ii

iii
Worldwide
Integral
Calculus
with innite series
David B. Massey
with select exercises and answers by Gabriel Merton and Ryan Malloy,
select illustrations by Eugene Saletan,
and select video solutions by Michael Garcia, Ryan Malloy, and Meryl Stav
iv
c _2009-2011, Worldwide Center of Mathematics, LLC
Contents
0.1 Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
1 Anti-dierentiation: the Indenite Integral 1
1.1 Basic Anti-Dierentiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.2 Special Trig. Integrals and Trig. Substitutions . . . . . . . . . . . . . . . . . . . 29
1.2.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
1.3 Integration by Partial Fractions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
1.3.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
1.4 Integration using Hyperbolic Sine and Cosine . . . . . . . . . . . . . . . . . . . . 57
1.4.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
2 Continuous Sums: the Denite Integral 67
2.1 Sums and Dierences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
2.1.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
2.2 Prelude to the Denite Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
2.2.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
2.3 The Denite Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
2.3.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
2.4 The Fundamental Theorem of Calculus . . . . . . . . . . . . . . . . . . . . . . . . 139
2.4.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
2.5 Improper Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
2.5.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
v
vi CONTENTS
2.6 Numerical Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
2.6.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
Appendix 2.A Technical Matters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
3 Applications of Integration 215
3.1 Displacement and Distance Traveled . . . . . . . . . . . . . . . . . . . . . . . . . 216
3.1.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
3.2 Area in the Plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
3.2.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
3.3 Distance Traveled in Space and Arc Length . . . . . . . . . . . . . . . . . . . . . 249
3.3.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
3.4 Area Swept Out and Polar Coordinates . . . . . . . . . . . . . . . . . . . . . . . 276
3.4.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
3.5 Volume . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294
3.5.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314
3.6 Surface Area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
3.6.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329
3.7 Mass and Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334
3.7.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343
3.8 Centers of Mass and Moments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 348
3.8.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 361
3.9 Work and Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 366
3.9.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 380
3.10 Hydrostatic Pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 385
3.10.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 389
4 Polynomials and Power Series 393
4.1 Approximating Polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 395
4.1.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 405
4.2 Approximation of Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 408
CONTENTS vii
4.2.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 422
4.3 Error in Approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 426
4.3.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 438
4.4 Functions as Power Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 444
4.4.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 458
4.5 Power Series as Functions I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 462
4.5.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 482
4.6 Power Series as Functions II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 487
4.6.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 510
4.7 Power Series Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 515
4.7.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 526
Appendix 4.A Technical Matters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 528
5 Theorems on Sequences and Series 529
5.1 Theorems on Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 530
5.1.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 544
5.2 Basic Theorems on Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 548
5.2.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 566
5.3 Non-negative Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 572
5.3.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 592
5.4 Series with Positive and Negative Terms . . . . . . . . . . . . . . . . . . . . . . . 596
5.4.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 611
Appendix A An Introduction to Vectors and Motion 615
Appendix B Tables of Integration Formulas 623
Appendix C Answers to Odd-Numbered Exercises 627
Bibliography 649
Index 651
viii CONTENTS
About the Author 657
0.1. PREFACE ix
0.1 Preface
Welcome to the Worldwide Integral Calculus textbook; the second textbook from the Worldwide
Center of Mathematics.
Our goal with this textbook is, of course, to help you learn Integral Calculus (and power
series methods) the Calculus of integration. But why publish a new textbook for this purpose
when so many already exist? There are several reasons why we believe that our textbook is a
vast improvement over those already in existence.
Even if this textbook is used as a classic printed text, we believe that the exposition, expla-
nations, examples, and layout are superior to every other Calculus textbook. We have tried to
write the text as we would speak the material in class; though, of course, the book contains far
more details than we would normally present in class. In the book, we emphasize intuitive ideas
in conjunction with rigorous statements of theorems, and provide a large number of illustrative
examples. Where we think it will be helpful to you, we include proofs, or sketches of proofs,
in the midst of the sections, but the extremely technical proofs are contained in the Technical
Matters appendices to chapters, or are contained in referenced external sources. This greatly
improves the overall readability of our textbook, while still allowing us to give mathematically
precise denitions and statements of theorems.
Our textbook is an Adobe pdf le, with linked/embedded/accompanying video content, an-
notations, and hyperlinks. With the videos contained in the supplementary les, you eectively
possess not only a textbook, but also an online/electronic version of a course in Integral Cal-
culus. Depending on the version of the les that you are using, clicking on the video frame to
the right of each section title will either open an online, or an embedded, or a locally installed
video lecture on that section. The annotations replace classic footnotes, without aecting the
readability or formatting of the other text. The hyperlinks enable you to quickly jump to a
reference elsewhere in the text, and then jump back to where you were.
The pdf format of our textbook makes it incredibly portable. You can carry it on a laptop
computer, on many handheld devices, e.g., an iPad, or can print any desired pages.
Rather than force you to buy new editions of textbooks to obtain corrections and minor
revisions, updates of this textbook are distributed free of cost.
Because we have no print or dvd costs for the electronic version of this book and/or videos,
we can make them available for download at an extremely low price. In addition, the printed,
bound copies of this text and/or disks with the electronic les are priced as low as possible, to
help reduce the burden of excessive textbook prices.
x PREFACE
In this book, we assume you are already familiar with Dierential Calculus. Specically,
we assume that you know the denition of the derivative of a function, that it represents the
instantaneous rate of change, and that you know the rules for calculating derivatives. We will
also need lHopitals Rule and parameterized curves. Referring to the Worldwide Dierential
Calculus textbook [2], this means that you should know the contents of Chapters 1 and 2, Section
3.5, and Appendix A.
Our discussion of denite integrals, and their applications, is fairly traditional. However, our
approach to innite series is somewhat unusual. Our approach is motivated by two factors. First,
we believe that the primary use that students will have for innite series, outside of a Calculus
class, is that many important functions have convergent power series representations, and these
power series representations allow the student to mathematically manipulate and estimate the
functions involved, in ways that would be dicult/impossible without power series. Second,
statistical data that we collected over several years has made it clear that, in general, students
do not grasp the basic idea that, when x is close to zero, smaller powers of x are more signicant
than larger powers of x in a power series or, even, in a polynomial function.
Consequently, we place emphasis on polynomial approximations and power series represen-
tations for functions, and, in a sense, view the classic convergence tests for sequences and series
of constants as the technical details required to understand power series. We still include a
chapter, Chapter 5, on sequences and series of constants, but that chapter comes after Chap-
ter 4, which is on power series and approximating functions with polynomials. We rmly believe
that this ordering of topics is better for the student and for applications, even though it may
seem a bit awkward not to have the rigorous mathematical foundations of sequences and series
come before their use in discussing power series.
This book is organized as follows:
Other than the Technical Matters sections, each section is accompanied by a video le, which
is either a separate le, or an embedded video. Each video contains a classroom lecture of the
essential contents of that section; if the student would prefer not to read the section, he or
she can receive the same basic content from the video. Each non-technical section ends with
exercises. The answers to all of the odd-numbered exercises are contained in Appendix C, at
the end of the book.
Important denitions are boxed in green, important theorems are boxed in blue. Remarks,
especially warnings of common misconceptions or mistakes, are shaded in red. Important con-
ventions or fundamental principles, that will be used throughout the book, are boxed in black.
Very technical denitions and proofs from each section are contained in the Technical Matters
appendices at the ends of some chapters, or in external sources. Our favorite external technical
source is the excellent textbook by William F. Trench, Introduction to Real Analysis, [4], which
PREFACE xi
is available, courtesy of the author, as a free pdf. For producing answers to various exercises or
for help with examples or visualization, you may nd the free web site wolframalpha.com very
useful.
Internal references through the text are hyperlinked; simply click on the boxed-in link to
go to the appropriate place in the textbook. If you have activated the forward and back
buttons in your pdf-viewer software, clicking on the back button will return you to where you
started, before you clicked on the hyperlink.
Some terms or names are annotated; these are clearly marked in the margins by little blue
balloons. Comments will pop up when you click on such annotated items.
Occasionally, when looking at approximations, we write an equals sign in quotes, as in =.
We use this to denote equal as far as a calculator is concerned, i.e., equal to the precision of
many/most/all calculators.
We sincerely hope that you nd using our modern, multimedia textbook to be as enjoyable
as using a mathematics textbook can be.
David B. Massey
August 2009
xii PREFACE
Chapter 1
Anti-dierentiation: the
Indenite Integral
In this chapter, we discuss anti-dierentiation, which is also called indenite integration. This
is the process for undoing dierentiation. In the rst section, we start with the basic tech-
niques/results, and then in the remaining sections, we include some more-complicated methods.
The indenite integral should not be confused with the denite integral, which is the topic
of the next chapter. The denite integral is the mathematically precise notion of what it means
to take a continuous sum of innitesimal contributions. The reason that both indenite and
denite integration are referred to as integration is because calculating continuous sums and
nding anti-derivatives are related by the Fundamental Theorem of Calculus, Theorem 2.4.10.
1
2 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
1.1 Basic Anti-Dierentiation
This section is about the process and formulas involved in un-doing dierentiation, that is, in
anti-dierentiating. This means that you are given a function f(x) and are asked to produce
some/all functions F(x) which have f(x) as their derivative. This comes up often in applications,
such as when youre given the acceleration a(t) of an object and want the velocity v(t), or when
calculating denite integrals via the Fundamental Theorem of Calculus (see Section 2.3 and
Theorem 2.4.10).
Since you know dierential Calculus, you know what it means to have a function F(x) and
then be asked to calculate its derivative F

(x). For instance, if F(x) = x

3
, then F

(x) = 3x
2
.
But what about the reverse question? What if you are given the function f(x) = 3x
2
and
asked to produce an anti-derivative of f(x), that is, if you are asked to nd a function F(x)
whose derivative equals the given f(x)?
Certainly, F(x) = x
3
is one anti-derivative of 3x
2
. Are there any others? According to
a corollary to the Mean Value Theorem, the only other anti-derivatives of 3x
2
are functions
that dier by a constant from the one anti-derivative that we produced, i.e., every other anti-
derivative F(x) of f(x) = 3x
2
is of the form F(x) = x
3
+C, for some constant C.
Denition 1.1.1. Given a function f(x), dened on an open interval I, a function F(x),
on I, such that F

(x) = f(x) is called an anti-derivative of f(x), with respect to x.

Thus, an anti-derivative y = F(x) of f(x) is a solution to the dierential equation
dy/dx = f(x).
If F(x) is an anti-derivative of f(x), on an open interval, then every anti-derivative of
f(x), on that interval, is given by y = F(x) + C, where C is a constant. The collection
y = F(x) + C is called the (general) anti-derivative of f(x), with respect to x; it is the
general solution y to the dierential equation dy/dx = f(x).
The notation for the general anti-derivative of f(x), with respect to x, is
_
f(x) dx.
This is also called the (indenite) integral of f(x), with respect to x.
1.1. BASIC ANTI-DIFFERENTIATION 3
Remark 1.1.2. We have several important comments to make.
First, it is important that
_
f(x) dx is not one particular function, but it almost is;
_
f(x) dx
is actually a collection, or set, of functions, any two of which dier by a constant.
We write
_
3x
2
dx = x
3
+ C,
where including the +C is extremely important, for changing the value of C changes which
element of the set of all anti-derivatives of 3x
2
you are talking about. Technically, we ought to
write
_
3x
2
dx =
_
x
3
+C [ C R
_
,
but this is very cumbersome, and no one (well...no one that we know of) ever writes this.
Second, you should notice that it follows from the denition that the units of
_
f(x) dx are
the units of f(x) times the units of x.
For instance, if f(x) is in kilograms per cubic meter, and x is in cubic meters, then
_
f(x) dx
is in kilograms.
Third, you should think of the anti-dierentiation, with respect to x, operator,
_ _ _
dx,
as essentially being the inverse operator of
d
dx
_ _
, dierentiation with respect to x. That is,
the anti-dierentiation operator is a compound symbol; it starts with a
_
, and ends with a
dierential, like dx, which, together, tell you to anti-dierentiate whatever is in-between with
respect to the variable which appears in the dierential.
We wrote essentially above because, if you rst dierentiate and then anti-dierentiate,
you get what you started with, except that there is an additional +C; that is, you end up with
a collection of functions that all dier by constants, instead of simply the one function that you
started with.
We should also comment on the term indenite integral. There is another notion, called
the denite integral of a function over a closed interval; see Section 2.3. The denite integral is
dened in such a way that it agrees with ones intuitive idea of what a continuous sum of in-
nitesimal contributions should mean. This would seem to be unrelated to anti-dierentiating.
However, there is a theorem, the Fundamental Theorem of Calculus, which tells us: i)
every continuous function possesses an anti-derivative (Theorem 2.4.7), and ii) the primary step
used to obtain a nice formula for a denite integral is to produce an anti-derivative of the given
function (Theorem 2.4.10).
Hence, anti-dierentation is frequently referred to simply as integration, and denite in-
tegration is also simply referred to as integration; the context should always make it clear
4 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
whether the meaning is anti-dierentiation or denite integration. In addition, the symbols for
anti-dierentiating
_ _ _
dx are essentially the same as the symbols used for denite integration.
All of the dierentiation formulas which you have learned yield corresponding anti-dierent-
iation formulas; its just a matter of reading things in reverse, for, if F

(x) = f(x), then the

corresponding integration rule is
_
f(x) dx = F(x) +C, where C denotes an arbitrary constant.
In this context, f(x) is frequently referred to as the integrand.
For instance, we have a Power Rule for Integration:
Theorem 1.1.3. For all x in an open interval for which the functions involved are dened,
1.
_
0 dx = C;
2.
_
1 dx = x +C;
3. if p ,= 1,
_
x
p
dx =
x
p+1
p + 1
+C; and
4.
_
x
1
dx =
_
1
x
dx = ln [x[ +C.
Remark 1.1.4. In many books, only the third formula above is referred to as the Power Rule
for Integration.
As is frequently the case, you should try to remember this rule not in symbols, but in words;
it says that, as long as the exponent is not 1, you obtain the anti-derivative of x raised to a
constant exponent by adding one to the exponent, and dividing by the new exponent (and then
adding a C).
Remark 1.1.5. There are two fairly common, horric mistakes associated with the inte-
gration rule
_
x
1
dx =
_
1
x
dx = ln [x[ +C.
1.1. BASIC ANTI-DIFFERENTIATION 5
The rst big mistake is to treat the p = 1 case in the same manner as the cases where
p ,= 1. If you were to do this, you would obtain
_
x
1
dx =
x
1+1
1 + 1
+C =
x
0
0
+C.
The undened division by 0 should immediately tell you that youve done something wrong,
and remind you that you must treat the p = 1 case dierently.
The second big mistake may come later, when we have more integration rules. It will then
be tempting to look at the formula
_
1
x
dx = ln [x[ +C and think that it implies that
_
1
anything
dx = ln [anything[ +C.
This is completely wrong (in general); it is not true that the derivative of the expression
on the right would be the integrand. The problem is that, if you dierentiate the expression
on the right, you do, in fact, get a 1/anything factor, but then the Chain Rule tells you that
that is multiplied by d(anything)/dx.
Example 1.1.6. Find the function P(r), with domain r > 0, such that
dP
dr
=

r and P(9) = 7.
Solution:
We nd that
P =
_
r
1/2
dr =
r
3/2
3/2
+C =
2
3
r
3/2
+C.
We need to determine C. We have
7 = P(9) =
2
3
(9)
3/2
+C = 18 +C.
6 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
Therefore, C = 25, and so
P =
2
3
r
3/2
25.
The linearity of the derivative gives us the linearity of the anti-derivative.
Theorem 1.1.7. (Linearity of Anti-dierentiation) If a and b are constants, not both
zero, then
_
af(x) +bg(x) dx = a
_
f(x) dx + b
_
g(x) dx.
Remark 1.1.8. The prohibition against a = b = 0 in Theorem 1.1.7 is there for just one reason:
we do not want both of the arbitrary constants on the right to be eliminated by multiplying by
zero. We will explain this more fully.
Since each indenite integral actually yields a set, or collection, of functions, there may be
some question in your mind about what it means to multiply a set of functions by a constant,
like a or b, and what it means to add two such sets. In other words, you may wonder exactly
what the right-hand side of the equality in Theorem 1.1.7 means.
For instance, what does it mean to write that
_
(5x
3
7

x) dx = 5
_
x
3
dx 7
_

xdx?
We know, from the Power Rule for Integration, that
_
x
3
dx = x
4
/4 +C
1
and
_

xdx =
_
x
1/2
dx =
x
3/2
3/2
+C
2
= 2x
3/2
/3 +C
2
,
where we have used C
1
and C
2
, in place of using simply C twice, since we dont want to assume
that we have to pick the two arbitrary constants to be the same thing.
So, what does 5
_
x
3
dx mean? It means the collection of functions obtained by taking 5
times any function from the collection of functions x
4
/4+C
1
; that is, the collection of functions
1.1. BASIC ANTI-DIFFERENTIATION 7
5x
4
/4 + 5C
1
, where C
1
could be any constant. But, if C
1
can be anything, then 5C
1
can be
anything, and we might as well just call it B
1
, where B
1
can be any number. Thus, we can
write the collection of functions 5
_
x
3
dx as 5x
4
/4 +B
1
.
However, heres the part that can be confusing; instead of using a new constant name, like
B
1
, it is fairly standard to just use the name C
1
again, i.e., to use C
1
to now denote 5 times
the old value of C
1
. Assuming that we had not determined some value for the old C
1
, there is
no harm in doing this, but it certainly can make things look confusing, for you frequently see
calculations like
5
_
x
3
dx = 5
_
x
4
/4 +C
1
_
= 5x
4
/4 +C
1
,
where the C
1
on the far right above is actually 5 times the C
1
in the middle.
Similarly, we write
7
_

xdx =
_
x
1/2
dx = 7
_
2x
3/2
/3 +C
2
_
= 14x
3/2
/3 +C
2
.
Therefore,
5
_
x
3
dx 7
_

xdx = 5x
4
/4 +C
1
14x
3/2
/3 +C
2
=
5x
4
/4 14x
3/2
/3 + (C
1
+C
2
) = 5x
4
/4 14x
3/2
/3 +C,
where C = C
1
+C
2
can be any real number.
Using this example as a guide, you can see what to do more generally: whenever you have
a linear combination of indenite integrals, i.e., a sum of constants multiplied times indenite
integrals, you do not include an arbitrary constant for each individual indenite integral; instead,
for each indenite integral you write one particular anti-derivative, and then put in a single +C
at the end.
8 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
The fact that an indenite integral is actually a collection of functions can lead to seemingly
bizarre results. For instance, while its true that
_
xdx =
_
xdx, it would nonetheless be
bad to write that
_
xdx
_
xdx = 0. Why? Because our operations on sets of functions
tell us that the correct calculation is
_
xdx
_
xdx = x
2
/2 +C
1

_
x
2
/2 +C
2
_
= C
1
C
2
= C,
which is the same as
_
0 dx.
This agrees with what we wrote above: when you have a linear combination of indenite
integrals, you should use one particular anti-derivative for each integral, and then add a C
at the end. Thus, in the above calculation, you should get x
2
/2x
2
/2+C, which is just C.
Example 1.1.9. Calculate the indenite integral
_ _
5
w
3 + 7w
3
+ 5
9

w
_
dw.
Solution:
We calculate
_ _
5
w
3 + 7w
3
+ 5
9

w
_
dw =
_ _
5
1
w
3 + 7w
3
+ 5w
1/9
_
dw =
5 ln [w[ 3w + 7
w
4
4
+ 5
w
(1/9)+1
(1/9) + 1
+C =
5 ln [w[ 3w +
7w
4
4
+
9w
10/9
2
+C.
Example 1.1.10. Suppose that an object moves in a straight line with constant acceleration a
meters per second per second. Show that the position of the object p = p(t), in meters, is given
by
p = at
2
/2 +v
0
t +p
0
,
1.1. BASIC ANTI-DIFFERENTIATION 9
where p
0
is the initial position of the object (i.e., the position at t = 0), in meters, v
0
is the
initial velocity in m/s, and t is the time in seconds.
Solution:
Acceleration a is the derivative, with respect to t, of the velocity v, i.e., a = dv/dt. This is
exactly the same as writing that v is an anti-derivative of a, with respect to t, i.e., v =
_
a dt.
Since a is a constant, we nd
v =
_
a dt = a
_
1 dt = at +C m/s,
for some constant C. Therefore, v = at + C, but we would like to give some physical meaning
to the constant C.
How do we do this? We are not given any other data. The answer is that we use tautological
initial data; that is, we use initial data that is simply obviously true. We use that, when t = 0,
the velocity v equals v
0
. Why is this true? Because it says something thats clearly true: at time
0, the velocity equals the velocity at time 0. It may seem strange, but using this tautological
initial data actually gets us somewhere.
We have v = at + C. Now, plug in that, when t = 0, v = v
0
. You nd v
0
= a 0 + C, i.e.,
C = v
0
. Thus, we conclude that v = at +v
0
. Notice that no one has to give you any extra data
to conclude that C = v
0
; it follows from the equation v = at +C.
Now, we have that v = at +v
0
, and we know that the velocity v equals the rate of change of
p, with respect to time, i.e., v = dp/dt. This is the same as writing that p =
_
v dt. Therefore,
we have
p =
_
v dt =
_
(at +v
0
) dt = a
__
t dt
_
+v
0
__
1 dt
_
=
at
2
/2 +v
0
t +C meters,
where this C is denitely not the same C that we used in the equation for v.
How do we nd this new C? We plug in more tautological initial data, namely that p = p
0
when t = 0. We nd that p
0
= a(0)
2
/2 +v
0
(0) +C, and so C = p
0
. Thus, we conclude
p = at
2
/2 +v
0
t +p
0
meters.
Other integration formulas obtained at once from dierentiation formulas are:
10 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
Theorem 1.1.11. As functions on the entire real line (, ), we have
1. _
cos xdx = sin x +C;
2. _
sin xdx = cos x +C; and
3. _
e
x
dx = e
x
+C.
Note that the integration formulas for sin and cos have the negative sign in the opposite
place from the dierentiation formulas. This frequently leads to confusion. It shouldnt.
Remember: you are nding anti-derivatives. This means that the derivative of what you
end up with should be what you started with (i.e., the integrand).
Lets look at another example in which you are given the acceleration of an object, and are
asked to nd the velocity and position, but, this time, we have an acceleration which is not
constant.
Example 1.1.12. Suppose that the acceleration, in m/s
2
, of an object moving in a straight
line is a = sin t, where t is the time in seconds. Find the velocity and position of the object, as
functions of time, in terms of the initial velocity and initial position.
Solution:
We nd that
v =
_
a dt =
_
sin t dt = cos t +C,
and, plugging in the tautological initial data, we nd v
0
= cos(0) + C = 1 + C. Thus,
C = v
0
+ 1, and
v = cos t +v
0
+ 1 m/s.
Now, we integrate the velocity to nd the position:
p =
_
v dt =
_
(cos t +v
0
+ 1) dt =
_
cos t dt + (v
0
+ 1)
_
1 dt =
sin t + (v
0
+ 1)t +C.
1.1. BASIC ANTI-DIFFERENTIATION 11
Finally, we plug in the tautological initial data, in order to give physical meaning to this last
C:
p
0
= sin(0) + (v
0
+ 1)(0) +C.
Therefore, C = p
0
, and we nd
p = sin t + (v
0
+ 1)t +p
0
meters.
We shall not rewrite, as anti-dierentiation formulas, every one of the dierentiation formulas
that you should know; the standard anti-dierentiation formulas are contained in Appendix B.
However, well go ahead and give two more, before looking at the Chain Rule and the Product
Rule in their indenite integral forms.
Theorem 1.1.13. As functions on the open interval (1, 1),
_
1

1 x
2
dx = sin
1
x +C.
As functions on the open interval (, ),
_
1
1 +x
2
dx = tan
1
x +C.
Example 1.1.14. Calculate
_
7z
2
+ 5
z
2
+ 1
dz.
Solution:
12 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
We begin by simplifying via long division of polynomials, except the division is not so long
in this case. We get clever and write 7z
2
= 7(z
2
+ 1) 7, and nd
7z
2
+ 5
z
2
+ 1
=
7(z
2
+ 1) 7 + 5
z
2
+ 1
= 7 2
1
z
2
+ 1
.
Thus,
_
7z
2
+ 5
z
2
+ 1
dz =
_ _
7 2
1
z
2
+ 1
_
dz = 7
_
1 dz 2
_
1
z
2
+ 1
dz =
7z 2 tan
1
z +C.
Substitution: the Chain Rule in anti-derivative form:
The Chain Rule for dierentiation tells you how to dierentiate a composition of functions.
If f and g are dierentiable, then
_
f(g(x))
_

= f

(g(x))g

(x) or, letting u = g(x), the Chain

Rule can be written as
d
dx
_
f(u)
_
= f

(u)
du
dx
.
As an anti-derivative formula, this becomes
Theorem 1.1.15. (Integration by Substitution) If f and g are dierentiable functions,
then _
f

(g(x))g

(x) dx = f(g(x)) +C,

or, letting u = g(x),
_
f

(u)
du
dx
dx =
_
f

(u) du = f(u) +C.

