This action might not be possible to undo. Are you sure you want to continue?

# A Summary of Calculus

**Karl Heinz Dovermann
**

Professor of Mathematics

University of Hawaii

July 28, 2003

c Copyright 2003 by the author. All rights reserved. No part of this

publication may be reproduced, stored in a retrieval system, or transmit-

ted, in any form or by any means, electronic, mechanical, photocopying,

recording, or otherwise, without the prior written permission of the author.

Printed in the United States of America.

This publication was typeset using A

M

S-T

E

X, the American Mathemat-

ical Society’s T

E

X macro system, and L

A

T

E

X2

ε

. The graphics were produced

with the help of Mathematica

1

.

This is an incomplete draft which will undergo further changes.

1

Mathematica Version 2.2, Wolfram Research, Inc., Champaign, Illinois (1993).

Contents

Preface v

1 Basic Concepts 1

1.1 Real Numbers and Functions . . . . . . . . . . . . . . . . . . 1

1.2 Limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.2.1 Two important estimates . . . . . . . . . . . . . . . . 4

1.3 More Limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

1.4 Continuous Functions . . . . . . . . . . . . . . . . . . . . . . 8

1.5 Lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

1.6 Tangent Lines and the Derivative . . . . . . . . . . . . . . . . 10

1.6.1 Derivatives without Limits . . . . . . . . . . . . . . . 12

1.7 Secant Lines and the Derivative . . . . . . . . . . . . . . . . . 13

1.8 Diﬀerentiability implies Continuity . . . . . . . . . . . . . . . 14

1.9 Basic Examples of Derivatives . . . . . . . . . . . . . . . . . . 15

1.10 The Exponential and Logarithm Functions . . . . . . . . . . . 18

1.11 Diﬀerentiability on Closed Intervals . . . . . . . . . . . . . . . 20

1.12 Other Notations for the Derivative . . . . . . . . . . . . . . . 21

1.13 Rules of Diﬀerentiation . . . . . . . . . . . . . . . . . . . . . 22

1.13.1 Linearity of the Derivative . . . . . . . . . . . . . . . . 22

1.13.2 Product and Quotient Rules . . . . . . . . . . . . . . . 23

1.13.3 Chain Rule . . . . . . . . . . . . . . . . . . . . . . . . 26

1.13.4 Hyperbolic Functions . . . . . . . . . . . . . . . . . . 29

1.13.5 Derivatives of Inverse Functions . . . . . . . . . . . . . 30

1.13.6 Implicit Diﬀerentiation . . . . . . . . . . . . . . . . . . 34

1.14 Related Rates . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

1.15 Exponential Growth and Decay . . . . . . . . . . . . . . . . . 41

1.16 More Exponential Growth and Decay . . . . . . . . . . . . . 43

1.17 The Second and Higher Derivatives . . . . . . . . . . . . . . . 48

1.18 Numerical Methods . . . . . . . . . . . . . . . . . . . . . . . . 48

i

1.18.1 Approximation by Diﬀerentials . . . . . . . . . . . . . 48

1.18.2 Newton’s Method . . . . . . . . . . . . . . . . . . . . . 51

1.18.3 Euler’s Method . . . . . . . . . . . . . . . . . . . . . . 53

1.19 Table of Important Derivatives . . . . . . . . . . . . . . . . . 63

2 Global Theory 65

2.1 Cauchy’s Mean Value Theorem . . . . . . . . . . . . . . . . . 65

2.2 Unique Solutions of Diﬀerential Equations . . . . . . . . . . . 67

2.3 The First Derivative and Monotonicity . . . . . . . . . . . . . 69

2.3.1 Monotonicity on Intervals . . . . . . . . . . . . . . . . 69

2.3.2 Monotonicity at a Point . . . . . . . . . . . . . . . . . 75

2.4 The Second Derivative and Concavity . . . . . . . . . . . . . 76

2.4.1 Concavity on Intervals . . . . . . . . . . . . . . . . . . 77

2.4.2 Concavity at a Point . . . . . . . . . . . . . . . . . . . 80

2.5 Local Extrema and Inﬂection Points . . . . . . . . . . . . . . 81

2.6 Detection of Local Extrema . . . . . . . . . . . . . . . . . . . 83

2.7 Detection of Inﬂection Points . . . . . . . . . . . . . . . . . . 88

2.8 Absolute Extrema of Functions . . . . . . . . . . . . . . . . . 90

2.9 Optimization Story Problems . . . . . . . . . . . . . . . . . . 92

2.10 Sketching Graphs . . . . . . . . . . . . . . . . . . . . . . . . . 98

3 Integration 105

3.1 Properties of Areas . . . . . . . . . . . . . . . . . . . . . . . . 105

3.2 Partitions and Sums . . . . . . . . . . . . . . . . . . . . . . . 107

3.2.1 Upper and Lower Sums . . . . . . . . . . . . . . . . . 108

3.2.2 Riemann Sums . . . . . . . . . . . . . . . . . . . . . . 111

3.3 Limits and Integrability . . . . . . . . . . . . . . . . . . . . . 112

3.3.1 The Darboux Integral and Areas . . . . . . . . . . . . 112

3.3.2 The Riemann Integral . . . . . . . . . . . . . . . . . . 115

3.4 Integrable Functions . . . . . . . . . . . . . . . . . . . . . . . 116

3.5 Some elementary observations . . . . . . . . . . . . . . . . . . 118

3.6 Areas and Integrals . . . . . . . . . . . . . . . . . . . . . . . . 121

3.7 Anti-derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . 122

3.8 The Fundamental Theorem of Calculus . . . . . . . . . . . . . 124

3.8.1 Some Proofs . . . . . . . . . . . . . . . . . . . . . . . 126

3.9 Substitution . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128

3.9.1 Substitution and Deﬁnite Integrals . . . . . . . . . . . 131

3.10 Areas between Graphs . . . . . . . . . . . . . . . . . . . . . . 133

3.11 Numerical Integration . . . . . . . . . . . . . . . . . . . . . . 136

3.12 Applications of the Integral . . . . . . . . . . . . . . . . . . . 142

ii

3.13 The Exponential and Logarithm Functions . . . . . . . . . . . 144

3.13.1 Other Bases . . . . . . . . . . . . . . . . . . . . . . . . 148

4 Trigonometric Functions 151

iii

iv

Preface

In these notes we like to summarize calculus.

v

vi PREFACE

Chapter 1

Basic Concepts

Introduction

In this chapter we introduce limits and derivatives. These are basic concepts

of calculus. We provide some rules for their computations.

1.1 Real Numbers and Functions

We assume that the reader is familiar with the real numbers (denoted by R)

and the operations of addition and multiplication. A real number is either

positive, negative, or zero. This allows us to order the real numbers. If

x and y are real numbers, then x is larger than y (i.e., x > y) if x − y is

positive.

Until further notice, we will work with real valued functions in one real

variable. Their domains, the sets on which these functions are deﬁned, are

subsets of the real numbers, and they take values in R. The range of a

function is a set in which the function takes values. The image of a function

f consists of all those points y in the range for which there exists an x in

the domain of f, such that f(x) = y.

We will make frequent use of the absolute value function.

|x| =

x if x > 0

−x if x < 0

0 if x = 0

The distance between two points a and b on the real line is |a − b|, and

{x ∈ R | |x − a| < } is the set of all real numbers whose distance from

1

2 CHAPTER 1. BASIC CONCEPTS

a is less than . Expressed as an interval this set is (a − , a + ). For

computations with absolute values it is worth noting that, for any two real

numbers x and x

|x · y| = |x| · |y|, |x +y| ≤ |x| +|y|, and ||x| −|y|| ≤ |x −y|. (1.1)

The ﬁrst inequality is referred to as triangle inequality, and the last one is

a variation of it.

Every now and then we will allude to the completeness of the real line,

which means that every bounded subset of the real line has a least upper

bound. This property is crucial for calculus, but arguments using it are too

diﬃcult for an introductory course on the subject.

1.2 Limits

Limits are a central tool in calculus and other areas of mathematics. We

discuss them in this section.

Deﬁnition 1.1. Let f be a function and L a real number. We say that

L = lim

x→a

f(x) (1.2)

if for all > 0 there exists a δ > 0, such that |f(x) − L| < whenever x is

in the domain of f and 0 < |x −a| < δ.

The equation in (1.2) reads as L is the limit of f(x) as x approaches

a. We also say that f(x) approaches or converges to L as x approaches a.

An intuitive interpretation is that the expected value of f(x) at x = a is L,

based on the values of f(x) for x near a.

In all but a few degenerate cases, limits are unique if they exist.

Proposition 1.2. Suppose that f(x) has a limit at x = a, then this limit is

unique, provided that the domain of the function f contains points arbitrarily

close to a.

1 2

The latter assumption in the proposition is satisﬁed if the domain of f

contains an interval, and either a belongs to this interval or a is an end point

1

Expressed in mathematical language this means, for all δ > 0 there is a point b in the

domain of f, such that 0 < |b −a| < δ.

2

Some authors do not apply the concept of a limit at isolated points of the domain of

a function, points for which there are no other arbitrarily close points in the domain of

the function.

1.2. LIMITS 3

of it. To avoid intricate language, we make this kind of an assumption for

the remainder of this section. Taking limits is compatible with the basic

algebraic operations in the following sense.

Proposition 1.3. Assume that the domains of the functions f(x) and g(x)

both contain an interval of the form (d, a) or (a, e) where d < a < e. Suppose

that

lim

x→a

f(x) = L and lim

x→a

g(x) = M.

and that c is a constant. Then

lim

x→a

(f +g)(x) = M +L

lim

x→a

cf(x) = cM

lim

x→a

(f · g)(x) = M · L

lim

x→a

(f/g)(x) = M/L provided that L = 0.

As a special case we obtain the following useful observation:

lim

x→c

f(x) = L if and only if lim

x→c

(f(x) −L) = 0. (1.3)

Proposition 1.4 (Pinching Theorem). Assume that the domains of the

functions f(x), g(x), and h(x) all contain an interval of the form (d, a) or

(a, e) where d < a < e and that f(x) ≤ h(x) ≤ g(x). If

lim

x→a

f(x) = L = lim

x→a

g(x),

then the limit of h(x) exists as x approaches a, and it is equal to L.

For many functions the computation of limits is no challenge.

Proposition 1.5. If f(x) is a polynomial, a rational function, or a trigono-

metric function and f(a) is deﬁned, then

lim

x→a

f(x) = f(a).

The following limits are important in the calculations of some derivatives.

lim

x→0

cos x −1

x

= 0, lim

x→0

sin x

x

= 1. and lim

x→a

x

n

−a

n

x −a

= na

n−1

. (1.4)

Hints: The ﬁrst two limits follow easily from the estimates in Theo-

rem 1.7, discussed in the following subsection. The last assertion can be

proved using synthetic division, at least if n is an integer.

4 CHAPTER 1. BASIC CONCEPTS

1.2.1 Two important estimates

In preparation of the proof of Theorem 1.7 we show

Theorem 1.6. If h ∈ [−π/4, π/4], then

| sin h| ≤ |h| ≤ | tan h|. (1.5)

Proof. In Figure 1.1 you see part of the unit circle. For h ∈ [−π/4, π/4] we

set C = (cos h, sin h). Given two points X and Y in the plane, the distance

between them is denoted by XY . We denote by

¯

BC the length of the arc

(part of the unit circle) between B and C.

O A B

C

D

E

Figure 1.1: The unit circle

We ﬁnd that | sin h| = AC ≤ |h| =

¯

BC because going from C straight

down to the x-axis is shorter than following the circle from C to the x-axis.

Secondly, to show that |h| =

¯

BC ≤ | tan h| = BD, imagine that you roll

the circle along the vertical line through B until the point C touches it in the

point E. We use the process of rolling the circle along the line to measure

|h|. In particular, |h| = BE. It appears to be clear

3

that BE ≤ BD. This

3

Here our argument relies on intuition. A rigorous argument requires work. One can

show that the area of a disk with radius one is π. From this is follows by elementary

geometry that the area of the slice of the disk with vertices O, B and C has area |h|/2.

This slice is contained in the triangle with vertices O, B and D, and the area of the slice

is (tan |h|)/2. It follows that |h| ≤ tan |h|.

1.2. LIMITS 5

veriﬁes that |h| ≤ tan |h|, the second inequality which we claimed in the

theorem.

Theorem 1.7. If h ∈ [−π/4, π/4], then

4

|1 −cos h| ≤

h

2

2

and |h −sin h| ≤

h

2

2

. (1.6)

Proof of Theorem 1.7. In Figure 1.2 you see half of a circle of radius 1

centered at the origin, and a triangle with vertices A, B, and C. Let

h ∈ [−π/4, π/4] be the number for which we want to show the inequal-

ity and C = (cos h, sin h). Denote by XY the length of the straight line

segment between the points X and Y . Let

¯

BC be the length of the arc

(part of the unit circle) between B and C.

A B

C

D

Figure 1.2: The unit circle

From the picture we read oﬀ that

AB = 2, DB = (1 −cos h),

¯

BC = |h|, and BC ≤

¯

BC.

Using similar triangles we see AB/BC = BC/DB and (BC)

2

= AB×DB.

In other words

2(1 −cos h) = AB ×DB = (BC)

2

≤ (

¯

BC)

2

= h

2

.

4

The inequalities hold without the restriction on h, but we only need them on an

interval around zero. Restricting ourselves to this interval simpliﬁes the proofs somewhat.

6 CHAPTER 1. BASIC CONCEPTS

The ﬁrst estimate in (1.6) is an immediate consequence.

If h = 0, then both sides of the second inequality in (1.6) are zero,

verifying the assertion in this case. If 0 = h ∈ [−π/4, π/4], then Theorem 1.6

tells us that

| sin h| ≤ |h| ≤ | tan h| =

| sin h|

cos h

hence 0 ≤ cos h ≤

sin h

h

≤ 1.

Subtracting the terms in this inequality from 1 we ﬁnd

0 ≤ 1 −

sin h

h

≤ 1 −cos h ≤ 1.

Using our previous estimate for |1 − cos h| and our assumption that |h| ≤

π/4 < 1, we conclude that

h −sin h

h

≤ |1 −cos h| ≤

h

2

2

≤

h

2

.

The second estimate claimed in the theorem is an immediate consequence.

1.3 More Limits

The material in the previous, ﬁrst section about limits suﬃces for a while.

In some situations one would like to modify the deﬁnition in Section 1.2,

and we do so in this section. The ﬁrst two limits express how the function

behaves as we approach a point a from the right or left. They are called the

right and left hand limits. The next two limits express what happens as the

variable tends to plus or minus inﬁnity. We call them limits at inﬁnity. The

last two limits allow us to express that the values of a function tend to plus

or minus inﬁnity. We call them inﬁnite limits.

Deﬁnition 1.8. Let f be a function and L a real number. We say that

L = lim

x→a

+

f(x)

if for all > 0 there exists a δ > 0, such that |f(x) − L| < whenever x is

in the domain of f and a < x < a +δ.

Deﬁnition 1.9. Let f be a function and L a real number. We say that

L = lim

x→a

−

f(x)

if for all > 0 there exists a δ > 0, such that |f(x) − L| < whenever x is

in the domain of f and a −δ < x < a.

1.3. MORE LIMITS 7

For example, if f(x) = sign(x) = x/|x|, then

lim

x→a

+

f(x) = 1 and lim

x→a

−

f(x) = −1.

We can consider what happens to the values of a function f(x) as x

approaches ∞ or −∞.

Deﬁnition 1.10. Let f be a function and L a real number. We say that

L = lim

x→∞

f(x)

if for all > 0 there exists a number M, such that |f(x) −L| < whenever

x is in the domain of f and x > M.

Deﬁnition 1.11. Let f be a function and L a real number. We say that

L = lim

x→−∞

f(x)

if for all > 0 there exists a number M, such that |f(x) −L| < whenever

x is in the domain of f and x < M.

For example

lim

x→∞

1

x

= 0 and lim

x→−∞

1

1 +x

2

= 1.

Deﬁnition 1.12. Let f be a function and a a real number. We say that

lim

x→a

f(x) = ∞

if for all M there exists a δ > 0 such that f(x) > M whenever x is in the

domain of f and 0 < |a −x| < δ.

In other words, we can make sure that the value of f(x) is larger than

any given number M, no matter how large, by taking x close to a.

Deﬁnition 1.13. Let f be a function and a a real number. We say that

lim

x→a

f(x) = −∞

if for all M there exists a δ > 0 such that f(x) < M whenever x is in the

domain of f and 0 < |a −x| < δ.

In the last two deﬁnitions a may be replaced by a

±

, so that we approach

a from the left or right, and a can be replaced by ±∞.

For example

lim

x→0

+

1

x

= ∞ and lim

x→∞

√

x = ∞.

8 CHAPTER 1. BASIC CONCEPTS

1.4 Continuous Functions

We deﬁne continuous functions and discuss a few of their basic properties.

The class of continuous functions will play a central role later.

Deﬁnition 1.14. Let f be a function and c a point in its domain. The

function is said to be continuous at c if for all > 0 there exists a δ > 0,

such that |f(c) − f(x)| < whenever x belongs to the domain of f and

|x−c| < δ. A function f is continuous if it is continuous at all points in its

domain.

In most cases the condition in Deﬁnition 1.14 says that

lim

x→c

f(x) = f(c). (1.7)

In fact, this equation holds whenever there are points in the domain of f

arbitrarily close to c. See the footnote to Proposition 1.2. If c is an isolated

point in the domain of f, i.e., there are no other points in the domain of f

arbitrarily close to c, then the function is always continuous at c.

Polynomials, rational functions, and trigonometric functions are contin-

uous. One can produce many more continuous functions through standard

operations on functions.

Proposition 1.15. Let f and g be continuous functions. Then f +g, f · g,

f/g and f ◦ g are continuous, wherever these functions are deﬁned.

The clarify the remark about the domain in the proposition, we note that

the function (f +g)(x) = f(x) +g(x) is deﬁned for those x for which both f

and g are deﬁned. The same statement holds for (f · g)(x) = f(x) · g(x). To

determine the domain of f/g one needs to exclude those points where g is

zero. For the composition (f ◦ g)(x) = f(g(x)) on needs that g takes values

in the domain of f.

One may also reverse the order of applying a continuous function and

calculating a limit:

lim

x→c

f(g(x)) = f

lim

x→c

g(x)

, (1.8)

provided the natural technical assumption hold, i.e., g is deﬁned at points

arbitrarily close to c, f is deﬁned for all g(x) where x is in the domain of g

and close to c, and f is continuous at lim

x→c

g(x).

Theorem 1.16 (Intermediate Value Theorem). Suppose that f is de-

ﬁned and continuous on the closed interval [a, b]. If C is in between f(a)

and f(b), then there exists a c ∈ [a, b], such that f(c) = 0.

1.5. LINES 9

E.g., suppose that p(x) = x

3

−x

2

+2x −1. The polynomial is certainly

a continuous function, p(0) = −1 and p(1) = 1. According to the theorem

there exists some c ∈ (0, 1), such that p(c) = 0.

Theorem 1.17 (Extreme Value Theorem). Let f be deﬁned and con-

tinuous on the closed interval [a, b]. Then there exist points c and d in [a, b],

such that f(c) ≤ f(x) ≤ f(d) for all x ∈ [a, b].

Expressed in words, the theorem says that a continuous function on a

closed interval assumes a smallest and largest value.

The Intermediate Value and Extreme Value theorem are typically proved

in an introductory analysis course. They are equivalent to the completeness

of the real line. We mentioned this property of the real numbers in Sec-

tion 1.1.

1.5 Lines

In general, a line consists of the points (x, y) in the plane which satisfy the

equation

ax +by = c (1.9)

for some given real numbers a, b and c, where it is assumed that a and b

are not both zero. The line is vertical if and only if b = 0. If b = 0 we may

rewrite the equation as

y = −

a

b

x +

c

b

= mx +B. (1.10)

The number m is called the slope of the line, and B is the point in which the

line intersects the y-axis, also called the y-intercept. Given any two points

(x

1

, y

1

) and (x

2

, y

2

) in the plane, the line through them has slope

m =

y

2

−y

1

x

2

−x

1

.

For our purposes, the most useful version of the equation of a line is its

point-slope formula. The equation of a line with slope m through the point

(x

1

, y

1

) is

y = m(x −x

1

) +y

1

. (1.11)

10 CHAPTER 1. BASIC CONCEPTS

1.6 Tangent Lines and the Derivative

We like to introduce the concept of tangent lines. To be able to express

ourselves concisely, let us say

Deﬁnition 1.18. A point c is an interior point of a subset B of R if there

is an open interval I, such that c ∈ I ⊆ B.

We give a ﬁrst deﬁnition for a tangent line.

Deﬁnition 1.19. Suppose f(x) is a function and c is an interior point of

its domain. We call a line t(x) the tangent line to the graph of f(x) at x = c

if t(x) is the best linear approximation of f(x) on some open interval around

c, i.e., the line t(x) is closer to the graph of f(x) than any other line for all

x in some open interval around c.

For a given function and an interior point c in its domain there may or

may not be a tangent line, but it there is a tangent line, then it is unique.

Although the term ‘best linear approximation near c’ gives an excellent

intuitive picture what a tangent line is, this deﬁnition is hard to work with.

It is easier to work with a more concrete deﬁnition.

Deﬁnition 1.20. Suppose f(x) is a function and c is an interior point of

its domain. We call a line t(x) the tangent line to the graph of f(x) at x = c

if

lim

x→c

f(x) −t(x)

x −c

= 0. (1.12)

The equation in (1.12) expresses in a precise form in which sense the

tangent line is close to the graph of f(x) near c. Not only does f(x) −t(x)

converge to zero as x approaches c, it does so even when divided by x −c.

We use tangent lines to deﬁne the concept of diﬀerentiability and the

derivative.

Deﬁnition 1.21. Suppose f(x) is a function and c is an interior point of

its domain, and assume that there is a tangent line to the graph of f(x) at

x = c. Then we say that f(x) is diﬀerentiable at c. We call the slope of the

tangent line the derivative of f(x) at c, and we denote it by f

(c).

Utilizing the notation in the previous deﬁnition we can write down the

equation of the tangent line to the graph of f(x) at x = c in point-slope

form:

t(x) = f

(c)(x −c) +f(c). (1.13)

1.6. TANGENT LINES AND THE DERIVATIVE 11

To diﬀerentiate a function means to ﬁnd its derivative.

By deﬁnition, an open set is a set, such that each of its points is an

interior point.

Deﬁnition 1.22. Suppose the domain of the function f(x) is an open set.

Then say that f(x) is diﬀerentiable if it is diﬀerentiable at each point of

its domain. We consider f

**(x) as a function, whose domain consists of all
**

those points where f(x) is diﬀerentiable.

Example 1.23. Let p(x) = 2x

4

−3x

2

+5. Find the tangent line t(x) to the

graph of p(x) at x = −2 and p

(−2).

Solution: As a ﬁrst step we expand p in powers of u = (x + 2). To do

so, we substitute u−2 for x and expand p in powers of u. You are expected

to ﬁll in some of the arithmetic steps.

p = 2(u −2)

4

−3(u −2)

2

+ 5

= 2(u

4

−8u

3

+ 24u

2

−32u + 16) −3(u

2

−4u + 4) + 5

= 2u

4

−16u

3

+ 45u

2

−52u + 25

Reversing the substitution, replacing u by (x + 2), we ﬁnd:

p(x) = 2(x + 2)

4

−16(x + 2)

3

+ 45(x + 2)

2

−52(x + 2) + 25.

We assert that t(x) = −52(x + 2) + 25 and p

(−2) = −52.

For t(x) as proposed, we see that

p(x) −t(x)

x −c

=

p(x) −t(x)

x + 2

= |2(x + 2)

3

−16(x + 2)

2

+ 45(x + 2)|

≤ 65|x + 2| (provided |x + 2| ≤ 1)

This estimate shows that (p(x) − t(x))/(x − c) converges to zero as x ap-

proaches c = −2. By deﬁnition, this means that t(x) is the desired tangent

line. Its slope is p

(−2) = −52. ♦

The example is generic. We can use any polynomial p(x) and point x = c

and write p(x) in powers of (x −c). Say, the result is

p(x) = A

n

(x −c)

n

+· · · +A

1

(x −c) +A

0

.

The technique used in the example, suitably generalized, shows that

t(x) = A

1

(x −c) +A

0

12 CHAPTER 1. BASIC CONCEPTS

is the tangent line to the graph of p(x) at x = c, and p

(c) = A

1

. Eventually

we will ﬁnd a more eﬃcient method for diﬀerentiating polynomials, but we

have shown that

Proposition 1.24. Polynomials are diﬀerentiable.

1.6.1 Derivatives without Limits

Without a doubt, the deﬁnition of a limit is the most diﬃcult one in a

ﬁrst semester of calculus, and it is interesting to explore ways to develop

calculus, rigorously, without the limit concept. One can do this by replacing

the condition in (1.12) by a slightly stronger one.

Deﬁnition 1.25. Suppose f(x) is a function and c is an interior point of

its domain. We call a line t(x) the tangent line to the graph of f(x) at x = c

if there exists and open interval I around c and a number A, such that

|f(x) −t(x)| ≤ A(x −c)

2

(1.14)

for all x ∈ I.

With this deﬁnition fewer functions will be diﬀerentiable than with the

one given in Deﬁnition 1.20, but this is not crucial.

The inequality in (1.14) can be rewritten as

q(x) = t(x) −A(x −c)

2

≤ f(x) ≤ t(x) +A(x −c)

2

= p(x),

where the parabolas q(x) and p(x) are deﬁned by the expressions they are

adjacent to. All four function f(x), q(x), p(x), and t(x) have the same value

at x = c. In an example, this situation is shown in Figure 1.3. There you

see the function f(x) = sin x, the parabola p(x) (dotted and open upwards),

the parabola q(x) (dotted and open downwards), and the tangent line t(x)

(dashed). The parabolas p(x) and q(x) touch each other without crossing,

and the picture shows how they ‘hug’ each other. There is very little space

left between p(x) and q(x), and f(x) and t(x) are squeezed in between them.

In this sense, the graphs of f(x) and t(x) have to be close to each other near

x = c.

A pedagogical advantage of the approach is that one does not have to

understand limits before one can understand the deﬁnition of the derivative.

There is also a geometric picture which illustrates the concept of closeness,

tangent line, and derivative. The condition in (1.14) is also more accessible

to computer assisted algebra than the limit deﬁnition. In terms of algebraic

geometry (1.14) at least alludes to a divisibility condition.

1.7. SECANT LINES AND THE DERIVATIVE 13

0.5 1 1.5 2

0.25

0.5

0.75

1

1.25

1.5

Figure 1.3: Sine Function and Tangent Line between two Parabolas

1.7 Secant Lines and the Derivative

Often a diﬀerent approach is taken to motivate and introduce the derivative.

Theorem 1.26. Suppose f is a function and c is an interior point of its

domain. If f is diﬀerentiable at c, then

f

(c) = lim

x→c

f(c) −f(x)

c −x

.

Proof. This is obvious once one uses the expression for the tangent line in

(1.13) and substitutes it in the expression in (1.12) inside the limit.

f(x) −t(x)

x −c

=

f(x) −f(c)

x −c

−f

(c). (1.15)

Apply limits to both sides of the equation and the assertion follows.

Let us explain the situation geometrically. Suppose a and b are distinct

points in the domain of the function f. The line through (a, f(a)) and

(b, f(b)) is called a secant line, and its slope (f(a) − f(b))/(a − b) is called

14 CHAPTER 1. BASIC CONCEPTS

the average rate of change of f over the interval [a, b]. In (1.15) we are

considering the slopes of secant lines through (c, (f(c)) and (x, f(x)), and

then we take the limit as x approaches c. The theorem asserts that for a

diﬀerentiable function this limit of the slopes of secant lines is the slope of

the tangent line. For the obvious reason f

**(c) is called the rate of change or
**

instantaneous rate of change of f at c.

Many authors introduce the derivative as the limit of the slopes of secant

lines, call t(x) = f

**(x − c) + f(c) the tangent line, and possibly illustrate
**

that the tangent line is close to the graph in the sense of Deﬁnition 1.20.

-0.1 -0.05 0.05 0.1

-0.0075

-0.005

-0.0025

0.0025

0.005

0.0075

Figure 1.4: f(x) = x

2

sin(1/x)

-0.01 -0.005 0.005 0.01

-0.0001

-0.00005

0.00005

0.0001

Figure 1.5: f(x) = x

2

sin(1/x)

It is misleading to say that the graph of f(x) looks like, or resembles, a

line near c. Eventually you will be able to show that the function

f(x) =

x

2

sin(1/x) if x = 0

0 if x = 0

is diﬀerentiable everywhere on the real line. You see part of its graph over

two diﬀerent intervals in Figure 1.4 and 1.5. By no stretch of imagination

will you say that the graph of the function looks like a line.

1.8 Diﬀerentiability implies Continuity

It is worth pointing out that

Theorem 1.27. If a function is diﬀerentiable at a point, then it is contin-

uous at this point.

1.9. BASIC EXAMPLES OF DERIVATIVES 15

Proof. Denote the function by f(x) and the point of diﬀerentiability by c.

By assumption we have the derivative f

(c) and

lim

x→c

¸

f(x) −f(c)

x −c

−f

(c)

= 0.

Then certainly

lim

x→c

[(f(x) −f(c)) −f

(c)(x −c)] = 0.

Because f

**(c)(x−c) converges to zero as x approaches c, so does (f(x)−f(c)).
**

This implies that lim

x→c

f(x) = f(c) and that f(x) is continuous at c.

-2 -1 1 2

0.5

1

1.5

2

Figure 1.6: The absolute value function

The converse of the theorem is false. There are continuous functions

which are not diﬀerentiable. E.g., the function f(x) = |x| is continuous,

but it is not diﬀerentiable at x = 0. It is apparent from the graph (see

Figure 1.6) that there is not line close to the graph of this function near

x = 0.

We can also give an analytic argument. According to the deﬁnition of

diﬀerentiability, we have to study the diﬀerence quotients (|x|−|0|)/(x−0) =

|x|/x. They are 1 if x > 0 and −1 if x < 0. There is no number these

diﬀerence quotients converge to, and f(x) = |x| is not diﬀerentiable at x = 0.

1.9 Basic Examples of Derivatives

Let us use the deﬁnitions and work out a few derivatives.

16 CHAPTER 1. BASIC CONCEPTS

Example 1.28. If f(x) = x

n

and n is a non-negative integer, i.e., n = 0,

1, 2, . . . , then f

(x) = nx

n−1

.

Proof. Suppose that n ≥ 2. Then

lim

x→c

x

n

−c

n

x −c

= lim

x→c

(x

n−1

+x

n−2

c +· · · xc

n−2

+c

n−1

) = nc

n−1

The cases n = 0 and n = 1 are even easier and left to the reader.

Example 1.29. If f(x) = 1/x, then f

(x) = −1/x

2

.

Proof. Suppose c = 0.

lim

x→c

1

x

−

1

c

x −c

= lim

x→c

c −x

xc(x −c)

= −

1

c

2

.

Example 1.30. If f(x) =

√

x and x > 0, then f

(x) = 1/(2

√

x).

Proof.

lim

x→c

√

x −

√

c

x −c

= lim

x→c

x −c

(x −c)(

√

x +

√

c)

=

1

2

√

c

.

Remark 1. Eventually we will see that if f(x) = x

a

for any real number

a, then f

(x) = ax

a−1

, generalizing all of the examples above.

Exercise 1. Suppose that f(x) =

√

ax +b and ax +b > 0. Show that

f

(x) =

a

2

√

ax +b

.

The tangent line to the graph of f(x) at x = c is then

t(x) =

a

2

√

ac +b

(x −c) +

√

ac +b.

Verify that

|f(x) −t(x)| ≤

a

2

2(

√

ac +b)

3

(x −c)

2

. (1.16)

1.9. BASIC EXAMPLES OF DERIVATIVES 17

The estimate in (1.16) shows diﬀerentiability in the sense of Deﬁni-

tion 1.25, and provides an explicit error estimate, a bound on the diﬀerence

between the function and its tangent line.

Example 1.31. Show that sin

**x = cos x. For this equation to hold, the
**

angle x needs to be measured in radians.

Proof. Below we will set x = c +h and x −c = h.

lim

x→c

¸

sin x −sin c

x −c

= lim

h→0

¸

sin(c +h) −sin c

h

= lim

h→0

¸

sinc cos h + cos c sin h −sin c

h

= lim

h→0

¸

sinc(cos h −1) + cos c sin h

h

= sin c · lim

h→0

cos h −1

h

+ cos c · lim

h→0

sin h

h

= cos c.

For computation of the limits in the second to last line see (1.4).

The tangent line to the graph of the sine function at x = c is

t(x) = cos c(x −c) + sin c.

It is left as an exercise for the reader to show that

| sin x −t(x)| ≤ (x −c)

2

(1.17)

The steps are essentially the same as in the proof above. The estimate in

(1.17) does not only show diﬀerentiability in the sense of Deﬁnition 1.25,

but it provides an explicit error estimate, a bound on the diﬀerence between

the function and its tangent line.

Exercise 2. If f(x) = cos x, then f

**(x) = −sin x. The details are similar
**

to the ones in Example 1.31. Furthermore, if

t(x) = sin c(x −c) + cos x

is the tangent line to the graph of f(x) at x = c, then

|f(x) −t(x)| ≤ (x −c)

2

.

18 CHAPTER 1. BASIC CONCEPTS

1.10 The Exponential and Logarithm Functions

The exponential and logarithm are of great importance and we do not want

to delay their introduction any further. Still, technically we are not quite

prepared for it and at a later point we have to revisit the introduction to ﬁll

in details.

Suppose a is a positive real number and a = 1. For any rational number

r = p/q (p and q are integers) one can deﬁne a

r

=

q

√

a

p

. First we take a p-th

power and then a q-root. In this sense we have a function h(r) = a

r

, whose

domain consists of all rational numbers. This function is monotonic. More

precisely, h(r) is increasing if a > 1 and decreasing when 0 < a < 1.

Theorem-Deﬁnition 1.32. Let a be a positive number, a = 1. There

exists exactly one monotonic function, called the exponential function with

base a and denoted by exp

a

(x), which is deﬁned for all real numbers x such

that exp

a

(x) = a

x

whenever x is a rational number. Furthermore, a

x

> 0 for

all x, so that the domain of the exponential function is (−∞, ∞). For every

number y > 0 there exists exactly one number x, such that exp

a

(x) = y, so

we use (0, ∞) as the range of the exponential function exp

a

(x).

It is common, and we will follow this convention, to use the notation

a

x

for exp

a

(x) also if x is not rational. The arithmetic properties of the

exponential function, also called the exponential laws, are collected in our

next theorem. The theorem just says that the exponential laws, which you

previously learned for rational exponents, also hold in the generality of our

current discussion.

Theorem 1.33 (Exponential Laws). For any positive real number a and

all real numbers x and y

a

x

a

y

= a

x+y

a

x

/a

y

= a

x−y

(a

x

)

y

= a

xy

If x is the unique solution of the equation a

x

= y, then we set

log

a

(y) = x. (1.18)

We just deﬁned a function log

a

(y). It is called the logarithm function with

base a, and by construction it is the inverse of the exponential function

exp

a

(x). More explicitly,

a

log

a

y

= y and log

a

(a

x

) = x

1.10. THE EXPONENTIAL AND LOGARITHM FUNCTIONS 19

for all x ∈ R and all y > 0. The domain of the logarithm function is

(0, ∞) and its range is (−∞, ∞). It is increasing if a > 1 and decreasing if

0 < a < 1.

Corresponding to the exponential laws in Theorem 1.33 we have the laws

of logarithms. One set of laws implies the other one, and vice versa.

Theorem 1.34 (Laws of Logarithms). For any positive real number a =

1, for all positive real numbers x and y, and any real number z

log

a

(xy) = log

a

(x) + log

a

(y)

log

a

(x/y) = log

a

(x) −log

a

(y)

log

a

(x

z

) = z log

a

(x)

In Figures 1.7 and 1.8 you see parts of the graphs of the exponential and

logarithm functions with base 2.

-1 -0.5 0.5 1 1.5

0.5

1

1.5

2

2.5

Figure 1.7: exp

2

(x)

0.5 1 1.5 2 2.5 3

-1

-0.5

0.5

1

1.5

Figure 1.8: log

2

(x)

The Euler number e as base

There is one number which is preferrable as base over the others. This

irrational number is called the Euler number (named after Leonard Euler)

and denoted by e, and e ≈ 2.718281828. We will deﬁne it precisely later.

Deﬁnition 1.35. The exponential function is the exponential function for

the base e. It is denoted by exp(x) or e

x

. Its inverse is the natural logarithm

function. It is denoted by ln(x). So exp(x) = exp

e

(x) and ln(x) = log

e

(x).

20 CHAPTER 1. BASIC CONCEPTS

Eventually we will see

exp

(x) = exp(x) and ln

(x) =

1

x

. (1.19)

The derivative of the exponential function is the exponential function, and

the derivative of the natural logarithm function is 1/x.

Other Bases

Finally, let us relate the exponential and logarithm functions for diﬀerent

bases to those with base e. For any positive number a (a = 1),

Theorem 1.36.

a

x

= e

x lna

and log

a

x =

ln x

ln a

.

These identities follow from the exponential laws and the laws of loga-

rithms.

1.11 Diﬀerentiability on Closed Intervals

In Deﬁnition 1.22 we deﬁned what it means that a function is diﬀerentiable

on an open set. There are situations in which one would like to apply the

notion of diﬀerentiability to functions with other kinds of domains. Let us

formalize the idea of extending functions.

Deﬁnition 1.37. Suppose that I and J are subsets of the real line R and

I ⊆ J, that I is the domain of a function f, and that J is the domain of

a function F. We call F an extention of f if it agrees with f on I, i.e.,

F(x) = f(x) for all x ∈ I.

Deﬁnition 1.38. A function f is said to be diﬀerentiable on a subset I

of R if it extends to a diﬀerentiable function F on an open set. We set

f

(x) = F

(x) for all x ∈ I.

Without some restrictions on I, a function may be diﬀerentiable without

the derivative being well deﬁned. The least technical and for our purposes

suﬃcient solution is captured in

Proposition 1.39. Suppose the function f is deﬁned on an interval I, the

interval is neither empty nor a single point, and f extends to a diﬀerentiable

function F on an open interval containing I, then f

(x) = F

(x) is unique

for all x ∈ I.

1.12. OTHER NOTATIONS FOR THE DERIVATIVE 21

We are mostly concerned with deﬁning diﬀerentiability for functions

whose domain is a closed interval [a, b], where a < b. Some authors use

one-sided limits and one-sided derivatives to contemplate derivatives at the

end points of the interval. Our discussion is less painful, and it lends itself

more to generalizations in higher dimensions.

Let us discuss two examples. The function f(x) = x

2

with domain [0, 1]

is diﬀerentiable. It extends to the diﬀerentiable function F(x) = x

2

with

the open set (−∞, ∞) as its domain. In contrast, the function g(x) =

√

x

is not diﬀerentiable on the interval [0, ∞). The only sensible candidate for

the tangent line to the graph of g(x) at the point (0, 0) is a vertical line.

The slope of this line is not a real number and we do not have a derivative.

(The function g(x) is diﬀerentiable if we use (0, ∞) as domain.)

1.12 Other Notations for the Derivative

There are diﬀerent notations for the derivative of a function. Physicists will

indicate a derivative with respect to time by a dot. E.g., if x is a function

of time, then they will write ˙ x(t) instead of x

**(t). Leibnitz’ notation for the
**

derivative of a function f of a variable x is

df

dx

. We will use it frequently. Ex-

pressing the derivatives of the exponential and natural logarithm functions

this way (see (1.19)) we have:

If y(x) = e

x

, then

dy

dx

= y = e

x

, and if y(x) = ln x, then

dy

dx

=

1

x

.

This notation is not always speciﬁc enough. The expression dy/dx stands

for the derivative of y with respect to x, and that is a function. The ex-

pression does not tell where dy/dx is evaluated. To be speciﬁc about this

aspect, it makes sense to write (compare Example 1.31):

If y(x) = sinx, then

dy

dx

(x) = cos x.

In this notation x plays two roles. It is the name of the variable of y as

well as the name of the variable of the derivative of y. This in acceptable

because it won’t lead to confusion. Instead of

df

dx

(x) we also write

d

dx

f(x).

This is particularly convenient if f stands for a larger expression as in

d

dx

sin x = cos x or

d

dx

e

x

= e

x

.

22 CHAPTER 1. BASIC CONCEPTS

1.13 Rules of Diﬀerentiation

We discuss formulas for calculating the derivative of a composite function

from the derivatives of its constituents. These formulas, together with the

knowledge of the derivatives of some basic functions, turn the process of

diﬀerentiation for many functions into an algorithm, a rather mechanical

process. You can do it even on the computer, which means that no “under-

standing” is required. You are expected to learn the basic rules, be able to

apply the accurately, and practice many examples. In the last section of this

chapter we summarize the computational results of this section. We collect

the rules established in this section and tabulate the derivatives of many of

the important functions which we considered.

1.13.1 Linearity of the Derivative

Diﬀerentiation is compatible with addition of functions and multiplication

with a constant. In a more mathematical language one says that diﬀeren-

tiation is linear. Let f and g be functions, and assume that both of them

are diﬀerentiable at x. Let c be a real number. Then f + g and cf are

diﬀerentiable at x and their derivatives are given by

(f +g)

(x) = f

(x) +g

(x) and (cf)

(x) = cf

(x). (1.20)

In Leibnitz’ notation this reads

d

dx

(f +g)(x) =

df

dx

(x) +

dg

dx

(x) and

d

dx

(cf)(x) = c

df

dx

(x). (1.21)

In words, the derivative of a sum of functions is the sum of the derivatives,

and the derivative of a multiple of a function is the multiple of the derivative.

Example 1.40. Diﬀerentiate

h(x) = x

2

+ 3e

x

.

Solution: Set f(x) = x

2

, g(x) = e

x

and c = 3. Then h(x) = f(x) + 3g(x).

Previously we found that f

(x) = 2x and that g

(x) = e

x

, see (1.19). We

conclude that

h

(x) =

dh

dx

(x) = 2x + 3e

x

. ♦

1.13. RULES OF DIFFERENTIATION 23

Example 1.41. Diﬀerentiate log

a

x, the logarithm functions for an arbi-

trary positive base a, a = 1.

Solution: Recall that log

a

x =

ln x

ln a

, see Theorem 1.36. In this sense

log

a

x = cf(x) where c = 1/ ln a and f(x) = ln x. We stated previously that

ln

**x = 1/x, see (1.19). Using the linearity of the derivative, we ﬁnd
**

log

a

x =

d

dx

ln x

ln a

=

1

ln a

ln

x =

1

ln a

×

1

x

=

1

xln a

. ♦

Suppose f and g are deﬁned and diﬀerentiable on an set. Thinking of f

and g more as functions, and not so much as functions evaluated at a point,

we may omit (x) from the notation. Then the diﬀerentiation rules are

(f +g)

= f

+g

or

d

dx

(f +g) =

df

dx

+

dg

dx

(1.22)

and

(cf)

= cf

or

d

dx

(cf) = c

df

dx

. (1.23)

Example 1.42. Find the derivative of an arbitrary polynomial.

Solution: A polynomial is a ﬁnite sum of multiples of non-negative

powers of the variable, i.e., a function of the form

f(x) = a

n

x

n

+a

n−1

x

n−1

+· · · +a

1

x +a

0

,

where the a

i

are constants. Using Example 1.28 and the linearity of the

derivative we see right away that

f

(x) = na

n

x

n−1

+ (n −1)a

n−1

x

n−2

+· · · +a

1

.

Here is a speciﬁc example, a special case of the formula which we just

derived.

If f(x) = 4x

5

−3x

2

+ 4x + 5, then f

(x) = 20x

4

−6x + 4. ♦

1.13.2 Product and Quotient Rules

Next we state the product and the quotient rule. They allow us to calculate

the derivatives of products and quotients of functions. Again, let f and g

be functions, and assume that both of them are diﬀerentiable at x. For the

quotient rule assume in addition that g(x) = 0. Then the product fg and

the quotient f/g are diﬀerentiable at x and their derivatives are given by

24 CHAPTER 1. BASIC CONCEPTS

(fg)

(x) = f

(x)g(x) +f(x)g

(x) (1.24)

f

g

(x) =

f

(x)g(x) −f(x)g

(x)

[g(x)]

2

. (1.25)

In Leibnitz’ notation these formulas become

d

dx

(fg)(x) =

df

dx

(x)g(x) +f(x)

dg

dx

(x) (1.26)

d

dx

f

g

(x) =

df

dx

(x)g(x) −f(x)

dg

dx

(x)

[g(x)]

2

. (1.27)

Example 1.43. Diﬀerentiate the function h(x) = x

2

ln x.

Solution: Write h(x) = f(x)g(x) with f(x) = x

2

and g(x) = ln x.

Then f

(x) = 2x and g

**(x) = 1/x, see (1.19). Putting this into the product
**

formula yields

h

(x) = f

(x)g(x) +f(x)g

(x) = 2xln x +x

2

1

x

= x(2 ln x + 1). ♦

Example 1.44. Find the derivative of the rational function.

r(x) =

x

2

−5

x

3

+ 1

.

Solution: We set p(x) = x

2

− 5 and q(x) = x

3

+ 2. Then p

(x) = 2x

and q

(x) = 3x

2

. According to the quotient rule

r

(x) =

2x(x

3

+ 1) −(x

2

−5)3x

2

(x

3

+ 1)

2

=

−x

4

+ 15x

2

+ 2x

(x

3

+ 1)

2

. ♦

Example 1.45. The formula

d

dx

x

n

= nx

n−1

for all integer powers n. If n ≤ −1, then we domain of the function is R\{0},

the real line with the origin removed.

Solution: We veriﬁed this formula for n ≥ 0 in Example 1.28. Let n be

a negative integer and m = −n. Then

d

dx

x

n

=

d

dx

¸

1

x

m

=

0 · x

m

−1 · mx

m−1

x

2m

=

−m

x

m+1

= nx

n−1

.

1.13. RULES OF DIFFERENTIATION 25

Example 1.46. Find the derivative of

f(x) = tan x.

Solution: We express f(x) as a quotient of two functions, f(x) =

sin x/ cos x, and apply the quotient rule. Use that sin

**x = cos x (see Exam-
**

ple 1.31) and cos

**x = −sinx (see Exercise 2 on page 17). We ﬁnd
**

tan

x =

sin

xcos x −sin xcos

x

cos

2

x

=

cos

2

x + sin

2

x

cos

2

x

=

1

cos

2

x

= sec

2

x.

(1.28)

Some books and computer programs will give this result in a diﬀerent form.

Based on the relevant trigonometric identity, they write

tan

x = 1 + tan

2

x. (1.29)

That draws our attention to the fact that the function f(x) = tan x satisﬁes

the diﬀerential equation

f

(x) = 1 +f

2

(x). ♦

Example 1.47. Diﬀerentiate the function

f(x) = sec x.

Solution: We write the function as a quotient: f(x) = 1/ cos x. The

function is deﬁned for all x for which cos x = 0, i.e., for x not of the form

nπ + 1/2, where n is an integer. We apply the quotient rule, using that

cos

**x = −sin x (see Exercise 2 on page 17), and that the derivative of a
**

constant vanishes. We ﬁnd

sec

x =

sin x

cos

2

x

=

sin x

cos x

·

1

cos x

= tan xsec x. ♦ (1.30)

Suppose f and g are deﬁned and diﬀerentiable on an open set. Thinking

of f and g again more as functions, and not so much as functions evaluated

at a point, we may once more omit (x) from the notation. Then the product

rule and quotient rule become

(fg)

= f

g +fg

or

d

dx

(fg) =

df

dx

g +f

dg

dx

(1.31)

and, wherever g(x) = 0,

f

g

=

f

g −fg

g

2

or

d

dx

f

g

=

df

dx

g −f

dg

dx

g

2

. (1.32)

Here g

2

is the square of the function g, given by g

2

(x) = [g(x)]

2

.

26 CHAPTER 1. BASIC CONCEPTS

1.13.3 Chain Rule

Let f and g be functions, and suppose that the domain of f contains the

range of g, so that the composition (f ◦ g)(x) = f(g(x)) is deﬁned for all x

in the domain of g. Set h = f ◦ g, so that h(x) = f(g(x)). The chain rule

says that whenever g is diﬀerentiable at x and f is diﬀerentiable at g(x),

then h(x) is diﬀerentiable at x and

h

(x) = (f ◦ g)

(x) = f

(g(x))g

(x). (1.33)

In Leibnitz’ notation the chain rule says that

dh

dx

(x) =

d

dx

f(g(x)) =

df

du

(g(x))

dg

dx

(x). (1.34)

Example 1.48. Diﬀerentiate the function

h(x) = e

x

2

+1

.

Solution: We write h = f ◦ g as a composition of two functions, with

g(x) = x

2

+ 1 and f(u) = e

u

. Remember that f

(u) = f(u) = e

u

and

g

(x) = 2x. In particular, f

(g(x)) = e

x

2

+1

. The chain rule tells us that

h

(x) = f

(g(x))g

(x) = 2xe

x

2

+1

.

In the last expression we reversed the order of the factors to make the

expression more readable. ♦

Example 1.49. Let u(x) be a diﬀerentiable function.

If f(x) = e

u(x)

then f

(x) = u

(x)e

u(x)

.

Here are some speciﬁc examples:

d

dx

e

2x+5

= 2 e

2x+5

d

dx

e

sin x

= cos x e

sin x

d

dx

e

tan x

= sec

2

x e

tan x

. ♦

Example 1.50. Combining Example 1.45 with the chain rule we ﬁnd

d

dx

u

n

(x) = nu

(x)u

n−1

(x)

1.13. RULES OF DIFFERENTIATION 27

for all integers n, assuming only that u is diﬀerentiable at x and u(x) = 0 if

n ≤ −1.

Solution: Set g(x) = u(x) and f(u) = u

n

. Then

h(x) = f(g(x)) = u

n

(x).

According to Example 1.45, f

(u) = nu

n−1

. The chain rule tells us now that

h

(x) = f

(g(x))g

(x) = n(g(x))

n−1

g

(x) = nu

(x)u

n−1

(x).

We reordered the expressions so that the expression is more readable.

To be speciﬁc, here are concrete examples:

d

dx

(3x + 5)

8

= 8(3x + 5)

8−1

· 3 = 24(3x + 5)

7

d

dx

(x

2

+ 1)

25

= 25(x

2

+ 1)

24

· 2x = 50x(x

2

+ 1)

24

d

dx

tan

3

x = 3 sec

2

xtan

2

x

d

dx

cos

2

x = 2 cos x(−sin x) = −2 cos xsin x

d

dx

sec

5

x = 5 sec

4

xsec xtan x = 5 sec

5

xtan x. ♦

Example 1.51. Diﬀerentiate the function ln |u| for u = 0.

Solution: We asserted that ln

**u = 1/u for positive values of u, see
**

(1.19). So, suppose that u < 0. Then u = −|u| and ln |u| = ln(−u). The

chain rule tells us that, for u < 0,

d

du

ln |u| =

1

|u|

d

du

(−u) = (−1)

1

−u

=

1

u

.

This means that for all non-zero u

d

du

ln |u| =

1

u

. ♦ (1.35)

More generally, envoking the chain rule

d

dx

ln|u(x)| =

u

(x)

u(x)

, (1.36)

assuming that u is diﬀerentiable and nowhere zero on its domain. E.g.,

d

dx

ln |x

2

−4| =

2x

x

2

−4

28 CHAPTER 1. BASIC CONCEPTS

for all x = ±2.

We push matters a bit further. We use the formulae for diﬀerentiating

the exponential and natural logarithm functions. Eventually we will verify

them independently.

Consider a function u which is diﬀerentiable and nowhere zero on its

domain and q any real number. Then

If f(x) = |u(x)|

q

then f

(x) = q

u

(x)

u(x)

|u(x)|

q

. (1.37)

The assertion follows from (1.36), Example 1.49 and the exponential laws.

f

(x) =

d

dx

e

ln f(x)

=

d

dx

e

ln(|u(x)|

q

)

=

d

dx

e

q ln|u(x)|

=

¸

d

dx

(q ln |u(x)|)

e

q ln |u(x)|

= q

u

(x)

u(x)

|u(x)|

q

.

Here is a concrete example:

d

dx

1

2

−sin x

5

= 5

−cos x

1

2

−sin x

1

2

−sin x

5

whenever sin x = 1/2. Speciﬁcally, we have to exclude all x of the form

π

6

+ 2nπ and

5π

6

+ 2nπ, where n is an arbitrary integer.

For diﬀerentiable functions which are everywhere positive on their do-

main and any real number q the diﬀerentiation formula in (1.37) specializes

to

d

dx

u

q

(x) = qu

(x)u

q−1

(x). (1.38)

For example:

d

dx

(sin x)

1/2

=

cos x

2

√

sin x

for x ∈ (0, π) and

d

dx

(sec

2

x + 5)

π

= 2π sec

2

xtan x(sec

2

x + 5)

π−1

for x ∈ (−π/2, π/2).

1.13. RULES OF DIFFERENTIATION 29

Using the tricks from above, we get the following derivatives:

d

dx

a

x

= a

x

ln a (Assume a > 0. Hint: a

x

= e

x lna

)

d

dx

x

x

= (1 + ln x)x

x

(Assume x > 0, x = 1. Hint: x

x

= e

x lnx

)

d

dx

x

sinx

=

sin x

x

+ cos xln x

x

sinx

(Assume x ∈ (0, π/4)).

To diﬀerentiate a composition of more than two diﬀerentiable functions

we apply the chain rule repeatedly. E.g.,

d

dx

f(g(h(x))) = f

(g(h(x))

d

dx

g(h(x)) = f

(g(h(x)) · g

(h(x)) · h

(x).

For example

d

dx

e

√

x

2

+1

= e

√

x

2

+1

·

1

2

√

x

2

+ 1

· 2x =

xe

√

x

2

+1

√

x

2

+ 1

d

dx

tan

3

(5x

2

−x + 5) = 3 tan

2

(5x

2

−x + 5) sec

2

(5x

2

−x + 5) · (10x −1)

1.13.4 Hyperbolic Functions

The exponential function may be used to deﬁne the hyperbolic sine and

cosine.

sinhx =

1

2

e

x

−e

−x

& cosh x =

1

2

e

x

+e

−x

(1.39)

You are invited to verify that

cosh

2

x −sinh

2

x = 1.

Conversely, one can show that any point (u, v) on the hyperbola

u

2

−v

2

= 1

can be expressed as (±cosh x, sinh x) for some x ∈ (−∞, ∞). These obser-

vations motive the attribute ‘hyperbolic’.

It is elementary to compute the derivatives of the hyperbolic functions:

sinh

x = cosh x and cosh

x = sinhx.

30 CHAPTER 1. BASIC CONCEPTS

One may also deﬁne other hyperbolic functions

tanh x =

sinhx

cosh x

, coth x =

cosh x

sinhx

, sech x =

1

cosh x

, and csch x =

1

sinh x

.

As a routine application of the rules of diﬀerentiation, you may calculate

the derivatives of these functions. There are identities for these hyperbolic

functions, comparable to the identities for the trigonometric functions. You

can ﬁnd them in any table of mathematical formulas, or you can work them

out yourself.

1.13.5 Derivatives of Inverse Functions

Let us recall. Two functions f and g are said to be inverses of each other

(or each function is the inverse of the other one) if the domain of f is equal

to the range of g, the domain of g is equal to the range of f, and

g(f(x)) = x and f(g(y)) = y (1.40)

for all x in the domain of f and all y in the domain of g. A few essential

properties of inverse functions are listed in

Proposition 1.52. Suppose f and g are inverses of each other.

1. The graph of g is obtained from the graph of f by reﬂection at the

diagonal.

2. If f is increasing, then so is g. If f is decreasing, then so is g.

3. If f is continuous, then it is monotonic (increasing or decreasing) on

any interval in its domain.

4. If f is continuous and I is an interval in the domain of f, then J =

f(I), the image of I under the map f, is an interval. If I is an open

interval, then J is an open interval.

Some parts of this proposition are elementary, others are consequences

of the intermediate value theorem.

For example, the function f(x) = cos x maps the interval [0, π] to the

inteval [−1, 1]. The function f(x) = e

x

maps the interval (−∞, ∞) to the

interval (0, ∞). The function tan x maps the interval (−π/2, π/2) to the

interval (−∞, ∞). It is customary to deﬁne its inverse arctan x as a func-

tion from (−∞, ∞) to (−π/2, π/2). The function sin x maps the interval

[−π/2, π/2] to the interval [−1, 1]. Its inverse arcsin x is typically used with

1.13. RULES OF DIFFERENTIATION 31

-1.5 -1 -0.5 0.5 1 1.5

-1

-0.5

0.5

1

Figure 1.9: sin x on [−π/2, π/2]

-1 -0.5 0.5 1

-1.5

-1

-0.5

0.5

1

1.5

Figure 1.10: arcsin y on [−1, 1]

domain [−1, 1], and its range is [−π/2, π/2]. You see the graph of these two

functions in Figures 1.9 and 1.10.

The relation between the derivative of a function and its inverse is spelled

out in our next theorem.

Theorem 1.53. Let f be a diﬀerentiable and invertible function which is

deﬁned on an open interval (a, b), and denote the image of f by (A, B).

Denote the inverse of f by g. Then g is diﬀerentiable at all points y ∈ (A, B)

for which f

**(g(y)) = 0. For these values of y and for x such that f(x) = y
**

the derivative is given by:

g

(y) =

1

f

(g(y))

or g

(f(x)) =

1

f

(x)

.

Proof. We will not give a formally complete proof of the diﬀerentiability

assertion. Still, if the line t(x) is close to the graph of the function f(x) at

the point (x, f(x)) and y = f(x), then its reﬂection T(x) at the diagonal

is close to the graph of the function g(x) at the point (f(x), x) = (y, g(y)).

We need that T(x) is not vertical, and this is assured by the assumption

that t(x) is not horizontal. With the role of x and y being interchanged,

the slope of t(x) is the reciprocal of the slope of T(x). This provides the

formula for the derivative. Actually, this is also easy to calculate.

By deﬁnition we have f(g(y)) = y for all y ∈ (A, B). Diﬀerentiate both

sides of the equation. We ﬁnd

f

(g(y))g

(y) = 1 and g

(y) =

1

f

(g(y))

,

32 CHAPTER 1. BASIC CONCEPTS

as claimed. If y = f(x), then g(y) = g(f(x)) = x, and we obtain the second

version of the formula for the derivative of the inverse of the function:

g

(f(x)) =

1

f

(x)

.

We apply the theorem to ﬁnd some important derivatives.

Example 1.54. Assume that the natural logarithm function is diﬀeren-

tiable and that ln

**x = 1/x, as asserted in (1.19). Show that the exponential
**

function is diﬀerentiable and that

d

dy

e

y

= e

y

.

Solution: By deﬁnition, the exponential function is the inverse of the

natural logarithm function ln. Set f(x) = ln x and g(y) = e

y

in Theo-

rem 1.53. We note that ln

**(x) = 0 for all x in (0, ∞), the domain of the
**

natural logarithm. The theorem says that the exponential function is dif-

ferentiable and provides the formula for the derivative:

d

dy

e

y

=

1

ln

(e

y

)

=

1

1/e

y

= e

y

,

as claimed. ♦

Example 1.55. Show that the function g(y) = arctan y (the inverse of

f(x) = tan x) is diﬀerentiable, and that

d

dy

arctan y =

1

1 +y

2

.

According to standard conventions we use (−∞, ∞) as the domain and

(−π/2, π/2) as the range for arctan.

Solution: The function f(x) = tan x is diﬀerentiable on its entire do-

main, and f

(x) = sec

2

x is nowhere zero. Theorem 1.53 tells us that

g(y) = arctan y is diﬀerentiable on its entire domain (−∞, ∞). The the-

orem also provides us with the formula for the derivative:

arctan

(y) =

1

tan

(arctan y)

=

1

sec

2

(arctan y)

= cos

2

(arctan y).

All we need to do now is to ﬁgure out what cos

2

(arctan y) is. To do this

we draw a triangle in which we identify the available data. We refer to the

notation in Figure 1.11.

1.13. RULES OF DIFFERENTIATION 33

y

1

u

A B

Figure 1.11: An informative triangle

There you see a rectangular triangle, the right angle is at the vertex

B. The angle at the vertex A is called u. The adjacent side to this angle is

chosen to be of length 1, and the opposing side of length y. So, by deﬁnition,

tan u = y and arctan y = u.

By the theorem of Pythagoras, the length of the hypotenuse is

1 +y

2

.

Then

cos u =

1

1 +y

2

and cos

2

(arctan y) =

1

1 +y

2

.

The conclusion is that

arctan

(y) =

1

1 +y

2

. (1.41)

This is exactly what we claimed. ♦

Combined with the chain rule, and assuming the diﬀerentiability of u(x),

we ﬁnd a slightly more general formula:

d

dx

arctan(u(x)) =

u

(x)

1 +u

2

(x)

. (1.42)

34 CHAPTER 1. BASIC CONCEPTS

For example:

d

dx

arctan(x

2

+ 5) =

2x

1 + (x

2

+ 5)

2

d

dx

arctan(sin x) =

cos x

1 + sin

2

x

.

The reader is invited to verify the formulas for the other inverse trigono-

metric functions arcsin x, arccos x, arccot x, and arcsec x as they are given

in Table 1.3 on page 63. For example

Exercise 3. It is customary to think of arcsin x as a function from [−1, 1]

to [−π/2, π/2]. Show that arcsin x is diﬀerentiable on (−1, 1), and that its

derivative is

d

dx

arcsin x =

1

√

1 −x

2

.

We may once more improve on this formula. Let u(x) be a diﬀerentiable

function which is deﬁned on an open interval, and suppose that |u(x)| < 1.

Then, using the chain rule, we ﬁnd that

d

dx

arcsin(u(x)) =

u

(x)

1 −u

2

(x)

. (1.43)

For example:

d

dx

arcsin(3x) =

3

√

1 −9x

2

if x ∈ (−1/3, 1/3)

d

dx

arcsin(x

2

) =

2x

√

1 −x

4

if x ∈ (−1, 1)

1.13.6 Implicit Diﬀerentiation

Until now we considered functions which were given explicitly. I.e., we were

given an equation y = f(x), where f(x) is some instruction which assigns a

value to x. The points on the graph of f are the points which satisfy the

equation. Consider the equation

(x

2

+y

2

)

2

= x

2

−y

2

. (1.44)

The solutions of this equation form a curve

5

in the plane called a lemniscate,

see Figure 1.12. Parts of this curve look like the graph of a function, such

1.13. RULES OF DIFFERENTIATION 35

-1 -0.5 0.5 1

-0.3

-0.2

-0.1

0.1

0.2

0.3

Figure 1.12: Lemniscate

as the points for which y ≥ 0. Without solving the equation for y, we still

like to calculate the slope of curve at one of its points. This process is called

implicit diﬀerentiation.

Let us start out with an example which we have studied before.

Example 1.56. The unit circle consists of all points which satisfy the equa-

tion x

2

+y

2

= 1. Find the slope of the tangent line to the unit circle at the

point (1/2,

√

3/2).

Solution: We write y = y(x) to emphasize that y as a function of x.

Diﬀerentiating both sides of the equation of the circle we get

2x + 2y

dy

dx

= 0 or

dy

dx

=

−x

y

.

Plugging in the coordinates of the speciﬁed point, we ﬁnd that

dy

dx

(1/2,

√

3/2)

=

−1

√

3

.

We used a diﬀerent way to indicate at which point we evaluate the derivative

because we had to specify the x and the y coordinate of the point. ♦

5

We will rely on the readers intuitive idea of a curve in the plane.

36 CHAPTER 1. BASIC CONCEPTS

Example 1.57. Find the slope of the tangent line to the lemniscate

(x

2

+y

2

)

2

= x

2

−y

2

,

and ﬁnd the coordinates of the points where the tangent line is horizontal.

Solution: You see a picture of the lemniscate in Figure 1.12. As in

Example 1.56, we consider y as a function of x and diﬀerentiate both sides

of the equation. We ﬁnd

2(x

2

+y

2

)(2x + 2y

dy

dx

) = 2x −2y

dy

dx

.

Bring all terms with a factor dy/dx to the left hand side of the equation and

those without to the right hand side.

(2y(x

2

+y

2

) +y)

dy

dx

= x(1 −2(x

2

+y

2

)).

Finally we get an explicit expression for

dy

dx

in terms of x and y:

dy

dx

=

x(1 −2(x

2

+y

2

))

2y(x

2

+y

2

) +y

=

x(1 −2(x

2

+y

2

))

y(2(x

2

+y

2

) + 1)

.

Given any point (x, y) with y = 0 on the lemniscate, we can plug it into the

expression for

dy

dx

and we get the slope of the curve at this point.

