem, 14
for sequences, 342
quotient of power series Stirling’s formula, 344
MATLAB, 410 substitution
definite integral, 212
range, 1 indefinite integral, 211
Ratio Test, 372 summation notation, 188
MATLAB, 375 surface area of revolution, 297
reciprocal rule, 64 general formula, 299
recursive formula, 221 MATLAB, 301
related rates, 84 parametric curve, 302
relationship between differentiability and MATLAB, 305
continuity, 61
relationship between the limit and the table
one-sided limits, 16 indefinite integral, 206
remainder Taylor expansion, 396
Taylor series, 396 Taylor polynomials, 393
integral form, 399 asymptotic approximation, 394
Lagrange, 399 MATLAB, 396
remainder estimate uniqueness, 393
Alternating Series Test, 380 Taylor series, 396
Integral Test, 367 convergence theorem, 397
Taylor series, 400 integral form of the remainder, 399
Riemann integral, 188 Lagrange remainder, 400
448 Index

remainder, 396 l’Hospital’s Rule, 134

telescoping series, 352 Leibniz Formula, 93
MATLAB, 355 Limit Comparison Test, 370
test Limit Laws for sequences, 341
concavity, 130 linear approximation, 86
MATLAB, 131 Mean Value, 108
first derivative, 114 monotone test, 111
MATLAB, 119 monotonic sequence, 347
increasing/decreasing, 111 properties of exponential functions, 33
MATLAB, 113 properties of logarithmic functions, 41
monotone, 111 Ratio Test, 372
MATLAB, 113 relationship between differentiability and
second derivative, 117 continuity, 61
MATLAB, 120 relationship between the limit and the
the number e, 34 one-sided limits, 16
theorem Root Test, 376
Alternating Series Test rules of differentiation, 61
remainder estimate, 380 second derivative test, 117
analytic function, 417 sequential limit of continuous function,
arithmetic rules of limits, 13 346
chain rule, 68 squeeze, 14
Comparison Test, 369 Taylor series
concavity test, 130 convergence theorem, 397
constant function, 112 integral form of the remainder, 399
continuity of basic elementary functions, Lagrange remainder, 400
6, 47 theorem of analytic function, 417
continuity of composition of continuous transcendental function, 4
functions, 31 trapezoidal rule, 238
continuity of the inverse function, 39 error bound, 240
continuity with arithmetic operations, 28 MATLAB, 245
derivatives of exponential functions, 75 trigonometric functions, 5
derivatives of inverse trigonometric MATLAB, 46
functions, 73 trigonometric substitutions, 216
derivatives of logarithmic functions, 77
unit-impulse signal, 12
derivatives of trigonometric functions, 65
upper limit, 188
differentiation of inverse function, 71
extreme value theorem, 102 vector
Fermat’s Theorem, 99, 101 MATLAB, 149
first derivative test, 114 vertical asymptote, 25
Fundamental Theorem of Calculus, 201 volume, 275
geometric sequences, 345 by cross sections, 275
growth rates of sequences, 343 by cylindrical shells, 281
horizontal line test, 36 MATLAB, 279
integrable functions, 189 using shells, 281
Integral Test, 359 using slices, 276
remainder estimate, 367
integration by parts, 218, 227 work, 327
intermediate value, 32 MATLAB, 331
 449

zero derivative constant function, 112