The second formula for substitution is particularly easy to use; it looks as though the dxs
cancel, as in multiplying fractions. This is not whats happening, but it does make substitution
1.1. BASIC ANTI-DIFFERENTIATION 13
easier to remember. It also means that the dierential notation
du =
du
dx
dx,
which we introduced in [2], yields correct formulas in integrals, and, hence, we use it extensively.
Example 1.1.16. Calculate the integral
_
cos(e
x
+ 7) e
x
dx.
Solution:
How should you approach an integral like
_
cos(e
x
+ 7) e
x
dx?
First, you should realize that its not just a linear combination of integrals of specic functions
that youve memorized. You should then think Well...cos(e
x
+7) is a composition of functions.
What if I make a substitution for the inside function? Ill let u = e
x
+ 7, so that cos(e
x
+ 7)
becomes cos u, and see if the remaining part of the integrand looks like du.
In fact, for easy cases, you can do this in your head. If u = e
x
+ 7, then
du =
du
dx
dx = e
x
dx,
and you see that this is the remaining factor in the integral. Thus, by substitution, our original
integral is transformed into an integral in terms of u that is very simple:
_
cos(e
x
+ 7) e
x
dx =
_
cos udu = sin u +C = sin(e
x
+ 7) +C.
Nice.
14 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
Example 1.1.17. Calculate the integral
_
t
1 +t
2
dt.
Solution:
You once again realize that this integral is not some linear combination of basic integrals
that have memorized, and theres no obvious composition of functions this time. You might
think that tan
1
t is involved somehow, since
_
1/(1 + t
2
) dt = tan
1
t + C, but the t in the
numerator causes a problem.
You might think that you can factor the integrand and use that in some way:
_
t
1 +t
2
dt =
_
t
1
1 +t
2
dt
?
=
?
_
t dt
_
1
1 +t
2
dt =
t
2
2
tan
1
t +C.
THIS IS COMPLETELY WRONG. You cant just integrate each factor in a product
and then multiply the results; dealing with products in an integral is more complicated than
that.
We will get to the integral form of the Product Rule shortly, but, for now, you should
dierentiate t
2
tan
1
t/2 (using the Product Rule) and verify that you dont get anything close
to our integrand t/(1 +t
2
).
Great. Now we know one way NOT to nd the integral
_
t/(1 +t
2
) dt. We also know that
were discussing substitution here, and so you should suspect that a substitution is involved.
With practice, you should actually see relatively quickly that if you let w = 1 + t
2
, then
dw = 2t dt, and we have a t dt in the integral, so this substitution might be good. We can
always x multiplying by a constant, like the 2 in dw = 2t dt. We divide by 2 to get
1
2
dw = t dt,
and
_
t
1 +t
2
dt =
_
1
1 +t
2
t dt =
_
1
w

1
2
dw =
1.1. BASIC ANTI-DIFFERENTIATION 15
1
2
_
1
w
dw =
1
2
ln [w[ +C =
1
2
ln [1 +t
2
[ +C =
1
2
ln(1 +t
2
) +C,
where the last equality follows from the fact that 1 +t
2
0 (in fact, 1 +t
2
1).
It would be a good exercise for you to dierentiate our nal answer above, and see how the
Chain Rule comes into play to produce our initial integrand.
By making the substitution u = x/a, so that x = au, we easily obtain:
Theorem 1.1.18. Suppose that a > 0 is a constant. As functions on the open interval
(a, a),
_
1

a
2
x
2
dx = sin
1
_
x
a
_
+C.
As functions on the open interval (, ),
_
1
a
2
+x
2
dx =
1
a
tan
1
_
x
a
_
+C.
Integration by Parts: the Product Rule in anti-derivative form:
The Product Rule as an anti-derivative formula is
16 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
Theorem 1.1.19. (Integration by Parts) If f and g are dierentiable functions, then
_
f(x)g

(x) dx +
_
g(x)f

(x) dx = f(x)g(x) +C,

or, letting u = f(x) and v = g(x),
_
udv +
_
v du = uv +C
or _
udv = uv
_
v du.
It is the last formula above that most people memorize as THE formula for integration by
parts.
Heres how a basic integration by parts attack on a problem goes. You look at your integral,
and make a choice that lets you write the integral in the form
_
udv. Then, you apply the
integration by parts formula to obtain that your integral equals uv
_
v du. Then, you hope
that the new integral
_
v du is easier (or, at least, no harder) to integrate than the integral that
you started with.
Example 1.1.20. Calculate
_
ze
z
dz and
_
z
2
e
z
dz.
Solution:
You should look at
_
ze
z
dz, and realize quickly that ze
z
does not result from one of the
basic derivative formulas that you should have memorized, and so this integration is not a basic
one that you should know immediately. In addition, if you think for a minute or so, you should
be able to convince yourself that theres no substitution that will help. However, we do see that
the integrand is the product of two very dierent-looking functions, z and e
z
; this is a hint that
integration by parts may be good to use.
1.1. BASIC ANTI-DIFFERENTIATION 17
Now what do you do? You identify some factor in the integrand that will be u; that factor
should not contain the dierential (here, dz). The remaining part of the integrand, together
with the dierential, should be dv. With u and dv determined, you calculate du and v =
_
dv.
You do not need to include an arbitrary constant in your calculation of v; we need some v, not
all possible vs. Then, you write uv
_
v du, look at your new integral, and hope that its easier.
Lets see how this works for
_
ze
z
dz. There are two obvious choices for u: either u = z
or u = e
z
. Either one of these will lead to a formula that is true, but only one will lead to
something useful.
Lets look at the bad choice rst, so that you can see how you can tell when youve made
a bad choice. Lets try u = e
z
. That leaves z dz to be dv. Now, if u = e
z
and dv = z dz, then
du = e
z
dz and v =
_
dv =
_
z dz = z
2
/2, where, as we discussed, we dont include a +C in our
calculation of v. Applying the integration by parts formula, we nd
_
ze
z
dz =
_
e
z
z dz =
_
udv = uv
_
v du =
(e
z
)(z
2
/2)
_
(z
2
/2)e
z
dz. (1.1)
This is true, but not particularly helpful for calculating
_
ze
z
dz. The power of z went up, and,
after moving the constant 1/2 outside the integral, the rest of the integral, the e
z
, is the same
as what we started with. You should realize that the new integral is harder to deal with than
the original. We could try to integrate by parts again, but, we leave it as an exercise for you to
verify, depending on your new choice of u, that either the power of z goes up again (and the e
z
remains), or the power of z goes back down to 1, but you end up with exactly the integral that
we started with.
So, lets make the other choice for u in integrating by parts to calculate
_
ze
z
dz. Let u = z,
which means that dv = e
z
dz. Then, we nd that du = dz and v =
_
dv =
_
e
z
dz = e
z
. Thus,
we obtain
_
ze
z
dz =
_
udv = uv
_
v du =
ze
z

_
e
z
dz = ze
z
e
z
+C.
Why did this choice of u work better than our earlier one? Before, when we picked u to not
include the power of z, the power of z was left in dv; when we integrated dv to get v, the power
of z went up. Now, when we pick u to be the power of z, namely z
1
, the power of z goes down in
the calculation of du. For this reason, it is frequently (but not always see the next example) a
good idea in integration by parts problems which include powers of the variable to let the power
of the variable be u.
18 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
With this in mind, how do you integrate
_
z
2
e
z
dz? You integrate by parts, letting u = z
2
,
which means dv = e
z
dz. What you should get is actually exactly what we got in Formula 1.1,
except you need to multiply Formula 1.1 by 2, and rearrange, to obtain:
_
z
2
e
z
dz = z
2
e
z
2
_
ze
z
dz.
Now, even if we had not already calculated
_
ze
z
dz, you should realize that the new integral on
the right above is easier than the one you started with, and you would calculate it by a second
integration by parts (if we had not already done so). What we nd is
_
z
2
e
z
dz = z
2
e
z
2
_
ze
z
dz = z
2
e
z
2 (ze
z
e
z
+C) = z
2
e
z
2ze
z
+ 2e
z
+C.
It might be instructive to dierentiate the nal result above, on the right, and verify that you
obtain z
2
e
z
.
Example 1.1.21. Calculate
_
t
5
ln t dt and
_
ln t dt.
Solution:
You look at
_
t
5
ln t dt, and you realize immediately that this doesnt come from one of our
basic derivative/integral formulas. You might think about a substitution, like w = ln t, for a
minute or so, but then realize that it doesnt get you anywhere. Then, you decide that, since the
integrand is the product of two dierent kinds of functions, maybe integration by parts would
be a good thing to try.
If you look at our previous example, youd probably be tempted to let u be the power of
t, i.e., let u = t
5
, which would lead to dv = ln t dt. However, this is one of those times when
picking the power of the variable to be u is a bad idea. Perhaps the most obvious reason why
this is bad is because we dont know how to calculate v =
_
dv =
_
ln t dt. You may think that
weve discussed this integral, and that it equals 1/t + C. This is very wrong. We know that
(ln t)

= 1/t or, whats the same thing,

_
1/t dt = ln t +C (for t > 0), but we dont know
_
ln t dt
1.1. BASIC ANTI-DIFFERENTIATION 19
(its possible that you do, but it hasnt been discussed in the book up to this point). In fact,
calculating
_
ln t dt is the second part of this example.
So, to calculate
_
t
5
ln t dt,
well try integration by parts with u = ln t, which means that dv = t
5
dt.
We nd that du = (1/t) dt and v =
_
dv =
_
t
5
dt = t
6
/6 (remember, we dont need a +C
here). Applying the integration by parts formula, we nd
_
t
5
ln t dt =
_
udv = uv
_
v du = (ln t)(t
6
/6)
_
(t
6
/6)(1/t) dt =
t
6
ln t
6

1
6
_
t
5
dt =
t
6
ln t
6

1
6

t
6
6
+C =
t
6
ln t
6

t
6
36
+C.
Now, lets look at
_
ln t dt. How could this possibly be an integration by parts problem?
Theres no product in the integrand! Admittedly, this does not look like an integration by parts
problem. Nonetheless, it is. Let u = ln t, which means that dv = dt. Then, du = (1/t) dt, and
v =
_
dv =
_
dt = t. Applying the integration by parts formula, we obtain
_
ln t dt =
_
udv = uv
_
v du = (ln t)(t)
_
t(1/t) dt =
t ln t
_
1 dt = t ln t t +C.
We wish to look at one more integration by parts example, a complicated example.
Example 1.1.22. Calculate
_
e
x
cos xdx.
Solution:
We will calculate this integral via integration by parts. We could use either u = e
x
or
u = cos x; this time, it actually makes little dierence. We will pick u = e
x
, which means that
dv = cos xdx. (As an exercise, you should try starting with u = cos x instead.) We nd, then,
20 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
that du = e
x
dx and v =
_
dv =
_
cos xdx = sin x. The integration by parts formula tells us
that we have
_
e
x
cos xdx =
_
udv = uv
_
v du = e
x
sin x
_
(sin x)e
x
dx. (1.2)
But our new integral,
_
(sin x)e
x
dx is clearly just as dicult to integrate as the original integral.
Are we getting anywhere? The answer is yes, but its not obvious yet. We will integrate
_
(sin x)e
x
dx by parts also.
You have to be careful. If you make the choice u = sin x here, you will obtain that
_
(sin x)e
x
dx equals e
x
sin x
_
e
x
cos xdx. If you substitute this into Formula 1.2, you will
nd that you have undone our original integration by parts, and you will come to the stunning
conclusion that
_
e
x
cos xdx =
_
e
x
cos xdx. Its true, but not very helpful.
When integrating
_
(sin x)e
x
dx by parts, you need to make the choice of u that is analogous
to your choice of u for the original integration by parts; you let u = e
x
again, and so dv = sin xdx.
Then, du = e
x
dx and v =
_
dv =
_
sin xdx = cos x. Applying integration by parts, we nd
_
(sin x)e
x
dx =
_
udv = uv
_
v du = e
x
(cos x)
_
(cos x)e
x
dx =
e
x
cos x +
_
e
x
cos xdx.
If you insert this result into Formula 1.2, you obtain
_
e
x
cos xdx = e
x
sin x +e
x
cos x
_
e
x
cos xdx.
At this point, you may be thinking to yourself Aaaaagggghhhhhh! We spent all that time
integrating by parts, only to end up with the same integral that we started with! However, the
fact that the new occurrence of the original integral has a negative sign in front of it saves us. If
you simply add
_
e
x
cos xdx to each side of this last equation (and x the loss of the arbitrary
constant on the right), you obtain
2
_
e
x
cos xdx = e
x
sin x +e
x
cos x +C,
1.1. BASIC ANTI-DIFFERENTIATION 21
and so
_
e
x
cos xdx =
e
x
sin x +e
x
cos x
2
+C.
We end this section with a possibly surprising complication that exists for anti-dierentiation;
a type of complication which does not occur for dierentiation.
Remark 1.1.23. From the derivative formulas in [2], we see that the derivative of any elemen-
tary functionis again an elementary function. You might hope that anti-derivatives/integrals
would behave equally as well. They do not. It is easy to give elementary functions f(x) for
which it is possible to prove that there is no elementary function F(x) such that F

(x) = f(x),
i.e., f(x) has no elementary anti-derivative. Such functions f(x) include e
x
2
, e
x
2
, sin(x
2
), and
cos(x
2
). This was rst proved by Liouville in 1835.
The Fundamental Theorem of Calculus, Theorem 2.4.7, guarantees that the functions e
x
2
,
e
x
2
, sin(x
2
), and cos(x
2
), and, in fact, all continuous functions, have some anti-derivative, but
those anti-derivatives need not be elementary functions.
1.1.1 Exercises
Calculate the general anti-derivatives in Exercises 1 through 21.
1.
_
(4x
2
+ 4x + 9) dx
2.
_ _
5
w
7e
w
+ 6
3

w
_
dw
3.
_ _
5 sin t
3

1 t
2
_
dt
4.
_
1 +v +

v
v
2
dv
5.
_ _
1
y
+
1
y
2
+ 1
_
dy
6.
_
5
3z 7
dz
7.
_
cos(2 1) d
22 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
8.
_
e
p+4
dp
9.
_
r
r
2
4
dr
10.
_ _
x

x
2
1
+
1
[x[

x
2
1
_
dx
11.
_
(5 3)
100
d
12.
_
_
2 cos(2t + 5) + 3 sin(9t)
_
dt
13.
_
ln
_
(x + 2)
x+5

dx
14.
_
e
1/x
x
3
dx
15.
_
5
xln x
dx
16.
_
e
1/x
6x
2
dx
17.
_
tan d
18.
_
5
4 +x
2
dx
19.
_
t
4
_
t
5
+ 6 dt
20.
_
1
x
2
+ 4x + 5
dx Hint: x
2
+ 4x + 5 = (x + 2)
2
+ 1.
21.
_
1

x
2
+ 6x 8
dx
In each of Exercises 22 through 31, nd the anti-derivative of the given function
which satises the given initial condition. The anti-derivative of each function with
a lower-case name is denoted by the upper-case version of the same letter.
22. h(x) = 4x
2
+ 4x + 9, such that H(1) = 2.
23. p(w) =
5
w
7e
w
+ 6
3

w, such that P(1) = 0.

24. q(t) = 5 sin t
3

1 t
2
, such that Q(0) = 7.
1.1. BASIC ANTI-DIFFERENTIATION 23
25. k(v) =
1 +v +

v
v
2
, such that K(1) = 2.
26. b(y) =
1
y
+
1
y
2
+ 1
, such that B(1) = 0.
27. f(x) = x
2
+xsin x, such that F() = 2.
28. s(t) =
2
t(ln t)
2
, such that S(e
2
) = 5.
29. g(x) = x

x + 1, such that G(0) = 1.

30. w(y) =
tan
1
_
y
2
_
4 +y
2
, such that W(2) = .
31. r(t) = te
1t
2
t, such that R(1) =

2.
In each of Exercises 32 through 41, use integration by parts to nd the indicated
anti-derivative.
32.
_
xe
3x
dx
33.
_
(x 5)
2
e
x
dx
34.
_
t sin(2t) dt
35.
_
t
2
cos t dt
36.
_

p ln p dp
37.
_
ln t
t
2
dt
38.
_
e
x
sin xdx
39.
_
e
2x
sin(5x) dx
40.
_
tan
1
wdw
41.
_
wtan
1
wdw Hint: At some point, you may want to use that w
2
= (1 +w
2
) 1.
42. Suppose that the net force F, acting on an object of mass m, pushes the object along the
x-axis with an acceleration function, in m/s
2
, of
a(t) = sin(2t),
where t 0 is measured in seconds.
24 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
a. Recall that F = ma and that momentum p = mv. Find the momentum of the object,
as a function of time, if the mass of the object is 10 kilograms, and momentum at
time t = 0 is 20 kilogram meters per second.
b. What is the momentum of the object at time t = 4 seconds?
43. Repeat the preceding problem with the new acceleration function a(t) = t sin(2t).
44. For each positive integer n, dene f
n
() = sin
n
cos and g
n
() = cos
n
sin
a. Find
_
f
n
() d and
_
g
n
() d.
b. Find the specic anti-derivatives F
n
() and G
n
() that satisfy the initial conditions
F
n
(0) = 5 and G
n
_

2
_
= 4.
In Exercises 45 through 48, you are given the velocity of a particle at time t, and
the position p(t
0
) of the particle at a specic time t
0
. Find the position function.
45. v(t) = 3t
2
4t + 3, p(1) = 4.
46. v(t) = t + cos(t), p(0) = 0.
47. v(t) = 2t

18 + 7t
2
, p(1) = 8.
48. v(t) = at +b, p(2) = 5. Leave your answer in terms of a and b.
49. In the following steps, you will calculate the general anti-derivatives for sin
2
x and cos
2
x.
a. Apply integration by parts to
_
sin x sin xdx (written suggestively).
b. Integration by parts yields a new anti-derivative. Use a trigonometric identity to
write this new anti-derivative in terms of sin
2
x, and solve your integration by parts
equation for
_
sin
2
xdx.
c. What is
_
cos
2
xdx? Hint: Use your answer to part (b).
d. From the cosine double angle formula, cos(2x) = 2 cos
2
x 1. Use this to integrate
cos
2
x, and explain why the dierent-looking answer that you obtain is, in fact, the
same as your answer from part (c).
Exercises 50 and 51 show that the argument in Exercise 49 can be generalized to
calculate anti-derivatives of higher powers of sin and cos.
1.1. BASIC ANTI-DIFFERENTIATION 25
50. a. Use integration by parts to prove that
_
sin
n
t dt =
1
n
cos t sin
n1
t +
n 1
n
_
sin
n2
t dt.
Assume n 2.
b. Use this formula to calculate
_
sin
2
t dt. Check your answer by comparing to the
previous problem.
51. a. Use integration by parts to prove that
_
cos
n
t dt =
1
n
cos
n1
t sin t +
n 1
n
_
cos
n2
t dt.
b. Use the formula in part (a) to determine
_
cos
2
t dt.
52. Suppose that instantaneous rate of change, with respect to time, of a population of an
island at time t, measured in years, where t = 0 corresponds to the year 2000, is given by
1
2

6, 250, 000 +t

500
(t + 1)
2
.
The population of the island in 2000 is 3000.
a. Find an explicit formula for the population at time t.
b. What is the predicted population in 2050?
53. Prove that the argument used to calculate
_
ln t dt can be generalized. Assume that f(t)
is dierentiable and prove that
_
f(t) dt = tf(t)
_
tf

(t) dt.
54. Calculate
_
e
x
sinh xdx. Hint: do not use integration by parts.
26 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
55. Consider the following logic in calculating
_
e
x
sinh xdx using integration by parts.
_
e
x
sinh xdx = e
x
sinh x
_
e
x
cosh xdx
= e
x
sinh x
_
e
x
cosh x
_
e
x
sinh xdx
_

0 = e
x
sinh x e
x
cosh x
e
x
sinh x = e
x
cosh x.
Since e
x
> 0, we can divide and conclude that sinh x = cosh x. What is the aw in this
argument?
56. Prove the formula
_
t
n
e
t
dt = t
n
e
t
n
_
t
n1
e
t
dt.
Assume that n 1.
57. Calculate
_
1
x
2
+a
2
dx, where a ,= 0 is a constant. Hint: use Theorem 4.2.14 and an
appropriate substitution.
58. Prove that
_
dx

x
2
+a
2
= sinh
1
_
x
a
_
+C.
59. Calculate
_
1

a
2
x
2
dx, where a > 0 is a constant.
60. Calculate
_
1
[x[

x
2
1
dx. Hint: consider sec
1
x.
61. Calculate
_
e
x
e
e
x
dx. Hint: use substitution.
62. Let f
1
(x) = e
x
and, for all integers n 2, let f
n
(x) = e
fn1(x)
. Prove that
_
f
1
(x) f
2
(x) ... f
n
(x) dx = f
n
(x) +C.
63. Calculate
_
2
x
dx. Hint: rewrite 2
x
as an exponential expression with base e.
64. Calculate
_
x + 4

x + 2
dx.
65. Calculate
_
ln(1 +x
2
) dx.
1.1. BASIC ANTI-DIFFERENTIATION 27
66. Calculate
_
e
3x
e
2x
e
2x
1
dx.
Consider a simple electric circuit with an inductor, but no resistor or capacitor.
A battery supplies voltage V (t). If inductance is constantly L, in henrys, then
Kirchos Law from Example 2.7.18 of [2] tells us that the current i at time t
satises the dierential equation L
di
dt
= V (t). In Exercises 67 through 70, nd an
explicit formula for i(t), given the condition i(t
0
) = i
0
.
67. L = 12, V (t) = sin t, i(0) = 0.
68. L = 9, V (t) = 12, i(3) = 12.
69. L = 3, V (t) =
t
t
2
+ 1
, i(0) = 0.
70. L = 13, V (t) =
ln(1/t)
t
2
, i(1) = 2.
For each of the functions in Exercises 71 through 74, verify that (a)
d
dx
__
f(x) dx
_
=
f(x) and (b)
_ _
d
dx
f(x)
_
dx = f(x) +C. Assume an appropriate domain for f(x).
71. f(x) = x
4
.
72. f(x) = 3 cos(2x).
73. f(x) = ln x.
74. f(x) =
1
1 +x
2
.
75. In the following steps, you will nd the general anti-derivatives for functions of the form
t
n
ln t.
a. Suppose that n ,= 1. Apply integration by parts to nd
_
t
n
ln t dt.
b. Find
_
t
1
ln t dt.
76. Suppose that water ows out of a hole 0.1 square meters in area from the bottom of a
cylindrical tank with a base radius of 2 meters and an initial height of 10 meters at a rate
dV
dt
= 0.007798t 1.3999 cubic meters per second.
28 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
a. If the tank starts out full, what is the function V (t) for the volume in the tank at
time t?
b. Calculate the amount of water remaining in the tank one minute after the leak starts.
c. Verify that
dV
dt
= 0 precisely when the tank is empty.
77. A particle is traveling along the curve y = x
2
, so that its x-coordinate is a function of
time t, measured in minutes. Suppose that the horizontal velocity (i.e., velocity in only
the horizontal direction) is given by dx/dt = 0.5 cos
3
t miles per minute.
a. What is the maximum horizontal speed (absolute value of velocity) on the time in-
terval 0 t 20?
b. Find the function x(t), the x-coordinate of the particle at time t, subject to the
condition that the particle is at the origin at time t = 0.
c. Find the function y(t), the y-coordinate of the particle at time t.
d. What is the vertical velocity of the particle at time t = minutes?
78. An enclosed room is built in order to experiment with the eects of pressure changes on
objects. Suppose that the equipment is capable of decreasing the pressure in the room
at a rate of
2
t+1
0.04t kilopascals, kPa, per second, and that it can run for up to one
minute before overheating.
a. If the internal pressure in the room is 101.325 kPa (equal to the standard atmospheric
pressure or atm), nd an expression for P(t).
b. How long will it take to reach 50 kPa (this is approximately the atmospheric pressure
ve and a half kilometers above the surface of the Earth)?
c. What is the lowest pressure that can be reached in the room?
In Exercises 79 through 82, you are given the acceleration, a = a(t) in m/s
2
, of an
object moving in a straight line, where t is the time in seconds. Find the velocity
v and position p of the object, as functions of time, in terms of the initial velocity
v
0
and initial position p
0
.
79. a = t + 1
80. a = sin t + cos t
81. a = e
3t
82. a = te
t
1.2. SPECIAL TRIG. INTEGRALS AND TRIG. SUBSTITUTIONS 29
1.2 Special Trigonometric Integrals
and Trigonometric Substitutions
In this section, we will discuss some special, important, trigonometric integrals, and then give
three examples of integration by trigonometric substitution.
Proposition 1.2.1.
_
tan d = ln [ cos [ +C = ln [ sec [ +C.
Proof. We have
_
tan d =
_
sin
cos
d.
Let u = cos , so that du = sin d, i.e., du = sin d. We nd
_
sin
cos
d =
_
du
u
= ln [u[ +C = ln [ cos [ +C.
Proposition 1.2.2.
_
sec d = ln [ sec + tan [ +C.
Proof. As (tan )

= sec
2
and (sec )

= sec tan , it follows that

(tan + sec )

= sec
2
+ sec tan = (sec )(tan + sec ).
This means that, if u = tan + sec , then du/d = usec , i.e.,
_
sec d =
_
1
u
du = ln [u[ +C = ln [ sec + tan [ +C.
30 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
Proposition 1.2.3.
_
sec
3
d =
sec tan + ln [ sec + tan [
2
+C.
Proof. This is a fun integration by parts problem.
_
sec
3
d =
_
sec (1 + tan
2
) d =
_
sec d +
_
sec tan
2
d =
ln [ sec + tan [ +
_
(tan )(sec tan ) d.
We approach this last integral by parts; let u = tan and dv = (sec tan ) d. Then, du =
sec
2
d and v =
_
dv = sec . We nd
_
(tan )(sec tan ) d =
_
udv = uv
_
v du = tan sec
_
sec
3
d.
Combining this with our previous equality, we obtain
_
sec
3
d = sec tan + ln [ sec + tan [
_
sec
3
d.
Adding
_
sec
3
d to each side, xing the missing +C, and dividing by 2 yields the desired
result.
Remark 1.2.4. You may wonder why we didnt give integral formulas for the co-functions in
the previous three propositions. The reason for this is that you should memorize fundamental
integral formulas, and then use general techniques to quickly derive others, if possible.
Since replacing by /2 changes any trig function into the corresponding co-function, the
substitution u = /2 tells us that the integrals of the co-functions of those in Proposition 1.2.1,
1.2. SPECIAL TRIG. INTEGRALS AND TRIG. SUBSTITUTIONS 31
Proposition 1.2.2, and Proposition 1.2.3 are obtained by negating what we obtained and replacing
all of the trig functions by their co-functions, i.e.,
_
cot d = ln [ sin [ +C,
_
csc d = ln [ csc + cot [ +C
and
_
csc
3
d =
csc cot + ln [ csc + cot [
2
+C.
The following two iteration formulas for integrating powers of sine and cosine are frequently
useful. They are proved by using integration by parts.
Proposition 1.2.5. If n 2 is an integer, then
_
sin
n
d =
1
n
sin
n1
cos +
n 1
n
_
sin
n2
d,
and _
cos
n
d =
1
n
cos
n1
sin +
n 1
n
_
cos
n2
d.
Proof. These are obtained from integration by parts, and the two demonstrations are entirely
similar. We will derive the
_
cos
n
d formula, and leave the other as an exercise.
We have
_
cos
n
d =
_
cos
n1
cos d.
Let u = cos
n1
and dv = cos d. Then, du = (n 1) cos
n2
sin d and v =
_
dv = sin .
We nd
_
cos
n
d =
_
udv = uv
_
v du =
cos
n1
sin
_
sin ((n 1) cos
n2
sin ) d =
cos
n1
sin + (n 1)
_
cos
n2
sin
2
d =
cos
n1
sin + (n 1)
_
(cos
n2
)(1 cos
2
) d =
32 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
cos
n1
sin + (n 1)
_
cos
n2
d (n 1)
_
cos
n
d.
Therefore, we have concluded that
_
cos
n
d = cos
n1
sin + (n 1)
_
cos
n2
d (n 1)
_
cos
n
d.
Adding (n1)
_
cos
n
d to each side of the equation, dividing by n, and replacing the missing
+C yields the desired result.
Integration by trigonometric substitution:
Integration by trigonometric substitution (trig substitution) refers to having an integral
involving some variable, say x, and letting x equal an expression involving a trig function,
e.g., x = a sin or x = a tan , and then using properties of trigonometric functions to produce
a manageable integral in terms of . We placed the word letting in quotes in the previous
sentence because we dont really get to let x be anything; it is what it is.
What we can do, however, is dene a new variable in terms of x. Hence, what we really do
are things like let = sin
1
(x/a) or let = tan
1
(x/a), which actually imply more than
x = a sin and x = a tan , respectively; they also imply that the values of are restricted to
the intervals [/2, /2] and (/2, /2), respectively.
Lets look at three examples.
Example 1.2.6. Evaluate the integral
_
_
a
2
x
2
dx, where a > 0 is a constant.
Solution:
As the integrand is

a
2
x
2
, we must have that a x a; it follows that 1 x/a 1.
This is important since the closed interval [1, 1] is the domain of sin
1
.
We now let = sin
1
(x/a), so that x = a sin and /2 /2. Then,
dx = a cos d,
and
_
a
2
x
2
=
_
a
2
a
2
sin
2
= a
_
1 sin
2
= a

cos
2
= a cos ,
1.2. SPECIAL TRIG. INTEGRALS AND TRIG. SUBSTITUTIONS 33
where we used that

a
2
= a, since a > 0, and that

cos
2
= cos , since cos 0, because
/2 /2.
Therefore, we obtain:
_
_
a
2
x
2
dx =
_
a cos a cos d = a
2
_
cos
2
d.
Now, either by using the trig identity that cos
2
= (1 + cos(2))/2, or by using the cosine
iteration formula (which is proved using integration by parts) from Appendix B, we know that
_
cos
2
d =
1
2
_
(sin cos ) +

+C.
Thus,
_
_
a
2
x
2
dx =
a
2
2
_
(sin cos ) +

+C =
1
2
_
(a sin a cos ) +a
2

+C =
1
2
_
x
_
a
2
x
2
+a
2
sin
1
_
x
a
__
+C.
Example 1.2.7. Evaluate the integral
_
_
a
2
+x
2
dx, where a > 0 is a constant.
Solution: You might suspect that this integral would turn out to be something similar to the
previous answer. After all, all that we changed was a minus sign to a plus sign. However, this
seemingly simple change alters the problem dramatically.
Let = tan
1
(x/a), so that x = a tan and /2 < < /2. Then, dx = a sec
2
d, and
_
a
2
+x
2
=
_
a
2
+a
2
tan
2
= a
_
1 + tan
2
= a

sec
2
= a sec ,
where

a
2
= a, since a > 0, and

sec
2
= sec , since sec > 0 because /2 < < /2.
Thus, we nd
_
_
a
2
+x
2
dx =
_
a sec a sec
2
d = a
2
_
sec
3
d =
34 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
a
2

sec tan + ln [ sec + tan [

2
+C =
x

a
2
+x
2
+a
2
ln

a
2
+x
2
a
+
x
a

2
+C =
x

a
2
+x
2
+a
2
ln

a
2
+x
2
+ x

2
+C,
where, in the last step, we used the property of the natural logarithm that ln(w/a) = ln wln a,
and absorbed the resulting constant a
2
ln a/2 into the constant C.
Example 1.2.8. Evaluate the integral
_
1
(a
2
+x
2
)
n
dx,
where a > 0 is a constant and n 1 is an integer.
Solution:
Again, we let = tan
1
(x/a), so that x = a tan and /2 < < /2. Then, dx =
a sec
2
d, and
a
2
+x
2
= a
2
+a
2
tan
2
= a
2
(1 + tan
2
) = a
2
sec
2
.
Thus,
_
1
(a
2
+x
2
)
n
dx =
_
1
(a
2
sec
2
)
n
a sec
2
d =
a
12n
_
1
sec
2n2

d = a
12n
_
cos
2n2
d.
This nal integral can be calculated using the cosine iteration formula in Proposition 1.2.5.
For instance, when n = 1, we recover a formula that we already knew:
_
1
a
2
+x
2
dx = a
12
_
1 d =
1
a
+C =
1
a
tan
1
_
x
a
_
+C.
When n = 2, we obtain
_
1
(a
2
+x
2
)
2
dx = a
14
_
cos
2
d =
1
a
3
_
1
2
cos sin +
1
2

_
+C =
1.2. SPECIAL TRIG. INTEGRALS AND TRIG. SUBSTITUTIONS 35
1
2a
3
_
ax
a
2
+x
2
+ tan
1
_
x
a
_
_
+C.
1.2.1 Exercises
In each of Exercise 1 through 19, calculate the anti-derivative.
1.
_
tan(3) d.
2.
_
2 sec(5t) dt.
3.
_
sec(4y) + tan(3y) dy.
4.
_
2z cot(z
2
) dz.
5.
_
(cos x) (cot(sin x)) dx.
6.
_
sec
4
u 1
sec u 1
du. Hint: factor before anti-dierentiating.
7.
_
csc t(1 + csc
2
t) dt.
8.
_
_
25
2
d.
9.
_