E.g, the point (x, y) = (

1

2

,

1

2

−3 + 2

√

3) is a point on the lemniscate,

and at this point the slope of the tangent line is

dy

dx

=

2 −

√

2

√

3

−3 + 2

√

3

.

This speciﬁc calculation takes a bit of arithmetic skill and eﬀort to carry

out.

The tangent line is horizontal whenever

dy

dx

= 0. A quick look at Fig-

ure 1.12 tells us that we may ignore points where x = 0 or y = 0. That

means that

dy

dx

= 0 whenever

1 −2(x

2

+y

2

) = 0 or x

2

+y

2

=

1

2

.

Substitute x

2

+ y

2

=

1

2

, and y

2

=

1

2

− x

2

into the equation of the curve.

Then we get an equation in one variable:

1

4

= x

2

−

1

2

−x

2

or x

2

=

3

8

and y

2

=

1

8

.

1.13. RULES OF DIFFERENTIATION 37

The points at which the tangent line to the lemniscate is horizontal are

(x, y) = (±

√

6

4

, ±

√

2

4

) ≈ (±.6124, ±.3536). ♦

Example 1.58. Suppose you drop a circle of radius 1 into a parabola with

the equation y = 2x

2

. At which points will the circle touch the parabola?

6

-1 -0.5 0.5 1

0.5

1

1.5

2

2.5

3

Figure 1.13: Ball in a Cup.

Solution: You see a picture of the problem in Figure 1.13. The crucial

observation in this example is, that the tangent line to the parabola and the

circle will be the same at the point of contact.

Suppose the coordinates of the center of the circle are (0, a), then its

equation is x

2

+ (y − a)

2

= 1. Diﬀerentiating the equation of the parabola

with respect to x, we ﬁnd that

dy

dx

= 4x. Diﬀerentiating the equation of the

circle with respect to x, we get

2x + 2(y −a)

dy

dx

= 0.

Assuming that

dy

dx

is the same for both curves at the point of contact, we

substitute

dy

dx

= 4x into the second equation. After some impliﬁcations we

6

More sensibly, drop a ball of radius 1 into a cup whose vertical cross section is the

parabola y = 2x

2

.

38 CHAPTER 1. BASIC CONCEPTS

ﬁnd:

x(1 + 4(y −a)) = 0.

The ball it too large to ﬁt into the parabola and touch at (0, 0). So we may

assume that x = 0. Solving the equation 1 + 4(y − a) = 0 for y, we ﬁnd

that the y coordinate of the point of contact is y = a −

1

4

. We substitute

this expression into the equation of the circle and ﬁnd that the x coordinate

of the point of contact is x = ±

√

15

4

. Substituting this into the equation of

the parabola, we ﬁnd that y =

15

8

at the point of contact. In summary, the

circle touches the parabola in the points

(x, y) =

±

√

15

4

,

15

8

. ♦

Exercise 4. Consider the curve given by the equation

x

3

+y

3

= 1 + 3xy

2

.

Find the slope of the curve at the point (x, y) = (2, −1).

Exercise 5. Consider the curve given by the equation x

2

= sin y. Find the

slope of the curve at the point with coordinates x = 1/

4

√

2 and y = π/4.

Exercise 6. Repeat Example 1.57 with the curve given by the equation

y

2

−x

2

(1−x

2

) = 0. You ﬁnd a picture of this Lissajous ﬁgure in Figure 1.14.

1.14 Related Rates

Many times you encounter situations in which you have two related variables,

you know at which rate one of them changes, and you like to know at which

rate the other one changes. In this section we treat such problems.

Example 1.59. Suppose the radius of a ball changes at a rate of 2 cm/min.

At which rate does its volume change when r = 20 cm?

Solution: Denote the volume of the ball by V and its radius by r. We

use t to denote the time variable. We consider V as a function of r as well

as t. The formula for the volume of a ball is V (r) =

4π

3

r

3

. According the

the chain rule:

dV

dt

=

dV

dr

dr

dt

= 4πr

2

dr

dt

.

With r = 20 and

dr

dt

= 2 we get

dV

dt

= 3200π cm

3

/min. This is the rate at

which the volume of the ball changes with respect to time. ♦

1.14. RELATED RATES 39

-1 -0.5 0.5 1

-0.4

-0.2

0.2

0.4

Figure 1.14: y

2

−x

2

(1 −x

2

) = 0

Example 1.60. Suppose a particle moves on a circle of radius 10 cm and

centered at the origin (0, 0) in the Cartesian plane. At some time the particle

is at the point (5, 5

√

3) and moves downwards at a rate of 3 cm/min. At

which rate does it move in the horizontal direction?

Solution: The equation of the circle is x

2

+y

2

= 100. We consider both

variables, x and y, as functions of the time variable t. Implicit diﬀerentiation

of the equation of the circle gives us the equation

2x

dx

dt

+ 2y

dy

dt

= 0.

In the given situation x = 5, y = 5

√

3, and

dy

dt

= −3. We ﬁnd that

dx

dt

= 3

√

3,

so that the particle is moving to the right at a rate of 3

√

3 cm/min. ♦

Example 1.61. Pressure (P) and volume (V ) of air at room temperature

are related by the equation

7

PV

1.4

= C.

7

Boyle-Mariotte described the relation between the pressure and volume of a gas. They

derived the equation PV

γ

= C. It is called the adiabatic law. The constant γ depends on

the molecular structure of the gas and the temperature. For the purpose of this problem,

we suppose that γ = 1.4 for air at room temperature.

40 CHAPTER 1. BASIC CONCEPTS

Here C is a constant. At some instant t

0

the pressure of the gas is 25 kg/cm

2

and the volume is 200 cm

3

. Find the rate of change of P if the volume

increases at a rate of 10 cm

3

/min.

Solution: We consider P as a function of V . Diﬀerentiation of the

equation yields

dP

dV

V

1.4

+ 1.4PV

.4

= 0 or

dP

dV

= −

1.4P

V

.

According to the chain rule

dP

dt

=

dP

dV

dV

dt

= −

1.4P

V

dV

dt

.

Substituting the given information we ﬁnd that the pressure decreases at a

rate of 1.75 kg/cm

2

sec. ♦

Example 1.62. The mass M of a particle at velocity v, as perceived by an

observer in resting position, is

M(v) =

m

1 −v

2

/c

2

,

where m is that mass at rest and c is the speed of light. This formula is from

Einstein’s special theory of relativity. At which rate is the mass changing

when the particle’s velocity is 90% of the speed of light, and increasing at

.001c per second?

Solution: According to our rules of diﬀerentiation

dM

dv

=

mvc

(c

2

−v

2

)

3/2

.

Applying the chain rule and substituting the values, we ﬁnd

dM

dt

=

dM

dv

dv

dt

=

mvc

2

1000(c

2

−v

2

)

3/2

=

9

√

19m

3610

≈ .010867m.

The perceived mass increases at a rate of approximately 1% of its mass at

rest. ♦

Exercise 7. A ladder, 7 m long, is leaning against a wall. Right now the

foot of the latter is 1 m away from the wall. You are pulling the foot of the

ladder further away from the wall at a rate of .1 m/sec. At which rate is

the top of the ladder sliding down the wall?

1.15. EXPONENTIAL GROWTH AND DECAY 41

1.15 Exponential Growth and Decay

An idealistic, but very useful model for population growth is the Malthusian

Law

A

(t) = aA(t). (1.45)

It says that the rate of change of a population is proportional to its size. We

denoted the proportionality factor by a. We saw that the functions A(t) =

Ce

at

are solutions of this equation, and it can be shown that on an interval

any solution is of this form. We also say that A(t) grows exponentially and

a is the relative growth rate.

The equation in (1.45) is an example of a diﬀerential equation, an equa-

tion which involves a function and its derivatives, and the unknown is a

function.

We may specify the value of A at some time t

0

, say A

0

= A(t

0

). Then

we have an initial value problem

A

(t) = aA(t) and A

0

= A(t

0

). (1.46)

Theorem 1.63. On an interval which contains t

0

the function

A(t) = A

0

e

a(t−t

0

)

is the unique solution

8

of the initial value problem in (1.46).

The essential aspects of dealing with (1.46) are addressed in

Example 1.64. Suppose the size of a population of bacteria in a laboratory

experiment is C

1

= 5, 000 at time t

1

= 2 and C

2

= 7, 000 at time t

2

= 5.

Here time is measured in hours since the beginning of the experiment.

1. Find the relative growth rate a of the population.

2. Find the formula for the size of the population at any time t ≥ 0.

3. Predict the size of the population at time t = 10.

4. Find the time at which the population reaches 8, 000.

5. Find the time within which the population doubles

9

.

8

That the function satisﬁes the diﬀerential equation follows from (1.19), which we still

need to prove. The uniqueness assertion follows from Proposition 2.9 on page 68.

9

Note that the doubling time depends only on the relative growth rate a.

42 CHAPTER 1. BASIC CONCEPTS

Solution: We denote the size of the population at time t by A(t). The

theorem tells us that A(t) = A

0

e

a(t−t

0

)

, where A

0

= A(t

0

).

1. To calculate the relative growth rate a observe that

C

2

C

1

=

A

0

e

a(t

2

−t

0

)

A

0

e

a(t

1

−t

0

)

= e

a(t

2

−t

1

)

and ln

C

2

C

1

= a(t

2

−t

1

).

We ﬁnd that

a =

ln C

2

−ln C

1

t

2

−t

1

=

ln 1.4

3

≈ .11.

The population grows at a rate of about 11% per hour.

2. and 3. The size of the population at any time t ≥ 0 is

A(t) = 5000e

a(t−2)

,

where a is as above. Substituting t = 10 we ﬁnd that A(10) ≈ 12, 264.

4. Suppose the size of the population reaches 8, 000 at time t

1

, then

8000 = 5000e

a(t

1

−2)

or ln(8/5) = a(t

1

−2) or t

1

=

ln(1.6)

a

+ 2 ≈ 6.2

The size of the population reaches 8, 000 about 6.2 hours into the experiment.

5. Suppose at some time t

0

the size of the population is A

0

= A(t

0

) and

T hours later the size of the population is 2A

0

= A(t

0

+T). Then

A(t

0

+T) = A

0

e

aT

= 2A

0

or e

aT

= 2 and aT = ln 2.

Thus the doubling time is T =

ln 2

a

≈ 6.18 hours.

Consider a radioactive substance. Experiments have shown that the

rate at which radioactive decays occur is proportional to the amount of

radioactive material present. This rate is proportional to the rate at which

the amount of the material decreases. Suppose t denotes time and A(t) the

amount of radioactive substance at time t. The experience which we just

described can be expressed as a diﬀerential equation

A

(t) = −kA(t). (1.47)

The minus sign in the equation is included so that k will be positive. The

half-life T of a radioactive substance is the time within which half of it

decays. As in the computation of the doubling time in the previous example,

one ﬁnds

T =

ln 2

k

. (1.48)

1.16. MORE EXPONENTIAL GROWTH AND DECAY 43

In the late 1940ies Willard Libby invented (or discovered) the method of

carbon-14 dating. He was awarded the Nobel price for it. In brief, the idea is

as follows. Carbon-14 occurs naturally in the atmosphere, and the amount

is believed to have been essentially constant for a long time (until recent

nuclear testing). All living organisms absorb it. Within a living organism

there is an equilibrium. The amount which is absorbed equals the amount

which decays. The level of the equilibrium is characteristic for the organism,

or a part thereof (e.g. wood from an oak or a human bone). After death

no more carbon-14 is absorbed, and the carbon-14 which was present at the

time of death decays. The half-life of carbon-14 has been determined to be

about 5568 years. For many organisms one also knows how many carbon-14

decays to expect at the time of death. Measuring the number of decays in

a dead organism allows us to determine the time of death, approximately.

We explain the process in a numerical example.

Example 1.65. Suppose we measure 6.68 carbon-14 decays per minute and

gram in a certain kind of wood at the time of death of the tree. Suppose

dead wood of the same kind shows 1.8 decays per minute and gram. How

long ago did the tree die?

Solution: Let t

0

= 0 be the time of death of the tree, and t

1

the present

time, measured in years. The number of decays to be expected t years after

death is

A(t) = 6.68e

−

ln 2

5568

t

.

We have that A(t

1

) = 1.8. From this we calculate:

ln

1.8

6.68

= −

ln 2

5568

t

1

or t

1

= −

5568

ln 2

ln

1.8

6.68

≈ 10, 534. (1.49)

The tree died approximately 10, 500 years ago.

1.16 More Exponential Growth and Decay

More generally than in (1.45), consider the diﬀerential equation

f

(t) = af(t) +b, (1.50)

where a and b are constants, and a = 0. A time independent solution (steady

state solution) of this equation is f(t) = −b/a.

44 CHAPTER 1. BASIC CONCEPTS

Theorem 1.66. Functions of the form

f(t) = ce

at

−

b

a

are solutions of the diﬀerential equation in (1.50). Here c denotes an arbi-

trary constant. On an interval every solution of (1.50) is of this form.

We obtain a unique solution if we add an initial condition to the diﬀerential

equation in (1.50).

Theorem 1.67. On an interval which contains t

o

, the function

f(t) =

y

0

+

b

a

e

a(t−t

0

)

−

b

a

is the unique solution of the initial value problem

f

(t) = af(t) +b and f(t

0

) = y

0

.

Remark 2. It is not hard to verify that the given functions are solutions of

the respective problems. The uniqueness assertion is a minor modiﬁcation

of Proposition 2.9 on page 68.

Let us apply these ideas to solve some problems. The important aspects

are to translate the given information into a mathematical equation. The

rest will be routine calculation.

Example 1.68. On graduation day the balance of your student loan is

$15,000. Interest is added at a rate of .5% per month, and you are repaying

the loan at a rate of $ 200.00 per month. Analyze the future of the loan.

Solution: As variable we use time, denoted by t and measured in

months. We set t = 0 at the time of graduation. This is the time at

which you start to repay the loan. Denote the balance of your loan at time

t by B(t). The balance increases at a rate of .005B(t) due to interest being

added and decreases at a rate of $200.00 per month due to payments which

you make. In summary, we have the initial value problem

B

**(t) = .005B(t) −200 and B(0) = 15, 000.
**

According to Theorem 1.67 the solution of the initial value problem is

B(t) =

15, 000 +

−200

.005

e

.005t

−

−200

.005

= −25, 000e

.005t

+ 40, 000.

1.16. MORE EXPONENTIAL GROWTH AND DECAY 45

For example, B(T) = 0 if

T =

1

.005

ln

40

25

≈ 94.

After approximately 94 months (7 years and 10 months) you repaid the

loan. Your total payments were $18,800, so that you paid the principal plus

$3,800 in interest. ♦

Example 1.69. You are absorbing a medication at a rate of 3 mg per hour.

(You can keep this rate constant with a skin patch.) The liver metabolizes

the medication at a rate of 4% per hour. Analyze the amount of medication

in your body at any time.

Solution: We use time as independent variable, denote it by t and

measure it in hours. We denote by t = 0 the time when we start taking

the medication. Let A(t) denote the amount of medication in your body,

measured in milligrams. Then A(t) increases at a rate of 3 mg per hour

because you are taking in medication and at the same time A(t) decreases

at a rate of .04A(t) due to your liver metabolizing the medication. We have

the initial value problem

A

(t) = −.04A(t) + 3 and A(0) = 0.

The solution of this problem is

A(t) = −75e

−.04t

+ 75.

For example, after 12 hours there will be about 28.6 milligram of medication

in your body. It will take slightly more than 40 hours before the amount of

medication in your body reaches 60 milligrams. The steady state solution

of the problem is A(t) = 75. The amount of medication will stabilize at this

amount with time. ♦

Example 1.70 (Newton’s Law of Cooling). Suppose you have an ob-

ject whose temperature is diﬀerent from the temperature of its surround-

ings. With time, the temperature of the object will approach the one of

its surroundings. We discuss how this happens, at least under idealized

circumstances.

Think of a cup of coﬀee. You stir the coﬀee gently so that the tem-

perature in the cup remains homogeneous and almost no energy is added

46 CHAPTER 1. BASIC CONCEPTS

through the process of stirring.

10

Denote the temperature of the coﬀee by

T. It is a function of time, so that we write T(t). Newton’s law of cooling

says that the rate at which the heat is transferred, and with this the rate

of change of temperature of the coﬀee, is proportional to the temperature

diﬀerence. If K is the temperature of the surroundings, then

T

(t) = a(T(t) −K) = aT(t) −aK. (1.51)

Setting b = −aK, this is the diﬀerential equation in (1.50).

Let us work out a numerical example. At time t = 0, just after you

poured the coﬀee into your cup, its temperature is 95 degrees Celsius. Five

minutes later the temperature has dropped to 80 degree, while you stir it

slightly and patiently. The room temperature is 25 degrees.

1. Determine the function T(t).

2. Find t

1

, such that T(t

1

) = 70 degrees Celsius.

Solution: To apply Theorem 1.67, we set t

0

= 0, y

0

= 95, and K = 25.

Note that −b/a = K. Putting all of this into the formula for the solution of

the initial value problem, we get that

T(t) = (95 −25)e

at

+ 25 = 70e

at

+ 25.

To determine a we use that

T(5) = 80 = 70e

5a

+ 25,

and we conclude that a =

1

5

ln

55

70

**≈ −.0482. Using these data, Equa-
**

tion (1.51) says that the temperature of the coﬀee drops at a rate of about

.048 degrees per minute for each degree of diﬀerence between the tempera-

ture of the coﬀee and the room temperature. Having a numerical value for

a gives us an explicit expression for the temperature T as a function of t:

T(t) = 70e

−.0482t

+ 25.

10

The physics of heat transfer changes substantially if you take a solid object, such

as a turkey in the oven. The temperature in the solid object will not be homogenous,

the outside warms up much faster than the inside. In addition, the speciﬁc heat (the

amount of energy needed to increase the temperature of one unit of the material by one

degree) varies. It is diﬀerent for fat, protein, and bone. Furthermore, the speciﬁc heat

is highly temperature dependent for substances like protein. That means, a in (1.51)

depends on the temperature T. All of this leads to a signiﬁcantly diﬀerent development

of the temperature inside a turkey as you roast it for your Thanksgiving dinner.

1.16. MORE EXPONENTIAL GROWTH AND DECAY 47

We like to ﬁnd out the time t

1

for which

T(t

1

) = 70e

−.0482t

1

+ 25 = 70.

Solving the equation for t

1

, we ﬁnd that t

1

≈ 9.17. That means that the

temperature drops to 70 degrees approximately 9.17 minutes after pouring

it. ♦

Exercise 8. A chemical factory is located on the banks of a river. Down

stream from the factory is a lake, and the river is the only contributor to

the lake. Assume that the amount of water carried by the river is the same

all year around, and the amount of water in the lake is 10 times the amount

of water carried by the river per year. In negotiations which the EPA, the

owner has agreed to an acceptable level of 2.5 mg per m

3

of a pollutant in

the lake. After a major accident the level has risen to 15 mg per m

3

. As a

remedy, the factory owner proposes to reduce the emission of pollution so

that the level of pollutant in the river is only 1.5 mg per m

3

. It is assumed

that the pollutant is distributed uniformly in the lake at any time.

1. Let P(t) denote the amount of pollutant (measured in mg per m

3

) in

the lake at time t. Let t

0

= 0 be the time just after the accident and

at which the clean-up strategy is implemented. State the initial value

problem for P(t).

2. Find the function P(t).

3. At which time will the level of pollution be back to 2.5 mg per m

3

?

Exercise 9. The population of an endangered species of birds on Kauai

decreases at a relative rate of 25% per year. Currently, at time t

0

= 0, the

population is estimated to be 700 birds. A government agency raises the

species in captivity and releases birds into the wild at a rate of 80 birds per

year. Denote the size of the population at time t by P(t), where t denotes

time and is measured in years.

1. State the initial value problem for P(t).

2. Find the function P(t).

3. At which time will the population drop to 500 birds?

4. What is the long term estimate for the population of this species in

the wild?

48 CHAPTER 1. BASIC CONCEPTS

1.17 The Second and Higher Derivatives

Let f(x) be a function which is deﬁned on an open set. If the function is

diﬀerentiable at each point of its domain, then f

**(x) is again a function with
**

the same domain as f(x). We may ask whether the function f

(x) is dif-

ferentiable. Its derivative, wherever it exists, is called the second derivative

of f. It is denoted by f

**(x). This process can be iterated. The deriva-
**

tive of the second derivative is called the third derivative, and denoted by

f

**(x), etc. We will make use of the second derivative. Leibnitz’s notation
**

for the second derivative of a function f(x) is d

2

f/dx

2

. Here is a sample

computation in which you are invited to ﬁll in the details:

d

2

dx

2

e

sin x

=

d

dx

cos xe

sin x

= (−sin x + cos

2

x)e

sin x

.

Exercise 10. Find the second derivatives of the following functions:

(1) f(x) = 3x

3

+ 5x

2

(2) g(x) = sin 5x

(3) h(x) =

x

2

+ 2

(4) i(x) = e

5x

(5) j(x) = tan x

(6) k(x) = cos(x

2

)

(7) l(x) = ln 2x

(8) m(x) = ln(x

2

+ 3)

(9) n(x) = arctan 3x

(10) o(x) = sec(x

3

)

(11) p(x) = ln

2

(x + 4)

(12) q(x) = e

cos x

(13) r(x) = ln(tan x)

(14) s(x) = e

x

2

−1

(15) t(x) = sin

3

x.

1.18 Numerical Methods

In this section we introduce some methods for numerical computations.

Their common feature is, that for a diﬀerentiable function we do not make

a large error when we use the tangent line to the graph instead of the graph

itself. This rather casual statement will become clearer when you look at

the individual methods.

1.18.1 Approximation by Diﬀerentials

Suppose x

0

is an interior point of the domain of a function f(x) and f(x)

is diﬀerentiable at x

0

. Assume also that f(x

0

) and f

(x

0

) are known. The

method of approximation by diﬀerentials provides an approximate values

f(x

1

) if x

1

is near x

0

. We use the symbol ‘≈’ to stand for ‘is approximately’.

One uses the formula

f(x

1

) ≈ f(x

0

) +f

(x

0

)(x

1

−x

0

). (1.52)

1.18. NUMERICAL METHODS 49

On the right hand side in (1.52) we have l(x

1

), the tangent line to the graph

of f(x) at (x

0

, f(x

0

)) evaluated at x

1

. In the sense of the deﬁnition of the

tangent line in Section 1.6, f(x

1

) is close to l(x

1

) for x

1

near x

0

.

Example 1.71. Find an approximate value for

3

√

9.

Solution: We set f(x) =

3

√

x, so we are supposed to ﬁnd f(9). Note

that

f

(x) =

1

3

x

−2/3

, f(8) = 2, and f

(8) =

1

12

.

Formula (1.52), applied with x

1

= 9 and x

0

= 8, says that

3

√

9 = f(9) ≈ 2 +

1

12

(9 −8) =

25

12

≈ 2.0833.

Your calculator will give you

3

√

9 ≈ 2.0801. ♦

Example 1.72. Find an approximate value for tan 46

◦

.

Solution: We carry out the calculation in radial measure. Note that

46

◦

= 45

◦

+ 1

◦

, and this corresponds to π/4 + π/180. Use the function

f(x) = tan x. Then f

(x) = sec

2

x, f(π/4) = 1, and f

(π/4) = 2. Formula

(1.52), applied with x

1

= (π/4 +π/180) and x

0

= π/4 says

tan 46

◦

= tan

π

4

+

π

180

≈ tan

π

4

+ sec

2

π

4

π

180

= 1 +

π

90

≈ 1.0349.

Your calculator will give you tan 46

◦

≈ 1.0355. ♦

Exercise 11. Use approximation by diﬀerentials to ﬁnd approximate values

for

(1)

5

√

34 (2) tan 31

◦

(3) ln 1.2 (4) arctan 1.1.

In each case, compare your answer with one found on your calculator.

We have been causal in (1.52) insofar as we have not estimated (provided

an upper bound for) the error which we make using the right hand side of

(1.52) instead of of the actual value of the function on the left hand side.

The inequality in Deﬁnition 1.25 on page 12 provides us with an estimate.

According to this slightly more demanding deﬁnition, diﬀerentiability of the

function f(x) means that there exist numbers A and d > 0, such that

|f(x

1

) −[f(x

0

) +f

(x

0

)(x

1

−x

0

)| ≤ A(x

1

−x

0

)

2

whenever |x

1

−x

0

| < d. Thus, if we know A and d, then we can approximate

the error as long as |x

1

−x

0

| < d.

50 CHAPTER 1. BASIC CONCEPTS

Example 1.73. Find an approximate value for sin31

◦

and estimate the

error.

Solution: Set f(x) = sinx. Then f

(x) = cos x, f(π/6) = 1/2, and

f

(π/6) =

√

3/2. Measuring angles in radians we set x

0

= π/6 and x

1

=

π/6 +π/180. Applying the formula in (1.52), we ﬁnd

sin 31

◦

≈ sin

π

6

+

π

180

cos

π

6

=

1

2

1 +

√

3

π

180

≈ .515115.

The calculator will tell that sin 31

◦

≈ .515038.

From the computation in Example 1.31 on page 17 we also know that we

may use A = 1 and d = π/4 in the diﬀerentiability estimate. The estimate

assures us that the error is at most

(x

1

−x

0

)

2

=

π

180

2

≤ .000305.

Comparison of the actual and approximate value conﬁrm this. ♦

Example 1.74. Use approximation by diﬀerentials to ﬁnd an approximate

value of

√

10 and give an upper bound for the error.

Solution: We use f(x) =

√

x and x

0

= 9. The f

(x) = 1/(2

√

x),

f(x

0

) = 3, and f

(x

0

) = 1/6. The formula in (1.52) tells us that

√

10 = f(10) ≈ f(9) +f

(9)(10 −9) = 3 +

1

6

≈ 3.16666.

The calculator will give you

√

10 ≈ 3.16228.

For the error estimate we may use

A =

1

2(

√

x

0

)

3

and any d > 0, see (1.16). The estimate assures us that the error is at most

1

2(

√

x

0

)

3

(x

1

−x

0

)

2

=

1

54

.

The actual error is again substantially less than this. ♦

Exercise 12. Use approximation by diﬀerentials to ﬁnd approximate values

for

(1) cos 28

◦

(2)

√

26 (3) sin 47

◦

.

In each case, estimate also the maximal error which you may have made by

using the method of approximation by diﬀerentials.

1.18. NUMERICAL METHODS 51

1.18.2 Newton’s Method

Newton’s method is designed to ﬁnd the zeros of a function. You have learned

how to solve linear and quadratic equations, i.e., ﬁnding the zeros of func-

tions of degree 1 and 2. More sophisticated methods allow you to ﬁnd the

exact solutions of polynomial equations of degree three and four. For poly-

nomials of degree greater or equal to 5 and most other functions there are

no general methods for ﬁnding their roots.

Newton’s method works as follows. Suppose we want to ﬁnd a zero of a

diﬀerentiable function f(x), i.e., we want to ﬁnd some x, such that f(x) = 0.

Suppose that by some means we know that such an x exists, and that x

0

is

not far from x. Then we set

x

1

= x

0

−

f(x

0

)

f

(x

0

)

, x

2

= x

1

−

f(x

1

)

f

(x

1

)

, x

3

= x

2

−

f(x

2

)

f

(x

2

)

, etc. (1.53)

and in general

x

n+1

= x

n

−

f(x

n

)

f

(x

n

)

. (1.54)

Geometry of Newton’s Method: Let us give a geometric explanation

for the formlas. Given any x

0

at which f is deﬁned and diﬀerentiable, we

obtain the tangent line l(x) to the graph of f at this point. Then x

1

, as

given in (1.53), is the point at which l(x) intersects the x-axis. Speciﬁcally,

l(x) = f

(x

0

)(x −x

0

) +f(x

0

), and l(x

1

) = 0 if x

1

= x

0

−

f(x

0

)

f

(x

0

)

.

This means that we accept that the tangent line is close to the graph of

the function, and instead of ﬁnding the zero of the function itself, we ﬁnd

the zero of the tangent line. The process is then iterated.

Let us calculate

√

A, i.e., the positive root of the function f(x) = x

2

−A.

Then f

(x) = 2x, and

x

n+1

= x

n

−

f(x

n

)

f

(x

n

)

= x

n

−

x

2

n

−A

2x

n

=

x

2

n

+A

2x

n

=

1

2

x

n

+

A

x

n

(1.55)

If we use A = 3 and x

0

= 2, then we ﬁnd

x

1

=

1

2

2 +

3

2

=

7

4

, x

2

=

1

2

7

4

+

12

7

=

97

56

and x

3

=

18817

10864

.

We summarize the computation in Table 1.1. In the ﬁrst column you ﬁnd

the subscript n. In the following two columns you ﬁnd the values of x

n

, once

52 CHAPTER 1. BASIC CONCEPTS

expressed as a fraction of integers, once in decimal form. In the last column

you see the square of x

n

. At least x

2

3

is rather close to 3. Your calculator

will give you 1.73205080757 as an approximate value of

√

3. You see that

our value for x

3

is rather precise. In fact, if you carry the calculation one

step further and ﬁnd x

4

, then the accuracy of this approximation of

√

3 will

exceed the accuracy of most calculators. The numbers in the last column

show that we are making rapid progress in ﬁnding a good approximation of

√

3.

n x

n

x

n

x

2

n

0 2 2.0000000000 4.0000000000

1 7/4 1.7500000000 3.0625000000

2 97/56 1.7321428571 3.0003188775

3 18817/10864 1.7320508100 3.0000000085

Table 1.1: The Babylonian Method

More than 4000 years ago the Babylonians used the outermost expres-

sions in (1.55)

x

n+1

=

1

2

x

n

+

A

x

n

**to ﬁnd good approximations of square roots, expressed as rational numbers.
**

We refer to the described procedure as the Babylonian method.

Let us consider one more example to illustrate Newton’s method. Find

a solution of the equation

xsin x = cos x.

Equivalently, we may say, ﬁnd a root of the function f(x) = xsinx −cos x.

Step 1: Let us make sure that there is a root of the function to be found.

Observe that f(0) = −1 < 0 and f(π/2) = π/2 > 0. The intermediate value

theorem tells us that f(x), as a continuous funtion, has a root in the interval

(0, π/2). Let us call this root x.

Step 2: Let us come up with a ﬁrst guess for a root. Considering the

values of f at the end points of the interval, we guess that x

0

= 1 is not

1.18. NUMERICAL METHODS 53

too far away from the root, which we know to exist by Step 1. Actually

f(1) ≈ .3.

Step 3: Let us improve the guess: Set

x

1

= x

0

−

f(x

0

)

f

(x

0

)

≈ .8645.

Your calculator will tell you that f(x

1

) ≈ .00874. You see that f(x

1

) is

much closer to zero than f(x

0

), and in this sense we expect that x

1

is much

closer to the root x of f(x) than x

0

. We made progress ﬁnding x.

Step 4: Repeat Step 3 and calculate x

2

, x

3

, . . . . The distance between

x and x

n

will decrease rapidly as n increases.

We explained Newton’s method because we want to illustrate the power

of the concept ‘tangent line.’ A full discussion of Newton’s method requires

mathematical tools which are not available to us at this time. In general,

many interesting phenomena can occur. Still, the principle problem is as

follows. Suppose that f(x) is a diﬀerentiable function and f(x) = 0 and x

0

is given. Suppose that x

n

for n ≥ 1 are computed according to (1.54). Do

the x

n

tend (converge) to x, and how fast?

For completeness sake, we give an answer. Consider an interval I =

[x−a, x +a], and suppose that |f

(x)| ≥ m and |f

(x)| ≤ M on I. Suppose

x

n

∈ I and |x −x

n

| <

2m

M

. A theorem from advanced calculus asserts that

|x −x

n+1

| ≤

M

2m

(x −x

n

)

2

. (1.56)

We illustrate the theorem by applying it to the previous example. Ob-

serve that f(.8) < 0 and f(.9) > 0. This tells us that x ∈ [.8, .9]. Let us

set a = .2, so that I ⊂ J = [.6, 1.1]. On J, and with this also on I, we

have that |f

(x)| ≥ m = 1.5 and |f

**(x)| ≤ M = 2.5. You are invited to
**

verify these estimates using technology. In (1.56) we use that

M

2m

< 1. As

ﬁrst guess we used x

0

= 1, so that we know that |x − x

0

| < .2. The quoted

theorem asserts that |x − x

1

| < .04. If we repeat the process, then we see

that |x −x

2

| < .0016 and |x −x

3

| < .00000256. This illustrates that the x

n

approach x rapidly.

There is one feature of Newton’s method which helps. You may say

that with each iteration you make a fresh start, and in this sense previous

round-oﬀ errors don’t carry over.

1.18.3 Euler’s Method

Euler’s method is designed to ﬁnd, by numerical means, an approximate

solution of the following kind of problem:

54 CHAPTER 1. BASIC CONCEPTS

Problem 1. Find a function y(t) which satisﬁes

y

=

dy

dt

= F(t, y) and y(t

0

) = y

0

. (1.57)

Here F(t, y) denotes a given function in two variables, and t

0

and y

0

are

given numbers.

The ﬁrst condition on y in (1.57) is a ﬁrst order diﬀerential equation. It is

an equation which involves a function and its derivative, and the unknown is

the function. The second condition is called an initial condition. It speciﬁes

the value of the function at one point. For short, the problem in (1.57) is

called an initial value problem.

Approach in one step: Suppose you want to ﬁnd y(T) for some T = t

0

.

Then you might try the formula

y(T) ≈ y(t

0

) +y

(t

0

)(T −t

0

) = y

0

+F(t

0

, y

0

)(T −t

0

). (1.58)

The tangent line to the graph of y at (t

0

, y

0

) is

l(t) = y(t

0

) +y

(t

0

)(t −t

0

),

so that the middle term in (1.58) is just l(T). The ﬁrst, approximate equality

in (1.58) expresses the philosophy that the graph of a diﬀerentiable function

is close to its tangent line, at least as long as T is close to t

0

. To get the

second equality in (1.58) we use the diﬀerential equation and initial condition

in (1.57), which tell us that

y

(t

0

) = F(t

0

, y(t

0

)) = F(t

0

, y

0

).

The Logistic Law

The diﬀerential equation in our next example is known as the logistic law

of population growth. In the equation, t denotes time and y(t) the size of

a population, which depends on t. The constants a and b are called the

vital coeﬃcients of the population. The equation was ﬁrst used in popula-

tion studies by the Dutch mathematician-biologist Verhulst in 1837. The

equation reﬁnes the Malthusian law for population growth (see (1.45)).

In the diﬀerential equation, the term ay expresses that population growth

is proportional to the size of the population. In addition, the members of

the population meet and compete for food and living space. The probability

of this happening is proportional to y

2

, so that it is assumed that population

growth is reduced by a term which is proportional to y

2

.

1.18. NUMERICAL METHODS 55

Example 1.75. Consider the initial value problem:

dy

dt

= ay −by

2

and y(t

0

) = y

0

, (1.59)

where a and b are given constants. Find an approximate value for y(T).

Remark 3. An exact solution of the initial value problem in (1.59) is given

by the equation

y(t) =

ay

0

by

0

+ (a −by

0

)e

−a(t−t

0

)

(1.60)

This is not the time to derive this exact solution, though you are invited to

verify that it satisﬁes (1.59). We are providing the exact solution, so that

we can see how well our approximate values match it.

Solution: Setting F(t, y) = ay −by

2

, you see that the diﬀerential equa-

tion in this example is a special case of the one in (1.57). According to the

formula in (1.58) we ﬁnd

y(T) ≈ y

0

+ (ay

0

−by

2

0

)(T −t

0

). (1.61)

We expect a close approximation only for T close to t

0

. ♦

Let us be even more speciﬁc and give a numerical example.

Example 1.76. Consider the initial value problem.

dy

dt

=

1

10

y −

1

10000

y

2

and y(0) = 300. (1.62)

Find approximate values for y(1) and y(10).

Solution: Substituting a = 1/10, b = 1/10000, t

0

= 0, and y

0

= 300

into the solution in (1.61), we ﬁnd that

y(1) ≈ 300 +

300

10

−

300

2

10000

(1 −0) = 321.

According to the exact solution in (1.60), we ﬁnd that

y(t) =

3000

3 + 7e

−t/10

.

Substituting t = 1, we ﬁnd the exact value y(1) = 321.4; this number is

rounded oﬀ. So, our approximate value is close.

For T = 10 the formula suggests that y(10) ≈ 510. According to the

exact solution for this initial value problem, y(10) = 538.1. For this larger

value of T, the formula in (1.61) gives us a less satisfactory result. ♦

56 CHAPTER 1. BASIC CONCEPTS

Multi-step approach: We like to ﬁnd a remedy for the problem which

we discovered in Example 1.76 for T further away from t

0

. Consider again

Problem 1 on page 54. We want to get an approximate value for y(T). For

notational convenience we assume that T > t

0

. Pick several t

i

between t

0

and T:

t

0

< t

1

< t

2

< · · · < t

n

= T.

Starting out with t

0

and y(t

0

), we use the one step method from above to get

an approximate value for y(t

1

). Then we pretend that y(t

1

) is exact, and we

repeat the process. We use t

1

and y(t

1

) to calculate an approximate value for

y(t

2

). Again we pretend that y(t

2

) is exact and use t

2

and y(t

2

) to calculate

y(t

3

). Iteratively, we calculate [t

i+1

, y(t

i+1

)] from [t

i

, y(t

i

)] according to the

formula in (1.58):

[t

i+1

, y(t

i+1

)] = [t

i+1

, y(t

i

) +F(t

i

, y(t

i

))(t

i+1

−t

i

)] (1.63)

We continue this process until we reach T.

For reasonably nice

11

expressions F(t, y) the accuracy of the value which

we get for y(T) will increase with n, the number of steps we make (at least if

all steps are of the same length). On the other hand, in an actual numerical

computation we also make round-oﬀ errors in each step, and the more steps

we make the worse the result might get. Experience will guide you in the

choice of the step length.

Example 1.77. Consider the initial value problem

dy

dt

=

1

10

y −

1

10000

y

2

and y(0) = 10. (1.64)

1. Apply the multi-step method to ﬁnd approximate values for y(t) at

t

1

= 5, t

2

= 10, t

3

= 15, . . . , t

20

= 100. Arrange them in a table.

2. Graph the points found in the previous step together with the actual

solution of the initial value problem.

Solution: As points in the multi-step process we use

t

0

= 0, t

1

= 5, t

2

= 10, t

3

= 15, t

4

= 20, . . . , t

20

= 100.

11

We do not want to make this term precise, but the F(t, y) in Example 1.75 is of this

kind.

1.18. NUMERICAL METHODS 57

t y(t) & t y(t) & t y(t)

0 10.00 35 153.96 70 857.73

5 14.95 40 219.09 75 918.74

10 22.31 45 304.62 80 956.07

15 33.22 50 410.55 85 977.07

20 49.28 55 531.55 90 988.27

25 72.70 60 656.05 95 994.07

30 106.41 65 768.87 100 997.02

Table 1.2: Solution of Problem 1.77

For each t

i

(0 ≤ i ≤ 19) we use the formula

y(t

i+1

) = y(t

i

) + 5

y(t

i

)

10

−

y

2

i

(t

i

)

10000

and calculate y(t

1

), y(t

2

), y(t

3

), . . . , y(t

20

) consecutively. We summarize

the calculation in Table 1.2.

In Figure 1.15 you see the graph of the exact solution of the initial value

problem. You also see the points from Table 1.2. The points suggest a graph

which does follow the actual one reasonably closely. But you see that we are

deﬁnitely making errors, and they get worse as t increases

12

. You may try

a shorter step length. The points will follow the curve much more closely if

you use t

1

= 1, t

2

= 2, t

3

= 3, . . . , t

100

= 100 in your calculation. ♦

Steady States: Let us consider some very speciﬁc solutions of our initial

value problem in (1.57):

y

=

dy

dt

= F(t, y) and y(t

0

) = y

0

.

Suppose F(y

0

, t) = 0 for all t. Then the constant function y(t) = y

0

is a

solution of the problem. Such a solution is called a steady state solution.

12

It is incidental that the points eventually get closer to the graph again. This is due

to the speciﬁc problem, and will not occur in general.

58 CHAPTER 1. BASIC CONCEPTS

20 40 60 80 100

200

400

600

800

1000

Figure 1.15: Illustration of Euler’s Method

Example 1.78. Find the steady states of the diﬀerential equation (see

(1.50) in Section 1.16)

f

(t) = af(t) +b. (1.65)

Solution: Apparently, f

**(t) = 0 if and only if f(t) = −b/a. So the
**

constant function f(t) = −b/a is the only steady state of this diﬀerential

equation. ♦

In review of Example 1.68 in Section 1.16, you see that the steady state

in that example is B(t) = 40, 000. I.e., if your loan balance is $40,000.00, the

bank charges you interest at a rate of .5% per month, and you are repaying

the loan at a rate of $ 200.00 per month, then the principal balance of your

account will stay unchanged. Your payments cover exactly the occuring

interest charges.

Example 1.79. For the logistic law (see Equation (1.59))

dy

dt

= F(y, t) = ay −by

2

= y(a −by)

1.18. NUMERICAL METHODS 59

we ﬁnd that F(y, t) = 0 if and only if y = 0 or y = a/b. There are two

steady state solutions: y

u

(t) = 0 and y

s

(t) = a/b.

Let us interpret these steady state solutions for the speciﬁc numerical

values of a = 1/10 and b = 1/10, 000 in Example 1.77. If the initial value

y

0

of the population is positive, then the population size will tend to and

stabilize

13

at y(t) = a/b = 1, 000. In this sense, y

s

(t) = a/b = 1, 000 is a

stable steady state solution. It is also referred to as the carrying capacity.

It tells you which size population of the given kind the speciﬁc habitat will

support.

If the initial value y

0

is negative, then y(t) will tend to −∞ as time

increases. If y

0

= 0, then y(t) will not tend to the steady state y(t) = 0. In

this sense, y(t) = 0 is an unstable steady state. ♦

Exercise 13. Consider the initial value problem

y

(t) = −50 +

1

2

y(t) −

1

2000

y

2

(t) and y

0

= y(0) = 200. (1.66)

To make the problem explicit, you should think of a population of deer in a

protected wildlife preserve. There are no predators. The deer are hunted at

a rate of 50 animals per year. The population has a growth rate of 50% per

year. Reproduction takes place at a constant rate all year round. Finally,

the last term in the diﬀerential equation accounts for the competition for

space and food.

1. Use Euler’s method to ﬁnd the population size over the next 30 years.

Proceed in 1 year steps. Tabulate and plot your results.

2. Guess at which level the population stabilizes.

3. Repeat the ﬁrst two steps of the problem if hunting is stopped.

4. Repeat the ﬁrst two steps of the problem if the initial population is

100 animals.

5. Find the steady states of the original equation in which hunting takes

place. I.e., ﬁnd for which values of y you have that y

= 0? You will

ﬁnd two values. Call the smaller one of them Y

u

and the larger one

Y

s

. Experiment with diﬀerent initial values to see which of the steady

states is stable, and which one is unstable.

13

The common language meaning of these expressions suﬃces for the purpose of our

discussion, and the mathematical deﬁnition of ‘tends to’ and ‘stabilizes at’ only make these

terms precise.

60 CHAPTER 1. BASIC CONCEPTS

Orthogonal Trajectories

Let us explore a diﬀerent kind of application. Suppose we are given a family

F(x, y, a) = 0 of curves. In Figure 1.16 you see a family of ellipses

C

a

: F(x, y, a) = x

2

+ 3y

2

−a = 0. (1.67)

There is one ellipse for each a > 0. We like to ﬁnd curves D

b

which in-

tersect the curves C

a

perpendicularly. (We say that D

b

and C

a

intersect

perpendicularly in a point (x

1

, y

1

), if the tangent lines to the curves at this

point intersect perpendicularly.) We call such a curve D

b

an orthogonal tra-

jectory to the family of the C

a

’s. You also see one orthogonal trajectory in

Figure 1.16.

-6 -4 -2 0 2 4 6

-6

-4

-2

0

2

4

6

Figure 1.16: Orthogonal Trajectory to Level Curves

Let us explain where this type of situation occurs. Suppose the curves

C

a

are the level curves in a crater. Here a represents the elevation, so that

the elevation is constant along each curve C

a

. The orthogonal trajectory

gives a path of steepest descent. A new lava ﬂow which originates at some

point in the crater will follow this path.

Suppose that each ellipse represents an equipotential line of an electro-

magnetic ﬁeld. The orthogonal trajectory provides you with a path which

is always in the direction of the most rapid change of the ﬁeld. A charged

particle will move along an orthogonal trajectory.

1.18. NUMERICAL METHODS 61

Suppose a stands for temperature, so that along each ellipse the tem-

perature is constant. In this case the curves are called isothermal lines

14

.

A heat seeking bug will, at any time, move in the direction in which the

temperature increases most rapidly, i.e., along an orthogonal trajectory to

the isothermal lines.

Suppose a stands for the concentration of a nutrient in a solution. It is

constant along each curve C

a

. On their search for food, bacteria will follow

a path in the direction in which the concentration increases most rapidly.

They will move along an orthogonal trajectory.

Example 1.80. Find orthogonal trajectories for the family of ellipses

C

a

: F(x, y, a) = x

2

+ 3y

2

−a = 0. (1.68)

Solution: Diﬀerentiating the equation for the ellipses, we get

2x + 6y

dy

dx

= 0 or

dy

dx

=

−x

3y

.

The slope of the tangent line to a curve C

a

at a point (x

1

, y

1

) is

−x

1

3y

1

. If

a curve D

b

intersects C

a

in (x

1

, y

1

) perpendicularly, then we need that the

slope of the tangent line to D

b

at this point is

3y

1

x

1

. Thus, to ﬁnd an orthog-

onal trajectory to the family of the C

a

’s we need to ﬁnd functions which

satisfy this diﬀerential equation. If we also require that the orthogonal tra-

jectory goes through a speciﬁc point (x

0

, y

0

), then we end up with the initial

value problem

dy

dx

=

3y

x

and y(x

0

) = y

0

.

This is exactly the kind of problem which we solved with Euler’s method. In

this particular example it is not diﬃcult to ﬁnd solutions for the diﬀerential

equation. They are functions of the form y(x) = bx

3

. The orthogonal

trajectory shown in Figure 1.16 has the equation y = x

3

/25. There is one

orthogonal trajectory which does not have this form, and this is the curve

x = 0.

Let us apply Euler’s method to solve the problem. Let us ﬁnd approxi-

mate values for the initial value problem

dy

dx

=

3y

x

and y(1) =

1

25

.

14

The idea of isothermal lines, and with this the method in all of these applications,

was pioneered by Alexander von Humbold (1769–1859).

62 CHAPTER 1. BASIC CONCEPTS

Use x

0

= 1, x

1

= 1.2, x

2

= 1.4, . . . , x

20

= 5.

We set (x

0

, y

0

) = (1, 1/25) and calculate (x

n

, y

n

) according to the for-

mula

y

n

= y

n−1

+.2

3y

n−1

x

n−1

for n = 1, 2, . . . , 20.

Without recording the results of this calculation, we graphed the points in

Figure 1.16. ♦

Exercise 14. Consider the family of hyperbolas:

C

a

: x

2

−5y

2

+a = 0.

There is one hyperbola for each value of a, only for a = 0 the hyperbola

degenerates into two intersecting lines.

1. Graph several of the curves C

a

.

2. Find the diﬀerential equation for an orthogonal trajectory.

3. Use Euler’s method to ﬁnd points on the orthogonal trajectory through

the point (3, 4). Use the points x

0

= 3, x

1

= 3.2, x

2

= 3.4, . . . ,

x

20

= 7. Plot the points (x

n

, y

n

) in your ﬁgure.

4. Check that the graph of y(x) = bx

−5

is an orthogonal trajectory to the

family of hyperbolas for every b. Determine b, so that the orthogonal

trajectory passes through the point (3, 4), and add this graph to your

ﬁgure.

1.19. TABLE OF IMPORTANT DERIVATIVES 63

1.19 Table of Important Derivatives

f(x) f

(x) Assumptions

x

q

qx

q−1

q a natural number, or x > 0

e

x

e

x

x ∈ (−∞, ∞)

ln |x| 1/x x ∈ (−∞, ∞), x = 0

sin x cos x x ∈ (−∞, ∞)

cos x −sin x x ∈ (−∞, ∞)

tan x sec

2

x all x for which tan x is deﬁned

cot x −csc

2

x all x for which cot x is deﬁned

sec x sec xtan x all x for which sec x is deﬁned

csc x −csc xcot x all x for which csc x is deﬁned

arctan x

1

1+x

2

x ∈ (−∞, ∞)

arcsin x

1

√

1−x

2

x ∈ (−1, 1), arcsin x ∈ (−π/2, π/2)

arccos x

−1

√

1−x

2

x ∈ (−1, 1), arccos x ∈ (0, π)

arccot x

−1

1+x

2

x ∈ (−∞, ∞), arccot x ∈ (0, π)

arcsec x

1

|x|

√

x

2

−1

x < −1 or x > 1, arcsec x ∈ (0, π/2) ∪ (π/2, π)

arccsc x

−1

|x|

√

x

2

−1

x < −1 or x > 1, arcsec x ∈ (−π/2, 0) ∪ (0, π/2)

Table 1.3: Some Derivatives

64 CHAPTER 1. BASIC CONCEPTS

Chapter 2

Global Theory

So far we studied the local behaviour of a function. All concepts related to

the behaviour of a function near a point. In this chapter we will use local in-

formation about a function to draw global conclusions. We will discuss some

uniqueness properties of solutions of diﬀerential equations. Then we discuss

geometric properties of graphs, their monotonicity and concavity. We apply

these ideas to the study of extrema of functions. With this information it is

possible to sketch graphs capturing their essential features.

The fundamental result which allows us to do this is referred to as

Cauchy’s mean value theorem. Augustin-Louis Cauchy (1789–1857) was

one of the great mathematicians of the 19-th century. He made major con-

tributions to make calculus a rigorous mathematical theory.

2.1 Cauchy’s Mean Value Theorem

It is useful to make the following

Deﬁnition 2.1. Let f(x) be a function which is deﬁned on the interval

[a, b]. Then we call

f(b) −f(a)

b −a

the average rate of change of f over the interval [a, b].

For example, the average rate of change of f(x) = x

2

over the interval

[0, 2] is 2. The average rate of change of f(x) = sin x over [0, π/2] is 2/π

and over [0, π] it is 0.

65

66 CHAPTER 2. GLOBAL THEORY

Theorem 2.2 (Chauchy’s Mean Value Theorem). Let f be a real val-

ued function which is deﬁned and continuous on the interval [a, b] and dif-

ferentiable on (a, b), where a < b. Then there exists a number c ∈ (a, b) such

that

f

(c) =

f(b) −f(a)

b −a

.

In words, the theorem asserts that the average rate of change over an

interval is equal to the rate of change at some point in the interval. For

example, the average rate of change of f(x) = x

2

over the interval [−2, 1] is

−1, and f

(−1/2) = −1.

The following special case of the theorem, called Rolle’s theorem (named

after Michel Rolle (1652–1719)), is of particular interest.

Theorem 2.3 (Rolle’s Theorem). Let f be a real valued function which

is deﬁned and continuous on the interval [a, b] and diﬀerentiable on (a, b),

where a < b. If f(a) = f(b), then there exists a number c between a and b

(i.e., a < c < b) such that

f

(c) = 0.

We are not going to say anything about the proof of these two theorems,

except that Cauchy’s theorem and Rolle’s theorem are equivalent (each is

an easy consequence of the other one), and that the proof of both of them

depends heavily on the completeness

1

of the real numbers. We are also not

interested in ﬁnding the points c, as they occur in the two theorems. We

are interested in more general consequences.

Corollary 2.4. Let f be a real valued function which is deﬁned and contin-

uous on an interval I. If f

**(x) = 0 for all interior points x of I, then f is
**

constant on this interval. In other words, there exists a number d such that

f(x) = d for all x ∈ I.

Proof. A diﬀerent formulation of the claim is that f(a) = f(b) for all a,

b ∈ I. We prove this statement using Cauchy’s theorem. If f(a) = f(b),

then a = b and there exists some c ∈ (a, b), such that

f

(c) =

f(b) −f(a)

b −a

= 0.

But this contradicts the assumption that f

**(c) = 0 for all c ∈ I, and the
**

corollary is proved.

1

We discussed this property of the real numbers in Section 1.1.

2.2. UNIQUE SOLUTIONS OF DIFFERENTIAL EQUATIONS 67

We are going to use the following corollary frequently.

Corollary 2.5. Let h and g be functions which are deﬁned and continuous

on an interval I. If h

(x) = g

**(x) for all x ∈ I, then h and g diﬀer by a
**

constant, i.e., there exists a number d such that

h(x) = g(x) +d

for all x ∈ I.

Proof. Apply the previous corollary to f(x) = h(x) −g(x).

Deﬁnition 2.6. Suppose the function f(x) is deﬁned on the interval I. We

call a function F(x) with domain I an antiderivative of f if F

(x) = f(x)

for all x ∈ I.

Using this notion, we can reformulate Corollary 2.5.

Corollary 2.7. Suppose h and g are antiderivatives of a function f, deﬁned

on an interval. Then h and g diﬀer by a constant.

2.2 Unique Solutions of Diﬀerential Equations

Corollary 2.4 implies

Proposition 2.8. If the function F(x) is deﬁned on an interval I and

F

**(x) = 0 for all x ∈ I, then F(x) is constant on I.
**

In other words, on intervals the only solutions of the diﬀerential equation

F

**(x) = 0 are the constant functions.
**

More generally, if you like to ﬁnd all antiderivatives F(x) of a function

f(x) on an interval, then it suﬃces to ﬁnd one antiderivative H(x). Any

antiderivative F(x) is of the form H(x) + c where c is a constant. The

constant c is referred to as integration constant. For the time being you

depend on being able to guess such a function H(x). By diﬀerentiating

H(x) you can check whether you guessed right.

For example, any antiderivative F(x) of the function f(x) = 2x on the

real line (−∞, ∞) is of the form F(x) = x

2

+ c where c is a constant. Any

antiderivative F(x) of the function f(x) = sec

2

x on the interval (−π/2, π/2)

is of the form F(x) = tan x +c.

Typically, the integration constant is determined by an initial condition.

Suppose we like to solve the initial value problem

f

(x) = cos x and f(0) = 1.

68 CHAPTER 2. GLOBAL THEORY

Our ﬁrst conclusion is that f(x) = sin x + c. This follows from the above

because (sin x + c)

**= cos x. Next we substitute x = 0 in the equation.
**

Then we see that f(0) = c = 1. The solution of the initial value problem is

f(x) = sin x + 1.

Of particular importance to our discussion of expenential growth and

decay is

Proposition 2.9. Every solution f(x) of the diﬀerential equation

f

(x) = af(x)

on an interval is of the form f(x) = ce

ax

for some constant c.

Proof. We asserted in (1.19), and will eventually prove, that all functions of

the form f(x) = ce

ax

satisfy the diﬀerential equation. We want to see that

these are the solutions.

Let f(x) be any function which satisﬁes the diﬀerential equation on some

interval. Consider the function

h(x) = f(x)e

−ax

.

As a product of diﬀerentiable functions, h is diﬀerentiable. Its derivative is

h

(x) = f

(x)e

−ax

−af(x)e

−ax

= af(x)e

−ax

−af(x)e

−ax

= 0.

Corollary 2.4 tells us that h(x) is a constant function. Calling the constant

c we ﬁnd that

f(x) = ce

ax

.

This means that all solutions of the diﬀerential equation f

(x) = af(x) are

of the form f(x) = ce

ax

, where c is a constant.

Proposition 2.10. The initial value problem

f

(x) = af(x) and f(x

0

) = C

has a unique solution on an interval containing x

0

. In fact

f(x) = Ce

a(x−x

0

)

.

Proof. By the previous proposition we know that the solution is of the form

f(x) = ce

ax

for some c. Substituting the initial condition we obtain

C = f(x

0

) = ce

ax

0

.

Thus c = Ce

−ax

0

and f(x) = ce

ax

= Ce

−ax

0

e

ax

= Ce

a(x−x

0

)

.

2.3. THE FIRST DERIVATIVE AND MONOTONICITY 69

Remark 4. The uniqueness of the solution of an initial value problem as in

the previous proposition is not only of theoretical importance. Imagine that

you study the growth rate of a strain of bacteria, as we did in Example 1.64

on page 41. Before you can publish your result, it must be certain that your

experiment can be reproduced at a diﬀerent time in a diﬀerent location.

That is a requirement which any experiment in science must satisfy. If there

is more than one mathematical solution to your problem, then you have to

expect that the experiment can go either way, and this would invalidate your

experiment.

2.3 The First Derivative and Monotonicity

One of the interesting properties of a function is whether it is increasing or

decreasing. We might want to ﬁnd out whether the part of a population

which is infected with a disease is increasing or decreasing. We might want

to know how the level of pollution in a body of water is changing. The ﬁrst

derivative of a function gives us information of this kind.

2.3.1 Monotonicity on Intervals

Recall that a function f is called increasing if f(b) > f(a) whenever b > a.

It is called decreasing if f(b) < f(a) whenever b > a. A function is called

monotonic if it is either increasing or decreasing.

Theorem 2.11. Suppose that the function f is deﬁned and continuous on

the interval I.

1. If f

**(x) > 0 for all x ∈ I, then f is increasing on I.
**

2. If f

**(x) < 0 for all x ∈ I, then f is decreasing on I.
**

3. More generally, the conclusions in (1) and (2) still hold if in each

ﬁnite interval J ⊂ I there are only ﬁnitely many points at which the

assumption on f

(x) is not satisﬁed.

2

Proof. We show (1). Let a and b be points in I, and suppose that a < b.

Cauchy’s theorem says that there exists a point c, a < c < b, such that

f

(c) =

f(b) −f(a)

b −a

.

2

It is permissable that f is not diﬀerentiable at a few points in J, or that f

(x) = 0.

It is not possible that f

**(x) < 0 at some point in the interval, and f(x) is increasing on
**

the interval.

70 CHAPTER 2. GLOBAL THEORY

We have that f

**(c) > 0 and b − a > 0, and it follows that f(b) − f(a) > 0.
**

This means that f(b) > f(a). The proof of the second claim is similar. We

leave it and the generalization of both statements to the reader.

For example, log

2

x =

ln x

ln 2

=

1

x ln2

> 0 for all x ∈ (0, ∞). In the

computation we used (1.19), Theorem 1.36, and that ln2 > 0. It follows

from Theorem 2.11 that log

2

x is increasing on x ∈ (0, ∞). You see part of

the graph of the function in Figure 1.8.

The exponential function exp

a

x = a

x

is increasing on (−∞, ∞) if a > 1

and decreasing if 0 < a < 1. To see this, observe that a

x

= e

x ln a

and

d

dx

a

x

= (ln a)a

x

. Furthermore, a

x

> 0 and lna > 0 if a > 1 and ln a < 0

if 0 < a < 1. Now Theorem 2.11 implies our assertion. You may also

want to have a look at the graph of the exponential function with base 2 in

Figure 1.7.

The function f(x) = 1/x is deﬁned and diﬀerentiable on the set of all

nonzero real numbers, and its derivative is f

(x) = −1/x

2

. In particular

f

**(x) < 0 for all nonzero real numbers. According to Theorem 2.11, f(x)
**

is decreasing on the interval (−∞, 0), and that f(x) is decreasing on the

interval (0, ∞). The function is not decreasing on the union of the two

intervals. The example illustrates that it is crucial in Theorem 2.11 that we

deal with functions which are deﬁned and diﬀerentiable on an interval.

The function f(x) = tan x, deﬁned on (−π/2, π/2), has as its deriva-

tive f

(x) = sec

2

x, and the derivative is positive. Consequently, f(x) is

increasing on (−π/2, π/2). Its inverse g(x) = arctan x, deﬁned on (−∞, ∞),

has as its derivative g

(x) =

1

1+x

2

, which is positive on (−∞, ∞), so that

g(x) = arctan x is increasing on (−∞, ∞). As a general priciple, one may

show that the inverse of an increasing function is increasing.

Example 2.12. For a three dimensional solid we set E = A/V , where

A denotes the surface area and V the volume. For example, for a ball

E(r) = (4πr

2

)/(

4

3

πr

3

) = 3/r, where r denotes the radius. Then E

(r) < 0.

The same principle holds for other shapes, E decreases as we enlarge the

solid without changing its shape. What does this have to do with the size

of animals?

Warm blooded animals living in cold climates need to preserve their body

temperature. The total amount of heat stored in the body is proportional to

the volume, while the heat loss is proportional to the surface area. The ratio

of volume to surface area increases as the animal gets larger, so that for warm

blooded animals it is of advantage to be large if they live in cold climates.

In hot climates they need to give oﬀ heat, so that it is of advantage to be

2.3. THE FIRST DERIVATIVE AND MONOTONICITY 71

small. Natural selection (Darwinism) should favor the larger specimens of

a warm blooded species in a cold climate and smaller ones in a hot climate.

You can observe this phenomenon in real life.

For cold blooded animals the converse holds. They absorbe heat so

that they body reaches a temperature at which they can be active. In cold

climates it helps to be small, because then the surface area is relatively large,

compared to the volume. In hot climates cold blooded animals can aﬀord to

be large, as it is easy to reach and maintain the temperature at which they

can be active. The argument is again consistent with real life.

Needless to say, there are other mechanisms to increase the surface area

of a body than decreasing its size, and the maintenance of the body temper-

ature is only one factor which inﬂuences the size of specimens of a species.

Larger animals need more food, are stronger, cannot hide so well, and are of-

ten less agile. All of these factors need to be taken into account to determine

the optimal size of an animal. ♦

So far we have only discussed examples where we used (1) and (2) of

Theorem 2.11. Let us show how to use the conclusion in (3). To apply it

we need to determine intervals on which a function does not change signs.

We recall a procedure which works well for continuous functions.

Deﬁnition 2.13. Suppose f(x) is a function. We call a point x

0

on the

real line exceptional if either f(x

0

) = 0 or f(x

0

) is not deﬁned.

The following result is an immediate consequence of the Intermediate

Value Theorem, see Theorem 1.16 on page 8. Expressed casually it says

that a continuous function can change signs only at exceptional points.

Proposition 2.14. Suppose f(x) is continuous and f(x) has no exceptional

points in the interval (x

0

, x

1

). Then f(x) > 0 for all points in the interval

(x

0

, x

1

), or f(x) < 0 for all points in the interval (x

0

, x

1

). In particular, if

f(x) is positive at one point in the interval, then it is positive at all points

in the interval. If f(x) is negative at one point in the interval, then it is

negative at all points in the interval.

Example 2.15. For example, consider the function

f(x) =

x

2

(x

2

−4)

x

2

+ 2x −15

=

x

2

(x −2)(x + 2)

(x −3)(x + 5)

.

The zeros of the numerator, and with this the zeros of f(x), are x = 0,

x = 2, and x = −2. The zeros of the denominator, i.e., the points where

f(x) is not deﬁned, are x = 3 and x = −5.

72 CHAPTER 2. GLOBAL THEORY

According to the proposition, the sign of f(x) remains unchanged on

each of the intervals (−∞, −5), (−5, −2), (−2, 0), (0, 2), (2, 3) and (3, ∞).

Counting signs of the factors in the expression for f(x), we see f(x) is

positive on the interval (−∞, −5), negative on (−5, −2), positive on (−2, 0)

and on (0, 2), negative on (2, 3), and positive on (3, ∞). You see that the

sign changes at some, but not all, exceptional numbers. ♦

Exercise 15. Find intervals on which the following functions do not change

signs. Decide whether the functions are positive or negative on these inter-

vals.

(1) f(x) = x

3

−x

2

−5x −3 (2) g(x) =

x

x

3

+ 5x

2

−4x −20

.

We are ready to discuss the monotonicity of functions whose derivative

vanishes at some points.

Example 2.16. Find intervals of monotonicity for the function

f(x) = 3x

2

+ 5x −4.

-3 -2 -1 1 2 3 4

-5

5

10

15

20

25

30

Figure 2.1: A quadratic polyno-

mial, f(x) = 3x

2

+ 5x −4

-2 2 4

-20

-15

-10

-5

5

Figure 2.2: A cubic polynomial,

p(x) = x

3

−3x

2

−9x + 3

Solution: We graphed the function in Figure 2.1. Its derivative is

f

(x) = 6x + 5. In particular, f

(x) > 0 if x ∈ (−5/6, ∞). So f

(x) > 0

for all points x ∈ [−5/6, ∞), except at x = −5/6. Theorem 2.11 (3) says

that f is increasing on the interval [−5/6, ∞). By a similar argument, f is

decreasing on the interval (−∞, −5/6]. ♦

2.3. THE FIRST DERIVATIVE AND MONOTONICITY 73

Example 2.17. Find intervals of monotonicity for the degree three poly-

nomial (for a graph see Figure 2.2)

p(x) = x

3

−3x

2

−9x + 3

Solution: The function is deﬁned and diﬀerentiable on the real line. Its

derivative is

p

(x) = 3x

2

−6x −9 = 3(x

2

−2x −3) = 3(x −3)(x + 1).