_
25
2
d. Hint: do not use a trig substitution. Compare your answer to that of
the previous problem.
10.
_
_
36 +y
2
dy.
11.
_
dk
(121 +k
2
)
2
.
12.
_
dx
x
4
+ 18a
2
x
2
+ 81
.
13.
_
_
10 49v
2
dv.
14.
_
_
x
2
+ 6x + 21 dx. Hint: complete the square.
36 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
15.
_
dz
(1 +z
2
)
3
.
16.
_
tan
2
y dy.
17.
_
cot
2
v dv.
18.
_
e
x
_
16 e
2x
dx.
19.
_
_
e
2x
16 dx.
20. Prove the sine iteration formula:
_
sin
n
d =
1
n
sin
n1
cos +
n 1
n
_
sin
n2
d.
Assume n 2.
21. The reduction formula below gives a method for calculating integrals of powers of the
tangent function. Verify this formula in the case n = 2.
_
tan
n
d =
1
n 1
tan
n1

_
tan
n2
d.
Use the angle addition identities to calculate anti-derivatives in the next three
exercises. Assume in each exercise that a and b are positive integers. These integrals
are extremely important in Fourier analysis.
22.
_
sin(ax) sin(bx) dx.
23.
_
sin(ay) cos(by) dy.
24.
_
cos(aw) cos(bw) dw.
25. Explain why the results in the last three problems do not hold in the case a = b.
26. Assume that n 1 and show that:
_
z
n
sin z dz = z
n
cos z +n
_
z
n1
cos z dz.
1.2. SPECIAL TRIG. INTEGRALS AND TRIG. SUBSTITUTIONS 37
27. Assume that n 1 and show that:
_
z
n
cos z dz = z
n
sin z n
_
z
n1
sin z dz.
Use the integral formulas from the previous problems to calculate the anti-derivatives
in Exercises 28 - 38.
28.
_
x
2
sin xdx.
29.
_
y
2
cos y dy.
30.
_
w
3
sin wdw.
31.
_
z
3
cos z dz.
32.
_
sin
3
d.
33.
_
sin
4
d.
34.
_
tan
3
t dt.
35.
_
tan
4
udu.
36.
_
sin(9x) sin(2x) dx.
37.
_
cos(4x) cos xdx.
38.
_
sin(3t) cos(5t) dt.
Recall that the momentum, p, of a mass m moving with velocity v is given by p = mv.
In Exercises 39 - 42 you are given the mass, acceleration, and the momentum at
one specic time. Find the momentum function, p(t). The units of acceleration are
meters per second per second. The units of momentum are
kg m
sec
.
39. m = 12 kg, a(t) = 2t
3
_
16 t
2
, p(0) = 8.
38 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
40. m = 9 kg, a(t) =
1
t
2

t
2
15
, p(4) = 6.
41. m = 5 kg, a(t) =
6t
3

t
2
+ 9
, p(0) = 0.
42. m = 20 kg, a(t) =
t

20 8t t
2
, p(1) = 1.
Integrate the following products of trig. functions.
43.
_
sin
3
xcos
2
xdx.
44.
_
sin
3
wcos
3
wdw.
45.
_
tan y sec
2
y dy.
46.
_
tan
3
z sec
2
z dz.
Find the general solution to the separable dierential equation.
47.
dx
dt
=

9 t
2

4 +x
2
t
2
x
.
48.
dy
dx
=
3
_
y
2
16
2x
2

x
2
+ 4
.
49.
dx
dz
=
z
3
x

z
2
+ 9 9x
2
x
2
z
2
.
50.
dy
dx
=
_
2y y
2
2x
2
y +x
2
y
2
y 1
. Assume [x[ < 1 and [y[ <

2.
1.3. INTEGRATION BY PARTIAL FRACTIONS 39
1.3 Integration by partial fractions
A rational function is one dened by the quotient of two polynomial functions, e.g.,
x
3
7x +
x + 5
and
x
2
5x + 1
x
3
+x
2
,
where the domains exclude roots of the denominators. This section describes the fundamental
techniques for integrating rational functions. Integration by partial fractions refers to an al-
gebraic technique for obtaining the partial fractions decomposition of a rational function, and
then integrating the resulting, easier summands in the decomposition. The partial fractions
decomposition is essentially what you get by undoing the work you have to do to add rational
functions and write the result as a single rational function, i.e., you have to undo nding a
(least) common denominator and simplifying the numerator after writing all of the fractions
over the common denominator.
Example 1.3.1. For instance, if you want to write
3
x + 7
+
5
x 2
as a rational function, that is, as the quotient of two polynomials, you would use the common
denominator (x + 7)(x 2) and obtain
3
x + 7
+
5
x 2
=
3(x 2)
(x + 7)(x 2)
+
5(x + 7)
(x + 7)(x 2)
=
3x 6 + 5x + 35
(x + 7)(x 2)
=
8x + 29
x
2
+ 5x 14
.
The corresponding partial fractions problem is to start with
8x + 29
x
2
+ 5x 14
40 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
and produce its partial fractions decomposition, i.e., to nd that this rational function equals
3
x + 7
+
5
x 2
.
How does this help with integration? It means that, if we want to calculate the integral
_
8x + 29
x
2
+ 5x 14
dx,
then, what we need to do is calculate
_ _
3
x + 7
+
5
x 2
_
dx = 3
_
1
x + 7
dx + 5
_
1
x 2
dx =
3 ln [x + 7[ + 5 ln [x 2[ + C,
where the nal two integrals were accomplished via the easy substitutions u = x + 7 and
w = x 2, respectively.
So, how do you nd partial fractions decompositions, and how do you integrate all of the
resulting summands? Before we state the general process/technique, we shall rst give a few
examples of integration by partial fractions, and work through them slowly, in order to discuss
most of the cases and issues that you will typically encounter.
Example 1.3.2. Lets look at the example above, in reverse. We know how to integrate the
summands that we know will appear. The question is: how do you produce the partial fractions
decomposition of
8x + 29
x
2
+ 5x 14
?
(We are, of course, pretending that we dont already know the answer.)
First, you note that the numerator has smaller degree than denominator. If this were not
the case, the rst step would be to (long) divide the numerator by the denominator. We shall
look at such a case in the next example.
1.3. INTEGRATION BY PARTIAL FRACTIONS 41
The next step is to show either that the denominator factors into a product of two degree
one (i.e., linear) real polynomials, or you complete the square to show that x
2
+ 5x 14 is an
irreducible quadratic polynomial, that is, a quadratic polynomial that does not factor into two
linear (real) polynomials. Hopefully, you see quickly (even if you have forgotten our work prior
to this example) that x
2
+ 5x 14 factors as
x
2
+ 5x 14 = (x + 7)(x 2).
However, if you did not notice this factorization, you would need to complete the square.
Since the leading coecient (the coecient in front of x
2
) is a 1, we complete the square by
squaring half of the coecient in front of the x term, and then adding and subtracting the
resulting quantity; in this example, this means we add and subtract (5/2)
2
= 25/4. What do
we mean by adding and subtracting 25/4? Exactly what we wrote; its true that this adds
zero in the end, but it adds zero in a very clever way. We obtain
x
2
+ 5x 14 = x
2
+ 5x +
25
4

25
4
14.
The point is that x
2
+ 5x +
25
4
is a perfect square; it equals
_
x +
5
2
_
2
. Thus, we nd that
x
2
+ 5x 14 =
_
x +
5
2
_
2

25
4
14 =
_
x +
5
2
_
2

81
4
. (1.3)
The fact that we end up with something squared minus a positive quantity means that this
result can be factored (over the real numbers) as the dierence of squares; hence, as we already
knew, x
2
+ 5x 14 factors as
x
2
+ 5x 14 =
_
x +
5
2
_
2

81
4
=
_
x +
5
2
_
2

_
9
2
_
2
=
_
x +
5
2
+
9
2
__
x +
5
2

9
2
_
= (x + 7)(x 2).
Had we completed the square and ended up with something squared plus a positive quantity
in Formula 1.3, then the quadratic polynomial would have no real roots and, hence, would be
42 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
an irreducible quadratic polynomial. We shall look at such an example in Example 1.3.4.
Discussing what to do in other cases is making this example much longer than it really is.
All that we have really done to help us with our current problem is to factor the denominator,
i.e., so far, whats relevant to this example is that
8x + 29
x
2
+ 5x 14
=
8x + 29
(x + 7)(x 2)
.
We would like to undo the obtaining of the common denominator and write an equality of
functions:
8x + 29
(x + 7)(x 2)
=
?
x + 7
+
?
x 2
, (1.4)
but what goes in place of the question marks?
It can be shown that there are unique polynomials that go where the question marks are
and, because the numerator on the lefthand side has smaller degree than the denominator, the
degrees of the numerators on the righthand side of Formula 1.4 must be less than degrees of
the denominators. However, the denominators have degree 1, and so the numerators must be
polynomials of degree 0 (or the zero polynomial), i.e., the numerators must be constants.
Therefore, we know that there exist unique constants A and B such that, for all x unequal
to 7 and 2, we have the equality
8x + 29
(x + 7)(x 2)
=
A
x + 7
+
B
x 2
, (1.5)
but how do we gure out what A and B are?
The rst step is to clear the denominators by multiplying both sides of the equation by
the big denominator on the left, and canceling factors. We obtain the equality
8x + 29 = A(x 2) + B(x + 7), (1.6)
where, initially, this equality is required to hold for all x, except x = 7 and x = 2. However,
the polynomial functions on each side of the equality are dened and continuous everywhere;
hence, if they are equal for all x, other than 7 and 2, then they must, in fact, be equal when
x = 7 and when x = 2.
It is important to emphasize that the equality in Formula 1.6 must now hold at those x values
1.3. INTEGRATION BY PARTIAL FRACTIONS 43
which we couldnt use earlier, those which made the original denominators in Formula 1.5 equal
zero. The reason that this is important is because the easiest way to determine A and B is to
plug x = 7 and x = 2 into Formula 1.6. We nd:
When x = 7:
8(7) + 29 = A(7 2) +B(0), and so 27 = 9A. Hence, A = 3.
When x = 2:
8(2) + 29 = A(0) +B(2 + 7), and so 45 = 9B. Hence, B = 5.
Thus, plugging A = 3 and B = 5 back into Formula 1.5, we nally obtain the partial fractions
decomposition
8x + 29
x
2
+ 5x 14
=
8x + 29
(x + 7)(x 2)
=
3
x + 7
+
5
x 2
.
If you now wish to integrate this, you proceed as we indicated before this example, by making
easy substitutions, and you obtain
_
8x + 29
x
2
+ 5x 14
dx = 3 ln [x + 7[ + 5 ln [x 2[ + C.
We refer to our method above for determining A and B by plugging in exceptional values
of x as the exceptional values method. There are at least two other methods for determining A
and B, which are very similar to each other.
One method is simply to substitute any two, distinct, x values into Formula 1.6; this will
yield two linear equations, involving A and B. You solve these equations simultaneously, and
you will still nd A = 3 and B = 5. (Try it.)
The other method is one that we refer to as matching coecients. You multiply out
Formula 1.6, then collect the powers of x, and match the coecients of the corresponding
powers of x on each side of the equation, since two polynomial functions (in the variable x) are
equal if and only if the coecients in front of each power of x agree.
44 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
Following this method, we obtain
8x + 29 = Ax 2A+Bx + 7B = (A+B)x + (2A+ 7B).
The constant terms on each side of the equation must agree, and so must the coecients in front
of x. Therefore, we nd that 29 = 2A+ 7B and 8 = A+B. Solving these equations, we once
again nd that A = 3 and B = 5.
Example 1.3.3. Consider the integral
_ _
x
3
+ 2x
2
21x + 71
x
2
+ 5x 14
_
dx.
True, its disgusting-looking, but its really not so bad. We want to produce the partial fractions
decomposition of the integrand, and integrate the resulting pieces.
This time, the degree of numerator is greater than or equal to the degree of the denominator.
The rst step is thus to long divide x
3
+ 2x
2
21x + 71 by x
2
+ 5x 14.
x 3
x
2
+ 5x 14
_
x
3
+ 2x
2
21x + 71
x
3
5x
2
+ 14x
3x
2
7x + 71
3x
2
+ 15x 42
8x + 29
What does this division tell us? It tells us that there is an equality
x
3
+ 2x
2
21x + 71
x
2
+ 5x 14
= x 3 +
8x + 29
x
2
+ 5x 14
.
The remaining fractional part is precisely what we integrated in the last example. Hence, we
conclude that
_ _
x
3
+ 2x
2
21x + 71
x
2
+ 5x 14
_
dx =
_
(x 3) dx +
_ _
8x + 29
x
2
+ 5x 14
_
dx =
1.3. INTEGRATION BY PARTIAL FRACTIONS 45
x
2
2
3x + 3 ln [x + 7[ + 5 ln [x 2[ + C.
Example 1.3.4. Consider the integral
_
x + 1
(x 5)(x
2
+ 6x + 11)
dx.
The numerator has smaller degree than the denominator, so theres no need to long divide. We
either need to factor x
2
+ 6x + 11, or complete the square. A few seconds of thought should
convince you that it doesnt factor into linear terms with integer coecients; still, it could factor
with real, but irrational coecients, or be irreducible over the real numbers. Completing the
square will tell us, either way. We add and subtract (6/2)
2
= 9, and nd
x
2
+ 6x + 11 = x
2
+ 6x + 9 9 + 11 = (x + 3)
2
+ 2.
This polynomial has no real roots, and so is irreducible (over the real numbers). What do we
do?
Now that we know that x
2
+6x+11 is irreducible, we temporarily put o using that it equals
(x + 3)
2
+ 2. What we want to do now is write an equality of functions
x + 1
(x 5)(x
2
+ 6x + 11)
=
?
x 5
+
?
x
2
+ 6x + 11
,
but what should the question marks be this time?
As before, it can be shown that there are unique polynomials that take the places of these
question marks and, because the rational function on the left has a numerator of smaller degree
than the denominator, the question marks are polynomials of smaller degree than the respective
denominators. Thus, there exist unique constants A, B, and C such that
x + 1
(x 5)(x
2
+ 6x + 11)
=
A
x 5
+
Bx +C
x
2
+ 6x + 11
. (1.7)
We wish to determine A, B, and C.
Again, we clear the denominators by multiplying both sides of the equation by the big
denominator on the left, and canceling. We obtain
46 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
x + 1 = A(x
2
+ 6x + 11) + (Bx +C)(x 5),
which now must hold for all x. Plugging in x = 5 immediately tells us that 6 = A 66 +B 0, so
that A = 1/11, but there are no easy (real) values to plug in for x to immediately yield B and
C. We will multiply things out and match coecients.
We nd
x + 1 = Ax
2
+ 6Ax + 11A+Bx
2
5Bx +Cx 5C =
(A+B)x
2
+ (6A5B +C)x + (11A5C).
Therefore, noting that 0 is the coecient of x
2
in the polynomial x + 1, we obtain the three
simultaneous equations
A+B = 0, 6A5B +C = 1, and 11A5C = 1.
As we already know that A = 1/11, we really only need two of these equations, say the rst
and the last, to nd that B = 1/11 and C = 0. Substituting these values into Formula 1.7,
we obtain
x + 1
(x 5)(x
2
+ 6x + 11)
=
1
11

_
1
x 5

x
x
2
+ 6x + 11
_
.
Before integrating, we wish to rewrite x/(x
2
+ 6x + 11) in a dierent form. Recall that we
completed the square to nd that x
2
+ 6x + 11 = (x + 3)
2
+ 2. Thus,
x
x
2
+ 6x + 11
=
x
(x + 3)
2
+ 2
=
(x + 3) 3
(x + 3)
2
+ 2
,
where we added and subtracted a 3 in the numerator in order to have the same quantities in
parentheses in the numerator and denominator. We shall see momentarily why this is useful.
Putting together all of our work above, we have
x + 1
(x 5)(x
2
+ 6x + 11)
=
1
11
_
1
x 5

x + 3
(x + 3)
2
+ 2
+ 3
1
(x + 3)
2
+ 2
_
.
This, nally, is the partial fractions decomposition that we need. We now discuss how to
integrate the main pieces of the righthand side above, and will then put these pieces together
to obtain the nal answer.
1.3. INTEGRATION BY PARTIAL FRACTIONS 47
Certainly, the integral of 1/(x 5) is easy. By making the substitution u = x 5, we nd
quickly that
_
1
x 5
dx = ln [x 5[ +C
1
.
By making the substitution w = (x + 3)
2
+ 2, we nd that dw = 2(x + 3) dx, and so
dw/2 = (x + 3) dx. Hence,
_
x + 3
(x + 3)
2
+ 2
dx =
_
1
2

dw
w
=
1
2
ln [w[ +C
2
=
1
2
ln
_
(x + 3)
2
+ 2
_
+C
2
,
where the absolute value signs disappeared since (x + 3)
2
+ 2 0.
If we let v = x + 3, then dv = dx, and we nd
_
1
(x + 3)
2
+ 2
dx =
_
1
v
2
+ (

2)
2
dv,
which, by the second formula in Theorem 1.1.18, is equal to
1

2
tan
1
_
v

2
_
+C
3
=
1

2
tan
1
_
x + 3

2
_
+C
3
.
Combining all of our work above, we, at long last, conclude that
_
x + 1
(x 5)(x
2
+ 6x + 11)
dx =
1
11
_
ln [x 5[
1
2
ln
_
(x + 3)
2
+ 2
_
+
3

2
tan
1
_
x + 3

2
__
+C.
Example 1.3.5. As our nal example of integration by partial fraction, we will calculate the
integral
_
1
(x + 2)(x 1)
3
(x
2
+ 4)
dx.
Note that the denominator of the integrand has a repeated linear factor, (x1)
3
. This factor
is, of course, a polynomial of degree 3, but it is important that it is a repeated linear term. As
48 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
you would expect, we want to nd the partial fractions decomposition:
1
(x + 2)(x 1)
3
(x
2
+ 4)
=
A
x + 2
+
?
(x 1)
3
+
Bx +C
x
2
+ 4
,
where x
2
+ 4 is clearly an irreducible quadratic polynomial, and the question mark should be
a general polynomial of degree 2, one degree less than the denominator. The slightly tricky
point is that, instead of writing the polynomial in the numerator of the (x 1)
3
term in powers
of x, it is best to write it in powers on (x 1). Thus, we replace the question mark with
D(x 1)
2
+E(x 1) +F. This yields a summand
D(x 1)
2
+E(x 1) +F
(x 1)
3
=
D
x 1
+
E
(x 1)
2
+
F
(x 1)
3
.
This leads to the general rule that, when looking for the partial fractions decomposition
of a rational function, when there are repeated linear factors in the denominator, you have
summands of the form a constant over each repeated power of the repeated linear factor, up to
the power that appears in the original denominator. In a similar fashion, if we have repeated
powers of an irreducible quadratic factor in the denominator, we have summands of the form a
general linear polynomial over each repeated power of the repeated irreducible quadratic factor,
up to the power that appears in the original denominator.
Hence, we want to nd constants A through F such that
1
(x + 2)(x 1)
3
(x
2
+ 4)
=
A
x + 2
+
D
x 1
+
E
(x 1)
2
+
F
(x 1)
3
+
Bx +C
x
2
+ 4
. (1.8)
However, even before nding the constants, we can easily write the integral of this rational
function, with the unknown constants still to be determined. You should be able to show quickly
that
_ _
A
x + 2
+
D
x 1
+
E
(x 1)
2
+
F
(x 1)
3
+
Bx
x
2
+ 4
+
C
x
2
+ 4
_
dx =
Aln [x + 2[ + Dln [x 1[
E
x 1

F
2(x 1)
2
+
B
2
ln(x
2
+ 4) +
C
2
tan
1
_
x
2
_
+K, (1.9)
where K is an arbitrary constant (which cannot be determined, unlike A through F).
We need to nd the constants A through F to nish the problem. You clear all the de-
nominators in Formula 1.8 by multiplying both sides by the big denominator on the left, and
1.3. INTEGRATION BY PARTIAL FRACTIONS 49
obtain
1 = A(x 1)
3
(x
2
+ 4) + D(x + 2)(x 1)
2
(x
2
+ 4) +
E(x + 2)(x 1)(x
2
+ 4) + F(x + 2)(x
2
+ 4) + (Bx +C)(x + 2)(x 1)
3
.
What a mess! Still A and F are easy to nd; plug in x = 2 and x = 1. We nd 1 =
A(2 1)
3
((2)
2
+ 4) so that A = 1/216, and 1 = F(1 + 2)(1
2
+ 4), so that F = 1/15. To
nd the remaining constants in the partial fractions decomposition, we leave it as an exercise (a
horrible exercise) for you to expand the terms, collect the powers of x, match the coecients, and
solve the resulting simultaneous equations. Alternatively, many calculators, computer programs,
and the web site wolframalpha.com can produce partial fractions decompositions. What you
should nd is that B = 9/1000, C = 13/500, D = 46/3375, and E = 11/225. Now, you
substitute the values of A through F into Formula 1.9, and youre nished.
As we saw in the examples above, there are always two major steps in using partial fractions
to integrate a rational function f(x):
Find the partial fractions decomposition of f(x).
Integrate each summand in the partial fractions decomposition.
Of course, we have yet to state exactly what the partial fractions decomposition for an
arbitrary rational function is. We shall do so now.
Suppose n(x) and q(x) are (real) polynomial functions, and that the leading coecient of
q(x) is a 1, i.e., q(x) is not the zero function, and the coecient of the largest power of x
appearing in q(x), with a non-zero coecient, is a 1. (If the leading coecient of q(x) were some
other (non-zero) constant c, immediately factor out the c, follow the procedure given below, and
then, in the end, multiply by 1/c.)
We want to dene the partial fractions decomposition of the rational function f(x) =
n(x)/q(x), whose domain is all x such that q(x) ,= 0.
1. If the degree of n(x) is greater than, or equal to, the degree of q(x), then you rst (long)
divide q(x) into n(x); if what you obtain from the division is a polynomial p(x) and a
remainder of r(x), then exactly what that means is that
f(x) =
n(x)
q(x)
= p(x) +
r(x)
q(x)
,
50 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
where r(x) = 0 or has degree less than the degree of q(x). As it is easy to integrate the
polynomial portion p(x), we will concentrate on the portion r(x)/q(x). In addition, if
r(x) = 0, then there is nothing left to do, so we will assume that r(x) ,= 0.
So, at this point, we want to dene and integrate the partial fractions decomposition of
r(x)/q(x), where r(x) and q(x) are polynomials, r(x) ,= 0, q(x) ,= 0, the degree of r(x) is
strictly less than the degree of q(x), and the leading coecient of q(x) equals 1.
2. Factor q(x) into, possibly repeated, linear terms, each of the form (x r)
m
, and, possibly
repeated, irreducible quadratic terms, which, after completing the square, can each be
written in the form [(x r)
2
+b
2
]
m
.
3. r(x)/q(x) is equal to a sum of contributions, where you have a contribution from each,
possibly repeated, irreducible factor of q(x). For a factor of the form (x r)
m
, the
contribution to the partial fractions decomposition is a sum of the form
A
1
x r
+
A
2
(x r)
2
+ +
A
m
(x r)
m
,
where all of the A
i
s are constants (to be determined).
For a factor of the form [(x r)
2
+b
2
]
m
, the contribution to the partial fractions decom-
position is a sum of the form
B
1
(x r) +C
1
(x r)
2
+b
2
+
B
2
(x r) +C
2
[(x r)
2
+b
2
]
2
+ +
B
m
(x r) +C
m
[(x r)
2
+b
2
]
m
,
where all of the B
i
s and C
i
s are constants (to be determined).
4. Now, set r(x)/q(x) equal to the sum of all of the contributions described above, and clear
the denominators by multiplying both sides of the equation by q(x), and canceling factors.
The result is that the polynomial r(x) equals a sum of polynomials with (possibly many)
unknown constants; these polynomial functions must be equal for all x.
5. Determine the unknown constants above by plugging in exceptional values (roots of q(x))
for x, or by multiplying out the entire righthand side, gathering the powers of x together,
and matching coecients with the coecients of r(x). A combination of these two methods
may be easiest.
6. After determining all of the constants, the sum of the contributions described above, with
the all of the values of the A
i
s, B
i
s, and C
i
s inserted is the partial fractions decomposition
of r(x)/q(x).
7. The individual summands in the partial fractions decomposition are integrated as follows.
We will omit the arbitrary constants produced by the indenite integral.
The integrals with powers of a linear factor are easy:
1.3. INTEGRATION BY PARTIAL FRACTIONS 51
_
A
x r
dx = Aln [x r[,
and if m 2,
_
A
(x r)
m
dx =
A(x r)
m+1
m+ 1
The integrals with powers of an irreducible quadratic term are harder:
_
B(x r) +C
[(x r)
2
+b
2
]
m
dx =
_
B(x r)
[(x r)
2
+b
2
]
m
dx +
_
C
[(x r)
2
+b
2
]
m
dx.
In the rst integral, the substitution u = (x r)
2
+b
2
yields du = 2(x r) dx, so that we
have
_
B(x r)
[(x r)
2
+b
2
]
m
dx =
B
2
_
du
u
m
,
which, if m = 1, is
B
2
ln [u[ =
B
2
ln((x r)
2
+b
2
)
and, if m 2, is
B
2

u
m+1
m+ 1
=
B
2

((x r)
2
+b
2
)
m+1
m+ 1
.
Finally, we need to integrate terms of the form
_
C
[(x r)
2
+b
2
]
m
dx.
After pulling the constant C outside the integral, and making the trivial substitution
u = x r, we are reduced to an integral of the form in Example 1.2.8; see that example
for the technique.
1.3.1 Exercises
Use partial fractions to calculate the anti-derivatives in Exercises 1-25.
1.
_
11x + 26
x
2
+ 5x + 6
dx
2.
_
t + 58
t
2
+ 6t 16
dt
52 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
3.
_
8u 35
u
2
7u
du
4.
_
42
y
2
49
dy
5.
_
11z 2
2z
2
z 1
dz
6.
_
2w
3
+ 6w
2
4w + 11
w
2
+ 3w 4
dw
7.
_
p
2
+p + 3
p + 3
dp
8.
_
2
3
+ 4
2
+ 6 2

2
1
d
9.
_
x
3
+ 12x
2
+ 51x + 74
x
2
+ 8x + 15
dx
10.
_
s
3
+ 7s
2
+s + 10
s + 7
ds
11.
_
7m
2
+ 12m+ 8
(m+ 3)(m
2
+m+ 1)
dm
12.
_
6n
2
4n + 20
(n 2)(n
2
+ 5)
dn
13.
_
3v
2
+ 4v 66
(v 6)(v
2
4v + 18)
dv
14.
_
3r
r
4
+ 5r
2
+ 4
dr
15.
_
12t
2
+ 52t + 150
(t
2
+ 5t + 30)(2t + 3)
dt
16.
_
3x
2
+ 8x + 5
(x
2
+ 2x + 5)(x + 3)
dx
17.
_
2y
2
60
(y
2
+ 4y + 20)(y + 12)
dy
18.
_
1 2w
(w 1)(w
2
+w + 1)
dw
19.
_
3
2
30
(
2
+ + 5)( + 2)
d
20.
_
4s
2
+ 4s + 6
(s
2
s + 3)(s + 3)
ds
1.3. INTEGRATION BY PARTIAL FRACTIONS 53
21.
_
5u
2
3u 29
(u + 3)(u 2)
2
du
22.
_
4t
2
10t 14
(t + 1)
2
(t 1)
dt
23.
_
3x
2
17x 26
(x + 3)
2
(x + 2)
dx
24.
_
2j
2
+ 18j + 43
(j + 5)
3
dj
25.
_
7r
3
2r
2
+ 9r 22
(r + 3)(r 1)
3
dr
26. Prove that
_
(A+C)x +AD +BC
x
2
+ (B +D)x +BD
dx = Aln [x +B[ +C ln [x +D[ +K where A, B, C,
D, and K are constants and neither A nor C is zero.
27. Suppose A, B, C and D are non-zero constants and B and D are both positive. What is
_
(A+C)x
2
+AD +BC
x
4
+ (B +D)x
2
+BD
dx?
Solve the separable dierential equations in Exercises 28 - 32. If an initial condition
is given, solve for the integration constant.
28.
dy
dx
=
7x
2
+ 44x + 80
(x + 5)
2
(x 2)
, y(3) = 5/8.
29.
dy
dx
=
2x
3
3x
2
32x 23
x
2
2x 15
, y(6) = 42.
30.
dy
dx
=
(6x
3
17x
2
27x + 16)(y
2
+ 2y + 10)(y + 3)
(x 3)
3
(x + 5)(5y
2
+ 16y 10)
.
31.
dy
dx
=
(x 2)(y
3
+ 9y)
(3y
2
+ 2y + 27)(x
2
+ 11x + 30)
.
32.
dy
dx
=
(3x
3
23x
2
+ 53x 32)(y
2
3y + 10)(y + 4)
(x
2
6x + 8)(3y
2
+ 8y + 22)
.
33. Recall that the logistic model for population growth is given by
dP
dt
= kP(M P)
where M and k are non-zero constants and P = P(t) is the population at time t. Suppose
P
0
= P(0) is the initial population. Use the fact that this dierential equation is separable
54 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
and integrate using partial fractions to arrive at the solution to the dierential equation:
P =
MP
0
(M P
0
)e
Mkt
+P
0
.
Integrals involving square-roots can often be solved by a technique involving partial
fractions and a rationalizing substitution. The idea is to make a substitution that
allows the integrand to be written as a rational function, without a square-root
symbol, and then to integrate using partial fractions. The method is outlined in
Exercises 34. Use the method to calculate the anti-derivatives in Exercises 35 -40.
34. Consider
_
dx
x

x + 1
.
a. Let u =

x + 1. Show that the integral can than be written as

_
2 du
u
2
1
.
b. Use partial fractions to conclude that the integral in part (a) is ln [u1[ln [u+1[+C.
c. Rewrite the anti-derivative in terms of x, thus establishing:
_
dx
x

x + 1
= ln [

1 +x
1[ ln [

1 +x + 1[ +C.
35.
_
x
x 9
dx.
36.
_
dx
x

x + 30
.
37.
_
dx
3

x 1
.
38.
_
dx

x +
3

x
.
39.
_
x
(x + 2)(

x + 3)
dx.
40.
_
dx
a

x +b
where a and b are positive constants.
A common application of partial fraction integration is the reaction rate of two chemicals that
combine to form a third chemical. This is called a bimolecular reaction and is justied by the
law of mass action.
Specically, assume chemicals A and B have initial concentrations of a and b in mols per unit
volume and react to form chemical C. Then under certain conditions, the rate of increase in the
1.3. INTEGRATION BY PARTIAL FRACTIONS 55
concentration, c, of chemical C is given by:
dc
dt
= k(a c)(b c)
where k is some constant. Assume throughout the next four problems that c < a < b.
41. Find an explicit solution for the concentration of chemical C, c(t), if a = 3, b = 4, k = 2
and c(0) = 0.
42. Find an explicit solution for the concentration of chemical C, c(t), if a = 8, b = 10, k = 1
and c(0) = 0.
43. What is the concentration of chemical C after 15 seconds if a = 2, b = 5, k = 4.5 and
c(0) = 0? Assume that t is measured in minutes.
44. Assume that a ,= b and prove that a general solution to the dierential equation is
c(t) =
ab(1 e
(ab)kt
)
b ae
(ab)kt
.
Assume that c(0) = 0.
The 18th century mathematician Leonhard Euler developed a quick way of deter-
mining the coecients of partial fraction decompositions of rational functions. Let
P(x)/Q(x) is a rational function and assume a+bx is a factor of Q(x), but that (a+bx)
2
is not a factor of Q(x). Then,
P(x)
Q(x)
=
k
a +bx
+R(x), and wed like to determine k.
Multiplying by a +bx and using lHopitals Rule, Euler proved that
k = lim
xa/b
bP(x)
Q

(x)
.
Use this method to answer Exercises 45-49.
45. We considered the partial fraction decomposition of
x + 1
(x 5)(x
2
+ 6x + 11)
in Example 1.3.4.
The coecient A of the term
A
x 5
was found to be 1/11. Lets use Eulers method to
nd this term.
a. Show that a = 5, b = 1 and that Q

(x) = 3x
2
+ 2x 19.
56 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
b. Show that based on Eulers method, A = lim
x5
x + 1
3x
2
+ 2x 10
.
c. Show that this limit is indeed 1/11.
46. What is the coecient of x + 1 in the partial fraction decomposition of
x
13
x
23
+ 1
?
47. Determine A, B and C in the equation
x
2
+ 3x + 5
(x + 2)(x + 3)(x + 5)
=
A
x + 2
+
B
x + 3
+
C
x + 5
using Eulers method. Note that (x + 2)(x + 3)(x + 5) = x
3
+ 10x
2
+ 31x + 30.
48. Use Eulers method to nd the partial fraction decomposition of
3x
3
+ 4x
2
+ 4x + 5
x
4
1
. Hint:
use Euler to nd the A/(x1) term. Subtract this term from the original function to nd
the remaining term.
49. Let m and n be positive integers. Let f(x) =
x
m
+ (2n 1)x
1 x
2n
. What is the coecient A
of the term A/(x 1) in the partial fraction decomposition?
50. Consider
_
1
x
4
+ 1
dx. According to Courant, Even Leibnitz found this a troublesome
integration.
a. Complete the square to show that x
4
+ 1 = (x
2
+ 1)
2
2x
2
.
b. Notice that the expression in (a) is the dierence of squares. Use this fact to prove
x
4
+ 1 = (x
2
+ 1 +

2x)(x
2
+ 1

2x).
c. Use the methods in this chapter to prove that
1
x
4
+ 1
= K
_
x +

2
x
2
+

2x + 1

2
x
2

2x + 1
_
,
where K =

2/4.
d. Finally, conclude that
_
dx
x
4
+ 1
=
K
_
1
2
ln [x
2
+

2x+1[
1
2
ln [x
2

2x+1[ +tan
1
(

2x+1) +tan
1
(

2x1)
_
+C.
1.4. INTEGRATION USING HYPERBOLIC SINE AND COSINE 57
1.4 Integration using
Hyperbolic Sine and Cosine
Corresponding to sine and cosine, we will dene the hyperbolic sine and hyperbolic cosine func-
tions, f(x) = sinh x and g(x) = cosh x, respectively; when speaking, hyperbolic cosine is usually
referred to by pronouncing cosh phonetically, while hyperbolic sine is usually referred to by
pronouncing sinh as though it were the word cinch.
Once we have hyperbolic sine and cosine, we could, but will not, dene the remaining four
hyperbolic trig. functions, in analogy with how the other trig. functions are dened in terms of
sine and cosine, e.g., hyperbolic tangent is hyperbolic sine divided by hyperbolic cosine.
Instead, we will concentrate on the properties of sinh and cosh which make them so useful
for anti-dierentiating certain types of functions.
The denitions of hyperbolic sine and cosine do not look analogous in any way to the deni-
tions of the usual sine and cosine functions. Later, in Example 4.6.21, we will see that Eulers
Formula provides a close relationship between the hyperbolic and standard trig. functions, but
you have to be willing to deal with complex numbers.
Denition 1.4.1. The hyperbolic sine function, sinh : R R, is dened by
sinh x =
e
x
e
x
2
,
and the hyperbolic cosine function, cosh : R R, is dened by
cosh x =
e
x
+e
x
2
.
The following algebraic/graphical/Calculus properties of sinh and cosh are easy to verify,
and we leave them as exercises.
58 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
-3 -2 -1 0 1 2 3
-3
-2
-1
1
2
3
Figure 1.1: The graph of y = sinh x.
-3 -2 -1 0 1 2 3
-3
-2
-1
1
2
3
Figure 1.2: The graph of y = cosh x.
Proposition 1.4.2. Hyperbolic sine and cosine have the following properties:
1. sinh x is an odd function;
2. cosh x is an even function;
3. sinh x is one-to-one and its range is the entire real line;
4. cosh x is not one-to-one, but is one-to-one when restricted to the domain [0, ), and
its range is the interval [1, );
5. sinh x is strictly increasing, is negative when x is negative, and positive when x is
positive;
6. cosh x is strictly decreasing on the interval (, 0], strictly increasing on the interval
[0, ), and obtains its global minimum value of 1 when x = 0;
7. sinh

x = cosh x;
8. cosh

x = sinh x;
9. 1 + sinh
2
x = cosh
2
x.
The last property above is where the term hyperbolic comes from; a point (x, y) = (cosh t, sinh t)
lies on the hyperbola given by x
2
y
2
= 1. You should think of this as being analogous to the
fact that a point (x, y) = (cos t, sin t) lies on the circle given by x
2
+y
2
= 1; hence, the usual
trigonometric functions are sometimes referred to as the circular trigonometric functions.
As hyperbolic sine is one-to-one, onto, and possesses an everywhere non-zero derivative,
the inverse function sinh
1
exists and is dierentiable everywhere. However, as cosh is not
one-to-one, we restrict its domain to [0, ) before producing an inverse function. We can nd
explicit formulas for these inverse functions by writing x = (e
y
e
y
)/2 and x = (e
y
+e
y
)/2,
1.4. INTEGRATION USING HYPERBOLIC SINE AND COSINE 59
respectively, and solving for y in each case, while paying attention to the restrictions on the
domain and codomain of cosh
1
. The algebra involved is simply the quadratic formula, and
applying the natural logarithm, but the variable to which you apply the quadratic formula is
e
y
; we leave this algebra as an exercise also.
You should nd the following formulas for inverse hyperbolic sine and inverse hyperbolic
cosine.
Proposition 1.4.3. The function sinh
1
: R R is given by
sinh
1
x = ln
_
x +
_
x
2
+ 1
_
,
and the function cosh
1
: [1, ) [0, ) is given by
cosh
1
x = ln
_
x +
_
x
2
1
_
.
By using either the algebraic formulas above, or the general formula for the derivative of an
inverse function, we nd (via another exercise for you):
Proposition 1.4.4. The function sinh
1
: R R is dierentiable (everywhere), and
(sinh
1
x)

=
1

x
2
+ 1
.
The function cosh
1
: [1, ) [0, ) is continuous; for all x > 1, cosh
1
is dierentiable,
and
(cosh
1
x)

=
1

x
2
1
.
We could immediately rewrite the formulas above as anti-derivative formulas, but we want
to emphasize that the real value of hyperbolic sine and cosine when anti-dierentiating stems
from three properties: that 1 + sinh
2
x = cosh
2
x, sinh

x = cosh x, and cosh

x = sinh x.
60 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
Example 1.4.5. Suppose that you dont remember the derivatives of inverse hyperbolic sine
and cosine that we gave above, but you need to nd the anti-derivative
_
1

x
2
+a
2
dx,
where a > 0 is a constant.
When you see

x
2
+a
2
, you should realize that there are (at least) two reasonable choices
for methods of attack; these rely on your knowing that 1+tan
2
= sec
2
and that 1+sinh
2
t =
cosh
2
t.
We will rst use tan and see that the calculation takes a while. We will then look at how
much easier the integral is using sinh t.
Using circular trig. functions:
Using tangent, you should think I want x to be a tan , so that x
2
+a
2
= a
2
tan
2
+a
2
=
a
2
(tan
2
+ 1) = a
2
sec
2
, and then

x
2
+a
2
= a sec . Thats how you think about the
problem, but, being more careful, what you really do is not let x = a tan ; after all, x is
already dened in the problem. You dene the new variable by letting = tan
1
(x/a), so
that, yes, x = a tan , but now we also know that /2 < < /2. This restriction on is
important, because, what we really get is that
_
x
2
+a
2
=

a
2
sec
2
= [a sec [.
To know that this last quantity is equal to a sec , you have to use that a > 0 and that, since
/2 < < /2, sec > 0.
Therefore, since x = a tan , dx = a sec
2
d and
_
1

x
2
+a
2
dx =
_
1
a sec
a sec
2
d =
_
sec d =
ln [ sec + tan [ +C = ln(sec + tan ) +C,
where, to eliminate the absolute value signs, we used that our restrictions on imply that
sec + tan 0. To nish the problem, you now use that tan = x/a and that sec =

x
2
+a
2
/a.
1.4. INTEGRATION USING HYPERBOLIC SINE AND COSINE 61
Our nal result:
_
1

x
2
+a
2
dx = ln
_
x
a
+

x
2
+a
2
a
_
+C = ln
_
x +
_
x
2
+a
2
_
+C,
where, in the last step, we used that ln
__
x +

x
2
+a
2
_
/a

= ln
_
x+

x
2
+a
2
_
ln a, and then
we absorbed the ln a into the constant C.
Using hyperbolic trig. functions:
Using hyperbolic sine, you should think I want x to be a sinh t, so that x
2
+a
2
= a
2
sinh
2
t+
a
2
= a
2
(sinh
2
t + 1) = a
2
cosh
2
t, and then

x
2
+a
2
= a cosh t. Again, thats how you think
about the problem, but what you really do is not let x = a sinh t; you dene the new variable
t by letting t = sinh
1
(x/a). However, we dont need any special restrictions on t to know that

a
2
cosh
2
t = a cosh t, because cosh t 1 0.
As x = a sinh t, dx = a cosh t dt, and we nd
_
1

x
2
+a
2
dx =
_
1
a cosh t
a cosh t dt =
_
1 dt = t +C = sinh
1
_
x
a
_
+C,
which is the nicest form in which to leave the answer. On the other hand, if you wish to see
that this answer agrees with the answer that we obtained above, using tangent, then you use
the formula for sinh
1
in Proposition 1.4.3.
Example 1.4.6. Suppose that a > 0, and that we want the general anti-derivative of the
function f(x) =
1

x
2
a
2
with domain x > a, i.e., we want to nd
_
1

x
2
a
2
dx,
where x > a > 0.
We want x = a cosh t, so that
_
x
2
a
2
=
_
a
2
cosh
2
t a
2
= a
_
sinh
2
t = a[ sinh t[.
62 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
Thus, we let t = cosh
1
(x/a), which implies both that x = a cosh t and, as x/a > 1, that t > 0.
Since t > 0, [ sinh t[ = sinh t.
Therefore, dx = a sinh t dt and
_
1

x
2
a
2
dx =
_
1
a sinh t
a sinh t dt =
_
1 dt = t +C = cosh
1
_
x
a
_
+C.
The results from the last two examples are worth recording in a proposition.
Proposition 1.4.7. Suppose that a > 0. Then,
_
1

x
2
+a
2
dx = sinh
1
_
x
a
_
+C,
and, for x > a,
_
1

x
2
a
2
dx = cosh
1
_
x
a
_
+C.
1.4.1 Exercises
Calculate the anti-derivatives of the functions.
1.
_
7

x
2
+ 9
dx.
2.
_
12

z
2
625
dz.
3.
_
dy
_
4y
2
+ 49
.
4.
_
dt

7t
2
81
.
5.
_
cosh(3x) dx.
1.4. INTEGRATION USING HYPERBOLIC SINE AND COSINE 63
6.
_
s sinh s ds.
7.
_
2x

x
4
+ 9
dx.
8.
_
cosh
2
z sinh
2
z dz. Hint: why is this trivial?
9.
_
dx

4x
2
+ 4x + 9
.
10.
_
dy
_
y
2
10y + 9
.
11.
_
1 +

t
3
+ 3t

t
2
+ 3
dt, t > 0.
12.
_

u
2
4 +

u
2
4

u
4
16
du.
13.
_

x
2
4 +

x
2
+ 9

x
4
+ 5x
2
36
dx.
14.
_
u
2
(u
2
+ 10)
3/2
du.
15.
a. Calculate
_
sinh xcosh xdx using integration by parts.
b. Calculate
_
sinh xcosh xdx directly by writing sinh x and cosh x in terms of their
denitions using the exponential function.
16. Calculate
_
_
x
2
+ 1 dx.
In each of Exercises 17 through 25, prove the given statements from Proposi-
tion 1.4.2.
17. sinh x is an odd function.
18. cosh x is an even function.
19. sinh x is one-to-one and its range is the entire real line.
20. cosh x is not one-to-one, but is one-to-one when restricted to the domain [0, ), and its
range is the interval [1, ).
64 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
21. sinh x is strictly increasing, is negative when x is negative, and positive when x is positive.
22. cosh x is strictly decreasing on the interval (, 0], strictly increasing on the interval
[0, ), and obtains its global minimum value of 1 when x = 0.
23. sinh

x = cosh x.
24. cosh

x = sinh x.
25. 1 + sinh
2
x = cosh
2
x.
26. What is
_
_
x
2
1 dx?
27. What is
_
x
2
_
x
2
1 dx?
28. What is
_
x
2
_
x
2
+ 1 dx?
29. More generally, what is
_
x
2
_
x
2
+a
2
dx?
Use the previous four problems to calculate the integrals in the next four problems.
30.
_
_
9x
2
16 dx.
31.
_
25x
2
_
25x
2
1 dx.
32.
_
_
x
6
+ 100x
4
dx.
33.
_
cx
2
_
b
2
x
2
+a
2
dx, b > 0, a > 0.
In exercises 34 though 37, you are given the acceleration function of a particle and
the velocity at one specic time. Find the velocity function.
34. a(t) =
1

t
2
+ 16t + 25
, v(0) = 10. Assume that t > 8 +

39.
35. a(t) =
1

t
2
6t + 16
, v(6) = 8.
36. a(t) =
12

6t
2
+ 24
, v(3) = 6.
37. a(t) = 2t cosh(t
2
+ 2), v(0) = 4.
1.4. INTEGRATION USING HYPERBOLIC SINE AND COSINE 65
38. What is
_
sinh
1
xdx? Hint: use a technique similar to the one used in calculating
_
ln xdx.
39. What is
_
cosh
1
xdx?
In the next four exercises you are given the acceleration function of a particle. Find
a formula for the position vector. Make sure to retain any integration constants.
Use the previous two problems.
40. a(t) =
1

t
2
+ 9
.
41. a(t) =
1

t
2
16
.
42. a(t) =
5

4t
2
+ 25
, v(0) = 0.
43. a(t) =
6

9t
2
11
, v(2) = 8.
Solve the following separable dierential equations.
44.
dx
dt
=

12x
2
8
4 + 6t
2
.
45.
dx
dt
=
(x
2
+ 1)
3/2
(

t
2
3)
.
46.
dx
dt
=
e
t(x+1)
e
t(x1)
e
x(t+1)
+e
x(t1)
.
47.
dx
dt
=
_
x
4
+x
2
+t
2
+t
2
x
2
t
4
t
2
x
2
+t
2
x
2
. Hint: look for a factor common to the numerator and
denominator.
Calculate the integrals of the following products of hyperbolic and trigonometric
functions. You may nd it easier write the hyperbolic functions in terms of expo-
nential functions.
48.
_
sinh xcos xdx.
49.
_
cosh xcos xdx.
50.
_
cosh xsinh xsin xcos xdx.
66 CHAPTER 1. ANTI-DIFFERENTIATION: THE INDEFINITE INTEGRAL
Chapter 2
Continuous Sums: the Denite
Integral
As we wrote at the beginning of the previous chapter, the denite integral is the mathematically
precise notion of what it means to take a continuous sum of innitesimal contributions. It
is the denition of integration as a continuous sum that yields all of its applications, many of
which we will look at in Chapter 3. However, the basic method of calculating denite integrals
is to use the Fundamental Theorem of Calculus, Theorem 2.4.10, to conclude that the problem
of calculation of an integral (typically) boils down to producing an anti-derivative. Thus, as we
wrote in the previous chapter, anti-dierentiation is also referred to as calculating an indenite
integral.
We begin this chapter with a discussion of basic properties, notations, and techniques involv-
ing summations; not surprisingly, dierences are involved in crucial ways. We then move on to
the problem of how to approximate a continuous sum; of course, part of the problem is that
we dont (yet) have a mathematically rigorous denition of what a continuous sum is. After
we have our approximations, we then use limits of our approximations to dene the denite
integral.
67
68 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
2.1 Sums and Dierences
In this section, we need to recall some notation and properties related to summations. We shall
also dene the dierence operator, and use it to derive some useful summation formulas.
Recall the sigma notation for summations.
Denition 2.1.1. Suppose that we have two integers m and n, where m n, we let [m..n]
denote the set of integers between, or equal to, m and n. We call such a set an integer
interval.
Suppose we have a function f, whose domain includes [m..n]. Then, we write
n

k=m
f(k) for
the summation, as k goes from m to n, of f(k). This means that
n

k=m
f(k) = f(m) +f(m+ 1) + +f(n 1) +f(n).
In this context k is frequently referred to as the index of summation, and f(k) is frequently
written as f
k
. The integer interval [m..n] is called the range of the index of summation.
For example,
3

k=1
k
2
= (1)
2
+ 0
2
+ 1
2
+ 2
2
+ 3
2
= 15,
or, as another example,
p(x) = a
0
+a
1
x + +a
n
x
n
=
n

k=0
a
k
x
k
,
where x
0
in the summation is to be interpreted as equaling 1, even if x = 0.
Note that the indexing variable k is a dummy variable; if we replaced it with an i or j, or
2.1. SUMS AND DIFFERENCES 69
any other variable (which is not already present), the summation would not change, e.g.,
3

j=1
j
2
= (1)
2
+ 0
2
+ 1
2
+ 2
2
+ 3
2
= 15,
Summations can be split. For instance, in the example above,
3

j=1
j
2
= (1)
2
+ 0
2
+ 1
2
+ 2
2
+ 3
2
=
_
(1)
2
+ 0
2

+
_
1
2
+ 2
2
+ 3
2

=
0

j=1
j
2
+
3

j=1
j
2
.
In general, we have
Proposition 2.1.2. (splitting summations) If m, n, and p are integers, with m n p,
and f is a function whose domain includes the integer interval [m..p], then
p

k=m
f
k
=
n

k=m
f
k
+
p

k=n+1
f
k
,
where, if n = p, the last summation should be interpreted as being 0.
Looking once again at
3

j=1
j
2
= (1)
2
+ 0
2
+ 1
2
+ 2
2
+ 3
2
,
we note that we can replace each term j
2
by
_
(j + 6) 6
_
2
(which may seem like a silly thing
to do, but it leads us to an important property); this means that we have the easy equality
(1)
2
+ 0
2
+ 1
2
+ 2
2
+ 3
2
= (6 7)
2
+ (7 7)
2
+ (8 7)
2
+ (9 7)
2
+ (10 7)
2
,
70 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
which equals
10

k=6
(k 7)
2
.
In general,
Proposition 2.1.3. (shifting indices) Suppose we have a function f, whose domain in-
cludes the integer interval [m..n]. Let p be an integer.
Then, by making the substitution k = j +p, so that j = k p, we obtain
n

j=m
f
j
=
n+p

k=m+p
f
kp
.
Thus, you get the same sum if you add p to each of the bounds of the summation, and simulta-
neously replace each occurrence of the index k by k minus p.
Consider now
9

k=7
_
1.7k
2

3 sin k
_
=
_
1.7(7)
2

3 sin(7)
_
+
_
1.7(8)
2

3 sin(8)
_
+
_
1.7(9)
2

3 sin(9)
_
=
1.7
_
7
2
+ 8
2
+ 9
2
_
+ (

3)
_
sin 7 + sin 8 + sin 9
_
=
1.7
9

k=7
k
2
+ (

3)
9

k=7
sin k.
More generally, the algebraic properties of real numbers (i.e., associativity, commutativity,
and distributivity) immediately imply:
Proposition 2.1.4. (linearity of summation) If a and b are constants, and f and g are
functions whose domain includes the integer interval [m..n], then
n

k=m
_
af
k
+bg
k
_
= a
n

k=m
f
k
+ b
n

k=m
g
k
.
2.1. SUMS AND DIFFERENCES 71
Example 2.1.5. The properties of summations described above may seem simple, but they
allow us to derive formulas involving sums that really dont look so obvious. Consider, for
instance, the problem of simplifying
2
50

k=3
k(k 1)
50

k=1
(k + 1)(k + 2). (2.1)
We would like to combine the two summations, using linearity; however, the ranges of the
indices of summation would need to be the same, and they are not. We will x this problem
in two dierent ways. Understand that the point of this example is not that you will necessarily
agree that what we end up with is simpler than what we started with; the point is for you to
understand the types of manipulations that we use.
First approach: We split the second summation, and have
2
50

k=3
k(k 1)
50

k=1
(k + 1)(k + 2) =
2
50

k=3
k(k 1)
50

k=3
(k + 1)(k + 2)
2

k=1
(k + 1)(k + 2).
Now we can use linearity on the rst two of the three summations above to obtain
50

k=3
_
2k(k 1) (k + 1)(k + 2)

k=1
(k + 1)(k + 2) =
_
50

k=3
_
2k
2
2k k
2
3k 2
_
_
6 12 =
18 +
50

k=3
_
k
2
5k 2
_
Second approach: We rst shift the index in the second summation in Formula 2.1; we add 2
to each of the bounds and replace k by (k 2) in the summation. We obtain
2
50

k=3
k(k 1)
50

k=1
(k + 1)(k + 2) =
72 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
2
50

k=3
k(k 1)
52

k=3
_
(k 2) + 1
__
(k 2) + 2
_
=
2
50

k=3
k(k 1)
52

k=3
(k 1)k.
The range of the index of summation on the right above still does not match that of the
summation on the left, but we can split o part of the sum
2
50

k=3
k(k 1)
52

k=3
(k 1)k.
2
50

k=3
k(k 1)
_
50

k=3
k(k 1)
_
(50)(51) (51)(52) =
(50)(51) (51)(52) +
50

k=3
_
2k(k 1) k(k 1)

=
(50)(51) (51)(52) +
50

k=3
_
k(k 1)

,
which does not look very similar to our answer from the rst approach, but is nonetheless equally
correct.
We now want to dene notation related to dierences.
Denition 2.1.6. Suppose that m and n are integers, and m < n, and suppose that we have
a real function f, whose domain is [m..n]. Then, we dene the nite dierence function
f to be the function with domain [(m+ 1)..n] given by
(f)(k) = f(k) f(k 1).
2.1. SUMS AND DIFFERENCES 73
Remark 2.1.7. We have been very formal with our notation above. Typically, we write things
like k
2
= k
2
(k 1)
2
= k
2
(k
2
2k + 1) = 2k 1, in place of writing that, if f(k) = k
2
,
then (f)(k) = 2k 1.
We should also remark that, in the notation [m..n], we mean to allow the cases where
m = and/or n = . To be precise, we mean that, if m and n are integers, then
[..n] = k [ k is an integer, and k n,
[m..] = k [ k is an integer, and m k,
and [..] is the entire set of integers.
Finally, in Denition 2.1.6, if m = , then the m+ 1 which appears in the domain of f
should also be taken to equal .
Like summations, the nite dierence operator is linear.
Proposition 2.1.8. (linearity of dierences) If a and b are constants, and f and g are
functions whose domain is [m..n], then

_
af(k) +bg(k)
_
= a f(k) + b g(k).
Despite the fact that it is trivial to prove, the following result turns out to be very useful.
Proposition 2.1.9. (telescoping sums) Suppose that m and n are integers, and m < n.
If f is a real function, whose domain is [m..n], then
n

k=m+1
f(k) = f(n) f(m).
74 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
Proof. A rigorous proof would use mathematical induction. However, it is easy to see why this
is true.
n

k=m+1
f(k) =
_
f(m+ 1) f(m)
_
+
_
f(m+ 2) f(m+ 1)
_
+
_
f(m+ 3) f(m+ 2)
_
+ . . .
. . . +
_
f(n 2) f(n 3)
_
+
_
f(n 1) f(n 2)
_
+
_
f(n) f(n 1)
_
.
Note that every term, except f(n) and f(m), occurs once with a plus sign and once with a
minus sign. Thus, all those intermediate terms collapse to 0. The result follows.
Proposition 2.1.10. Suppose that b is a constant. We have the following formulas for
nite dierences:
1. k = 1;
2. k
2
= 2k 1;
3. k
3
= 3k
2
3k + 1; and
4. b
k+1
= b
k
(b 1).
Proof. These are all easy computations. For instance,
k
3
= k
3
(k 1)
3
= k
3
(k
3
3k
2
+ 3k 1) = 3k
2
3k + 1.
We leave the proofs of the remaining items as exercises.
Corollary 2.1.11. Suppose that b is a constant. We have the following formulas for nite
dierences:
1.
k =
_
k(k + 1)
2
_
;
2.
k
2
=
_
k(k + 1)(2k + 1)
6
_
; and
3. if b ,= 1,
b
k
=
_
b
k+1
b 1
_
.
2.1. SUMS AND DIFFERENCES 75
Consequently,
a. if n 1,
n

k=1
k =
n(n + 1)
2
;
b. if n 1,
n

k=1
k
2
=
n(n + 1)(2n + 1)
6
; and
c. if n 0 and b ,= 1, then
n

k=0
b
k
=
b
n+1
1
b 1
,
provided that b
0
is interpreted as equaling 1 when b = 0.
Proof. Items (a), (b), and (c) follow immediately from Items (1), (2), and (3), respectively, by
applying Proposition 2.1.9. We shall prove Items (1), (2), and (3).
From Proposition 2.1.10, we have
k
2
= 2k k,
and so, by the linearity of nite dierences,
k =
1
2
(k
2
+k) =
_
k(k + 1)
2
_
.
From Proposition 2.1.10 and Item (1), we have
k
3
= 3k
2
3
_
k(k + 1)
2
_
+ k.
Therefore, linearity gives us
k
2
=
_
1
3

_
k
3
+
3k(k + 1)
2
k
_
_
=
_
2k
3
+ 3k(k + 1) 2k
6
_
=
76 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL

_
k(k + 1)(2k + 1)
6
_
.
Finally, Proposition 2.1.10 and linearity immediately yield Item 3.
We should remark that the summation

n
k=0
b
k
that appears in Item (c) of Corollary 2.1.11
is one which will be very important to us in Chapter 4 and Chapter 5; it is called a geometric
sum.
2.1.1 Exercises
Calculate the sums.
1.
7

k=3
(3k + 2).
2.
5

j=2
_
j
2
j
_
.
3.
5

t=1
ln t.
4.
7

m=4
sin(m).
5.
5

k=1
(k 1)(k 2)(k 3).
6.
5

t=5
cosh t.
7.
5

s=5
sinh s.
8.
6

k=0
(sin(k) + cos(k)).
9.
5

j=0
2
j
.
2.1. SUMS AND DIFFERENCES 77
10.
3

m=2
1
m
.
Evaluate the sums by reindexing.
11.
12

j=6
(3j 18).
12.
8

k=4
(k 3)
2

j=1
(9j + 20).
13.
_
9

k=6
_
k
2
10k + 25
_
_

_
_
0

j=3
_
j
2
+ 8j + 16
_
_
_
.
Prove the following statements of Proposition 2.1.10
14. k = 1.
15. k
2
= 2k 1.
16. b
k+1
= b
k
(b 1).
Calculate f(k) for the following functions.
17. f(k) = 5k
18. f(k) = C, where C is a constant.
19. f(k) = 3k
2
4k + 7.
20. f(k) = ln k.
21. f(k) = sin(2k).
22. f(k) = 3
k
.
23. Prove that if f(x) = sin x, then (f)(x) = (sin x)(1 cos 1) + (sin 1) cos x.
24. Prove that if g(x) = cos x, then (g)(x) = (cos x)(1 cos 1) + (sin 1) sin x.
25. a. What is
U

n=0
(1)
n
when U is even?
b. What is
U

n=0
(1)
n
when U is odd?
78 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
26. Is the following statement true? Assume f and g are dierentiable functions.
_
_
3

j=1
f(j)
_
_
_
_
3

j=1
g(j)
_
_
=
3

j=1
f(j)g(j)
Prove or give a counterexample.
27. What is
100

k=1
1
k(k + 1)
? Hint: use partial fractions and nd a telescoping sum.
28. What is
J

k=J
f(k) if f is an odd function, f(0) = y
0
, and J is some positive integer?
29. What is
25

j=17
_
_
j
_
j 1
_
?
30. Suppose that f is a dierentiable function. Prove that there exist numbers c
i
in each open
interval (i 1, i), for i = 1, 2, 3, such that f(3) f(0) = f