Counting the signs of the factors we see that p

(x) is positive on (−∞, −1)

and on (3, ∞). We conclude that p(x) is increasing on the interval [3, ∞)

and that it is increasing on the interval (−∞, −1]. The derivative is negative

on the interval (−1, 3). The theorem implies that p(x) is decreasing on the

interval [−1, 3]. ♦

Example 2.18. Find intervals of monotonicity for the rational function

f(x) =

x

2

+ 3x

x −1

.

Solution: The simpliﬁed expression for the derivative of f is

f

(x) =

(x + 1)(x −3)

(x −1)

2

.

We see that the exceptional points for f

(x) are x = 1, x = −1 and x = 3. We

conclude that f

**(x) does not change signs on the intervals (−∞, −1), (−1, 1),
**

(1, 3), and (3, ∞). Counting the signs of the factors of f

(x), we conclude

that f

(x) > 0 on the intervals (−∞, −1) and (3, ∞), and f

(x) < 0 on the

intervals (−1, 1) and (1, 3). Observe that f(x) is deﬁned and diﬀerentiable

on the entire real line with the only exception of x = 1. We conclude that

f(x) is increasing on the (−∞, −1] and [3, ∞). The function is decreasing

on the intervals [−1, 1) and (1, 3]. ♦

Example 2.19. Find intervals on which the function

f(x) = sin 2x + 2 sin x

is monotonic. Restrict your discussion to the interval [0, 2π].

Solution: We diﬀerentiate the function and rewrite the expression for

the derivative so that it is easier to ﬁnd its exceptional points.

f

(x) = 2 cos 2x + 2 cos x

= 2[2 cos

2

x + cos x −1]

= 4(cos x + 1)

cos x −

1

2

.

74 CHAPTER 2. GLOBAL THEORY

To see the second equality we used that cos 2x = 2 cos

2

x − 1. Then we

solved the quadratic equation in terms of cos x. We ﬁnd exceptional points

where cos x = −1 (i.e., x = π) and where cos x =

1

2

(i.e., x =

π

3

and x =

5π

3

).

1 2 3 4 5 6

-2

-1

1

2

3

4

Figure 2.3: A function and its derivative.

Observe that f is diﬀerentiable on [0, 2π], and that f

(x) = 0 at the end

points of this interval. This provides us with the intervals [0, π/3), (π/3, π),

(π, 5π/3) and (5π/3, 2π] on which f

**does not change sign. Checking the
**

sign of f

(at one point) in each of the intervals, we ﬁnd that f

(x) > 0 for

x ∈ [0, π/3) and x ∈ (5π/3, 2π], and f

(x) < 0 for x ∈ (π/3, π) and (π, 5π/3).

We conclude that f is increasing on the interval [0, π/3] and [5π/3, 2π]. The

function is decreasing on the interval [π/3, 5π/3], and in this interval there

are three points at which f

(x) is not positive.

You may conﬁrm the calculation by having a look at Figure 2.3. There

you see the graph of the function (solid line) and the graph of its derivative

(dashed line). As you see, wherever f

**(x) is positive, there f(x) is increasing.
**

Wherever f

**(x) is negative, there f(x) is decreasing. ♦
**

Exercise 16. Find intervals on which the function f increases and intervals

on which f decreases. In the last two problems, (g) and (h), restrict yourself

2.3. THE FIRST DERIVATIVE AND MONOTONICITY 75

to the interval [0, 2π].

(a) f(x) = 3x

2

+ 5x + 7

(b) f(x) = x

3

−3x

2

+ 6

(c) f(x) = (x + 3)/(x −7)

(d) f(x) = x + 1/x

(e) f(x) = x

3

(1 +x)

(f) f(x) = x/(1 +x

2

)

(g) f(x) = cos 2x + 2 cos x

(h) f(x) = sin

2

x −

√

3 sin x

2.3.2 Monotonicity at a Point

It is quite natural to ask what it means that a function is increasing at a

point, and how this concept is related to the one of being increasing on an

interval. We address both questions in this subsection.

Deﬁnition 2.20. Suppose f is a function and c is an interior point of its

domain. We say that f is increasing at c if, for some d > 0,

f(x) < f(c) for all x ∈ (c −d, c) and f(x) > f(c) for all x ∈ (c, c +d).

We say that f is decreasing at c if this statement holds with the inequalities

reversed.

Expressed informally, to the left of c the function is smaller and to the

right of c it is larger than at c, at least for a while.

Being increasing or decreasing at a point c is a local property. We are

making a statement about the behavior of the function on some open interval

which contains c. Being increasing on an interval is a global property. For

the global property the interval is given to us. For the local property we may

chose the, possibly rather small, interval. The global property has to hold

for any two points in the given interval. For the local property we compare

f(x) to f(c) where c is ﬁxed and x is any point in an open interval around

c which we may chose.

Theorem 2.21. Suppose f is a function which is deﬁned on an open inter-

val I. Then f is increasing (decreasing) on I if and only it it is increasing

(decreasing) at each point in I.

This theorem establishes the relation between the local and the global

property. The ‘only if’ part is not diﬃcult to show, but the ‘if’ part uses

some deeper facts about ﬁnite closed intervals. Our second result gives us a

valuable tool to detect monotonicity of functions at a point.

76 CHAPTER 2. GLOBAL THEORY

Proposition 2.22. Let f be a function and c an interior point of its do-

main. If f is diﬀerentiable at c and f

(c) > 0, then f is increasing at c. If

f

(c) < 0, then f is decreasing at c.

Remark 5. A function does not have to be diﬀerentiable to be increasing.

Graph the function f(x) = 2x+|x| to convince yourself of this fact. A func-

tion can be diﬀerentiable and increasing at a point x, even if the assumptions

of Proposition 2.22 do not hold, i.e., f(x) = x

3

is increasing at x = 0, but if

f

**(0) = 0. A function can also be increasing at a point x, but there is not
**

open interval which contains x such that the function is increasing on this

interval.

Remark 6. The ideas of of a function being increasing or decreasing at

a point may be generalized to cover domains of functions which are half-

closed or closed intervals, and where we like to make a statement about the

behavior of a function at an endpoint. We have no speciﬁc needs for such

statements, but the motivated reader is encouraged to explore them.

2.4 The Second Derivative and Concavity

We like to capture the property of a graph being bent upwards or downwards.

Secant lines will either be required to lie above or below the graph, and the

rates of change will be either increasing of decreasing. These properties can

be described globally over intervals and locally at points. You may use the

graphs in Figures 2.4 and 2.5 as illustrations of the discussion.

-2 -1 1 2

2

4

6

8

10

Figure 2.4: Concave Up

-2 -1 1 2 3 4

-8

-6

-4

-2

2

4

Figure 2.5: Concave Down

2.4. THE SECOND DERIVATIVE AND CONCAVITY 77

2.4.1 Concavity on Intervals

Let f(x) be a function and let (a, f(a)) and (b, f(b)) be two distinct points

on its graph. The line through these two points is

l(x) = f(a) +

f(b) −f(a)

b −a

(x −a).

If we restrict l(x) to x ∈ [a, b], then we get the secant line through the two

point, i.e., the line segment joining the two points.

Deﬁnition 2.23. Let f be a function which is deﬁned on an interval I.

We say that f is concave up on I if f(c) < l(c) for all a, b in I and

c ∈ (a, b). Here l(x) is the secant line through (a, f(a)) and (b, f(b)). The

inequality expresses that between the points a and b the secant line lies above

the graph. We say that f is concave down on I if f(c) > l(c) for all a, b

in I and c ∈ (a, b). The inequality expresses that between the points a and b

the secant line lies below the graph.

We state a theorem which provides you with assumptions under which a

function is concave up or down. We will not provide a proof of the theorem.

Theorem 2.24. Let f be a function which is deﬁned on an interval I.

1. Suppose that f(x) is diﬀerentiable on I. If f

(x) is increasing on I,

then f(x) is concave up on I. If f

(x) is decreasing on I, then f(x) is

concave down on I.

2. Suppose that f(x) is twice diﬀerentiable

3

on I. If f

(x) > 0 for all x

in I, then f(x) is concave up on I. If f

**(x) < 0 for all x in I, then
**

f(x) is concave down on I.

3. More generally, the conclusions in (2) still hold if in each ﬁnite interval

J ⊂ I there are only ﬁnitely many points at which the assumption

f

(x) > 0, resp. f

(x) < 0, is not satisﬁed.

For example, the function shown in Figure 2.4 is q(x) = x

2

−2x +3. Its

second derivative is q

**(x) = 2 > 0. Theorem 2.24 (2) says that q is concave
**

3

Strictly speaking, so far we can consider being ‘twice diﬀerentiable’ only for functions

which are deﬁned on open intervals. More generally, we proceed as in Section 1.11. We say

that f(x) is twice diﬀerentiable on I, if f(x) extends to a function F(x) which is deﬁned

on an open interval J which contains I, and F(x) is twice diﬀerentiable on J. The second

derivative will be unique at all points in I if I is not empty and does not consist of exactly

one point.

78 CHAPTER 2. GLOBAL THEORY

up on (−∞, ∞). The function shown in Figure 2.5 is g(x) = −x

2

+ 5x −1,

and its second derivative is g

**(x) = −2 < 0. Theorem 2.24 (2) says that q
**

is concave down on (−∞, ∞).

The function ln x is concave down on the interval (0, ∞). To see this, you

may use that ln

(x) = −1/x

2

< 0 on (0, ∞) and apply Theorem 2.24 (2).

Alternatively, you may note that the derivative ln

x = 1/x is decreasing on

(0, ∞) and apply Theorem 2.24 (1). The exponential function exp(x) = e

x

is

concave up on (−∞, ∞). To see this, you may note that exp

(x) = exp(x) >

0 and apply Theorem 2.24 (2). You may also use that exp

(x) is increasing

on the real line, and then quote Theorem 2.24 (1) to derive the desrired

conclusion. Finally, you may observe that a function is concave up if its

inverse is convave down

4

. So, ln x being concave down implies that exp(x)

is concave up.

Let us look at examples where we apply condition Theorem 2.24 (3).

Example 2.25. Study the concavity properties of the function

p(x) = x

3

−3x

2

−9x + 3.

Solution: You ﬁnd the graph of this function in Figure 2.2. Its second

derivative is p

(x) = 6x−6 = 6(x−1). We see that p

(x) > 0 for x ∈ (1, ∞),

and p

(x) < 0 for x ∈ (−∞, 1). This means that p

(x) > 0 for all x ∈ [1, ∞)

with only one exception, x = 1. Theorem 2.24 (3) tells us that p(x) is

concave up on the interval [1, ∞). Similarly, p

(x) < 0 for x ∈ [−∞, 1) with

only one exception, x = 1. One deduces that f(x) is concave down on the

interval (−∞, 1]. ♦

Consider the function tan x. You may verify that tan

x = 2 sec

2

xtan x.

In particular, tan

x < 0 for x ∈ (−π/2, 0) and tan

x > 0 for x ∈ (0, π/2).

Theorem 2.24 (3) implies that tan x is concave down on (−π/2, 0] and con-

cave up on [0, π/2). You may conﬁrm these statements visually by inspecting

a graph of the tangent function. You are invited to study the concavity of

the other trigonometric and hyperbolic functions.

Remark 7. You may consider the spread of a desease. Denote the number

of infected people by I(t). It may be scary if I

(t) > 0, i.e., I(t) increases.

It is worse, and often true in the early stages of an epedemic, if I

(t) > 0.

4

If f and g are inverses of each other, then the graph of one of the functions is obtained

from the one of the other one by reﬂection at the diagonal x = y. In this process, secant

lines which are above the graph turn into secant lines below the graph. Thus, if f is

concave up, then g is concave down, and vice versa.

2.4. THE SECOND DERIVATIVE AND CONCAVITY 79

This means that I

**(t) increases, and the desease spreads at an increasing
**

rate. Medical professional will not necessarily wait for the time when I(t),

the number of infected people, starts decreasing. When I

(t) turns negative,

then I

**(t) decreases. The spread or the desease slows. One may hope that
**

eventually I

**(t) becomes negative, so that the actual number of sick people
**

decreases. The point at which I

**(t) changes signs from being positive to
**

being negative may be considered the turning point in the spread of the

desease. One of the recent presidents was confused by a subtle argument of

this kind

5

.

Let us look at this phenomena in a concrete example. Earlier we con-

sidered the logistic equation

y

= ay −by

2

.

See Example 1.75 and the graph of a solution of this diferential equation in

Figure 1.15. Use implicit diﬀerentiation to ﬁnd the second derivative:

y

= ay

−2byy

= (a −2by)y

.

We see that y

= 0 if y

**= 0 or y = a/(2b). The ﬁrst case occurs if y = 0
**

or y = a/b. We called y = a/b the carrying capacity of the system, and it

was the stable equilibrium point. The inﬂection occurs when y is half the

carrying capacity. As long as y is less than a/(2b), the population grows at

an increasing rate. If a/(2b) < y < a/b, then growth slows. You see the

turning point in the graph in Figure 1.15. For a while the population seems

to explode, but after a while it levels oﬀ so that it does not exceed a the

carrying capacity.

Exercise 17. Find intervals on which the following functions are concave

up, resp., concave down.

1. f(x) = x

3

−4x

2

+ 8x −7

2. g(x) = x

4

+ 2x

3

−3x

2

+ 5x −2

3. h(x) = x + 1/x

4. i(x) = 2x

4

−x

2

5. j(x) = x/(x

2

−1)

6. k(x) = 2 cos

2

x −x

2

for x ∈ [0, 2π].

5

During a televised presidential debate, one of the candidates said (see the New York

Times from October 8th, 1984, page B6): “Some of these facts and ﬁgures just don’t add

up. Yes, there has been an increase in poverty but it is a lower rate of increase than it was

in the preceding years before we got here. It has begun to decline, but it is still going up.”

80 CHAPTER 2. GLOBAL THEORY

2.4.2 Concavity at a Point

The notion of being concave up or down was deﬁned for functions which are

deﬁned on intervals. Still, we got a picture how the function has to look

like near a point, and this is the behavior which we like to capture in a

deﬁnition.

Deﬁnition 2.26. Let f be a function and c an interior point

6

of its do-

main. We say that f is concave up, resp., concave down, at c if there exists

an open interval I and a line l, called a support line, such that l(c) = f(c)

and

f(x) > l(x), resp., f(x) < l(x),

for all x ∈ I with x = c.

0.5 1 1.5 2

1

2

3

4

Figure 2.6: Concave up at •

1 2 3 4 5

-0.5

0.5

1

1.5

2

Figure 2.7: Concave down at •

In other words, we are asking for a line l(x), such that the graph lies

on one side of the graph, at least near c. If the graph is above the line,

then the function is concave up, if it is below, then the function is concave

down. We assume that the graph and the line agree at c. You see this

situation illustrated in two generic pictures in Figures 2.6 and 2.7. One

shows a function which is concave up at the indicated point, one shows a

function which is concave down.

Our next theorem tells us how to detect concavity, and it tells us how

to ﬁnd the support line if the function is diﬀerentiable.

6

The idea of an interior point was deﬁned in Deﬁnition 1.18 on page 10.

2.5. LOCAL EXTREMA AND INFLECTION POINTS 81

Theorem 2.27. Let f be a function and c an interior point of its domain.

1. If f

is increasing at c or if f

(c) > 0, then f is concave up at c.

2. If f

is decreasing at c or if f

**(c) < 0, then f is concave down at c.
**

3. If f is diﬀerentiable and concave up or down at c, then there is only

one support line, and this line is the tangent line to the graph of f at

c.

The sign of the second derivative of a functions tells us whether a function

is concave up or down at a point. If the second derivative is zero, then the

test is inconclusive. The function can be concave up, down, or neither.

In general, there can be many support lines at any given point, but if the

function is diﬀerentiable at c, then the support line is unique. It is the

tangent line. So, for a diﬀerentiable function which is concave up or down

at a point, we can draw the tangent line easily. We just hold the ruler

against the graph.

For example, the function f(x) = x

5

−7x

4

+2x

3

+2x

2

−5x+4 is concave

down at x = 2 because f

(2) = −148 < 0.

To relate concavity properties on an interval to those at each point in

the interval we state, without proof, the following theorem.

Theorem 2.28. Let f be a function which is deﬁned on an open interval

(a, b). Then f is concave up (resp., down) on (a, b) if and only if f is concave

up (resp., down) at each point in (a, b).

2.5 Local Extrema and Inﬂection Points

We are going to discuss two types of points which are particularly important

in the discussion of (graphs of) functions. As we like to apply local properties

of the function, we focus on interior points is the domain of the function.

Deﬁnition 2.29 (Local Extrema). Let f be a function and c an interior

point in its domain

7

. We say that f has a local maximum, resp. minimum,

at c if

f(c) ≥ f(x), resp. f(c) ≤ f(x),

for all x in some open interval I around c. In this case we call f(c) a local

maximum, resp. minimum, of f. A local extremum is a local maximum or

minimum.

82 CHAPTER 2. GLOBAL THEORY

-2 -1 1 2

-2

-1

1

2

3

Figure 2.8: A local minimum

-1 -0.5 0.5 1

-1.5

-1

-0.5

0.5

1

1.5

Figure 2.9: An Inﬂection Point

In other words, f has a local maximum of f(c) at c, if f(c) is the largest

value compared the values at points near c. The function shown in Figure 2.8

has a local minimum at x = −1. We will study tests which allow us ﬁnd local

extrema soon. We do not need any test to see that f(x) = |x| has a local

minimum at x = 0, and f(x) = −(x − 1)

2

has a local maximum at x = 1.

The vertex of a parabola is always a local extremum, a local minimum if the

coeﬃcient of x

2

is positive, and a local maximum if the coeﬃcient of x

2

is

negative.

Deﬁnition 2.30 (Inﬂection Points). Let f be a function and c an inte-

rior point of its domain. We call c an inﬂection point of f if the concavity

of f changes at c. I.e., for some numbers a and b with a < c < b, we have

that f is concave up on the interval (a, c] and concave down on [c, b), or vice

versa.

Soon we will develop tests which detect inﬂections points. No test is

required to see that f(x) = tan x has an inﬂection point at x = 0. The

function is concave down on the interval (−π/2, 0] and concave up on the

interval [0, π/2). So the concavity changes at x = 0 and that means that

there is an inﬂection point at x = 0. You see the graph of this function in

Figure 2.9.

7

According to Deﬁnition 1.18 on page 10 this means that f(x) is deﬁned for all x in

some open interval around c.

2.6. DETECTION OF LOCAL EXTREMA 83

2.6 Detection of Local Extrema

We will discuss how to detect local extrema. The ﬁrst result excludes many

points. Typically, there are very few points where local extrema can occur.

Theorem 2.31. Let f be a function and c an interior point of its domain. If

f is diﬀerentiable at c and f

**(c) = 0, then f does not have a local extremum
**

at c. In other words, if f has a local extremum at c, then f is either not

diﬀerentiable at c or f

(c) = 0.

To have an abbreviation for the points which are recognized as important

in this theorem, it is customary to say:

Deﬁnition 2.32 (Critical Points). Let f be a function and c an interior

point of its domain. We say that c is a critical point of f if f is diﬀerentiable

at c and f

(c) = 0, or if f is not diﬀerentiable at c.

Theorem 2.31 provides us with a necessary condition. If a function has

a local extremum at c, then c is a critical point of the function. No local

extrema can occur at points which are not critical. The test does not give

a suﬃcient condition for a local extremum. If c is a critical point of the

function, then the function need not have a local extremum at c. It makes

sense to introduce one more word.

Deﬁnition 2.33 (Saddle Points). Let f be a function and c an interior

point of its domain. We say that c is a saddle point of f if f is diﬀerentiable

at c and f

**(c) = 0, but f does not have a local extremum at c.
**

Proof of Theorem 2.31. Suppose that f is diﬀerentiable at c and f

(c) > 0.

Proposition 2.22 on page 76 tells us that there exists some positive number

d, such that f(x) < f(c) for all x ∈ (c − d, c), and f(x) > f(c) for all

x ∈ (c, c +d). So, there are points x to the left of and arbitrarily close to c

such that f(x) < f(c), and there are points x to the right of and arbitrarily

close to c such that f(x) > f(c). This means, by deﬁnition, that f does not

have a local extremum at c. If f

**(x) < 0, then the same argument applies
**

with inequalities reversed. If f

(c) = 0, then either f

(c) > 0 or f

(c) < 0,

and in neither case we have an extremum at c.

Neither the exponential function nor the logarithm function have local

extrema. To see this, observe that these functions are diﬀerentiable on their

domain, and their derivatives exp

x = exp x and ln

x = 1/x are every-

where nonzero. These functions have no critical points, and according to

Theorem 2.31 they have no local extrema.

84 CHAPTER 2. GLOBAL THEORY

Example 2.34. Find the local extrema of the function

q(x) = x

2

−2x + 3.

Solution: The function is diﬀerentiable for all real numbers x, and

q

(x) = 2x −2 = 2(x −1).

So q

**(x) = 0 if x = 1. The only point at which we can have a local extremum,
**

i.e., the only critical point, is x = 1. If we write the function in the form

q(x) = (x −1)

2

+ 2,

then we see that q does indeed that a local minimum at x = 1. You should

conﬁrm this result by having a look at Figure 2.10, where this function is

graphed. ♦

-1 1 2 3

2

4

6

8

Figure 2.10: A local minimum

-2 -1 1 2

-8

-6

-4

-2

2

4

6

8

Figure 2.11: A saddle point

Example 2.35. Show that the function g(x) = x

3

has a saddle point at

x = 0.

Solution: The function g(x) is everywhere diﬀerentiable, and its only

critical point is at x = 0, which is the only zero of g

(x) = 3x

2

. Obviously,

g(x) > 0 for all x ∈ (0, ∞) and g(x) < 0 for all x ∈ (−∞, 0). This means

that there is no local extremum at x = 0. As g

(0) = 0 and there is no

local extremum at x = 0, the function has a saddle point at this point. This

saddle point is shown in Figure 2.11. ♦

2.6. DETECTION OF LOCAL EXTREMA 85

Let us formulate a criterion which conﬁrms that a function has a local

extremum at a point c. It gives us a suﬃcient condition for a local extremum

at c.

Theorem 2.36. Suppose c is an interior point of the domain of a function

f, and suppose that for some d > 0 the function is increasing on (c−d, c] and

decreasing on [c, c + d). Then f has a local maximum at c. If the function

is decreasing on (c − d, c] and increasing on [c, c + d), then f has a local

minimum at c.

Taking advantage of the information provided by the ﬁrst derivative, we

obtain the following test.

Theorem 2.37 (First Derivative Test). Suppose f is a function which

is deﬁned and diﬀerentiable on (c − d, c + d) for some d > 0, and c is a

critical point.

1. If f

(x) > 0 for all x ∈ (c − d, c) and f

(x) < 0 for all x ∈ (c, c + d),

then f has a local maximum at c.

2. If f

(x) < 0 for all x ∈ (c − d, c) and f

(x) > 0 for all x ∈ (c, c + d),

then f has a local minimum at c.

3. If f

**(x) > 0 for all x ∈ (c−d, c)∪(c, c+d), then f has a saddle point at
**

c. This conclusion also holds if f

**(x) < 0 for all x ∈ (c−d, c)∪(c, c+d).
**

Let us illustrate the use of the theorem with an example.

Example 2.38. Find the local extrema of the function

f(x) = x

3

−3x

2

+ 2x + 2.

Solution: We diﬀerentiate f(x) and express f

(x) as a product of linear

factors:

f

(x) = 3x

2

−6x + 2 = 3

¸

x −

1 +

√

3

3

¸¸

x −

1 −

√

3

3

¸

It is easy to determine where the factors are zero, positive and negative. We

conclude that f

(x) = 0 if x = 1 ±

√

3/3, f

**(x) is positive on the intervals
**

(−∞, 1 −

√

3/3) and (1 +

√

3/3, ∞), and f

**(x) is negative on the interval
**

(1 −

√

3/3, 1 +

√

3/3). You can see graphs of f and f

in Figures 2.12 and

2.13

86 CHAPTER 2. GLOBAL THEORY

0.5 1 1.5 2

1.2

1.4

1.6

1.8

2.2

2.4

Figure 2.12: f(x) = x

3

− 3x

2

+

2x + 2

0.5 1 1.5 2

-1

1

2

3

4

Figure 2.13: f

(x) = 3x

2

−6x +2

The only only critical points of f are at x = 1±

√

3/3, and these are the

only points where a local extremum can occur. Based on the sign of f

(x) on

intervals to the left and right of these two critical points we see that f has

a local maximum at x = 1 −

√

3/3 and a local minimum at x = 1 +

√

3/3.

♦

Exercise 18. Find the local extrema of the following function:

(1) f(x) =

x

2

+ 3x

x −1

(2) g(x) = sin2x + 2 sin x for x ∈ [0, 2π].

Hint: We discussed the monotonicity properties of these functions in Exam-

ples 2.18 and 2.19.

Exercise 19. Find the local extrema of the following functions. In the last

two problems, (g) and (h), restrict yourself to the interval [0, 2π].

(a) f(x) = 3x

2

+ 5x + 7

(b) f(x) = x

3

−3x

2

+ 6

(c) f(x) = (x + 3)/(x −7)

(d) f(x) = x + 1/x

(e) f(x) = x

3

(1 +x)

(f) f(x) = x/(1 +x

2

)

(g) f(x) = cos 2x + 2 cos x

(h) f(x) = sin

2

x −

√

3 sin x

Hint: You discussed the intervals of monotonicity for these functions in

Exercise 16.

2.6. DETECTION OF LOCAL EXTREMA 87

We may use the second derivative to detect the change of sign of the ﬁrst

derivative, as it is called for in the assumptions in Theorem 2.37.

Theorem 2.39 (Second Derivative Test). Let f be a function and c an

interior point in its domain. Assume also that f

(c) and f

**(c) exist and that
**

f

(c) = 0. If f

(c) > 0, then f has a local minimum at c. If f

(c) < 0, then

f has a local maximum at c.

To apply the theorem to the detection of the local extrema of a diﬀer-

entiable function f(x), we diﬀerentiate f and ﬁnd the critical points, the

zeros of f

(x). Then we diﬀerentiate f

(x). The sign of f

at the critical

points tells us whether we found a local minimum or a local maximum. If

f

(c) = f

**(c) = 0, then the test is inconclusive. There may or may not be a
**

local extremum at c. Furthermore, the function f can have a local extremum

at c, and the assumptions of the test are not satisﬁed. In this sense, the test

provides us with a suﬃcient condition for the existence of a local extremum

at a point. It does not provide us with a necessary condition.

Example 2.40. Find the local extrema of the function (for a graph, see

Figure 2.2 on page 72)

p(x) = x

3

−3x

2

−9x + 3.

Solution: We calculated the ﬁrst derivative,

p

(x) = 3x

2

−6x −9 = 3(x + 1)(x −3).

The critical points of the function are x = −1 and x = 3. Furthermore,

p

(x) = 6x −6 = 6(x −1).

In particular, p

(−1) = −12 and p

**(3) = 12. The second derivative test
**

tells us that we have a local maximum at x = −1, because this is a critical

point and p

**(−1) < 0. We also have a local minimum at x = 3 because at
**

this critical point the second derivative of the function is positive. ♦

Proof of the Second Derivative Test. First, let us assume that f

(c) = 0 and

f

**(c) > 0. We will show that f has a local minimum at c. The assumption
**

that f

**(c) = 0 means that the tangent line to the graph of f at (c, f(c))
**

is horizontal. Its equation is l(x) = f(c). The assumption that f

(c) > 0

means that f is concave up at c (see Theorem 2.27 (1)). Spelled out explicitly

this means that

f(x) > l(x) = f(c)

88 CHAPTER 2. GLOBAL THEORY

for some positive number d and for all x ∈ (c − d, c) ∪ (c, c + d). In other

words, f has a local minimum at c.

The proof that f has a local maximum at c if f

(c) = 0 and f

(c) < 0 is

similar. We leave it to the reader.

Exercise 20. Find the critical points and the local extrema.

(a) f(x) = 4x

2

−7x + 13

(b) f(x) = x

3

−3x

2

+ 6

(c) f(x) = x + 3/x

(d) f(x) = x

2

(1 −x)

(e) f(x) = |x

2

−16|

(f) f(x) = x

2

/(1 +x

2

).

2.7 Detection of Inﬂection Points

We deﬁned an inﬂection point to be a point at which the concavity of a

function changes. If we know where the function is concave up and down,

then we can just answer this question. We want to detect inﬂection points

more eﬃciently. A theorem provides a necessary and a suﬃcient condition

for the existence of an inﬂection point. Let us start out with an example.

Example 2.41. Find the the inﬂection points of the function

g(x) = x

3

−4x

2

+ 3x −5.

-0.5 0.5 1 1.5 2 2.5

-0.5

0.5

1

1.5

2

2.5

Figure 2.14: The graph of g.

-0.5 0.5 1 1.5 2 2.5

-2

2

4

6

8

Figure 2.15: The graph of g

.

2.7. DETECTION OF INFLECTION POINTS 89

You see the graph of g in Figure 2.14 and the one of g

in Figure 2.15.

We calculate the ﬁrst and second derivative of g:

g

(x) = 3x

2

−8x + 3 and g

(x) = 6x −8.

From the formula for the second derivative we conclude that

g

(x) < 0 if x ∈ (−∞, 4/3) and that g

(x) > 0 if x ∈ (4/3, ∞).

This means that g is concave down on the interval (−∞, 4/3] and concave

up on [4/3, ∞). By deﬁnition, we have an inﬂection point at x = 4/3. You

see the inﬂection point indicated as a dot in Figure 2.14. You also see that

g

**(x) has a local extremum at the same point. ♦
**

Theorem 2.42. Let f be a function and c an interior point of its domain.

Suppose that the ﬁrst and second derivatives of f exist at c.

1. If f has an inﬂection point at c, then f

(c) = 0.

2. If f

(c) = 0, f

(c) exists and f

**(c) = 0, then f has an inﬂection
**

point at c.

Example 2.43. Find the inﬂection points of

f(t) = 2t

4

−6t

3

+ 5t

2

−7t + 4.

Solution: We calculate the second derivative of the function and ﬁnd

f

(t) = 24t

2

−36t + 10.

According to the theorem, we have to ﬁnd the zeros of f

(x) to determine

where an inﬂection point can be. The roots are

t =

3

4

±

1

12

√

21 =

9 ±

√

21

12

.

Now, let us check whether there are inﬂection points at either of these values

for t. We calculate the third derivative of f:

f

(t) = 48t −36.

We could plug t = (9 ±

√

21)/12 into the expression for f

, but this is a

bit cumbersome. We see right away that f

(t) = 0 exactly if t = 3/4, and

this means that f

(9 ±

√

21)/12) = 0. The theorem says that the inﬂection

points of f(t) are at t = (9 ±

√

21)/12. ♦

90 CHAPTER 2. GLOBAL THEORY

Apparently, our ability it ﬁnd inﬂection points of a function is limited by

our ability to ﬁnd the zeros of its second derivative. If we are given graphical

information, then this quite easy.

Example 2.44. Find the inﬂection points of the function

f(x) =

1.2 +x

2

−3(sin x)

3

.

-3 -2 -1 0 1 2 3

0.5

1

1.5

2

2.5

3

Figure 2.16: The graph of f.

-3 -2 -1 1 2 3

-2

2

4

6

8

Figure 2.17: The graph of f

.

Apparently, it will take an eﬀort to calculate the second derivative of

this function, and it will be nearly impossible to ﬁnd the zeros of f

. Any

reasonable software has no problem with this. We asked the computer to

graph f and f

**for x ∈ [−3, 3]. You see the graphs in Figures 2.16 and 2.17.
**

A look at the graph of f barely reveals some of the inﬂection points, but

the graph of f

**shows them clearly. Zooming in on parts of the graph f will
**

not improve this. At least in this example, the graph of f

tells us much

more about the concavity of the function f than its own graph. ♦

Exercise 21. Discuss the relation between the inﬂection points of a function

f and the local extrema of its derivative f

.

2.8 Absolute Extrema of Functions

We said that a function f has a local maximum at c if its value at c is largest

in comparison to the values at point near c. In many cases we like to ﬁnd the

maximal value of a function, and where it occurs, anywhere in the domain

of the function. This concept is captured in

2.8. ABSOLUTE EXTREMA OF FUNCTIONS 91

Deﬁnition 2.45. Let f be a function, and c a point in its domain. We say

that f has an absolute maximum at c if f(x) ≤ f(c) for all x in the domain

of f. Then we call f(c) the absolute maximum of f. If f(x) ≥ f(c) for all

x in the domain of f, then we say that f has an absolute minimum at c,

and we call f(c) the absolute minimum of f.

A diﬀerent expression is to say that the function assumes its absolute

extremum at c.

Theorem 2.46. A continuous function on a closed interval [a, b] assumes

its absolute maximum and minimum either at a critical point or at an end-

point of the interval.

Proof. In Theorem 1.17 we asserted that a continuous function assumes

its absolute maximum at some point in the interval. If the function does

not assume its absolute maximum at an endpoint, then it does so at some

interior point c, and the function has a local maximum at c. If f is not

diﬀerentiable at c, the c is critical. If f is diﬀerentiable at c, then f

(c) = 0

by Theorem 2.31, and c is critical as well. The argument for the absolute

minimum is left to the reader.

Example 2.47. Find the absolute extrema of the function

f(x) = x

3

−5x

2

+ 6x + 1

for x ∈ [0, 4].

1 2 3 4

2

4

6

8

Figure 2.18: x

3

−5x

2

+ 6x + 1.

1 2 3 4

-2.5

2.5

5

7.5

10

12.5

Figure 2.19: 3x

2

−10x + 6

Solution: According to Theorem 2.46, the absolute extrema of the func-

tion occur either at one of the end points x = 0, x = 4, or at a critical

92 CHAPTER 2. GLOBAL THEORY

point. In fact, f(0) = 1 and f(4) = 9. The critical points, i.e., the zeros of

f

(x) = 3x

2

− 10x + 6, are x = (5 ±

√

7)/3. Approximate values of these

roots are 2.5486 and .7848. You may also check that f

(x) = 6x −10, and

f

((5 +

√

7)/3) > 0 and f

((5 −

√

7)/3) < 0.

The second derivative test tells us that the function has a local minimum at

x = (5 +

√

7)/3 and a local maximum at x = (5 −

√

7)/3. The approximate

values of the function at these points are

f((5 +

√

7)/3) = 3.1126 and f((5 −

√

7)/3) = .3689.

Comparing the values of f(x) at these four points, we conclude that the

function assumes its absolute maximum of 9 at x = 4, and its absolute

minimum of approximately .3689 at x = (5 −

√

7)/3.

You may compare our calculation with the graphs of f in Figure 2.18

and the one of f

in Figure 2.19. ♦

Exercise 22. Find the absolute extrema of the functions on the indicated

intervals.

(a) f(x) = x

2

−5x + 2 for x ∈ [0, 5]

(b) f(x) = x

3

+ 3x

2

−5x + 2 for x ∈ [−3, 2.5]

(c) f(x) =

√

2 +x/

√

1 +x for x ∈ [0, 5]

(d) f(x) = cos 2x + 2 cos x for 0 ≤ x ≤ 2π

(e) f(x) = sinx + cos x for 0 ≤ x ≤ 2π

2.9 Optimization Story Problems

Many real-life problems are formulated as optimization problems. Calculus

helps us to solve these optimization problems. To avoid lenghty introduc-

tions to real-life problems, we content ourselves with problems of an algebraic

or geometric nature. We consider a few examples and give some problems

for practice.

Example 2.48. Cut a string of length 50 centimeters into two pieces. Use

one piece as the perimeter of an equilateral triangle and the other one as the

perimeter of a disk. How long should each piece be, so that the combined

2.9. OPTIMIZATION STORY PROBLEMS 93

area of the triangle and the circle is minimal? How long should each piece

be, so that the combined area of the triangle and the circle is maximal?

In our solution we will go through several steps.

Introduction of notation: There are many ways to set up the notation

to solve this problem. Among them we say that the side length of the triangle

is a and the radius of the circle is r.

Express information as equations: The perimeter of the triangle

will be 3a and the perimeter of the circle will be 2πr. This means that

3a + 2πr = 50 and a =

50 −2πr

3

.

The height of the triangle is h =

a

√

3

2

, and its area is

a

2

√

3

4

. The area of the

disk is πr

2

. The combined area of the triangle and disk is

A =

a

2

√

3

4

+πr

2

=

√

3

4

50 −2πr

3

2

+πr

2

.

For this to make sense, we need that 0 ≤ r ≤ 25/π.

Formulate the problem mathematically: Find the absolute mini-

mum (maximum) of the function

A(r) =

√

3

4

50 −2πr

3

2

+πr

2

.

for r ∈ [0, 25/π].

Solve the mathematical problem: The derivative of A(r) is

A

(r) = −

4π

3

·

√

3

4

50 −2πr

3

+ 2πr =

−π

√

3

50 −2πr

3

+ 2πr,

and A

(r) = 0 if and only if r = 50/(2π + 6

√

3). We note that A(r) is a

parabola which is open upwards. The critical point, which we just found, is

where the local minimum occurs. It is also the absolute minimum of A(r) on

any interval which contains the critical point. For the end points we have:

A(0) ≈ 120.28 amd A(25/π) ≈ 198.94.

Answer the original question: The combined area of the disk and

the triangle will be minimal if r = 50/(2π + 6

√

3), and it will be maximal

of r = 25/π. In the latter case, all string is used for the circle. ♦

Exercise 23. Repeat the previous example with

1. a disk and a square.

94 CHAPTER 2. GLOBAL THEORY

2. an equilateral triangle and a square.

3. a regular hexagon and a square.

4. a disk and half an equilateral triangle (the angles at 30, 60 and 90

degrees).

5. two geometric shapes of your own choice.

Example 2.49. Construct an open box from a rectangular piece of card

board of length L and width W. What are the dimensions of the box with

the largest possible volume?

In our solution we will go through several steps.

Clariﬁcation and introduction of notation: We construct the box

by making an incision at a 45 degree angle at each corner. Then we fold

up a strip of width x along each side

8

. For yourself, draw a picture of

this production process, and convince yourself that any box obtained by a

diﬀerent process will have smaller volume. To simplify matters, we call the

longer side of the rectangle L and the shorter one W.

Express information as equations: As we folded up a strip of width

x, the box will have width W −2x, length L −2x, height x, and volume

V (x) = (W −2x)(L −2x)x = WLx −2(L +W)x

2

+ 4x

3

.

By construction, x ≥ 0, x ≤ W/2, and x ≤ L/2, in fact x ≤ W/2.

Formulate the problem mathematically: Find the absolute maxi-

mum of the function

V (x) = WLx −2(L +W)x

2

+ 4x

3

for x ∈ [0, W/2].

Solve the mathematical problem: At the end points of the interval

V vanishes, i.e., V (0) = V (W/2) = 0. On the interior of the interval the

function is positive. The derivative of V is

V

(x) = WL −4(W +L)x + 12x

2

.

The zeros of V

are at

x =

1

6

(L +W) ±

L

2

+W

2

−LW

.

8

You could have cut out a square of size x ×x at each corner.

2.9. OPTIMIZATION STORY PROBLEMS 95

The function has an inﬂection point at (W +L)/6, to the right of which V

**is positive and V is concave up, and to the left of which V is concave down.
**

We conclude that V has a local maximum at

x =

1

6

(L +W) −

L

2

+W

2

−LW

.

As the function V (x) has only one local maximum in the interval, the local

maximum is the same as the absolute maximum.

Answer the original question: The box with the largest volume will

have a height of

x =

1

6

(L +W) −

L

2

+W

2

−LW

.

Its width will be W −2x and its length L −2x. ♦

Exercise 24. Repeat the previous example with speciﬁc numbers for the

width and length of the piece of card board.

Exercise 25. Start out with an equilateral piece of card board with side

length a. Make incisions at the corners, and fold up strips along the edges.

You will get an open box whose base is an equilateral triangle. How broad

should the folded up strips be, so that the volume of the box is maximal?

Exercise 26. Modify the problem from above, constructing a box with a

round base from a circular piece of card board.

Exercise 27. What is the largest possible volume for a right circular cone

of slant height a?

Example 2.50. Determine the rectangle of maximal area which can be

placed between the x-axis and the graph of the function f(x) = sin x.

Solution: Draw a graph of sin x so that you can follow the discussion.

Convince yourself that the vertices of the rectangle should be (x, 0), (π −

x, 0), (x, sin x) and (π − x, sin x) for some x ∈ [0, π/2]. The width of the

rectangle is π −2x and its height is sin x, so that its area is

A(x) = (π −2x) sin x.

We need to ﬁnd the absolute maximum for this function for x ∈ [0, π/2].

The ﬁrst derivative of this function is A

(x) = −2 sin x + (π − 2x) cos x.

After a simple algebraic simpliﬁcation, you ﬁnd that

A

**(x) = 0 if and only if tan x =
**

π −2x

2

.

96 CHAPTER 2. GLOBAL THEORY

Find an approximate solution of the equation using Newton’s method or your

calculator. A fairly good approximation of the zero of A

(x) is x

0

= .710462.

Convince yourself

9

that this is the only zero of A

(x) for x ∈ [0, π/2]. We

conclude that x

0

is the only critical point of A(x).

You may calculate A

(x). Substituting x

0

you will see that A

(x

0

) < 0.

It follows from the second derivative test that A(x) has a local maximum

at x

0

. Apparently A(x) = 0 at the end points x = 0 and x = π/2 of the

interval. This tells us that A(x) assumes its absolute maximum at x

0

.

With this, the ﬁnal answer to our problem is: The rectangle of maximal

area which can be placed between the x-axis and the graph of the sine

function will have a width of approximately π −2x

0

= 1.72066 and a height

of sinx

0

= .652183. Its area will be about 1.12218. ♦

To ﬁnd the absolute extrema of a continuous function on an interval of

the form [a, b] we could inspect the values of the function at the critical

points and at a and b. It allows us to decide whether a local extremum is

also an abolute one. Our next result allows us to do the same even if the

interval is not closed and bounded. The assumptions of this theorem are

satisﬁed in many applied problems.

Theorem 2.51. Suppose f is deﬁned on an interval I.

(a) If f is concave up on I and has a local minimum at x

0

, then f assumes

its absolute minimum at x

0

.

(b) If f is concave down on I and has a local maximum at x

0

, then f

assumes its absolute maximum at x

0

.

Example 2.52. Find the absolute minimum of the function

f(x) = x +

1

x

for x ∈ (0, ∞).

Solution: We calculate the ﬁrst and second derivative of f(x):

f

(x) = 1 −

1

x

2

and f

(x) =

2

x

3

.

We ﬁnd that f

(x) = 0 if x = 1, and that f

(x) > 0 for all x in (0, ∞). So f

has a local minimum at x = 1, and f is concave up on (0, ∞). Theorem 2.51

tells us that the absolute minimum of the function is f(1) = 2. ♦

9

One possible argument is that tan x is increasing on the interval [0, π/2), and that

π−2x

2

is decreasing. So these functions can intersect in only one point.

2.9. OPTIMIZATION STORY PROBLEMS 97

Exercise 28. Find the largest possible area for a rectangle with base on

the x-axis and upper vertices on the curve y = 4 −x

2

.

Exercise 29. A rectangular warehouse will have 5000 m

2

of ﬂoor space and

will be separated into two rectangular rooms by an interior wall. The cost of

the exterior walls is $ 1,000.00 per linear meter and the cost of the interior

wall is $ 600.00 per linear meter. Find the dimensions of the warehouse that

minimizes the construction cost.

Exercise 30. One side of a rectangular meadow is bounded by a cliﬀ, the

other three sides by straight fences. The total length of the fence is 600

meters. Determine the dimensions of the meadow so that its area is maximal.

Exercise 31. Draw a rectangle with one vertex at the origin (0, 0) in the

plane, one vertex on the positive x-axis, one vertex on the positive y-axis,

and one vertex on the line 3x + 5y = 15. What are the dimensions of a

rectangle of this kind with maximal area?

Exercise 32. Two hallways, one 8 feet wide and one 6 feet wide, meet at a

right angle. Determine the length of the longest ladder that can be carried

horizontally from one hallway into the other one.

Exercise 33. Inscribe a right circular cylinder into a right circular cone of

height 25 cm and radius 6 cm. Find the dimensions of the cylinder if its

volume is the be a maximum.

Exercise 34. A right circular cone is inscribed in a sphere of radius R.

Find the dimensions of the cone if its volume is to be maximal.

Exercise 35. Find the dimensions of a right circular cone of minimal vol-

ume, so that a ball of radius 10 centimeters can be inscribed.

Exercise 36. Consider a triangle in the plane with vertices (0, 0), (a, 0),

and (0, b). Suppose that a and b are positive, and that (2, 5) lies on the line

through the points (a, 0), and (0, b). What should the slope of the line be,

so that the area of the triangle is minimal?

Exercise 37. Minimize the cost of the material needed to make a round

drum with a volume of 200 liter (i.e., .2 m

3

) if

(a) the drum has a bottom and a top, and the same material is used for

the top, bottom and sides.

(b) the drum has no top (but a bottom) and the same material is used for

the bottom and sides.

98 CHAPTER 2. GLOBAL THEORY

(c) the drum has a bottom and a top, the same material is used for the

top and bottom, and the material for the top and bottom is twice as

expensive as the material for the sides.

(d) the situation is as in the previous case, but the top and the bottom

are cut out of squares, and the left over material is recycled for half

its value.

Exercise 38. Consider a box with a round base and no lid whose interior

is subdivided into six wedge shaped sectors. Which shape should it have,

so that its volume is maximal, assuming you are allowed a ﬁxed amount of

material? More speciﬁcally determine the ratio of radius and height which

will maximize the volume.

Exercise 39. Design a roman window with a perimeter of 4 m which admits

the largest amount of light. (A roman window has the shape of a rectangle

capped by a semicircle.)

Exercise 40. A rectangular banner has a red border and a white center.

The width of the border at top and bottom is 15 cm, and along the sides

10 cm. The total area is 1 m

2

. What should be the dimensions of the banner

if the area of the white area is to be maximized?

Exercise 41. A power line is needed to connect a power station on the

shore line to an island 2 km oﬀ shore. The point on the coast line closest to

the island is 6 km from the power station, and, for all practical purposes, you

may suppose that the shore line is straight. To lay the cable costs $40,000

per kilometer under ground and $70,000 under water. Find the minimal

cost for laying the cable.

Exercise 42. Consider the distance D(x) between a point P(x) = (x, f(x))

on the graph of a diﬀerentiable function f(x) and a point Q = (x

0

, y

0

)

not on this graph. Suppose D(x) has a local minimum at x

1

. Then the

tangent line to the graph of f at x

1

intersects the line joining P(x

1

) and Q

perpendicularly.

2.10 Sketching Graphs

The techiques which we developed so far provide us with some valuable tools

for graphing functions. Let us make a list of data which we may determine,

so that we can sketch a graph rather precisely. Going through the following

2.10. SKETCHING GRAPHS 99

program is also a good review of the material which we developed in this

chapter.

Useful information for graphing a function: We call the function

f(x).

(a) Plot some points on the graph, such as the y-intercept. If the function

is given on a closed interval, plot the values at its endpoints.

(b) Plot the zeros of the function. If you cannot ﬁnd the zeros by analytical

means, try it numerically (Newton’s method).

(c) If possible, decide on which intervals the function is positive, resp.,

negative.

(d) Find the ﬁrst derivative f

(x) of f(x).

(e) Repeat (b) and (c) with f

**(x) in place of f(x). Intervals on which
**

f

**(x) is positive give you intervals on which f(x) is increasing, and
**

intervals on which f

**(x) is negative give you intervals on which f(x)
**

is decreasing. The zeros of f

**(x) provide you with the critical points
**

of f(x). Plot the critical points (x and y value), and keep track of the

intervals on which the function is increasing, resp., decreasing.

(f) Find the second derivative f

(x) of f(x).

(g) Repeat (b) and (c) with f

**(x) in place of f(x). Intervals on which
**

f

**(x) is positive give you intervals on which f(x) is concave up, and
**

intervals on which f

**(x) is negative give you intervals on which f(x)
**

is concave down. Find the inﬂection points of the function, i.e., the

points where the concavity changes. Plot the inﬂection points (x and y

value), and keep track of the intervals on which the function is concave

up, resp., concave down.

(h) Decide at which critical points of f(x) the function has a saddle point

or local extremum, and whether it is a minimum or a maximum.

If you now draw a graph which exhibits all of the properties which you

gathered in the course of the suggested program, then your graph will look

very much like the graph of f(x). More importantly, the graph will have all

of the essential features of the graph of f(x). Let us go through the program

in an example.

100 CHAPTER 2. GLOBAL THEORY

Example 2.53. Discuss the graph of the function

f(x) = x

4

−2x

3

−3x

2

+ 8x −4 for x ∈ [−3, 3].

Solution: To make the discussion a little easier, we note that

f(x) = (x −1)

2

(x

2

−4) = (x −1)

2

(x −2)(x + 2). (2.1)

You should verify this by multiplying out the expression for f(x) in (2.1).

(a): Plot the y intercept of the function and its values at the end points

of the given interval: f(−3) = 80, f(0) = −4 and f(3) = 20.

(b): As a polynomial, the function f(x) is diﬀerentiable on the given

interval. The only exceptional points are its zeros. Having written f(x) as

in (2.1), we see right away that f(x) = 0 if and only if x = −2, x = 1, or

x = 2. Plot these x-intercepts.

(c): Counting the signs of the factors of f(x), we see that f(x) is positive

on the intervals [−3, −2) and (2, 3], and negative on (−2, 1) and (1, 2).

(d): We calculate the derivative of f(x):

f

(x) = 2(x −1)(x

2

−4) + (x −1)

2

2x = 2(x −1)(2x

2

−x −4).

We based the calculation on the description of f(x) in (2.1). In the ﬁrst

step we applied the product rule, and then we used elementary algebra.

(e): We use the quadratic formula to ﬁnd the zeros of the factor 2x

2

−x−4

in the expression for f

(x). They are (1 ±

√

33)/4. This allows us to factor

the expression for f

(x), and we ﬁnd:

f

(x) = 4(x −1)

x −

1

4

[1 +

√

33]

x −

1

4

[1 −

√

33]

.

We conclude that:

• f

**(x) is negative on the interval [−3, (1−
**

√

33)/4) and f(x) is decreas-

ing on [−3, (1 −

√

33)/4].

• f

**(x) is positive on the interval ((1−
**

√

33)/4, 1) and f(x) is increasing

on [(1 −

√

33)/4, 1].

• f

**(x) is negative on the interval (1, (1+
**

√

33)/4) and f(x) is decreasing

on [1, (1 −

√

33)/4].

• f

**(x) is positive on the interval ((1+
**

√

33)/4, 3] and f(x) is increasing

on [(1 +

√

33)/4, 3].

2.10. SKETCHING GRAPHS 101

• f(x) has a critical point and local minimum at (1 −

√

33)/4 ≈ −1.19,

a critical point and local maximum at x = 1, and a critical point and

local minimum at (1 +

√

33)/4 ≈ 1.69.

The values of the function at its three critical points are approximately:

f(

1 −

√

33

4

) ≈ −12.39 & f(1) = 0 & f(

1 +

√

33

4

) ≈ −.54.

Plot these points.

(f): We rewrite the ﬁrst derivative as f

(x) = 4x

3

− 3x

2

− 3x + 4, and

ﬁnd

f

(x) = 12x

2

−12x −6.

(g): We use the quadratic formula to ﬁnd the zeros on f

(x) and factor

it:

f

(x) = 12

x −

1

2

[1 +

√

3]

x −

1

2

[1 −

√

3]

.

We conclude that:

• f

**(x) is positive on the interval [−3, (1 −
**

√

3)/2) and f(x) is concave

up on [−3, (1 −

√

3)/2]

• f

**(x) is negative on the interval ((1 −
**

√

3)/2, (1 +

√

3)/2) and f(x) is

concave down on [(1 −

√

3)/2, (1 +

√

3)/2]

• f

**(x) is positive on the interval ((1+
**

√

3)/2, 3] and f(x) is concave up

on [(1 +

√

3)/2, 3]

• f(x) has inﬂection points at x = (1 −

√

3)/2 ≈ −.37 and at x =

(1 +

√

3)/2 ≈ 1.37.

The values of the function at its inﬂection points is approximately:

f(

1 −

√

3

2

) ≈ −7.21 & f(

1 −

√

3

2

) ≈ −.29.

Plot these points.

(h): At this point we could use the second derivative test to ﬁnd at which

critical points the function has local extrema, but we decided this already

based on ﬁrst derivative behaviour in (e).

102 CHAPTER 2. GLOBAL THEORY

Let us gather and organize our information. We consider the interval:

I

1

= [−3, −2]

I

2

=

¸

−2,

1 −

√

33

4

¸

I

3

=

¸

1 −

√

33

4

,

1 −

√

3

2

¸

I

4

=

¸

1 −

√

3

2

, 1

¸

I

5

=

¸

1,

1 +

√

3

2

¸

I

6

=

¸

1 +

√

3

2

,

1 +

√

33

4

¸

I

7

=

¸

1 +

√

33

4

, 2

¸

I

8

= [2, 3] .

We tabulate the which properties hold on which interval. It should be

understood, that at some end points of intervals the function is zero.

Property I

1

I

2

I

3

I

4

I

5

I

6

I

7

I

8

Sign pos neg neg neg neg neg neg pos

Monotonicity dec dec inc inc dec dec inc inc

Concavity up up up down down up up up

Table 2.1: Properties of the Graph

In Figure 2.20 you see the graph of the function. We have shown it on

a slightly smaller interval, as the values at the endpoint a comparetively

large. Showing all of the graph would show less clearly what happens near

the intercept, extrema, and inﬂection points. The dots indicate the points

which we suggests to plot.

In Figure 2.21 you see the graph of f on an even smaller interval, and

parts of the graphs of f

and f

**. You can use them to see that f is decreasing
**

where f

is negative, f is concave down where f

is negative, etc. ♦

Exercise 43. In analogy with the previous example, discuss the function

f(x) = (x −1)(x −2)(x + 2) = x

3

−x

2

−4x + 4

on the interval [−3, 2.5]. In addition, ﬁnd the absolute extrema of this

function.

2.10. SKETCHING GRAPHS 103

-3 -2 -1 1 2 3

-10

-5

5

10

Figure 2.20: The Graph

-2 -1 1 2

-10

-5

5

10

Figure 2.21: f, f

, f

**Exercise 44. In analogy with the previous example, discuss the function
**

f(x) = x

3

−3x + 2

on the interval [−2, 2]. In addition, ﬁnd the absolute extrema of this func-

tion.

Exercise 45. In analogy with the previous example, discuss the function

f(x) = 2 sin x + cos 3x

on the interval [0, 2π]. In addition, ﬁnd the absolute extrema of this function.

You may have to apply Newton’s method to ﬁnd zeros of f, f

, and f

.

104 CHAPTER 2. GLOBAL THEORY

Chapter 3

Integration

We will introduce the ideas of the deﬁnite and the indeﬁnite integral. Sup-

pose that f is a function which is deﬁned and bounded on the interval [a, b].

If it exists, then the deﬁnite integral of f over the interval [a, b] is a real

number. It is denoted by

b

a

f(x) dx.

The deﬁnition is set up, so that for a non-negative function it makes sense

to think of the integral as the area of the region bounded by the graph of

the function, the x-axis, and the lines x = a and x = b.

The indeﬁnite integral of a function f is the family (set) of all antideriva-

tives of f, i.e. all functions whose derivative is f. For important classes of

functions one may utilize deﬁnite integrals to construct antiderivatives. The

Fundamental Theorem of Calculus relates deﬁnite integrals and antideriva-

tives.

To be concrete, consider the function f(x) = x

2

e

−x

, shown in Figure 3.1,

and ﬁnd the area of the region Ω bounded by the graph of f(x), the lines

x = 1 and x = 5, and the x-axis.

3.1 Properties of Areas

So far, we only know the area of some simple regions, like rectangles. We

will denote the area of a region Ω by Area(Ω). Whatever concept of area we

have in mind, it should have the following properties:

• The area of a rectangle is the product of the lengths of its sides.

105

106 CHAPTER 3. INTEGRATION

1 2 3 4 5 6

0.1

0.2

0.3

0.4

0.5

0.6

Figure 3.1: f(x) = x

2

e

−x

• Suppose that Ω

1

and Ω

2

are regions in the plane, and that the area of

each of them is deﬁned.

If Ω

1

⊆ Ω

2

, then Area(Ω

1

) ≤ Area(Ω

2

).

• Suppose that Ω

1

and Ω

2

are regions in the plane, and that the area

of each of them is deﬁned. If the regions Ω

1

and Ω

2

do not intersect,

then the area of the union Ω

1

∪ Ω

2

of Ω

1

and Ω

2

is deﬁned, and

Area(Ω

1

∪ Ω

2

) = Area(Ω

1

) + Area(Ω

2

).

Suppose for a moment, that the region under the graph shown in Fig-

ure 3.1 has an area. In Figure 3.2 you see a rectangle R

l

with area .6, which

is contained in Ω. In Figure 3.3 you see a rectangle R

u

with area 2.24 which

contains Ω. The ﬁrst two principles tell us that

Area(R

l

) = .6 ≤ Area(Ω) ≤ Area(R

u

) = 2.24.

From above principles one may derive another one, which occurs fre-

quently in our upcoming constructions:

3.2. PARTITIONS AND SUMS 107

1 2 3 4 5 6

0.1

0.2

0.3

0.4

0.5

0.6

Figure 3.2: A rectangle R

l

con-

tained in Ω

1 2 3 4 5 6

0.1

0.2

0.3

0.4

0.5

0.6

Figure 3.3: A rectangle R

u

con-

taining Ω

• Suppose the region R in the plane is the union of a ﬁnite number of

rectangles R

1

, . . . , R

n

and any two of them intersect at most in an

edge. Then Area(R) is deﬁned, and it is equal to the sum of the areas

of the regions R

1

, . . . , R

n

:

Area(R) = Area(R

1

) +· · · + Area(R

n

).

3.2 Partitions and Sums

We like to reﬁne the approach to calculating areas of regions which we

started in the previous section. We do so by partitioning the interval before

applying the ideas from above, and then we add up what we get over the

individual intervals.

A partition of an interval [a, b] is of a collection is points {x

j

| 0 ≤ j ≤ n},

such that

a = x

0

≤ x

1

≤ · · · ≤ x

n−1

≤ x

n

= b.

The interval [a, b] is partitioned into n intervals [x

j−1

, x

j

] with 1 ≤ j ≤ n.

108 CHAPTER 3. INTEGRATION

3.2.1 Upper and Lower Sums

As before, f denotes a function which is deﬁned and bounded on [a, b]. On

each interval we pick numbers m

j

and M

j

, such that

m

j

≤ f(x) ≤ M

j

for all x ∈ [x

j−1

, x

j

].

We deﬁne the lower sum to be

S

l

= m

1

(x

1

−x

0

) +m

2

(x

2

−x

1

) +· · · +m

n

(x

n

−x

n−1

). (3.1)

and the upper sum to be

S

u

= M

1

(x

1

−x

0

) +M

2

(x

2

−x

1

) +· · · +M

n

(x

n

−x

n−1

). (3.2)

These sums depend on the choice of partition and the choices for the m

j

and M

j

.

1 2 3 4 5 6

0.1

0.2

0.3

0.4

0.5

0.6

Figure 3.4: A union of rectangles

contained in Ω

1 2 3 4 5 6

0.1

0.2

0.3

0.4

0.5

0.6

Figure 3.5: A union of rectangles

containing Ω

Let us return to the example of the function f(x) = x

2

e

−x

on the interval

[1, 4]. In the computation of the lower sum we use the partition

x

0

= 1 < x

1

= 2 < x

2

= 3 < x

3

= 4 < x

4

= 5

of the interval. We also pick m

1

= .35, m

2

= .43, m

3

= .28 and m

4

= .16.

This leads to a lower sum S

l

= 1.22. In the computation of the upper sum

we use the partition

x

0

= 1 < x

1

= 3 < x

2

= 4 < x

3

= 5

3.2. PARTITIONS AND SUMS 109

of the interval. We also pick M

1

= .55, M

2

= .45 and M

3

= .3. This leads

to an upper sum S

u

= 1.85. The m

j

and M

j

represent the heights of the

rectangles in Figures 3.4 and 3.5, and we trust these ﬁgures to show that

m

j

≤ f(x) and f(x) ≤ M

j

on the respective interval.

As before, let Ω denote the region under the graph. Then the union of

the rectangles shown in Figure 3.4 is contained in Ω, and the union of the

rectangles shown in Figure 3.5 contains Ω. Thus, if Ω has an area, the our

principles tell us that

S

l

= 1.22 ≤ Area(Ω) ≤ S

u

= 1.85.

In fact the only number greater or equal to all lower sums and smaller

or equal to all upper sums is

5

e

−

37

e

5

, and this will be the area of the region

Ω. Here e is the Euler number.

Example 3.1. Let us ﬁnd upper and lower sums for the function

f(x) = x

3

−7x

2

+ 14x −8

for x ∈ [.5, 4.5]. In contrast to the function in the previous example, this

function is not non-negative.

1 2 3 4

-3

-2

-1

1

2

3

4

Figure 3.6: Rectangles for calcu-

lating an upper sum.

1 2 3 4

-3

-2

-1

1

2

3

4

Figure 3.7: Rectangles for calcu-

lating a lower sum.

Solution: For the purpose of calculating an upper sum, we partitioned

the interval [.5, 4.5] using the intermediate points x

0

= .5, x

1

= 1.1, x

2

= 2.4,

x

3

= 3.8, and x

4

= 4.5. As numbers M

i

(so that M

i

≥ f(x) for x ∈ [x

i−1

, x

i

])

110 CHAPTER 3. INTEGRATION

we chose M

1

= .3, M

2

= .7, M

3

= −.9, and M

4

= 4.4. These data are shown

in Figure 3.6. With these choices, the upper sum is

S

u

= .3(1.1 −.5) +.7(2.4 −1.1) + (−.9)(3.8 −2.4) + 4.4(4.5 −3.8)

= 2.91.

In Figure 3.6 you see four rectangles. Their areas are combined to calculate

the upper sum. The areas of the ones above the x-axis are added, the ones

below the axis are subtracted, in accordance with the sign of the M

i

.

In the calculation of the lower sum we partitioned [.5, 4.5] using x

0

= .5,

x

1

= .8, x

2

= 2.3, x

3

= 4.2, and x

4

= 4.5. As numbers m

i

(so that

m

i

≤ f(x) for x ∈ [x

i−1

, x

i

]) we chose m

1

= −2.7, m

2

= −.8, m

3

= −2.2,

and m

4

= 1.3. These data are shown in Figure 3.7. With these choices we

calculate a lower sum of

S

l

= −2.7(.8 −.5) + (−.8)(2.3 −.8) + (−2.2)(4.2 −2.3) + 1.3(4.5 −4.2)

= −5.8.

In Figure 3.7 you see four rectangles. Their areas are combined to calculate

the lower sum. The areas of the ones above the x-axis are added, the ones

below the axis are subtracted, in accordance with the sign of the m

i

.

In summary, you see that we still combine areas of rectangles in the

calculation of the upper and lower sum, only that, depending on the sign

of the M

i

or m

i

, these rectangles are either above or below the x-axis, and

depending on this, their areas are either added or subtracted. ♦

Let us make a simple albeit important observation:

Theorem 3.2. Let f be a function which is deﬁned and bounded on a closed

interval [a, b]. Let S

l

be any lower sum of f and S

u

any upper sum. Then

S

l

≤ S

u

.

Let us repeat the statement of the theorem to emphasize its meaning.

Whichever partition of the interval [a, b] and whichever m

i

we use in the

calculation of the lower sum S

l

and whichever partition of the interval and

whichever M

i

we use in the calculation of the upper sum S

u

, the lower sum

is always smaller or equal to the upper sum. To see this, one reﬁnes the

partitions for the upper and lower sum computation so that they become

the same. Then one notes that m

i

≤ M

i

for all i.

3.2. PARTITIONS AND SUMS 111

3.2.2 Riemann Sums

Suppose once again that f(x) is a function which is deﬁned on the interval

[a, b]. Pick once more a partition

a = x

0

≤ x

1

≤ · · · ≤ x

n−1

≤ x

n

= b

of the interval. In each subinterval, pick a point x

j

∈ [x

j−1

, x

j

]. Then we

deﬁne the Riemann Sum

S

R

= f(x

1

)(x

1

−x

0

) +f(x

2

)(x

2

−x

1

) +· · · +f(x

n

)(x

n

−x

n−1

). (3.3)

We leave it to the reader to contemplate

Proposition 3.3. Let f be a function which is deﬁned and bounded on a

closed interval [a, b]. Let S

l

be any lower sum of f, S

u

any upper sum, and

S

R

any Riemann sum. Then

S

l

≤ S

R

≤ S

u

.

1 2 3 4 5 6

0.1

0.2

0.3

0.4

0.5

0.6

Figure 3.8: Representing a Riemann Sum

To be more concrete, let us return to the example f(x) = x

2

e

−x

on the

interval [1, 5]. Let us use the partition

x

0

= 1 < x

1

=

√

3 < x

2

= 5.

In the two interval of this subdivision we pick the points x

1

=

√

2 ∈ [1,

√

3]

and x

2

= π ∈ [

√

3, 5]. As Riemann sum we obtain

S

R

= f(x

1

)(x

1

−x

0

) +f(x

2

)(x

2

−x

1

) ≈ 1.749741.

In Figure 3.8 you see the picture illustrating the computation. There are

two rectangles, their bases are the intervals in the subdivision, and their

heights are f(x

1

) and f(x

2

). The sum of the areas of these rectangles is the

Riemann sum.

112 CHAPTER 3. INTEGRATION

3.3 Limits and Integrability

The idea is to reﬁne the partitions in our previous construction, so that in

the limit our sums can be justiﬁably called the area of the region under the

graph, if the function is non-negative. The speciﬁcs depend on which sums

we are working with.

3.3.1 The Darboux Integral and Areas

As we discussed earlier, whatever choices we make in the calculation of

lower and upper sums S

l

and S

u

, we always have that S

l

≤ S

u

. A crucial

additional fact is stated in the next result.

Theorem 3.4. Let f be a function which is deﬁned and bounded on a closed

interval [a, b]. There exists a real number Y , such that

S

l

≤ Y ≤ S

u

for all lower sums S

l

and upper sums S

u

of f.

Idea of Proof. To deduce the theorem from the completeness of the real

numbers, one observes that the set of all lower sums of f has a least upper

bound. Call it Y

l

. The set of all upper sums of f has a greatest lower

bound. Call it Y

u

. Apparently, Y

l

≤ Y

u

. Then Y is any number such that

Y

l

≤ Y ≤ Y

u

.

We are now prepared to deﬁne the concept of integrability of a function.

Deﬁnition 3.5. Let f be a function which is deﬁned and bounded on a

closed interval [a, b]. If there is exactly one number Y , such that

S

l

≤ Y ≤ S

u

for all lower sums S

l

and all upper sums S

u

of f, then we say that f is

integrable over the interval [a, b]. In this case, the number Y is called the

integral

1

of f for x between a and b. It is also denoted by

b

a

f(x) dx.

1

To distinguish it from the result of a diﬀerent, but typically equivalent, construction

we should Y the Darboux integral.

3.3. LIMITS AND INTEGRABILITY 113

Remark 8. For completeness sake and later use, let us explain what hap-

pens when a function is not integrable. In this case there are at least two

diﬀerent numbers, and with this an entire interval, between all upper and

lower sums. So, a function over a closed interval [a, b] is not integrable if

and only if the exists a positive number D such that S

u

− S

l

≥ D for any

lower sum S

u

and any upper sum S

u

.

On the other hand, a function is integrable if for every positive number

D there is an upper sum S

u

and a lower sum S

l

such that S

u

−S

l

< D.

Example 3.6. Explore upper sums, lower sums, and integrability for the

function f(x) = x

2

on the interval [0, 1].

Solution: Fix a natural number n and set

x

0

= 0 < x

1

=

1

n

< x

2

=

2

n

< · · · < x

n−1

=

n −1

n

< x

n

=

n

n

= 1.

This is an equidistant partition of the interval [0, 1], all subintervals have

the same length 1/n.

For the upper sums we pick

M

1

= f(x

1

) =

1

n

2

, M

2

= f(x

2

) =

2

n

2

, M

3

= f(x

3

) =

3

n

2

, . . .

and M

j

= f(x

j

) =

j

n

2

in general. Apparently, M

j

≥ f(x) for all x ∈

[x

j−1

, x

j

] because f(x) is increasing on [0, 1]. Without proof, we use that

1

2

+ 2

2

+ 3

2

+· · · +n

2

=

n(n + 1)(2n + 1)

6

.

We calculate the upper sum

S

u

= M

1

(x

1

−x

0

) +M

2

(x

2

−x

1

) +· · · +M

n

(x

n

−x

n−1

)

=

1

n

2

×

1

n

+

2

n

2

×

1

n

+· · · +

n

n

2

×

1

n

=

1

n

3

1

2

+ 2

2

+· · · n

2

=

n(n + 1)(2n + 1)

6n

3

=

1

3

+

1

2n

+

1

6n

2

For the lower sums we pick

m

1

= f(x

0

) = 0, m

2

= f(x

1

) =

1

n

2

, m

3

= f(x

2

) =

2

n

2

, . . .

114 CHAPTER 3. INTEGRATION

-0.2 0.2 0.4 0.6 0.8 1 1.2

0.25

0.5

0.75

1

1.25

Figure 3.9: Rectangles for calcu-

lating a lower sum.

-0.2 0.2 0.4 0.6 0.8 1 1.2

0.25

0.5

0.75

1

1.25

Figure 3.10: Rectangles for calcu-

lating an upper sum.

and m

j

= f(x

j−1

) =

j−1

n

2

in general. The resulting lower sum is

S

l

=

1

3

−

1

2n

+

1

6n

2

For n = 5 you see the rectangles whose areas are the summands in the

lower and upper sums in Figures 3.9 and 3.10.

Using the expressions for S

u

and S

l

you see that S

u

− S

l

= 1/n. We

do not only see that S

l

≤

1

3

≤ S

u

, but also that Y = 1/3 is the only real

number, so that S

l

≤ Y ≤ S

u

for all natural numbers n. According to the

deﬁnition this means, that f(x) = x

2

is integrable over the interval [0, 1]

and that

1

0

x

2

dx =

1

3

. ♦

We motivated our introduction of upper and lower sums by our quest to

deﬁne the concept of area. Our answer is formulated as a

Deﬁnition 3.7. Let f be a function which is deﬁned, bounded, and non-

negative on a closed interval [a, b]. Let Ω be the region bounded by the graph

of f, the x-axis, and the lines x = a and x = b. If f is integrable over this

interval, then we say that the region Ω has an area and

Area(Ω) =

b

a

f(x) dx.

3.3. LIMITS AND INTEGRABILITY 115

The upper and lower sum were constructed such that if there is any

justiﬁcation to assigning an area to Ω then

S

l

≤ Area(Ω) ≤ S

u

.

For an integrable function there is exactly one real number between the

lower and upper sums, so this is the only number which we can call the area

of Ω.

For example, the area of the region Ω bounded by the graph of the

function f(x) = x

2

, the x-axis, and the lines x = 0 and x = 1 is

Area(Ω) =

1

0

x

2

dx =

1

3

. ♦

3.3.2 The Riemann Integral

Earlier we introduced the idea of a Riemann sum. Consider an interval [a, b]

and a function f(x) deﬁned on it. We picked a partition

P : a = x

0

≤ x

1

≤ x

2

≤ · · · x

n−1

≤ x

n

= b,

which broke [a, b] up into smaller interval [x

j−1

, x

j

]. In each of the subin-

tervals we picked a point x

j

∈ [x

j−1

, x

j

], and set

S

R

= f(x

1

)(x

1

−x

0

) +f(x

2

)(x

2

−x

1

) +· · · +f(x

n

)(x

n

−x

n−1

).

We want to consider a limit Riemann sums. This is trickier than for

functions, because there are a lot of choices which we make to deﬁne such a

sum. We deﬁne the norm of the partition P to be

|P| = max{x

j

−x

j−1

| 1 ≤ j ≤ n},

in other words, the norm of P is the length of the longest of the intervals

[x

j−1

, x

j

].

Deﬁnition 3.8 (Limit for Riemann Sums). Suppose the function f(x)

is deﬁned on [a, b]. We say that

L = lim

|P|→0

S

R

if for all > 0 there exists a δ > 0, such that |L−S

R

| < whenever |P| < δ.

If the limit of the S

R

exists, then we say that f is Riemann integrable over

[a, b], call L the Riemann integral of f, and write

L = lim

|P|→0

S

R

=

b

a

f(x) dx.

116 CHAPTER 3. INTEGRATION

Thus L = limS

R

if we can force S

R

to be close to L, as close as we like,

by making the partition ﬁne, by making each subinterval no longer that

some number.

It is worth pointing out and not very diﬃcult to show the following

proposition.

Proposition 3.9. Suppose the function f is deﬁned on the interval [a, b].

Then f is Riemann integrable if and only if it is Darboux integrable. If

deﬁned, the Riemann and the Darboux integral are the same.

3.4 Integrable Functions

We like to provide a supply of integrable functions. Our ﬁrst result is typi-

cally proved in an analysis course.

Theorem 3.10. Suppose f is deﬁned and continuous on [a, b]. Then f is

integrable over [a, b].

According to this theorem, polynomials are integrable over any interval

of the form [a, b]. Rational functions (i.e., functions of the form p(x)/q(x)

where p(x) and q(x) are polynomials) are integrable over intervals of the

form [a, b] as long as q does not vanish anywhere on the interval. The

trigonometric functions (sin, cos, tan, cot, sec, and csc) are integrable on

intervals where the functions are deﬁned. Arbitrary powers of a variable,

f(x) = x

α

, are integrable. One just needs to make sure that the function is

deﬁned on the interval [a, b]. For any real number α it suﬃces to assume that

a > 0. For any real α ≥ 0, it suﬃces to assume a ≥ 0. For rational numbers

α = p/q, where p and q are integers and q is odd, it suﬃces to assume

0 ∈ [a, b]. For non-negative integers α no assumption needs to be made on a

and b. Just making sure that the resulting functions are deﬁned everywhere

on [a, b], the functions just mentioned may be added, subtracted, multiplied,

divided, and composed, and one still ends up with integrable functions.

Let us introduce another class of functions for which we can prove that

they are integrable.

Deﬁnition 3.11. Suppose f(x) is a function. We say that f(x) is non-

decreasing if f(x

1

) ≤ f(x

2

) whenever x

1

and x

2

are in the domain of f(x)

and x

1

≤ x

2

. We say that f(x) is non-increasing if f(x

1

) ≥ f(x

2

) whenever

x

1

≤ x

2

.

Proposition 3.12. Let [a, b] be a closed interval and let f be deﬁned and

non-increasing or non-decreasing on [a, b]. Then f is integrable on [a, b]. In

particular, monotonic (increasing or decreasing) functions are integrable.

3.4. INTEGRABLE FUNCTIONS 117

Proof. We will use Darboux integrability. Let us assume that the function

f is non-decreasing on the interval. The non-increasing case is left as an

exercise. Take any partition of the interval:

a = x

0

< x

1

< · · · < x

n

= b.

The reader may justify why we can use the same partition in the computation

of the upper and lower sum. For i = 1, . . . , n we set

m

i

= f(x

i−1

) & M

i

= f(x

i

).

Then, because f is non-decreasing,

m

i

≤ f(x) ≤ M

i

for all x ∈ [x

i−1

, x

i

].

We use the m

i

and M

i

to compute upper and lower sums. Let ∆ be the

largest value of the x

i

−x

i−1

. Then

S

u

−S

l

= [M

1

(x

1

−x

0

) +· · · +M

n

(x

n

−x

n−1

)]

−[m

1

(x

1

−x

0

) +· · · +m

n

(x

n

−x

n−1

)]

= (M

1

−m

1

)(x

1

−x

0

) +· · · + (M

n

−m

n

)(x

n

−x

n−1

)

≤ [(M

1

−m

1

) + (M

2

−m

2

) +· · · + (M

n

−m

n

)] ∆

= (M

n

−m

1

)∆

= [f(b) −f(a)]∆

The inequality in the computation follows from the choice of ∆. The

second to last equality follows because M

i−1

= m

i

for all i = 2, . . . , n.

Many terms in the computation cancel. Given any positive number D, we

can make the partition ﬁne enough so that [f(b) −f(a)]∆ < D. According

to our Remark 8 this means that f is integrable over the interval, as we

claimed.

We illustrate the steps in the proof in a concrete example. In Figure 3.11

you see the upper and lower sum. The lower sum is the sum of the areas

of the darkly shaded rectangles. The upper sum is the sum of the areas of

the lightly and darkly shaded rectangles. The diﬀerence between the upper

and the lower sum is the sum of the lightly shaded rectangles shown in

Figure 3.12. We can combine these areas by sliding the rectangles sideways

so that they form one column. Its height will be f(b) −f(a). Its width may

vary, but in the widest place it is no wider than ∆, the width of the largest

interval in the partition of [a, b]. That means, the diﬀerence between the

upper and the lower sum is at most [f(b) −f(a)]∆. As above, we conclude

that the function is integrable.

118 CHAPTER 3. INTEGRATION

0.2 0.4 0.6 0.8 1

-0.2

0.2

0.4

0.6

0.8

1

1.2

Figure 3.11: Rectangles for calcu-

lating a lower and an upper sum.

0.2 0.4 0.6 0.8 1

-0.2

0.2

0.4

0.6

0.8

1

1.2

Figure 3.12: Rectangles for calcu-

lating the diﬀerence between an

upper and a lower sum.

Remark 9. There are functions which are not integrable over any interval

of the form [a, b] with a < b.

Remark 10. Here we only discuss integrability of function over closed ﬁnite

intervals, i.e., intervals of the form [a, b]. The discussion of integrability of

functions over intervals which are not of this form, e.g., half-open intervals

like [a, b) or unbounded closed intervals like [a, ∞), requires additional ideas

and techniques which we are not ready to discuss yet.

3.5 Some elementary observations

In spite of our success calculating some integrals using upper and lower

sums and the deﬁnition, this is certainly not the way to go in general. To

integrate “well behaved” functions we want a theory which allows us to

calculate integrals more easily. We have to develop a few basic tools. These

are fairly straight forward consequences of the deﬁnition of the integral.

Proposition 3.13. If the function f is deﬁned at a, then

a

a

f(x) dx = 0 (3.4)

Proof. The reader should contemplate the proposition.

3.5. SOME ELEMENTARY OBSERVATIONS 119

Proposition 3.14. Let [a, b] be a closed interval, c a point between a and

b, and f a function which is deﬁned on the interval. Then

c

a

f(x) dx +

b

c

f(x) dx =

b

a

f(x) dx. (3.5)

Implicitly in the formulation of the proposition is the statement that f

is integrable over [a, b] if and only if it is integrable over the intervals [a, c]

and [c, b]. If one of the sides of Equation (3.5) exists, then so does the other

one.

Idea of Proof. Use c as one of the points in the partition. The remaining

details are left to the reader.

As an immediate consequence of Propositions 3.12 and 3.14 we ﬁnd

Corollary 3.15. Let f be deﬁned on the interval [a, b]. Suppose that we

can partition the interval into a ﬁnite number of intervals such that f is

non-increasing or non-decreasing on each of them. Then f is integrable on

[a, b].

We can also extend Theorem 3.10.

Deﬁnition 3.16. Suppose that f is deﬁned on an interval [a, b]. We call f

piecewise continuous if there is a partition

a = x

0

< x

1

< · · · < x

n−1

< x

n

= b

such that f is continuous on the open intervals (x

j−1

, x

j

) for all 1 ≤ j ≤ n,

and the one-sided limits (see Section 1.3)

lim

x→x

+

j−1

f(x) and lim

x→x

−

j

f(x).

exist and are ﬁnite.

Corollary 3.17. If f is a piecewise contiuous function on [a, b], then f is

integrable on [a, b].

Idea of Proof. According to Proposition 3.14 we may break the problem up,

and consider it over each of the intervals [x

j−1

, x

j

] separately. On each of

these smaller intervals, we can change the deﬁntion of the function at a point

or two and make it continuous. This changes neither the integrability nor

the value of the integral. So the assertion follows from Theorem 3.10.

120 CHAPTER 3. INTEGRATION

Deﬁnition 3.18. Let f be deﬁned and integrable on the interval [a, b]. Then

b

a

f(x) dx = −

a

b

f(x) dx.

This deﬁnition is convenient and consistent with what we have said so

far about the integral. The approach to integrals via lower and upper sums

could also be generalized to include integrals

b

a

where b < a, leading to

exactly this formula.

Using the deﬁnition of the integral it is not diﬃcult to show:

Proposition 3.19. Let [a, b] be a closed interval and c a scalar. Suppose

that f and g are integrable over the interval. Then f +g and cf are integrable

over [a, b] and

b

a

(f(x) +g(x)) dx =

b

a

f(x) dx +

b

a

g(x) dx

and

b

a

cf(x) dx = c

b

a

f(x) dx.

We mention a few useful estimates for integrals.

Proposition 3.20. If f is integrable over [a, b], and f(x) ≥ 0 for all x ∈

[a, b], then

b

a

f(x) dx ≥ 0.

Proof. The proof is left to the reader.

Corollary 3.21. If h and g are integrable over [a, b], and g(x) ≥ h(x) for

all x ∈ [a, b], then

b

a

g(x) dx ≥

b

a

h(x) dx.

Proof. Use that f(x) = g(x) −h(x) ≥ 0 for all x ∈ [a, b].

Proposition 3.22. Let [a, b] be a closed interval and f integrable over [a, b].

Then the absolute value of f is integrable over [a, b], and

b

a

f(x) dx

≤

b

a

|f(x)| dx. (3.6)

The proof of this proposition is elementary, though a bit tricky.

3.6. AREAS AND INTEGRALS 121

3.6 Areas and Integrals

Let us return to the relation between areas and integrals. Suppose f(x) is

a non-negative integrable function over an interval [a, b]. If Ω is the area

bounded by the graph of f(x), the x-axis, and the lines x = a and x = b,

then

Area(Ω) =

b

a

f(x) dx.

The question is, what happens if f(x) is not non-negative?

Let f be a function which is deﬁned and bounded on a closed interval

[a, b] and Ω the set of points which lie between the graph of f(x) and the

x-axis for a ≤ x ≤ b. We decompose Ω into the union of two sets, Ω

+

and Ω

−

. Speciﬁcally, Ω

+

consist of those points (x, y) in the plane for which

a ≤ x ≤ b and 0 ≤ y ≤ f(x), and Ω

−

of those points for which a ≤ x ≤ b and

f(x) ≤ y ≤ 0. Then Ω is the union of the sets Ω

+

and Ω

−

. We decompose

the region between the x-axis and the graph into the part Ω

+

above the

x-axis and the part Ω

−

below it. Making use of this notation, we have:

Proposition 3.23. If f is integrable, then the areas of the regions Ω

+

and

Ω

−

are deﬁned

2

and

b

a

f(x) dx = Area(Ω

+

) −Area(Ω

−

). (3.7)

Idea of Proof. We deﬁne two functions:

f

+

(x) =

f(x) if f(x) ≥ 0

0 if f(x) ≤ 0

and f

−

(x) =

f(x) if f(x) ≤ 0

0 if f(x) ≥ 0

It is elementary, though a bit tricky, to show that the integrability of f(x)

implies the integrability of f

+

(x) and f

−

(x). Apparently, f = f

+

+ f

−

, so

that the additivity of the integral implies that

b

a

f(x) dx =

b

a

f

+

(x) dx +

b

a

f

−

(x) dx. (3.8)

According to Deﬁnition 3.7 we have

Area(Ω

+

) =

b

a

f

+

(x) dx. (3.9)

2

If you want to be formal, then you have to ﬂip the region Ω

−

to lie above the x-axis.

Only then have we addressed the question of it having an area.

122 CHAPTER 3. INTEGRATION

Let −Ω

−

be the area obtained by ﬂipping Ω

−

up, i.e., we take its mirror

image along the x-axis. This process does not change areas, so Area(Ω

−

) =

Area(−Ω

−

). The function −f

−

(x) is non-negative, and −Ω

−

is bounded by

the graph of −f

−

(x), the x-axis, and the lines x = a and x = b. According

to Deﬁnition 3.7 and our elementary properties of the integral we have

Area(Ω

−

) = Area(−Ω

−

) =

b

a

−f

−

(x) dx = −

b

a

f

−

(x) dx. (3.10)

Our claim follows now by substituting the results in (3.9) and (3.10) into

(3.8).

For example,

π/2

−π/2

sin x dx = 0 because the graph bounds congruent

regions above and below the x-axis.

3.7 Anti-derivatives

Consider a function f(x) with domain I. In Deﬁnition 2.6 we called a func-

tion F(x) with domain I an antiderivative of f(x) if F

(x) = f(x). Having

an anti-derivative of a function will (typically) make it easy to integrate it

over a closed interval.

Remember that any antiderivatives F

1

and F

2

of a function f on an

interval I diﬀer only by a constant (see Corollary 2.5). In other words,

there exists a constant c, such that

F

1

(x) = F

2

(x) +c for all x ∈ I.

Deﬁnition 3.24. Let f be a function which is deﬁned on an interval I, and

suppose that f has an antiderivative. The set of all antiderivatives of f is

called the indeﬁnite integral of f. It is denoted by

f(x) dx.

Given a function f and an antiderivative F of it, we typically write

f(x) dx = F(x) +c. (3.11)

In this expression c stands for an arbitrary constant. Diﬀerent values for c

result in diﬀerent functions. Allowing all real numbers as possible values for

c, we understand the the right hand side of (3.11) as a set of functions. The

constant c in the expression is referred to as integration constant.

3.7. ANTI-DERIVATIVES 123

Example 3.25. Given a function f(x) we might know or guess a function

F(x), such that F

**(x) = f(x). Then we can write down the indeﬁnite
**

integral of f in the form F(x) + c. You can check the correctness of your

guess by diﬀerentiation. You may want to consult Table 1.3 on page 63 to

come up with ideas for antiderivatives. Here are some examples.

1 dx = x +c

√

x dx =

2

3

x

3/2

+c

sin x dx = −cos x +c

cos x dx = sin x +c

dx

1 +x

2

= arctan x +c

x dx =

1

2

x

2

+c

x

n

dx =

1

n + 1

x

n+1

(n = −1)

sec

2

x dx = tan x +c

sec xtan x dx = sec x +c

dx

√

1 −x

2

= arcsin x +c

Using the linearity of the diﬀerentiation (see the diﬀerentiation rules

in (1.20)), it is easy to produce more examples. E.g.

5x

2

−2 cos x dx =

5

3

x

3

−2 sin x +c.

Occasionally, an additional idea is required before we can see the anti-

derivative. E.g., using the trigonometric identity cos

2

x = (1 + cos 2x)/2,

we ﬁnd that

cos

2

x dx =

1

2

(1 + cos(2x)) dx =

1

2

¸

x +

1

2

sin(2x)

+c.

Using a diﬀerent trigonometric identity we ﬁnd

(1 + cot

2

x) dx =

csc

2

x dx = −cot x +c. ♦

We shall explore additional ideas for ﬁnding antiderivatives at a later.

The reader may practice ﬁnding some antiderivatives for the functions in

the next exercise. As you go through them you are expected to learn, or

pick up some new ideas as you go along.

124 CHAPTER 3. INTEGRATION

Exercise 46. Find the following indeﬁnite integrals:

(a)

3 dx

(b)

(x + 4) dx

(c)

(x

2

−5) dx

(d)

cos 2x dx

(e)

(3 +x)

3

dx

(f)

(3 + 2x)

5

dx

(g)

1

x

3

dx

(h)

csc

2

x dx

(i)

(1 + tan

2

x) dx

(j)

csc xcot x dx

(k)

sin

2

x dx

(l)

sec

2

(3x) dx

(m)

e

x/3

dx

(n)

2x

x

2

+ 1

dx

(o)

(4 −3x)

5

dx

(p)

cos(4 −3x) dx

(q)

2x

(x

2

+ 3)

2

dx

(r)

xsec

2

(x

2

+ 5) dx

3.8 The Fundamental Theorem of Calculus

Our ﬁrst result provides us with a large class of functions which have an-

tiderivatives.

Theorem 3.26. Continuous functions, deﬁned over intervals, have anti-

derivatives. More speciﬁcally, suppose that a function f is deﬁned and con-

tinuous over the interval I. Let a ∈ I. Then

f(x) =

d

dx

x

a

f(t) dt

for all x ∈ I.

The major tool for calculating integrals, and the grand conclusion of our

discussion of antiderivatives is the Fundamental Theorem of Calculus.

Theorem 3.27 (Fundamental Theorem of Calculus). Suppose that f

is a continuous function over a closed interval [a, b] and that F is an an-

tiderivative of f. Then

b

a

f(x) dx = F(b) −F(a).

For example, F(x) = −cos x is an antiderivative of f(x) = sinx, so that

the Fundamental Theorem of Calculus tells us that

π

0

sinx dx = −cos(π) −(−cos(0)) = −(−1) −(−1) = 2.

3.8. THE FUNDAMENTAL THEOREM OF CALCULUS 125

As another example, note that F(x) = tan x is an anti-derivative of f(x) =

sec

2

x, so that the Fundamental Theorem of Calculus tells us that

π/4

0

sec

2

x dx = tan(π/4) −tan(0) = 1.

Remark 11 (Notational Convention). One commonly uses the nota-

tion

F(x)

b

a

= F(b) −F(a).

This is quite convenient. E.g., we write

sin x

π

0

= sinπ −sin0.

If there are ambiguities due to the length of the expression to which this

construction is applied, we also use the notation shown in the following

example:

x

3

−5x

2

+ 2x −8

5

3

= p(5) −p(3)

where p(x) = x

3

−5x

2

+ 2x −8.

Using this notation, we calculate that

3

−2

(x

2

−2x + 5) dx =

¸

x

3

3

−x

2

+ 5x

3

−2

=

95

3

.

Other examples are

π/4

0

sec xtan x dx = sec x

π/4

0

=

√

2 −1

and

π/3

π/4

csc xcot x dx = −csc x

π/3

π/4

=

−2

√

3

3

−(−

√

2) =

√

2 −

2

√

3

3

.

The reader is invited to practice a few examples.

126 CHAPTER 3. INTEGRATION

Exercise 47. Evaluate the following deﬁnite integrals:

(a)

1

0

(3x + 2) dx

(b)

2

1

6 −t

t

3

dt

(c)

5

2

2

√

x −1 dx

(d)

0

1

(t

3

−t

2

) dt

(e)

π/4

π/6

csc xcot x dx

(f)

−1

−1

7x

6

dx

(g)

π

0

1

2

cos x dx

(h)

π

0

cos(x/2) dx

(i)

2

−2

|x

2

−1| dx

(j)

π/2

0

cos

2

x dx

(k)

π/2

0

sin

2

(2x) dx

(l)

π/4

0

sec

2

x dx

3.8.1 Some Proofs

Because of their importance, we like to prove Theorem 3.26 and the Funda-

mental Theorem of Calculus.

Proof of the Fundamental Theorem of Calculus. Essentially, the desired re-

sult is an easy consequence of Theorem 3.26. Let F(x) be any anti-derivative

of f(x) on I, and H(x) =

x

a

f(t) dt the one provided by Theorem 3.26. In

particular, F

(x) = H

**(x) = f(x). Cauchy’s Theorem (see its application in
**

Corollary 2.5) tells us that F and H diﬀer by a constant. For some constant

c and all x ∈ I:

H(x) =

x

a

f(t) dt = F(x) +c (3.12)

We can ﬁnd out the value for c by substituting x = a in this equation. In

particular, we ﬁnd that

a

a

f(t) dt = 0 = F(a) +c or c = −F(a).

Using this calculation of c and substituting x = b in (3.12), we obtain

b

a

f(t) dt = F(b) −F(a),

as claimed.

3.8. THE FUNDAMENTAL THEOREM OF CALCULUS 127

Proof of Theorem 3.26. Because we assumed continuity of f on the interval

I, it follows from Theorem 3.10 that

F(x) =

x

a

f(t) dt

exists. So it is our task to show that F is diﬀerentiable at x, and that

F

**(x) = f(x). Using Theorem 1.26, after adjusting the notation to ﬁt the
**

current setting, the task becomes to show that

f(x) = lim

h→0

F(x +h) −F(x)

h

= lim

h→0

1

h

¸

x+h

a

f(t) dt −

x

a

f(t) dt

= lim

h→0

1

h

x+h

x

f(t) dt.

Here we assume that x is not an endpoint of I, so that x and x+h are both

in I. We omit (leave to the reader) the modiﬁcations of the proof which are

required in the case where x is an endpoint of I.

According to the Extreme Value Theorem (see Theorem 1.17) there are

points c and d between x and x +h, such that

f(c) = m(h) ≤ f(x) ≤ f(d) = M(h) (3.13)

for all t between x and x + h. The points c and d may not be uniquely

determined by h, but m and M are. It follows from (3.13) and Corollary 3.21

that

m(h) · h =

x+h

x

m(h) dt ≤

x+h

x

f(t) dt ≤

x+h

x

M(h) dt = M(h) · h,

and with this that

m(h) ≤

1

h

x+h

x

f(t) dt ≤ M(h).

Continuity of f(x) implies that

lim

h→0

m(h) = f(x) = lim

h→0

M(h).

It follows from a pinching argument (see Proposition 1.4) that

lim

h→0

x+h

x

f(t) dt = f(x),

and this is exactly what we needed to show.

128 CHAPTER 3. INTEGRATION

3.9 Substitution

In some cases it is not that easy to ‘see’ an antiderivative of the function one

likes to integrate. Substitution is a method which, when applied correctly,

will simplify the expression for the function you like to integrate. You hope

that you can ﬁnd an antiderivative for the simpliﬁed expression. The method

is based on the chain rule for diﬀerentiation. Sometimes this method is

helpful, other times it is not. Your success with this method depends greatly

on experience, i.e., practice.

We explain the method. Let F and g be functions which are deﬁned and

diﬀerentiable on an interval I. Set F

**= f. Then, according to the chain
**

rule,

d

dx

F(g(x)) = f(g(x))g

(x).

Assume that f and g

are continuous on I. Then f(g(x))g

(x) is continuous

as well. We may take antiderivatives of both sides of our previous equation,

and conclude that

f(g(x))g

(x) dx = F(g(x)) +c. (3.14)

The variable for the functions f and F is often called u, and this means in

context that u = g(x).

Let us give a few examples to illustrate how this method can be put

to use. There are no general rules what substitution must be used, rather

success justiﬁes the means. Working through the examples will teach you

how to apply this method in some typical situations. It will give you at least

some experience which you may then rely on in similar examples.

For example,

(2x −3)

3

dx =

1

2

(2x −3)

3

· 2dx =

1

8

(2x −3)

4

+c.

Here we used g(x) = 2x −3, g

(x) = 2, f(u) = u

3

, and F(u) =

u

4

4

.

There is a pattern, a way to use the notation, which can be applied to

write down the steps in an integration using substitution eﬃciently. Setting

u = g(x) we write

du = g

(x)dx,

3.9. SUBSTITUTION 129

instead of g

(x) = du/dx

3

. Suppose also that F is an anti-derivative of f,

so F

= f. Then the pattern for calculating an integral via substitution is

f(g(x))g

(x) dx =

f(u) du = F(u) +c = F(g(x)) +c. (3.15)

In the ﬁrst step of this calculation we carry out the substitution, in the

second one we ﬁnd the anti-derivative, and in the third one we reverse the

substitution. We make use of this notation in our next example.

For example, we calculate that

x

x

2

+ 2 dx =

1

2

x

2

+ 2 · 2xdx

=

1

2

√

u du

=

1

3

u

3/2

+c

=

1

3

(x

2

+ 2)

3/2

+c.

We used the substitution u = x

2

+ 2. Then

du

dx

= 2x, or du = 2xdx.

We calculate that

t

2

(t + 1)

7

dt =

(u −1)

2

u

7

du

=

(u

2

−2u + 1)u

7

du

=

(u

9

−2u

8

+u

7

) du

=

1

10

u

10

−

2

9

u

9

+

1

8

u

8

+c

=

1

10

(t + 1)

10

−

2

9

(t + 1)

9

+

1

8

(t + 1)

8

+c.

Here we used the substitution u = t + 1. Then du = dx and t = u −1.

We may have to use a substitution and a trigonometric identity to solve

3

We do not attach any particular meaning to the symbols dx and du in their own right.

The equation du = g

**(x)dx helps us to write down what happens when we perform the
**

substitution as in the ﬁrst equality in (3.15). Thought of as inﬁnitesimals or diﬀerentials,

these symbols have a meaning, but this is beyond the scope of these notes.

130 CHAPTER 3. INTEGRATION

an integration problem:

2xsin

2

(x

2

+ 5) dx =

sin

2

u du

=

1

2

[1 −cos 2u] du

=

1

2

¸

u −

1

2

sin2u

+c

=

1

2

¸

(x

2

+ 5) −

1

2

sin[2(x

2

+ 5)]

+c.

We used the substitution u = x

2

+ 5, so that du = 2xdx and the identity

sin

2

α = [1 −cos 2α]/2.

Find the substitution which we used in the following computation, and

check the details:

sec

2

xtan x dx =

**sec x · sec xtan x dx
**

=

u du

=

1

2

u

2

+c

=

1

2

sec

2

x +c.

Sometimes we have to apply the method of substitution twice, or more

often, to work out an integral. Here is an example.

(x

2

+ 1) sin

3

(x

3

+ 3x −2) cos(x

3

+ 3x −2) dx =

1

3

sin

3

ucos u du

=

1

3

v

3

dv

=

1

12

v

4

+c

=

1

12

sin

4

u +c

=

sin

4

(x

3

+ 3x −2)

12

+c

In the computation we used the substitution u = x

3

+ 3x − 2. Then du =

3(x

2

+1) dx. In a second substitution we set v = sin u. Then dv = cos u du.

3.9. SUBSTITUTION 131

Here are two examples, which are important in the context of integrating

rational functions. In the ﬁrst example we assume that a = 0, and we use

the subsitution x = au. The dx = adu.

dx

x

2

+a

2

dx =

adu

a

2

u

2

+a

2

=

1

a

du

u

2

+ 1

=

1

a

arctan(u) +c

=

1

a

arctan

x

a

+c

Adding another idea, we calculate

dx

x

2

+ 2x + 5

=

dx

(x + 1)

2

+ 4

=

du

u

2

+ 4

=

1

2

arctan

x + 1

2

+c.

We used the substitution u = x+1, and then we proceeded as in the previous

example.

3.9.1 Substitution and Deﬁnite Integrals

Let us now explore how substitution is used to calculate deﬁnite integrals.

Assuming as before that f and g

**are continuous on the interval [a, b], we
**

have

b

a

f(g(x))g

(x) dx =

g(b)

g(a)

f(u) du. (3.16)

To see this, observe that f has an anti-derivative, which we again denote

by F. Then

b

a

f(g(x))g

(x) dx = F(g(x))

b

a

= F(u)

g(b)

g(a)

=

g(b)

g(a)

f(u) du.

The ﬁrst identity is obtained as a combination of the Fundamental Theorem

of Calculus and (3.14). The second one is obvious, and the third one is

another application of the Fundamental Theorem of Calculus.

Let us apply this formula in a few examples.

1

0

(x

2

−1)(x

3

−3x + 5)

3

dx =

1

3

3

5

u

3

du =

1

12

u

4

3

5

= −

136

3

.

132 CHAPTER 3. INTEGRATION

We used the substitution u = x

3

− 3x + 5. Then du = (3x

2

− 3) dx, and

1

3

du = (x

2

−1) dx. To obtain the limits for the integral we calculate u(0) = 5

and u(1) = 3.

Another example is

π/4

0

cos

2

xsin x dx = −

√

2/2

1

u

2

du = −

1

3

u

3

√

2/2

1

= −

1

3

¸

1 −

√

2

4

¸

.

We use the substitution u = cos x. Then −du = sinx dx. If x = 0, then

u = 1, and if x = π/4, then u =

√

2/2.

Incorporating one of our previous techniques, we calculate

2

0

x(x + 1)

6

dx =

3

1

(u −1)u

6

du =

3

1

u

7

−u

6

du =

3554

7

.

We use the substitution u = x + 1. Then du = dx and x = u − 1. If x = 0,

then u = 1, and if x = 2, the u = 3.

Similarly,

√

8

0

x

3

x

2

+ 1 dx =

1

2

9

1

(u −1)

√

u du =

1

2

9

1

u

3/2

−u

1/2

du =

596

15

.

We use the substitution u = x

2

+ 1. Then

1

2

du = x dx and x

2

= u −1. For

the limits we calculate, if x = 0, then u = 1, and if x =

√

8, then u = 9.

Finally,

1

0

1 −x

2

dx =

π/2

0

1 −sin

2

ucos u du =

π/2

0

cos

2

u du =

π

4

.

We use the substitution x = sin u. Then dx = cos u du. If x = 0, then

u = 0, and if x = 1, then u = π/2. For our given values of x, there are other

possible values for u, but they will lead to the same results.

Remark 12. The graph of f(x) =

√

1 −x

2

is the northern part of a circle.

Using x ∈ [0, 1] means that we calculated the area under this graph in the

ﬁrst quadrant, i.e., the area of one forth of the disk of radius 1. You were

told long time ago in school, that the area of this unit disk is π, so that the

result of the calculation is hardly surprising.

There is a more serious matter. Is the example genuine, or did we assume

the answer previously? By deﬁnition, π is the ratio of the circumference of

a circle by its diameter. In our calculation of the derivative of the sine and

cosine functions we used the estimate that | sin h − h| ≤ h

2

/2. When we

3.10. AREAS BETWEEN GRAPHS 133

showed this, we used that |h| ≤ | tan h| for h ∈ [−π/4, π/4]. A typical proof

of the latter inequality starts out by ﬁrst showing that the area of the unit

disk is π. This means, we assumed the result in the example, we did not

derive it.

Exercise 48. Find the following integrals:

(a)

dx

√

2x + 1

(b)

t

(4t

2

+ 9)

2

dt

(c)

t(1 +t

2

)

3

dt

(d)

2s

3

√

6 −5s

2

ds

(e)

b

3

x

3

√

1 −a

4

x

4

dx

(f)

π

0

xcos x

2

dx

(g)

x

2

√

x + 1 dx

(h)

x + 3

√

x + 1

dx

(i)

sin

2

(3x) dx

(j)

π/2

0

cos

2

x dx

(k)

π/4

π/6

sec(2x) tan(2x) dx

(l)

1/2

0

dx

4 +x

2

(m)

sec

2

x

√

1 + tan x

dx

(n)

√

1 + sinxcos x dx

(o)

r

0

r

2

−x

2

dx

3.10 Areas between Graphs

Previously we related the integral to areas of a region under a graph. This

idea can be generalized to the discussion of areas of regions between two

graphs. Let us look at an example.

Example 3.28. Calculate the area of the region between the graphs of the

functions f(x) = x

2

and g(x) =

√

1 −x

2

.

Solution To get a better understanding, we draw the two graphs, see

Figure 3.13. Now you see the region between the two graphs whose area we

want to calculate. We call the region Ω.

The graphs intersect in two points. To ﬁnd their x-coordinates, we solve

the equation

f(x) = x

2

= g(x) =

1 −x

2

.

After squaring the equation and solving it for x

2

, we ﬁnd x

2

=

−1±

√

5

2

.

Only the + sign occurs as x

2

≥ 0. Taking the square root, we ﬁnd the

x-coordinates of the points where the curves intersect:

A = −

−1 +

√

5

2

and B =

−1 +

√

5

2

.

134 CHAPTER 3. INTEGRATION

-1 -0.5 0.5 1

0.2

0.4

0.6

0.8

1

Figure 3.13: Region between two

graphs

0.5 1 1.5 2 2.5 3

-1

-0.5

0.5

1

Figure 3.14: Region between two

graphs

To get the area of the region under the graph of f(x) and g(x) over the

interval [A, B] we can calculate the appropriate integrals. To get the area of

the region Ω between the graphs, we take the area of the region under the

graph of g(x) and subtract the area of the region under the graph of f(x).

Concretely:

Area(Ω) =

B

A

g(x) dx −

B

A

f(x) dx =

B

A

(g(x) −f(x)) dx ≈ 1.06651.

The numerical value was obtained by computer. You are invited to work

out the integral with the help of the Fundamental Theorem of Calculus to

verify the result. ♦

Some problems are a bit more subtle.

Example 3.29. Find the area of the region between the graphs of the func-

tions f(x) = cos x and g(x) = sin x for x between 0 and π.

Solution: The region Ω between the graphs is shown in Figure 3.14.

The region breaks up into two pieces, the region Ω

1

over the interval [0, π/4]

on which f(x) ≥ g(x), and the region Ω

2

over the interval [π/4, π] where

g(x) ≥ f(x). We calculate the areas of the regions Ω

1

and Ω

2

separately.

3.10. AREAS BETWEEN GRAPHS 135

In each case, we proceed as in the previous example:

Area(Ω

1

) =

π/4

0

(cos x −sin x) dx = (sin x + cos x)

π/4

0

=

√

2 −1

Area(Ω

2

) =

π

π/4

(sin x −cos x) dx = −(sin x + cos x)

π

π/4

= 1 +

√

2.

In summary we ﬁnd:

Area(Ω) = Area(Ω

1

) + Area(Ω

2

) = 2

√

2.

An additional remark may be in place. When we compared integrals and

areas, we had to take into account where the function is non-negative, resp.,

non-positive. Here we did not. We took care of this aspect by breaking up

the interval into the part where f(x) ≥ g(x) and the part where g(x) ≥ f(x).

♦

Our general deﬁnition for the area between two graphs is as follows.

Deﬁnition 3.30. Suppose f(x) and g(x) are integrable functions over an

interval [a, b]. Let Ω be the region between the graphs of f(x) and g(x) for

x between a and b. The area of Ω is

Area(Ω) =

b

a

|f(x) −g(x)| dx.

This deﬁnition generalizes Deﬁnition 3.7 on page 114. The deﬁnition is

also consistent with the intuitive idea of the area of a region, and it incor-

porates and generalizes Proposition 3.23 on page 121. Taking the absolute

value of the diﬀerence of f(x) and g(x) allows us avoid the question where

f(x) ≥ g(x) and where g(x) ≥ f(x). Typically this problem gets addressed

when the integral is calculated. In some problems a and b are explicitly

given, in others you have to determine them from context. In all cases it is

good to graph the functions before calculating the area of the region between

them. Having the correct picture in mind helps you to avoid mistakes.

Exercise 49. Sketch and ﬁnd the area of the region bounded by the curves:

(a) y = x

2

and y = x

3

.

(b) y = 8 −x

2

and y = x

2

(c) y = x

2

and y = 3x + 5.

(d) y = sin x and y = πx −x

2

.

(e) y = sin x and y = 2 sin xcos x for x between 0 and π.

136 CHAPTER 3. INTEGRATION

3.11 Numerical Integration

The Fundamental Theorem of Calculus provided us with a highly eﬃcient

method for calculating deﬁnite integrals. Still, for some functions we have

no good expression for its anti-derivative. In such cases we may have to

rely on numerical methods for integrating. Let us take such a function, and

show some methods for ﬁnding an approximate value for the integral.

We describe diﬀerent ways to ﬁnd, by numerical means, approximate

values for the integral of a function f(x) over the interval [a, b]:

b

a

f(x) dx.

In all of the diﬀerent approaches we partition the interval into smaller ones:

a = x

0

< x

1

< · · · < x

n−1

< x

n

= b.

Left and Right Endpoint Method: In the left endpoint method we

ﬁnd the value of the function at each left endpoint of the intervals of the

partition. We multiply it with the length of the associated interval, and

then add up the terms. Explicitly, we calculate

I

L

= f(x

0

)(x

1

−x

0

) +f(x

1

)(x

2

−x

1

) +· · · +f(x

n−1

)(x

n

−x

n−1

). (3.17)

In the right endpoint method we proceed as we did on the left endpoint

method, only we use the value of the function at the right endpoint instead

of the left endpoint:

I

R

= f(x

1

)(x

1

−x

0

) +f(x

2

)(x

2

−x

1

) +· · · +f(x

n

)(x

n

−x

n−1

). (3.18)

Both expressions provide us with speciﬁc examples of Riemann sums.

Example 3.31. Use the left and right endpoint method to ﬁnd approximate

values for

2

0

e

−x

2

dx.

Solution: Set f(x) = e

−x

2

and choose the partition:

x

0

= 0 < x

1

=

1

2

< x

2

= 1 < x

3

=

3

2

< x

4

= 2.

3.11. NUMERICAL INTEGRATION 137

Then x

k

−x

k−1

= 1/2 for k = 1, 2, 3, and 4. Formula (3.17) for I

L

specializes

to

I

L

=

f(0) +f(1/2) +f(1) +f(3/2)

2

≈ 1.126039724.

Formula (3.18) for I

R

specializes to

I

R

=

f(1/2) +f(1) +f(3/2) +f(2)

2

≈ .6351975438.

0.5 1 1.5 2

0.2

0.4

0.6

0.8

1

Figure 3.15: Use left end points

0.5 1 1.5 2

0.2

0.4

0.6

0.8

1

Figure 3.16: Use right end points

Apparently, I

L

and I

R

are calculated by combining the areas of certain

rectangles. In our case the values of f(x) are all positive and all of the

rectangles are above the x axis, so the areas of the rectangles are all added.

Note also, that our speciﬁc function f(x) is decreasing on the interval [0, 2],

so that I

L

is an upper sum for the function f(x) over the interval [0, 2], and

I

R

is a lower sum. In this sense, we have

I

R

= .6351975438 ≤

2

0

e

−x

2

dx ≤ I

L

= 1.126039724.

The function and the rectangles whose areas are added to give us I

L

and I

R

are shown in Figure 3.15 and Figure 3.16. ♦

Midpoint and Trapezoid Method: We may try and improve on the

endpoint methods. In the midpoint methods, we use the value of the function

at the midpoints of the intervals of the partition. That should be less bias.

138 CHAPTER 3. INTEGRATION

We use the same partition and notation as above. Then the formula for the

midpoint method is:

I

M

= f

x

0

+x

1

2

(x

1

−x

0

) +· · · +f

x

n

+x

n−1

2

(x

n

−x

n−1

). (3.19)

In the trapezoid method we do not take the function at the average (i.e.

midpoint) of the end points of the intervals in the partition, but we average

the values of the function at the end points. Speciﬁcally, the formula is

I

T

=

f(x

0

) +f(x

1

)

2

(x

1

−x

0

) +· · · +

f(x

n−1

) +f(x

n

)

2

(x

n

−x

n−1

). (3.20)

It is quite easy to see that

I

T

=

I

L

+I

R

2

. (3.21)

Let us explain the reference to the word trapezoid. For simplicity, sup-

pose that f(x) is non-negative on the interval [a, b]. Consider the trapezoid

of width (x

1

− x

0

) which has height f(x

0

) at its left and f(x

1

) at its right

edge. The area of this trapezoid is

f(x

0

)+f(x

1

)

2

(x

1

− x

0

). This is the ﬁrst

summand in the formula for I

T

, see (3.20). We have such a trapezoid over

each of the intervals in the partition, and their areas are added to give I

T

.

Expressed diﬀerently, we can draw a secant line through the points

(x

0

, f(x

0

)) and (x

1

, f(x

1

)). This gives us the graph of a function T(x)

over the interval [x

0

, x

1

]. Over the interval [x

1

, x

2

] the graph of T(x) is the

secant line through the points (x

1

, f(x

1

)) and (x

2

, f(x

2

)). Proceeding in

the fashion, we use appropriate secant lines above all of the intervals in the

partition to deﬁne the function T(x) over the entire interval [a, b]. Then

I

T

=

b

a

T(x) dx.

This integral is easily computed by the formula in (3.20).

Example 3.32. Use the midpoint and trapezoid method to ﬁnd approxi-

mate values for

2

0

e

−x

2

dx.

Solution: We use the same partition of [0, 2] as in Example 3.31. The

formula for I

M

(see (3.19)) specializes to

I

M

=

f(.25) +f(.75) +f(1.25) +f(1.75)

2

≈ .8827889485.

3.11. NUMERICAL INTEGRATION 139

As for the endpoint methods, I

M

is the combined area of certain rectangles.

Their heights are the values f(x

i

) at the midpoints of the intervals of the

partition. Their width are the lengths of the intervals of the partition. You

see the rectangles for this calculation in Figure 3.17.

0.5 1 1.5 2

0.2

0.4

0.6

0.8

1

Figure 3.17: Use midpoints

0.5 1 1.5 2

0.2

0.4

0.6

0.8

1

Figure 3.18: Trapezoid Method

Based on our previous calculations and Formula (3.21) we ﬁnd

I

T

=

I

L

+I

R

2

≈ .8806186341.

We illustrated this calculation in Figure 3.18. There you see the function

f(x) = e

−x

2

and ﬁve dots on the graph. The dots are connected by straight

line segments. These line segments form the graph of a function T(x), and

I

T

is the area of the region under this graph. So

I

T

=

2

0

T(x) dx. ♦

Simpson’s Method: In Simpson’s method we combine the endpoint

and midpoint methods in a weighted fashion. Again, we use the same no-

tation for the function and the partition as above. The speciﬁc formula for

an approximate value of the integral of f(x) over [a, b] is

I

S

=

1

6

¸

f(x

0

) + 4f

x

0

+x

1

2

+f(x

1

)

(x

1

−x

0

) +· · ·

+

1

6

¸

f(x

n−1

) + 4f

x

n−1

+x

n

2

+f(x

n

)

(x

n

−x

n−1

)

(3.22)

140 CHAPTER 3. INTEGRATION

It is quite easy to see that

I

S

=

I

L

+ 4I

M

+I

R

6

=

I

T

+ 2I

M

3

.

Let us explain the background to Simpson’s method. We deﬁne a func-

tion P(x) over the interval [a, b] by deﬁning a degree 2 polynomial on each

of the intervals of the partition. The polynomial over the interval [x

k−1

, x

k

]

is chosen so that it agrees with f(x) at the end points and at the midpoint

of this interval. Simpson’s method is a reﬁnement of the Trapezoid method.

In one method we use two points on the graph and connect them by a

straight line segment. In the other one we use three points on the graph and

construct a parabola through them. With some work one can show that

I

S

=

b

a

P(x) dx.

0.5 1 1.5 2

0.2

0.4

0.6

0.8

1

Figure 3.19: Simpson’s Method

Example 3.33. Use Simpson’s method to ﬁnd an approximate value for

2

0

e

−x

2

dx.

3.11. NUMERICAL INTEGRATION 141

Solution: We use the same partition of [0, 2] as in Example 3.31. The

formula for I

S

(see the special case of (3.22)) specializes to

I

S

=

I

L

+ 4I

M

+I

R

6

≈ .88206555104,

where I

L

, I

M

and I

R

are as above.

You see the method illustrated in Figure 3.19. There you see the graphs

of two functions, the function f(x) = e

−x

2

and the function P(x) from the

discussion of Simpson’s method. Only the thickness of the line suggests that

there are two graphs of almost identical functions. ♦

Example 3.34. Compare the accuracy of the various approximate values

of

2

0

e

−x

2

dx.

Solution: We compare the approximate values for the integral obtained

by the diﬀerent formulas. We partition the interval [0, 2] into n intervals

of the same length, and vary n. We tabulate the results. They should be

compared with an approximate value for the integral of

0.882081390762421.

n = 1 n = 10 n = 100 n = 1000

I

L

2.0000000 0.9800072469 0.891895792451 0.883063050702697

I

R

0.0366313 0.7836703747 0.872262105229 0.881095681980474

I

M

0.7357589 0.8822020700 0.882082611663 0.882081402972833

I

T

1.0183156 0.8818388108 0.882078948840 0.882081366341586

I

S

0.8299445 0.8820809836 0.882081390722 0.882081390762417

Table 3.1: Approximate Values of the Integral

Simpson’s method is more accurate than the other ones. E.g., Simpson’s

method with n = 4 gives a result which is better than the left and right

endpoint method with n = 1000. Even if you use the midpoint and trapezoid

method with n = 1000, then the result is far less accurate that Simpson’s

method with n = 100. ♦

142 CHAPTER 3. INTEGRATION

Remark 13. It is important that we keep the number n of intervals into

which we partition [a, b] small. It does not only keep the number of overall

computations small. In each computational step we expect to make a round-

oﬀ error, and these may add up. The fewer computations we make, the

smaller the cummulative round-oﬀ error will be.

Exercise 50. Proceed as in Example 3.34 and compare the diﬀerent meth-

ods applied to the calculation of

π/2

0

sin x dx = 1.

3.12 Applications of the Integral

In Deﬁnition 3.7 and Proposition 3.23 we related deﬁnite integrals to areas.

Based on the context, this can have a more concrete meaning. Consider a

function f(t) on an interval [a, b] and the integral

I =

b

a

f(t) dt.

If f(t) stands for the rate at which a drug is absorbed, then I is the total

amount of the drug which has been absorbed in the time interval [a, b]. If

f(t) stands for the speed with which you travel, then I stands for the total

distance which you traveled during the time interval [a, b]. You are invited

to come up with more interpretations. In addition, the following deﬁnition

expresses the common notion of the average value of a function.

Deﬁnition 3.35. Suppose that f(t) is an integrable function over the in-

terval [a, b]. Then the quantity

f

av

:=

1

b −a

b

a

f(t) dt

is called the average value of f(t) over the interval [a, b].

For example, the average value of the sine function f(x) = sinx over the

interval [0, π] is 2/π.

Let us explore the diﬀerent aspects of integration in an example.

Example 3.36. The river Little Brook ﬂows into a reservoir, referred to

as Beaver Pond by the locals. The amount of water carried by the river

depends on the season. As a function of time, it is

g(t) = 2 + sin

πt

180

.

3.12. APPLICATIONS OF THE INTEGRAL 143

We measure time in days, and t = 0 corresponds to New Year. The units

of g(t) are millions of liter of water per day. Water is released from Beaver

Pond at a constant rate of 2 million liters per day. At the beginning of the

year, there are 200 million liters of water in the reservoir.

(a) How many liter of water are in Beaver Pond by the end of April?

(b) Suppose F(t) tells how much water there is in the reservoir on day t

of the year. Find F(t).

(c) At which rate does the amount of water in the reservoir change at the

beginning of September?

(d) On which days will there be 250 million liters of water in Beaver Pond?

(e) At which amount of water will the reservoir crest?

(f) On the average, by how much has the amount of water in Beaver Pond

increased per day during the ﬁrst three months of the year?

Solution: Water enters and leaves the pond. The net rate entering is

f(t) = g(t) −2 = sin

πt

180

**millions of liters per day.
**

We obtain the total change of the amount of water in the reservoir by inte-

grating f(t). Set

A(T) =

T

0

f(t) dt.

On the T-th day of the year, the total amount of water in Beaver Pond is

F(T) = 200 +

T

0

f(t) dt = 200 +

180

π

¸

1 −cos

πT

180

millions of liters.

This answers (b). By the end of April, after 120 days, there are

F(120) = 200 +

180

π

¸

1 −cos

2π

3

≈ 238.2

millions of liters of water in the pond. This answers (a).

The rate at which the amount of water in the pond changes is F

(t) =

f(t). At the beginning of September, after 240 days, the rate of change is

144 CHAPTER 3. INTEGRATION

f(240) ≈ −.866. The pond is losing water at a rate of 866, 000 liters per

day.

To answer (d), we like to know for which T we have F(T) = 250. We

solve the equation for T:

250 = 200 +

180

π

¸

1 −cos

πT

180

or cos

πT

180

= 1 −

5π

18

.

We apply the function arccos to both sides of the last equation and ﬁnd

T =

180

π

arccos

1 −

5π

18

≈ 88, or 272.

On the 88-th and 272-nd day of the year there will be 250 millions of liters

of water in the reservoir.

To ﬁnd at which amount the reservoir crests, we have to ﬁnd the max-

imum value of F(t). This occurs apparently when cos(πt/180) = −1 or

t = 180. The pond crests at mid-year, and then the amount of water in it

is about 314.6 millions of liters of water. This answers (e).

After three months or 90 days there are about 257.3 millions of liters

of water in Beaver Pond. Within this time, the amount of water has in-

creased by 57.3 millions of liters. On the average, the amount of water in

the reservoir increased by about 640,000 liters per day. ♦

Exercise 51. A pain reliever has been formulated such that it is absorbed

at a rate of 600 sin(πt) (mg/hr) by the body. Here t measures time in hours,

t = 0 at the time you take the medication, and the absorption process is

complete at time t = 1.

(a) What is the total amount of the drug which is absorbed?

(b) Find a function F(t), such that F(t) tells how much medication has

been absorbed at time t.

(c) A total of 150 mg of the medication has to be absorbed before the

drug is eﬀective. How long does it take until this threshold is reached?

3.13 The Exponential and Logarithm Functions

In Section 1.10 we introduced the exponential function exp(x) = e

x

and the

natural logarithm function ln x. At the time we only stated that they exist

because we did not have the tools to properly deﬁne them. We will now ﬁll

in the details. Many of the routine calculations are formulated as exercises.

3.13. THE EXPONENTIAL AND LOGARITHM FUNCTIONS 145

Deﬁnition 3.37. Let x ∈ (0, ∞). The natural logarithm of x is deﬁned as

ln x =

x

1

dt

t

. (3.23)

Theorem 3.38. The natural logarithm function is diﬀerentiable on its en-

tire domain (0, ∞), its derivative is

ln

x =

1

x

,

and ln x is increasing on (0, ∞).

Proof. The function 1/x is deﬁned and continuous on (0, ∞). According

to Theorem 3.10 this means that lnx is deﬁned for all x in (0, ∞). Theo-

rem 3.26 tells us that ln

**x = 1/x. According to Theorem 2.11, the function
**

is increasing because its derivative ln

x > 0 for all x > 0.

Let us also verify one of the central equations for calculating with loga-

rithms, the third rule in Theorem 1.34.

Proposition 3.39. For any x, y > 0,

ln(xy) = ln x + ln y. (3.24)

Proof. We need a short calculation. Here x and y are ﬁxed positive numbers.

We use the substitution u =

t

x

, so that du =

1

x

dt. For the adjustment of

the limits of integration, observe that t/x = u = 1 when t = x, and that

t/x = u = y when t = xy. Then

xy

x

dt

t

=

xy

x

1

(t/x)

1

x

dt =

y

1

du

u

= ln y.

Using this calculation we deduce that

ln(xy) =

xy

1

dt

t

=

x

1

dt

t

+

xy

x

dt

t

= ln x + ln y.

This is exactly our claim.

Exercise 52. Show:

(1) ln 1 = 0.

(2) ln(1/y) = −ln y for all y > 0.

146 CHAPTER 3. INTEGRATION

(3) ln(x/y) = ln x −ln y for all x, y > 0.

Exercise 53. Show that ln 4 > 1. Hint: Using the partition

1 = x

0

< 2 = x

1

< 3 = x

2

< 4 = x

3

,

ﬁnd a lower sum S

l

for the function 1/t over the interval [1, 4] so that S

l

> 1.

We can now deﬁne the Euler number:

Deﬁnition 3.40. The number Euler number e is the unique number such

that

ln e = 1 or, equivalently,

e

1

dt

t

= 1.

For this deﬁnition to make sense, we have to show that there is a number

e which has the property used in the deﬁnition. To see this, observe that

ln 1 = 0 < 1 < ln 4. Because ln x is diﬀerentiable, it follows from the

Intermediate Value Theorem (see Theorem 1.16) that there is a number e

for which ln e = 1. It also follows that 1 < e < 4.

Proposition 3.41. For every real number x there exists exactly one positive

number y, such that

ln y = x (3.25)

Proof. Observe that ln(e

n

) = n and ln(1/e

n

) = −n for all natural num-

bers n. So all integers (whole numbers) are values of the natural logarithm

function. Every real number x lies between two integers. According to the

Intermediate Value Theorem, every real number is a value of the function

ln y. We saw that ln y is an increasing function. This means that, for any

given x, the equation ln y = x has at most one solution. Taken together it

means that it has a unique solution.

Exercise 54. Show that

ln(a

r

) = r ln a

for all positive numbers a and all rational numbers r, i.e., numbers of the

form r = p/q where p and q are integers and q = 0.

In summary, we have seen that

Corollary 3.42. The natural logarithm function ln x is a diﬀerentiable, in-

creasing function with domain (0, ∞) and range (−∞, ∞), and ln

x = 1/x.

3.13. THE EXPONENTIAL AND LOGARITHM FUNCTIONS 147

We are now ready to deﬁne the exponential function.

Deﬁnition 3.43. Given any real number x, we deﬁne exp(x) to be the

unique number for which

ln(exp(x)) = x, (3.26)

i.e., y = exp(x) is the unique solution of the equation ln(y) = x. This

assignment (mapping x to exp(x)) deﬁnes a function, called the exponential

function, with domain (−∞, ∞) and range (0, ∞).

Exercise 55. Show that the exponential function exp and the natural log-

arithm function ln are inverses of each other. In addition to the equation in

(3.26), you need to show that

exp(ln(y)) = y (3.27)

for all y ∈ (0, ∞).

Summarizing this discussion, and adding some observations which we

have made elsewhere, we have:

Proposition 3.44. The exponential function exp(x) is a diﬀerentiable, in-

creasing function with domain (−∞, ∞) and range (0, ∞), and the exponen-

tial function is its own derivative, i.e., exp

(x) = exp(x).

Exercise 56. Show for all real numbers x and y that:

(1) exp(0) = 1

(2) exp(1) = e

(3) exp(x) exp(y) = exp(x +y)

(4) 1/ exp(y) = exp(−y)

(5) exp(x)/ exp(y) = exp(x −y).

Hint: Use the results of Exercise 52, the deﬁnition of e in Deﬁnition 3.40,

and that the exponential and logarithm functions are inverses of each other.

Exercise 57. Show that exp(r) = e

r

for all rational numbers r. Hint: Use

Exercise 54 and that the exponential and logarithm functions are inverses

of each other.

148 CHAPTER 3. INTEGRATION

The expression e

r

makes sense only if r is a rational number. If r = p/q

then we raise e to the r-th power and take the q-th root of the result. For

an arbitrary real number we set

e

x

= exp(x). (3.28)

This is consistent with the meaning of the expression for rational exponents

due to Exercise 57, and it deﬁnes what we mean by raising e to any real

power.

3.13.1 Other Bases

So far we discussed the natural logarithm function and the exponential func-

tion with base e. We now expand the discussion to other bases.

Deﬁnition 3.45. Let a be a positive number, a = 1. Set

log

a

x =

ln x

ln a

and exp

a

(x) = exp(xln a). (3.29)

We call log

a

(x) the logarithm function with base a and exp

a

(x) the expo-

nential function with base a. For the function log

a

we use the domain (0, ∞)

and range (−∞, ∞). For the exponential function exp

a

we use the domain

(−∞, ∞) and range (0, ∞).

Exercise 58. Show

(1) ln a > 0 if a > 1 and lna < 0 if 0 < a < 1.

(2) log

a

(x) and exp

a

(x) are diﬀerentiable functions.

(3) log

a

(x) and exp

a

(x) are increasing functions if a > 1.

(4) log

a

(x) and exp

a

(x) are decreasing functions if 0 < a < 1.

Exercise 59. Suppose a > 0 and a = 1. Show that

(a) exp

a

(log

a

(y)) = y for all y > 0.

(b) log

a

(exp

a

(x)) = x for all real numbers x.

Taken together, the speciﬁcations for the domains and ranges for the

functions exp

a

and log

a

and the results from Exercise 59 tell us that

Corollary 3.46. Suppose a > 0 and a = 1. The functions exp

a

and log

a

are inverses of each other.

3.13. THE EXPONENTIAL AND LOGARITHM FUNCTIONS 149

Exercise 60. Suppose a > 0 and a = 1. Show the laws of logarithms:

(a) log

a

1 = 0 and log

a

a = 1.

(b) log

a

(xy) = log

a

x + log

a

y for all x, y > 0.

(c) log

a

(1/y) = −log

a

y for all y > 0.

(d) log

a

(x/y) = log

a

x −log

a

y for all x, y > 0.

Exercise 61. Suppose a > 0 and a = 1. Show the exponential laws:

(1) exp

a

(0) = 1 and exp

a

(1) = a

(2) exp

a

(x) exp

a

(y) = exp

a

(x +y)

(3) 1/ exp

a

(y) = exp

a

(−y)

(4) exp

a

(x)/ exp

a

(y) = exp

a

(x −y).

Exercise 62. Suppose a > 0, a = 1, and r is a rational number. Show

log

a

(a

r

) = r and exp

a

(r) = a

r

.

We rephrase a convention which we made previously for e. Suppose

a > 0 and a = 1. The expression a

r

makes sense if r is a rational number.

If r = p/q then we raise a to the r-th power and take the q-th root of the

result. For an arbitrary real number we set

a

x

= exp

a

(x). (3.30)

This is consistent with the meaning of the expression for rational exponents

due to Exercise 62, and it deﬁnes what we mean by raising a to any real

power. Equation 3.30 specializes to the one in Equation 3.28 if we set a = e.