(c
1
) +f

(c
2
) +f

(c
3
).
31. Recall that the average of n numbers is given by the formula:
x =

n
i=1
x
n
.
Prove that
n

i=1
(x x) = 0.
The sample standard deviation of a data set, s, measures how spread out a data
set is. For example, if A = 84, 84, 84 and B = 80, 84, 88, then the sample standard
deviation of B will be larger than the sample standard deviation of A since there the
points of B are more dispersed than those in set A. The sample standard deviation
of a set with n data points is
s =

_
1
n 1
n

i=1
(x x)
2
.
The units of s are the same as those of the data.
2.1. SUMS AND DIFFERENCES 79
32. A Calculus test is given to two sections of students. The grades of the students in section A
are 85, 65, 73, 40, 64, 90 and the grades of the students in section B are 84, 90, 96, 88, 100.
What are s
A
and s
B
?
33. Suppose that the week-to-week changes in a stocks price, in dollars per share, over a six
week period are: +2.5, +4, 7, 0, +3. What is s for this set? The standard deviation of
a stocks price measures the volatility of the stock.
34. The recorded annual rainfall, in inches, for a city over a ve year span is 23, 47, 35, 42, 29.
What is the sample standard deviation?
35. We dene the sample variance to be the square of the sample standard deviation:
s
2
=

n
i=1
(x
i
x)
2
n 1
.
Prove the following useful alternative formula:
s
2
=
_
x
2
i
_
(1/n) (

x
i
)
2
(n 1)
.
All sums are taken from 1 to n where n is the number of items in the data set.
Oftentimes statistics are used to measure the relationship between two variables.
For example, researchers could be interested in the relationship between drug
dosage and cancer cell counts. Basketball executives may be interested in the
relationship between a teams free throw percentage, and the teams overall win-
ning percentage. The linear correlation coecient, r, quanties the relationship.
Specically, r helps answer the question: to what extant are the two variables lin-
early related? r is always between 1 and +1. If r is close to +1 (resp. 1), then
a strong positive (resp. negative) linear relationship exists between the variables
in the sense that as one variable increases, the second variable tends to increase
(resp. decrease) proportionally. If x
i
and y
i
are two data sets with means x and y
and standard deviations s
x
and s
y
, then r is given by:
r =

n
i=1
(x
i
x)(y
i
y)
(n 1)s
x
s
y
.
36. Suppose you work for a large retail store and your manager asks you study the rela-
80 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
tionship between revenue (sales) and the unemployment rate. Assume you assemble the
unemployment rates and sales over the last six months in the table below.
Unemp (%) Sales ($)
5 50350
7 37570
8 33140
9 25550
8 38750
5 55450
Let x be the unemployment rate and y the sales.
a. What is x?
b. What is y?
c. What is s
x
?
d. What is s
y
?
e. What is r?
37. Prove the following useful formula for the numerator of the formula for r:

(x
i
x)(y
i
y) =
_

x
i
y
i
_

1
n

x
i

y
i
.
38. Another common question in statistics is, What is the line that best ts the data? As-
suming you believe that two variables have a linear relationship, youd like to know the
what line in the form y = mx + b appears to best describe the data. Here we use a hat
to emphasize that the output of the above formula is a predicted value, as opposed to an
actual value. In classical linear regression, formulas for the parameters m and b are:
m =

n
i=1
(x
i
x)(y
i
y)

n
i=1
(x x)
2
b = y m x.
a. Use these formulas to calculate the regression line of the above data.
b. What sales does the line predict for an unemployment rate of 6%?
2.1. SUMS AND DIFFERENCES 81
Linear Regression FAQ
You probably have three good questions now:
Q In what sense is this the line of best t?
A The line of best t should be close to the data. We quantify closeness by the square
of the dierence between the predicted and actual data. More specically, the quantity

( y y)
2
is minimized.
Q Where do the formulas for m and b come from?
A We nd m and b by minimizing the function:
f(m, b) =

( y y)
2
=

(mx +b y)
2
.
This is an elementary exercise in multivariable Calculus. In single variable Calculus, we
nd minima of a function by locating points where the derivative is zero. This procedure
has a natural analog in higher dimensions.
Q When is it appropriate to use linear regression?
A There are several very important technical assumptions lurking behind linear regression
that are seldom checked. One of these assumptions is that the data should be normally
distributed about its mean at each y value. Many important economic models, including
the Capital Asset Pricing Model and the Nobel Prize winning Black-Scholes option pricing
method rely on the normality assumption. The extremely non-normal market uctuations
sparked by the burst of the housing market bubble in 2008-2009 has caused many people
to question the validity of the normality assumption.
Exercises 39 - 43 are based on the lyrics of the song The Twelve Days of Christ-
mas. You may nd it helpful to have the lyrics in hand.
39. Let f(m) be the total number of gifts received on the mth day. On the third day of
Christmas, for example, a total of six gifts are given. Three french hens, two turtle doves,
and a partridge in a pair tree. Write a formula for f(m) using summation notation.
40. What is f(12)?
82 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
41. The cumulative number of gifts received after m days, call this function F(m), is also
growing. By the end of the third day, for example, 10 gifts have been received: 6 from
the 3rd day, 3 from the 2nd day and 1 from the 1st day. Express F(m) using summation
notation.
42. What is F(12)?
43. Research the terms triangular numbers and tetrahedral numbers on the internet. How do
these terms relate to these exercises?
44. By how much does ln 2 dier from
4

n=1
(1)
n1
n
? Does the summation get closer to ln 2 if
the upper limit is changed from 4 to 5?
45. Let A
n
=
n

u=1
_
u
2
+ 1
2
_
2
and B
n
=
n

u=1
_
u
2
1
2
_
2
. Find an expression for A
n
+ B
n
.
Hint: combine the sums immediately rather than dealing with each one individually.
The Fibonacci sequence is dened recursively. Set F
0
= 0, F
1
= 1 and, for k 2,
F
k
= F
k1
+F
k2
. This gives a sequence 0, 1, 1, 2, 3, 5, 8, .... Prove the following facts
about the Fibonacci sequence.
46. F
k
= F
k2
.
47.
2n

k=0
F
k
= F
2n+2
1.
48.
n

k=1
F
2k1
= F
2n
.
49.
n

k=0
F
2k
= F
2n+1
1.
50.
n

k=0
F
2
k
= F
n
F
n+1
.
2.2. PRELUDE TO THE DEFINITE INTEGRAL 83
2.2 Prelude to the Denite Integral:
Riemann Sums
In dierential Calculus, the instantaneous rate of change of a function is dened as follows: we
have a notion of the average rate of change of the function, and we believe, for small changes
in the independent variable, that the average rate of change approximates something. We
then take limits of the average rates of change to dene the instantaneous rate of change, the
derivative.
In this section, we do something analogous: we will dene Riemann sums, which are supposed
to be approximations to continuous sums of an innite number of innitesimal contributions.
As with the passage from average rates of change to instantaneous ones, we pass from Riemann
sums to continuous sums by taking limits. The continuous sum that we end up with is called
the denite integral.
Lets begin with an extended example.
Example 2.2.1. Suppose that a car is traveling along a straight road, on which a coordinate
axis has been laid out in meters. Let p(t) denote the position of the car (on the axis), at time t
seconds (after some initial starting time), and let v(t) denote the velocity of the car, in meters
per second, at time t seconds.
If the driver of the car were to brake hard, or step down hard on the accelerator, he or she
could easily change the velocity of the car by a noticeable amount in 2 seconds. However, it
would be dicult to change the velocity of the car in a signicant way in a substantially smaller
time interval, like 0.25 seconds.
We want to discuss how you would estimate the change in the position, the displacement, of
the car, between times 0 and 2 seconds, if you were given some velocity measurements. First,
we should clarify that, when we have two numbers a and b, where a < b, and we write that x
is between a and b, we mean to allow for the possibility that x equals a or b, i.e., x is between
a and b means a x b. If a < b, and we write that x is strictly between a and b, we mean
to disallow the possibility that x equals a or b, i.e., x is strictly between a and b means
a < x < b.
Now, suppose that we know that,
at some time between 0 and 0.3 seconds, the car is moving with velocity 30 m/s;
84 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
at some time between 0.3 and 0.8 seconds, the car is moving with velocity 20 m/s;
at some time between 0.8 and 1.2 seconds, the car is moving with velocity 10 m/s; and,
at some time between 1.2 and 2 seconds, the car is moving at 2 m/s (i.e., at 2 m/s in the
negative direction).
How can we estimate the displacement of the car between times t = 0 and t = 2 seconds, i.e.,
how can we estimate p(2) p(0)?
Note that we have the equality
p(2) p(0) = (2.2)
_
p(2) p(1.2)
_
+
_
p(1.2) p(0.8)
_
+
_
p(0.8) p(0.3)
_
+
_
p(0.3) p(0)
_
,
which results from the fact that all of the terms, other than p(2) and p(0), cancel out, that is,
the sum telescopes, as in Proposition 2.1.9. Thus, to estimate p(2) p(0), we can estimate the
four sub-displacements appearing in Formula 2.2, and add them together.
Okay. Fine. So how do we approximate each of the sub-displacements?
The answer is that, for each of the four intervals of time [0, 0.3], [0.3, 0.8], [0.8, 1.2], and
[1.2, 2] (subintervals of [0, 2]), we use, as an approximation, that the velocity is constant on
the given interval. Why do we do this? For two reasons. First, the subintervals of time are
fairly small, small enough so that we believe that the velocity cannot change too much on each
subinterval. Second, we dont really have much of a choice, considering that the only data that
we are given is the velocity at some time in each of the given subintervals.
Thus, as an approximation, we assume/pretend that, on the subintervals [0, 0.3], [0.3, 0.8],
[0.8, 1.2], and [1.2, 2], the velocity is constantly 30, 20, 10, and 2 m/s, respectively. Now, if the
velocity is a constant during an interval of time, then that velocity is also the average velocity
during the interval of time, and we know that the average velocity on an interval of time is the
change in position on the interval divided by the change in time on the interval. Thus, we have
the four approximations (in m/s):
p(2) p(1.2)
2 1.2
2,
p(1.2) p(0.8)
1.2 0.8
10,
p(0.8) p(0.3)
0.8 0.3
20, and
p(0.3) p(0)
0.3 0
30.
Therefore, we approximate that the displacement of the car, between times t = 0 and t = 2
seconds:
p(2) p(0) =
2.2. PRELUDE TO THE DEFINITE INTEGRAL 85
_
p(2) p(1.2)
_
+
_
p(1.2) p(0.8)
_
+
_
p(0.8) p(0.3)
_
+
_
p(0.3) p(0)
_

2(2 1.2) + 10(1.2 0.8) + 20(0.8 0.3) + 30(0.3 0) =
1.6 + 4 + 10 + 9 = 21.4 meters.
Therefore, we conclude that the displacement of the car, between times t = 0 and t = 2 seconds,
is approximately 21.4 meters. Note that this is an estimate of the change in the position; we
cannot estimate (in a reasonable manner) the actual position of the car at t = 2 seconds, unless
we have p(0) (or, at least, an estimate of p(0)).
Our data and approximations of the displacements on the subintervals look best in a table.
subinterval (sec.) [0, 0.3] [0.3, 0.8] [0.8, 1.2] [1.2, 2]
time (sec.) t
1
t
2
t
3
t
4
velocity (m/s) 30 20 10 -2
Approx. p (meters) 9 10 4 -1.6
Approx. total p = 9 + 10 + 4 + (1.6) = 21.4 meters.
Note that we gave the names t
1
, t
2
, t
3
, and t
4
to the times in the corresponding subintervals
at which we were given the velocities that appear in the 3rd row. Our approximation of the
total displacement is the sum of the entries in the 4th row of the table.
How could we obtain a better approximation of the displacement p(2) p(0) or, at least,
an approximation that wed expect to be better? We could subdivide the subintervals above
into even smaller subintervals of time, and be given velocities at some time in each of the new
subintervals.
For instance, if we thought that 0.8 seconds was too large of a change in time over which
to approximate the velocity as being constant, we might subdivide the last subinterval [1.2, 2]
above into two subintervals, say [1.2, 1.6] and [1.6, 2]. We would then need to know which of
these new subintervals contains the time t
4
(both would, if t
4
= 1.6), and wed need to be given
the velocity at some time in the other new subinterval. For this example, lets assume that t
4
is
in the subinterval [1.2, 1.6], and that we know, at some time t
5
in the interval [1.6, 2], that the
velocity is 5 m/s. What approximation do we get for the total displacement now?
Our new table is:
subinterval (sec.) [0, 0.3] [0.3, 0.8] [0.8, 1.2] [1.2, 1.6] [1.6, 2]
time (sec.) t
1
t
2
t
3
t
4
t
5
velocity (m/s) 30 20 10 -2 -5
Approx. p (meters) 9 10 4 -0.8 -2
Approx. total p = 9 + 10 + 4 + (0.8) + (2) = 20.2 meters.
86 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
Our new approximation for the displacement p(2) p(0) is no longer 21.4 meters, but is now
20.2 meters.
Do we really know that this new approximation is better than the rst one? Not really. For
all we know, in the last subinterval [1.6, 2], the velocity is actually closer to 2 m/s most
of time, and is only close to 5 m/s for a tiny subinterval around t
5
. We can try to appeal
to our intuition and/or physical experience to say that, in 0.4 seconds, the velocity cannot
change a signicant amount, but thats no proof, and it also leaves us with the question of:
whats a signicant amount? Nonetheless, we suspect that subdividing our subintervals
into smaller subintervals leads to better approximations, provided we take our subdivisions
small enough.
Despite the remark/warning above, it is nonetheless true that there is a strong sense in which
using small enough subintervals guarantees a close approximation; see the remark below.
Remark 2.2.2. It is possible to make precise in what way the approximation gets better
as the time intervals get smaller. First, we need to assume that the velocity function v(t)
is continuous on the interval [0, 2]. Now, call a subdivision of the interval [0, 2] into a nite
collection of subintervals a partition of [0, 2] (technically, the partition is just the set of endpoints
of the subintervals; see Denition 2.2.3). Call the length of the longest (any one of the longest)
subintervals the mesh of the partition. Thus, if the mesh of the partition is less than some (small)
positive constant , then every subinterval in the partition has length less than ; informally,
this means that saying that the mesh of a partition is small implies that every subinterval in
the partition is small.
Now, we can state carefully in what sense picking small subintervals of time, and being given
the velocity at one time per time interval, allows you to accurately approximate the actual
displacement. Given any > 0 (think: could be an arbitrarily small positive number), we can
guarantee that our approximation for the displacement is within of the actual displacement
by using partitions with a small enough mesh, i.e., there exists a number > 0 such that, for
every partition of [0, 2], with mesh less than , for every choice of one time (a sample point) per
subinterval at which to be given the velocity, the sum of the products of the lengths of each
subinterval with the given velocities for the subintervals will be within of the displacement
p(2) p(0).
The above statement is a result of the denition of the denite integral, Denition 2.3.1,
Theorem 2.3.8, and the Fundamental Theorem of Calculus, Theorem 2.4.10.
2.2. PRELUDE TO THE DEFINITE INTEGRAL 87
We wish to discuss partitions, meshes, sample points, and the summing process above in a
more general context.
Suppose that a < b. It turns out to be convenient to dene a partition of the interval [a, b]
by giving the endpoints of the subintervals that we chop [a, b] into, rather than dening the
partition to be the collection of subintervals themselves.
Denition 2.2.3. A partition T of the interval [a, b], into n subintervals, is an ordered
set of numbers x
0
, x
1
, . . . , x
n
such that x
0
= a, x
n
= b, and x
0
< x
1
< < x
n
.
If 1 i n, then the closed interval [x
i1
, x
i
] is the i-th subinterval of the partition T;
note that the nite dierence x
i
equals x
i
x
i1
, which is the length of the i-th subinterval
of the partition.
The mesh of a partition T, denoted [[ T [[, is the maximum length of a subinterval of the
partition, i.e., [[ T [[ = max
_
x
i
[ 1 i n
_
.
A set o of sample points for a partition T is an ordered set of points, one for each
subinterval of the partition, i.e., an ordered set of elements s
1
, s
2
, . . . , s
n
such that, for all
i, where 1 i n, s
i
is in the i-th subinterval of the partition, that is, x
i1
s
i
x
i
.
A sampled partition of the interval [a, b] is an ordered pair (T, o), consisting of a partition
T of the interval [a, b] and a set o of sample points for T.
For instance, in Example 2.2.1, we had the partition T = 0, 0.3, 0.8, 1.2, 2. The 1st subin-
terval is [0, 0.3], the 2nd is [0.3, 0.8], the 3rd subinterval is [0.8, 1.2], and the 4th is [1.2, 2]. The
mesh is [[ T [[ = 2 1.2 = 0.8. We were not given explicit sample points; we were simply told
that, in each subinterval, there was a sample point (a time) at which we knew the velocity, and
we were given those velocities.
Now, we need to dene the general setup for the type of summation that we used in Exam-
ple 2.2.1. We continue to assume that a < b.
88 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
Denition 2.2.4. Let T = x
0
, . . . , x
n
be a partition of [a, b], into n subintervals, let
o = s
1
, . . . , s
n
be a set of sample points for T, and let f be a real-valued function whose
domain includes the set of sample points o.
Then, the Riemann sum, !
S
P
(f), of f, with respect to T and o, is dened to be
n

i=1
f(s
i
)x
i
=
f(s
1
)(x
1
x
0
) +f(s
2
)(x
2
x
1
) + +f(s
n1
)(x
n1
x
n2
) +f(s
n
)(x
n
x
n1
).
For instance, in Example 2.2.1, for each of our sampled partitions, the associated Riemann
sums of the velocity are precisely our approximations of the displacement p(2) p(0).
We need some terminology for when one partition has smaller subintervals than another,
and for when we enlarge our set of sample points. It will be helpful to recall the notion of a
subset; a set A is a subset of a set B, denoted A B, if and only if every element of A is also
an element of B.
Denition 2.2.5. A partition Q of an interval [a, b] is a renement of a partition T of
[a, b] if and only if every point in T is in Q, i.e., if and only if T Q.
If (T, o) is a sampled partition of [a, b], then another sampled partition (Q, T ) of [a, b] is a
renement of (T, o) if and only if T Q and o T .
Remark 2.2.6. As an example, our second sampled partition in Example 2.2.1, the one given
in the second table, is a renement of the original sampled partition in that example.
Understand the point of a renement: the subintervals of a renement are obtained from the
subintervals of the original partition by subdividing some of the original subintervals into more,
smaller, subintervals. Thus, if T and Q are partitions of an interval, and T Q, then we have
an inequality of meshes m(Q) [[ T [[.
For a renement of a sampled partition, we not only subdivide some of the original subin-
tervals, but we keep the original sample points and throw in some new ones, to give us sample
points in each new subinterval that does not contain an old sample point (or, if an old sample
point is the endpoint of a subinterval in the rened partition, then we assign that point to one
2.2. PRELUDE TO THE DEFINITE INTEGRAL 89
of the two adjacent subintervals, and must select a new sample point for the other adjacent
subinterval).
Our interest in Riemann sums is not limited to velocity and displacement.
Example 2.2.7. Suppose that a balloon is being inated between times t = 0 and t = 120
seconds, and the ination rates, measured in in
3
/s, are 8, 7, 5, and 2 at times 50, 70, 90, and
110 seconds, respectively.
Estimate the change in the volume V = V (t) of air in the balloon between times t = 40 and
t = 120 seconds.
Solution:
We will begin by selecting a partition of [40, 120] that allows us to use 50, 70, 90, and 110
as sample points. An obvious choice is to subdivide [40, 120] into 4 subintervals of equal length
t = (120 40)/4 = 20. Then, the subintervals of the partition would be [40, 60], [60, 80],
[80, 100], and [100, 120], and the points 50, 70, 90, and 110 would, in fact, be sample points for
this partition (in fact, they are the midpoints of the subintervals).
subinterval (sec.) [40,60] [60,80] [80,100] [100, 120]
time (sec.) 50 70 90 110
ination rate (in
3
/s) 8 7 5 2
Approx. V (in
3
) 160 140 100 40
Approx. total V = 160 + 140 + 100 + 40 = 440 in
3
.
We should emphasize that 440 in
3
is the approximate change in the volume between times
t = 40 and t = 120 seconds. With no knowledge of the volume of air in the balloon at time t = 40
seconds, we cannot reasonably approximate the actual volume V (120) of air in the balloon at
time t = 120 seconds.
Lets look at another example.
Example 2.2.8. Suppose that a circular rod, of length 1 meter, and cross-sectional area 0.01
m
2
(i.e., of radius 0.1/

meters) is lying along the x-axis.

90 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
Figure 2.1: A rod of varying density.
Suppose that, for all x such that 0 x 1, at each point in the cross section of the rod at
x meters, the density of the rod is (x) = (1 + x) kg/m
3
. We would like to estimate the total
mass of the rod, by partitioning the interval [0, 1] into 5 intervals of equal length.
Thus, each subinterval of our partition will have length x = (1 0)/5 = 0.2 meters, and so
our partition is T = 0, 0.2, 0.4, 0.6, 0.8, 1. While we shall not use it, we remark that the mesh
of T is [[ T [[ = 0.2.
Lets rst use the left endpoint of each subinterval of our partition as the sample point for
that subinterval. Thus, our left sample set is L = 0, 0.2, 0.4, 0.6, 0.8.
This will yield the left Riemann sum, once we determine what function were nding Riemann
sums of. Average density is mass per volume. In an analogous way to how we dealt with velocity,
position, and time in Example 2.2.1, we will assume that, for a small x-interval, the average
density is approximated fairly well by the (instantaneous) density at any x-coordinate in the
given interval. Therefore, if we take the density (x) at a point, and multiply times a little
chunk of volume around that point, we will obtain an approximation of the mass of the chunk.
A nice, manageable chunk of volume surrounding a given x-coordinate is given simply by
taking an x-interval, containing the given x-coordinate, and multiplying by the cross-sectional
area, 0.01 m
2
. Thus, if were going to multiply by lengths of subintervals in our Riemann sums,
then the function that we want to nd Riemann sums of, in order to approximate the mass, is
cross-sectional area times density, i.e.,
f(x) = 0.01(1 +x) kg/m.
For 0 x 1 meter, let M(x) denote the mass of the rod between 0 and x, so that, if
0 a < b 1, then M(b) M(a) is the mass of the rod between x = a and x = b. We obtain a
table of data that looks like:
2.2. PRELUDE TO THE DEFINITE INTEGRAL 91
[x
i1
, x
i
] (m) [0, 0.2] [0.2, 0.4] [0.4,0.6] [0.6, 0.8] [0.8,1]
s
i
(m) 0 0.2 0.4 0.6 0.8
f(s
i
) (kg/m) 0.01 0.01(1.2) 0.01(1.4) 0.01(1.6) 0.01(1.8)
M (kg) 0.01(0.2) 0.01(1.2)(0.2) 0.01(1.4)(0.2) 0.01(1.6)(0.2) 0.01(1.8)(0.2)
Approx. total M = !
L
P
(f) = (0.01)(0.2)(1 + 1.2 + 1.4 + 1.6 + 1.8) = 0.014 kg.
Now lets look at the Riemann sum, where we use the same partition, but we use right
endpoints of the subintervals as sample points. Thus, we let R = 0.2, 0.4, 0.6, 0.8, 1, and will
calculate !
R
P
(f), the right Riemann sum of the same function f(x) that we used above.
[x
i1
, x
i
] (m) [0, 0.2] [0.2, 0.4] [0.4,0.6] [0.6, 0.8] [0.8,1]
s
i
(m) 0 0.2 0.4 0.6 0.8
f(s
i
) (kg/m) 0.01(1.2) 0.01(1.4) 0.01(1.6) 0.01(1.8) 0.01(2)
M (kg) 0.01(1.2)(0.2) 0.01(1.4)(0.2) 0.01(1.6)(0.2) 0.01(1.8)(0.2) 0.01(2)(0.2)
Approx. total M = !
R
P
(f) = (0.01)(0.2)(1.2 + 1.4 + 1.6 + 1.8 + 2) = 0.016 kg.
There is a third common set of sample points that frequently get used: the midpoints of
the intervals; we would normally denote this set by M, but, in our current example, M is
already being used to denote the mass. So, we let C = 0.1, 0.3, 0.5, 0.7, 0.9, and call !
C
P
(f)
the midpoint Riemann sum. We calculate:
[x
i1
, x
i
] (m) [0, 0.2] [0.2, 0.4] [0.4,0.6] [0.6, 0.8] [0.8,1]
s
i
(m) 0 0.2 0.4 0.6 0.8
f(s
i
) (kg/m) 0.01(1.1) 0.01(1.3) 0.01(1.5) 0.01(1.7) 0.01(1.9)
M (kg) 0.01(1.1)(0.2) 0.01(1.3)(0.2) 0.01(1.5)(0.2) 0.01(1.7)(0.2) 0.01(1.9)(0.2)
Approx. total M = !
C
P
(f) = (0.01)(0.2)(1.1 + 1.3 + 1.5 + 1.7 + 1.9) = 0.015 kg.
The fact that the midpoint Riemann sum ends up in-between the left and right Riemann
sums, and is, in fact, the average of the left and right Riemann sums, is not generally true. These
things are true in this example because our function f is linear, which also causes the midpoint
Riemann sum to be exactly the value of what we will later call the denite integral of f over [0, 1].
In the completely general setting, the midpoint Riemann sums can yield worse approximations
of the denite integral than the left and right Riemann sums. However, generally, we expect
midpoint Riemann sums to yield better approximations than the left and right Riemann sums.
92 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
What we would like to see is that, if we use partitions of [0, 1] with arbitrarily small (positive)
meshes, then the associated Riemann sums of f get arbitrarily close to some value; that value
must then be the total mass of the rod.
Suppose that n is a natural number. Divide the interval [0, 1] into n subintervals of equal
length, so that the associated partition is
T
n
= 0, 1/n, 2/n, . . . , (n 1)/n, 1.
We will calculate the left, right and midpoint Riemann sums of f with respect to T
n
, and see
that all three of them approach the same limit as n . (Here, we are using the limit of a
sequence, which means that n takes on only integer values. We discussed this in [2]; also see
Denition 4.5.1.).
The left sample set is L
n
= 0, 1/n, 2/n, . . . , (n 1)/n, where s
i
= (i 1)/n. The right
sample set is R
n
= 1/n, 2/n, . . . , 1, where s
i
= i/n. The midpoint sample set is C
n
=
(1/(2n), 3/(2n), . . . , (2n 1)/(2n), where s
i
= (i 1/2)/n = (2i 1)/(2n).
We have:
!
Ln
Pn
(f) =
n

i=1
f
_
i 1
n
_

1
n
=
n

i=1
0.01
_
1 +
i 1
n
_

1
n
=
0.01
n
__
n

i=1
1
_
+
1
n

_
n

i=1
i
_

1
n

_
n

i=1
1
__
.
As
n

i=1
1 = n, and
n

i=1
i =
n(n + 1)
2
, by Corollary 2.1.11, we nd that
!
Ln
Pn
(f) =
0.01
n
_
n +
1
n

n(n + 1)
2

1
n
n
_
=
0.01
n
_
2n + (n + 1) 2
2
_
=
0.01
_
3 1/n
2
_
.
Thus, as n , we see that the left Riemann sums approach 0.01(3/2) = 0.015 kg.
Similarly, we nd
2.2. PRELUDE TO THE DEFINITE INTEGRAL 93
!
Rn
Pn
(f) =
n

i=1
f
_
i
n
_

1
n
=
n

i=1
0.01
_
1 +
i
n
_

1
n
=
0.01
n
__
n

i=1
1
_
+
1
n

_
n

i=1
i
__
=
0.01
n
_
n +
1
n

n(n + 1)
2
_
=
0.01
n
_
2n + (n + 1)
2
_
=
0.01
_
3 + 1/n
2
_
.
Thus, as n , we see that the right Riemann sums also approach 0.01(3/2) = 0.015 kg.
Finally, for the midpoint Riemann sums, we have
!
Cn
Pn
(f) =
n