It is also a standard convention to set

1

x

= 1 and 0

x

= 0

for any real number x. Typically 0

0

is set 1.

We can now state an equation which is typically considered to be one of

the laws of logarithms:

Exercise 63. Suppose a > 0, a = 1, x > 0, and z is any real number. Then

log

a

(x

z

) = z log

a

(x).

150 CHAPTER 3. INTEGRATION

We are now ready to ﬁll in the details for one of the major statements

which we made in Section 1.10. We are ready to prove

Theorem 3.47. Let a be a positive number, a = 1. There exists exactly one

monotonic function, called the exponential function with base a and denoted

by exp

a

(x), which is deﬁned for all real numbers x such that exp

a

(x) = a

x

whenever x is a rational number.

Proof. In this section we constructed the function exp

a

(x), and this function

has all of the properties called for in the theorem. That settles the existence

statement. We have to show the uniqueness statement, i.e., there in only

one such function.

Suppose f(x) is any monotonic function and f(r) = a

r

= exp

a

(r) for

all rational numbers r. We have to show that f(x) = exp

a

(x) for all real

numbers x. We leave the veriﬁcation of this assertion to the reader. Here one

uses that f(x) and exp

a

(x) are monotonic, and that exp

a

(x) is continuous.

Chapter 4

Trigonometric Functions

In this section we discuss the radian measure of angles and introduce the

trigonometric functions. These are the functions sine, cosine, tangent, et. al.

We collect some formulas relating these functions.

Arc Length and Radian Measure of Angles: Consider the unit

circle (a circle with radius 1) centered at the origin in the Cartesian plane.

It is shown in Figure 4.1. We take a practical approach to measuring the

length of an arc on this circle. We imagine that we can straighten it out,

and measure how long it is. It requires some work to introduce the idea of

the length of a curve in a mathematically rigorous fashion.

Deﬁnition 4.1. The number π is the ratio between the circumference of a

circle and its diameter.

This deﬁnition goes back to the Greeks. Stated diﬀerently it says, that

the circumference of a circle of radius r is 2πr. Observe that the ratio

referred to in the deﬁnition does not depend on the radius of the circle.

Consider an angle α between the positive x-axis and a ray which origi-

nates at the origin of the coordinate system and intersects the unit circle in

the point p. We like to ﬁnd the radian measure of the angle α. Consider an

arc on the unit circle which starts out at the point (1, 0) and ends at p, and

suppose its length is s. Then

α = ±s (radians). (4.1)

The + sign is used if the arc goes counter clockwise around the circle. The

− sign is used if it proceeds clockwise. We may also consider arcs which

wrap around the circle several times before they end at p. In this sense, the

radian measure of the angle α is not unique, but any two radian measures

of the angle diﬀer by an integer multiple of 2π.

151

152 CHAPTER 4. TRIGONOMETRIC FUNCTIONS

-1 -0.5 0.5 1

-1

-0.5

0.5

1

(cos t, sin t)

Figure 4.1: The unit circle

Conversely, let t be any real number. We construct the angle with radian

measure t. Starting at the point (1, 0) we travel the distance |t| along the

unit circle (here |t| denotes the absolute value of t). By convention, we travel

counter clockwise if t is positive and clockwise if t is negative. In this way

we reach a point p on the circle. Let α be the angle between the positive

x-axis and the ray which starts at the origin and intersects the unit circle

in p. This angle has radian measure t.

Comparison of Angles in Degrees and Radians: We suppose that

you are familiar with measuring angles in degrees. The measure of half a

revolution (a straight angle) comprises π radians and 180 degrees. So, one

degree corresponds to π/180 ≈ 0.017453293 radians, and one radian corre-

sponds to 180/π ≈ 57.29577951 degrees. We have the conversion formula

x degrees =

π

180

x radians. (4.2)

Trigonometric Functions: Let t be once more a real number. Starting

at the point (1, 0) we travel the distance |t| along the unit circle, counter

clockwise if t is positive and clockwise if t is negative. In this way we reach

a point p = (x(t), y(t)) on the circle, and we set

x(t) = cos t and y(t) = sin t. (4.3)

153

This deﬁnes the functions sint and cos t. You see the construction im-

plemented in Figure 4.1. You can ﬁnd the graphs of the sine and cosine

functions on the interval [0, 2π] in Figures 4.2 and 4.3.

1 2 3 4 5 6

-1

-0.5

0.5

1

Figure 4.2: f(x) = sin x

1 2 3 4 5 6

-1

-0.5

0.5

1

Figure 4.3: f(x) = cos x

The other trigonometric functions, tangent (tan), cotangent (cot), secant

(sec), and cosecant (csc) are deﬁned as follows:

tan x =

sin x

cos x

cot x =

cos x

sin x

sec x =

1

cos x

csc x =

1

sin x

(4.4)

To make sure you have some idea about the behavior of the tangent

and cotangent function we provided two graphs for each of them. They

are drawn over diﬀerent parts of the domain to show diﬀerent aspects. See

Figure 4.4 to Figure 4.7. You can see the graphs of the secant and cosecant

functions in Figure 4.8 and 4.9.

A small table with angles given in degrees and radians, as well as the

associated values for the trigonometric functions is given in Table 4.1. If the

functions are not deﬁned at some point, then this is indicated by ‘n/a’. Older

calculus books may still contain tables with the values of the trigonometric

functions, and there are books which were published for the speciﬁc purpose

of providing these tables. This is really not necessary anymore because any

scientiﬁc calculator gives those values to you with rather good accuracy.

Trigonometric Functions deﬁned at a right triangle: Occasionally

it is more convenient to use a right triangle to deﬁne the trigonometric

functions. To do this we return to Figure 4.1. You see a right triangle with

vertices (0, 0), (x, 0) and (x, y). We may use a circle of any radius r. The

154 CHAPTER 4. TRIGONOMETRIC FUNCTIONS

-3 -2 -1 1 2 3

-40

-20

20

40

Figure 4.4: tan x on [−π, π]

-1 -0.5 0.5 1

-2

-1

1

2

Figure 4.5: tan x on [−1.1, 1.1]

-3 -2 -1 1 2 3

-40

-20

20

40

Figure 4.6: cot x on [−π, π]

1.5 2 2.5

-1.5

-1

-0.5

0.5

1

1.5

Figure 4.7: cot x on [

π

2

−1,

π

2

+1]

-3 -2 -1 1 2 3

-10

10

20

Figure 4.8: sec x on [−π, π]

-3 -2 -1 1 2 3

-20

-10

10

20

Figure 4.9: csc x on [−π, π]

155

degrees radians sin x cos x tan x cot x sec x csc x

0 0 0 1 0 n/a 1 n/a

30 π/6

1

2

√

3

2

√

3

3

√

3

2

√

3

3

2

45 π/4

√

2

2

√

2

2

1 1

√

2

√

2

60 π/3

√

3

2

1

2

√

3

√

3

3

2

2

√

3

3

90 π/2 1 0 n/a 0 1 n/a

120 2π/3

√

3

2

−

1

2

−

√

3 −

√

3

3

−2

2

√

3

3

135 3π/4

√

2

2

−

√

2

2

−1 −1 −

√

2

√

2

150 5π/6

1

2

−

√

3

2

−

√

3

3

−

√

3 −

2

√

3

3

2

180 π 0 −1 0 n/a −1 n/a

Table 4.1: Values of Trigonometric Functions

right angle is at the vertex (x, 0) and the hypotenuse has length r. Let α

be the angle at the vertex (0, 0). In the following the words adjacent and

opposing are in relation to α. Then

sin α =

opposing side

hypothenuse

cos α =

adjacent side

hypothenuse

tan α =

opposing side

adjacent side

cot α =

adjacent side

opposing side

sec α =

hypothenuse

adjacent side

csc α =

hypothenuse

opposing side

Trigonometric Identities: There are several important identities for

the trigonometric functions. Some of them you should know, others you

should be aware of, so that you can look them up whenever needed. From

the theorem of Pythagoras and the deﬁnitions you obtain

sin

2

x + cos

2

x = 1, sec

2

x = 1 + tan

2

x, csc

2

x = 1 + cot

2

x. (4.5)

The following identities are obtained from elementary geometric observa-

156 CHAPTER 4. TRIGONOMETRIC FUNCTIONS

tions using the unit circle.

sin x = sin(x + 2π) = sin(π −x) = −sin(−x)

cos x = cos(x + 2π) = −cos(π −x) = cos(−x)

cos x = sin(x +

π

2

) = −cos(x +π) = −sin(x +

3π

2

)

sin x = −cos(x +

π

2

) = −sin(x +π) = cos(x +

3π

2

)

You should have seen, or even derived, the following addition formulas in

precalculus.

sin(α +β) = sin αcos β + cos αsin β (4.6)

sin(α −β) = sin αcos β −cos αsin β (4.7)

cos(α +β) = cos αcos β −sinαsin β (4.8)

cos(α −β) = cos αcos β + sinαsin β (4.9)

tan(α +β) =

tan α + tan β

1 −tan αtan β

(4.10)

tan(α −β) =

tan α −tan β

1 + tan αtan β

(4.11)

These formulas specialize to the double angle formulas

sin2α = 2 sin αcos α and cos 2α = cos

2

α −sin

2

α (4.12)

From the addition formulas we can also obtain

sinαsin β =

1

2

[cos(α −β) −cos(α +β)] (4.13)

sinαcos β =

1

2

[sin(α −β) + sin(α +β)] (4.14)

cos αcos β =

1

2

[cos(α −β) + cos(α +β)] (4.15)

which specialize to the the half-angle formulas

sin

2

α =

1

2

[1 −cos 2α] and cos

2

α =

1

2

[1 + cos 2α] (4.16)

c Copyright 2003 by the author. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior written permission of the author. Printed in the United States of America. This publication was typeset using AMS-TEX, the American MathematA ical Society’s TEX macro system, and L TEX 2ε . The graphics were produced 1. with the help of Mathematica This is an incomplete draft which will undergo further changes.

1

Mathematica Version 2.2, Wolfram Research, Inc., Champaign, Illinois (1993).

Contents

Preface 1 Basic Concepts 1.1 Real Numbers and Functions . . . . . . . . 1.2 Limits . . . . . . . . . . . . . . . . . . . . . 1.2.1 Two important estimates . . . . . . 1.3 More Limits . . . . . . . . . . . . . . . . . . 1.4 Continuous Functions . . . . . . . . . . . . 1.5 Lines . . . . . . . . . . . . . . . . . . . . . . 1.6 Tangent Lines and the Derivative . . . . . . 1.6.1 Derivatives without Limits . . . . . 1.7 Secant Lines and the Derivative . . . . . . . 1.8 Diﬀerentiability implies Continuity . . . . . 1.9 Basic Examples of Derivatives . . . . . . . . 1.10 The Exponential and Logarithm Functions . 1.11 Diﬀerentiability on Closed Intervals . . . . . 1.12 Other Notations for the Derivative . . . . . 1.13 Rules of Diﬀerentiation . . . . . . . . . . . 1.13.1 Linearity of the Derivative . . . . . . 1.13.2 Product and Quotient Rules . . . . . 1.13.3 Chain Rule . . . . . . . . . . . . . . 1.13.4 Hyperbolic Functions . . . . . . . . 1.13.5 Derivatives of Inverse Functions . . . 1.13.6 Implicit Diﬀerentiation . . . . . . . . 1.14 Related Rates . . . . . . . . . . . . . . . . . 1.15 Exponential Growth and Decay . . . . . . . 1.16 More Exponential Growth and Decay . . . 1.17 The Second and Higher Derivatives . . . . . 1.18 Numerical Methods . . . . . . . . . . . . . . i v 1 1 2 4 6 8 9 10 12 13 14 15 18 20 21 22 22 23 26 29 30 34 38 41 43 48 48

. . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Monotonicity on Intervals . . . . . . 3. . . . . . . . . . . . . . 3. . . . . .3. . . . 3. . . . . . . . . . . . .1 Approximation by Diﬀerentials 1. . .4.12 Applications of the Integral . 2.2. . .6 Areas and Integrals . . .4 Integrable Functions . . . . . . . . . . . . . . . 3. . . . . . . . . 3. . . . . . . . . . . . 2. . . . . . . 3. . . . . . . . . . . . . . 3 Integration 3. .3. . . . . . .2. .3. . . . . . . . .3 The First Derivative and Monotonicity . . . . .1 Substitution and Deﬁnite Integrals 3. ii . . . . . . . . . .10 Areas between Graphs .1 Upper and Lower Sums . . . . . . . . . .18. . . . . . . . . . . . . . . . . . . . . . 2. . . . .1 Properties of Areas . . . . . . 3. . 48 51 53 63 65 65 67 69 69 75 76 77 80 81 83 88 90 92 98 105 105 107 108 111 112 112 115 116 118 121 122 124 126 128 131 133 136 142 2 Global Theory 2. . . . . . . . . . 3. . . . . . . . . . 2. . . . . . . . 3. . . .1 Concavity on Intervals . . . . . . . . . .18. . 1. . . . . . . . . . . . .6 Detection of Local Extrema .3.1. . . .4 The Second Derivative and Concavity . . . . . . .2 Newton’s Method . . . . . . . . .2 Unique Solutions of Diﬀerential Equations 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5 Local Extrema and Inﬂection Points .1 Cauchy’s Mean Value Theorem . . . . 3. . . . . . 1. . 3. . . . . . 2. . . . . . . . . . . . . . . . . . . . 2. . . 2. . . . . . . . . . . . . 2. . .9. . . . .2 Partitions and Sums . . . . . .10 Sketching Graphs . . . . . . . .11 Numerical Integration . . . . . . . . . . 2. . . . 2. . . 3. . . . . . . . . . . .8. . .2 Monotonicity at a Point . . . . . . .2 The Riemann Integral . . . . . . . . . . . . . .18. . . . . . . . . . 3. . .19 Table of Important Derivatives . . . . . . . . . . . . . . . . .7 Detection of Inﬂection Points . . . . .7 Anti-derivatives . .1 Some Proofs . . . . . . . . . . . . . . . . . . . .2 Concavity at a Point . 3. . . . . . 2. . . .3 Limits and Integrability . .8 Absolute Extrema of Functions . . . . . . . . . . . . .9 Substitution . . . . . . . . . . . . . . . . . . . . 2. . . . . . . . . . .4. . . . . . . . . . . . . . . . . . . . . . . . . 3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Riemann Sums . . . . . . .8 The Fundamental Theorem of Calculus . . . . . . . . . .9 Optimization Story Problems . . . . . . . . . . . . . . . .1 The Darboux Integral and Areas . .3 Euler’s Method . . . . . . . . . 3. . .5 Some elementary observations . . . . .

. . . .1 Other Bases . . . . . .13. . . . . . . .13 The Exponential and Logarithm Functions . . . . . . . . . . 148 4 Trigonometric Functions 151 iii . .3. . . . . . . . 144 3.

iv

Preface

In these notes we like to summarize calculus.

v

vi

PREFACE

or zero. and {x ∈ R | |x − a| < } is the set of all real numbers whose distance from 1 .Chapter 1 Basic Concepts Introduction In this chapter we introduce limits and derivatives. The image of a function f consists of all those points y in the range for which there exists an x in the domain of f .. Until further notice. If x and y are real numbers. and they take values in R.1 Real Numbers and Functions We assume that the reader is familiar with the real numbers (denoted by R) and the operations of addition and multiplication. A real number is either positive. These are basic concepts of calculus. x if x > 0 |x| = −x if x < 0 0 if x = 0 The distance between two points a and b on the real line is |a − b|. 1. The range of a function is a set in which the function takes values. negative. then x is larger than y (i. are subsets of the real numbers. Their domains. We provide some rules for their computations. We will make frequent use of the absolute value function. we will work with real valued functions in one real variable. x > y) if x − y is positive.e. the sets on which these functions are deﬁned. such that f (x) = y. This allows us to order the real numbers.

In all but a few degenerate cases. which means that every bounded subset of the real line has a least upper bound. but arguments using it are too diﬃcult for an introductory course on the subject.2 Limits Limits are a central tool in calculus and other areas of mathematics.2) L = lim f (x) x→a if for all > 0 there exists a δ > 0. Every now and then we will allude to the completeness of the real line. a + ).1) |x · y| = |x| · |y|. for all δ > 0 there is a point b in the domain of f . points for which there are no other arbitrarily close points in the domain of the function. Proposition 1. 2 Some authors do not apply the concept of a limit at isolated points of the domain of a function. Let f be a function and L a real number. then this limit is unique. Suppose that f (x) has a limit at x = a. |x + y| ≤ |x| + |y|.1. This property is crucial for calculus. and ||x| − |y|| ≤ |x − y|. for any two real numbers x and x (1. whenever x is The equation in (1. such that 0 < |b − a| < δ.2. 1 .1 2 The latter assumption in the proposition is satisﬁed if the domain of f contains an interval. We also say that f (x) approaches or converges to L as x approaches a. For computations with absolute values it is worth noting that. limits are unique if they exist. provided that the domain of the function f contains points arbitrarily close to a. such that |f (x) − L| < in the domain of f and 0 < |x − a| < δ. Deﬁnition 1. based on the values of f (x) for x near a.2 CHAPTER 1. We say that (1. We discuss them in this section. 1. Expressed as an interval this set is (a − . and either a belongs to this interval or a is an end point Expressed in mathematical language this means. and the last one is a variation of it. BASIC CONCEPTS a is less than . The ﬁrst inequality is referred to as triangle inequality. An intuitive interpretation is that the expected value of f (x) at x = a is L.2) reads as L is the limit of f (x) as x approaches a.

The last assertion can be proved using synthetic division.3. If x→a lim f (x) = L = lim g(x).7. a) or (a. at least if n is an integer. x→0 x lim sin x = 1. Taking limits is compatible with the basic algebraic operations in the following sense. and h(x) all contain an interval of the form (d. To avoid intricate language. Proposition 1. Assume that the domains of the functions f (x) and g(x) both contain an interval of the form (d. then x→a lim f (x) = f (a). Assume that the domains of the functions f (x). As a special case we obtain the following useful observation: (1. or a trigonometric function and f (a) is deﬁned. For many functions the computation of limits is no challenge.4 (Pinching Theorem). and it is equal to L.4) cos x − 1 = 0. The following limits are important in the calculations of some derivatives. . x→a x − a Hints: The ﬁrst two limits follow easily from the estimates in Theorem 1. g(x).3) x→c lim f (x) = L if and only if x→c lim (f (x) − L) = 0. a) or (a. Proposition 1.1.2. LIMITS 3 of it. (1.5. a rational function. we make this kind of an assumption for the remainder of this section. Then x→a lim (f + g)(x) = M + L x→a x→a lim cf (x) = cM lim (f · g)(x) = M · L x→a lim (f /g)(x) = M/L provided that L = 0. If f (x) is a polynomial. and that c is a constant. e) where d < a < e and that f (x) ≤ h(x) ≤ g(x). Proposition 1. x→0 x lim and lim xn − an = nan−1 . x→a then the limit of h(x) exists as x approaches a. e) where d < a < e. Suppose that x→a lim f (x) = L and x→a lim g(x) = M. discussed in the following subsection.

Proof.2. imagine that you roll the circle along the vertical line through B until the point C touches it in the point E. π/4]. then (1.6. This Here our argument relies on intuition. |h| = BE.1 you see part of the unit circle. If h ∈ [−π/4. It follows that |h| ≤ tan |h|. D E C O A B Figure 1. We denote by BC the length of the arc (part of the unit circle) between B and C. π/4] we set C = (cos h. sin h).1 Two important estimates In preparation of the proof of Theorem 1. A rigorous argument requires work. to show that |h| = BC ≤ | tan h| = BD. This slice is contained in the triangle with vertices O. It appears to be clear3 that BE ≤ BD.1: The unit circle We ﬁnd that | sin h| = AC ≤ |h| = BC because going from C straight down to the x-axis is shorter than following the circle from C to the x-axis.5) | sin h| ≤ |h| ≤ | tan h|. BASIC CONCEPTS 1. B and C has area |h|/2. For h ∈ [−π/4. From this is follows by elementary geometry that the area of the slice of the disk with vertices O. One can show that the area of a disk with radius one is π. We use the process of rolling the circle along the line to measure |h|. In particular. the distance between them is denoted by XY . Secondly. Given two points X and Y in the plane. B and D. 3 . In Figure 1. and the area of the slice is (tan |h|)/2.4 CHAPTER 1.7 we show Theorem 1.

Let BC be the length of the arc (part of the unit circle) between B and C.1. If h ∈ [−π/4. In Figure 1. then4 h2 h2 and |h − sin h| ≤ . (1. B.7. sin h). Restricting ourselves to this interval simpliﬁes the proofs somewhat. 2 2 Proof of Theorem 1.6) |1 − cos h| ≤ C A D B Figure 1. 4 The inequalities hold without the restriction on h. and C. and a triangle with vertices A. and BC ≤ BC. π/4]. the second inequality which we claimed in the theorem.2: The unit circle From the picture we read oﬀ that AB = 2.2. Let h ∈ [−π/4. Denote by XY the length of the straight line segment between the points X and Y . .2 you see half of a circle of radius 1 centered at the origin. LIMITS 5 veriﬁes that |h| ≤ tan |h|. Using similar triangles we see AB/BC = BC/DB and (BC)2 = AB × DB. but we only need them on an interval around zero. DB = (1 − cos h). Theorem 1. In other words 2(1 − cos h) = AB × DB = (BC)2 ≤ (BC)2 = h2 . π/4] be the number for which we want to show the inequality and C = (cos h.7. BC = |h|.

The last two limits allow us to express that the values of a function tend to plus or minus inﬁnity. In some situations one would like to modify the deﬁnition in Section 1.8. such that |f (x) − L| < in the domain of f and a − δ < x < a.6 CHAPTER 1. such that |f (x) − L| < in the domain of f and a < x < a + δ. h 2 2 The second estimate claimed in the theorem is an immediate consequence. We call them inﬁnite limits.6 tells us that | sin h| sin h | sin h| ≤ |h| ≤ | tan h| = hence 0 ≤ cos h ≤ ≤ 1. If 0 = h ∈ [−π/4. then both sides of the second inequality in (1. and we do so in this section. If h = 0. ﬁrst section about limits suﬃces for a while. whenever x is Deﬁnition 1. The next two limits express what happens as the variable tends to plus or minus inﬁnity. We call them limits at inﬁnity. They are called the right and left hand limits. Let f be a function and L a real number. π/4].2.6) is an immediate consequence.9. Let f be a function and L a real number. whenever x is .3 More Limits The material in the previous. cos h h Subtracting the terms in this inequality from 1 we ﬁnd 0 ≤1− sin h ≤ 1 − cos h ≤ 1. The ﬁrst two limits express how the function behaves as we approach a point a from the right or left. verifying the assertion in this case.6) are zero. then Theorem 1. 1. Deﬁnition 1. We say that L = lim f (x) x→a− if for all > 0 there exists a δ > 0. h Using our previous estimate for |1 − cos h| and our assumption that |h| ≤ π/4 < 1. we conclude that h − sin h h2 h ≤ |1 − cos h| ≤ ≤ . BASIC CONCEPTS The ﬁrst estimate in (1. We say that L = lim f (x) x→a+ if for all > 0 there exists a δ > 0.

then x→a+ 7 lim f (x) = 1 and x→a− lim f (x) = −1. We can consider what happens to the values of a function f (x) as x approaches ∞ or −∞. if f (x) = sign(x) = x/|x|. MORE LIMITS For example. whenever Deﬁnition 1.3. In the last two deﬁnitions a may be replaced by a± . Let f be a function and a a real number. so that we approach a from the left or right. Let f be a function and a a real number. and a can be replaced by ±∞. In other words.11. For example √ 1 lim = ∞ and lim x = ∞. We say that x→a lim f (x) = −∞ if for all M there exists a δ > 0 such that f (x) < M whenever x is in the domain of f and 0 < |a − x| < δ. We say that L = lim f (x) x→∞ if for all > 0 there exists a number M . Deﬁnition 1. + x x→∞ x→0 . We say that x→a lim f (x) = ∞ if for all M there exists a δ > 0 such that f (x) > M whenever x is in the domain of f and 0 < |a − x| < δ. we can make sure that the value of f (x) is larger than any given number M . such that |f (x) − L| < x is in the domain of f and x > M . 1 + x2 Deﬁnition 1. Let f be a function and L a real number. no matter how large. We say that L = lim f (x) x→−∞ if for all > 0 there exists a number M . by taking x close to a. such that |f (x) − L| < x is in the domain of f and x < M .13. For example x→∞ whenever lim 1 = 0 and x x→−∞ lim 1 = 1.1. Let f be a function and L a real number. Deﬁnition 1.12.10.

rational functions. f · g. Theorem 1. then there exists a c ∈ [a..e. g is deﬁned at points arbitrarily close to c. f /g and f ◦ g are continuous. If C is in between f (a) and f (b). b]. Proposition 1. and trigonometric functions are continuous. The clarify the remark about the domain in the proposition. One may also reverse the order of applying a continuous function and calculating a limit: (1. If c is an isolated point in the domain of f . .8) x→c lim f (g(x)) = f lim g(x) . x→c provided the natural technical assumption hold. For the composition (f ◦ g)(x) = f (g(x)) on needs that g takes values in the domain of f .2. One can produce many more continuous functions through standard operations on functions. The class of continuous functions will play a central role later.16 (Intermediate Value Theorem). and f is continuous at limx→c g(x). i.14 says that (1.4 Continuous Functions We deﬁne continuous functions and discuss a few of their basic properties. then the function is always continuous at c.15. wherever these functions are deﬁned. we note that the function (f + g)(x) = f (x) + g(x) is deﬁned for those x for which both f and g are deﬁned. such that f (c) = 0. Deﬁnition 1. The function is said to be continuous at c if for all > 0 there exists a δ > 0. Polynomials. there are no other points in the domain of f arbitrarily close to c. this equation holds whenever there are points in the domain of f arbitrarily close to c. i. Then f + g. Suppose that f is deﬁned and continuous on the closed interval [a. f is deﬁned for all g(x) where x is in the domain of g and close to c. b].14. See the footnote to Proposition 1.8 CHAPTER 1. In most cases the condition in Deﬁnition 1.7) x→c lim f (x) = f (c). To determine the domain of f /g one needs to exclude those points where g is zero. Let f and g be continuous functions. The same statement holds for (f · g)(x) = f (x) · g(x). Let f be a function and c a point in its domain.e.. A function f is continuous if it is continuous at all points in its domain. such that |f (c) − f (x)| < whenever x belongs to the domain of f and |x − c| < δ. BASIC CONCEPTS 1. In fact.

17 (Extreme Value Theorem). According to the theorem there exists some c ∈ (0. Given any two points (x1 . and B is the point in which the line intersects the y-axis. Then there exist points c and d in [a. b]. Theorem 1.10) a c y = − x + = mx + B. They are equivalent to the completeness of the real line. a line consists of the points (x.1. . such that f (c) ≤ f (x) ≤ f (d) for all x ∈ [a. We mentioned this property of the real numbers in Section 1.g. y) in the plane which satisfy the equation (1.5 Lines In general. also called the y-intercept.1. 1. The line is vertical if and only if b = 0. the line through them has slope m= y2 − y1 . b b The number m is called the slope of the line. y1 ) is (1.9) ax + by = c for some given real numbers a. x2 − x 1 For our purposes. b]. such that p(c) = 0. the most useful version of the equation of a line is its point-slope formula. the theorem says that a continuous function on a closed interval assumes a smallest and largest value. y2 ) in the plane.11) y = m(x − x1 ) + y1 . The Intermediate Value and Extreme Value theorem are typically proved in an introductory analysis course. where it is assumed that a and b are not both zero. Expressed in words.. The polynomial is certainly a continuous function. The equation of a line with slope m through the point (x1 .5. 1). b]. p(0) = −1 and p(1) = 1. b and c. If b = 0 we may rewrite the equation as (1. y1 ) and (x2 . suppose that p(x) = x3 − x2 + 2x − 1. LINES 9 E. Let f be deﬁned and continuous on the closed interval [a.

this deﬁnition is hard to work with.e. We call a line t(x) the tangent line to the graph of f (x) at x = c if t(x) is the best linear approximation of f (x) on some open interval around c. Deﬁnition 1. Suppose f (x) is a function and c is an interior point of its domain. Deﬁnition 1.18.12) expresses in a precise form in which sense the tangent line is close to the graph of f (x) near c. Suppose f (x) is a function and c is an interior point of its domain.13) t(x) = f (c)(x − c) + f (c).10 CHAPTER 1. For a given function and an interior point c in its domain there may or may not be a tangent line. it does so even when divided by x − c. Not only does f (x) − t(x) converge to zero as x approaches c. .19.12) x→c lim f (x) − t(x) = 0. Although the term ‘best linear approximation near c’ gives an excellent intuitive picture what a tangent line is.20.6 Tangent Lines and the Derivative We like to introduce the concept of tangent lines. Suppose f (x) is a function and c is an interior point of its domain.. We use tangent lines to deﬁne the concept of diﬀerentiability and the derivative. i. such that c ∈ I ⊆ B. then it is unique. Then we say that f (x) is diﬀerentiable at c. It is easier to work with a more concrete deﬁnition. the line t(x) is closer to the graph of f (x) than any other line for all x in some open interval around c.21. We call a line t(x) the tangent line to the graph of f (x) at x = c if (1. BASIC CONCEPTS 1. Deﬁnition 1. To be able to express ourselves concisely. but it there is a tangent line. We call the slope of the tangent line the derivative of f (x) at c. and assume that there is a tangent line to the graph of f (x) at x = c. let us say Deﬁnition 1. x−c The equation in (1. We give a ﬁrst deﬁnition for a tangent line. A point c is an interior point of a subset B of R if there is an open interval I. and we denote it by f (c). Utilizing the notation in the previous deﬁnition we can write down the equation of the tangent line to the graph of f (x) at x = c in point-slope form: (1.

The technique used in the example. this means that t(x) is the desired tangent line. TANGENT LINES AND THE DERIVATIVE 11 To diﬀerentiate a function means to ﬁnd its derivative. By deﬁnition.1. You are expected to ﬁll in some of the arithmetic steps. Say. we see that p(x) − t(x) x−c = p(x) − t(x) x+2 (provided |x + 2| ≤ 1) = |2(x + 2)3 − 16(x + 2)2 + 45(x + 2)| ≤ 65|x + 2| This estimate shows that (p(x) − t(x))/(x − c) converges to zero as x approaches c = −2. Deﬁnition 1. We consider f (x) as a function. To do so. an open set is a set.6. p = 2(u − 2)4 − 3(u − 2)2 + 5 = 2(u4 − 8u3 + 24u2 − 32u + 16) − 3(u2 − 4u + 4) + 5 = 2u4 − 16u3 + 45u2 − 52u + 25 Reversing the substitution. Solution: As a ﬁrst step we expand p in powers of u = (x + 2). Then say that f (x) is diﬀerentiable if it is diﬀerentiable at each point of its domain. Example 1. By deﬁnition. whose domain consists of all those points where f (x) is diﬀerentiable. Find the tangent line t(x) to the graph of p(x) at x = −2 and p (−2). shows that t(x) = A1 (x − c) + A0 . the result is p(x) = An (x − c)n + · · · + A1 (x − c) + A0 .22. replacing u by (x + 2). For t(x) as proposed. suitably generalized. such that each of its points is an interior point. We assert that t(x) = −52(x + 2) + 25 and p (−2) = −52. Suppose the domain of the function f (x) is an open set. Its slope is p (−2) = −52. we substitute u − 2 for x and expand p in powers of u. We can use any polynomial p(x) and point x = c and write p(x) in powers of (x − c). ♦ The example is generic.23. we ﬁnd: p(x) = 2(x + 2)4 − 16(x + 2)3 + 45(x + 2)2 − 52(x + 2) + 25. Let p(x) = 2x4 − 3x2 + 5.

and t(x) have the same value at x = c. In an example. but we have shown that Proposition 1. Suppose f (x) is a function and c is an interior point of its domain. With this deﬁnition fewer functions will be diﬀerentiable than with the one given in Deﬁnition 1. There is very little space left between p(x) and q(x).20. and f (x) and t(x) are squeezed in between them. and the tangent line t(x) (dashed). There you see the function f (x) = sin x. tangent line. this situation is shown in Figure 1. In terms of algebraic geometry (1.12 CHAPTER 1. the deﬁnition of a limit is the most diﬃcult one in a ﬁrst semester of calculus. 1. the graphs of f (x) and t(x) have to be close to each other near x = c. BASIC CONCEPTS is the tangent line to the graph of p(x) at x = c. and derivative.24. The parabolas p(x) and q(x) touch each other without crossing.14) is also more accessible to computer assisted algebra than the limit deﬁnition.14) at least alludes to a divisibility condition. All four function f (x). but this is not crucial. Deﬁnition 1. without the limit concept. |f (x) − t(x)| ≤ A(x − c)2 . The inequality in (1. We call a line t(x) the tangent line to the graph of f (x) at x = c if there exists and open interval I around c and a number A.25. and the picture shows how they ‘hug’ each other.14) can be rewritten as q(x) = t(x) − A(x − c)2 ≤ f (x) ≤ t(x) + A(x − c)2 = p(x). There is also a geometric picture which illustrates the concept of closeness. such that (1. p(x). the parabola q(x) (dotted and open downwards). and it is interesting to explore ways to develop calculus. One can do this by replacing the condition in (1.1 Derivatives without Limits Without a doubt. q(x).3. rigorously. the parabola p(x) (dotted and open upwards). Eventually we will ﬁnd a more eﬃcient method for diﬀerentiating polynomials.12) by a slightly stronger one. The condition in (1.6.14) for all x ∈ I. where the parabolas q(x) and p(x) are deﬁned by the expressions they are adjacent to. Polynomials are diﬀerentiable. and p (c) = A1 . A pedagogical advantage of the approach is that one does not have to understand limits before one can understand the deﬁnition of the derivative. In this sense.

Suppose f is a function and c is an interior point of its domain. If f is diﬀerentiable at c.5 2 Figure 1.25 1 0. (1. x−c x−c Apply limits to both sides of the equation and the assertion follows.26.5 1. SECANT LINES AND THE DERIVATIVE 13 1. c−x x→c Proof.15) f (x) − t(x) f (x) − f (c) = − f (c). then f (c) = lim f (c) − f (x) .7. Let us explain the situation geometrically. and its slope (f (a) − f (b))/(a − b) is called . The line through (a. Theorem 1. Suppose a and b are distinct points in the domain of the function f .5 1 1.75 0.7 Secant Lines and the Derivative Often a diﬀerent approach is taken to motivate and introduce the derivative.25 0.3: Sine Function and Tangent Line between two Parabolas 1. f (a)) and (b.5 0. f (b)) is called a secant line. This is obvious once one uses the expression for the tangent line in (1.13) and substitutes it in the expression in (1.1.12) inside the limit.

5: f (x) = x2 sin(1/x) It is misleading to say that the graph of f (x) looks like.00005 0.4: f (x) = x2 sin(1/x) Figure 1.01 -0.20. . In (1.01 0.00005 Figure 1. 0. The theorem asserts that for a diﬀerentiable function this limit of the slopes of secant lines is the slope of the tangent line.005 -0. f (x)).005 0. Many authors introduce the derivative as the limit of the slopes of secant lines. You see part of its graph over two diﬀerent intervals in Figure 1.0075 0.8 Diﬀerentiability implies Continuity It is worth pointing out that Theorem 1.005 0.0001 -0. (f (c)) and (x. For the obvious reason f (c) is called the rate of change or instantaneous rate of change of f at c. and then we take the limit as x approaches c.0001 0. Eventually you will be able to show that the function f (x) = x2 sin(1/x) 0 if x = 0 if x = 0 is diﬀerentiable everywhere on the real line.0025 -0.15) we are considering the slopes of secant lines through (c. then it is continuous at this point.1 -0. By no stretch of imagination will you say that the graph of the function looks like a line. and possibly illustrate that the tangent line is close to the graph in the sense of Deﬁnition 1.27. or resembles.005 0. If a function is diﬀerentiable at a point. b].05 -0. call t(x) = f (x − c) + f (c) the tangent line.5.4 and 1. 1.14 CHAPTER 1.05 0. BASIC CONCEPTS the average rate of change of f over the interval [a.1 -0. a line near c.0075 -0.0025 -0.

. It is apparent from the graph (see Figure 1. There are continuous functions which are not diﬀerentiable. E. x−c Then certainly x→c lim [(f (x) − f (c)) − f (c)(x − c)] = 0. we have to study the diﬀerence quotients (|x|−|0|)/(x−0) = |x|/x. We can also give an analytic argument. This implies that limx→c f (x) = f (c) and that f (x) is continuous at c.5 1 0. Denote the function by f (x) and the point of diﬀerentiability by c. but it is not diﬀerentiable at x = 0.g. BASIC EXAMPLES OF DERIVATIVES 15 Proof. According to the deﬁnition of diﬀerentiability.9 Basic Examples of Derivatives Let us use the deﬁnitions and work out a few derivatives.6) that there is not line close to the graph of this function near x = 0. There is no number these diﬀerence quotients converge to. Because f (c)(x−c) converges to zero as x approaches c. the function f (x) = |x| is continuous. By assumption we have the derivative f (c) and x→c lim f (x) − f (c) − f (c) = 0.6: The absolute value function The converse of the theorem is false. 1.1. so does (f (x)−f (c)). and f (x) = |x| is not diﬀerentiable at x = 0.5 -2 -1 1 2 Figure 1. They are 1 if x > 0 and −1 if x < 0. 2 1.9. .

2 ac + b Verify that (1.16) |f (x) − t(x)| ≤ a2 √ (x − c)2 . √ Exercise 1. then f (x) = axa−1 . lim 1 x x→c −1 c−x 1 c = lim = − 2. Suppose that f (x) = ax + b and ax + b > 0. If f (x) = 1/x. Suppose that n ≥ 2. then f (x) = nxn−1 . then f (x) = −1/x2 .16 CHAPTER 1. .30. Proof. Example 1.28. 1. 3 2( ac + b) . . . Example 1. 2.. generalizing all of the examples above. = lim x→c x − c x→c (x − c)( x + c) 2 c lim Remark 1. BASIC CONCEPTS Example 1. 2 ax + b The tangent line to the graph of f (x) at x = c is then √ a t(x) = √ (x − c) + ac + b. √ √ x− c x−c 1 √ √ = √ . . then f (x) = 1/(2 x). Eventually we will see that if f (x) = xa for any real number a. If f (x) = Proof. Proof. Suppose c = 0. n = 0. x→c xc(x − c) x−c c √ √ x and x > 0. Then xn − cn = lim (xn−1 + xn−2 c + · · · xcn−2 + cn−1 ) = ncn−1 x→c x − c x→c lim The cases n = 0 and n = 1 are even easier and left to the reader. i. If f (x) = xn and n is a non-negative integer.e.29. Show that a f (x) = √ .

lim sin x − sin c x−c = = = = = sin(c + h) − sin c h→0 h sin c cos h + cos c sin h − sin c lim h→0 h sin c(cos h − 1) + cos c sin h lim h→0 h cos h − 1 sin h sin c · lim + cos c · lim h→0 h→0 h h cos c.9. Example 1.4).17) does not only show diﬀerentiability in the sense of Deﬁnition 1. and provides an explicit error estimate.1. if t(x) = sin c(x − c) + cos x is the tangent line to the graph of f (x) at x = c.25. then |f (x) − t(x)| ≤ (x − c)2 . BASIC EXAMPLES OF DERIVATIVES 17 The estimate in (1. The details are similar to the ones in Example 1.16) shows diﬀerentiability in the sense of Deﬁnition 1. The estimate in (1.25. Proof. but it provides an explicit error estimate. Show that sin x = cos x. lim x→c For computation of the limits in the second to last line see (1.31. If f (x) = cos x. Furthermore.31. a bound on the diﬀerence between the function and its tangent line. the angle x needs to be measured in radians. a bound on the diﬀerence between the function and its tangent line. It is left as an exercise for the reader to show that (1. .17) | sin x − t(x)| ≤ (x − c)2 The steps are essentially the same as in the proof above. The tangent line to the graph of the sine function at x = c is t(x) = cos c(x − c) + sin c. then f (x) = − sin x. Exercise 2. Below we will set x = c + h and x − c = h. For this equation to hold.

The arithmetic properties of the exponential function. The theorem just says that the exponential laws. which you previously learned for rational exponents. aloga y = y and loga (ax ) = x . and we will follow this convention. a = 1.32. More precisely. also hold in the generality of our current discussion. ∞) as the range of the exponential function expa (x). Still.33 (Exponential Laws). so that the domain of the exponential function is (−∞. ax > 0 for all x. For any positive real number a and all real numbers x and y ax ay = ax+y ax /ay = ax−y (ax )y = axy If x is the unique solution of the equation ax = y. First we take a p-th power and then a q-root. BASIC CONCEPTS 1. There exists exactly one monotonic function. are collected in our next theorem. also called the exponential laws. For every number y > 0 there exists exactly one number x. It is common. to use the notation ax for expa (x) also if x is not rational.18 CHAPTER 1. In this sense we have a function h(r) = ar . Theorem 1. such that expa (x) = y. More explicitly. technically we are not quite prepared for it and at a later point we have to revisit the introduction to ﬁll in details. We just deﬁned a function loga (y). called the exponential function with base a and denoted by expa (x). Theorem-Deﬁnition 1. and by construction it is the inverse of the exponential function expa (x). ∞). It is called the logarithm function with base a. Suppose a is a positive real number and a = 1. This function is monotonic.18) loga (y) = x. then we set (1. Furthermore.10 The Exponential and Logarithm Functions The exponential and logarithm are of great importance and we do not want to delay their introduction any further. h(r) is increasing if a > 1 and decreasing when 0 < a < 1. so we use (0. √ any rational number For r = p/q (p and q are integers) one can deﬁne ar = q ap . which is deﬁned for all real numbers x such that expa (x) = ax whenever x is a rational number. Let a be a positive number. whose domain consists of all rational numbers.

5 2 1. Deﬁnition 1.7 and 1. Theorem 1. The domain of the logarithm function is (0. 2. and vice versa. This irrational number is called the Euler number (named after Leonard Euler) and denoted by e.8: log2 (x) The Euler number e as base There is one number which is preferrable as base over the others.7: exp2 (x) Figure 1. It is denoted by ln(x).5 0. Its inverse is the natural logarithm function. It is increasing if a > 1 and decreasing if 0 < a < 1.34 (Laws of Logarithms). THE EXPONENTIAL AND LOGARITHM FUNCTIONS 19 for all x ∈ R and all y > 0.718281828.5 -1 -1 -0.35.5 1 0. ∞). So exp(x) = expe (x) and ln(x) = loge (x). for all positive real numbers x and y. Corresponding to the exponential laws in Theorem 1.5 1. and any real number z loga (xy) = loga (x) + loga (y) loga (x/y) = loga (x) − loga (y) loga (xz ) = z loga (x) In Figures 1.5 3 Figure 1.5 0.33 we have the laws of logarithms.5 0. .5 1 1.10. It is denoted by exp(x) or ex . The exponential function is the exponential function for the base e.5 1 -0.8 you see parts of the graphs of the exponential and logarithm functions with base 2. and e ≈ 2. ∞) and its range is (−∞.5 1 1. We will deﬁne it precisely later. One set of laws implies the other one.1.5 2 2. For any positive real number a = 1.

Suppose that I and J are subsets of the real line R and I ⊆ J.11 Diﬀerentiability on Closed Intervals In Deﬁnition 1. 1. i. F (x) = f (x) for all x ∈ I.19) exp (x) = exp(x) CHAPTER 1. ln a These identities follow from the exponential laws and the laws of logarithms. BASIC CONCEPTS and ln (x) = 1 . Without some restrictions on I. A function f is said to be diﬀerentiable on a subset I of R if it extends to a diﬀerentiable function F on an open set. We call F an extention of f if it agrees with f on I.39. We set f (x) = F (x) for all x ∈ I. Other Bases Finally.38. Suppose the function f is deﬁned on an interval I. Deﬁnition 1.22 we deﬁned what it means that a function is diﬀerentiable on an open set. and that J is the domain of a function F .37. then f (x) = F (x) is unique for all x ∈ I. . and f extends to a diﬀerentiable function F on an open interval containing I. the interval is neither empty nor a single point. x The derivative of the exponential function is the exponential function. The least technical and for our purposes suﬃcient solution is captured in Proposition 1.. Deﬁnition 1. and the derivative of the natural logarithm function is 1/x.36. let us relate the exponential and logarithm functions for diﬀerent bases to those with base e. Theorem 1. ax = ex ln a and loga x = ln x . Let us formalize the idea of extending functions. There are situations in which one would like to apply the notion of diﬀerentiability to functions with other kinds of domains.e.20 Eventually we will see (1. For any positive number a (a = 1). that I is the domain of a function f . a function may be diﬀerentiable without the derivative being well deﬁned.

∞). This is particularly convenient if f stands for a larger expression as in d sin x = cos x dx or d x e = ex . and if y(x) = ln x. 1] is diﬀerentiable. The function f (x) = x2 with domain [0. ∞) as domain. the function g(x) = x is not diﬀerentiable on the interval [0. This in acceptable df d because it won’t lead to confusion. then they will write x(t) instead of x (t). E.g. 0) is a vertical line. dx . then dy dy 1 = y = ex .31): If y(x) = sin x. it makes sense to write (compare Example 1.12 Other Notations for the Derivative There are diﬀerent notations for the derivative of a function. where a < b. dx dx x This notation is not always speciﬁc enough. Leibnitz’ notation for the ˙ df derivative of a function f of a variable x is dx . Some authors use one-sided limits and one-sided derivatives to contemplate derivatives at the end points of the interval.1. ∞) as its domain. and it lends itself more to generalizations in higher dimensions. (The function g(x) is diﬀerentiable if we use (0. It is the name of the variable of y as well as the name of the variable of the derivative of y. Physicists will indicate a derivative with respect to time by a dot. Instead of dx (x) we also write dx f (x). Expressing the derivatives of the exponential and natural logarithm functions this way (see (1. Let us discuss two examples. then = . In contrast. The only sensible candidate for the tangent line to the graph of g(x) at the point (0. then dy (x) = cos x. The expression dy/dx stands for the derivative of y with respect to x. To be speciﬁc about this aspect. The slope of this line is not a real number and we do not have a derivative. We will use it frequently.12.19)) we have: If y(x) = ex . OTHER NOTATIONS FOR THE DERIVATIVE 21 We are mostly concerned with deﬁning diﬀerentiability for functions whose domain is a closed interval [a.. if x is a function of time. dx In this notation x plays two roles. It extends to the diﬀerentiable function F (x) = x2 with √ the open set (−∞. b]. The expression does not tell where dy/dx is evaluated. and that is a function. Our discussion is less painful.) 1.

BASIC CONCEPTS 1. We conclude that h (x) = dh (x) = 2x + 3ex . the derivative of a sum of functions is the sum of the derivatives.20) (f + g) (x) = f (x) + g (x) and (cf ) (x) = cf (x).22 CHAPTER 1. In Leibnitz’ notation this reads d df dg (f + g)(x) = (x) + (x) dx dx dx d df (cf )(x) = c (x). dx dx (1. and assume that both of them are diﬀerentiable at x. In a more mathematical language one says that diﬀerentiation is linear. g(x) = ex and c = 3. ♦ dx . together with the knowledge of the derivatives of some basic functions. which means that no “understanding” is required. turn the process of diﬀerentiation for many functions into an algorithm. Solution: Set f (x) = x2 . see (1. You can do it even on the computer. Then f + g and cf are diﬀerentiable at x and their derivatives are given by (1.19). and the derivative of a multiple of a function is the multiple of the derivative. In the last section of this chapter we summarize the computational results of this section.13 Rules of Diﬀerentiation We discuss formulas for calculating the derivative of a composite function from the derivatives of its constituents. 1. Diﬀerentiate h(x) = x2 + 3ex . We collect the rules established in this section and tabulate the derivatives of many of the important functions which we considered.21) and In words. Let f and g be functions.40. Then h(x) = f (x) + 3g(x). Example 1.13. Let c be a real number. Previously we found that f (x) = 2x and that g (x) = ex . and practice many examples. These formulas. a rather mechanical process.1 Linearity of the Derivative Diﬀerentiation is compatible with addition of functions and multiplication with a constant. You are expected to learn the basic rules. be able to apply the accurately.

where the ai are constants.13. Solution: Recall that loga x = ln x .36.23) (cf ) = cf or d df (cf ) = c . Thinking of f and g more as functions.41. Diﬀerentiate log a x. Solution: A polynomial is a ﬁnite sum of multiples of non-negative powers of the variable. a special case of the formula which we just derived. Again.13.1. RULES OF DIFFERENTIATION 23 Example 1. For the quotient rule assume in addition that g(x) = 0. Using the linearity of the derivative. Find the derivative of an arbitrary polynomial. They allow us to calculate the derivatives of products and quotients of functions.e. see (1. see Theorem 1.19). and not so much as functions evaluated at a point. Using Example 1. we ﬁnd loga x = d dx ln x ln a = 1 1 1 1 ln x = × = . We stated previously that ln x = 1/x. the logarithm functions for an arbitrary positive base a. and assume that both of them are diﬀerentiable at x. i. a = 1.22) and (1.42. a function of the form f (x) = an xn + an−1 xn−1 + · · · + a1 x + a0 . dx dx (f + g) = f + g or d df dg (f + g) = + dx dx dx Example 1. In this sense ln a loga x = cf (x) where c = 1/ ln a and f (x) = ln x. then f (x) = 20x4 − 6x + 4.28 and the linearity of the derivative we see right away that f (x) = nan xn−1 + (n − 1)an−1 xn−2 + · · · + a1 . let f and g be functions. ♦ ln a ln a x x ln a Suppose f and g are deﬁned and diﬀerentiable on an set.2 Product and Quotient Rules Next we state the product and the quotient rule.. Then the product f g and the quotient f /g are diﬀerentiable at x and their derivatives are given by . Here is a speciﬁc example. Then the diﬀerentiation rules are (1. If f (x) = 4x5 − 3x2 + 4x + 5. we may omit (x) from the notation. ♦ 1.

Find the derivative of the rational function.26) d (f g)(x) = dx d dx f g (x) = df dg (x)g(x) + f (x) (x) dx dx df dx (x)g(x) dg − f (x) dx (x) . then we domain of the function is R\{0}. Putting this into the product formula yields 1 = x(2 ln x + 1). [g(x)]2 (1. BASIC CONCEPTS (1.44. According to the quotient rule r (x) = 2x(x3 + 1) − (x2 − 5)3x2 −x4 + 15x2 + 2x = . the real line with the origin removed. If n ≤ −1. see (1.24) (f g) (x) = f (x)g(x) + f (x)g (x) f g f (x)g(x) − f (x)g (x) . ♦ (x3 + 1)2 (x3 + 1)2 Example 1. Solution: We veriﬁed this formula for n ≥ 0 in Example 1.43.28. Then d n 0 · xm − 1 · mxm−1 d −m 1 = = m+1 = nxn−1 .45.25) (x) = In Leibnitz’ notation these formulas become (1. Solution: Write h(x) = f (x)g(x) with f (x) = x2 and g(x) = ln x.19). Then p (x) = 2x and q (x) = 3x2 . Diﬀerentiate the function h(x) = x2 ln x. The formula d n x = nxn−1 dx for all integer powers n. x = m 2m dx dx x x x . h (x) = f (x)g(x) + f (x)g (x) = 2x ln x + x2 r(x) = x2 − 5 .27) Example 1. x3 + 1 ♦ Solution: We set p(x) = x2 − 5 and q(x) = x3 + 2. Let n be a negative integer and m = −n. Then f (x) = 2x and g (x) = 1/x. x Example 1. [g(x)]2 (1.24 CHAPTER 1.

wherever g(x) = 0. i.e. we may once more omit (x) from the notation. and apply the quotient rule. 2x 2x cos cos cos2 x Some books and computer programs will give this result in a diﬀerent form. The function is deﬁned for all x for which cos x = 0. . We apply the quotient rule. ♦ 2x cos cos x cos x Suppose f and g are deﬁned and diﬀerentiable on an open set.47. Solution: We write the function as a quotient: f (x) = 1/ cos x. and not so much as functions evaluated at a point. Diﬀerentiate the function f (x) = sec x. We ﬁnd (1. given by g2 (x) = [g(x)]2 . Use that sin x = cos x (see Example 1. and that the derivative of a constant vanishes. (1. g2 ♦ and. using that cos x = − sin x (see Exercise 2 on page 17)..28) sin x cos x − sin x cos x cos2 x + sin2 x 1 = = = sec2 x. where n is an integer.31) and cos x = − sin x (see Exercise 2 on page 17). 25 Solution: We express f (x) as a quotient of two functions. RULES OF DIFFERENTIATION Example 1.13. for x not of the form nπ + 1/2.30) = · = tan x sec x. Thinking of f and g again more as functions. That draws our attention to the fact that the function f (x) = tan x satisﬁes the diﬀerential equation f (x) = 1 + f 2 (x). Then the product rule and quotient rule become (1. Example 1. We ﬁnd sin x sin x 1 sec x = (1.1.32) f g f g − fg = g2 or = Here g2 is the square of the function g. they write tan x = (1. Based on the relevant trigonometric identity. f (x) = sin x/ cos x. Find the derivative of f (x) = tan x.31) (f g) = f g + f g or d df dg (f g) = g+f dx dx dx d dx f g df dx g dg − f dx .46.29) tan x = 1 + tan2 x.

so that h(x) = f (g(x)).33) h (x) = (f ◦ g) (x) = f (g(x))g (x). Diﬀerentiate the function h(x) = ex 2 +1 . Combining Example 1. Solution: We write h = f ◦ g as a composition of two functions.50. Let u(x) be a diﬀerentiable function.45 with the chain rule we ﬁnd d n u (x) = nu (x)un−1 (x) dx . ♦ Example 1. ♦ e dx Example 1. then h(x) is diﬀerentiable at x and (1. so that the composition (f ◦ g)(x) = f (g(x)) is deﬁned for all x in the domain of g. with g(x) = x2 + 1 and f (u) = eu .3 Chain Rule Let f and g be functions. BASIC CONCEPTS 1.26 CHAPTER 1.34) dh d df dg (x) = f (g(x)) = (g(x)) (x).48. In the last expression we reversed the order of the factors to make the expression more readable. f (g(x)) = ex +1 . Here are some speciﬁc examples: d 2x+5 = 2 e2x+5 e dx d sin x = cos x esin x e dx d tan x = sec2 x etan x .13.49. The chain rule tells us that h (x) = f (g(x))g (x) = 2xex 2 +1 . In Leibnitz’ notation the chain rule says that (1. Remember that f (u) = f (u) = eu and 2 g (x) = 2x. The chain rule says that whenever g is diﬀerentiable at x and f is diﬀerentiable at g(x). dx dx du dx Example 1. and suppose that the domain of f contains the range of g. In particular. If f (x) = eu(x) then f (x) = u (x)eu(x) . Set h = f ◦ g.

19). envoking the chain rule (1. The chain rule tells us now that h (x) = f (g(x))g (x) = n(g(x))n−1 g (x) = nu (x)un−1 (x). The chain rule tells us that. suppose that u < 0. assuming only that u is diﬀerentiable at x and u(x) = 0 if n ≤ −1..g.13. According to Example 1. f (u) = nun−1 . ♦ du u More generally. here are concrete examples: d (3x + 5)8 dx d 2 (x + 1)25 dx d tan3 x dx d cos2 x dx d sec5 x dx = 8(3x + 5)8−1 · 3 = 24(3x + 5)7 = 25(x2 + 1)24 · 2x = 50x(x2 + 1)24 = 3 sec2 x tan2 x = 2 cos x(− sin x) = −2 cos x sin x = 5 sec4 x sec x tan x = 5 sec5 x tan x. Solution: We asserted that ln u = 1/u for positive values of u.51.35) d 1 ln |u| = . d 2x ln |x2 − 4| = 2 dx x −4 . To be speciﬁc. E. ♦ Example 1.36) d u (x) ln |u(x)| = . Diﬀerentiate the function ln |u| for u = 0. Then h(x) = f (g(x)) = un (x). So.1. We reordered the expressions so that the expression is more readable. for u < 0. d 1 d 1 1 ln |u| = (−u) = (−1) = . see (1. dx u(x) assuming that u is diﬀerentiable and nowhere zero on its domain. RULES OF DIFFERENTIATION 27 for all integers n. Solution: Set g(x) = u(x) and f (u) = un .45. du |u| du −u u This means that for all non-zero u (1. Then u = −|u| and ln |u| = ln(−u).

38) For example: d (sin x)1/2 = dx cos x √ 2 sin x for x ∈ (0. Speciﬁcally. where n is an arbitrary integer. Example 1. Consider a function u which is diﬀerentiable and nowhere zero on its domain and q any real number.49 and the exponential laws. π) and for x ∈ (−π/2.36).28 CHAPTER 1. f (x) = = = = = Here is a concrete example: d 1 − sin x dx 2 5 d ln f (x) e dx d ln(|u(x)|q ) e dx d q ln |u(x)| e dx d (q ln |u(x)|) eq ln |u(x)| dx u (x) q |u(x)|q . BASIC CONCEPTS for all x = ±2. Eventually we will verify them independently. u(x) The assertion follows from (1.37) specializes to (1. We push matters a bit further. we have to exclude all x of the form π 5π 6 + 2nπ and 6 + 2nπ. We use the formulae for diﬀerentiating the exponential and natural logarithm functions. u(x) − cos x 1 =51 − sin x 2 2 − sin x 5 whenever sin x = 1/2. For diﬀerentiable functions which are everywhere positive on their domain and any real number q the diﬀerentiation formula in (1. Then (1. π/2). dx d (sec2 x + 5)π = 2π sec2 x tan x(sec2 x + 5)π−1 dx . d q u (x) = qu (x)uq−1 (x).37) If f (x) = |u(x)|q then f (x) = q u (x) |u(x)|q .

13. These observations motive the attribute ‘hyperbolic’.39) sinh x = 1 x e − e−x 2 & cosh x = 1 x e + e−x 2 You are invited to verify that cosh2 x − sinh2 x = 1. Hint: ax = ex ln a ) dx d x x = (1 + ln x)xx (Assume x > 0. we get the following derivatives: d x a = ax ln a (Assume a > 0. one can show that any point (u.13. RULES OF DIFFERENTIATION Using the tricks from above.. dx x 29 To diﬀerentiate a composition of more than two diﬀerentiable functions we apply the chain rule repeatedly. Conversely. dx dx For example √ d √x2 +1 1 xe x +1 2 = e x +1 · √ e · 2x = √ dx 2 x2 + 1 x2 + 1 √ 2 d tan3 (5x2 − x + 5) = 3 tan2 (5x2 − x + 5) sec2 (5x2 − x + 5) · (10x − 1) dx 1. x = 1. E. sinh x) for some x ∈ (−∞. π/4)).g. Hint: xx = ex ln x ) dx d sin x sin x = x + cos x ln x xsin x (Assume x ∈ (0. v) on the hyperbola u2 − v 2 = 1 can be expressed as (± cosh x. (1. .1. d d f (g(h(x))) = f (g(h(x)) g(h(x)) = f (g(h(x)) · g (h(x)) · h (x). It is elementary to compute the derivatives of the hyperbolic functions: sinh x = cosh x and cosh x = sinh x.4 Hyperbolic Functions The exponential function may be used to deﬁne the hyperbolic sine and cosine. ∞).

the image of I under the map f . For example. 2. and csch x = . ∞) to (−π/2. then it is monotonic (increasing or decreasing) on any interval in its domain. sech x = . then so is g. or you can work them out yourself. Suppose f and g are inverses of each other.13. 3. Some parts of this proposition are elementary. π/2) to the interval (−∞. If I is an open interval. Two functions f and g are said to be inverses of each other (or each function is the inverse of the other one) if the domain of f is equal to the range of g. ∞) to the interval (0. 1. 4.5 Derivatives of Inverse Functions Let us recall. The function sin x maps the interval [−π/2. If f is decreasing. 1]. and (1. There are identities for these hyperbolic functions. Its inverse arcsin x is typically used with . You can ﬁnd them in any table of mathematical formulas. If f is increasing. comparable to the identities for the trigonometric functions. then J is an open interval. others are consequences of the intermediate value theorem. you may calculate the derivatives of these functions. coth x = . A few essential properties of inverse functions are listed in Proposition 1. π/2). 1]. then so is g.30 CHAPTER 1. π/2] to the interval [−1. BASIC CONCEPTS One may also deﬁne other hyperbolic functions tanh x = sinh x cosh x 1 1 . the domain of g is equal to the range of f . the function f (x) = cos x maps the interval [0. If f is continuous. cosh x sinh x cosh x sinh x As a routine application of the rules of diﬀerentiation. The graph of g is obtained from the graph of f by reﬂection at the diagonal. 1. π] to the inteval [−1. The function f (x) = ex maps the interval (−∞. If f is continuous and I is an interval in the domain of f . The function tan x maps the interval (−π/2. then J = f (I).40) g(f (x)) = x and f (g(y)) = y for all x in the domain of f and all y in the domain of g. ∞). ∞). is an interval. It is customary to deﬁne its inverse arctan x as a function from (−∞.52.

b).5 1 1. By deﬁnition we have f (g(y)) = y for all y ∈ (A. Denote the inverse of f by g. B) for which f (g(y)) = 0. f (g(y)) . if the line t(x) is close to the graph of the function f (x) at the point (x. The relation between the derivative of a function and its inverse is spelled out in our next theorem.5 0. Let f be a diﬀerentiable and invertible function which is deﬁned on an open interval (a.5 -0. this is also easy to calculate.9 and 1. Still.5 -1 -1 -1. f (x)) and y = f (x). 1]. and denote the image of f by (A. g(y)). the slope of t(x) is the reciprocal of the slope of T (x).1. Then g is diﬀerentiable at all points y ∈ (A.5 -1 -0.10: arcsin y on [−1. π/2]. You see the graph of these two functions in Figures 1. then its reﬂection T (x) at the diagonal is close to the graph of the function g(x) at the point (f (x). Actually.5 1 -0.9: sin x on [−π/2. and its range is [−π/2.5 -1.10. For these values of y and for x such that f (x) = y the derivative is given by: g (y) = 1 f (g(y)) or g (f (x)) = 1 . RULES OF DIFFERENTIATION 1 1.5 -1 -0. π/2] Figure 1.5 0. We will not give a formally complete proof of the diﬀerentiability assertion. This provides the formula for the derivative.5 31 1 0. We ﬁnd f (g(y))g (y) = 1 and g (y) = 1 . B). With the role of x and y being interchanged. Theorem 1. f (x) Proof.13. 1] domain [−1. and this is assured by the assumption that t(x) is not horizontal. x) = (y.5 0.53. B). Diﬀerentiate both sides of the equation. We need that T (x) is not vertical.5 Figure 1.

If y = f (x). arctan y = dy 1 + y2 According to standard conventions we use (−∞. ∞). The theorem also provides us with the formula for the derivative: arctan (y) = 1 1 = = cos2 (arctan y). π/2) as the range for arctan. Theorem 1. the exponential function is the inverse natural logarithm function ln. .19). Example 1.32 CHAPTER 1. ∞). as asserted in (1. f (x) We apply the theorem to ﬁnd some important derivatives. Show that the function g(y) = arctan y (the inverse of f (x) = tan x) is diﬀerentiable. Assume that the natural logarithm function is diﬀerentiable and that ln x = 1/x. Set f (x) = ln x and g(y) = ey in rem 1. and that d 1 . tan (arctan y) sec2 (arctan y) All we need to do now is to ﬁgure out what cos2 (arctan y) is. Show that the exponential function is diﬀerentiable and that d y e = ey . The theorem says that the exponential function ferentiable and provides the formula for the derivative: d y 1 1 = ey . We note that ln (x) = 0 for all x in (0.11. e = = y) dy 1/ey ln (e as claimed. the domain natural logarithm.53. dy Solution: By deﬁnition. We refer to the notation in Figure 1. and f (x) = sec2 x is nowhere zero. and we obtain the second version of the formula for the derivative of the inverse of the function: 1 g (f (x)) = . ∞) as the domain and (−π/2.53 tells us that g(y) = arctan y is diﬀerentiable on its entire domain (−∞.54. ♦ of the Theoof the is dif- Example 1.55. Solution: The function f (x) = tan x is diﬀerentiable on its entire domain. BASIC CONCEPTS as claimed. then g(y) = g(f (x)) = x. To do this we draw a triangle in which we identify the available data.