i=1
f
_
2i 1
2n
_

1
n
=
n

i=1
0.01
_
1 +
2i 1
2n
_

1
n
=
0.01
n
__
n

i=1
1
_
+
1
n

_
n

i=1
i
_

1
2n
_
n

i=1
1
__
=
0.01
n
_
n +
1
n

n(n + 1)
2

1
2n
n
_
=
0.01
n

3n
2
= 0.015.
Thus, for each n, !
Cn
Pn
(f) = 0.015 kg, and so, certainly, as n , we see that the limit of the
midpoint Riemann sums is also 0.015 kg.
In dierential Calculus, it is extremely helpful to picture instantaneous rates of change
graphically. This is accomplished by noting that the slope of a secant line yields the average rate
of change, and then we take limits to arrive at the notion of a tangent line; we then visualize
the instantaneous rate of change as the slope of the appropriate tangent line.
Our question now is: can we do something similar for Riemann sums and their
limits, in order to picture these things geometrically?
The answer is (as you may have suspected): YES. We discuss this in the example below.
94 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
Example 2.2.9. How can we graphically represent Riemann sums? As a specic example, how
can we graphically represent the left, right, and midpoint Riemann sums from Example 2.2.8?
Recall that the function we were nding Riemann sums of was
y = f(x) = 0.01(1 +x).
(We shall omit the units throughout this example.) The graph is, of course, a straight line.
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
Figure 2.2: The graph of y = 0.01(1 +x).
Lets rst look at the partition of [0, 1] given by T = 0, 0.2, 0.4, 0.6, 0.8, 1, and the left
sample set L = 0, 0.2, 0.4, 0.6, 0.8. How can we visualize !
L
P
(f)? We do it in terms of the
areas of rectangles.
The Riemann sum is
!
L
P
(f) = f(s
1
)x
1
+f(s
2
)x
2
+f(s
3
)x
3
+f(s
4
)x
4
+f(s
5
)x
5
=
f(0) 0.2 +f(0.2) 0.2 +f(0.4) 0.2 +f(0.6) 0.2 +f(0.8) 0.2.
As f(x) is non-negative on the interval [0, 1], we can interpret f(s
i
)x
i
as the area of a rectangle
of height f(s
i
) and width x
i
; we draw this rectangle over the i-th subinterval on the x-axis.
Thus, we represent f(0)x
1
= f(0) 0.2 by the rectangle in Figure 2.3, and the entire left
Riemann sum is equal to the total area of all ve of the inscribed rectangles in Figure 2.4
2.2. PRELUDE TO THE DEFINITE INTEGRAL 95
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
Figure 2.3: The rst left summand.
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
Figure 2.4: The left Riemann sum.
What about the right Riemann sum? We have
!
R
P
(f) = f(s
1
)x
1
+f(s
2
)x
2
+f(s
3
)x
3
+f(s
4
)x
4
+f(s
5
)x
5
=
f(0.2) 0.2 +f(0.4) 0.2 +f(0.6) 0.2 +f(0.8) 0.2 +f(1) 0.2.
Again, we interpret f(s
i
)x
i
as the area of a rectangle of height f(s
i
) and width x
i
. Now,
however, the corresponding rectangles are above the line y = 0.01(1+x). Thus, the entire right
Riemann sum is equal to the total area of all ve of the superscribed rectangles in Figure 2.5.
In order to compare the left and right Riemann sums visually, we have given both collections
of relevant rectangles in Figure 2.6.
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
Figure 2.5: The right Riemann sum.
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
Figure 2.6: Combined Riemann sums.
What happens when we rene our partition into 10 subintervals of equal length, and look at
96 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
both the left and right Riemann sums in terms of area of rectangles? We obtain the collections
of inscribed and circumscribed rectangles in Figure 2.7.
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
Figure 2.7: The rectangles for the rened Riemann sums.
As you can see, the total areas of the left and right rectangles have gotten closer together,
that is, the dierence between the area of the circumscribed rectangles and the area of the
inscribed rectangles has gotten smaller; both areas have gotten closer to being the actual area of
the trapezoid under the graph of f(x) = 0.01(1 +x), and above the interval [0, 1] in the x-axis.
Therefore, we see that, as we rene our partition, the total area of the rectangles
representing the Riemann sums approaches the actual area under the graph and
above the closed interval on the x-axis.
Of course, we know how to calculate the area of a trapezoid; its one half the sum of the
lengths of the bases times the height, where the bases are the two parallel sides. Thus, the area
of our trapezoid is
1
2
_
f(0) +f(1)
_
1 =
1
2
(0.01 + 0.02) = 0.015,
which agrees with what we found for the limit of our Riemann sums in Example 2.2.8
We have looked at the Riemann sums graphically only for the left and right Riemann sums.
What about the midpoint (Figure 2.8) Riemann sums? What about Riemann sums with arbi-
trary sample sets?
2.2. PRELUDE TO THE DEFINITE INTEGRAL 97
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
Figure 2.8: The midpoint Riemann sum.
Of course, we cant sketch rectangles corresponding to every choice that you might make
for sample points. However, it is true that, since our f(x) is continuous, all limits of
Riemann sums, with arbitrary sample sets approach the same value, so long as the
meshes of the partitions approach zero; that same value, for a non-negative function, is
the area under the graph and above the closed interval under consideration. This follows from
Denition 2.3.1 and Theorem 2.3.8.
In fact, for f(x) = 0.01(1 + x), it is easy to see that, since the left and right Riemann
sums approach the area under the graph (as we take partitions with arbitrarily small mesh),
so must Riemann sums using any sample sets whatsoever. Why is it easy to see this? Because
f(x) = 0.01(1 +x) is monotonically increasing, which implies that the smallest that f ever gets
on a closed subinterval is at the left endpoint of the subinterval, and largest that f ever gets on
a closed subinterval is at the right endpoint of the subinterval.
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
Figure 2.9: Left, right, and midpoint Riemann sums.
98 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
Thus, for all sets of sample points for a given partition, the rectangles representing the
Riemann sums are trapped between the left and right Riemann sum rectangles. See Figure 2.9,
where we have included the midpoint Riemann sum, with our previous partition of the interval
[0, 1] into 5 subintervals of equal length.
You may be thinking: Ah, so, we can calculate limits of Riemann sums simply by calculating
areas, like that of a trapezoid. This is easy. The problem is that we dont know the areas under
the graphs of functions, at least, not by applying easy formulas from basic geometry.
Suppose, for instance, that our function f has been given, not by a function whose graph is
a straight line, but rather by
f(x) = 0.01(1 +x
2
).
Then, exactly how do we calculate the area under the graph of this f and over the interval [0, 1]?
0.2 0 -0.2 0.4 0.6 0.8 1.0 1.2
x
y
0.01
0.02
Figure 2.10: Area under a parabola.
The answer is that we calculate the limit of Riemann sums to nd the area, not the other
way around. Only in very few cases can we calculate the limit of Riemann sums from known
formulas for the area.
Before we leave this example, we should comment on how we would have graphically repre-
sented the Riemann sums had f(x) been negative, or sometimes positive and sometimes negative.
Suppose that we have a continuous function f such that f(x) < 0 for all x in the interval
[0, 1]. Then, given a partition T = x
0
, . . . , x
n
of [0, 1], and a sample set o = s
1
, . . . , s
n
for
T, the summand f(s
i
)x
i
in the Riemann sum !
S
P
(f) cannot be represented by area, since it
will be a negative quantity. However, we can consider a rectangle under the subinterval [x
i1
, x
i
]
of height f(s
i
), and then f(s
i
)x
i
will be equal to negative the area of this rectangle beneath
the x-axis. See Figure 2.11.
2.2. PRELUDE TO THE DEFINITE INTEGRAL 99
Thus, the limit of the Riemann sums of our negative function f, as the mesh of the partitions
approaches zero, will be equal to negative the area above the graph and under the
interval [0, 1].
x
i
x
y
x
i-1
s
i
}
_
height f ( ) = s
i
x
y
)
(
f

=
y
x
A
1
A
2
Figure 2.11: A graph below the x-axis.
Finally, what if f is continuous, but is negative sometimes and positive other times? As we
shall see in Theorem 2.3.16, the limit of Riemann sums of such an f is a sum of contributions
from where f 0 and where f 0 and, thus, in terms, of area, the limit of Riemann sums, as
the mesh of the partitions approaches zero, is the equal to the area under the graph and above
the x-axis minus the area above the graph and under the x-axis.
Thus, for a function f such as that in Figure 2.12, the limit of !
S
P
(f), as the mesh of the
partition T of [0, 1] approaches zero, will equal the dierence of areas A
2
A
1
.
In no way are we claiming that the area under the x-axis and above the graph is negative;
we are simply saying that the contribution to the limit of the Riemann sums from that
portion of the graph is A
1
, where A
1
itself is positive.
x
i
x
y
x
i-1
s
i
}
_
height f ( ) = s
i
x
y
)
(
f

=
y
x
A
1
A
2
Figure 2.12: A graph below and above the x-axis.
100 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
2.2.1 Exercises
In each of Exercises 1 through 5, calculate the mesh of the given partition.
1. T = 2, 0, 3, 5, 7, 9.
2. T = ln 1, ln 2, ln 3, ln 4, ln 5.
3. T = 0, 1/4, 1/3, 1/2.
4. T = 0, 1/2, 2/3, 3/4, 4/5.
5. T =
_
0,
1
2
n
,
2
2
n
, ...,
2
n
1
2
n
, 1
_
, where n is a positive integer.
6. Explain why x
0
= 0, x
1
= 2, x
2
= 1, x
3
= 3, and x
4
= 4 cannot be used as a partition.
For the following true/false questions, in Exercises 7 through 13, assume all parti-
tions mentioned are on the same interval, [0, 1].
7. If [[Q[[ [[P[[, then Q is a renement of T.
8. If Q is a renement of T, then [[Q[[ [[P[[.
9. Given a specic mesh 1, where is a rational number, there exists a partition T of
[0, 1], which has mesh , where every point of T is a rational number.
10. If (Q, T ) is a renement of (T, o), then T is a renement of Q.
11. If T is a renement of Q, then (Q, T ) is a renement of (T, o).
12. The sampled points of a partition must all be distinct.
13. If f is a continuous function and T is a partition of some interval, then !
M
P
(f) is always
between !
R
P
(f) and !
L
P
(f).
In each of Exercises 14 through 16, say (a) whether Q is a renement of partition
T and (b) whether (Q, T ) is a renement of (T, o).
14. T = 0, 1/4, 1/2, 1, Q = 0, 1/2, 1, o = 1/8, 2/3, 3/4, T = 1/8, 3/4.
15. T = 0, 2/5, 1, Q = 0, 1/3, 2/5, 3/5, 1, o = 1/3, 2/3, T = 1/4, 1/3, 2/3, 3/4.
2.2. PRELUDE TO THE DEFINITE INTEGRAL 101
16. T = 0, 1/8, 2/5, 3/4, 1, Q = 0, 1/8, 3/13, 2/5, 5/9, 3/4, 4/5, 1, o = 1/16, 2/5, 5/9, 3/4,
T = 1/7, 4/21, 5/18, 1/2, 8/13, 7/9, 9/10.
17. Suppose a car is traveling with variable velocity for three minutes, and that
at some point between 0 and 40 seconds, the car is moving with velocity 15 m/s;
at some point between 40 and 90 seconds, the car is moving with velocity 20 m/s;
at some point between 90 and 120 seconds, the car is moving with velocity 25 m/s;
at some point between 120 and 180 seconds, the car is moving with velocity 15 m/s.
Use Riemann sums to approximate the total displacement of the car from its initial posi-
tion.
18. Suppose a car is traveling with variable velocity for one hour, and that
at some point between 0 and 12 minutes, the car is moving with velocity 40 miles/hour;
at some point between 12 and 30 minutes, the car is moving with velocity -10
miles/hour;
at some point between 30 and 40 minutes, the car is moving with velocity 20 miles/hour;
at some point between 40 and 60 minutes, the car is moving with velocity 0 miles/hour.
Use Riemann sums to approximate the total displacement of the car from its initial posi-
tion.
19. Suppose a car is traveling with variable acceleration for 30 minutes, and that
at some point between 0 and 10 minutes, the car is accelerating at a rate of 5 miles
per hour per hour.
at some point between 10 and 15 minutes, the car is accelerating at a rate of -3 miles
per hour per hour.
at some point between 15 and 25 minutes, the car is accelerating at a rate of 1 mile
per hour per hour.
at some point between 25 and 30 minutes, the car is accelerating at a rate of 4 miles
per hour per hour.
Use Riemann sums to approximate the total change in velocity of the car from its initial
velocity.
20. Suppose that a balloon starts with no air in it, and is inated between t = 0 and t = 90
seconds. Suppose further that,
102 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
at some time between 0 and 20 seconds, the balloon is inating at a rate of 2 in
3
/s.
at some time between 20 and 50 seconds, the balloon is inating at a rate of 5 in
3
/s.
at some time between 50 and 75 seconds, the balloon is inating at a rate of -3 in
3
/s.
Perhaps someone lost their grip on the balloon momentarily.
at some time between 75 and 90 seconds, the balloon is inating at a rate of 4 in
3
/s.
Use Riemann sums to approximate the total volume of air in the balloon at t = 90 seconds.
21. If f(t) = c on the interval [a, b], prove that the Riemann sum of f on [a, b] using any
sample set and any partition is c(b a). Give a physical interpretation of this fact if f is
a velocity function.
22. Let f(x) = mx be a line through the origin dened on [0, k]. Assume m > 0. What is the
limiting Riemann sum of this function over this interval? Use the fact that Riemann sums
approximate the area under a curve. Your answer may be non-rigorous.
23. Suppose f is a continuous function on [a, b]. Then f achieves its minimum m and maximum
M. Let T be a partition of [a, b]. Prove that m(b a) !
P
(f) M(b a).
24. Consider the following function with domain [0, 1].
f(x) =
_
1, if x is rational;
0, , if x is irrational.
Show that no matter what partition is chosen, and no matter how small its mesh, we can
always nd two sets of sample point o = s
i
and o

= s

i
for that partition such that
the Riemann sum is 1 on o and 0 on o

.
In each of Exercises 25 through 27, calculate the left Riemann sum for the given
function and partition.
25. h(x) = x, T = 3, 2, 1, 0, 1, 2, 3.
26. g(x) = sin x, T =
_
0,

4
,

2
,
3
4
,
_
.
27. f(x) = x
2
, T = 0, 0.1, 0.2, ..., 0.9, 1.
In each of Exercises 28 through 30, calculate the right Riemann sum for the given
function and partition.
2.2. PRELUDE TO THE DEFINITE INTEGRAL 103
28. k(x) = e
x
/2, T = 0, ln 2, ln 3, ln 4.
29. l(x) = e
x
/2, T = 0, ln 2, ln 3, ln 4.
30. j(x) = cosh x, T = 0, ln 2, ln 3, ln 4. Hint: Use the previous two problems and the fact
that Riemann sums are linear, a fact you will prove in a later exercise.
In each of Exercises 31 through 34, calculate the midpoint Riemann sums for the
following functions and partitions.
31. y(x) = cos x, T =
3
2
,

2
,

2
,
3
2

32. r(x) = 1/x
2
, T = 1/10, 1/8, 1/5, 1/2, 1.
33. g(x) = tan
1
x, T = 0, 1, 2, 3, 4.
34. h(x) = sin xcos x, T = 0, /2, , 3/2, 2.
35. Let f(x) =
_
1 x
2
. Calculate the left Riemann sums for the following partitions.
a. T
1
= 1, 0, 1.
b. T
2
= 1, 0.5, 0, 0.5, 1.
c. T
3
= 1, 2/3, 1/3, 0, 1/3, 2/3, 1.
d. Based on geometric intuition, what is the limiting Riemann sum, as the mesh ap-
proaches 0, using partitions of the interval [1, 1]?
36. Suppose f(x) is a continuous odd function. What is the midpoint Riemann sum using the
partition T = a
3
, a
2
, a
1
, 0, a
1
, a
2
, a
3
, where 0 < a
1
< a
2
< a
3
?
The work done on an object experiencing a constant force F, as the object travels
along the x-axis, from point a to b, is W = F(b a) (provided that the force acts
parallel to the x-axis).
The total work done on an object experiencing a variable force F = F(x) can
be approximated by Riemann sums. Namely, suppose T = x
0
, x
1
, . . . , x
n
is a par-
tition of [a, b] and that o = s
1
, ..., s
n
is a sample set for the partition. Then the
approximate work done is J
S
P
(F) =
n

i=1
F(s
i
)(x
i
x
i1
). The metric units of work are
Newton-meters, or joules. C
In each of Exercises 37 through 39, assume an object moves along the interval
[0, 1], that the partition is T = 0, 1/4, 1/2, 3/4, 1, and that the sample set is o =
1/4, 1/2, 3/4, 1; calculate J
S
P
(F).
104 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
37. F(x) = x
2
.
38. F(x) = sin(x).
39. F(x) = 3x.
40. Let f(x) = 1/x on the interval I
n
= [
1
n
, 1], where n = 2, 3, . . . . Let T =
_
1
n
,
1
n 1
, , ...,
1
2
, 1
_
.
Calculate the left Riemann sum of f over this partition using the left endpoints for n = 2, 3
and 4.
41. Suppose that f and g are two continuous functions on the interval [a, b]. Let (T, o) be
a partition and sample set for this interval. Prove that the Riemann sum is linear in the
sense that
!
S
P
(cf +g) = c !
S
P
(f) +!
S
P
(g)
where c is any real number.
42. Consider the function,
h(x) =
_
1
x
sin
_

x
_
, if 0 < x 1;
0, if x = 0,
and suppose that T is a partition of the interval [0, 1].
a. Show that the Riemann sums can be made arbitrarily large, i.e., for all N > 0, there
exists a sample set o such that !
S
P
> N.
b. Similarly, show that the Riemann sums can be made arbitrarily small, i.e., for all
N > 0, there exists a sample set o such that !
S
P
< N.
43. Consider the function
f(x) =
_
3, if 0 x < 1 or 1 < x 2;
100 x = 1.
a. Let T
n
= 0,
1
2
n
,
2
2
n
, ..., 2 be the uniform mesh where the distance between consecu-
tive points is 1/2
n
. Calculate !
R
Pn
(f),
b. Show that lim
n
!
R
Pn
(f) = 6.
44. Let f(x) = c be a constant function on the interval [0, 1] with c > 0. Let
g(x) =
_
c, if x ,= 1/2;
0, if x = 1/2.
2.2. PRELUDE TO THE DEFINITE INTEGRAL 105
Let T
n
be the partition on [0, 1] where the distance between each sample point is 1/n.
Show that lim
n
!
R
P
(f) = lim
n
!
R
P
(g) = c.
The idea is that the Riemann sum of a function with a single discontinuity approaches the
Riemann sum of the same function with the discontinuity removed.
Suppose that a partition T = x
0
, x
1
, ..., x
n
of the interval [a, b] has been given, and
that f is continuous on [a, b]. The Extreme Value Theorem tells us that f attains
maximum and minimum values on any closed subinterval of [a, b].
Let U(T) be a sample set chosen in such a way that, for each s
i
in the sample
set U(T), f(s
i
) equals the maximum value of f on [x
i
, x
i+1
]. Similarly, let L(T) be a
sample set chosen in such a way that, for each s
i
in the sample set L(T), f(s
i
) equals
the minimum value of f on [x
i
, x
i+1
].
45. Show that !
U(P)
P
!
L(P)
P
.
46. Show that if Q is a renement of T, then !
U(Q)
Q
!
U(P)
P
and !
L(Q)
Q
!
L(P)
P
.
47. Show that if T

is another partition, not necessarily a renement of T or vice versa, then

!
L(P)
P
!
U(P

)
P
.
We can use Riemann sums to approximate lengths as well as areas. Suppose wed
like to approximate the length of the graph of a function f(x) on the interval [a, b].
If T = x
0
, x
1
, . . . , x
n
is a partition of the interval [a, b], then we can measure the
lengths of the line segments connecting the points (x
i1
, f(x
i1
)), (x
i
, f(x
i
)). Recall
that this length is given by
_
(x
i
x
i1
)
2
+ (f(x
i
) f(x
i1
))
2
.
48. Let f(x) =

1 x
2
.
a. Consider the partition T
1
= 1, 0, 1 of [1, 1]. Approximate the length of the
graph by calculating
2

i=1
_
(x
i
x
i1
)
2
+ (f(x
i
) f(x
i1
))
2
.
b. Approximate the length of the graph, via the same technique, but using the rened
partition T
2
= 1, 1/2, 0, 1/2, 1.
c. Based on classical geometry, what is the length of this graph? Are your approxima-
tions close?
106 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
In each of Exercises 49 through 51, use the technique in the previous problem to
approximate the lengths of the graphs over the specied interval using the given
partition.
49. g(x) = sin x, [0, ], T = 0, /4, /2, 3/4, .
50. h(x) = x
2
, [0, 1], T = 0, 1/4, 1/2, 3/4, 1.
51. j(x) = 1/x, [1, 5], T = 1, 2, 3, 4, 5.
52. Given a partition x
0
, ..., x
n
of [a, b], weve been using
n

i=1
_
(x
i
x
i1
)
2
+ (f(x
i
) f(x
i1
))
2
to approximate the length of the graph of f on the interval [a, b]. This doesnt look like a
Riemann sum, but we can x it.
Prove that, if f is dierentiable on an open interval containing [a, b], and no two x
i
s are
equal, then there exists s
i
in the open interval (x
i1
, x
i
) such that
n

i=1
_
(x
i
x
i1
)
2
+ (f(x
i
) f(x
i1
))
2
=
n

i=1
_
_
1 + [f

(s
i
)]
2
_
x
i
.
2.3. THE DEFINITE INTEGRAL 107
2.3 The Denite Integral
As we saw in the previous section, the process of taking Riemann sums, and using partitions
with meshes that get arbitrarily small, arises in a number of dierent contexts. We might be
given the velocity of an object over an interval of time, and want to nd the displacement. We
might be given the rate at which a balloon is being inated over an interval of time, and want to
know how much the volume of air in the balloon changes. We might be given a rod of variable
density, and want to know its total mass. We might want to know how much area is trapped
between the graph of a function and the x-axis.
In this section, we dene the denite integral as the limit of Riemann sums, and discuss some
of the basic properties of denite integrals. We do not give many applications of denite integrals
here; such applications will be the sole topic of the entire next chapter, Chapter 3. It is also
somewhat dicult to look at serious applications before we have the basic tool for calculating
denite integrals, the second part of the Fundamental Theorem of Calculus, Theorem 2.4.10.
Many of the proofs in this section are extremely technical and we have deferred them to the
Technical Matters section, Section 2.A.
In Example 2.2.8 and Example 2.2.9 in the previous section, we saw that our Riemann sums

n
i=1
f(s
i
)x
i
approached a limit as we let the mesh of the partition approach zero, but we
used special partitions and sample sets; we used subintervals which all had the same length, and
we used the left and right Riemann sums. What we would like to know is that other manners of
choosing partitions and sample sets would have approximated the same quantities, that is, we
would like to know that we could have used arbitrary partitions and arbitrary sample sets, and
still obtained the same limit, so long as the meshes of the partitions approach zero. Technically,
this means that we would like for there to exist a limit L that our Riemann sums get arbitrarily
close to (within , for arbitrary > 0) if we make the mesh of our partitions small enough (less
than some > 0, where would typically depend on how was chosen).
Thus, we make the denition below, though it should not be clear at this point that
many/any functions satisfy the strong requirements.
108 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
Denition 2.3.1. (The Denite Integral) Suppose that f is dened on the closed interval
[a, b], where a < b, and that there exists a real number L such that, for all > 0, there exists
> 0 such that, for all partitions T = x
0
, . . . , x
n
of [a, b], with mesh less than , and for
all sample sets o = s
1
, . . . , s
n
for T,

i=1
f(s
i
)x
i
L

< .
Then, we write that
lim
|| P ||0
!
S
P
(f) = lim
|| P ||0
n

i=1
f(s
i
)x
i
= L,
and we say that f is Riemann integrable on [a, b].
When f is Riemann integrable, the limit above is usually written
_
b
a
f(x) dx,
and is called the (denite) integral of f on [a, b], or the integral of f(x), with respect
to x, as x goes from a to b.
In this context, f is referred to as the integrand, and a and b are the limits of integration.
Remark 2.3.2. There are several points that we need to make now.
First, you should understand the point of the denition of the integral. A function is Riemann
integrable if and only if the Riemann sums

n
i=1
f(s
i
)x
i
approach a specic limit, regardless
of how you choose the partitions and/or sample points, as long as you take small enough
subintervals in your partitions. You are allowed to use partitions in which the subintervals have
dierent lengths and choose sample points randomly. For f to be Riemann integrable means
that none of this matters, as long as the mesh of the partitions approaches 0.
2.3. THE DEFINITE INTEGRAL 109
Its the strongest property that you ask for, given our discussion in the previous section, and
its a little dicult to believe that many interesting functions satisfy such a strong condition,
but we shall see that most of the functions with which youre familiar are, in fact, Riemann
integrable. (We are deliberately leaving you in suspense for the moment about which functions
those are.)
We should also mention that it is customary in Calculus textbooks to drop the modier
Riemann from the term Riemann integrable, and simply say that a function is integrable,
whenever they mean Riemann integrable. The justication given for this is always that
Riemann integrability is the only type of integrability that will be used in this textbook.
Strangely, this is not usually the case; every, or almost every, Calculus textbook includes a
discussion of improper integrals (see Section 2.5).
Improper integrals specically involve dening an integral
_
b
a
f(x) dx in certain cases where
f is not Riemann integrable on [a, b].
For this reason, we shall not drop the modier Riemann from Riemann integrable. We
shall reserve the simpler term integrable for functions on intervals for which the Riemann
integral exists or for which the improper integral exists.
It is important to note that the units on the integral are the units of the Riemann sums, i.e.,
the units of f times the units of x.
Finally, you should realize that the variable x, which appears in
_
b
a
f(x) dx, is a dummy
variable, which means that this variable name is irrelevant in the determining the actual value
of the integral. For instance, we have the following equalities:
_
b
a
f(x) dx =
_
b
a
f(t) dt =
_
b
a
f(u) du.
All of these integrals mean the same thing: you partition the interval [a, b] into arbitrarily small
subintervals, you evaluate f at a sample point in each subinterval, you multiply the values of f
at the sample points times the lengths of the respective subintervals, you add, and you take the
limit of totals that you get, as the mesh of the partitions approaches zero. The point of writing
all of that, without referring to a variable, was to make it clear that it is irrelevant what the
variable is.
However, it is true that, in applications, there will typically be variables of physical signi-
cance, like t for time or x for position, and it would, in fact, be a little odd to change the variable
name when writing integrals related to these quantities.
110 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
The fact that Denition 2.3.1 denes the Riemann integral as one single thing follows from
the standard type of proof (see Theorem 2.A.1) that limits are unique; thus, there arent two
(or more) dierent limits L
1
and L
2
that satisfy the - condition that we required. We state
this as a theorem.
Theorem 2.3.3. The limit of Riemann sums, the denite integral, given in Denition 2.3.1
is unique, if it exists.
In addition, if f is Riemann integrable on the interval [a, b],
_
b
a
f(x) dx = L, and
(T
n
, o
n
) is a sequence of sampled partitions of [a, b] such that lim
n
[[ T
n
[[ = 0, then
lim
n
!
Sn
Pn
(f) = L.
Well, this is all great, but do we know any functions which are Riemann integrable?
For that matter, do we know any functions which arent Riemann integrable? How you calculate
denite integrals, when they exist, is a separate question, and our best answer to that will have
to wait until Section 2.4, but, for now, wed at least like to know some results on when integrals
exist and when they dont.
There is an easy, fundamental, criterion which implies that a function is not Riemann
integrable. If a function is arbitrarily large in absolute value on an interval, then the function
will not be Riemann integrable. We give a careful denition before stating this result as theorem.
Denition 2.3.4. A set E of real numbers is bounded if and only if there exist real
numbers p and q such that, for all x in E, p x q. Equivalently, E is bounded if and
only if there exists a real number M 0 such that, for all x in E, M x M, i.e.,
[x[ M.
Suppose that f is a real function and that a set E is contained in the domain of f. We say
that f is bounded on E if and only if the set of values of f on E is bounded, i.e., there
exists M 0 such that, for all x in E, [f(x)[ M.
If f is not bounded on E, then we say that f is unbounded on E.
2.3. THE DEFINITE INTEGRAL 111
Example 2.3.5. As a particular case of a bounded set, note that closed, bounded intervals are
intervals of the form [a, b].
As an example of an unbounded function on a closed, bounded interval, consider
f(x) =
_
_
_
0 , if x = 0;
1
x
, if 0 < x 1.
This function is unbounded on [0, 1], since f becomes arbitrarily large as x approaches 0 from
the right.
On the other hand, the Extreme Value Theorem (see [2], or [4]) tells us that every contin-
uous function on a closed, bounded interval is bounded.
The most basic way in which a function can fail to be Riemann integrable is given by:
Theorem 2.3.6. If f is unbounded on the interval [a, b], then f is not Riemann integrable
on [a, b].
Proof. Suppose that f is unbounded on [a, b], and let T = x
0
, . . . , x
n
be a partition of [a, b].
Then, f must be unbounded on at least one of the subintervals [x
i1
, x
i
]. Therefore, by changing
the sample point s
i
in this subinterval, you can make [f(s
i
)x
i
[ arbitrarily large. This means
that, as the mesh of the partitions approaches 0, the Riemann sums do not approach a limit that
is independent of how the sample points are chosen, i.e., the denite integral does not exist.
Example 2.3.7. Theorem 2.3.6 tells us immediately that the function f(x) from Example 2.3.5
is not Riemann integrable.
Alright. Now we have a basic way in which functions can fail to be Riemann integrable. But,
we want a theorem that tells us that lots of functions are Riemann integrable. There is such a
112 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
theorem. Informally, what it says is that a function on a closed, bounded interval is Riemann
integrable if and only if the function is bounded and the set of points where the function is
discontinuous is small. The technical requirement for small is that the set of points where
f is discontinuous has to have measure zero. A discussion of this result, in full generality, is
beyond the scope of the textbook. See [4] and [3].
However, all of the functions that we shall want to integrate in this book will either be
continuous, or have a nite number of discontinuities; the good news is that the empty set and
nite sets of points denitely have measure zero.
Thus, we have:
Theorem 2.3.8. Bounded functions on closed bounded intervals, which have, at most, a
nite number of discontinuities, are Riemann integrable.
In particular, as continuous functions on closed, bounded intervals are bounded, all con-
tinuous functions on closed, bounded intervals are Riemann integrable.
Proof. We give the proof in Theorem 2.A.5.
It is convenient to give a name to functions which may have, at most, a nite number of
discontinuities.
Denition 2.3.9. A real function f is piecewise-continuous on an interval I provided
that f is dened on I and is continuous at all, except (possibly) a nite number of, points
in I. In particular, a continuous function is also piecewise-continuous.
Example 2.3.10. The function
g(x) =
_
_
_
0 , if 0 x 0.1;
1
x
, if 0.1 < x 1
is piecewise-continuous and, hence, Theorem 2.3.8 tells that g(x) is Riemann integrable.
2.3. THE DEFINITE INTEGRAL 113
It is interesting to compare this with the function f(x) from Example 2.3.5:
f(x) =
_
_
_
0 , if x = 0;
1
x
, if 0 < x 1.
0 0.5 1
Figure 2.13: Graph of y = f(x).
0 0.5 1
Figure 2.14: Graph of y = g(x).
While the graphs of f and g may appear to be similar (or maybe they dont to you), it is
important to keep in mind that unbounded functions are very dierent from bounded piecewise-
continuous functions.
Before stating more theorems, we should translate our discussion about area in Example 2.2.9
into a proposition about denite integrals, a proposition which tells us how to visualize integrals
in terms of area. This proposition could, instead, be used as a denition for area below or
above graphs of certain types of functions; we choose to believe that you have a preconceived
notion of area, and that the following proposition tells you rigorously what that area equals in
terms of limits of Riemann sums, i.e., in terms of denite integrals.
114 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
Proposition 2.3.11. Suppose that a < b, and f is Riemann integrable on [a, b].
1. If f 0 on [a, b], then
_
b
a
f(x) dx is equal to the area under the graph of y = f(x) and
above the interval [a, b].
2. If f 0 on [a, b], then
_
b
a
f(x) dx =
_
b
a
[f(x)] dx is equal to the area under the
interval [a, b] and above the graph of y = f(x); thus,
_
b
a
f(x) dx is equal to negative
the area under the interval [a, b] and above the graph of y = f(x).
y=f(x)
Figure 2.15: Integral of a positive function is
area under the graph.
y=f(x)
Figure 2.16: Integral of a negative function is
negative the area above the graph.
Be careful: the area under the x-axis and above the graph is NOT, itself, negative; its
positive, as area always is. The point is that, to interpret the integral of a negative function
in terms of area, you negate the positive area above the graph, which, of course, yields a
negative number.
A special case of Proposition 2.3.11 is when y = f(x) = k, where k is a constant.
Proposition 2.3.12. Suppose that a < b, and k is a constant. Then, the function f(x) = k
is Riemann integrable and
_
b
a
k dx = k(b a).
2.3. THE DEFINITE INTEGRAL 115
Proof. Of course, constant functions are continuous and, hence, Riemann integrable by Theo-
rem 2.3.8. However, our proof of the formula will also prove that constant functions are Riemann
integrable.
We shall simply show that every Riemann sum of f(x) = k, regardless of the partition T
and the sample set o, has !
S
P
(f) equal to precisely k(b a). Thus, L = k(b a) clearly satises
the conditions for L in Denition 2.3.1.
Let T = x
0
, x
1
, . . . , x
n
be a partition of [a, b], and let o = s
1
, . . . , s
n
be a sample set for
T. Then,
!
S
P
(f) =
n