13. and the opposing side of length y. by deﬁnition. 1 + y2 1 1+ y2 and cos2 (arctan y) = 1 . tan u = y and arctan y = u. we ﬁnd a slightly more general formula: (1. Combined with the chain rule. the right angle is at the vertex B. 1 + y2 This is exactly what we claimed. The angle at the vertex A is called u. and assuming the diﬀerentiability of u(x). dx 1 + u2 (x) . So. the length of the hypotenuse is Then cos u = The conclusion is that (1. The adjacent side to this angle is chosen to be of length 1. By the theorem of Pythagoras.1.42) d u (x) arctan(u(x)) = .11: An informative triangle There you see a rectangular triangle. RULES OF DIFFERENTIATION 33 y u A 1 B Figure 1.41) arctan (y) = ♦ 1 . 1 + y2.

arccot x. and that its derivative is d 1 .. It is customary to think of arcsin x as a function from [−1.43) For example: d arcsin(3x) = dx d arcsin(x2 ) = dx 3 if x ∈ (−1/3. and suppose that |u(x)| < 1. 1) 1 − x4 √ d arcsin(u(x)) = dx u (x) 1 − u2 (x) .13. 1). I.44) (x2 + y 2 )2 = x2 − y 2 . For example Exercise 3. see Figure 1. Consider the equation (1. 1] to [−π/2. The points on the graph of f are the points which satisfy the equation. Then. we ﬁnd that (1. arcsin x = √ dx 1 − x2 We may once more improve on this formula.34 For example: CHAPTER 1. 1/3) 1 − 9x2 2x √ if x ∈ (−1. 1 + sin2 x The reader is invited to verify the formulas for the other inverse trigonometric functions arcsin x.12. The solutions of this equation form a curve5 in the plane called a lemniscate.3 on page 63. arccos x. 1. π/2].6 Implicit Diﬀerentiation Until now we considered functions which were given explicitly. where f (x) is some instruction which assigns a value to x. Let u(x) be a diﬀerentiable function which is deﬁned on an open interval. BASIC CONCEPTS d arctan(x2 + 5) = dx d arctan(sin x) = dx 2x 1 + (x2 + 5)2 cos x . and arcsec x as they are given in Table 1. using the chain rule.e. Show that arcsin x is diﬀerentiable on (−1. Parts of this curve look like the graph of a function. such . we were given an equation y = f (x).

Example 1. 3 We used a diﬀerent way to indicate at which point we evaluate the derivative because we had to specify the x and the y coordinate of the point.2 -0. we still like to calculate the slope of curve at one of its points. 3/2) −1 =√ .5 -0. The unit circle consists of all points which satisfy the equation x2 + y 2 = 1. This process is called implicit diﬀerentiation. Solution: We write y = y(x) to emphasize that y as a function of x. we ﬁnd that dy dx √ (1/2.13.12: Lemniscate as the points for which y ≥ 0. ♦ 5 We will rely on the readers intuitive idea of a curve in the plane.1 -0.1 -1 -0.56. Let us start out with an example which we have studied before. RULES OF DIFFERENTIATION 35 0. Diﬀerentiating both sides of the equation of the circle we get 2x + 2y dy = 0 or dx dy −x = . Find the slope of the tangent line to the unit circle at the √ point (1/2. Without solving the equation for y.5 1 Figure 1.3 0. .3 0.1. 3/2).2 0. dx y Plugging in the coordinates of the speciﬁed point.

the point (x. y) with y = 0 on the lemniscate. dx dy dx Finally we get an explicit expression for in terms of x and y: dy x(1 − 2(x2 + y 2 )) x(1 − 2(x2 + y 2 )) = = . BASIC CONCEPTS Example 1. we consider y as a function of x and diﬀerentiate both sides of the equation. As in Example 1. dx dx Bring all terms with a factor dy/dx to the left hand side of the equation and those without to the right hand side.12.g. dx 3 −3 + 2 3 This speciﬁc calculation takes a bit of arithmetic skill and eﬀort to carry out. A quick look at Figure 1. 8 .57. √ E. 2 2 and at this point the slope of the tangent line is √ dy 2− 2 =√ √ . we can plug it into the dy expression for dx and we get the slope of the curve at this point. and ﬁnd the coordinates of the points where the tangent line is horizontal. (2y(x2 + y 2 ) + y) dy = x(1 − 2(x2 + y 2 )). We ﬁnd 2(x2 + y 2 )(2x + 2y dy dy ) = 2x − 2y . 2 2 Then we get an equation in one variable: 1 = x2 − 4 1 − x2 2 or x2 = 3 8 and 1 y2 = . 1 −3 + 2 3) is a point on the lemniscate.56. and y 2 = 1 − x2 into the equation of the curve. Find the slope of the tangent line to the lemniscate (x2 + y 2 )2 = x2 − y 2 .12 tells us that we may ignore points where x = 0 or y = 0. y) = ( 1 . dy The tangent line is horizontal whenever dx = 0. 2 Substitute x2 + y 2 = 1 . That dy means that dx = 0 whenever 1 − 2(x2 + y 2 ) = 0 or 1 x2 + y 2 = .36 CHAPTER 1. dx 2y(x2 + y 2 ) + y y(2(x2 + y 2 ) + 1) Given any point (x. Solution: You see a picture of the lemniscate in Figure 1.

that the tangent line to the parabola and the circle will be the same at the point of contact.5 1 Figure 1. we get 2x + 2(y − a) dy = 0.5 -1 -0. drop a ball of radius 1 into a cup whose vertical cross section is the parabola y = 2x2 . Suppose you drop a circle of radius 1 into a parabola with the equation y = 2x2 .± ) ≈ (±. Suppose the coordinates of the center of the circle are (0.3536). . then its equation is x2 + (y − a)2 = 1. a). Diﬀerentiating the equation of the circle with respect to x. ♦ 4 4 37 Example 1.13: Ball in a Cup. we dy substitute dx = 4x into the second equation.5 2 1. ±.5 0. dx dy Assuming that dx is the same for both curves at the point of contact. At which points will the circle touch the parabola?6 3 2. RULES OF DIFFERENTIATION The points at which the tangent line to the lemniscate is horizontal are √ √ 6 2 (x.1. Diﬀerentiating the equation of the parabola dy with respect to x.13.5 1 0.6124. Solution: You see a picture of the problem in Figure 1. The crucial observation in this example is.58. After some impliﬁcations we 6 More sensibly.13. y) = (± . we ﬁnd that dx = 4x.

0). we ﬁnd that the y coordinate of the point of contact is y = a − 1 . ♦ 4 8 Exercise 4. the 8 circle touches the parabola in the points √ 15 15 (x. The ball it too large to ﬁt into the parabola and touch at (0. −1).14. Example 1. Exercise 5. Consider the curve given by the equation x2 = sin y. You ﬁnd a picture of this Lissajous ﬁgure in Figure 1. This is the rate at dt dt which the volume of the ball changes with respect to time. Exercise 6. In summary. We substitute 4 this expression into the equation √ the circle and ﬁnd that the x coordinate of 15 of the point of contact is x = ± 4 .59. Consider the curve given by the equation x3 + y 3 = 1 + 3xy 2 . ♦ . Find the √ slope of the curve at the point with coordinates x = 1/ 4 2 and y = π/4. Repeat Example 1. . you know at which rate one of them changes. BASIC CONCEPTS x(1 + 4(y − a)) = 0. Substituting this into the equation of the parabola. dt dr dt dt With r = 20 and dr = 2 we get dV = 3200π cm3 /min. According the 3 the chain rule: dV dr dV dr = = 4πr 2 . Suppose the radius of a ball changes at a rate of 2 cm/min. Solving the equation 1 + 4(y − a) = 0 for y. y) = (2. In this section we treat such problems. So we may assume that x = 0. Find the slope of the curve at the point (x. The formula for the volume of a ball is V (r) = 4π r 3 .57 with the curve given by the equation y 2 −x2 (1−x2 ) = 0. We use t to denote the time variable. 1. We consider V as a function of r as well as t.14 Related Rates Many times you encounter situations in which you have two related variables. and you like to know at which rate the other one changes. y) = ± . we ﬁnd that y = 15 at the point of contact. At which rate does its volume change when r = 20 cm? Solution: Denote the volume of the ball by V and its radius by r.38 ﬁnd: CHAPTER 1.

4 = C.14: y 2 − x2 (1 − x2 ) = 0 Example 1. as functions of the time variable t. y = 5 3. and dy = −3.4 for air at room temperature. For the purpose of this problem. x and y. At which rate does it move in the horizontal direction? Solution: The equation of the circle is x2 + y 2 = 100. At some time the particle √ is at the point (5. They derived the equation P V γ = C. 5 3) and moves downwards at a rate of 3 cm/min. It is called the adiabatic law.4 0. dt dt √ √ In the given situation x = 5. Suppose a particle moves on a circle of radius 10 cm and centered at the origin (0. Implicit diﬀerentiation of the equation of the circle gives us the equation dx dy + 2y = 0.5 1 -0. 0) in the Cartesian plane.2 -1 -0. ♦ 2x Example 1. Boyle-Mariotte described the relation between the pressure and volume of a gas.60. Pressure (P ) and volume (V ) of air at room temperature are related by the equation7 P V 1. RELATED RATES 39 0. We ﬁnd that dx = 3 3.5 -0.61. we suppose that γ = 1.14. We consider both variables. dt dt √ so that the particle is moving to the right at a rate of 3 3 cm/min. 7 .1. The constant γ depends on the molecular structure of the gas and the temperature.4 Figure 1.2 0.

BASIC CONCEPTS Here C is a constant. At which rate is the mass changing when the particle’s velocity is 90% of the speed of light.4P V . ♦ Example 1. and increasing at .001c per second? Solution: According to our rules of diﬀerentiation dM mvc . Right now the foot of the latter is 1 m away from the wall. At which rate is the top of the ladder sliding down the wall? . dt dv dt 3610 1000(c2 − v 2 )3/2 The perceived mass increases at a rate of approximately 1% of its mass at rest. A ladder. 7 m long. Solution: We consider P as a function of V .4P =− . = 2 dv (c − v 2 )3/2 Applying the chain rule and substituting the values.4P dV = =− .40 CHAPTER 1. Find the rate of change of P if the volume increases at a rate of 10 cm3 /min. Diﬀerentiation of the equation yields dP 1.1 m/sec. ♦ Exercise 7. is M (v) = m 1 − v 2 /c2 .4 = 0 or dV According to the chain rule dP dP dV 1. as perceived by an observer in resting position. is leaning against a wall.010867m.62. dt dV dt V dt Substituting the given information we ﬁnd that the pressure decreases at a rate of 1. You are pulling the foot of the ladder further away from the wall at a rate of . At some instant t0 the pressure of the gas is 25 kg/cm2 and the volume is 200 cm3 . This formula is from Einstein’s special theory of relativity. The mass M of a particle at velocity v.75 kg/cm2 sec. we ﬁnd √ dM 9 19m dM dv mvc2 = = = ≈ .4 V + 1. dV V where m is that mass at rest and c is the speed of light. dP 1.

Find the relative growth rate a of the population. 9 Note that the doubling time depends only on the relative growth rate a.9 on page 68. Find the time within which the population doubles9 . 1. Then we have an initial value problem (1. Suppose the size of a population of bacteria in a laboratory experiment is C1 = 5.15. 000. Find the formula for the size of the population at any time t ≥ 0. an equation which involves a function and its derivatives. EXPONENTIAL GROWTH AND DECAY 41 1.45) is an example of a diﬀerential equation. and it can be shown that on an interval any solution is of this form.45) A (t) = aA(t). The equation in (1. We may specify the value of A at some time t0 . We denoted the proportionality factor by a. We saw that the functions A(t) = Ceat are solutions of this equation. The essential aspects of dealing with (1. 5. but very useful model for population growth is the Malthusian Law (1. The uniqueness assertion follows from Proposition 2. Predict the size of the population at time t = 10. 000 at time t1 = 2 and C2 = 7. 8 .15 Exponential Growth and Decay An idealistic. 4. It says that the rate of change of a population is proportional to its size. We also say that A(t) grows exponentially and a is the relative growth rate.64.1. 2. Theorem 1.46) are addressed in Example 1. That the function satisﬁes the diﬀerential equation follows from (1. which we still need to prove. Find the time at which the population reaches 8. 000 at time t2 = 5.46) A (t) = aA(t) and A0 = A(t0 ). say A0 = A(t0 ). 3.46).19). Here time is measured in hours since the beginning of the experiment.63. and the unknown is a function. On an interval which contains t0 the function A(t) = A0 ea(t−t0 ) is the unique solution8 of the initial value problem in (1.

one ﬁnds (1. Suppose the size of the population reaches 8. Consider a radioactive substance. The population grows at a rate of about 11% per hour. t2 − t1 3 and ln C2 C1 = a(t2 − t1 ). where a is as above. The half-life T of a radioactive substance is the time within which half of it decays. 000 about 6. 000 at time t1 . BASIC CONCEPTS Solution: We denote the size of the population at time t by A(t).48) T = ln 2 . and 3.42 CHAPTER 1.11. The minus sign in the equation is included so that k will be positive. where A0 = A(t0 ). Suppose at some time t0 the size of the population is A0 = A(t0 ) and T hours later the size of the population is 2A0 = A(t0 + T ). 4. The theorem tells us that A(t) = A0 ea(t−t0 ) . 2. Substituting t = 10 we ﬁnd that A(10) ≈ 12.4 = ≈ . k . Suppose t denotes time and A(t) the amount of radioactive substance at time t.6) + 2 ≈ 6. The experience which we just described can be expressed as a diﬀerential equation (1. 1.47) A (t) = −kA(t). ≈ 6. To calculate the relative growth rate a observe that C2 A0 ea(t2 −t0 ) = = ea(t2 −t1 ) C1 A0 ea(t1 −t0 ) We ﬁnd that a= ln C2 − ln C1 ln 1. As in the computation of the doubling time in the previous example. Experiments have shown that the rate at which radioactive decays occur is proportional to the amount of radioactive material present. This rate is proportional to the rate at which the amount of the material decreases.2 a The size of the population reaches 8.18 hours. 264. The size of the population at any time t ≥ 0 is A(t) = 5000ea(t−2) .2 hours into the experiment. Then A(t0 + T ) = A0 eaT = 2A0 Thus the doubling time is T = ln 2 a or eaT = 2 and aT = ln 2. then 8000 = 5000ea(t1 −2) or ln(8/5) = a(t1 − 2) or t1 = ln(1. 5.

ln 2 The tree died approximately 10.g. and the carbon-14 which was present at the time of death decays. How long ago did the tree die? Solution: Let t0 = 0 be the time of death of the tree. Suppose we measure 6.8 6. the idea is as follows.49) ln 1.1. 500 years ago. MORE EXPONENTIAL GROWTH AND DECAY 43 In the late 1940ies Willard Libby invented (or discovered) the method of carbon-14 dating. approximately. We explain the process in a numerical example. From this we calculate: (1. Example 1. 534. Suppose dead wood of the same kind shows 1. wood from an oak or a human bone). In brief.50) f (t) = af (t) + b.68 carbon-14 decays per minute and gram in a certain kind of wood at the time of death of the tree. The number of decays to be expected t years after death is A(t) = 6. or a part thereof (e. and a = 0. He was awarded the Nobel price for it. A time independent solution (steady state solution) of this equation is f (t) = −b/a. and the amount is believed to have been essentially constant for a long time (until recent nuclear testing).8. The level of the equilibrium is characteristic for the organism.68e− 5568 t .65. All living organisms absorb it. For many organisms one also knows how many carbon-14 decays to expect at the time of death. where a and b are constants.68 ≈ 10.45). We have that A(t1 ) = 1.68 5568 or t1 = − 5568 ln ln 2 1. After death no more carbon-14 is absorbed. The amount which is absorbed equals the amount which decays. measured in years.16.8 ln 2 =− t1 6. consider the diﬀerential equation (1. Carbon-14 occurs naturally in the atmosphere. Measuring the number of decays in a dead organism allows us to determine the time of death. and t1 the present time. Within a living organism there is an equilibrium. 1.8 decays per minute and gram.16 More Exponential Growth and Decay More generally than in (1. . The half-life of carbon-14 has been determined to be about 5568 years.

Interest is added at a rate of .005t − −200 = −25. The rest will be routine calculation. Here c denotes an arbitrary constant. Theorem 1.9 on page 68.5% per month. Solution: As variable we use time.005 e. On an interval every solution of (1.005B(t) due to interest being added and decreases at a rate of $200. This is the time at which you start to repay the loan. 000. In summary. The balance increases at a rate of . Remark 2. denoted by t and measured in months.00 per month due to payments which you make.000. and you are repaying the loan at a rate of $ 200.00 per month. BASIC CONCEPTS Theorem 1. Analyze the future of the loan.50) is of this form. According to Theorem 1. On an interval which contains to .50). Example 1. We obtain a unique solution if we add an initial condition to the diﬀerential equation in (1.005t + 40.005 . The important aspects are to translate the given information into a mathematical equation. we have the initial value problem B (t) = .50). . Let us apply these ideas to solve some problems.67. We set t = 0 at the time of graduation. The uniqueness assertion is a minor modiﬁcation of Proposition 2. Functions of the form f (t) = ceat − b a are solutions of the diﬀerential equation in (1. 000.66. the function f (t) = y0 + b a ea(t−t0 ) − b a is the unique solution of the initial value problem f (t) = af (t) + b and f (t0 ) = y0 . On graduation day the balance of your student loan is $15. Denote the balance of your loan at time t by B(t). It is not hard to verify that the given functions are solutions of the respective problems.67 the solution of the initial value problem is B(t) = 15.005B(t) − 200 and B(0) = 15.68. 000e.44 CHAPTER 1. 000 + −200 .

800.005 40 25 ≈ 94. For example. so that you paid the principal plus $3. Think of a cup of coﬀee. ♦ Example 1. Let A(t) denote the amount of medication in your body. measured in milligrams. Analyze the amount of medication in your body at any time. It will take slightly more than 40 hours before the amount of medication in your body reaches 60 milligrams. We have the initial value problem A (t) = −.04A(t) + 3 and The solution of this problem is A(t) = −75e−. .69. The steady state solution of the problem is A(t) = 75.16.04A(t) due to your liver metabolizing the medication.70 (Newton’s Law of Cooling).6 milligram of medication in your body. denote it by t and measure it in hours. the temperature of the object will approach the one of its surroundings. Your total payments were $18.1. Solution: We use time as independent variable. ♦ Example 1.800 in interest.04t + 75. after 12 hours there will be about 28. at least under idealized circumstances. Then A(t) increases at a rate of 3 mg per hour because you are taking in medication and at the same time A(t) decreases at a rate of . We discuss how this happens. 45 After approximately 94 months (7 years and 10 months) you repaid the loan. With time. You stir the coﬀee gently so that the temperature in the cup remains homogeneous and almost no energy is added A(0) = 0. (You can keep this rate constant with a skin patch.) The liver metabolizes the medication at a rate of 4% per hour. MORE EXPONENTIAL GROWTH AND DECAY For example. We denote by t = 0 the time when we start taking the medication. Suppose you have an object whose temperature is diﬀerent from the temperature of its surroundings. You are absorbing a medication at a rate of 3 mg per hour. The amount of medication will stabilize at this amount with time. B(T ) = 0 if T = 1 ln .

the outside warms up much faster than the inside. its temperature is 95 degrees Celsius. It is a function of time. Determine the function T (t). All of this leads to a signiﬁcantly diﬀerent development of the temperature inside a turkey as you roast it for your Thanksgiving dinner. protein. just after you poured the coﬀee into your cup. a in (1. If K is the temperature of the surroundings. That means. then (1. To determine a we use that T (5) = 80 = 70e5a + 25.51) T (t) = a(T (t) − K) = aT (t) − aK.67. such as a turkey in the oven. Having a numerical value for a gives us an explicit expression for the temperature T as a function of t: T (t) = 70e−. while you stir it slightly and patiently. we get that T (t) = (95 − 25)eat + 25 = 70eat + 25. At time t = 0. y0 = 95. Note that −b/a = K.10 Denote the temperature of the coﬀee by T . the speciﬁc heat (the amount of energy needed to increase the temperature of one unit of the material by one degree) varies. Newton’s law of cooling says that the rate at which the heat is transferred. Furthermore. Solution: To apply Theorem 1. 2. 1.50).46 CHAPTER 1. and bone. Five minutes later the temperature has dropped to 80 degree. It is diﬀerent for fat. Putting all of this into the formula for the solution of the initial value problem. the speciﬁc heat is highly temperature dependent for substances like protein. such that T (t1 ) = 70 degrees Celsius. The room temperature is 25 degrees. Find t1 .0482. Using these data. The physics of heat transfer changes substantially if you take a solid object.51) says that the temperature of the coﬀee drops at a rate of about . and with this the rate of change of temperature of the coﬀee.0482t + 25. Equa5 70 tion (1.048 degrees per minute for each degree of diﬀerence between the temperature of the coﬀee and the room temperature. this is the diﬀerential equation in (1. and K = 25. Setting b = −aK.51) depends on the temperature T . In addition. and we conclude that a = 1 ln 55 ≈ −. we set t0 = 0. 10 . BASIC CONCEPTS through the process of stirring. The temperature in the solid object will not be homogenous. so that we write T (t). is proportional to the temperature diﬀerence. Let us work out a numerical example.

17 minutes after pouring it. 3. Currently. Denote the size of the population at time t by P (t). and the amount of water in the lake is 10 times the amount of water carried by the river per year. we ﬁnd that t1 ≈ 9. What is the long term estimate for the population of this species in the wild? .0482t1 + 25 = 70.5 mg per m3 of a pollutant in the lake. State the initial value problem for P (t). 2. the population is estimated to be 700 birds. the owner has agreed to an acceptable level of 2. As a remedy. State the initial value problem for P (t). Let t0 = 0 be the time just after the accident and at which the clean-up strategy is implemented. In negotiations which the EPA. the factory owner proposes to reduce the emission of pollution so that the level of pollutant in the river is only 1. Let P (t) denote the amount of pollutant (measured in mg per m3 ) in the lake at time t. where t denotes time and is measured in years.5 mg per m3 ? Exercise 9. Find the function P (t).17. ♦ Exercise 8. 2. 1. That means that the temperature drops to 70 degrees approximately 9. At which time will the level of pollution be back to 2. and the river is the only contributor to the lake. at time t0 = 0. Assume that the amount of water carried by the river is the same all year around. A chemical factory is located on the banks of a river.16.5 mg per m3 .1. A government agency raises the species in captivity and releases birds into the wild at a rate of 80 birds per year. After a major accident the level has risen to 15 mg per m3 . Find the function P (t). The population of an endangered species of birds on Kauai decreases at a relative rate of 25% per year. 47 Solving the equation for t1 . Down stream from the factory is a lake. At which time will the population drop to 500 birds? 4. 1. MORE EXPONENTIAL GROWTH AND DECAY We like to ﬁnd out the time t1 for which T (t1 ) = 70e−. It is assumed that the pollutant is distributed uniformly in the lake at any time. 3.

48 CHAPTER 1. Leibnitz’s notation for the second derivative of a function f (x) is d2 f /dx2 . BASIC CONCEPTS 1. The derivative of the second derivative is called the third derivative. is called the second derivative of f . that for a diﬀerentiable function we do not make a large error when we use the tangent line to the graph instead of the graph itself. We will make use of the second derivative.18 Numerical Methods In this section we introduce some methods for numerical computations. and denoted by f (x).52) f (x1 ) ≈ f (x0 ) + f (x0 )(x1 − x0 ).1 Approximation by Diﬀerentials Suppose x0 is an interior point of the domain of a function f (x) and f (x) is diﬀerentiable at x0 . Its derivative. . It is denoted by f (x). This process can be iterated. We use the symbol ‘≈’ to stand for ‘is approximately’. 2 dx dx Exercise 10. The method of approximation by diﬀerentials provides an approximate values f (x1 ) if x1 is near x0 . One uses the formula (1.18. 1. 1. This rather casual statement will become clearer when you look at the individual methods. Find the second derivatives of the following functions: (1) f (x) = 3x3 + 5x2 (2) g(x) = sin 5x (3) h(x) = x2 + 2 (4) i(x) = e5x (5) j(x) = tan x (6) k(x) = cos(x2 ) (7) l(x) = ln 2x (8) m(x) = ln(x2 + 3) (9) n(x) = arctan 3x (10) o(x) = sec(x3 ) (11) p(x) = ln2 (x + 4) (12) q(x) = ecos x (13) r(x) = ln(tan x) (14) s(x) = ex 2 −1 (15) t(x) = sin3 x. Here is a sample computation in which you are invited to ﬁll in the details: d2 sin x d e = cos xesin x = (− sin x + cos2 x)esin x . then f (x) is again a function with the same domain as f (x). Their common feature is. etc. We may ask whether the function f (x) is differentiable.17 The Second and Higher Derivatives Let f (x) be a function which is deﬁned on an open set. If the function is diﬀerentiable at each point of its domain. Assume also that f (x0 ) and f (x0 ) are known. wherever it exists.

52) we have l(x1 ).52) insofar as we have not estimated (provided an upper bound for) the error which we make using the right hand side of (1. Exercise 11. so we are supposed to ﬁnd f (9). √ Solution: We set f (x) = 3 x.1.6.72. f (x0 )) evaluated at x1 . applied with x1 = (π/4 + π/180) and x0 = π/4 says tan 46◦ = tan π π π π + ≈ tan + sec2 4 180 4 4 π π =1+ ≈ 1. Solution: We carry out the calculation in radial measure. the tangent line to the graph of f (x) at (x0 .52).0349. such that |f (x1 ) − [f (x0 ) + f (x0 )(x1 − x0 )| ≤ A(x1 − x0 )2 whenever |x1 − x0 | < d. f (π/4) = 1. According to this slightly more demanding deﬁnition. Note that f (x) = 1 −2/3 . Use the function 46 f (x) = tan x. diﬀerentiability of the function f (x) means that there exist numbers A and d > 0. Example 1. In the sense of the deﬁnition of the tangent line in Section 1. says that √ 3 9 = f (9) ≈ 2 + √ 3 1 25 (9 − 8) = ≈ 2. 12 Formula (1. . √ Example 1. In each case. Find an approximate value for 3 9.52). then we can approximate the error as long as |x1 − x0 | < d. f (x1 ) is close to l(x1 ) for x1 near x0 . Then f (x) = sec2 x.0801. Formula (1.25 on page 12 provides us with an estimate. and f (π/4) = 2.52) instead of of the actual value of the function on the left hand side. applied with x1 = 9 and x0 = 8. and this corresponds to π/4 + π/180. Thus. and f (8) = 1 . 180 90 ♦ Your calculator will give you tan 46◦ ≈ 1.0833. Note that ◦ = 45◦ + 1◦ . if we know A and d.71. Use approximation by diﬀerentials to ﬁnd approximate values for √ 5 (1) 34 (2) tan 31◦ (3) ln 1.0355. NUMERICAL METHODS 49 On the right hand side in (1. Find an approximate value for tan 46◦ . x 3 f (8) = 2.18. We have been causal in (1.1.2 (4) arctan 1. The inequality in Deﬁnition 1. compare your answer with one found on your calculator. 12 12 ♦ Your calculator will give you 9 ≈ 2.

000305.515115. Use approximation by diﬀerentials to ﬁnd approximate values for √ (1) cos 28◦ (2) 26 (3) sin 47◦ . 2( x0 ) 54 The actual error is again substantially less than this. √ √ Solution: We use f (x) = x and x0 = 9. and f (x0 ) = 1/6.16). Find an approximate value for sin 31◦ and estimate the error. The f (x) = 1/(2 x).73. From the computation in Example 1. 6 180 6 2 180 The calculator will tell that sin 31◦ ≈ . Use approximation by diﬀerentials to ﬁnd an approximate √ value of 10 and give an upper bound for the error. see (1. In each case. . f (x0 ) = 3. Then f (x) = cos x. Solution: Set f (x) = sin x. The estimate assures us that the error is at most 1 1 √ 3 (x1 − x0 )2 = . ♦ Exercise 12. we ﬁnd sin 31◦ ≈ sin √ π π π π 1 + cos = 1+ 3 ≈ . 6 √ The calculator will give you 10 ≈ 3. Applying the formula in (1.515038. and √ f (π/6) = 3/2.52) tells us that √ 1 10 = f (10) ≈ f (9) + f (9)(10 − 9) = 3 + ≈ 3.31 on page 17 we also know that we may use A = 1 and d = π/4 in the diﬀerentiability estimate.52).16228.16666. BASIC CONCEPTS Example 1. ♦ Comparison of the actual and approximate value conﬁrm this. estimate also the maximal error which you may have made by using the method of approximation by diﬀerentials. The formula in (1. For the error estimate we may use A= 1 √ 3 2( x0 ) and any d > 0. f (π/6) = 1/2. The estimate assures us that the error is at most (x1 − x0 )2 = π 180 2 ≤ .50 CHAPTER 1. Measuring angles in radians we set x0 = π/6 and x1 = π/6 + π/180. Example 1.74.

ﬁnding the zeros of functions of degree 1 and 2.18.54) xn+1 = xn − f (xn ) . and that x0 is not far from x. f (xn ) Geometry of Newton’s Method: Let us give a geometric explanation for the formlas. For polynomials of degree greater or equal to 5 and most other functions there are no general methods for ﬁnding their roots. we ﬁnd the zero of the tangent line..1.18. More sophisticated methods allow you to ﬁnd the exact solutions of polynomial equations of degree three and four. is the point at which l(x) intersects the x-axis.e. NUMERICAL METHODS 51 1. and (1. and l(x1 ) = 0 if x 1 = x0 − f (x0 ) . we want to ﬁnd some x.55) xn+1 = xn − f (xn ) x2 − A x2 + A 1 = n = = xn − n f (xn ) 2xn 2xn 2 xn + A xn If we use A = 3 and x0 = 2. then we ﬁnd x1 = 1 2 2+ 3 2 7 = .. 10864 We summarize the computation in Table 1. The process is then iterated.e. Newton’s method works as follows. l(x) = f (x0 )(x − x0 ) + f (x0 ). such that f (x) = 0. i. Suppose that by some means we know that such an x exists. In the following two columns you ﬁnd the values of xn . f (x0 ) x 2 = x1 − f (x1 ) . we obtain the tangent line l(x) to the graph of f at this point. etc.53) x1 = x0 − f (x0 ) . 4 x2 = 1 2 7 12 + 4 7 = 97 56 and x3 = 18817 .53). Suppose we want to ﬁnd a zero of a diﬀerentiable function f (x). Then f (x) = 2x. i.e. and instead of ﬁnding the zero of the function itself. f (x2 ) and in general (1. as given in (1.1. Then we set (1. You have learned how to solve linear and quadratic equations. f (x1 ) x 3 = x2 − f (x2 ) . Speciﬁcally. i.2 Newton’s Method Newton’s method is designed to ﬁnd the zeros of a function. once . f (x0 ) This means that we accept that the tangent line is close to the graph of the function. Then x1 . Given any x0 at which f is deﬁned and diﬀerentiable.. the positive root of the function f (x) = x2 − A. √ Let us calculate A. In the ﬁrst column you ﬁnd the subscript n.

In the last column you see the square of xn . Let us consider one more example to illustrate Newton’s method. we guess that x0 = 1 is not . At least x2 is rather close to 3. has a root in the interval (0. as a continuous funtion. then the accuracy of this approximation of 3 will exceed the accuracy of most calculators.0000000000 1. once in decimal form. we may say.0625000000 1.55) xn+1 = 1 2 xn + A xn to ﬁnd good approximations of square roots.7320508100 3. You see that our value for x3 is rather precise. Your calculator 3 √ will give you 1. The intermediate value theorem tells us that f (x). In fact. Step 2: Let us come up with a ﬁrst guess for a root. BASIC CONCEPTS expressed as a fraction of integers.0000000085 Table 1.0000000000 4. if you carry the calculation one √ step further and ﬁnd x4 . Find a solution of the equation x sin x = cos x. Considering the values of f at the end points of the interval.73205080757 as an approximate value of 3. x2 n n xn 0 1 2 3 2 7/4 97/56 18817/10864 xn 2. Step 1: Let us make sure that there is a root of the function to be found. We refer to the described procedure as the Babylonian method. The numbers in the last column show that we are making rapid progress in ﬁnding a good approximation of √ 3. Equivalently.7500000000 3. expressed as rational numbers. ﬁnd a root of the function f (x) = x sin x − cos x.7321428571 3.52 CHAPTER 1.1: The Babylonian Method More than 4000 years ago the Babylonians used the outermost expressions in (1. Observe that f (0) = −1 < 0 and f (π/2) = π/2 > 0.0003188775 1. π/2). Let us call this root x.

9) > 0. then we see that |x − x2 | < . Actually f (1) ≈ .18.8) < 0 and f (.04. f (x0 ) Your calculator will tell you that f (x1 ) ≈ .2. and how fast? For completeness sake.0016 and |x − x3 | < .6. There is one feature of Newton’s method which helps. Still. Observe that f (. Let us set a = . Do the xn tend (converge) to x. by numerical means. In general. . Suppose that xn for n ≥ 1 are computed according to (1. an approximate solution of the following kind of problem: . so that I ⊂ J = [. the principle problem is as follows. Suppose xn ∈ I and |x − xn | < 2m . Step 3: Let us improve the guess: Set x 1 = x0 − f (x0 ) ≈ . As ﬁrst guess we used x0 = 1. The distance between x and xn will decrease rapidly as n increases. Consider an interval I = [x − a. You see that f (x1 ) is much closer to zero than f (x0 ). x3 . The quoted theorem asserts that |x − x1 | < . and with this also on I.1].8645. On J. 1. You are invited to M verify these estimates using technology.’ A full discussion of Newton’s method requires mathematical tools which are not available to us at this time. You may say that with each iteration you make a fresh start. We explained Newton’s method because we want to illustrate the power of the concept ‘tangent line. . we have that |f (x)| ≥ m = 1. We made progress ﬁnding x. . A theorem from advanced calculus asserts that M M (x − xn )2 . we give an answer.2. NUMERICAL METHODS 53 too far away from the root. which we know to exist by Step 1.9].1. If we repeat the process. and in this sense we expect that x1 is much closer to the root x of f (x) than x0 .56) we use that 2m < 1. . and suppose that |f (x)| ≥ m and |f (x)| ≤ M on I.8.00874. Step 4: Repeat Step 3 and calculate x2 .18. (1. and in this sense previous round-oﬀ errors don’t carry over. .3 Euler’s Method Euler’s method is designed to ﬁnd. x + a]. 2m We illustrate the theorem by applying it to the previous example.3. This illustrates that the xn approach x rapidly.5. Suppose that f (x) is a diﬀerentiable function and f (x) = 0 and x0 is given.5 and |f (x)| ≤ M = 2. many interesting phenomena can occur. so that we know that |x − x0 | < .00000256.54). This tells us that x ∈ [.56) |x − xn+1 | ≤ 1. In (1.

which depends on t. y) dt and y(t0 ) = y0 . y0 )(T − t0 ). Here F (t. The ﬁrst. The Logistic Law The diﬀerential equation in our next example is known as the logistic law of population growth. the problem in (1.45)).54 CHAPTER 1. y(t0 )) = F (t0 . BASIC CONCEPTS Problem 1.57). In the diﬀerential equation.57) is a ﬁrst order diﬀerential equation. The constants a and b are called the vital coeﬃcients of the population. t denotes time and y(t) the size of a population. . The equation was ﬁrst used in population studies by the Dutch mathematician-biologist Verhulst in 1837. y0 ). For short. and the unknown is the function.58) y(T ) ≈ y(t0 ) + y (t0 )(T − t0 ) = y0 + F (t0 . To get the second equality in (1. y0 ) is l(t) = y(t0 ) + y (t0 )(t − t0 ).57) y = dy = F (t.58) expresses the philosophy that the graph of a diﬀerentiable function is close to its tangent line. The tangent line to the graph of y at (t0 . the term ay expresses that population growth is proportional to the size of the population. Then you might try the formula (1.58) is just l(T ). so that it is assumed that population growth is reduced by a term which is proportional to y 2 . at least as long as T is close to t0 . It is an equation which involves a function and its derivative. which tell us that y (t0 ) = F (t0 . y) denotes a given function in two variables.58) we use the diﬀerential equation and initial condition in (1. The ﬁrst condition on y in (1. Find a function y(t) which satisﬁes (1. In the equation. and t0 and y0 are given numbers. so that the middle term in (1.57) is called an initial value problem. The probability of this happening is proportional to y 2 . The second condition is called an initial condition. The equation reﬁnes the Malthusian law for population growth (see (1. approximate equality in (1. It speciﬁes the value of the function at one point. Approach in one step: Suppose you want to ﬁnd y(T ) for some T = t0 . the members of the population meet and compete for food and living space. In addition.

y(10) = 538. An exact solution of the initial value problem in (1. ♦ . According to the formula in (1.61) 2 y(T ) ≈ y0 + (ay0 − by0 )(T − t0 ). Remark 3. Consider the initial value problem: (1. you see that the diﬀerential equation in this example is a special case of the one in (1.61). we ﬁnd that y(t) = 3000 .59). We are providing the exact solution.60). According to the exact solution in (1. though you are invited to verify that it satisﬁes (1. Example 1. So.60) by0 + (a − by0 )e−a(t−t0 ) This is not the time to derive this exact solution. ♦ Let us be even more speciﬁc and give a numerical example. For T = 10 the formula suggests that y(10) ≈ 510.59) is given by the equation ay0 y(t) = (1. Solution: Substituting a = 1/10. so that we can see how well our approximate values match it.62) dy 1 1 = y− y2 dt 10 10000 and y(0) = 300.4. this number is rounded oﬀ.76. y) = ay − by 2 . t0 = 0.1. Solution: Setting F (t.61) gives us a less satisfactory result.58) we ﬁnd (1. Find an approximate value for y(T ). NUMERICAL METHODS Example 1. Consider the initial value problem. According to the exact solution for this initial value problem.18. (1. b = 1/10000. For this larger value of T .57). the formula in (1.1. we ﬁnd the exact value y(1) = 321. our approximate value is close.59) dy = ay − by 2 dt and y(t0 ) = y0 . Find approximate values for y(1) and y(10). we ﬁnd that y(1) ≈ 300 + 300 3002 − 10 10000 (1 − 0) = 321. We expect a close approximation only for T close to t0 . 3 + 7e−t/10 Substituting t = 1. 55 where a and b are given constants. and y0 = 300 into the solution in (1.75.

Apply the multi-step method to ﬁnd approximate values for y(t) at t1 = 5. We use t1 and y(t1 ) to calculate an approximate value for y(t2 ). 2. t2 = 10. Consider again Problem 1 on page 54. in an actual numerical computation we also make round-oﬀ errors in each step. . . t2 = 10. y) in Example 1. the number of steps we make (at least if all steps are of the same length). Arrange them in a table. . .77. Iteratively. .75 is of this kind.64) dy 1 1 = y− y2 dt 10 10000 and y(0) = 10. On the other hand. Example 1. and we repeat the process. . t3 = 15. y(ti )] according to the formula in (1. t1 = 5. BASIC CONCEPTS Multi-step approach: We like to ﬁnd a remedy for the problem which we discovered in Example 1. . Then we pretend that y(t1 ) is exact. we use the one step method from above to get an approximate value for y(t1 ). t3 = 15. For reasonably nice11 expressions F (t. 1. but the F (t. Again we pretend that y(t2 ) is exact and use t2 and y(t2 ) to calculate y(t3 ). t20 = 100.76 for T further away from t0 . y(ti ))(ti+1 − ti )] We continue this process until we reach T .56 CHAPTER 1. Graph the points found in the previous step together with the actual solution of the initial value problem. Starting out with t0 and y(t0 ). 11 We do not want to make this term precise.58): (1. t4 = 20. and the more steps we make the worse the result might get. y(ti ) + F (ti . t20 = 100. . Consider the initial value problem (1. Solution: As points in the multi-step process we use t0 = 0.63) [ti+1 . Experience will guide you in the choice of the step length. y(ti+1 )] from [ti . For notational convenience we assume that T > t0 . y(ti+1 )] = [ti+1 . y) the accuracy of the value which we get for y(T ) will increase with n. Pick several ti between t0 and T : t0 < t1 < t2 < · · · < tn = T. . We want to get an approximate value for y(T ). we calculate [ti+1 .

07 90 988. Suppose F (y0 . . t) = 0 for all t.95 22. Such a solution is called a steady state solution.55 55 531.55 60 656. Then the constant function y(t) = y0 is a solution of the problem. .15 you see the graph of the exact solution of the initial value problem.09 45 304. In Figure 1.05 65 768. t2 = 2. y(t2 ). t3 = 3. 12 It is incidental that the points eventually get closer to the graph again. You also see the points from Table 1. . y(t3 ).87 70 857.2: Solution of Problem 1.31 33. y) dt and y(t0 ) = y0 .28 72.77 For each ti (0 ≤ i ≤ 19) we use the formula y(ti+1 ) = y(ti ) + 5 2 y(ti ) yi (ti ) − 10 10000 and calculate y(t1 ).2.74 80 956.1.57): y = dy = F (t. y(t20 ) consecutively.62 50 410. But you see that we are deﬁnitely making errors.00 14. ♦ Steady States: Let us consider some very speciﬁc solutions of our initial value problem in (1.07 85 977.2. . .73 75 918. This is due to the speciﬁc problem. and will not occur in general.18. and they get worse as t increases12 . t100 = 100 in your calculation.07 100 997.70 t y(t) & t y(t) 35 153. .22 49. . .96 40 219. We summarize the calculation in Table 1.02 30 106.27 95 994. You may try a shorter step length. NUMERICAL METHODS 57 t 0 5 10 15 20 25 y(t) & 10. . The points suggest a graph which does follow the actual one reasonably closely. The points will follow the curve much more closely if you use t1 = 1.41 Table 1.

Find the steady states of the diﬀerential equation (see (1.65) f (t) = af (t) + b. then the principal balance of your account will stay unchanged. Solution: Apparently. and you are repaying the loan at a rate of $ 200.16) (1..5% per month.e.78.59)) dy = F (y.58 CHAPTER 1. the bank charges you interest at a rate of . So the constant function f (t) = −b/a is the only steady state of this diﬀerential equation. Example 1.50) in Section 1.00 per month. ♦ In review of Example 1.79. f (t) = 0 if and only if f (t) = −b/a.00.000. t) = ay − by 2 = y(a − by) dt . BASIC CONCEPTS 1000 800 600 400 200 20 40 60 80 100 Figure 1.16. 000. I. you see that the steady state in that example is B(t) = 40.68 in Section 1.15: Illustration of Euler’s Method Example 1. For the logistic law (see Equation (1. if your loan balance is $40. Your payments cover exactly the occuring interest charges.

There are two steady state solutions: yu (t) = 0 and ys (t) = a/b. then y(t) will not tend to the steady state y(t) = 0. The population has a growth rate of 50% per year. ys (t) = a/b = 1. t) = 0 if and only if y = 0 or y = a/b. Consider the initial value problem (1. 1. Let us interpret these steady state solutions for the speciﬁc numerical values of a = 1/10 and b = 1/10. the last term in the diﬀerential equation accounts for the competition for space and food. 000. Call the smaller one of them Yu and the larger one Ys . Repeat the ﬁrst two steps of the problem if hunting is stopped. 13 . Repeat the ﬁrst two steps of the problem if the initial population is 100 animals. Guess at which level the population stabilizes. 2. Experiment with diﬀerent initial values to see which of the steady states is stable. ﬁnd for which values of y you have that y = 0? You will ﬁnd two values. you should think of a population of deer in a protected wildlife preserve. It is also referred to as the carrying capacity. and the mathematical deﬁnition of ‘tends to’ and ‘stabilizes at’ only make these terms precise. 000 in Example 1. ♦ Exercise 13. and which one is unstable. Use Euler’s method to ﬁnd the population size over the next 30 years. I. In this sense. The common language meaning of these expressions suﬃces for the purpose of our discussion. If the initial value y0 is negative. In this sense. It tells you which size population of the given kind the speciﬁc habitat will support. then y(t) will tend to −∞ as time increases. Find the steady states of the original equation in which hunting takes place..e. 5. If the initial value y0 of the population is positive. 3. Reproduction takes place at a constant rate all year round.66) 1 1 2 y (t) = −50 + y(t) − y (t) 2 2000 and y0 = y(0) = 200. Finally. NUMERICAL METHODS 59 we ﬁnd that F (y. There are no predators. 4.18.77. Proceed in 1 year steps. y(t) = 0 is an unstable steady state. The deer are hunted at a rate of 50 animals per year. To make the problem explicit. If y0 = 0. then the population size will tend to and stabilize13 at y(t) = a/b = 1. 000 is a stable steady state solution. Tabulate and plot your results.1.

16: Orthogonal Trajectory to Level Curves Let us explain where this type of situation occurs. You also see one orthogonal trajectory in Figure 1.60 CHAPTER 1. A charged particle will move along an orthogonal trajectory. y. The orthogonal trajectory gives a path of steepest descent. 6 4 2 0 -2 -4 -6 -6 -4 -2 0 2 4 6 Figure 1. A new lava ﬂow which originates at some point in the crater will follow this path. Suppose we are given a family F (x. a) = 0 of curves.16. Here a represents the elevation. y1 ).) We call such a curve Db an orthogonal trajectory to the family of the Ca ’s. The orthogonal trajectory provides you with a path which is always in the direction of the most rapid change of the ﬁeld. if the tangent lines to the curves at this point intersect perpendicularly.16 you see a family of ellipses (1.67) Ca : F (x. so that the elevation is constant along each curve Ca . In Figure 1. BASIC CONCEPTS Orthogonal Trajectories Let us explore a diﬀerent kind of application. We like to ﬁnd curves Db which intersect the curves Ca perpendicularly. Suppose that each ellipse represents an equipotential line of an electromagnetic ﬁeld. Suppose the curves Ca are the level curves in a crater. a) = x2 + 3y 2 − a = 0. . y. (We say that Db and Ca intersect perpendicularly in a point (x1 . There is one ellipse for each a > 0.

to ﬁnd an orthogx onal trajectory to the family of the Ca ’s we need to ﬁnd functions which satisfy this diﬀerential equation. along an orthogonal trajectory to the isothermal lines. y0 ).16 has the equation y = x3 /25. Suppose a stands for the concentration of a nutrient in a solution. They are functions of the form y(x) = bx3 . This is exactly the kind of problem which we solved with Euler’s method. was pioneered by Alexander von Humbold (1769–1859). Solution: Diﬀerentiating the equation for the ellipses. we get 2x + 6y dy = 0 or dx dy −x = . There is one orthogonal trajectory which does not have this form. Thus. dx 3y The slope of the tangent line to a curve Ca at a point (x1 . Example 1. Let us ﬁnd approximate values for the initial value problem dy 3y = dx x and y(1) = 1 . In this particular example it is not diﬃcult to ﬁnd solutions for the diﬀerential equation. i. If we also require that the orthogonal trajectory goes through a speciﬁc point (x0 . The orthogonal trajectory shown in Figure 1. y1 ) perpendicularly. bacteria will follow a path in the direction in which the concentration increases most rapidly. so that along each ellipse the temperature is constant.. a) = x2 + 3y 2 − a = 0. Let us apply Euler’s method to solve the problem. In this case the curves are called isothermal lines14 . move in the direction in which the temperature increases most rapidly. and this is the curve x = 0.80. y1 ) is −x11 . then we end up with the initial value problem dy 3y = dx x and y(x0 ) = y0 . On their search for food. They will move along an orthogonal trajectory. Find orthogonal trajectories for the family of ellipses (1.1.e. NUMERICAL METHODS 61 Suppose a stands for temperature.68) Ca : F (x. then we need that the slope of the tangent line to Db at this point is 3y11 . at any time. 25 14 The idea of isothermal lines. y. . It is constant along each curve Ca .18. If 3y a curve Db intersects Ca in (x1 . A heat seeking bug will. and with this the method in all of these applications.

x20 = 7.4. . only for a = 0 the hyperbola degenerates into two intersecting lines. . 1/25) and calculate (xn . Use Euler’s method to ﬁnd points on the orthogonal trajectory through the point (3. x1 = 3. . . . ♦ Exercise 14. y0 ) = (1. and add this graph to your ﬁgure. 2. .2 3yn−1 xn−1 for n = 1.62 CHAPTER 1. .4. 4). Consider the family of hyperbolas: Ca : x2 − 5y 2 + a = 0. BASIC CONCEPTS Use x0 = 1. x1 = 1. 20. yn ) in your ﬁgure. 1. . . yn ) according to the formula yn = yn−1 + . Without recording the results of this calculation. Determine b. . Find the diﬀerential equation for an orthogonal trajectory. Use the points x0 = 3. . x2 = 1.2. . Plot the points (xn . There is one hyperbola for each value of a.16. 3. so that the orthogonal trajectory passes through the point (3. x2 = 3. 2. 4. We set (x0 .2. 4). x20 = 5. we graphed the points in Figure 1. Graph several of the curves Ca . . Check that the graph of y(x) = bx−5 is an orthogonal trajectory to the family of hyperbolas for every b.

0) ∪ (0. 1).19 Table of Important Derivatives f (x) xq ex ln |x| sin x cos x tan x cot x sec x csc x arctan x arcsin x arccos x arccot x arcsec x arccsc x f (x) qxq−1 ex 1/x cos x − sin x sec2 x − csc2 x sec x tan x 1 1+x2 √ 1 1−x2 √ −1 1−x2 −1 1+x2 √1 |x| x2 −1 −1 √ |x| x2 −1 Assumptions q a natural number. arccos x ∈ (0. ∞) all x for which tan x is deﬁned all x for which cot x is deﬁned all x for which sec x is deﬁned − csc x cot x all x for which csc x is deﬁned x ∈ (−∞. ∞) x ∈ (−∞. ∞) x ∈ (−∞. π) x < −1 or x > 1. ∞). π) x ∈ (−∞. arcsin x ∈ (−π/2. ∞). π/2) x ∈ (−1. arccot x ∈ (0. π) x < −1 or x > 1.1. 1). π/2) ∪ (π/2. π/2) Table 1. or x > 0 x ∈ (−∞. arcsec x ∈ (−π/2.3: Some Derivatives . arcsec x ∈ (0. x = 0 x ∈ (−∞. ∞) x ∈ (−1.19. TABLE OF IMPORTANT DERIVATIVES 63 1.

64 CHAPTER 1. BASIC CONCEPTS .

Augustin-Louis Cauchy (1789–1857) was one of the great mathematicians of the 19-th century. Then we discuss geometric properties of graphs. For example. Then we call f (b) − f (a) b−a the average rate of change of f over the interval [a. the average rate of change of f (x) = x2 over the interval [0. We will discuss some uniqueness properties of solutions of diﬀerential equations. π] it is 0. The average rate of change of f (x) = sin x over [0. In this chapter we will use local information about a function to draw global conclusions. He made major contributions to make calculus a rigorous mathematical theory. π/2] is 2/π and over [0.Chapter 2 Global Theory So far we studied the local behaviour of a function. All concepts related to the behaviour of a function near a point. With this information it is possible to sketch graphs capturing their essential features. The fundamental result which allows us to do this is referred to as Cauchy’s mean value theorem. 2. We apply these ideas to the study of extrema of functions. b].1. b]. 65 . their monotonicity and concavity.1 Cauchy’s Mean Value Theorem It is useful to make the following Deﬁnition 2. Let f (x) be a function which is deﬁned on the interval [a. 2] is 2.

the average rate of change of f (x) = x2 over the interval [−2. then there exists a number c between a and b (i. Theorem 2. If f (a) = f (b). We prove this statement using Cauchy’s theorem. where a < b. Then there exists a number c ∈ (a. and that the proof of both of them depends heavily on the completeness1 of the real numbers. The following special case of the theorem. and the corollary is proved.e. 1 We discussed this property of the real numbers in Section 1. as they occur in the two theorems. b). b). b ∈ I. then a = b and there exists some c ∈ (a. . Let f be a real valued function which is deﬁned and continuous on an interval I. We are not going to say anything about the proof of these two theorems.3 (Rolle’s Theorem).2 (Chauchy’s Mean Value Theorem). then f is constant on this interval. If f (a) = f (b). b−a But this contradicts the assumption that f (c) = 0 for all c ∈ I. Proof. such that f (c) = f (b) − f (a) = 0. the theorem asserts that the average rate of change over an interval is equal to the rate of change at some point in the interval. For example..66 CHAPTER 2. 1] is −1.1. Corollary 2. except that Cauchy’s theorem and Rolle’s theorem are equivalent (each is an easy consequence of the other one). Let f be a real valued function which is deﬁned and continuous on the interval [a. We are interested in more general consequences. b] and differentiable on (a. b] and diﬀerentiable on (a. and f (−1/2) = −1. b−a In words. is of particular interest. GLOBAL THEORY Theorem 2. b). a < c < b) such that f (c) = 0.4. where a < b. In other words. If f (x) = 0 for all interior points x of I. b) such that f (c) = f (b) − f (a) . called Rolle’s theorem (named after Michel Rolle (1652–1719)). there exists a number d such that f (x) = d for all x ∈ I. We are also not interested in ﬁnding the points c. A diﬀerent formulation of the claim is that f (a) = f (b) for all a. Let f be a real valued function which is deﬁned and continuous on the interval [a.

∞) is of the form F (x) = x2 + c where c is a constant. By diﬀerentiating H(x) you can check whether you guessed right. the integration constant is determined by an initial condition. Apply the previous corollary to f (x) = h(x) − g(x). there exists a number d such that h(x) = g(x) + d for all x ∈ I. Proof. Suppose we like to solve the initial value problem f (x) = cos x and f (0) = 1.6. UNIQUE SOLUTIONS OF DIFFERENTIAL EQUATIONS We are going to use the following corollary frequently.7. Any antiderivative F (x) is of the form H(x) + c where c is a constant. π/2) is of the form F (x) = tan x + c.8. if you like to ﬁnd all antiderivatives F (x) of a function f (x) on an interval. We call a function F (x) with domain I an antiderivative of f if F (x) = f (x) for all x ∈ I. on intervals the only solutions of the diﬀerential equation F (x) = 0 are the constant functions. Then h and g diﬀer by a constant.5. Typically. Suppose h and g are antiderivatives of a function f .2 Unique Solutions of Diﬀerential Equations Corollary 2. More generally. 67 Corollary 2.2. For the time being you depend on being able to guess such a function H(x). 2.. we can reformulate Corollary 2. i. If the function F (x) is deﬁned on an interval I and F (x) = 0 for all x ∈ I. Any antiderivative F (x) of the function f (x) = sec2 x on the interval (−π/2. deﬁned on an interval. Using this notion. . then it suﬃces to ﬁnd one antiderivative H(x). Corollary 2. then F (x) is constant on I. Let h and g be functions which are deﬁned and continuous on an interval I.e.4 implies Proposition 2. then h and g diﬀer by a constant. any antiderivative F (x) of the function f (x) = 2x on the real line (−∞. Deﬁnition 2. If h (x) = g (x) for all x ∈ I. Suppose the function f (x) is deﬁned on the interval I.5. The constant c is referred to as integration constant. In other words. For example.2.

68 CHAPTER 2. and will eventually prove. Next we substitute x = 0 in the equation. Corollary 2. .9. In fact f (x) = Cea(x−x0 ) . that all functions of the form f (x) = ceax satisfy the diﬀerential equation.4 tells us that h(x) is a constant function. This means that all solutions of the diﬀerential equation f (x) = af (x) are of the form f (x) = ceax . Proof. Substituting the initial condition we obtain C = f (x0 ) = ceax0 . By the previous proposition we know that the solution is of the form f (x) = ceax for some c. The initial value problem f (x) = af (x) and f (x0 ) = C has a unique solution on an interval containing x0 . We asserted in (1. The solution of the initial value problem is f (x) = sin x + 1. Consider the function h(x) = f (x)e−ax . Proof. Every solution f (x) of the diﬀerential equation f (x) = af (x) on an interval is of the form f (x) = ceax for some constant c. GLOBAL THEORY Our ﬁrst conclusion is that f (x) = sin x + c. As a product of diﬀerentiable functions. where c is a constant. Let f (x) be any function which satisﬁes the diﬀerential equation on some interval.10. Thus c = Ce−ax0 and f (x) = ceax = Ce−ax0 eax = Cea(x−x0 ) . We want to see that these are the solutions. h is diﬀerentiable. Its derivative is h (x) = f (x)e−ax − af (x)e−ax = af (x)e−ax − af (x)e−ax = 0. Calling the constant c we ﬁnd that f (x) = ceax . Then we see that f (0) = c = 1. This follows from the above because (sin x + c) = cos x. Of particular importance to our discussion of expenential growth and decay is Proposition 2.19). Proposition 2.

2.3. THE FIRST DERIVATIVE AND MONOTONICITY

69

Remark 4. The uniqueness of the solution of an initial value problem as in the previous proposition is not only of theoretical importance. Imagine that you study the growth rate of a strain of bacteria, as we did in Example 1.64 on page 41. Before you can publish your result, it must be certain that your experiment can be reproduced at a diﬀerent time in a diﬀerent location. That is a requirement which any experiment in science must satisfy. If there is more than one mathematical solution to your problem, then you have to expect that the experiment can go either way, and this would invalidate your experiment.

2.3

The First Derivative and Monotonicity

One of the interesting properties of a function is whether it is increasing or decreasing. We might want to ﬁnd out whether the part of a population which is infected with a disease is increasing or decreasing. We might want to know how the level of pollution in a body of water is changing. The ﬁrst derivative of a function gives us information of this kind.

2.3.1

Monotonicity on Intervals

Recall that a function f is called increasing if f (b) > f (a) whenever b > a. It is called decreasing if f (b) < f (a) whenever b > a. A function is called monotonic if it is either increasing or decreasing. Theorem 2.11. Suppose that the function f is deﬁned and continuous on the interval I. 1. If f (x) > 0 for all x ∈ I, then f is increasing on I. 2. If f (x) < 0 for all x ∈ I, then f is decreasing on I. 3. More generally, the conclusions in (1) and (2) still hold if in each ﬁnite interval J ⊂ I there are only ﬁnitely many points at which the assumption on f (x) is not satisﬁed.2 Proof. We show (1). Let a and b be points in I, and suppose that a < b. Cauchy’s theorem says that there exists a point c, a < c < b, such that f (c) =

2

f (b) − f (a) . b−a

It is permissable that f is not diﬀerentiable at a few points in J, or that f (x) = 0. It is not possible that f (x) < 0 at some point in the interval, and f (x) is increasing on the interval.

70

CHAPTER 2. GLOBAL THEORY

We have that f (c) > 0 and b − a > 0, and it follows that f (b) − f (a) > 0. This means that f (b) > f (a). The proof of the second claim is similar. We leave it and the generalization of both statements to the reader.

1 For example, log2 x = ln x = x ln 2 > 0 for all x ∈ (0, ∞). In the ln 2 computation we used (1.19), Theorem 1.36, and that ln 2 > 0. It follows from Theorem 2.11 that log2 x is increasing on x ∈ (0, ∞). You see part of the graph of the function in Figure 1.8. The exponential function expa x = ax is increasing on (−∞, ∞) if a > 1 and decreasing if 0 < a < 1. To see this, observe that ax = ex ln a and d x x x dx a = (ln a)a . Furthermore, a > 0 and ln a > 0 if a > 1 and ln a < 0 if 0 < a < 1. Now Theorem 2.11 implies our assertion. You may also want to have a look at the graph of the exponential function with base 2 in Figure 1.7. The function f (x) = 1/x is deﬁned and diﬀerentiable on the set of all nonzero real numbers, and its derivative is f (x) = −1/x2 . In particular f (x) < 0 for all nonzero real numbers. According to Theorem 2.11, f (x) is decreasing on the interval (−∞, 0), and that f (x) is decreasing on the interval (0, ∞). The function is not decreasing on the union of the two intervals. The example illustrates that it is crucial in Theorem 2.11 that we deal with functions which are deﬁned and diﬀerentiable on an interval. The function f (x) = tan x, deﬁned on (−π/2, π/2), has as its derivative f (x) = sec2 x, and the derivative is positive. Consequently, f (x) is increasing on (−π/2, π/2). Its inverse g(x) = arctan x, deﬁned on (−∞, ∞), 1 has as its derivative g (x) = 1+x2 , which is positive on (−∞, ∞), so that g(x) = arctan x is increasing on (−∞, ∞). As a general priciple, one may show that the inverse of an increasing function is increasing.

Example 2.12. For a three dimensional solid we set E = A/V , where A denotes the surface area and V the volume. For example, for a ball E(r) = (4πr 2 )/( 4 πr 3 ) = 3/r, where r denotes the radius. Then E (r) < 0. 3 The same principle holds for other shapes, E decreases as we enlarge the solid without changing its shape. What does this have to do with the size of animals? Warm blooded animals living in cold climates need to preserve their body temperature. The total amount of heat stored in the body is proportional to the volume, while the heat loss is proportional to the surface area. The ratio of volume to surface area increases as the animal gets larger, so that for warm blooded animals it is of advantage to be large if they live in cold climates. In hot climates they need to give oﬀ heat, so that it is of advantage to be

2.3. THE FIRST DERIVATIVE AND MONOTONICITY

71

small. Natural selection (Darwinism) should favor the larger specimens of a warm blooded species in a cold climate and smaller ones in a hot climate. You can observe this phenomenon in real life. For cold blooded animals the converse holds. They absorbe heat so that they body reaches a temperature at which they can be active. In cold climates it helps to be small, because then the surface area is relatively large, compared to the volume. In hot climates cold blooded animals can aﬀord to be large, as it is easy to reach and maintain the temperature at which they can be active. The argument is again consistent with real life. Needless to say, there are other mechanisms to increase the surface area of a body than decreasing its size, and the maintenance of the body temperature is only one factor which inﬂuences the size of specimens of a species. Larger animals need more food, are stronger, cannot hide so well, and are often less agile. All of these factors need to be taken into account to determine the optimal size of an animal. ♦ So far we have only discussed examples where we used (1) and (2) of Theorem 2.11. Let us show how to use the conclusion in (3). To apply it we need to determine intervals on which a function does not change signs. We recall a procedure which works well for continuous functions. Deﬁnition 2.13. Suppose f (x) is a function. We call a point x0 on the real line exceptional if either f (x0 ) = 0 or f (x0 ) is not deﬁned. The following result is an immediate consequence of the Intermediate Value Theorem, see Theorem 1.16 on page 8. Expressed casually it says that a continuous function can change signs only at exceptional points. Proposition 2.14. Suppose f (x) is continuous and f (x) has no exceptional points in the interval (x0 , x1 ). Then f (x) > 0 for all points in the interval (x0 , x1 ), or f (x) < 0 for all points in the interval (x0 , x1 ). In particular, if f (x) is positive at one point in the interval, then it is positive at all points in the interval. If f (x) is negative at one point in the interval, then it is negative at all points in the interval. Example 2.15. For example, consider the function f (x) = x2 (x2 − 4) x2 (x − 2)(x + 2) = . 2 + 2x − 15 x (x − 3)(x + 5)

The zeros of the numerator, and with this the zeros of f (x), are x = 0, x = 2, and x = −2. The zeros of the denominator, i.e., the points where f (x) is not deﬁned, are x = 3 and x = −5.

f (x) = 3x2 + 5x − 4 Figure 2. ∞).11 (3) says that f is increasing on the interval [−5/6.2: A cubic polynomial.72 CHAPTER 2. 0). Theorem 2. 30 5 25 20 15 10 5 -3 -2 -1 -5 1 2 3 4 -2 -5 -10 -15 -20 2 4 Figure 2. ♦ . 3) and (3. Counting signs of the factors in the expression for f (x). In particular. Find intervals of monotonicity for the function f (x) = 3x2 + 5x − 4. You see that the sign changes at some. (2. −2). f is decreasing on the interval (−∞. Decide whether the functions are positive or negative on these intervals. −2). ∞). (0. ∞). So f (x) > 0 for all points x ∈ [−5/6. Find intervals on which the following functions do not change signs. negative on (−5. the sign of f (x) remains unchanged on each of the intervals (−∞. −5/6]. but not all. positive on (−2.1: A quadratic polynomial. GLOBAL THEORY According to the proposition. ♦ Exercise 15. 0) and on (0. 3). we see f (x) is positive on the interval (−∞. −5). x + 5x2 − 4x − 20 We are ready to discuss the monotonicity of functions whose derivative vanishes at some points. ∞). −5). f (x) > 0 if x ∈ (−5/6. ∞).16. exceptional numbers.1. x (1) f (x) = x3 − x2 − 5x − 3 (2) g(x) = 3 . negative on (2. 2). p(x) = x3 − 3x2 − 9x + 3 Solution: We graphed the function in Figure 2. By a similar argument. 2). Example 2. except at x = −5/6. and positive on (3. (−2. Its derivative is f (x) = 6x + 5. (−5.