i=1
f(s
i
)x
i
=
n

i=1
kx
i
= k
n

i=1
x
i
,
and this last summation telescopes to yield x
n
x
0
= b a.
y=k>0
a
b 0
Area = k(b-a)
Figure 2.17: Area under y = k > 0.
y=k<0
a b 0
Area = (-k)(b-a)
Figure 2.18: Area above y = k < 0.
Remark 2.3.13. The denition of the denite integral looks very technical. Thinking in terms
of area can help you to see things in some cases, but many physical problems, which dont involve
areas, deal with subdividing, taking Riemann sums, and taking limits. So, more generally, you
may be wondering how you should think about
_
b
a
f(x) dx = lim
|| P ||0
!
S
P
(f) = lim
|| P ||0
n

i=1
f(s
i
)x
i
.
116 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
One possible way to think about denite integrals a way that we shall use extensively through-
out this textbook is informal, but is, nonetheless, extremely useful. We discuss the problems
innitesimally.
If we take a partition with mesh less than some positive , then all of the x
i
s are less than
, and any x-coordinate in the interval [a, b] is within of one of the sample points. Thus, as
the meshes approach zero, the sample points get arbitrarily close to each point in the interval
[a, b], and the subinterval(s) containing a given point become(s) arbitrarily small.
Thus, we think, intuitively, that, if we select an x in [a, b], as the mesh of the partitions
approaches 0, and we look at

n
i=1
f(s
i
)x
i
, in the limit, one of the summands becomes f(x)
times an innitesimal change in x, represented by dx. That is, in the limit, we think that, for
each x in the interval [a, b], one of the f(s
i
)x
i
approaches the innitesimal summand f(x) dx,
and the summation

n
i=1
becomes the continuous summation
_
b
a
.
While we always keep in the backs of our minds that we are really using partitions, sample
points, Riemann sums, and taking limits, in many physical applications it is very intuitive, and
time-saving, to think of the denite integral as the continuous sum of innitesimal
contributions, and to analyze problems using this informal terminology.
For instance, we arrived at Proposition 2.3.11 by looking back at our discussion in Exam-
ple 2.2.9 and using the denition of the denite integral in Denition 2.3.1. Thus, we consider
area as a limit of Riemann sums. But now, lets discuss area in terms of continuous sums of
innitesimal contributions. Hopefully, this will seem more intuitive to you, but, keep in mind,
that the real denition of the integral is the limit of Riemann sums.
Consider a function f(x), dened on the closed interval [a, b], where a < b, and, for now,
assume that f(x) 0, for all x in [a, b]. If we assume that f is Riemann integrable on [a, b],
how do we picture
_
b
a
f(x) dx?
x
y
x
dx
f ( ) x
a
b
}
x
y
x
dx
f ( ) x
a
b
}
x
y
x
dx
f ( ) x
a
b
Figure 2.19: A rectangle with innitesimal
width and area.
x
y
f ( ) x
a
b
x
dx
}
dx
E
n
la
rg
e
d
Figure 2.20: A magnied innitesimal rectan-
gle.
2.3. THE DEFINITE INTEGRAL 117
Pick some x-coordinate between a and b, and draw a rectangle of height f(x) and a very
small width, which we think of as the innitesimal dx; the innitesimal subinterval of width
dx should include x. See Figure 2.19 and Figure 2.20, which show the type of thin rectangle
that you might draw for yourself, in order to represent a rectangle of innitesimal width and
area, together with a fancier illustration of a really (think: innitesimally) thin rectangle
being magnied. Intuitively, we think the only x-coordinate in the innitesimal subinterval is
x. Why? Because if there were some other x-coordinate, say x, in the subinterval, then the
subinterval would have to have length equal to at least [x x[, and so would not be innitesimal.
Our innitesimally wide rectangle has innitesimal area dA, which equals the height times the
width, i.e., dA = f(x)dx.
Now, to obtain the total area under the graph of f and above the interval [a, b] on the x-axis,
we simply take the continuous sum of all of the innitesimal areas as x goes from a to b. Thus,
the total area is
_
x=b
x=a
dA =
_
b
a
f(x) dx.
Note that, on the rst integral above, we had to include x = in the limits of integration;
by default, the limits of integration refer to the variable that you have d of, that is, the
variable in the dierential. So, if we had not included x = in the limits of integration in
_
x=b
x=a
dA, then it would have meant that A was going from a to b, not x. As we have dx
in the righthand integral above, we do not need to explicitly state in that integral that the
limits of integration refer to x.
Understand we are not claiming that we have given actual denitions that make what we
have written mean anything rigorous; we are merely trying to get you used to this manner of
intuitive thinking, in terms of continuous sums of innitesimal contributions.
Of course, this is precisely what we concluded in Example 2.2.9, except there we didnt call
it the integral; we simply said the area under the graph was the limit of the Riemann sums.
Thats what we are saying here too, its just that we have new terminology, the denite integral,
together with our new notation for the integral, and instead of explicitly adding up Riemann
sums, we talk about taking a continuous sum of innitesimal contributions.
What do we do when f(x) 0 (and is still Riemann integrable)? We draw a picture like
that in Figure 2.19, except that now the innitesimal rectangle lies under the x-axis and above
the graph; thus, the rectangle looks like that in Figure 2.11, except that now we think of the
width as the innitesimal dx and the height is f(x). Therefore, the total area A between the
x-axis and the graph of y = f(x) is
A =
_
x=b
x=a
dA =
_
b
a
f(x) dx =
_
b
a
f(x) dx,
118 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
and, hence,
_
b
a
f(x) dx = A. (Note that we cheated a bit here, and used that we could pull
out the negative sign in the integral; we wont really know that this is legal until we have
Theorem 2.3.19.)
Finally, what do we do when f is Riemann integrable, and f(x) 0 on some (nite number
of) closed subintervals in [a, b], and f(x) 0 on (a nite number of) other closed subintervals
in [a, b]? In this case, as we discussed at the end of Example 2.2.9,
_
b
a
f(x) dx is a sum of
the contributions from where f 0 and where f 0, which means that
_
b
a
f(x) dx can be
interpreted as the area under the graph of y = f(x) and above the interval [a, b] minus the area
above the graph and under the interval [a, b]. (Again, we cheated a bit here, and used that we
could split the interval over [a, b] into the sum of integrals over subintervals whose union is [a, b];
we wont really know that this is legal until we have Theorem 2.3.16 and/or Theorem 2.3.18.)
We now need to state a number of theorem about denite integrals. Despite the fact that
all of the examples of Riemann integrable functions that we shall look at will be bounded,
piecewise-continuous functions, the remaining theorems of this section are true for completely
general Riemann integrable functions, and its just as easy to state and prove them in that
generality; so we will.
The following theorem says that, as far as integration is concerned, you can ignore what
happens at a nite set of points in the domain. We give the proof in the Technical Matters
section; see Theorem 2.A.8.
Theorem 2.3.14. Suppose that f and g are dened on a closed interval [a, b], and that,
except possibly for a nite set points in [a, b], f and g are equal at each point in [a, b].
Then, f is Riemann integrable on [a, b] if and only if g is, and when f and g are Riemann
integrable,
_
b
a
f(x) dx =
_
b
a
g(x) dx.
Example 2.3.15. The function given by
f(x) =
_
7 , if 1 x < 3 or 3 < x 5;
4 , if x = 3
2.3. THE DEFINITE INTEGRAL 119
is equal to the function that is constantly 7 on the interval [1, 5], except at the point x = 3.
Thus, by Theorem 2.3.14
_
5
1
f(x) dx =
_
5
1
7 dx = 7(5 1) = 28,
where the next-to-last equality follows from Proposition 2.3.12.
In terms of area, it seems reasonable that what happens at a nite number of points shouldnt
aect the integral, since the area under (or over) a single point and above (or below, respectively)
the x-axis should be zero, and so does not change the area being considered. See Figure 2.21.
-1 0 1 2 3 4 5 6
-1
1
2
3
4
5
6
7
8
Figure 2.21: Area under a point is zero.
The following theorem is about subdividing, or splitting, the interval over which youre
integrating. Thinking in terms of continuous sums, Item 1 of the theorem says that, if the
continuous sum on a bigger interval is dened, then so is the continuous sum on any smaller
subinterval; Item 2 simply says that the sum of the innitesimal contributions over all x in
the interval [a, b] is equal to sum of those contributions as x goes from a to some intermediate
x-coordinate c plus the sum of the contributions as x goes on from c to b.
120 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
Theorem 2.3.16. Suppose that a < b.
1. If f is Riemann integrable on [a, b] and a c < d b, then f is Riemann integrable on
[c, d], i.e., if f is Riemann integrable on a given closed interval, then f is Riemann integrable
on any closed subinterval of the given interval.
2. Suppose that a < c < b. Then, f is Riemann integrable on [a, b] if and only if f is
Riemann integrable on [a, c] and [c, b] and, when these equivalent conditions hold,
_
b
a
f(x) dx =
_
c
a
f(x) dx +
_
b
c
f(x) dx.
Proof. See Theorem 2.A.6 in the Technical Matters section, Section 2.A.
For a non-negative, Riemann integrable function f, Item 2 of Theorem 2.3.16 is easy to
picture in terms of area; see Figure 2.22.
a b
0
c
y=f(x)
Figure 2.22: Area over [a, b] equals area over [a, c] plus area over [c, b].
You may wonder about counting the line segment at x = c once in
_
b
a
f(x) dx versus twice
in the sum of integrals from a to c and from c to b, but, remember: the area of a line segment
is 0, and adding it or subtracting it has no aect on the area calculation.
2.3. THE DEFINITE INTEGRAL 121
The formula in Item 2 of Theorem 2.3.16 is so useful that we would like for it to be true
regardless of what inequalities (or equalities) are satised by a, b, and c. For instance, we would
like for the formula to hold if a = c. This means that we must have
_
b
a
f(x) dx =
_
a
a
f(x) dx +
_
b
a
f(x) dx,
and so, we would need for
_
a
a
f(x) dx to equal 0.
Also, if a = b, we should have
_
a
a
f(x) dx =
_
c
a
f(x) dx +
_
a
c
f(x) dx.
If
_
a
a
f(x) dx = 0, then we need for
_
a
c
f(x) dx to equal
_
c
a
f(x) dx.
Therefore, we make the following denitions:
Denition 2.3.17. If a is in the domain of f, we say that f is Riemann integrable on the
interval [a, a] and dene
_
a
a
f(x) dx = 0.
If a < b, and f is Riemann integrable on [a, b], then we dene
_
a
b
f(x) dx =
_
b
a
f(x) dx.
With the denitions above, and given Theorem 2.3.16, it is easy to check that
Theorem 2.3.18. Suppose that f is Riemann integrable on a closed interval containing a,
b, and c, then
_
b
a
f(x) dx =
_
c
a
f(x) dx +
_
b
c
f(x) dx,
regardless of the order of a, b, and c.
122 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
The following two theorems describe fundamental properties of integration, and are true
essentially because the corresponding results for Riemann sums are true.
Theorem 2.3.19. (Linearity of Integration) Denite integration over a closed interval
is a linear operation, i.e., if f and g are Riemann integrable on [a, b], then, for all constants
r and s, the function rf +sg is Riemann integrable on [a, b], and
_
b
a
_
rf(x) +sg(x)
_
dx = r
_
b
a
f(x) dx + s
_
b
a
g(x) dx.
Proof. This proof actually splits into two distinct parts; that the integral of a sum is the sum
of the integrals, and that you can move constants, multiplied in the integrand, outside of the
integral.
The statement about constant multiples is easy, and follows from the fact that constants
distribute over Riemann sums; we leave this part of the proof as an exercise, Exercise 52. The
fact that the integral of a sum is the sum of the integrals is harder; we prove this in Theorem 2.A.7
in the Technical Matters section, Section 2.A.
If you think of integrals in terms of areas, the following theorem seems fairly obvious; see
Figure 2.23.
Theorem 2.3.20. (Monotonicity of Integration) If f and g are Riemann integrable on
the interval [a, b] and, for all x in [a, b], f(x) g(x), then
_
b
a
f(x) dx
_
b
a
g(x) dx.
Proof. The inequality f(x) g(x) implies that the corresponding inequality holds on the Rie-
mann sums for each sampled partition. As we are assuming that f and g are Riemann integrable,
it is easy to conclude that the integral of f on [a, b] is less than or equal to the integral of g. We
give the details in Theorem 2.A.9.
2.3. THE DEFINITE INTEGRAL 123
New A1
x
y
a
b
f
(
)
x
g
(
)
x
Figure 2.23: Comparing areas under the two graphs.
Theorem 2.3.21. If f is Riemann integrable on the interval [a, b], then so is [f[, and,

_
b
a
f(x) dx

_
b
a
[f(x)[ dx.
Proof. The proof that f being Riemann integrable implies that [f[ is Riemann integrable is very
technical; we refer you to Theorem 3.3.5 of [4]. Assuming this fact, the given inequality follows
as a corollary to Theorem 2.3.20 and Theorem 2.3.19.
To see this, note that [f(x)[ f(x) [f(x)[. Applying Theorem 2.3.20 and Theo-
rem 2.3.19, we obtain that [f[ is Riemann integrable on [a, b] and

_
b
a
[f(x)[ dx
_
b
a
f(x) dx
_
b
a
[f(x)[ dx.
The inequality in the theorem is now immediate.
Now, consider the continuous function f(x) = x on a closed interval [a, b]. Theorem 2.3.8
tells us that
_
b
a
xdx
exists. Given that the integral exists, Theorem 2.3.3 tells us that we may calculate the value
of the integral by using the limit of Riemann sums over any sequence of sampled partitions, as
124 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
long as the meshes of the partitions approach zero.
So, we will let T
n
be the partition of [a, b] into n subintervals of equal length, i.e., letting
x = (b a)/n,
T
n
= a, a + x, a + 2x, . . . , a + (n 1)x, a +nx,
where x
i
= a +ix and, of course, b = a +nx.
We will use the right Riemann sums, so that our sample points are also given by s
i
= a+ix.
Lets denote the corresponding Riemann sum by simply !
n
. For f(x) = x, we obtain
!
n
=
n

i=1
f(s
i
)x
i
=
n

i=1
[(a +ix) x] .
Using linearity of summations, Proposition 2.1.4, and Item a from Corollary 2.1.11, we nd
!
n
= ax
n

i=1
1 + (x)
2
n

i=1
i = anx + (x)
2

n(n + 1)
2
=
a(b a) +
(b a)
2
2

n
2
+n
n
2
= a(b a) +
(b a)
2
2

_
1 +
1
n
_
.
Therefore, we nd
_
b
a
xdx = lim
n
!
n
= lim
n
_
a(b a) +
(b a)
2
2

_
1 +
1
n
__
=
a(b a) +
(b a)
2
2
= (b a)
_
a +
b a
2
_
= (b a)
b +a
2
=
b
2
a
2
2
.
Hence, we have shown
Proposition 2.3.22.
_
b
a
xdx =
b
2
2

a
2
2
.
2.3. THE DEFINITE INTEGRAL 125
Note that, even though we assumed a < b in our proof of Proposition 2.3.22, it follows from
Denition 2.3.17 that the proposition above remains true when b a.
Example 2.3.23. Using the area interpretation of the integral, it is easy to derive Proposi-
tion 2.3.22 geometrically.
First, suppose that b > 0. Then, the integral
_
b
0
xdx equals the area under the graph of
y = x and above the interval [0, b] on the x-axis. This is the area of a triangle of width b and
height b.
b
b
Figure 2.24: Area under y = x.
Thus,
_
b
0
xdx =
b
2
2
.
This formula holds even if b < 0 for, in that case,
_
b
0
xdx =
_
0
b
xdx,
and
_
0
b
xdx is equal to negative the area below the interval [b, 0] and above the graph of y = x;
this area is again b
2
/2.
126 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
b
b
Figure 2.25: Area above y = x.
Hence,
_
b
0
xdx =
_
0
b
xdx =
_

b
2
2
_
=
b
2
2
.
Of course, it follows that
_
a
0
xdx =
a
2
2
,
and so, using Theorem 2.3.18, we nd
_
b
a
xdx =
_
0
a
xdx +
_
b
0
xdx =
_
a
0
xdx +
_
b
0
xdx =
a
2
2
+
b
2
2
,
which agrees with what we obtained in Proposition 2.3.22.
Example 2.3.24. Now that we have Proposition 2.3.22 and Proposition 2.3.12, we can combine
them with linearity to calculate
_
b
a
(mx +c) dx,
where m and c are constants.
We nd
_
b
a
(mx +c) dx = m
_
b
a
xdx +
_
b
a
c dx = m
_
b
2
a
2
2
_
+ c(b a).
2.3. THE DEFINITE INTEGRAL 127
Example 2.3.25. Consider the integral
_
6
3
[x[ dx.
-7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6
-5
-4
-3
-2
-1
1
2
3
4
5
6
7
8
9
Figure 2.26: Area under y = [x[.
By Theorem 2.3.16, we have
_
6
3
[x[ dx =
_
0
3
[x[ dx +
_
6
0
[x[ dx.
Why split up the integral like this? Because [x[ = x, if x 0, and [x[ = x, if x 0. Thus,
[x[ = x on the interval [3, 0], and [x[ = x on the interval [0, 6]. Therefore, we obtain
_
6
3
[x[ dx =
_
0
3
[x[ dx +
_
6
0
[x[ dx =
_
0
3
xdx +
_
6
0
xdx =

_
0
3
xdx +
_
6
0
xdx =
_
0
2
(3)
2
2
_
+
_
6
2
0
2
2
_
=
9
2
+
36
2
=
45
2
.
Note that this agrees with what you would get from interpreting the integral in terms of
area; see Figure 2.26.
128 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
Example 2.3.26. The cosine function is continuous and, for all x, 1 cos x 1. If we want
bounds on the integral of cos x, we can apply Theorem 2.3.20 twice.
For instance, even though we do not yet know how to calculate
_
/2
0
cos xdx, we can conclude
that

2
=
_
/2
0
1 dx
_
/2
0
cos xdx
_
/2
0
1 dx =

2
.
In fact, as we shall see later,
_
/2
0
cos xdx = 1, which is, indeed, between /2 and /2.
Proposition 2.3.27.
_
b
a
x
2
dx =
b
3
3

a
3
3
.
Proof. We will prove that, if b > 0, then
_
b
0
x
2
dx = b
3
/3. The case where b < 0 is similar, and
we leave it as an exercise. Once you know that
_
b
0
x
2
dx = b
3
/3, the proposition follows from
applying Theorem 2.3.18:
_
b
a
x
2
dx =
_
0
a
x
2
dx +
_
b
0
x
2
dx =
_
a
0
x
2
dx +
_
b
0
x
2
dx =

a
3
3
+
b
3
3
.
So, assume that b > 0. How do we show that
_
b
0
x
2
dx = b
3
/3? As x
2
is continuous, we know
that the integral exists, and so we may calculate it by taking a limit of Riemann sums, in which
the meshes of our partitions approach 0. This was how we derived Proposition 2.3.22.
Let T
n
be the partition of [0, b] into n subintervals of equal length, i.e., the partition in which
x
i
= ib/n, and x
i
= b/n. If we once again use right Riemann sums, our sample set R
n
is the
one in which s
i
also equals ib/n.
Our corresponding Riemann sum for x
2
is
!
n
=
n

i=1
_
ib
n
_
2

b
n
=
b
3
n
3
n

i=1
i
2
.
2.3. THE DEFINITE INTEGRAL 129
By Item (b) of Corollary 2.1.11, we know that
n

i=1
i
2
=
n(n + 1)(2n + 1)
6
,
and so
!
n
=
b
3
n
3

n(n + 1)(2n + 1)
6
=
b
3
6

_
1 +
1
n
__
2 +
1
n
_
.
Therefore, as n , !
n
(b
3
/6)(2) = b
3
/3 and, hence,
_
b
0
x
2
dx = b
3
/3, as we wanted to
show.
Example 2.3.28. The area interpretation of the denite integral lets us calculate the denite
integral of f(x) in cases where the region between the x-axis and the graph of y = f(x) consists
of rectangles, triangles, and trapezoids.
r
r
Figure 2.27: Area under y =

r
2
x
2
.
Of course, we also know that the area inside a circle of radius r is r
2
, which enables us to
calculate integrals like:
_
r
0
_
r
2
x
2
dx,
where r > 0 is a constant. This integral represents the area inside the rst quadrant quarter of
a circle of radius r, centered at the origin; see Figure 2.27.
Thus,
_
r
0
_
r
2
x
2
dx =
r
2
4
.
130 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
Example 2.3.29. Lets return to the rod of varying density in Example 2.2.8, and discuss the
problem of calculating its total mass, but now we shall use our continuous sum of innitesimal
contributions language.
A circular rod, of length 1 meter, and cross-sectional area 0.01 m
2
(i.e., of radius 0.1/

meters) is lying along the x-axis.