17.18. 3]. Find intervals of monotonicity for the degree three polynomial (for a graph see Figure 2. ∞).2. −1). ∞) and that it is increasing on the interval (−∞. and (3. Find intervals of monotonicity for the rational function x2 + 3x . 3). ∞). f (x) = 2 cos 2x + 2 cos x = 2[2 cos2 x + cos x − 1] = 4(cos x + 1) cos x − 1 2 . Find intervals on which the function f (x) = sin 2x + 2 sin x is monotonic. THE FIRST DERIVATIVE AND MONOTONICITY 73 Example 2. We conclude that f (x) is increasing on the (−∞. 2π]. . x−1 Solution: The simpliﬁed expression for the derivative of f is f (x) = f (x) = (x + 1)(x − 3) . 1) and (1. The derivative is negative on the interval (−1. Counting the signs of the factors we see that p (x) is positive on (−∞. The function is decreasing on the intervals [−1. ∞). We conclude that f (x) does not change signs on the intervals (−∞. We conclude that p(x) is increasing on the interval [3. −1) and (3.2) p(x) = x3 − 3x2 − 9x + 3 Solution: The function is deﬁned and diﬀerentiable on the real line. 1) and (1. 3). 3]. −1] and [3. Counting the signs of the factors of f (x). ∞). −1]. and f (x) < 0 on the intervals (−1. (−1. Solution: We diﬀerentiate the function and rewrite the expression for the derivative so that it is easier to ﬁnd its exceptional points. The theorem implies that p(x) is decreasing on the interval [−1. we conclude that f (x) > 0 on the intervals (−∞. (x − 1)2 We see that the exceptional points for f (x) are x = 1. Restrict your discussion to the interval [0. (1. Its derivative is p (x) = 3x2 − 6x − 9 = 3(x2 − 2x − 3) = 3(x − 3)(x + 1). 3). Observe that f (x) is deﬁned and diﬀerentiable on the entire real line with the only exception of x = 1.3. 1). ♦ Example 2. x = −1 and x = 3.19. ♦ Example 2. −1) and on (3.

and that f (x) = 0 at the end points of this interval. we ﬁnd that f (x) > 0 for x ∈ [0. (g) and (h). We ﬁnd exceptional points where cos x = −1 (i. there f (x) is increasing. Checking the sign of f (at one point) in each of the intervals. 2π]. (π.3: A function and its derivative. Then we solved the quadratic equation in terms of cos x.e. In the last two problems.3. Find intervals on which the function f increases and intervals on which f decreases.. π/3). 5π/3) and (5π/3. π/3) and x ∈ (5π/3.74 CHAPTER 2. 5π/3). The function is decreasing on the interval [π/3. GLOBAL THEORY To see the second equality we used that cos 2x = 2 cos2 x − 1. There you see the graph of the function (solid line) and the graph of its derivative (dashed line). there f (x) is decreasing. restrict yourself . and f (x) < 0 for x ∈ (π/3.. wherever f (x) is positive. 2π] on which f does not change sign. 2π]. This provides us with the intervals [0. (π/3. π/3] and [5π/3. 2 3 3 4 3 2 1 1 -1 -2 2 3 4 5 6 Figure 2. x = π and x = 5π ). You may conﬁrm the calculation by having a look at Figure 2. Wherever f (x) is negative. x = π) and where cos x = 1 (i. π).e. As you see. 2π]. We conclude that f is increasing on the interval [0. π) and (π. 5π/3]. and in this interval there are three points at which f (x) is not positive. ♦ Exercise 16. Observe that f is diﬀerentiable on [0.

(a) f (x) = 3x2 + 5x + 7 (b) f (x) = x3 − 3x2 + 6 (c) f (x) = (x + 3)/(x − 7) (d) f (x) = x + 1/x (e) f (x) = x3 (1 + x) (f) f (x) = x/(1 + x2 ) (g) f (x) = cos 2x + 2 cos x √ (h) f (x) = sin2 x − 3 sin x 75 2. c) and f (x) > f (c) for all x ∈ (c. interval. For the global property the interval is given to us. to the left of c the function is smaller and to the right of c it is larger than at c. at least for a while. Suppose f is a function which is deﬁned on an open interval I. We address both questions in this subsection. Being increasing or decreasing at a point c is a local property. 2π]. and how this concept is related to the one of being increasing on an interval. Theorem 2. For the local property we may chose the. f (x) < f (c) for all x ∈ (c − d. possibly rather small.20.2. c + d). . The global property has to hold for any two points in the given interval. We say that f is increasing at c if. Being increasing on an interval is a global property. Deﬁnition 2. This theorem establishes the relation between the local and the global property.21. For the local property we compare f (x) to f (c) where c is ﬁxed and x is any point in an open interval around c which we may chose. for some d > 0. but the ‘if’ part uses some deeper facts about ﬁnite closed intervals. Expressed informally. Suppose f is a function and c is an interior point of its domain. The ‘only if’ part is not diﬃcult to show. We are making a statement about the behavior of the function on some open interval which contains c. Then f is increasing (decreasing) on I if and only it it is increasing (decreasing) at each point in I. THE FIRST DERIVATIVE AND MONOTONICITY to the interval [0.2 Monotonicity at a Point It is quite natural to ask what it means that a function is increasing at a point. Our second result gives us a valuable tool to detect monotonicity of functions at a point.3.3. We say that f is decreasing at c if this statement holds with the inequalities reversed.

f (x) = x3 is increasing at x = 0. A function does not have to be diﬀerentiable to be increasing.. 2. i.4: Concave Up Figure 2. then f is increasing at c. The ideas of of a function being increasing or decreasing at a point may be generalized to cover domains of functions which are halfclosed or closed intervals. GLOBAL THEORY Proposition 2. You may use the graphs in Figures 2. A function can be diﬀerentiable and increasing at a point x. We have no speciﬁc needs for such statements.5: Concave Down . Remark 5.e.4 and 2. then f is decreasing at c. Secant lines will either be required to lie above or below the graph. even if the assumptions of Proposition 2.5 as illustrations of the discussion. but if f (0) = 0. and the rates of change will be either increasing of decreasing.22.22 do not hold. If f is diﬀerentiable at c and f (c) > 0. These properties can be described globally over intervals and locally at points. A function can also be increasing at a point x. Graph the function f (x) = 2x + |x| to convince yourself of this fact. but the motivated reader is encouraged to explore them. Remark 6. but there is not open interval which contains x such that the function is increasing on this interval. and where we like to make a statement about the behavior of a function at an endpoint. Let f be a function and c an interior point of its domain.76 CHAPTER 2. If f (c) < 0. 10 8 6 4 -2 -1 4 2 1 -2 -4 2 3 4 2 -6 -2 -1 1 2 -8 Figure 2.4 The Second Derivative and Concavity We like to capture the property of a graph being bent upwards or downwards.

f (a)) and (b. Deﬁnition 2. We say that f (x) is twice diﬀerentiable on I. then f (x) is concave up on I. so far we can consider being ‘twice diﬀerentiable’ only for functions which are deﬁned on open intervals. we proceed as in Section 1. f (a)) and (b. then f (x) is concave down on I.4. THE SECOND DERIVATIVE AND CONCAVITY 77 2. We say that f is concave down on I if f (c) > l(c) for all a. b). if f (x) extends to a function F (x) which is deﬁned on an open interval J which contains I. The inequality expresses that between the points a and b the secant line lies below the graph.1 Concavity on Intervals Let f (x) be a function and let (a.24. 1. f (b)) be two distinct points on its graph. is not satisﬁed. Let f be a function which is deﬁned on an interval I. the function shown in Figure 2. b]. then f (x) is concave down on I. The inequality expresses that between the points a and b the secant line lies above the graph. If f (x) is decreasing on I. i. If f (x) < 0 for all x in I. the line segment joining the two points.4. f (x) < 0.23. 2. b in I and c ∈ (a. then we get the secant line through the two point. For example. Theorem 2. Let f be a function which is deﬁned on an interval I. and F (x) is twice diﬀerentiable on J. b). 3.11. The line through these two points is l(x) = f (a) + f (b) − f (a) (x − a). b in I and c ∈ (a. b−a If we restrict l(x) to x ∈ [a. Its second derivative is q (x) = 2 > 0. More generally. the conclusions in (2) still hold if in each ﬁnite interval J ⊂ I there are only ﬁnitely many points at which the assumption f (x) > 0. We say that f is concave up on I if f (c) < l(c) for all a. The second derivative will be unique at all points in I if I is not empty and does not consist of exactly one point.4 is q(x) = x2 − 2x + 3. Suppose that f (x) is diﬀerentiable on I. Suppose that f (x) is twice diﬀerentiable3 on I. resp. then f (x) is concave up on I. f (b)). If f (x) > 0 for all x in I. Here l(x) is the secant line through (a. We state a theorem which provides you with assumptions under which a function is concave up or down.24 (2) says that q is concave 3 Strictly speaking. . Theorem 2.2.e. If f (x) is increasing on I.. We will not provide a proof of the theorem. More generally.

The function shown in Figure 2. π/2). Theorem 2. you may use that ln (x) = −1/x2 < 0 on (0. Thus. you may note that the derivative ln x = 1/x is decreasing on (0. then g is concave down. I(t) increases.24 (2).78 CHAPTER 2.24 (1) to derive the desrired conclusion. 1). Remark 7. 4 . ♦ Consider the function tan x. ∞) and apply Theorem 2. Theorem 2.24 (2) says that q is concave down on (−∞. x = 1. Alternatively. It is worse. ∞). This means that p (x) > 0 for all x ∈ [1. You are invited to study the concavity of the other trigonometric and hyperbolic functions. The exponential function exp(x) = ex is concave up on (−∞.2. secant lines which are above the graph turn into secant lines below the graph. Its second derivative is p (x) = 6x−6 = 6(x−1).. 0] and concave up on [0.5 is g(x) = −x2 + 5x − 1.24 (3) tells us that p(x) is concave up on the interval [1. One deduces that f (x) is concave down on the interval (−∞. 1) with only one exception. 0) and tan x > 0 for x ∈ (0. if f is concave up. ∞). So. Study the concavity properties of the function p(x) = x3 − 3x2 − 9x + 3. 1]. You may verify that tan x = 2 sec2 x tan x. You may also use that exp (x) is increasing on the real line. Similarly. π/2). and often true in the early stages of an epedemic. you may observe that a function is concave up if its inverse is convave down4 . Solution: You ﬁnd the graph of this function in Figure 2. ∞).25. and then quote Theorem 2. To see this. You may consider the spread of a desease. ∞). Example 2. Theorem 2. p (x) < 0 for x ∈ [−∞. then the graph of one of the functions is obtained from the one of the other one by reﬂection at the diagonal x = y. x = 1. and vice versa. i. You may conﬁrm these statements visually by inspecting a graph of the tangent function. We see that p (x) > 0 for x ∈ (1. ∞). tan x < 0 for x ∈ (−π/2. ln x being concave down implies that exp(x) is concave up. If f and g are inverses of each other. Let us look at examples where we apply condition Theorem 2.24 (3) implies that tan x is concave down on (−π/2. It may be scary if I (t) > 0. you may note that exp (x) = exp(x) > 0 and apply Theorem 2.24 (2). In particular. and its second derivative is g (x) = −2 < 0. The function ln x is concave down on the interval (0. ∞). and p (x) < 0 for x ∈ (−∞. Finally. ∞) with only one exception.e. To see this. In this process. if I (t) > 0.24 (3). ∞) and apply Theorem 2.24 (1). Denote the number of infected people by I(t). GLOBAL THEORY up on (−∞.

there has been an increase in poverty but it is a lower rate of increase than it was in the preceding years before we got here. When I (t) turns negative. and it was the stable equilibrium point. the population grows at an increasing rate. One may hope that eventually I (t) becomes negative. See Example 1. so that the actual number of sick people decreases. and the desease spreads at an increasing rate. If a/(2b) < y < a/b. THE SECOND DERIVATIVE AND CONCAVITY 79 This means that I (t) increases. concave down.4. starts decreasing. 1984. then growth slows. The ﬁrst case occurs if y = 0 or y = a/b. Find intervals on which the following functions are concave up. During a televised presidential debate.15. i(x) = 2x4 − x2 5. Medical professional will not necessarily wait for the time when I(t). 1. h(x) = x + 1/x 4.15. It has begun to decline. but after a while it levels oﬀ so that it does not exceed a the carrying capacity. Use implicit diﬀerentiation to ﬁnd the second derivative: y = ay − 2byy = (a − 2by)y . 2π]. Earlier we considered the logistic equation y = ay − by 2 . f (x) = x3 − 4x2 + 8x − 7 2. For a while the population seems to explode. The inﬂection occurs when y is half the carrying capacity. then I (t) decreases.” 5 .2. page B6): “Some of these facts and ﬁgures just don’t add up. Exercise 17. j(x) = x/(x2 − 1) 6. the number of infected people.. The point at which I (t) changes signs from being positive to being negative may be considered the turning point in the spread of the desease. k(x) = 2 cos2 x − x2 for x ∈ [0. g(x) = x4 + 2x3 − 3x2 + 5x − 2 3. One of the recent presidents was confused by a subtle argument of this kind5 . You see the turning point in the graph in Figure 1. The spread or the desease slows.75 and the graph of a solution of this diferential equation in Figure 1. one of the candidates said (see the New York Times from October 8th. We called y = a/b the carrying capacity of the system. resp. Let us look at this phenomena in a concrete example. We see that y = 0 if y = 0 or y = a/(2b). but it is still going up. As long as y is less than a/(2b). Yes.

5 1 1. then the function is concave up. then the function is concave down. You see this situation illustrated in two generic pictures in Figures 2. at least near c.6: Concave up at • Figure 2.6 and 2. resp. Still. such that the graph lies on one side of the graph. Let f be a function and c an interior point 6 of its domain. called a support line...7: Concave down at • In other words. 4 2 3 1. such that l(c) = f (c) and f (x) > l(x).5 1 -0. We say that f is concave up. and it tells us how to ﬁnd the support line if the function is diﬀerentiable. resp. 6 The idea of an interior point was deﬁned in Deﬁnition 1. If the graph is above the line.5 1 2 0. f (x) < l(x). we are asking for a line l(x). if it is below. We assume that the graph and the line agree at c. Deﬁnition 2. Our next theorem tells us how to detect concavity.7.5 0.80 CHAPTER 2. concave down. One shows a function which is concave up at the indicated point. . at c if there exists an open interval I and a line l.4.2 Concavity at a Point The notion of being concave up or down was deﬁned for functions which are deﬁned on intervals. and this is the behavior which we like to capture in a deﬁnition.26.18 on page 10.5 2 1 2 3 4 5 Figure 2. GLOBAL THEORY 2. for all x ∈ I with x = c. one shows a function which is concave down. we got a picture how the function has to look like near a point.

As we like to apply local properties of the function. minimum. LOCAL EXTREMA AND INFLECTION POINTS 81 Theorem 2. If the second derivative is zero.28. Let f be a function which is deﬁned on an open interval (a.2. down. then f is concave down at c.5 Local Extrema and Inﬂection Points We are going to discuss two types of points which are particularly important in the discussion of (graphs of) functions. A local extremum is a local maximum or minimum.5. of f . minimum. . Then f is concave up (resp. It is the tangent line. but if the function is diﬀerentiable at c. then f is concave up at c. We say that f has a local maximum. We just hold the ruler against the graph. If f is diﬀerentiable and concave up or down at c. b) if and only if f is concave up (resp. resp. we can draw the tangent line easily. If f is decreasing at c or if f (c) < 0. Let f be a function and c an interior point of its domain. b). In general. Deﬁnition 2. then there is only one support line. resp. the following theorem. down) on (a. To relate concavity properties on an interval to those at each point in the interval we state. at c if f (c) ≥ f (x). f (c) ≤ f (x). The function can be concave up. then the support line is unique. there can be many support lines at any given point. without proof. or neither.27. the function f (x) = x5 − 7x4 + 2x3 + 2x2 − 5x + 4 is concave down at x = 2 because f (2) = −148 < 0. For example. we focus on interior points is the domain of the function. In this case we call f (c) a local maximum. for all x in some open interval I around c. The sign of the second derivative of a functions tells us whether a function is concave up or down at a point. and this line is the tangent line to the graph of f at c. 2.29 (Local Extrema). 2. down) at each point in (a. 1.. for a diﬀerentiable function which is concave up or down at a point. Let f be a function and c an interior point in its domain7 . 3. resp. Theorem 2. then the test is inconclusive. So. b).. If f is increasing at c or if f (c) > 0.

π/2). We will study tests which allow us ﬁnd local extrema soon.30 (Inﬂection Points). We call c an inﬂection point of f if the concavity of f changes at c. The function shown in Figure 2. and a local maximum if the coeﬃcient of x2 is negative. .5 3 1 2 0.8: A local minimum Figure 2. b). 7 According to Deﬁnition 1. The function is concave down on the interval (−π/2. if f (c) is the largest value compared the values at points near c. You see the graph of this function in Figure 2.e. We do not need any test to see that f (x) = |x| has a local minimum at x = 0. or vice versa. No test is required to see that f (x) = tan x has an inﬂection point at x = 0. So the concavity changes at x = 0 and that means that there is an inﬂection point at x = 0. c] and concave down on [c. Soon we will develop tests which detect inﬂections points.9: An Inﬂection Point In other words.82 CHAPTER 2.8 has a local minimum at x = −1. and f (x) = −(x − 1)2 has a local maximum at x = 1. for some numbers a and b with a < c < b. The vertex of a parabola is always a local extremum.5 -0. f has a local maximum of f (c) at c.5 1 Figure 2.5 -1 -1. 0] and concave up on the interval [0. Let f be a function and c an interior point of its domain.18 on page 10 this means that f (x) is deﬁned for all x in some open interval around c.9. GLOBAL THEORY 1.5 0.. I. we have that f is concave up on the interval (a. Deﬁnition 2.5 1 -1 -2 -1 -1 -2 1 2 -0. a local minimum if the coeﬃcient of x2 is positive.

Theorem 2. Let f be a function and c an interior point of its domain. If a function has a local extremum at c. and there are points x to the right of and arbitrarily close to c such that f (x) > f (c).22 on page 76 tells us that there exists some positive number d. that f does not have a local extremum at c. This means. These functions have no critical points. Deﬁnition 2. it is customary to say: Deﬁnition 2.31. If f is diﬀerentiable at c and f (c) = 0. If f (c) = 0. Proof of Theorem 2. So. there are very few points where local extrema can occur. c). if f has a local extremum at c. Let f be a function and c an interior point of its domain. Proposition 2. such that f (x) < f (c) for all x ∈ (c − d. but f does not have a local extremum at c. We say that c is a saddle point of f if f is diﬀerentiable at c and f (c) = 0. by deﬁnition.6 Detection of Local Extrema We will discuss how to detect local extrema. and f (x) > f (c) for all x ∈ (c.31 they have no local extrema. The test does not give a suﬃcient condition for a local extremum.31. To see this. . and according to Theorem 2. In other words.6. c + d). there are points x to the left of and arbitrarily close to c such that f (x) < f (c). then f does not have a local extremum at c.31 provides us with a necessary condition. Let f be a function and c an interior point of its domain. and in neither case we have an extremum at c. or if f is not diﬀerentiable at c.33 (Saddle Points). Neither the exponential function nor the logarithm function have local extrema. then the same argument applies with inequalities reversed.32 (Critical Points). observe that these functions are diﬀerentiable on their domain. then either f (c) > 0 or f (c) < 0. Suppose that f is diﬀerentiable at c and f (c) > 0. No local extrema can occur at points which are not critical. and their derivatives exp x = exp x and ln x = 1/x are everywhere nonzero. then f is either not diﬀerentiable at c or f (c) = 0. then the function need not have a local extremum at c. It makes sense to introduce one more word. The ﬁrst result excludes many points.2. To have an abbreviation for the points which are recognized as important in this theorem. If f (x) < 0. then c is a critical point of the function. DETECTION OF LOCAL EXTREMA 83 2. We say that c is a critical point of f if f is diﬀerentiable at c and f (c) = 0. If c is a critical point of the function. Typically. Theorem 2.

and its only critical point is at x = 0.84 CHAPTER 2. Solution: The function is diﬀerentiable for all real numbers x. ∞) and g(x) < 0 for all x ∈ (−∞.11: A saddle point Example 2. and q (x) = 2x − 2 = 2(x − 1). which is the only zero of g (x) = 3x2 . This means that there is no local extremum at x = 0.11. So q (x) = 0 if x = 1. Solution: The function g(x) is everywhere diﬀerentiable. This saddle point is shown in Figure 2. the only critical point. The only point at which we can have a local extremum. Obviously. the function has a saddle point at this point. is x = 1. As g (0) = 0 and there is no local extremum at x = 0. You should conﬁrm this result by having a look at Figure 2. If we write the function in the form q(x) = (x − 1)2 + 2.e. ♦ .34.35. Find the local extrema of the function q(x) = x2 − 2x + 3.10. where this function is graphed. ♦ 8 8 6 6 4 2 4 -2 -1 -2 1 2 2 -4 -6 -1 1 2 3 -8 Figure 2.10: A local minimum Figure 2. 0). then we see that q does indeed that a local minimum at x = 1. i. g(x) > 0 for all x ∈ (0. GLOBAL THEORY Example 2.. Show that the function g(x) = x3 has a saddle point at x = 0.

c+d). and f (x) is negative on the interval √ √ (1 − 3/3. Theorem 2. and suppose that for some d > 0 the function is increasing on (c−d. positive and negative.2.36. 1 − 3/3) and (1 + 3/3.12 and 2. We √ conclude that f (x) = 0 if √ = 1 ± 3/3. and c is a critical point. If f (x) > 0 for all x ∈ (c−d. Find the local extrema of the function f (x) = x3 − 3x2 + 2x + 2. c + d). c + d). f (x) is positive on the intervals x √ (−∞. If the function is decreasing on (c − d. then f has a local minimum at c. then f has a local maximum at c. c + d). 1 + 3/3). 1. c] and increasing on [c. Suppose f is a function which is deﬁned and diﬀerentiable on (c − d. c + d) for some d > 0.6. If f (x) < 0 for all x ∈ (c − d. ∞). 2.13 . c) and f (x) > 0 for all x ∈ (c. then f has a saddle point at c. Then f has a local maximum at c. This conclusion also holds if f (x) < 0 for all x ∈ (c−d. c) and f (x) < 0 for all x ∈ (c. Taking advantage of the information provided by the ﬁrst derivative. If f (x) > 0 for all x ∈ (c − d. then f has a local minimum at c. It gives us a suﬃcient condition for a local extremum at c.38. Suppose c is an interior point of the domain of a function f . c)∪(c. c + d). 3.37 (First Derivative Test). c] and decreasing on [c. we obtain the following test. c+d). Let us illustrate the use of the theorem with an example. Theorem 2. Example 2. c)∪(c. DETECTION OF LOCAL EXTREMA 85 Let us formulate a criterion which conﬁrms that a function has a local extremum at a point c. You can see graphs of f and f in Figures 2. Solution: We diﬀerentiate f (x) and express f (x) as a product of linear factors: √ √ 3 3 2 f (x) = 3x − 6x + 2 = 3 x − 1 + x− 1− 3 3 It is easy to determine where the factors are zero.

6 1. Exercise 19.5 2 4 3 2 1 0.18 and 2. (a) f (x) = 3x2 + 5x + 7 (b) f (x) = x3 − 3x2 + 6 (c) f (x) = (x + 3)/(x − 7) (d) f (x) = x + 1/x (e) f (x) = x3 (1 + x) (f) f (x) = x/(1 + x2 ) (g) f (x) = cos 2x + 2 cos x √ (h) f (x) = sin2 x − 3 sin x Hint: You discussed the intervals of monotonicity for these functions in Exercise 16.5 1.8 1. 2π].5 2 Figure 2. 2π]. and these are the only points where a local extremum can occur.4 1. Find the local extrema of the following function: (1) f (x) = x2 + 3x x−1 (2) g(x) = sin 2x + 2 sin x for x ∈ [0.2 0.4 2. In the last two problems. Hint: We discussed the monotonicity properties of these functions in Examples 2.5 -1 1 1.19. .86 CHAPTER 2.2 1 1. Based on the sign of f (x) on intervals to the left and right √ these two critical points we see that √ has of f a local maximum at x = 1 − 3/3 and a local minimum at x = 1 + 3/3.12: f (x) = x3 − 3x2 + 2x + 2 Figure 2. ♦ Exercise 18. GLOBAL THEORY 2.13: f (x) = 3x2 − 6x + 2 √ The only only critical points of f are at x = 1 ± 3/3. restrict yourself to the interval [0. (g) and (h). Find the local extrema of the following functions.

see Figure 2. Furthermore. The assumption that f (c) = 0 means that the tangent line to the graph of f at (c. the zeros of f (x). If f (c) > 0. Furthermore.2. The assumption that f (c) > 0 means that f is concave up at c (see Theorem 2. Theorem 2. Find the local extrema of the function (for a graph. the test provides us with a suﬃcient condition for the existence of a local extremum at a point. then f has a local maximum at c. We will show that f has a local minimum at c. p (−1) = −12 and p (3) = 12. p (x) = 6x − 6 = 6(x − 1). It does not provide us with a necessary condition. Spelled out explicitly this means that f (x) > l(x) = f (c) . Solution: We calculated the ﬁrst derivative. then the test is inconclusive. p (x) = 3x2 − 6x − 9 = 3(x + 1)(x − 3). The critical points of the function are x = −1 and x = 3. To apply the theorem to the detection of the local extrema of a diﬀerentiable function f (x). because this is a critical point and p (−1) < 0.40. as it is called for in the assumptions in Theorem 2. Example 2. ♦ Proof of the Second Derivative Test. the function f can have a local extremum at c. In this sense. Then we diﬀerentiate f (x).37. First. If f (c) < 0. The sign of f at the critical points tells us whether we found a local minimum or a local maximum. We also have a local minimum at x = 3 because at this critical point the second derivative of the function is positive. let us assume that f (c) = 0 and f (c) > 0. The second derivative test tells us that we have a local maximum at x = −1. There may or may not be a local extremum at c.2 on page 72) p(x) = x3 − 3x2 − 9x + 3. we diﬀerentiate f and ﬁnd the critical points. If f (c) = f (c) = 0. then f has a local minimum at c. DETECTION OF LOCAL EXTREMA 87 We may use the second derivative to detect the change of sign of the ﬁrst derivative. In particular. Let f be a function and c an interior point in its domain.6. f (c)) is horizontal.27 (1)). Assume also that f (c) and f (c) exist and that f (c) = 0. Its equation is l(x) = f (c). and the assumptions of the test are not satisﬁed.39 (Second Derivative Test).

The proof that f has a local maximum at c if f (c) = 0 and f (c) < 0 is similar. Figure 2. We leave it to the reader.5 1 1. f has a local minimum at c.5 2 1.7 Detection of Inﬂection Points We deﬁned an inﬂection point to be a point at which the concavity of a function changes. 2. Find the the inﬂection points of the function g(x) = x3 − 4x2 + 3x − 5.5 -0. then we can just answer this question. Example 2.5 -0. Find the critical points and the local extrema. c + d).5 1 2 0. . In other words.5 -2 0. GLOBAL THEORY for some positive number d and for all x ∈ (c − d.41.5 0. Exercise 20.5 1 1.5 6 4 Figure 2. If we know where the function is concave up and down. A theorem provides a necessary and a suﬃcient condition for the existence of an inﬂection point.14: The graph of g.5 -0. c) ∪ (c.88 CHAPTER 2. We want to detect inﬂection points more eﬃciently. Let us start out with an example. 8 2.5 2 2. (a) f (x) = 4x2 − 7x + 13 (b) f (x) = x3 − 3x2 + 6 (c) f (x) = x + 3/x (d) f (x) = x2 (1 − x) (e) f (x) = |x2 − 16| (f) f (x) = x2 /(1 + x2 ).5 2 2.15: The graph of g .

Suppose that the ﬁrst and second derivatives of f exist at c. From the formula for the second derivative we conclude that g (x) < 0 if x ∈ (−∞. If f has an inﬂection point at c.2. 4/3] and concave up on [4/3. By deﬁnition. then f has an inﬂection point at c. You also see that g (x) has a local extremum at the same point. We calculate the ﬁrst and second derivative of g: g (x) = 3x2 − 8x + 3 and g (x) = 6x − 8. You see the inﬂection point indicated as a dot in Figure 2. f (c) exists and f (c) = 0. Example 2. We see right away that f (t) = 0 exactly if t = 3/4. Find the inﬂection points of f (t) = 2t4 − 6t3 + 5t2 − 7t + 4. and √ this means that f (9 ± 21)/12) = 0.43.14. √ We could plug t = (9 ± 21)/12 into the expression for f . 4/3) and that g (x) > 0 if x ∈ (4/3. Solution: We calculate the second derivative of the function and ﬁnd f (t) = 24t2 − 36t + 10. According to the theorem. we have to ﬁnd the zeros of f (x) to determine where an inﬂection point can be.42. but this is a bit cumbersome. ∞). We calculate the third derivative of f : f (t) = 48t − 36. Let f be a function and c an interior point of its domain. ∞). ♦ Theorem 2. 1. let us check whether there are inﬂection points at either of these values for t. then f (c) = 0. we have an inﬂection point at x = 4/3.15.7. 4 12 12 Now. DETECTION OF INFLECTION POINTS 89 You see the graph of g in Figure 2. The theorem says that the inﬂection √ points of f (t) are at t = (9 ± 21)/12.14 and the one of g in Figure 2. 2. The roots are √ 3 9 ± 21 1√ t= ± 21 = . If f (c) = 0. ♦ . This means that g is concave down on the interval (−∞.

but the graph of f shows them clearly. We asked the computer to graph f and f for x ∈ [−3. and it will be nearly impossible to ﬁnd the zeros of f . In many cases we like to ﬁnd the maximal value of a function. it will take an eﬀort to calculate the second derivative of this function. Find the inﬂection points of the function f (x) = 1.5 1 0. our ability it ﬁnd inﬂection points of a function is limited by our ability to ﬁnd the zeros of its second derivative. A look at the graph of f barely reveals some of the inﬂection points.17.16: The graph of f . ♦ Exercise 21.8 Absolute Extrema of Functions We said that a function f has a local maximum at c if its value at c is largest in comparison to the values at point near c.17: The graph of f . Apparently. Figure 2. Any reasonable software has no problem with this. 3].16 and 2. Discuss the relation between the inﬂection points of a function f and the local extrema of its derivative f .5 8 6 4 2 -3 0 1 2 3 -2 -1 -2 1 2 3 -3 -2 -1 Figure 2. and where it occurs.5 2 1. then this quite easy. 3 2.44. This concept is captured in . the graph of f tells us much more about the concavity of the function f than its own graph. Example 2. anywhere in the domain of the function. Zooming in on parts of the graph f will not improve this.2 + x2 − 3(sin x)3 . GLOBAL THEORY Apparently. You see the graphs in Figures 2. 2.90 CHAPTER 2. If we are given graphical information. At least in this example.

18: x3 − 5x2 + 6x + 1.5 2 3 4 Figure 2. 8 12. or at a critical . In Theorem 1.47.45. x = 4. and the function has a local maximum at c.5 4 5 2.8.46. If f is diﬀerentiable at c. A continuous function on a closed interval [a. and c is critical as well. If the function does not assume its absolute maximum at an endpoint. then we say that f has an absolute minimum at c.31.2. 4]. the c is critical. A diﬀerent expression is to say that the function assumes its absolute extremum at c. and c a point in its domain. then f (c) = 0 by Theorem 2. the absolute extrema of the function occur either at one of the end points x = 0. ABSOLUTE EXTREMA OF FUNCTIONS 91 Deﬁnition 2. and we call f (c) the absolute minimum of f . Figure 2. If f (x) ≥ f (c) for all x in the domain of f . then it does so at some interior point c. Let f be a function. We say that f has an absolute maximum at c if f (x) ≤ f (c) for all x in the domain of f . Example 2. Then we call f (c) the absolute maximum of f . Find the absolute extrema of the function f (x) = x3 − 5x2 + 6x + 1 for x ∈ [0. Proof.19: 3x2 − 10x + 6 Solution: According to Theorem 2. b] assumes its absolute maximum and minimum either at a critical point or at an endpoint of the interval.46.5 2 1 1 2 3 4 -2. The argument for the absolute minimum is left to the reader.5 10 6 7. If f is not diﬀerentiable at c.17 we asserted that a continuous function assumes its absolute maximum at some point in the interval. Theorem 2.

. ♦ Exercise 22. we conclude that the function assumes its absolute maximum of 9√ x = 4. You may also check that f (x) = 6x − 10.e. 5] (b) f (x) = x3 + 3x2 − 5x + 2 for x ∈ [−3. Example 2. To avoid lenghty introductions to real-life problems. We consider a few examples and give some problems for practice. the zeros of 2 − 10x + 6. Approximate values of these roots are 2.7848.5486 and . (a) f (x) = x2 − 5x + 2 for x ∈ [0. GLOBAL THEORY point.3689 at x = (5 − 7)/3. and its absolute at minimum of approximately . we content ourselves with problems of an algebraic or geometric nature.√ The critical points. Use one piece as the perimeter of an equilateral triangle and the other one as the perimeter of a disk.48. How long should each piece be.92 CHAPTER 2. are x = (5 ± f (x) = 3x 7)/3. Find the absolute extrema of the functions on the indicated intervals.9 Optimization Story Problems Many real-life problems are formulated as optimization problems.18 and the one of f in Figure 2. i.5] √ √ (c) f (x) = 2 + x/ 1 + x for x ∈ [0.3689.19. Cut a string of length 50 centimeters into two pieces. Comparing the values of f (x) at these four points.1126 and f ((5 − 7)/3) = . In fact. You may compare our calculation with the graphs of f in Figure 2. 2. and √ √ f ((5 + 7)/3) > 0 and f ((5 − 7)/3) < 0. f (0) = 1 and f (4) = 9. The approximate values of the function at these points are √ √ f ((5 + 7)/3) = 3. 5] (d) f (x) = cos 2x + 2 cos x for 0 ≤ x ≤ 2π (e) f (x) = sin x + cos x for 0 ≤ x ≤ 2π 2. The second derivative test tells us that the function has a local minimum at √ √ x = (5 + 7)/3 and a local maximum at x = (5 − 7)/3. so that the combined . Calculus helps us to solve these optimization problems.

is where the local minimum occurs. + πr = 4 4 3 For this to make sense. and it will be maximal of r = 25/π. a disk and a square. In the latter case.2. 3 2 The height of the triangle is h = a 2 3 .28 amd A(25/π) ≈ 198.94.9. We note that A(r) is a parabola which is open upwards. For the end points we have: A(0) ≈ 120. The critical point. This means that 3a + 2πr = 50 and a = √ 50 − 2πr . which we just found. The combined area of the triangle and disk is √ √ a2 3 3 50 − 2πr 2 2 A= + πr 2 . Formulate the problem mathematically: Find the absolute minimum (maximum) of the function √ 3 50 − 2πr 2 A(r) = + πr 2 . It is also the absolute minimum of A(r) on any interval which contains the critical point. 3 4 3 3 3 √ and A (r) = 0 if and only if r = 50/(2π + 6 3). √ . Answer the original question: The combined area of the disk and √ the triangle will be minimal if r = 50/(2π + 6 3). 4 3 for r ∈ [0. all string is used for the circle. OPTIMIZATION STORY PROBLEMS 93 area of the triangle and the circle is minimal? How long should each piece be. Repeat the previous example with 1. Introduction of notation: There are many ways to set up the notation to solve this problem. Express information as equations: The perimeter of the triangle will be 3a and the perimeter of the circle will be 2πr. Among them we say that the side length of the triangle is a and the radius of the circle is r. The area of the disk is πr 2 . ♦ Exercise 23. 25/π]. Solve the mathematical problem: The derivative of A(r) is √ 3 50 − 2πr 4π −π 50 − 2πr A (r) = − · + 2πr = √ + 2πr. we need that 0 ≤ r ≤ 25/π. and its area is a 4 3 . so that the combined area of the triangle and the circle is maximal? In our solution we will go through several steps.

On the interior of the interval the function is positive.. we call the longer side of the rectangle L and the shorter one W . an equilateral triangle and a square. W/2]. x ≥ 0. i. Express information as equations: As we folded up a strip of width x. Example 2. height x.e. two geometric shapes of your own choice. What are the dimensions of the box with the largest possible volume? In our solution we will go through several steps. Clariﬁcation and introduction of notation: We construct the box by making an incision at a 45 degree angle at each corner. in fact x ≤ W/2. You could have cut out a square of size x × x at each corner. Then we fold up a strip of width x along each side8 . a regular hexagon and a square. The derivative of V is V (x) = W L − 4(W + L)x + 12x2 . Formulate the problem mathematically: Find the absolute maximum of the function V (x) = W Lx − 2(L + W )x2 + 4x3 for x ∈ [0. . draw a picture of this production process. Solve the mathematical problem: At the end points of the interval V vanishes. To simplify matters. and convince yourself that any box obtained by a diﬀerent process will have smaller volume. x ≤ W/2. The zeros of V are at x= 8 1 (L + W ) ± 6 L2 + W 2 − LW . For yourself. the box will have width W − 2x. 3. V (0) = V (W/2) = 0.94 CHAPTER 2.49. 5. 4. a disk and half an equilateral triangle (the angles at 30. GLOBAL THEORY 2. By construction. length L − 2x. and volume V (x) = (W − 2x)(L − 2x)x = W Lx − 2(L + W )x2 + 4x3 . Construct an open box from a rectangular piece of card board of length L and width W . 60 and 90 degrees). and x ≤ L/2.

As the function V (x) has only one local maximum in the interval. sin x) for some x ∈ [0. How broad should the folded up strips be. Modify the problem from above. π/2]. The width of the rectangle is π − 2x and its height is sin x.2. constructing a box with a round base from a circular piece of card board. We conclude that V has a local maximum at x= 1 (L + W ) − 6 L2 + W 2 − LW . you ﬁnd that A (x) = 0 if and only if tan x = π − 2x . 0). After a simple algebraic simpliﬁcation. Start out with an equilateral piece of card board with side length a. 2 . and to the left of which V is concave down. and fold up strips along the edges. to the right of which V is positive and V is concave up. You will get an open box whose base is an equilateral triangle. 0). Determine the rectangle of maximal area which can be placed between the x-axis and the graph of the function f (x) = sin x. The ﬁrst derivative of this function is A (x) = −2 sin x + (π − 2x) cos x. Repeat the previous example with speciﬁc numbers for the width and length of the piece of card board.9. OPTIMIZATION STORY PROBLEMS 95 The function has an inﬂection point at (W + L)/6. ♦ Its width will be W − 2x and its length L − 2x. the local maximum is the same as the absolute maximum. (π − x. Exercise 25. We need to ﬁnd the absolute maximum for this function for x ∈ [0. so that its area is A(x) = (π − 2x) sin x. Answer the original question: The box with the largest volume will have a height of x= 1 (L + W ) − 6 L2 + W 2 − LW . Make incisions at the corners. Exercise 24. sin x) and (π − x. (x. Convince yourself that the vertices of the rectangle should be (x. π/2]. What is the largest possible volume for a right circular cone of slant height a? Example 2. so that the volume of the box is maximal? Exercise 26. Solution: Draw a graph of sin x so that you can follow the discussion. Exercise 27.50.

It follows from the second derivative test that A(x) has a local maximum at x0 . (a) If f is concave up on I and has a local minimum at x0 . ∞). ♦ 9 π−2x 2 One possible argument is that tan x is increasing on the interval [0. then f assumes its absolute maximum at x0 .72066 and a height of sin x0 = . x3 We ﬁnd that f (x) = 0 if x = 1. (b) If f is concave down on I and has a local maximum at x0 . b] we could inspect the values of the function at the critical points and at a and b.652183. π/2). This tells us that A(x) assumes its absolute maximum at x0 . So these functions can intersect in only one point. Find the absolute minimum of the function f (x) = x + 1 x for x ∈ (0. The assumptions of this theorem are satisﬁed in many applied problems. ∞). We conclude that x0 is the only critical point of A(x).710462. Our next result allows us to do the same even if the interval is not closed and bounded. Example 2. A fairly good approximation of the zero of A (x) is x0 = . . then f assumes its absolute minimum at x0 . Solution: We calculate the ﬁrst and second derivative of f (x): f (x) = 1 − 1 x2 and f (x) = 2 . and that is decreasing. It allows us to decide whether a local extremum is also an abolute one.52.12218. So f has a local minimum at x = 1. You may calculate A (x). and that f (x) > 0 for all x in (0. ♦ To ﬁnd the absolute extrema of a continuous function on an interval of the form [a. Its area will be about 1. GLOBAL THEORY Find an approximate solution of the equation using Newton’s method or your calculator.51. Suppose f is deﬁned on an interval I. and f is concave up on (0. Apparently A(x) = 0 at the end points x = 0 and x = π/2 of the interval. Convince yourself9 that this is the only zero of A (x) for x ∈ [0. Theorem 2. π/2].96 CHAPTER 2. the ﬁnal answer to our problem is: The rectangle of maximal area which can be placed between the x-axis and the graph of the sine function will have a width of approximately π − 2x0 = 1. With this. Substituting x0 you will see that A (x0 ) < 0.51 tells us that the absolute minimum of the function is f (1) = 2. Theorem 2. ∞).

0). Suppose that a and b are positive. Exercise 30. .9. The total length of the fence is 600 meters. the other three sides by straight fences. What are the dimensions of a rectangle of this kind with maximal area? Exercise 32.. Draw a rectangle with one vertex at the origin (0. 5) lies on the line through the points (a.2. (b) the drum has no top (but a bottom) and the same material is used for the bottom and sides.2 m3 ) if (a) the drum has a bottom and a top. and (0. b). Determine the dimensions of the meadow so that its area is maximal. and that (2. Exercise 29. Find the dimensions of the warehouse that minimizes the construction cost. and the same material is used for the top. and one vertex on the line 3x + 5y = 15. meet at a right angle. OPTIMIZATION STORY PROBLEMS 97 Exercise 28.00 per linear meter. Exercise 36. Exercise 35. A rectangular warehouse will have 5000 m2 of ﬂoor space and will be separated into two rectangular rooms by an interior wall. Find the largest possible area for a rectangle with base on the x-axis and upper vertices on the curve y = 4 − x2 . Determine the length of the longest ladder that can be carried horizontally from one hallway into the other one. Find the dimensions of the cone if its volume is to be maximal. One side of a rectangular meadow is bounded by a cliﬀ. What should the slope of the line be. Exercise 31. Minimize the cost of the material needed to make a round drum with a volume of 200 liter (i.00 per linear meter and the cost of the interior wall is $ 600. one vertex on the positive x-axis. Find the dimensions of a right circular cone of minimal volume. The cost of the exterior walls is $ 1. Exercise 34. and (0. Inscribe a right circular cylinder into a right circular cone of height 25 cm and radius 6 cm. A right circular cone is inscribed in a sphere of radius R. . (a. one 8 feet wide and one 6 feet wide. Two hallways. so that the area of the triangle is minimal? Exercise 37. 0). Exercise 33.e. so that a ball of radius 10 centimeters can be inscribed. 0). bottom and sides. Consider a triangle in the plane with vertices (0. 0) in the plane. b).000. Find the dimensions of the cylinder if its volume is the be a maximum. one vertex on the positive y-axis.

000 per kilometer under ground and $70. and the left over material is recycled for half its value. A power line is needed to connect a power station on the shore line to an island 2 km oﬀ shore. 2. Find the minimal cost for laying the cable. Exercise 38. you may suppose that the shore line is straight. so that we can sketch a graph rather precisely. The width of the border at top and bottom is 15 cm.98 CHAPTER 2. The point on the coast line closest to the island is 6 km from the power station. Exercise 39. for all practical purposes. assuming you are allowed a ﬁxed amount of material? More speciﬁcally determine the ratio of radius and height which will maximize the volume.) Exercise 40. the same material is used for the top and bottom.000 under water. GLOBAL THEORY (c) the drum has a bottom and a top. Let us make a list of data which we may determine. Going through the following . y0 ) not on this graph. Design a roman window with a perimeter of 4 m which admits the largest amount of light. Suppose D(x) has a local minimum at x1 . Consider the distance D(x) between a point P (x) = (x. (A roman window has the shape of a rectangle capped by a semicircle. A rectangular banner has a red border and a white center. Exercise 42. so that its volume is maximal. Consider a box with a round base and no lid whose interior is subdivided into six wedge shaped sectors. and. What should be the dimensions of the banner if the area of the white area is to be maximized? Exercise 41. but the top and the bottom are cut out of squares. Which shape should it have. To lay the cable costs $40. f (x)) on the graph of a diﬀerentiable function f (x) and a point Q = (x0 . Then the tangent line to the graph of f at x1 intersects the line joining P (x1 ) and Q perpendicularly. and the material for the top and bottom is twice as expensive as the material for the sides. and along the sides 10 cm. (d) the situation is as in the previous case. The total area is 1 m2 .10 Sketching Graphs The techiques which we developed so far provide us with some valuable tools for graphing functions.

negative. (d) Find the ﬁrst derivative f (x) of f (x). resp. concave down. (b) Plot the zeros of the function. If you now draw a graph which exhibits all of the properties which you gathered in the course of the suggested program.10. (h) Decide at which critical points of f (x) the function has a saddle point or local extremum.e.. decide on which intervals the function is positive. Let us go through the program in an example. (f) Find the second derivative f (x) of f (x). Intervals on which f (x) is positive give you intervals on which f (x) is increasing.. If the function is given on a closed interval. the points where the concavity changes. resp.2. The zeros of f (x) provide you with the critical points of f (x). such as the y-intercept.. (e) Repeat (b) and (c) with f (x) in place of f (x). and keep track of the intervals on which the function is increasing. resp. i. then your graph will look very much like the graph of f (x). and whether it is a minimum or a maximum. (c) If possible. Intervals on which f (x) is positive give you intervals on which f (x) is concave up. Plot the critical points (x and y value). More importantly. Useful information for graphing a function: We call the function f (x).. plot the values at its endpoints. Find the inﬂection points of the function. If you cannot ﬁnd the zeros by analytical means. (a) Plot some points on the graph. decreasing. Plot the inﬂection points (x and y value). . try it numerically (Newton’s method). (g) Repeat (b) and (c) with f (x) in place of f (x). and keep track of the intervals on which the function is concave up. and intervals on which f (x) is negative give you intervals on which f (x) is concave down. SKETCHING GRAPHS 99 program is also a good review of the material which we developed in this chapter. the graph will have all of the essential features of the graph of f (x). and intervals on which f (x) is negative give you intervals on which f (x) is decreasing.

√ • f (x) is positive on the interval ((1 + 33)/4. Discuss the graph of the function f (x) = x4 − 2x3 − 3x2 + 8x − 4 for x ∈ [−3. −2) and (2. You should verify this by multiplying out the expression for f (x) in (2. and we ﬁnd: √ 1 f (x) = 4(x − 1) x − [1 + 33] 4 We conclude that: √ • f (x) is negative √ the interval [−3.100 CHAPTER 2. (b): As a polynomial. (1 − 33)/4]. 3]. Plot these x-intercepts. (a): Plot the y intercept of the function and its values at the end points of the given interval: f (−3) = 80. The only exceptional points are its zeros. 1) and (1. (1+ 33)/4) and f (x) is decreasing √ on [1. 1) and f (x) is increasing √ on [(1 − 33)/4. Solution: To make the discussion a little easier. 4 . 3]. (d): We calculate the derivative of f (x): f (x) = 2(x − 1)(x2 − 4) + (x − 1)2 2x = 2(x − 1)(2x2 − x − 4). (1 − 33)/4) and f (x) is decreason ing on [−3. we note that (2.53. 3]. we see that f (x) is positive on the intervals [−3. and then we used elementary algebra. √ • f (x) is positive on the interval ((1 − 33)/4.1). 1]. 2).1). In the ﬁrst step we applied the product rule. Having written f (x) as in (2. GLOBAL THEORY Example 2. and negative on (−2. we see right away that f (x) = 0 if and only if x = −2. They are (1 ± 33)/4. √ • f (x) is negative on the interval (1. the function f (x) is diﬀerentiable on the given interval. x = 1. f (0) = −4 and f (3) = 20. (c): Counting the signs of the factors of f (x). This allows us to factor the expression for f (x). (e): We use the quadratic formula to ﬁnd the zeros of the factor 2x2 −x−4 √ in the expression for f (x). √ 1 x − [1 − 33] . 3] and f (x) is increasing √ on [(1 + 33)/4.1). or x = 2. (1 − 33)/4].1) f (x) = (x − 1)2 (x2 − 4) = (x − 1)2 (x − 2)(x + 2). We based the calculation on the description of f (x) in (2.

(1 − 3)/2) and f (x) is concave on up on [−3.10. but we decided this already based on ﬁrst derivative behaviour in (e). (1 + 3)/2] √ • f (x) is positive on the interval ((1 + 3)/2. √ 1 x − [1 − 3] .37 and at x = has (1 + 3)/2 ≈ 1. (f): We rewrite the ﬁrst derivative as f (x) = 4x3 − 3x2 − 3x + 4. 2 2 Plot these points.21 & f ( ) ≈ −. 3] √ • f (x) √ inﬂection points at x = (1 − 3)/2 ≈ −. (h): At this point we could use the second derivative test to ﬁnd at which critical points the function has local extrema. The values of the function at its three critical points are approximately: √ √ 1 − 33 1 + 33 f( ) ≈ −12.29. SKETCHING GRAPHS 101 √ • f (x) has a critical point and local minimum at (1 − 33)/4 ≈ −1.2. and a critical point and √ local minimum at (1 + 33)/4 ≈ 1.37.54. 4 4 Plot these points. (1 − 3)/2] √ √ • f (x) is negative on the interval ((1 − 3)/2.39 & f (1) = 0 & f ( ) ≈ −.19. and ﬁnd f (x) = 12x2 − 12x − 6. The values of the function at its inﬂection points is approximately: √ √ 1− 3 1− 3 f( ) ≈ −7. 3] and f (x) is concave up √ on [(1 + 3)/2. 2 . (1 + 3)/2) and f (x) is √ √ concave down on [(1 − 3)/2.69. (g): We use the quadratic formula to ﬁnd the zeros on f (x) and factor it: √ 1 f (x) = 12 x − [1 + 3] 2 We conclude that: √ • f (x) is positive√ the interval [−3. a critical point and local maximum at x = 1.

Property Sign I1 I2 I3 I4 neg inc I5 neg dec I6 I7 I8 pos neg neg inc up neg neg pos dec up inc up inc up Monotonicity dec dec Concavity up up down down Table 2. ♦ Exercise 43.21 you see the graph of f on an even smaller interval. 4 2 √ 1− 3 I4 = . In addition. 2 √ √ 1 + 3 1 + 33 I6 = .1: Properties of the Graph In Figure 2. . discuss the function f (x) = (x − 1)(x − 2)(x + 2) = x3 − x2 − 4x + 4 on the interval [−3. −2] I2 = −2. f is concave down where f is negative. etc. GLOBAL THEORY Let us gather and organize our information. 2 4 √ 1 + 33 I7 = . We have shown it on a slightly smaller interval. We consider the interval: I1 = [−3. Showing all of the graph would show less clearly what happens near the intercept. and parts of the graphs of f and f . It should be understood. and inﬂection points.102 CHAPTER 2.1 2 We tabulate the which properties hold on which interval. 1− 1− √ √ 4 √ 1+ 3 I5 = 1. You can use them to see that f is decreasing where f is negative. In Figure 2. extrema. In analogy with the previous example. 33 √ 33 1 − 3 I3 = .2 4 I8 = [2. 3] .20 you see the graph of the function. The dots indicate the points which we suggests to plot. that at some end points of intervals the function is zero.5]. 2. ﬁnd the absolute extrema of this function. as the values at the endpoint a comparetively large.

ﬁnd the absolute extrema of this function. In analogy with the previous example. discuss the function f (x) = 2 sin x + cos 3x on the interval [0.2.10. Exercise 45. 2]. discuss the function f (x) = x3 − 3x + 2 on the interval [−2.20: The Graph Figure 2. In addition. f . . You may have to apply Newton’s method to ﬁnd zeros of f . 2π]. SKETCHING GRAPHS 103 10 10 5 5 -3 -2 -1 -5 1 2 3 -2 -1 -5 1 2 -10 -10 Figure 2. and f . ﬁnd the absolute extrema of this function. In addition.21: f . f . f Exercise 44. In analogy with the previous example.

104 CHAPTER 2. GLOBAL THEORY .

all functions whose derivative is f . so that for a non-negative function it makes sense to think of the integral as the area of the region bounded by the graph of the function.e. we only know the area of some simple regions. 3. it should have the following properties: • The area of a rectangle is the product of the lengths of its sides. and the lines x = a and x = b. It is denoted by b f (x) dx. the x-axis.1. b].1 Properties of Areas So far. To be concrete. The indeﬁnite integral of a function f is the family (set) of all antiderivatives of f . and ﬁnd the area of the region Ω bounded by the graph of f (x). Whatever concept of area we have in mind. 105 . the lines x = 1 and x = 5. consider the function f (x) = x2 e−x . We will denote the area of a region Ω by Area(Ω). and the x-axis. then the deﬁnite integral of f over the interval [a. a The deﬁnition is set up. i. like rectangles. For important classes of functions one may utilize deﬁnite integrals to construct antiderivatives.Chapter 3 Integration We will introduce the ideas of the deﬁnite and the indeﬁnite integral. shown in Figure 3. If it exists. Suppose that f is a function which is deﬁned and bounded on the interval [a. The Fundamental Theorem of Calculus relates deﬁnite integrals and antiderivatives. b] is a real number.

which is contained in Ω.1 has an area. then Area(Ω1 ) ≤ Area(Ω2 ).2 0. In Figure 3. and that the area of each of them is deﬁned. In Figure 3. The ﬁrst two principles tell us that Area(Rl ) = .1: f (x) = x2 e−x • Suppose that Ω1 and Ω2 are regions in the plane.4 0. If Ω1 ⊆ Ω2 .3 0.24.6 ≤ Area(Ω) ≤ Area(Ru ) = 2. INTEGRATION 0. then the area of the union Ω1 ∪ Ω2 of Ω1 and Ω2 is deﬁned. and Area(Ω1 ∪ Ω2 ) = Area(Ω1 ) + Area(Ω2 ). that the region under the graph shown in Figure 3. and that the area of each of them is deﬁned.106 CHAPTER 3.3 you see a rectangle Ru with area 2.1 1 2 3 4 5 6 Figure 3. • Suppose that Ω1 and Ω2 are regions in the plane.24 which contains Ω. From above principles one may derive another one.5 0.6. Suppose for a moment.2 you see a rectangle Rl with area . If the regions Ω 1 and Ω2 do not intersect. which occurs frequently in our upcoming constructions: .6 0.

3. . Then Area(R) is deﬁned.6 0.5 0. xj ] with 1 ≤ j ≤ n.3 0. Rn and any two of them intersect at most in an edge.2 Partitions and Sums We like to reﬁne the approach to calculating areas of regions which we started in the previous section.1 0. Rn : Area(R) = Area(R1 ) + · · · + Area(Rn ). and then we add up what we get over the individual intervals. A partition of an interval [a. .2: A rectangle Rl contained in Ω Figure 3.2 0.3 0. 3.3: A rectangle Ru containing Ω • Suppose the region R in the plane is the union of a ﬁnite number of rectangles R1 . b] is partitioned into n intervals [xj−1 . . .1 1 2 3 4 5 6 1 2 3 4 5 6 Figure 3. We do so by partitioning the interval before applying the ideas from above.6 0. . PARTITIONS AND SUMS 107 0. .2. such that a = x0 ≤ x1 ≤ · · · ≤ xn−1 ≤ xn = b.4 0. b] is of a collection is points {xj | 0 ≤ j ≤ n}.2 0. and it is equal to the sum of the areas of the regions R1 . . . .4 0. The interval [a.5 0.

108

CHAPTER 3. INTEGRATION

3.2.1

Upper and Lower Sums

As before, f denotes a function which is deﬁned and bounded on [a, b]. On each interval we pick numbers mj and Mj , such that mj ≤ f (x) ≤ Mj We deﬁne the lower sum to be (3.1) Sl = m1 (x1 − x0 ) + m2 (x2 − x1 ) + · · · + mn (xn − xn−1 ). for all x ∈ [xj−1 , xj ].

and the upper sum to be (3.2) Su = M1 (x1 − x0 ) + M2 (x2 − x1 ) + · · · + Mn (xn − xn−1 ).

These sums depend on the choice of partition and the choices for the mj and Mj .

0.6 0.5 0.4 0.3 0.2 0.1

0.6 0.5 0.4 0.3 0.2 0.1

1

2

3

4

5

6

1

2

3

4

5

6

Figure 3.4: A union of rectangles contained in Ω

Figure 3.5: A union of rectangles containing Ω

Let us return to the example of the function f (x) = x2 e−x on the interval [1, 4]. In the computation of the lower sum we use the partition x0 = 1 < x1 = 2 < x2 = 3 < x3 = 4 < x4 = 5 of the interval. We also pick m1 = .35, m2 = .43, m3 = .28 and m4 = .16. This leads to a lower sum Sl = 1.22. In the computation of the upper sum we use the partition x0 = 1 < x1 = 3 < x2 = 4 < x3 = 5

3.2. PARTITIONS AND SUMS

109

of the interval. We also pick M1 = .55, M2 = .45 and M3 = .3. This leads to an upper sum Su = 1.85. The mj and Mj represent the heights of the rectangles in Figures 3.4 and 3.5, and we trust these ﬁgures to show that mj ≤ f (x) and f (x) ≤ Mj on the respective interval. As before, let Ω denote the region under the graph. Then the union of the rectangles shown in Figure 3.4 is contained in Ω, and the union of the rectangles shown in Figure 3.5 contains Ω. Thus, if Ω has an area, the our principles tell us that Sl = 1.22 ≤ Area(Ω) ≤ Su = 1.85. In fact the only number greater or equal to all lower sums and smaller or equal to all upper sums is 5 − 37 , and this will be the area of the region e e5 Ω. Here e is the Euler number. Example 3.1. Let us ﬁnd upper and lower sums for the function f (x) = x3 − 7x2 + 14x − 8 for x ∈ [.5, 4.5]. In contrast to the function in the previous example, this function is not non-negative.

4 3 2 1 1 -1 -2 -3 2 3 4

4 3 2 1 1 -1 -2 -3 2 3 4

Figure 3.6: Rectangles for calculating an upper sum.

Figure 3.7: Rectangles for calculating a lower sum.

Solution: For the purpose of calculating an upper sum, we partitioned the interval [.5, 4.5] using the intermediate points x0 = .5, x1 = 1.1, x2 = 2.4, x3 = 3.8, and x4 = 4.5. As numbers Mi (so that Mi ≥ f (x) for x ∈ [xi−1 , xi ])

110

CHAPTER 3. INTEGRATION

we chose M1 = .3, M2 = .7, M3 = −.9, and M4 = 4.4. These data are shown in Figure 3.6. With these choices, the upper sum is Su = .3(1.1 − .5) + .7(2.4 − 1.1) + (−.9)(3.8 − 2.4) + 4.4(4.5 − 3.8) = 2.91. In Figure 3.6 you see four rectangles. Their areas are combined to calculate the upper sum. The areas of the ones above the x-axis are added, the ones below the axis are subtracted, in accordance with the sign of the Mi . In the calculation of the lower sum we partitioned [.5, 4.5] using x0 = .5, x1 = .8, x2 = 2.3, x3 = 4.2, and x4 = 4.5. As numbers mi (so that mi ≤ f (x) for x ∈ [xi−1 , xi ]) we chose m1 = −2.7, m2 = −.8, m3 = −2.2, and m4 = 1.3. These data are shown in Figure 3.7. With these choices we calculate a lower sum of Sl = −2.7(.8 − .5) + (−.8)(2.3 − .8) + (−2.2)(4.2 − 2.3) + 1.3(4.5 − 4.2) = −5.8. In Figure 3.7 you see four rectangles. Their areas are combined to calculate the lower sum. The areas of the ones above the x-axis are added, the ones below the axis are subtracted, in accordance with the sign of the mi . In summary, you see that we still combine areas of rectangles in the calculation of the upper and lower sum, only that, depending on the sign of the Mi or mi , these rectangles are either above or below the x-axis, and depending on this, their areas are either added or subtracted. ♦ Let us make a simple albeit important observation: Theorem 3.2. Let f be a function which is deﬁned and bounded on a closed interval [a, b]. Let Sl be any lower sum of f and Su any upper sum. Then Sl ≤ Su . Let us repeat the statement of the theorem to emphasize its meaning. Whichever partition of the interval [a, b] and whichever mi we use in the calculation of the lower sum Sl and whichever partition of the interval and whichever Mi we use in the calculation of the upper sum Su , the lower sum is always smaller or equal to the upper sum. To see this, one reﬁnes the partitions for the upper and lower sum computation so that they become the same. Then one notes that mi ≤ Mi for all i.

3] √ and x2 = π ∈ [ 3. √ √ In the two interval of this subdivision we pick the points x1 = 2 ∈ [1. xj ]. The sum of the areas of these rectangles is the Riemann sum.5 0.4 0. In each subinterval. We leave it to the reader to contemplate Proposition 3.2 0. b].2 Riemann Sums Suppose once again that f (x) is a function which is deﬁned on the interval [a. Let Sl be any lower sum of f . let us return to the example f (x) = x2 e−x on the interval [1. 5]. Pick once more a partition a = x0 ≤ x1 ≤ · · · ≤ xn−1 ≤ xn = b of the interval. 5]. There are two rectangles.2. PARTITIONS AND SUMS 111 3.749741. Let f be a function which is deﬁned and bounded on a closed interval [a. pick a point xj ∈ [xj−1 . Let us use the partition √ x0 = 1 < x1 = 3 < x2 = 5. b]. Su any upper sum.3.8 you see the picture illustrating the computation.2.6 0. and their heights are f (x1 ) and f (x2 ).3 0.3) SR = f (x1 )(x1 − x0 ) + f (x2 )(x2 − x1 ) + · · · + f (xn )(xn − xn−1 ). As Riemann sum we obtain SR = f (x1 )(x1 − x0 ) + f (x2 )(x2 − x1 ) ≈ 1.8: Representing a Riemann Sum To be more concrete. their bases are the intervals in the subdivision. . 0.1 1 2 3 4 5 6 Figure 3. Then we deﬁne the Riemann Sum (3. In Figure 3. and SR any Riemann sum. Then Sl ≤ SR ≤ Su .3.

There exists a real number Y . but typically equivalent. It is also denoted by b f (x) dx. In this case. Deﬁnition 3. we always have that Sl ≤ Su . such that S l ≤ Y ≤ Su for all lower sums Sl and all upper sums Su of f . if the function is non-negative.4. b].3 Limits and Integrability The idea is to reﬁne the partitions in our previous construction. Call it Yu . The speciﬁcs depend on which sums we are working with. 3. Let f be a function which is deﬁned and bounded on a closed interval [a. construction we should Y the Darboux integral. . Apparently. whatever choices we make in the calculation of lower and upper sums Sl and Su . We are now prepared to deﬁne the concept of integrability of a function. The set of all upper sums of f has a greatest lower bound. then we say that f is integrable over the interval [a.1 The Darboux Integral and Areas As we discussed earlier. Then Y is any number such that Yl ≤ Y ≤ Yu .112 CHAPTER 3. a 1 To distinguish it from the result of a diﬀerent. If there is exactly one number Y . INTEGRATION 3. Idea of Proof. so that in the limit our sums can be justiﬁably called the area of the region under the graph. the number Y is called the integral1 of f for x between a and b.3. such that Sl ≤ Y ≤ Su for all lower sums Sl and upper sums Su of f . Let f be a function which is deﬁned and bounded on a closed interval [a.5. To deduce the theorem from the completeness of the real numbers. one observes that the set of all lower sums of f has a least upper bound. Theorem 3. b]. A crucial additional fact is stated in the next result. Yl ≤ Yu . Call it Yl . b].

. On the other hand. For completeness sake and later use. all subintervals have the same length 1/n. let us explain what happens when a function is not integrable..6. M3 = f (x3 ) = 3 n 2 . xj ] because f (x) is increasing on [0. 1]. So. a function over a closed interval [a... Apparently. j and Mj = f (xj ) = n in general. 6 Su = M1 (x1 − x0 ) + M2 (x2 − x1 ) + · · · + Mn (xn − xn−1 ) 1 2 1 2 2 1 n = × + × + ··· + n n n n n 1 2 = 1 + 22 + · · · n2 n3 n(n + 1)(2n + 1) = 6n3 1 1 1 = + + 2 3 2n 6n For the lower sums we pick m1 = f (x0 ) = 0. between all upper and lower sums. b] is not integrable if and only if the exists a positive number D such that Su − Sl ≥ D for any lower sum Su and any upper sum Su . Without proof. Mj ≥ f (x) for all x ∈ [xj−1 . m3 = f (x2 ) = 2 n 2 . M2 = f (x2 ) = 2 2 n 2 . and with this an entire interval. a function is integrable if for every positive number D there is an upper sum Su and a lower sum Sl such that Su − Sl < D. In this case there are at least two diﬀerent numbers. Example 3. 1]. m2 = f (x1 ) = 1 n 2 2 × 1 n . lower sums. . we use that 12 + 22 + 32 + · · · + n2 = We calculate the upper sum n(n + 1)(2n + 1) .. n n n n This is an equidistant partition of the interval [0. .3. For the upper sums we pick M1 = f (x1 ) = 1 n 2 . and integrability for the function f (x) = x2 on the interval [0.3. Solution: Fix a natural number n and set 1 2 n−1 n x0 = 0 < x1 = < x2 = < · · · < xn−1 = < xn = = 1. Explore upper sums. 1]. LIMITS AND INTEGRABILITY 113 Remark 8.

9: Rectangles for calculating a lower sum.2 1. and mj = f (xj−1 ) = j−1 n in general.25 -0.8 1 1.75 0.25 1 0. so that Sl ≤ Y ≤ Su for all natural numbers n. b].25 1 0. bounded. 1] and that 1 0 x2 dx = 1 .2 Figure 3.75 0. We do not only see that Sl ≤ 1 ≤ Su .5 0.10.2 0. then we say that the region Ω has an area and b Area(Ω) = a f (x) dx. . Let f be a function which is deﬁned. but also that Y = 1/3 is the only real 3 number.2 0.7.4 0.2 0.10: Rectangles for calculating an upper sum. 2 Figure 3. and nonnegative on a closed interval [a. Let Ω be the region bounded by the graph of f .114 CHAPTER 3. ♦ 3 We motivated our introduction of upper and lower sums by our quest to deﬁne the concept of area. If f is integrable over this interval. INTEGRATION 1.25 -0. Using the expressions for Su and Sl you see that Su − Sl = 1/n. the x-axis.6 0.8 1 1. According to the deﬁnition this means.9 and 3.5 0. Our answer is formulated as a Deﬁnition 3.4 0.2 0. that f (x) = x2 is integrable over the interval [0.6 0. The resulting lower sum is Sl = 1 1 1 − + 2 3 2n 6n For n = 5 you see the rectangles whose areas are the summands in the lower and upper sums in Figures 3. and the lines x = a and x = b.

♦ 3 3. in other words. . b].3. Suppose the function f (x) is deﬁned on [a. b] and a function f (x) deﬁned on it. Deﬁnition 3. then we say that f is Riemann integrable over [a. b]. For example. Consider an interval [a.8 (Limit for Riemann Sums). because there are a lot of choices which we make to deﬁne such a sum. such that |L − SR | < whenever |P| < δ. For an integrable function there is exactly one real number between the lower and upper sums. the x-axis. We deﬁne the norm of the partition P to be |P| = max{xj − xj−1 | 1 ≤ j ≤ n}. If the limit of the SR exists. xj ]. so this is the only number which we can call the area of Ω. LIMITS AND INTEGRABILITY 115 The upper and lower sum were constructed such that if there is any justiﬁcation to assigning an area to Ω then Sl ≤ Area(Ω) ≤ Su . b] up into smaller interval [xj−1 . and set SR = f (x1 )(x1 − x0 ) + f (x2 )(x2 − x1 ) + · · · + f (xn )(xn − xn−1 ).3. xj ]. and the lines x = 0 and x = 1 is 1 Area(Ω) = 0 1 x2 dx = . We say that L = lim SR |P|→0 if for all > 0 there exists a δ > 0. and write b L = lim SR = |P|→0 a f (x) dx. the norm of P is the length of the longest of the intervals [xj−1 . We want to consider a limit Riemann sums. which broke [a.3. We picked a partition P : a = x0 ≤ x1 ≤ x2 ≤ · · · xn−1 ≤ xn = b. the area of the region Ω bounded by the graph of the function f (x) = x2 . call L the Riemann integral of f .2 The Riemann Integral Earlier we introduced the idea of a Riemann sum. In each of the subintervals we picked a point xj ∈ [xj−1 . xj ]. This is trickier than for functions.

116

CHAPTER 3. INTEGRATION

Thus L = lim SR if we can force SR to be close to L, as close as we like, by making the partition ﬁne, by making each subinterval no longer that some number. It is worth pointing out and not very diﬃcult to show the following proposition. Proposition 3.9. Suppose the function f is deﬁned on the interval [a, b]. Then f is Riemann integrable if and only if it is Darboux integrable. If deﬁned, the Riemann and the Darboux integral are the same.

3.4

Integrable Functions

We like to provide a supply of integrable functions. Our ﬁrst result is typically proved in an analysis course. Theorem 3.10. Suppose f is deﬁned and continuous on [a, b]. Then f is integrable over [a, b]. According to this theorem, polynomials are integrable over any interval of the form [a, b]. Rational functions (i.e., functions of the form p(x)/q(x) where p(x) and q(x) are polynomials) are integrable over intervals of the form [a, b] as long as q does not vanish anywhere on the interval. The trigonometric functions (sin, cos, tan, cot, sec, and csc) are integrable on intervals where the functions are deﬁned. Arbitrary powers of a variable, f (x) = xα , are integrable. One just needs to make sure that the function is deﬁned on the interval [a, b]. For any real number α it suﬃces to assume that a > 0. For any real α ≥ 0, it suﬃces to assume a ≥ 0. For rational numbers α = p/q, where p and q are integers and q is odd, it suﬃces to assume 0 ∈ [a, b]. For non-negative integers α no assumption needs to be made on a and b. Just making sure that the resulting functions are deﬁned everywhere on [a, b], the functions just mentioned may be added, subtracted, multiplied, divided, and composed, and one still ends up with integrable functions. Let us introduce another class of functions for which we can prove that they are integrable. Deﬁnition 3.11. Suppose f (x) is a function. We say that f (x) is nondecreasing if f (x1 ) ≤ f (x2 ) whenever x1 and x2 are in the domain of f (x) and x1 ≤ x2 . We say that f (x) is non-increasing if f (x1 ) ≥ f (x2 ) whenever x1 ≤ x2 . Proposition 3.12. Let [a, b] be a closed interval and let f be deﬁned and non-increasing or non-decreasing on [a, b]. Then f is integrable on [a, b]. In particular, monotonic (increasing or decreasing) functions are integrable.

3.4. INTEGRABLE FUNCTIONS

117

Proof. We will use Darboux integrability. Let us assume that the function f is non-decreasing on the interval. The non-increasing case is left as an exercise. Take any partition of the interval: a = x0 < x1 < · · · < xn = b. The reader may justify why we can use the same partition in the computation of the upper and lower sum. For i = 1, . . . , n we set mi = f (xi−1 ) Then, because f is non-decreasing, mi ≤ f (x) ≤ Mi for all x ∈ [xi−1 , xi ]. & Mi = f (xi ).

We use the mi and Mi to compute upper and lower sums. Let ∆ be the largest value of the xi − xi−1 . Then Su − Sl = [M1 (x1 − x0 ) + · · · + Mn (xn − xn−1 )] −[m1 (x1 − x0 ) + · · · + mn (xn − xn−1 )] = (M1 − m1 )(x1 − x0 ) + · · · + (Mn − mn )(xn − xn−1 ) ≤ [(M1 − m1 ) + (M2 − m2 ) + · · · + (Mn − mn )] ∆ = (Mn − m1 )∆ = [f (b) − f (a)]∆ The inequality in the computation follows from the choice of ∆. The second to last equality follows because Mi−1 = mi for all i = 2, . . . , n. Many terms in the computation cancel. Given any positive number D, we can make the partition ﬁne enough so that [f (b) − f (a)]∆ < D. According to our Remark 8 this means that f is integrable over the interval, as we claimed. We illustrate the steps in the proof in a concrete example. In Figure 3.11 you see the upper and lower sum. The lower sum is the sum of the areas of the darkly shaded rectangles. The upper sum is the sum of the areas of the lightly and darkly shaded rectangles. The diﬀerence between the upper and the lower sum is the sum of the lightly shaded rectangles shown in Figure 3.12. We can combine these areas by sliding the rectangles sideways so that they form one column. Its height will be f (b) − f (a). Its width may vary, but in the widest place it is no wider than ∆, the width of the largest interval in the partition of [a, b]. That means, the diﬀerence between the upper and the lower sum is at most [f (b) − f (a)]∆. As above, we conclude that the function is integrable.

118

CHAPTER 3. INTEGRATION

1.2 1.2 1 1 0.8 0.8 0.6 0.6 0.4 0.4 0.2 0.2 0.2 0.2 -0.2 0.4 0.6 0.8 1 -0.2 0.4 0.6 0.8 1

Figure 3.11: Rectangles for calculating a lower and an upper sum.

Figure 3.12: Rectangles for calculating the diﬀerence between an upper and a lower sum.

Remark 9. There are functions which are not integrable over any interval of the form [a, b] with a < b. Remark 10. Here we only discuss integrability of function over closed ﬁnite intervals, i.e., intervals of the form [a, b]. The discussion of integrability of functions over intervals which are not of this form, e.g., half-open intervals like [a, b) or unbounded closed intervals like [a, ∞), requires additional ideas and techniques which we are not ready to discuss yet.

3.5

Some elementary observations

In spite of our success calculating some integrals using upper and lower sums and the deﬁnition, this is certainly not the way to go in general. To integrate “well behaved” functions we want a theory which allows us to calculate integrals more easily. We have to develop a few basic tools. These are fairly straight forward consequences of the deﬁnition of the integral. Proposition 3.13. If the function f is deﬁned at a, then

a

(3.4)

a

f (x) dx = 0

Proof. The reader should contemplate the proposition.

b] if and only if it is integrable over the intervals [a. b]. b]. Idea of Proof.3.5) a f (x) dx + c f (x) dx = a f (x) dx. We can also extend Theorem 3. Suppose that f is deﬁned on an interval [a.15.3) x→x+ j−1 lim f (x) and x→x− j lim f (x). exist and are ﬁnite. Idea of Proof. then so does the other one. According to Proposition 3. Corollary 3. Then f is integrable on [a. b]. and f a function which is deﬁned on the interval.14. b]. Let [a.17. we can change the deﬁntion of the function at a point or two and make it continuous.10.10.16. As an immediate consequence of Propositions 3. We call f piecewise continuous if there is a partition a = x0 < x1 < · · · < xn−1 < xn = b such that f is continuous on the open intervals (xj−1 . Implicitly in the formulation of the proposition is the statement that f is integrable over [a. Then c b b (3. b] be a closed interval. c a point between a and b. On each of these smaller intervals. Use c as one of the points in the partition. b]. xj ] separately. This changes neither the integrability nor the value of the integral. and consider it over each of the intervals [xj−1 . So the assertion follows from Theorem 3. . If f is a piecewise contiuous function on [a.12 and 3.14 we ﬁnd Corollary 3.14 we may break the problem up.5) exists. then f is integrable on [a. Suppose that we can partition the interval into a ﬁnite number of intervals such that f is non-increasing or non-decreasing on each of them. The remaining details are left to the reader.5. Let f be deﬁned on the interval [a. xj ) for all 1 ≤ j ≤ n. Deﬁnition 3. b]. SOME ELEMENTARY OBSERVATIONS 119 Proposition 3. c] and [c. and the one-sided limits (see Section 1. If one of the sides of Equation (3.

20. then b a f (x) dx ≥ 0. Then f +g and cf are integrable over [a. Then the absolute value of f is integrable over [a. b]. This deﬁnition is convenient and consistent with what we have said so far about the integral. Suppose that f and g are integrable over the interval. Let f be deﬁned and integrable on the interval [a. If f is integrable over [a. and b (3. b] and b b b (f (x) + g(x)) dx = a a f (x) dx + a g(x) dx and b b cf (x) dx = c a a f (x) dx. Proof. b]. Using the deﬁnition of the integral it is not diﬃcult to show: Proposition 3. b] be a closed interval and f integrable over [a. Let [a. b]. The approach to integrals via lower and upper sums b could also be generalized to include integrals a where b < a. INTEGRATION Deﬁnition 3. b].22.18. and f (x) ≥ 0 for all x ∈ [a. Use that f (x) = g(x) − h(x) ≥ 0 for all x ∈ [a. b]. leading to exactly this formula.6) a f (x) dx ≤ a b |f (x)| dx. . If h and g are integrable over [a. Let [a. Then b a f (x) dx = − b a f (x) dx. We mention a few useful estimates for integrals. b]. though a bit tricky.120 CHAPTER 3. b]. b]. then b a g(x) dx ≥ a b h(x) dx. The proof is left to the reader.21. Corollary 3. b] be a closed interval and c a scalar.19. Proposition 3. and g(x) ≥ h(x) for all x ∈ [a. The proof of this proposition is elementary. Proof. Proposition 3.

We decompose the region between the x-axis and the graph into the part Ω + above the x-axis and the part Ω− below it. Ω+ and Ω− .3. .6 Areas and Integrals Let us return to the relation between areas and integrals. We deﬁne two functions: f + (x) = f (x) if f (x) ≥ 0 0 if f (x) ≤ 0 and f − (x) = f (x) if f (x) ≤ 0 0 if f (x) ≥ 0 It is elementary. to show that the integrability of f (x) implies the integrability of f + (x) and f − (x). we have: Proposition 3. the x-axis. b] and Ω the set of points which lie between the graph of f (x) and the x-axis for a ≤ x ≤ b. though a bit tricky. then the areas of the regions Ω+ and Ω− are deﬁned2 and b (3. y) in the plane for which a ≤ x ≤ b and 0 ≤ y ≤ f (x). b]. what happens if f (x) is not non-negative? Let f be a function which is deﬁned and bounded on a closed interval [a. Apparently.8) a f (x) dx = a f + (x) dx + a b f − (x) dx. Only then have we addressed the question of it having an area. b f + (x) dx. then b Area(Ω) = a f (x) dx.23. According to Deﬁnition 3. If Ω is the area bounded by the graph of f (x). Suppose f (x) is a non-negative integrable function over an interval [a.7) a f (x) dx = Area(Ω+ ) − Area(Ω− ).6. We decompose Ω into the union of two sets. If f is integrable.7 we have (3. AREAS AND INTEGRALS 121 3. Speciﬁcally. Idea of Proof. The question is. Ω+ consist of those points (x. then you have to ﬂip the region Ω− to lie above the x-axis. so that the additivity of the integral implies that b b (3. and Ω− of those points for which a ≤ x ≤ b and f (x) ≤ y ≤ 0. Making use of this notation. f = f + + f − . Then Ω is the union of the sets Ω+ and Ω− . and the lines x = a and x = b.9) Area(Ω+ ) = a 2 If you want to be formal.

and suppose that f has an antiderivative. Remember that any antiderivatives F1 and F2 of a function f on an interval I diﬀer only by a constant (see Corollary 2. Diﬀerent values for c result in diﬀerent functions. there exists a constant c.122 CHAPTER 3. In this expression c stands for an arbitrary constant.6 we called a function F (x) with domain I an antiderivative of f (x) if F (x) = f (x). Let f be a function which is deﬁned on an interval I. Deﬁnition 3.11) as a set of functions.e.5).7 Anti-derivatives Consider a function f (x) with domain I. Given a function f and an antiderivative F of it. Allowing all real numbers as possible values for c. In Deﬁnition 2.7 and our elementary properties of the integral we have (3. so Area(Ω− ) = Area(−Ω− ). i. The function −f − (x) is non-negative. Our claim follows now by substituting the results in (3. π/2 3.9) and (3.8). such that F1 (x) = F2 (x) + c for all x ∈ I. and −Ω− is bounded by the graph of −f − (x). INTEGRATION Let −Ω− be the area obtained by ﬂipping Ω− up.24. For example. Having an anti-derivative of a function will (typically) make it easy to integrate it over a closed interval. we take its mirror image along the x-axis. −π/2 sin x dx = 0 because the graph bounds congruent regions above and below the x-axis.. This process does not change areas.10) Area(Ω− ) = Area(−Ω− ) = a b −f − (x) dx = − a b f − (x) dx. we understand the the right hand side of (3. and the lines x = a and x = b. According to Deﬁnition 3. the x-axis.10) into (3. we typically write (3. The constant c in the expression is referred to as integration constant. In other words. It is denoted by f (x) dx.11) f (x) dx = F (x) + c. . The set of all antiderivatives of f is called the indeﬁnite integral of f .

3 on page 63 to come up with ideas for antiderivatives. E. an additional idea is required before we can see the antiderivative. or pick up some new ideas as you go along. 3 Occasionally. E.g.25.. Then we can write down the indeﬁnite integral of f in the form F (x) + c. using the trigonometric identity cos2 x = (1 + cos 2x)/2. ♦ We shall explore additional ideas for ﬁnding antiderivatives at a later.7. Given a function f (x) we might know or guess a function F (x). 5 5x2 − 2 cos x dx = x3 − 2 sin x + c.g. The reader may practice ﬁnding some antiderivatives for the functions in the next exercise. As you go through them you are expected to learn.20)). You can check the correctness of your guess by diﬀerentiation. we ﬁnd that cos2 x dx = 1 2 (1 + cos(2x)) dx = 1 1 x + sin(2x) + c. it is easy to produce more examples.3. such that F (x) = f (x). 2 2 Using a diﬀerent trigonometric identity we ﬁnd (1 + cot2 x) dx = csc2 x dx = − cot x + c. 1 dx = x + c √ x dx = 2 3/2 x +c 3 1 x dx = x2 + c 2 1 xn dx = xn+1 (n = −1) n+1 sec2 x dx = tan x + c sec x tan x dx = sec x + c √ dx = arcsin x + c 1 − x2 sin x dx = − cos x + c cos x dx = sin x + c dx = arctan x + c 1 + x2 Using the linearity of the diﬀerentiation (see the diﬀerentiation rules in (1. . You may want to consult Table 1. Here are some examples. ANTI-DERIVATIVES 123 Example 3.

Let a ∈ I. For example.27 (Fundamental Theorem of Calculus). . Theorem 3. Then f (x) = for all x ∈ I. Theorem 3. More speciﬁcally. and the grand conclusion of our discussion of antiderivatives is the Fundamental Theorem of Calculus. b] and that F is an antiderivative of f .124 CHAPTER 3. INTEGRATION Exercise 46. Continuous functions. Then b a d dx x f (t) dt a f (x) dx = F (b) − F (a). so that the Fundamental Theorem of Calculus tells us that π 0 sin x dx = − cos(π) − (− cos(0)) = −(−1) − (−1) = 2. Find the following indeﬁnite integrals: 1 dx x3 csc2 x dx (1 + tan2 x) dx csc x cot x dx sin2 x dx sec2 (3x) dx (a) (b) (c) (d) (e) (f) 3 dx (x + 4) dx (x2 − 5) dx cos 2x dx (3 + x)3 dx (3 + 2x)5 dx (g) (h) (i) (j) (k) (l) (m) (n) (o) (p) (q) (r) ex/3 dx x2 2x dx +1 (4 − 3x)5 dx cos(4 − 3x) dx 2x dx (x2 + 3)2 x sec2 (x2 + 5) dx 3. have antiderivatives.8 The Fundamental Theorem of Calculus Our ﬁrst result provides us with a large class of functions which have antiderivatives. deﬁned over intervals.26. suppose that a function f is deﬁned and continuous over the interval I. The major tool for calculating integrals. F (x) = − cos x is an antiderivative of f (x) = sin x. Suppose that f is a continuous function over a closed interval [a.

3. we also use the notation shown in the following example: x3 − 5x2 + 2x − 8 where p(x) = x3 − 5x2 + 2x − 8. One commonly uses the notation b F (x) a = F (b) − F (a). If there are ambiguities due to the length of the expression to which this construction is applied. .8. This is quite convenient.g. 3 Other examples are π/4 π/4 sec x tan x dx = sec x 0 0 = √ 2−1 and π/3 π/4 csc x cot x dx = − csc x π/3 π/4 = √ −2 3 3 √ √ √ 2 3 − (− 2) = 2 − . E. we calculate that 3 −2 5 3 = p(5) − p(3) (x2 − 2x + 5) dx = x3 − x2 + 5x 3 3 = −2 95 . Using this notation. THE FUNDAMENTAL THEOREM OF CALCULUS 125 As another example. note that F (x) = tan x is an anti-derivative of f (x) = sec2 x. 3 The reader is invited to practice a few examples. we write π sin x 0 = sin π − sin 0. Remark 11 (Notational Convention). so that the Fundamental Theorem of Calculus tells us that π/4 0 sec2 x dx = tan(π/4) − tan(0) = 1..

1 Some Proofs Because of their importance. Cauchy’s Theorem (see its application in Corollary 2.12) H(x) = a f (t) dt = F (x) + c We can ﬁnd out the value for c by substituting x = a in this equation.26. Essentially. Using this calculation of c and substituting x = b in (3. INTEGRATION Exercise 47.5) tells us that F and H diﬀer by a constant. we ﬁnd that a f (t) dt = 0 = F (a) + c a or c = −F (a). For some constant c and all x ∈ I: x (3. Let F (x) be any anti-derivative x of f (x) on I. . In particular.126 CHAPTER 3.26 and the Fundamental Theorem of Calculus. Evaluate the following deﬁnite integrals: 1 π (a) 0 2 (3x + 2) dx 6−t dt t3 1 5 √ 2 x − 1 dx 2 0 (g) 0 π 1 cos x dx 2 cos(x/2) dx (b) (c) (d) (h) 0 2 (i) (j) 0 1 (t3 − t2 ) dt csc x cot x dx −2 π/2 |x2 − 1| dx cos2 x dx sin2 (2x) dx sec2 x dx π/4 π/2 (e) π/6 −1 (k) 0 π/4 (f) −1 7x dx 6 (l) 0 3. we obtain b a f (t) dt = F (b) − F (a). and H(x) = a f (t) dt the one provided by Theorem 3. we like to prove Theorem 3. F (x) = H (x) = f (x).12). as claimed.26.8. Proof of the Fundamental Theorem of Calculus. the desired result is an easy consequence of Theorem 3. In particular.

We omit (leave to the reader) the modiﬁcations of the proof which are required in the case where x is an endpoint of I. The points c and d may not be uniquely determined by h. It follows from (3.10 that x F (x) = a f (t) dt exists. THE FUNDAMENTAL THEOREM OF CALCULUS 127 Proof of Theorem 3. but m and M are.26.13) f (c) = m(h) ≤ f (x) ≤ f (d) = M (h) for all t between x and x + h. So it is our task to show that F is diﬀerentiable at x.3. According to the Extreme Value Theorem (see Theorem 1. after adjusting the notation to ﬁt the current setting.21 that m(h) · h = x x+h m(h) dt ≤ x x+h f (t) dt ≤ x x+h M (h) dt = M (h) · h. it follows from Theorem 3. Continuity of f (x) implies that h→0 lim m(h) = f (x) = lim M (h). h→0 It follows from a pinching argument (see Proposition 1. so that x and x + h are both in I.26. and this is exactly what we needed to show. Using Theorem 1. . and that F (x) = f (x). and with this that m(h) ≤ 1 h x+h x f (t) dt ≤ M (h).17) there are points c and d between x and x + h. such that (3. x Here we assume that x is not an endpoint of I.8. Because we assumed continuity of f on the interval I.13) and Corollary 3.4) that x+h h→0 x lim f (t) dt = f (x). the task becomes to show that f (x) = F (x + h) − F (x) h x+h 1 = lim f (t) dt − h→0 h a h→0 lim x f (t) dt a = 1 h→0 h lim x+h f (t) dt.

. which can be applied to write down the steps in an integration using substitution eﬃciently. (2x − 3)3 dx = 1 2 (2x − 3)3 · 2dx = 1 (2x − 3)4 + c. Sometimes this method is helpful. Then f (g(x))g (x) is continuous as well. The variable for the functions f and F is often called u. Set F = f . Let F and g be functions which are deﬁned and diﬀerentiable on an interval I.14) f (g(x))g (x) dx = F (g(x)) + c. when applied correctly. according to the chain rule. It will give you at least some experience which you may then rely on in similar examples. and F (u) = u . rather success justiﬁes the means. Your success with this method depends greatly on experience. Working through the examples will teach you how to apply this method in some typical situations.e. . There are no general rules what substitution must be used. Setting u = g(x) we write du = g (x)dx. For example. Then. and conclude that (3. a way to use the notation. f (u) = u3 .9 Substitution In some cases it is not that easy to ‘see’ an antiderivative of the function one likes to integrate. other times it is not. g (x) = 2. i. You hope that you can ﬁnd an antiderivative for the simpliﬁed expression. INTEGRATION 3. We may take antiderivatives of both sides of our previous equation. will simplify the expression for the function you like to integrate. 8 4 Here we used g(x) = 2x − 3. We explain the method. dx Assume that f and g are continuous on I. The method is based on the chain rule for diﬀerentiation. practice. Let us give a few examples to illustrate how this method can be put to use.128 CHAPTER 3. Substitution is a method which. 4 There is a pattern. and this means in context that u = g(x). d F (g(x)) = f (g(x))g (x).

Then We calculate that t2 (t + 1)7 dt = = = = = (u − 1)2 u7 du = 2x.15) f (g(x))g (x) dx = f (u) du = F (u) + c = F (g(x)) + c. We may have to use a substitution and a trigonometric identity to solve We do not attach any particular meaning to the symbols dx and du in their own right. in the second one we ﬁnd the anti-derivative. so F = f . Suppose also that F is an anti-derivative of f .3. 10 9 8 Here we used the substitution u = t + 1. or du = 2xdx. We make use of this notation in our next example. Thought of as inﬁnitesimals or diﬀerentials. and in the third one we reverse the substitution. In the ﬁrst step of this calculation we carry out the substitution. 3 . 3 du dx We used the substitution u = x2 + 2. For example. Then the pattern for calculating an integral via substitution is (3.15). but this is beyond the scope of these notes. SUBSTITUTION 129 instead of g (x) = du/dx3 . these symbols have a meaning. Then du = dx and t = u − 1. The equation du = g (x)dx helps us to write down what happens when we perform the substitution as in the ﬁrst equality in (3. we calculate that x x2 + 2 dx = = = = 1 x2 + 2 · 2xdx 2 1 √ u du 2 1 3/2 u +c 3 1 2 (x + 2)3/2 + c.9. (u2 − 2u + 1)u7 du (u9 − 2u8 + u7 ) du 1 10 2 9 1 8 u − u + u +c 10 9 8 1 2 1 (t + 1)10 − (t + 1)9 + (t + 1)8 + c.

130 an integration problem: 2x sin2 (x2 + 5) dx = = = = CHAPTER 3. to work out an integral. so that du = 2xdx and the identity sin2 α = [1 − cos 2α]/2. and check the details: sec2 x tan x dx = = = = sec x · sec x tan x dx u du 1 2 u +c 2 1 sec2 x + c. Then dv = cos u du. Here is an example. 2 2 We used the substitution u = x2 + 5. Find the substitution which we used in the following computation. (x2 + 1) sin3 (x3 + 3x − 2) cos(x3 + 3x − 2) dx = = = = = 1 sin3 u cos u du 3 1 v 3 dv 3 1 4 v +c 12 1 sin4 u + c 12 sin4 (x3 + 3x − 2) +c 12 In the computation we used the substitution u = x3 + 3x − 2. INTEGRATION sin2 u du 1 [1 − cos 2u] du 2 1 1 u − sin 2u + c 2 2 1 1 (x2 + 5) − sin[2(x2 + 5)] + c. Then du = 3(x2 + 1) dx. 2 Sometimes we have to apply the method of substitution twice. . In a second substitution we set v = sin u. or more often.

we calculate dx = x2 + 2x + 5 dx = (x + 1)2 + 4 du 1 = arctan u2 + 4 2 x+1 2 + c. SUBSTITUTION 131 Here are two examples.9. which are important in the context of integrating rational functions. 3.1 Substitution and Deﬁnite Integrals Let us now explore how substitution is used to calculate deﬁnite integrals. To see this. 1 0 (x2 − 1)(x3 − 3x + 5)3 dx = 1 3 3 5 u3 du = 1 4 u 12 3 5 =− 136 .3. observe that f has an anti-derivative. Then b b g(b) g(b) f (g(x))g (x) dx = F (g(x)) a a = F (u) g(a) = g(a) f (u) du. In the ﬁrst example we assume that a = 0.9.16) a f (g(x))g (x) dx = g(a) f (u) du. and we use the subsitution x = au. 3 . The dx = adu. The ﬁrst identity is obtained as a combination of the Fundamental Theorem of Calculus and (3. we have b g(b) (3. and the third one is another application of the Fundamental Theorem of Calculus. The second one is obvious. Assuming as before that f and g are continuous on the interval [a.14). which we again denote by F . dx dx = x2 + a2 = = = Adding another idea. adu a2 u2 + a2 1 du a u2 + 1 1 arctan(u) + c a 1 x arctan +c a a We used the substitution u = x+1. Let us apply this formula in a few examples. and then we proceeded as in the previous example. b].

the u = 3. 7 We use the substitution u = x + 1. but they will lead to the same results. Incorporating one of our previous techniques. then u = 9.. and if x = π/4. To obtain the limits for the integral we calculate u(0) = 5 and u(1) = 3. You were told long time ago in school. there are other possible values for u. Then du = dx and x = u − 1. 4 We use the substitution x = sin u. if x = 0. so that the result of the calculation is hardly surprising. 3 1 3 4 0 1 We use the substitution u = cos x. If x = 0. Then 1 du = x dx and x2 = u − 1. Then du = (3x2 − 3) dx. For our given values of x. the area of one forth of the disk of radius 1. 1 0 1 − x2 dx = π/2 0 1 − sin2 u cos u du = π/2 0 cos2 u du = π . then u = π/2. When we . If x = 0. then u = 1. and 1 2 3 du = (x −1) dx. Then −du = sin x dx. If x = 0. i. Then dx = cos u du. π is the ratio of the circumference of a circle by its diameter. we calculate 2 0 x(x + 1)6 dx = 1 3 (u − 1)u6 du = 3 1 u7 − u6 du = 3554 . then u = 1. and if x = 8. √ 8 0 x3 x2 + 1 dx = 1 2 9 1 √ 1 (u − 1) u du = 2 9 1 u3/2 − u1/2 du = 596 . There is a more serious matter. In our calculation of the derivative of the sine and cosine functions we used the estimate that | sin h − h| ≤ h2 /2. then u = 2/2. Finally. The graph of f (x) = 1 − x2 is the northern part of a circle. that the area of this unit disk is π. 1] means that we calculated the area under this graph in the ﬁrst quadrant.e. then u = 0. For 2 √ the limits we calculate. and if x = 2. or did we assume the answer previously? By deﬁnition. √ Remark 12. INTEGRATION We used the substitution u = x3 − 3x + 5. Another example is √ √ √ π/4 2/2 2 1 3 2/2 1 2 2 cos x sin x dx = − u du = − u =− 1− . Similarly. 15 We use the substitution u = x2 + 1.132 CHAPTER 3. Using x ∈ [0. and if x = 1. then √ u = 1. Is the example genuine.

The graphs intersect in two points. Now you see the region between the two graphs whose area we want to calculate. Find the following integrals: dx 2x + 1 t dt 2 + 9)2 (4t √ t(1 + t2 )3 dt 2s ds 6 − 5s2 b3 x3 √ dx 1 − a4 x4 √ 3 π π/4 (a) (b) (c) (d) (e) (f) 0 x cos x dx √ x2 x + 1 dx x+3 √ dx x+1 sin2 (3x) dx π/2 2 (k) π/6 1/2 sec(2x) tan(2x) dx dx 4 + x2 0 sec2 x √ dx 1 + tan x √ 1 + sin x cos x dx r (g) (h) (i) (j) 0 (l) (m) (n) (o) cos2 x dx 0 r 2 − x2 dx 3. This idea can be generalized to the discussion of areas of regions between two graphs. We call the region Ω. π/4]. we solve the equation f (x) = x2 = g(x) = 1 − x2 .3. we used that |h| ≤ | tan h| for h ∈ [−π/4. This means. A typical proof of the latter inequality starts out by ﬁrst showing that the area of the unit disk is π.13. we assumed the result in the example.10 Areas between Graphs Previously we related the integral to areas of a region under a graph. we ﬁnd x2 = −1± 5 .10. we draw the two graphs. we did not derive it. AREAS BETWEEN GRAPHS 133 showed this. see Figure 3. √ After squaring the equation and solving it for x2 . Taking the square root. Exercise 48. Calculate the area of the region between the graphs of the √ functions f (x) = x2 and g(x) = 1 − x2 . To ﬁnd their x-coordinates. Example 3. Solution To get a better understanding. . we ﬁnd the x-coordinates of the points where the curves intersect: A=− −1 + 2 √ 5 and B = −1 + 2 √ 5 .28. Let us look at an example. 2 Only the + sign occurs as x2 ≥ 0.

13: Region between two graphs Figure 3.5 0.134 CHAPTER 3.14: Region between two graphs To get the area of the region under the graph of f (x) and g(x) over the interval [A.6 0.2 -1 -1 -0. You are invited to work out the integral with the help of the Fundamental Theorem of Calculus to verify the result. . To get the area of the region Ω between the graphs. Solution: The region Ω between the graphs is shown in Figure 3. Example 3. π] where g(x) ≥ f (x). we take the area of the region under the graph of g(x) and subtract the area of the region under the graph of f (x).5 2 2. INTEGRATION 1 1 0. B] we can calculate the appropriate integrals. Concretely: B Area(Ω) = A g(x) dx − B B f (x) dx = A A (g(x) − f (x)) dx ≈ 1.29.5 3 0. ♦ Some problems are a bit more subtle.5 -0.5 1 Figure 3.06651.5 0.8 0. The numerical value was obtained by computer. the region Ω1 over the interval [0.4 0. and the region Ω2 over the interval [π/4. Find the area of the region between the graphs of the functions f (x) = cos x and g(x) = sin x for x between 0 and π.14. π/4] on which f (x) ≥ g(x). The region breaks up into two pieces.5 1 1. We calculate the areas of the regions Ω1 and Ω2 separately.

. (b) y = 8 − x2 and y = x2 (c) y = x2 and y = 3x + 5. We took care of this aspect by breaking up the interval into the part where f (x) ≥ g(x) and the part where g(x) ≥ f (x). Suppose f (x) and g(x) are integrable functions over an interval [a. ♦ Our general deﬁnition for the area between two graphs is as follows. we had to take into account where the function is non-negative. resp. (d) y = sin x and y = πx − x2 . in others you have to determine them from context.7 on page 114. b]. In all cases it is good to graph the functions before calculating the area of the region between them. An additional remark may be in place.. Deﬁnition 3. (e) y = sin x and y = 2 sin x cos x for x between 0 and π. This deﬁnition generalizes Deﬁnition 3. Taking the absolute value of the diﬀerence of f (x) and g(x) allows us avoid the question where f (x) ≥ g(x) and where g(x) ≥ f (x).23 on page 121. Sketch and ﬁnd the area of the region bounded by the curves: (a) y = x2 and y = x3 . When we compared integrals and areas. and it incorporates and generalizes Proposition 3. Exercise 49. The deﬁnition is also consistent with the intuitive idea of the area of a region. Here we did not.10. Having the correct picture in mind helps you to avoid mistakes. we proceed as in the previous example: π/4 135 Area(Ω1 ) = Area(Ω2 ) = 0 π π/4 (cos x − sin x) dx = (sin x + cos x) π/4 0 π π/4 = √ 2−1 √ 2. (sin x − cos x) dx = −(sin x + cos x) =1+ In summary we ﬁnd: √ Area(Ω) = Area(Ω1 ) + Area(Ω2 ) = 2 2. non-positive. In some problems a and b are explicitly given. The area of Ω is b Area(Ω) = a |f (x) − g(x)| dx. Typically this problem gets addressed when the integral is calculated. Let Ω be the region between the graphs of f (x) and g(x) for x between a and b. AREAS BETWEEN GRAPHS In each case.3.30.

17) IL = f (x0 )(x1 − x0 ) + f (x1 )(x2 − x1 ) + · · · + f (xn−1 )(xn − xn−1 ). 2 Solution: Set f (x) = e−x and choose the partition: x0 = 0 < x1 = 1 3 < x2 = 1 < x3 = < x4 = 2. 2 2 2 . only we use the value of the function at the right endpoint instead of the left endpoint: (3.31. Example 3. and then add up the terms. We multiply it with the length of the associated interval. In the right endpoint method we proceed as we did on the left endpoint method.136 CHAPTER 3. b]: b f (x) dx. INTEGRATION 3.18) IR = f (x1 )(x1 − x0 ) + f (x2 )(x2 − x1 ) + · · · + f (xn )(xn − xn−1 ). by numerical means. Let us take such a function. Still. Left and Right Endpoint Method: In the left endpoint method we ﬁnd the value of the function at each left endpoint of the intervals of the partition. Use the left and right endpoint method to ﬁnd approximate values for 2 0 e−x dx. we calculate (3. In such cases we may have to rely on numerical methods for integrating. and show some methods for ﬁnding an approximate value for the integral. Explicitly.11 Numerical Integration The Fundamental Theorem of Calculus provided us with a highly eﬃcient method for calculating deﬁnite integrals. for some functions we have no good expression for its anti-derivative. Both expressions provide us with speciﬁc examples of Riemann sums. We describe diﬀerent ways to ﬁnd. approximate values for the integral of a function f (x) over the interval [a. a In all of the diﬀerent approaches we partition the interval into smaller ones: a = x0 < x1 < · · · < xn−1 < xn = b.

2].8 0.4 0.3. so that IL is an upper sum for the function f (x) over the interval [0. 2].5 1 1. Formula (3.126039724. IL and IR are calculated by combining the areas of certain rectangles.17) for IL specializes to IL = f (0) + f (1/2) + f (1) + f (3/2) ≈ 1.16: Use right end points Apparently. we have IR = .4 0. 2.2 1 0. 2 Formula (3. 3. NUMERICAL INTEGRATION 137 Then xk −xk−1 = 1/2 for k = 1.2 0.11.16.6 0.6351975438 ≤ 2 0 e−x dx ≤ IL = 1.126039724. .18) for IR specializes to IR = f (1/2) + f (1) + f (3/2) + f (2) ≈ .8 0. In this sense. In our case the values of f (x) are all positive and all of the rectangles are above the x axis.15 and Figure 3.5 2 0. That should be less bias. that our speciﬁc function f (x) is decreasing on the interval [0.6351975438. 2 The function and the rectangles whose areas are added to give us IL and IR are shown in Figure 3.5 1 1. Note also. and 4.15: Use left end points Figure 3.6 0. we use the value of the function at the midpoints of the intervals of the partition. In the midpoint methods. 2 1 0. so the areas of the rectangles are all added. ♦ Midpoint and Trapezoid Method: We may try and improve on the endpoint methods. and IR is a lower sum.5 2 Figure 3.

We have such a trapezoid over each of the intervals in the partition.8827889485. The formula for IM (see (3. suppose that f (x) is non-negative on the interval [a. This integral is easily computed by the formula in (3.19)) specializes to IM = f (.75) ≈ . Speciﬁcally. f (x1 )). In the trapezoid method we do not take the function at the average (i. and their areas are added to give IT . b]. f (x0 )) and (x1 .20).20). For simplicity. see (3.25) + f (. The area of this trapezoid is f (x0 )+f (x1 ) (x1 − x0 ). b]. Consider the trapezoid of width (x1 − x0 ) which has height f (x0 ) at its left and f (x1 ) at its right edge. Over the interval [x1 . Use the midpoint and trapezoid method to ﬁnd approximate values for 2 0 e−x dx. the formula is f (x0 ) + f (x1 ) f (xn−1 ) + f (xn ) (x1 − x0 ) + · · · + (xn − xn−1 ). Expressed diﬀerently. f (x1 )) and (x2 . This gives us the graph of a function T (x) over the interval [x0 . we use appropriate secant lines above all of the intervals in the partition to deﬁne the function T (x) over the entire interval [a. 2 Let us explain the reference to the word trapezoid.21) IT = b IT = a T (x) dx. 2 . we can draw a secant line through the points (x0 .75) + f (1. 2 2 It is quite easy to see that (3. x2 ] the graph of T (x) is the secant line through the points (x1 . but we average the values of the function at the end points. INTEGRATION We use the same partition and notation as above. 2 Solution: We use the same partition of [0. f (x2 )).138 CHAPTER 3. Proceeding in the fashion. Then (3.e. Then the formula for the midpoint method is: (3.20) IT = IL + IR . This is the ﬁrst 2 summand in the formula for IT .32. Example 3.19) IM = f x 0 + x1 2 (x1 − x0 ) + · · · + f xn + xn−1 2 (xn − xn−1 ). midpoint) of the end points of the intervals in the partition. 2] as in Example 3.25) + f (1.31. x1 ].

18: Trapezoid Method Based on our previous calculations and Formula (3.8 0. b] is IS = (3.5 1 1. 1 0. Their width are the lengths of the intervals of the partition. So 2 IT = T (x) dx.21) we ﬁnd IT = IL + IR ≈ .6 0.8 0.18.5 2 0. The dots are connected by straight line segments. Their heights are the values f (xi ) at the midpoints of the intervals of the partition.17: Use midpoints Figure 3.17. Again. 2 We illustrated this calculation in Figure 3.11.8806186341. NUMERICAL INTEGRATION 139 As for the endpoint methods.4 0. The speciﬁc formula for an approximate value of the integral of f (x) over [a.22) 1 x 0 + x1 f (x0 ) + 4f 6 2 1 + f (xn−1 ) + 4f 6 + f (x1 ) (x1 − x0 ) + · · · xn−1 + xn 2 + f (xn ) (xn − xn−1 ) .2 0. You see the rectangles for this calculation in Figure 3. and IT is the area of the region under this graph.4 0. There you see the function 2 f (x) = e−x and ﬁve dots on the graph.2 1 0.6 0.3. These line segments form the graph of a function T (x).5 2 Figure 3.5 1 1. we use the same notation for the function and the partition as above. IM is the combined area of certain rectangles. 0 ♦ Simpson’s Method: In Simpson’s method we combine the endpoint and midpoint methods in a weighted fashion.

8 0. 2 .33.4 0. INTEGRATION IL + 4IM + IR IT + 2IM = . In the other one we use three points on the graph and construct a parabola through them. We deﬁne a function P (x) over the interval [a. 6 3 Let us explain the background to Simpson’s method.5 1 1. 1 0.6 0. b] by deﬁning a degree 2 polynomial on each of the intervals of the partition. Simpson’s method is a reﬁnement of the Trapezoid method.5 2 Figure 3.2 0. In one method we use two points on the graph and connect them by a straight line segment. The polynomial over the interval [xk−1 . With some work one can show that b IS = a P (x) dx.19: Simpson’s Method Example 3. xk ] is chosen so that it agrees with f (x) at the end points and at the midpoint of this interval.140 It is quite easy to see that IS = CHAPTER 3. Use Simpson’s method to ﬁnd an approximate value for 2 0 e−x dx.

♦ .883063050702697 0.8822020700 1. then the result is far less accurate that Simpson’s method with n = 100.882078948840 0.19. Even if you use the midpoint and trapezoid method with n = 1000.882081390762417 2.88206555104. 2 Solution: We compare the approximate values for the integral obtained by the diﬀerent formulas.11.881095681980474 0. 6 where IL .7357589 0.872262105229 0. There you see the graphs 2 of two functions. ♦ IS = Example 3. We partition the interval [0.882081390722 n = 1000 0. IM and IR are as above. We tabulate the results.0366313 0. the function f (x) = e−x and the function P (x) from the discussion of Simpson’s method.9800072469 0. 2] as in Example 3.882081366341586 0.882081402972833 0. They should be compared with an approximate value for the integral of 0.1: Approximate Values of the Integral Simpson’s method is more accurate than the other ones..0000000 0.3.22)) specializes to IL + 4IM + IR ≈ .882081390762421.882082611663 0. Simpson’s method with n = 4 gives a result which is better than the left and right endpoint method with n = 1000. n=1 IL IR IM IT IS n = 10 n = 100 0. E. and vary n.8299445 0.8820809836 Table 3.7836703747 0. Only the thickness of the line suggests that there are two graphs of almost identical functions. The formula for IS (see the special case of (3.0183156 0.891895792451 0. 2] into n intervals of the same length.34. NUMERICAL INTEGRATION 141 Solution: We use the same partition of [0. Compare the accuracy of the various approximate values of 2 0 e−x dx. You see the method illustrated in Figure 3.8818388108 0.g.31.

23 we related deﬁnite integrals to areas.34 and compare the diﬀerent methods applied to the calculation of π/2 sin x dx = 1. this can have a more concrete meaning. The fewer computations we make. b].142 CHAPTER 3. π] is 2/π. In addition. referred to as Beaver Pond by the locals. Deﬁnition 3. b]. then I stands for the total distance which you traveled during the time interval [a. b] small. then I is the total amount of the drug which has been absorbed in the time interval [a. Based on the context.36. You are invited to come up with more interpretations. the following deﬁnition expresses the common notion of the average value of a function. The amount of water carried by the river depends on the season. It is important that we keep the number n of intervals into which we partition [a. Then the quantity fav := 1 b−a b f (t) dt a is called the average value of f (t) over the interval [a. Proceed as in Example 3.35.7 and Proposition 3. For example. the average value of the sine function f (x) = sin x over the interval [0. Consider a function f (t) on an interval [a. As a function of time. Suppose that f (t) is an integrable function over the interval [a.12 Applications of the Integral In Deﬁnition 3. INTEGRATION Remark 13. b]. . the smaller the cummulative round-oﬀ error will be. it is g(t) = 2 + sin πt 180 . In each computational step we expect to make a roundoﬀ error. b] and the integral b I= a f (t) dt. The river Little Brook ﬂows into a reservoir. Exercise 50. It does not only keep the number of overall computations small. b]. Example 3. If f (t) stands for the rate at which a drug is absorbed. Let us explore the diﬀerent aspects of integration in an example. and these may add up. 0 3. If f (t) stands for the speed with which you travel.

APPLICATIONS OF THE INTEGRAL 143 We measure time in days. At the beginning of September. the rate of change is .12. By the end of April. Find F (t). the total amount of water in Beaver Pond is T F (T ) = 200 + 0 f (t) dt = 200 + 180 1 − cos π πT 180 millions of liters. On the T -th day of the year. and t = 0 corresponds to New Year. after 120 days.2 π 3 millions of liters of water in the pond. Set T A(T ) = 0 f (t) dt. Water is released from Beaver Pond at a constant rate of 2 million liters per day. (c) At which rate does the amount of water in the reservoir change at the beginning of September? (d) On which days will there be 250 million liters of water in Beaver Pond? (e) At which amount of water will the reservoir crest? (f) On the average. there are 200 million liters of water in the reservoir. This answers (a).3. there are F (120) = 200 + 180 2π 1 − cos ≈ 238. The net rate entering is f (t) = g(t) − 2 = sin πt 180 millions of liters per day. This answers (b). by how much has the amount of water in Beaver Pond increased per day during the ﬁrst three months of the year? Solution: Water enters and leaves the pond. We obtain the total change of the amount of water in the reservoir by integrating f (t). (a) How many liter of water are in Beaver Pond by the end of April? (b) Suppose F (t) tells how much water there is in the reservoir on day t of the year. The rate at which the amount of water in the pond changes is F (t) = f (t). The units of g(t) are millions of liter of water per day. after 240 days. At the beginning of the year.

We will now ﬁll in the details. the amount of water has increased by 57. or 272. 000 liters per day. the amount of water in the reservoir increased by about 640. Within this time. This occurs apparently when cos(πt/180) = −1 or t = 180.144 CHAPTER 3.866. we have to ﬁnd the maximum value of F (t).10 we introduced the exponential function exp(x) = ex and the natural logarithm function ln x.3 millions of liters of water in Beaver Pond. t = 0 at the time you take the medication. To answer (d).13 The Exponential and Logarithm Functions In Section 1. we like to know for which T we have F (T ) = 250. and the absorption process is complete at time t = 1.000 liters per day.6 millions of liters of water. Here t measures time in hours. Many of the routine calculations are formulated as exercises. On the average.3 millions of liters. INTEGRATION f (240) ≈ −. The pond crests at mid-year. . such that F (t) tells how much medication has been absorbed at time t. A pain reliever has been formulated such that it is absorbed at a rate of 600 sin(πt) (mg/hr) by the body. We solve the equation for T : 250 = 200 + 180 1 − cos π πT 180 or cos πT 180 =1− 5π . (a) What is the total amount of the drug which is absorbed? (b) Find a function F (t). (c) A total of 150 mg of the medication has to be absorbed before the drug is eﬀective. At the time we only stated that they exist because we did not have the tools to properly deﬁne them. This answers (e). 18 We apply the function arccos to both sides of the last equation and ﬁnd T = 180 5π arccos 1 − π 18 ≈ 88. To ﬁnd at which amount the reservoir crests. On the 88-th and 272-nd day of the year there will be 250 millions of liters of water in the reservoir. After three months or 90 days there are about 257. and then the amount of water in it is about 314. How long does it take until this threshold is reached? 3. The pond is losing water at a rate of 866. ♦ Exercise 51.

24) ln(xy) = ln x + ln y. ∞). Let x ∈ (0. x Proof. so that du = x dt.11.26 tells us that ln x = 1/x. The natural logarithm of x is deﬁned as x (3. ∞). Proof. Proposition 3. Let us also verify one of the central equations for calculating with logarithms.3. ∞). its derivative is ln x = and ln x is increasing on (0.10 this means that ln x is deﬁned for all x in (0. Exercise 52. Then xy x dt = t xy x 1 1 dt = (t/x) x y 1 du = ln y. t 1 We use the substitution u = x . The function 1/x is deﬁned and continuous on (0. and that t/x = u = y when t = xy. the function is increasing because its derivative ln x > 0 for all x > 0. t This is exactly our claim. Show: (1) ln 1 = 0.23) ln x = 1 dt . ∞).38.34. For the adjustment of the limits of integration. u Using this calculation we deduce that xy ln(xy) = 1 dt = t x 1 dt + t xy x dt = ln x + ln y. According to Theorem 2. THE EXPONENTIAL AND LOGARITHM FUNCTIONS 145 Deﬁnition 3.37. For any x. y > 0. Theorem 3. (2) ln(1/y) = − ln y for all y > 0. observe that t/x = u = 1 when t = x. We need a short calculation.13. . (3. 1 . t Theorem 3. ∞).39. Here x and y are ﬁxed positive numbers. The natural logarithm function is diﬀerentiable on its entire domain (0. the third rule in Theorem 1. According to Theorem 3.

This means that. numbers of the form r = p/q where p and q are integers and q = 0. 1 dt = 1. In summary. we have seen that Corollary 3. i. Every real number x lies between two integers. Hint: Using the partition 1 = x 0 < 2 = x1 < 3 = x2 < 4 = x3 . We can now deﬁne the Euler number: Deﬁnition 3. To see this. ∞) and range (−∞. we have to show that there is a number e which has the property used in the deﬁnition. increasing function with domain (0. Show that ln 4 > 1. Observe that ln(en ) = n and ln(1/en ) = −n for all natural numbers n. 4] so that Sl > 1. for any given x. ﬁnd a lower sum Sl for the function 1/t over the interval [1. equivalently. . Exercise 54. such that (3. observe that ln 1 = 0 < 1 < ln 4. y > 0. CHAPTER 3. The number Euler number e is the unique number such that e ln e = 1 or. Because ln x is diﬀerentiable. So all integers (whole numbers) are values of the natural logarithm function.25) ln y = x Proof. It also follows that 1 < e < 4. For every real number x there exists exactly one positive number y. Taken together it means that it has a unique solution. every real number is a value of the function ln y.e. the equation ln y = x has at most one solution. t For this deﬁnition to make sense.. it follows from the Intermediate Value Theorem (see Theorem 1. ∞). We saw that ln y is an increasing function.40. Proposition 3.16) that there is a number e for which ln e = 1. According to the Intermediate Value Theorem.146 (3) ln(x/y) = ln x − ln y for all x. INTEGRATION Exercise 53.42. and ln x = 1/x. The natural logarithm function ln x is a diﬀerentiable.41. Show that ln(ar ) = r ln a for all positive numbers a and all rational numbers r.

and adding some observations which we have made elsewhere.44. THE EXPONENTIAL AND LOGARITHM FUNCTIONS We are now ready to deﬁne the exponential function. Hint: Use Exercise 54 and that the exponential and logarithm functions are inverses of each other.e. In addition to the equation in (3.26). and that the exponential and logarithm functions are inverses of each other.40.26) ln(exp(x)) = x. ∞).. ∞) and range (0. ∞).e. Exercise 55. exp(ln(y)) = y . called the exponential function. you need to show that (3. we have: Proposition 3. i.. ∞) and range (0. The exponential function exp(x) is a diﬀerentiable. Exercise 57. 147 Deﬁnition 3. and the exponential function is its own derivative. exp (x) = exp(x). y = exp(x) is the unique solution of the equation ln(y) = x. Show for all real numbers x and y that: (1) exp(0) = 1 (2) exp(1) = e (3) exp(x) exp(y) = exp(x + y) (4) 1/ exp(y) = exp(−y) (5) exp(x)/ exp(y) = exp(x − y). increasing function with domain (−∞. ∞). This assignment (mapping x to exp(x)) deﬁnes a function. we deﬁne exp(x) to be the unique number for which (3. Summarizing this discussion. i. Exercise 56. with domain (−∞. the deﬁnition of e in Deﬁnition 3.3.27) for all y ∈ (0. Hint: Use the results of Exercise 52. Show that the exponential function exp and the natural logarithm function ln are inverses of each other.43. Show that exp(r) = er for all rational numbers r.13. Given any real number x.

the speciﬁcations for the domains and ranges for the functions expa and loga and the results from Exercise 59 tell us that Corollary 3. (b) loga (expa (x)) = x for all real numbers x. For the function loga we use the domain (0.29) loga x = ln x ln a and expa (x) = exp(x ln a). 3. ∞) and range (0.1 Other Bases So far we discussed the natural logarithm function and the exponential function with base e. Taken together. Exercise 58.45. ∞) and range (−∞. (3) loga (x) and expa (x) are increasing functions if a > 1. Let a be a positive number. For an arbitrary real number we set (3. and it deﬁnes what we mean by raising e to any real power. Show that (a) expa (loga (y)) = y for all y > 0. (2) loga (x) and expa (x) are diﬀerentiable functions. If r = p/q then we raise e to the r-th power and take the q-th root of the result.28) ex = exp(x). ∞). We call loga (x) the logarithm function with base a and expa (x) the exponential function with base a. The functions expa and loga are inverses of each other. ∞). (4) loga (x) and expa (x) are decreasing functions if 0 < a < 1. This is consistent with the meaning of the expression for rational exponents due to Exercise 57. We now expand the discussion to other bases. Set (3. For the exponential function expa we use the domain (−∞. Suppose a > 0 and a = 1. Show (1) ln a > 0 if a > 1 and ln a < 0 if 0 < a < 1.148 CHAPTER 3. Suppose a > 0 and a = 1. Exercise 59.13. . Deﬁnition 3. a = 1.46. INTEGRATION The expression er makes sense only if r is a rational number.

149 We rephrase a convention which we made previously for e. Suppose a > 0. Suppose a > 0 and a = 1. The expression ar makes sense if r is a rational number.13. and r is a rational number. Show loga (ar ) = r and expa (r) = ar . For an arbitrary real number we set (3. Exercise 62. Suppose a > 0. It is also a standard convention to set 1x = 1 and 0x = 0 for any real number x. Typically 00 is set 1. Show the laws of logarithms: (a) loga 1 = 0 and loga a = 1. x > 0. (b) loga (xy) = loga x + loga y for all x. If r = p/q then we raise a to the r-th power and take the q-th root of the result.30) ax = expa (x). y > 0. . Equation 3.28 if we set a = e. Show the exponential laws: (1) expa (0) = 1 and expa (1) = a (2) expa (x) expa (y) = expa (x + y) (3) 1/ expa (y) = expa (−y) (4) expa (x)/ expa (y) = expa (x − y).3. a = 1. We can now state an equation which is typically considered to be one of the laws of logarithms: Exercise 63.30 specializes to the one in Equation 3. Then loga (xz ) = z loga (x). Suppose a > 0 and a = 1. and z is any real number. THE EXPONENTIAL AND LOGARITHM FUNCTIONS Exercise 60. (d) loga (x/y) = loga x − loga y for all x. Exercise 61. y > 0. and it deﬁnes what we mean by raising a to any real power. (c) loga (1/y) = − loga y for all y > 0. Suppose a > 0 and a = 1. a = 1. This is consistent with the meaning of the expression for rational exponents due to Exercise 62.

We have to show the uniqueness statement. In this section we constructed the function expa (x). Proof. and that expa (x) is continuous.47. Here one uses that f (x) and expa (x) are monotonic. We are ready to prove Theorem 3. which is deﬁned for all real numbers x such that expa (x) = ax whenever x is a rational number. and this function has all of the properties called for in the theorem. a = 1. called the exponential function with base a and denoted by expa (x). Suppose f (x) is any monotonic function and f (r) = ar = expa (r) for all rational numbers r. We leave the veriﬁcation of this assertion to the reader..e. That settles the existence statement. . Let a be a positive number. There exists exactly one monotonic function. We have to show that f (x) = expa (x) for all real numbers x.10. INTEGRATION We are now ready to ﬁll in the details for one of the major statements which we made in Section 1. i.150 CHAPTER 3. there in only one such function.

These are the functions sine. It is shown in Figure 4. Stated diﬀerently it says.1) α = ±s (radians). Observe that the ratio referred to in the deﬁnition does not depend on the radius of the circle. but any two radian measures of the angle diﬀer by an integer multiple of 2π. and suppose its length is s. and measure how long it is. The − sign is used if it proceeds clockwise. The number π is the ratio between the circumference of a circle and its diameter. Consider an arc on the unit circle which starts out at the point (1. It requires some work to introduce the idea of the length of a curve in a mathematically rigorous fashion. 0) and ends at p. We imagine that we can straighten it out. We like to ﬁnd the radian measure of the angle α. et. Consider an angle α between the positive x-axis and a ray which originates at the origin of the coordinate system and intersects the unit circle in the point p. tangent. This deﬁnition goes back to the Greeks. cosine. We collect some formulas relating these functions.Chapter 4 Trigonometric Functions In this section we discuss the radian measure of angles and introduce the trigonometric functions.1. the radian measure of the angle α is not unique. We may also consider arcs which wrap around the circle several times before they end at p. that the circumference of a circle of radius r is 2πr. Deﬁnition 4. Arc Length and Radian Measure of Angles: Consider the unit circle (a circle with radius 1) centered at the origin in the Cartesian plane. Then (4. al.1. The + sign is used if the arc goes counter clockwise around the circle. 151 . We take a practical approach to measuring the length of an arc on this circle. In this sense.

By convention. So.2) x degrees = π x radians. 0) we travel the distance |t| along the unit circle (here |t| denotes the absolute value of t). and one radian corresponds to 180/π ≈ 57.5 1 -0. We have the conversion formula (4. In this way we reach a point p = (x(t).017453293 radians. . Comparison of Angles in Degrees and Radians: We suppose that you are familiar with measuring angles in degrees. TRIGONOMETRIC FUNCTIONS 1 0.5 -1 Figure 4. This angle has radian measure t.152 CHAPTER 4. In this way we reach a point p on the circle. 180 Trigonometric Functions: Let t be once more a real number. We construct the angle with radian measure t. y(t)) on the circle. Starting at the point (1.5 0. Starting at the point (1. and we set (4.1: The unit circle Conversely. we travel counter clockwise if t is positive and clockwise if t is negative. counter clockwise if t is positive and clockwise if t is negative.29577951 degrees.3) x(t) = cos t and y(t) = sin t. 0) we travel the distance |t| along the unit circle. one degree corresponds to π/180 ≈ 0.5 (cos t. The measure of half a revolution (a straight angle) comprises π radians and 180 degrees. sin t) -1 -0. Let α be the angle between the positive x-axis and the ray which starts at the origin and intersects the unit circle in p. let t be any real number.

You can ﬁnd the graphs of the sine and cosine functions on the interval [0.5 -1 -1 Figure 4. You can see the graphs of the secant and cosecant functions in Figure 4. If the functions are not deﬁned at some point.4) tan x = sin x cos x cot x = cos x sin x sec x = 1 cos x csc x = 1 sin x To make sure you have some idea about the behavior of the tangent and cotangent function we provided two graphs for each of them.4 to Figure 4.2 and 4. tangent (tan). and cosecant (csc) are deﬁned as follows: (4. 0).3. A small table with angles given in degrees and radians.2: f (x) = sin x Figure 4.1. secant (sec). See Figure 4.7. They are drawn over diﬀerent parts of the domain to show diﬀerent aspects. You see a right triangle with vertices (0. The .9. Trigonometric Functions deﬁned at a right triangle: Occasionally it is more convenient to use a right triangle to deﬁne the trigonometric functions. (x. Older calculus books may still contain tables with the values of the trigonometric functions. 0) and (x.5 1 2 3 4 5 6 1 2 3 4 5 6 -0.3: f (x) = cos x The other trigonometric functions.5 0.1. cotangent (cot). as well as the associated values for the trigonometric functions is given in Table 4. then this is indicated by ‘n/a’.8 and 4. You see the construction implemented in Figure 4. and there are books which were published for the speciﬁc purpose of providing these tables.5 -0. We may use a circle of any radius r.1. 1 1 0. This is really not necessary anymore because any scientiﬁc calculator gives those values to you with rather good accuracy. To do this we return to Figure 4. 2π] in Figures 4. y).153 This deﬁnes the functions sin t and cos t.

5 1 -20 -1 -40 -2 Figure 4. π] . π] Figure 4.5 2 2.6: cot x on [−π.5 Figure 4. π] Figure 4.8: sec x on [−π.9: csc x on [−π.154 CHAPTER 4. 1. π] Figure 4.1] 40 1.5 1 20 0.5 0.7: cot x on [ π − 1.5 -40 -1.1.5: tan x on [−1.5 -3 -2 -1 -20 1 2 3 -0.4: tan x on [−π.5 -1 1. π + 1] 2 2 20 20 10 10 -3 -2 -1 1 2 3 -3 -2 -1 1 2 3 -10 -10 -20 Figure 4. TRIGONOMETRIC FUNCTIONS 40 2 20 1 -3 -2 -1 1 2 3 -1 -0.

155 degrees radians sin x cos x tan x cot x 0 30 45 60 90 120 135 150 180 0 π/6 π/4 π/3 π/2 2π/3 3π/4 5π/6 π 0 1 2 √ 2 2 √ 3 2 √ √ sec x √ 2 3 3 csc x n/a 2 √ 2 1 3 2 √ 2 2 1 2 √ 0 3 3 n/a √ 3 √ 1 1 √ 3 n/a √ − 3 √ − 33 1 3 3 √ √ 2 2 1 −2 √ − 2 −1 √ 2 3 3 1 0 √ 2 2 √ − 23 0 − n/a √ 2 3 3 3 2 √ 2 2 1 2 −1 2 3 3 − −1 −1 √ √ − 3 −233 n/a √ 2 2 n/a 0 −1 0 Table 4. Let α be the angle at the vertex (0.5) sin2 x + cos2 x = 1. sec2 x = 1 + tan2 x. csc2 x = 1 + cot2 x. others you should be aware of.1: Values of Trigonometric Functions right angle is at the vertex (x. 0) and the hypotenuse has length r. 0). The following identities are obtained from elementary geometric observa- . From the theorem of Pythagoras and the deﬁnitions you obtain (4. Then opposing side hypothenuse opposing side tan α = adjacent side hypothenuse sec α = adjacent side sin α = cos α = adjacent side hypothenuse adjacent side cot α = opposing side hypothenuse csc α = opposing side Trigonometric Identities: There are several important identities for the trigonometric functions. In the following the words adjacent and opposing are in relation to α. so that you can look them up whenever needed. Some of them you should know.

15) sin α sin β = sin α cos β = cos α cos β = 1 [cos(α − β) − cos(α + β)] 2 1 [sin(α − β) + sin(α + β)] 2 1 [cos(α − β) + cos(α + β)] 2 which specialize to the the half-angle formulas (4.10) (4. or even derived.13) (4. (4.11) sin(α + β) = sin α cos β + cos α sin β sin(α − β) = sin α cos β − cos α sin β cos(α + β) = cos α cos β − sin α sin β cos(α − β) = cos α cos β + sin α sin β tan α + tan β tan(α + β) = 1 − tan α tan β tan α − tan β tan(α − β) = 1 + tan α tan β These formulas specialize to the double angle formulas (4.9) (4. TRIGONOMETRIC FUNCTIONS tions using the unit circle.156 CHAPTER 4. the following addition formulas in precalculus. sin x cos x cos x sin x = sin(x + 2π) = cos(x + 2π) = sin(x + π ) 2 = − cos(x + π ) 2 = sin(π − x) = − cos(π − x) = − cos(x + π) = − sin(x + π) = − sin(−x) = cos(−x) = − sin(x + 3π ) 2 = cos(x + 3π ) 2 You should have seen.7) (4.8) (4.6) (4.16) sin2 α = 1 [1 − cos 2α] 2 and cos2 α = 1 [1 + cos 2α] 2 .14) (4.12) sin 2α = 2 sin α cos α and cos 2α = cos2 α − sin2 α From the addition formulas we can also obtain (4.