Figure 2.28: A rod of varying density.
For all x such that 0 x 1, at each point in the cross section of the rod at x meters, the
density of the rod is (x) = (1 +x) kg/m
3
. What is the total mass of the rod?
Solution:
We shall present the solution using the language of innitesimals. Consider the innitesimal
amount of volume dV lying along the innitesimal subinterval at x of innitesimal length dx.
The innitesimal volume dV is just equal to the cross-sectional area times the length dx, i.e.,
dV = 0.01 dx m
3
,
and the innitesimal mass dM of this innitesimal volume is simply the density at x times the
(innitesimal) volume dV , i.e.,
dM = (x) dV = (1 +x)0.01 dx kg.
Finally, the total mass M of the rod is the continuous sum of all of the innitesimal masses as
x goes from 0 to 1 meter, i.e.,
M =
_
x=1
x=0
dM =
_
1
0
(1 +x)0.01 dx kg.
2.3. THE DEFINITE INTEGRAL 131
Suppose that a < b. We now want to develop a reasonable notion of the mean, or average,
value of f on the interval [a, b].
What should the average value of f mean? Well...if there were a nite number of f-values,
we would simply add up all of the values and then divided by the number of values; this is what
is normally meant by the average or mean. The problem, of course, is that we have an
innite number of values f(x), one for each x in the interval [a, b]. How do we deal with this?
As you may have guessed, we take the average of a nite number of f-values, and then take
a limit. Let n 1 be an integer. Subdivide the interval [a, b] into n subintervals of equal length
x = (b a)/n. As usual, in the i-th subinterval, we pick a sample point s
i
. We may now take
the average value of f at the n sample points; we obtain
f(s
1
) +f(s
2
) + +f(s
n
)
n
=
n

i=1
f(s
i
)
n
.
So, what would be a reasonable notion of the average value of f on the entire interval? We
can try taking the limit of the average above, as n approaches , but why would this limit
exist, and what does it have to do with integration?
Things become more clear if we multiply the numerator and denominator of the above
fractions by x. Then, we obtain
n

i=1
f(s
i
)x
nx
=
n

i=1
f(s
i
)x
b a
.
If f is Riemann integrable on [a, b], then, as n , this last numerator approaches
_
b
a
f(x) dx,
and so, the limit of the average value of f at the sample points approaches
1
b a
_
b
a
f(x) dx.
Therefore, we make the following denition:
132 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
Denition 2.3.30. Suppose that a < b and that f is Riemann integrable on [a, b]. Then,
we dene the mean value, or average value, of f on [a, b] to be
1
b a
_
b
a
f(x) dx.
For continuous functions, we have the following theorem about the mean value of f.
Theorem 2.3.31. (Mean Value Theorem for Integration) Suppose that f is continu-
ous on the closed interval [a, b]. Then, there exists c in [a, b] such that
_
b
a
f(x) dx = (b a)f(c).
Thus, if a < b, there exists c in the interval [a, b] such that f(c) equals the mean value
of f on [a, b], i.e., f attains its mean value on [a, b] at some point in [a, b].
Proof. If a = b, then the theorem is obviously true. So, assume that a < b.
As f is continuous on [a, b], the Extreme Value Theorem (see [2]), tells us that f attains a
global minimum value m on [a, b] and a global maximum value M on [a, b].
Thus, for all x in [a, b], m f(x) M. By Theorem 2.3.20, this implies that
m(b a) =
_
b
a
mdx
_
b
a
f(x) dx
_
b
a
M dx = M(b a).
Since a < b, b a ,= 0, and so we may divide to obtain
m
1
b a
_
b
a
f(x) dx M.
Now, the function f is continuous on [a, b], attains the values m and M, and
1
ba
_
b
a
f(x) dx
is a value between m and M. The Intermediate Value Theorem (see [2]), implies that there
2.3. THE DEFINITE INTEGRAL 133
exists c in [a, b] such that
f(c) =
1
b a
_
b
a
f(x) dx.
Multiply each side of the above equality by b a to obtain the theorem.
Remark 2.3.32. The area interpretation of integration helps us visualize the mean value of a
function on an interval. For instance, if f 0 and continuous on the interval [a, b], then the
mean value of f on [a, b] is the height of a rectangle with base [a, b] such that the rectangle has
the same area as that below the graph of f and above the interval [a, b].
For instance, as we shall see in Example 2.4.13, the mean value of y = f(x) = 1/(1 +x
2
) on
the interval [1, 1] is /4.
-1 -0.5 0 0.5 1
-0.25
0.25
0.5
0.75
1
1.25
1.5
y=!/4
y=1/(1+x )
2
y=1/(1+x )
2
Figure 2.29: Equal areas under y = 1/(1 +x
2
) and y = /4.
The Mean Value Theorem for Integration guarantees that, since f(x) = 1/(1 + x
2
) is con-
tinuous, there exists at least one x value in [1, 1] at which f attains its mean value. In fact,
we see in Figure 2.31 that there are two such x values, approximately 0.5. We calculate these
values precisely in Example 2.4.13.
2.3.1 Exercises
Calculate the denite integrals.
134 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
1.
_
7
3
5 dx.
2.
_
6
3
5t 7 dt.
3.
_
8
4
[u 4[ + 2 du.
4.
_
7
3
2[4 8z[ dz.
5.
_
12
0
5w
2
dw.
6.
_
7
3
(v + 3) dv.
7.
_
3
2
x
2
+xdx.
8.
_
6
4
f(x) dx, where f(x) =
_
3 x < 1
5 x 1.
9.
_
4
4
e
x
2
dx.
Answer the following true/false questions. If the statement is true, why is it true?
If the statement is false, provide a counterexample.
10. If f is dierentiable on an open interval containing [a, b], then
_
b
a
f(x) dx exists.
11. If
_
b
a
f(x) dx
_
b
a
g(x) dx, then f(x) g(x) for all x [a, b].
12. If f is integrable, then f is continuous.
13.
_
b
a
f(x) dx =
_
a
b
f(x) dx for all integrable functions f.
In each of exercises 14 through 18, calculate the average value of the function over
the interval.
14. c(x) = 10, [11, 9].
15. g(x) = 12x 10, [4, 9].
16. h(x) =
x
2
6
, [0, 12].
2.3. THE DEFINITE INTEGRAL 135
17. m(t) = [t[ + 3, [3, 6].
18. p(x) =
_
2x x ,= 1
0 x = 1
, [2, 4].
19. Recall that the height of a projectile with initial velocity v
0
and h
0
is
h(t) =
1
2
gt
2
+v
0
t +h
0
.
Calculate the average height between times t = 0 and t = t
1
.
20. Prove that the function f(x) =
_
0, if x is rational;
1, if x is irrational
is not integrable.
21. Construct an example of a function that is integrable but that does not satisfy the mean
value property. That is, construct a function that does not achieve its average value over
its domain.
In each of Exercises 22 through 25, you are given the velocity function of a parti-
cle, in meters per second. Calculate the average velocity of the function over the
specied interval.
22. v(t) = 3t + 2, [4, 8].
23. v(t) = t
2
5t + 3, [2, 6].
24. v(t) = t
3
2t
2
+ 3t + 2, [1, 3].
25. v(t) = [t 4[, [2, 6].
In Exercises 26 - 29, you are given the same velocity function and interval as in the
previous four problems. For each problem, calculate (a) the acceleration function,
and (b) the average acceleration over the interval.
26. v(t) = 3t + 2, [4, 8].
27. v(t) = t
2
5t + 3, [2, 6].
28. v(t) = t
3
2t
2
+ 3t + 2, [1, 3].
29. v(t) = [t 4[, [2, 6].
30. Suppose p
1
(t) and p
2
(t) are the position functions of two particles and that p
2
(t) = p
1
(t)+C
where C is some constant. The position functions have common domain [a, b].
136 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
a. What is the relationship between the average positions of the two particles?
b. What is the relationship between the average velocities of the two particles?
c. What is the relationship between the average accelerations of the two particles?
31. Suppose that v
1
(t) and v
2
(t) are the velocity functions of two particles and that v
2
(t) =
v
1
(t) +C where C is some constant. The velocity functions have common domain [a, b].
a. What is the relationship between the average velocities of the two particles?
b. What is the relationship between the average accelerations of the two particles?
32. Is the function below integrable on [, ]?
f(x) =
_
sin(1/x) x ,= 0
0 x = 0.
33. Is the function below integrable on [, ]?
g(x) =
_
xsin(1/x) x ,= 0
0 x = 0.
Estimate the denite integrals using the given partitions using the left endpoint
sample set.
34.
_
1
0
sin(x) dx, T = 0, 1/4, 1/2, 3/4, 1.
35.
_
1
0

xdx, T = 0, 1/4, 1/3, 3/4, 1.

36.
_
1
1
1
1 +x
2
dx, T = 1, 1/2, 0, 1/2, 1.
37.
_
1
1
cosh xdx, T = 1, 1/2, 0, 1/2, 1.
Calculate the denite integrals below by determining the area enclosed between
their graphs and the x-axis. Remember that area below the x-axis is counted with
a minus sign in the integral.
38.
_
7
4
2x 3 dx.
2.3. THE DEFINITE INTEGRAL 137
39.
_
5
5
_
25 x
2
+ 4 dx.
40.
_
6
2
[x 3[ + 5 dx.
The idea of using integrals to calculate the area between a curve and the x-axis can
be generalized. To calculate the area of the region between the graphs of f(x) and
g(x), we just need to notice that if f > g 0, then the area between the graphs
of f and g is the area bounded by the graph of f and the x-axis minus the area
bounded by the graph of g and the x-axis. By the linearity of the integral, this area
is given by
_
b
a
_
f(x) g(x)
_
dx. In each of Exercises 41 through 44, calculate the area
bounded by the two graphs along the interval.
41. f(x) = 4x, g(x) = 2, [8, 12].
42. f(x) = 5x, g(x) = 3x, [0, 10].
43. f(x) = x, g(x) = x
2
, [0, 1].
44. f(x) =

25 x
2
, g(x) = 3, [4, 4].
45. Consider a triangle contained entirely in the rst quadrant with vertices P = (x
1
, y
1
),
Q = (x
2
, y
2
) and R = (x
2
, y
3
). Assume further that x
1
< x
2
and that y
3
> y
2
.
a. Show that the equations of the lines between P and R, and P and Q are
f(x) =
y
3
y
1
x
2
x
1
x
x
2
(y
3
y
1
)
x
2
x
1
+y
3
g(x) =
y
2
y
1
x
2
x
1
x
x
2
(y
2
y
1
)
x
2
x
1
+y
2
respectively.
b. Calculate
_
x2
x1
f(x) g(x) dx and conclude that the area of the triangle is
1
2
[x
2
y
3
x
2
y
2
+x
1
y
2
x
1
y
3
] =
1
2
(x
1
x
2
)(y
2
y
3
).
46. Suppose a 10 meter long rod lies along the x-axis The rod is a rectangular prism with
cross-sectional area 2 m
2
. The density of the rod is given by x = (x
2
+ 1) kg / m
3
.
Express the total mass of the rod in terms of a denite integral. You need not calculate
the integral.
138 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
47. Suppose a 10 meter metallic cone lies along the x-axis with its vertex at the origin. More
specically, the projection of the cone onto the xy plane is dened by the line segments
y = x and y = x for x [0, 10]. The density at each point in the cross section of the rod
is given by (x) = (12x) kg/m
3
. Express the total mass of the cone in terms of a denite
integral. You need not calculate the integral. Hint: in this case, dA, the innitesimal area,
is no longer constant.
Recall that a parameterized planar curve with domain [a, b] is a map (t) into the
xy-plane. That is, (t) = (x(t), y(t)). Denite integrals can be used to calculate the
length of curves. Specically, if (t) is a dierentiable curve on [a, b], the length of
the curve is
_
b
a
[

(t)[ dt. Calculate the lengths of the curves in Exercises 48 - 51.

48. (t) = (t, 3), [2, 4].
49. (t) = (t, 4t), [0, 5].
50. (t) = (t, mt), [a, b].
51. (t) = (Rcos t, Rsin t), [0, ], where R 0.
52. Prove the constant multiple portion of Theorem 2.3.19, i.e., prove that if f is Riemann
integrable on the closed interval [a, b] and c is a constant, then cf is Riemann integrable
on [a, b], and
_
b
a
cf(x) dx = c
_
b
a
f(x) dx.
2.4. THE FUNDAMENTAL THEOREM OF CALCULUS 139
2.4 The Fundamental Theorem of Calculus
The denite integral has many applications, all of which stem from the fact that the denition of
the denite integral as a limit of Riemann sums makes the denite integral a continuous sum of
innitesimal contributions. However, actually calculating denite integrals by taking limits of
Riemann sums is extraordinarily painful, and, for denite integrals to really be manageable, we
need an easy way to calculate these limits of Riemann sums that come up in so many physical
situations.
The second part of the Fundamental Theorem of Calculus, Theorem 2.4.10, provides this
easy method; it tells us that we may calculate denite integrals by anti-dierentiating. This
theorem is truly fundamental, for it tells us that the two basics ideas of Calculus, dierentiating
and integrating, are inextricably related.
In Proposition 2.3.12, Proposition 2.3.22, and Proposition 2.3.27, we gave formulas for the
denite integrals of f(x) = k, g(x) = x, and h(x) = x
2
over the interval [a, b]. Those formulas
may not have had an obvious pattern, but it may become clear whats going on after we introduce
some new notation.
Denition 2.4.1. If F(x) is a real function, dened at x = a and x = b, then we dene
the evaluation notation,

b
a
by
F(x)

b
a
= F(b) F(a).
We also write
F(x)

x=b
x=a
= F(b) F(a),
if we want to emphasize the name of the variable that takes on the values a and b.
We read this as: F evaluated from a to b, or F(x) evaluated from x = a to x = b.
Now that we have this new notation, lets rewrite our formulas from Proposition 2.3.12,
Proposition 2.3.22, and Proposition 2.3.27. We have
_
b
a
k dx = k(b a) = kx

b
a
,
140 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
_
b
a
xdx =
b
2
2

a
2
2
=
x
2
2

b
a
,
and
_
b
a
x
2
dx =
b
3
3

a
3
3
=
x
3
3

b
a
.
Do you see the pattern here? It may not be clear from just three examples, but you may be
able to guess what happens more generally.
In each of the three formulas above, the derivative of the function being evaluated from a to
b on the right is precisely the function being integrated from a to b on the left. In other words,
the function being evaluated from a to b on the right is an anti-derivative of the integrand.
Is this true more generally? YES. This result is part of the Fundamental Theorem of
Calculus, and it is the reason that the notation
_
f(x) dx and terminology indenite integral
for anti-derivatives are so similar to the notation and terminology for denite integrals.
Once we have the Fundamental Theorem, we will no longer need to compute denite integrals
by the cumbersome process of taking limits of Riemann sums. All of our anti-derivative formulas
from Chapter 1 will become formulas for computing denite integrals.
Understand: ALL of the applications of the denite integral are due to the fact that it
represents a continuous sum of innitesimal contributions, i.e., that the denite integral is
a limit of Riemann sums. The Fundamental Theorem of Calculus is NOT the denition of
the denite integral; the value of the Fundamental Theorem is that it allows us to calculate
these continuous sums much more easily.
We need a few results before we can prove the Fundamental Theorem of Calculus.
Suppose that f is a Riemann integrable function on the interval [a, b]. By Theorem 2.3.16
(and Denition 2.3.17), for all x in [a, b], the function f is also integrable on the interval [a, x].
Thus, we may make the following denition.
Denition 2.4.2. Suppose that f is a Riemann integrable function on the interval [a, b].
Then, the integral function of f on [a, b] is the function I
[a,b]
f
, with domain [a, b] and
codomain (, ), given by
I
[a,b]
f
(x) =
_
x
a
f(t) dt.
Note the use of the dummy variable t in the integrand; it would be confusing to use x in
the integrand, as x is the upper-limit of integration. We did not have to use t as our dummy
2.4. THE FUNDAMENTAL THEOREM OF CALCULUS 141
variable (though it is a standard choice); we could have used any variable name, other than x,
a, b, and f. The dummy variable d would also be a bit confusing.
It will be important to us that the following theorem is true, and its proof is instructive,
since it uses a number of basic properties of integration.
Theorem 2.4.3. Suppose that f is Riemann integrable on the interval [a, b]. Then, the
integral function, I
[a,b]
f
, of f on [a, b] is continuous.
Proof. Let c be in [a, b]. We need to show that the limit, as x approaches c, of I
[a,b]
f
(x) is equal
to I
[a,b]
f
(c), where, if c equals a or b, we mean that we take the corresponding one-sided limit, so
that we are always looking at xs in the interval [a, b]. For simplicity, we will use the two-sided
limit notation throughout the proof.
We will show that
lim
xc
I
[a,b]
f
(x) = I
[a,b]
f
(c)
by showing that
lim
xc

I
[a,b]
f
(x) I
[a,b]
f
(c)

= 0.
Note that

I
[a,b]
f
(x) I
[a,b]
f
(c)

I
[a,b]
f
(c) I
[a,b]
f
(x)

.
Now,
I
[a,b]
f
(x) I
[a,b]
f
(c) =
_
x
a
f(x) dx
_
c
a
f(x) dx =
_
x
c
f(x) dx
and, hence, we wish to show that
lim
xc

_
x
c
f(x) dx

= 0.
We accomplish this by using that, since f is Riemann integrable on [a, b], f is bounded on
[a, b], by Theorem 2.3.6. Thus, there exists M 0 such that, for all x in [a, b], M f(x) M.
Now, if x c, the monotonicity of integration, Theorem 2.3.20, implies that
M(c x)
_
c
x
M dt
_
c
x
f(t) dt
_
c
x
M dt = M(c x).
142 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
Similarly, if c x, then
M(x c)
_
x
c
M dt
_
x
c
f(t) dt
_
x
c
M dt = M(x c).
Hence, in either case, we obtain

_
x
c
f(x) dx

M[c x[.
As x approaches c, [c x[ approaches 0, and, thus, so does

_
x
c
f(x) dx

.
Remark 2.4.4. The conclusion of Theorem 2.4.3 is that the function on [a, b] that sends x to
_
x
a
f(t) dt is continuous.
An essentially identical argument shows that the function on [a, b] that sends x to
_
b
x
f(t) dt
is continuous.
In Denition 1.1.1, we dened what is meant by an anti-derivative of a function f on an
open interval I; its a function F on I such that F

= f (at all points in I). We also reminded

you that the Mean Value Theorem for Derivatives implies that any two anti-derivatives of the
same f must dier by a constant. That is, if F
1
and F
2
are both anti-derivatives of f on I, then
there exists a constant C such that, for all x in I, F
1
(x) = F
2
(x) +C.
It will be useful for us to dene the notion of an anti-derivative of a function on a closed
interval, or on any interval whatsoever, and to prove that any two such anti-derivatives dier
by a constant.
Denition 2.4.5. Let f be a function on the interval J. An anti-derivative of f on J
is a continuous function on J, which, on the interior of J (the open interval of points other
than possible endpoints of the original interval), is an anti-derivative of f, i.e., a continuous
function F on J such that F dierentiable on the interior of J, and such that, for all x in
the interior of J, F

(x) = f(x).
2.4. THE FUNDAMENTAL THEOREM OF CALCULUS 143
Theorem 2.4.6. Suppose that f is a function on the interval J. Then, any two anti-
derivatives of f on J dier by a constant.
Proof. This follows easily from the result on open intervals. If J consists of a single point, then
the result is trivially true. So, assume that is not the case, i.e., assume that the interior of J is
a non-empty open interval.
Suppose that both F
1
and F
2
are anti-derivatives of f on J. Then, we know that there exists
a constant C such that, for all x in the interior of J, F
1
(x) = F
2
(x) +C.
Suppose that a is a left endpoint of the interval J, and that a is contained in J. As F
1
and
F
2
are continuous at a, we have
F
1
(a) = lim
xa
+
F
1
(x) = lim
xa
+
(F
2
(x) +C) = F
2
(a) +C.
In the exact same fashion, we see that, if b is a right endpoint of J, which is contained in J,
then F
1
(b) = F
2
(b) +C.
We are now going to prove the Fundamental Theorem of Calculus. The theorem is usually
broken into two parts. The rst part tells us that the integral function I
[a,b]
f
of a continuous
function f on the interval [a, b] is an anti-derivative of f on [a, b]. The second part tells us that,
if we already know an anti-derivative of f on [a, b], then we can use that to evaluate
_
b
a
f(x) dx.
It is this second part that is used most often in applications.
Theorem 2.4.7. (Fundamental Theorem of Calculus, Part 1) Suppose that f is
Riemann integrable on [a, b] and is continuous at a point x
0
in (a, b). Then, the integral
function I
[a,b]
f
of f on [a, b] is dierentiable at x
0
and
_
I
[a,b]
f
_

(x
0
) = f(x
0
).
Thus, if f is continuous on [a, b], then I
[a,b]
f
is an anti-derivative of f on [a, b].
Proof. See Theorem 2.A.10.
144 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
If f is continuous on an interval J, then, for all a and x in J, f is continuous on the interval
[a, x] or [x, a] (depending on which interval is dened), and so, in either case,
_
x
a
f(t) dt exists.
It is easy to conclude the following corollary from Theorem 2.4.7.
Corollary 2.4.8. Suppose that f is continuous on an interval J, and that a is a point in
J. Then, the function of x, with domain J, given by
_
x
a
f(t) dt
is an anti-derivative of f on J.
Remark 2.4.9. The rst part of the Fundamental Theorem, and its corollary, are of great
theoretical importance; they tell us that all continuous functions have anti-derivatives.
For instance, quick...whats an anti-derivative of the continuous function f(x) = e
x
2
?
Thats easy: F(x) =
_
x
0
e
t
2
dt is an anti-derivative of e
x
2
.
Understand: this means that, if you want to calculate
d
dx
__
x
0
e
t
2
dt
_
,
you dont rst calculate
_
x
0
e
t
2
dt, and then take the derivative. You simply apply the Funda-
mental Theorem to conclude immediately that
d
dx
__
x
0
e
t
2
dt
_
= e
x
2
.
Whats another anti-derivative of e
x
2
? Easy again:
_
x
37
e
t
2
dt.
How can both
_
x
0
e
t
2
dt and
_
x
37
e
t
2
dt be anti-derivatives of e
x
2
? Because they dier by
a constant:
_
x
0
e
t
2
dt
_
x
37
e
t
2
dt =
_
37
0
e
t
2
dt,
2.4. THE FUNDAMENTAL THEOREM OF CALCULUS 145
which is just a xed constant.
We should make a nal comment. Recall Remark 1.1.23, where we stated that the function
f(x) = e
x
2
has no elementary anti-derivative. This is true. The Fundamental Theorem
guarantees that continuous functions possess anti-derivatives; it does not tell us that those
anti-derivatives have to be easily expressible in terms of familiar functions.
Anti-derivatives of elementary functions need not be elementary functions.
We now come to the second part of the Fundamental Theorem of Calculus; the part that we
discussed at the beginning of the section, the part that is most often used in applications. In a
way, the result that we will state, and prove, should seem kind of obvious.
Suppose that F(x) is an anti-derivative of f(x) on [a, b]. Then, for all x in (a, b), f(x) = F

(x),
and the derivative F

(x) is the limit of the change in F divided by the change in x, i.e., the limit,
as x approaches 0, of F/x. But whats the denite integral,
_
b
a
f(x) dx =
_
b
a
F

(x) dx? You

take limits of Riemann sums; you take the limit of what you get when you multiply values of
F

times small changes in x, and add these together. In other words, as x goes from a to b, we
add up a bunch of quantities of the form
F

(x)x
F
x
x = F,
and we take the limit of this as x approaches 0. Now we use that the sum

(F) telescopes,
just as in Proposition 2.1.9, to give us F(b) F(a). Therefore, the Riemann sums approxi-
mately equal F(b) F(a) and, as x approaches 0, the approximation should become an
equality. Technically, we need that f(x) is continuous, but our informal reasoning should make
the following theorem easy to believe.
Theorem 2.4.10. (Fundamental Theorem of Calculus, Part 2) Suppose that f is
continuous on an interval [a, b], and that F is an anti-derivative of f on [a, b].
Then,
_
b
a
f(x) dx = F(x)

b
a
= F(b) F(a).
Proof. This actually follows very quickly from our previous results.
146 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
By Theorem 2.4.7, I
[a,b]
f
is an anti-derivative of f on [a, b]. As F is also an anti-derivative of
f on [a, b], Theorem 2.4.6 implies that there exists a constant C such that, for all x in [a, b],
_
x
a
f(t) dt = F(x) +C.
When x = a, the left side of the above equality is 0, and we obtain that 0 = F(a) + C, i.e.,
C = F(a). Therefore, for all x in [a, b],
_
x
a
f(t) dt = F(x) F(a).
When x = b, we obtain
_
b
a
f(t) dt = F(b) F(a),
which, noting that we may now use x as our dummy variable of integration, is what we wanted
to show.
It is now easy for us to calculate denite integrals by using our anti-derivative formulas from
Chapter 1.
Example 2.4.11. Calculate
_
4
1
_
7x
3
+ 3

x +
5
x
_
dx.
Solution:
This calculation would be ridiculously complicated if we had to explicitly use limits of Rie-
mann sums. But, we dont have to use Riemann sums; we now have the Fundamental Theorem
of Calculus.
From our formulas in Section 1.1, we quickly nd that
7
x
4
4
+ 3
x
3/2
3/2
+ 5 ln x =
7x
4
4
+ 2x
3/2
+ 5 ln x
is an anti-derivative of 7x
3
+ 3

x +
5
x
on [1, 4] (actually, on all of (0, )).
2.4. THE FUNDAMENTAL THEOREM OF CALCULUS 147
Therefore, the Fundamental Theorem tells us that
_
4
1
_
7x
3
+ 3

x +
5
x
_
dx =
_
7x
4
4
+ 2x
3/2
+ 5 ln x
_

4
1
=
_
7 4
4
4
+ 2 4
3/2
+ 5 ln 4
_

_
7 1
4
4
+ 2 1
3/2
+ 5 ln 1
_
=
448 + 16 + 5 ln 4
7
4
2 0 =
1841
4
+ 5 ln 4 467.18147.
You may be thinking Wait wouldnt I get a dierent answer if I used a dierent anti-
derivative of 7x
3
+ 3

x +
5
x
? The answer had better be no, and it is.
Any anti-derivative of 7x
3
+3

x+
5
x
diers from the one we used by some constant C. Thus,
the only other possibilities for us to use for the anti-derivative are all of the form
7x
4
4
+ 2x
3/2
+ 5 ln x +C.
Does the C cause us to get a dierent answer? No, because when we evaluate from 1 to 4, the
C gets cancelled out:
_
7x
4
4
+ 2x
3/2
+ 5 ln x +C
_

4
1
=
_
7 4
4
4
+ 2 4
3/2
+ 5 ln 4 +C
_

_
7 1
4
4
+ 2 1
3/2
+ 5 ln 1 +C
_
=
_
7 4
4
4
+ 2 4
3/2
+ 5 ln 4
_

_
7 1
4
4
+ 2 1
3/2
+ 5 ln 1
_
,
which is what we had before.
The moral of the story is that, when evaluating denite integrals by use of the Fundamental
Theorem, you just use SOME anti-derivative; you dont need to ever put in a general +C.
Example 2.4.12. Find the area under the graph of y = sin x and above the interval [0, ].
148 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
-1
-0.5
0
0.5
1
!
Figure 2.30: The area under part of the graph of y = sin x.
Solution:
This area is simply
_

0
sin xdx = cos x

0
= cos (cos 0) = 1 + 1 = 2.
Example 2.4.13. Find the average value of
f(x) =
1
1 +x
2
on the interval [1, 1], and nd all values of c in [1, 1] that are guaranteed to exist by the
Mean Value Theorem for Integration, Theorem 2.3.31, i.e., all values of c in [1, 1] such that
f(c) equals the average value of f on [1, 1].
Solution:
By denition (see Denition 2.3.30), the average value of f on [1, 1] is equal to
1
1 (1)
_
1
1
1
1 +x
2
dx.
2.4. THE FUNDAMENTAL THEOREM OF CALCULUS 149
By the Fundamental Theorem and Theorem 1.1.13, we nd that this equals
1
2
_
tan
1
(1) tan
1
(1)
_
=
1
2
_

4

_

4
__
=

4
.
-1 -0.5 0 0.5 1
-0.25
0.25
0.5
0.75
1
1.25
1.5
y=!/4
y=1/(1+x )
2
y=1/(1+x )
2
Figure 2.31: Equal areas under y = 1/(1 +x
2
) and y = /4.
For what c in [1, 1] does f(c) = /4? We solve
1
1 +c
2
=

4
and nd that we must have
c
2
=
4

,
i.e.,
c =
_
4

0.5227232;
both of which are in the interval [1, 1].
Example 2.4.14. In Example 2.3.28, we discussed the fact that the denite integral
_
r
0
_
r
2
x
2
dx
150 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
yields one quarter of the area inside a circle of radius r.
r
r
Figure 2.32: Area under y =

r
2
x
2
.
Then, we used our knowledge from high school geometry about the area inside a circle of
radius r to conclude that
_
r
0
_
r
2
x
2
dx =
r
2
4
.
However, now that we have the Fundamental Theorem, we can use integration/anti-dierentiation
to verify that the area inside one quarter of a circle of radius r is r
2
/4, i.e., that the area inside
a circle of radius r is r
2
. In other words, we can show that the formula from high school
geometry is correct.
In Example 1.2.6, we found (replacing a with r) that
_
_
r
2
x
2
dx =
1
2
_
x
_
r
2
x
2
+r
2
sin
1
_
x
r
__
+C.
Thus, the Fundamental Theorem tells us
_
r
0
_
r
2
x
2
dx =
1
2
_
x
_
r
2
x
2
+r
2
sin
1
_
x
r
__

r
0
=
1
2
_
(0 +r
2
sin
1
(1)) (0 +r
2
sin
1
(0))
_
=
1
2
r
2

2
=
r
2
4
.
2.4. THE FUNDAMENTAL THEOREM OF CALCULUS 151
Substitution in Denite Integrals:
We discussed the use of substitutions in nding anti-derivatives in Theorem 1.1.15. Now that
we have denite integration and the Fundamental Theorem, we can state a similar substitution
for denite integrals.
Theorem 2.4.15. (Substitution in Denite Integrals) Suppose that g is continuously
dierentiable on an open interval which contains the closed interval [a, b], and that f is
continuous on the closed interval between the minimum and maximum values of g on [a, b].
Then,
_
b
a
f(g(x)) g

(x) dx =
_
g(b)
g(a)
f(u) du.
Proof. We will assume that g(a) g(b). The case where g(b) g(a) is essentially identical.
As f is continuous on [g(a), g(b)], f possesses an anti-derivative F on [g(a), g(b)] by Theo-
rem 2.4.7, i.e., there is a function F, which is continuous on [g(a), g(b)], and such that, for all x
in (g(a), g(b)), F

(x) = f(x).
Since the composition of continuous functions is continuous, and, for all x in (a, b)
(F g)

(x) = F

(g(x)) g

(x),
it follows that F g is an anti-derivative of (f g) g

on [a, b]. The hypotheses guarantee that

f(g(x)) g

(x) is continuous on [a, b].

Therefore, we may apply the Fundamental Theorem, Theorem 2.4.10, to both sides of the
desired equality, and conclude that they are both equal to F(g(b)) F(g(a)).
Example 2.4.16. Lets calculate
_
2
0
xe
x
2
dx in two slightly dierent-looking ways.
Our rst method is to note that xe
x
2
is continuous everywhere, and use the Fundamental
Theorem, Theorem 2.4.10, to calculate the value of integral. We proceed as we did for indenite
integrals in Section 1.1, except that we carry along the limits of integration throughout the
process.
Let u = x
2
, so that du = 2xdx. Then, xdx = du/2, and obtain
_
2
0
xe
x
2
dx =
_
x=2
x=0
e
u
(du/2) =
1
2
_
x=2
x=0
e
u
du,
152 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
where we need to explicitly write x = 0 and x = 2 for the limits of integration, not just 0 and 2.
This is important: if we wrote simply
_
2
0
e
u
du for the integral, then the limits of inte-
gration would be telling us values of u, and then the Fundamental Theorem would give us
(e
2
1)/2 = (1 e
2
)/2 for the value of the integral. This is wrong.
Proceeding correctly, we nd
1
2
_
x=2
x=0
e
u
du =
1
2
e
u

x=2
x=0
=
1
2
e
x
2

2
0
=
e
4
2

1
2
=
1 e
4
2
.
Notice that, in the above process, all that we were really doing was calculating an anti-derivative,
as we did in Section 1.1, and carrying the limits of integration x = 0 and x = 2 throughout the
calculation.
Whats the other method? To apply Theorem 2.4.15, and actually change the limits of
integration to describe what u does. This saves us cumbersome notation, and means that we
dont need to reinsert what u was, in terms of x, at the end of the calculation. Aside from that,
the process looks the same.
So, as before, let u = x
2
, so that du = 2xdx. Then, of course, xdx = du/2, but, now,
we also note the values of u that we should use for the limits of integration in the u integral;
when x = 0, u = 0
2
= 0 and, when x = 2, u = 2
2
= 4. Therefore, we obtain
_
2
0
xe
x
2
dx =
_
4
0
e
u
(du/2) = =
1
2
e
u

4
0
=
e
4
2

1
2
=
1 e
4
2
.
Thus, we see that, when substituting into a denite integral, changing the limits of integration
to describe what your substitution variable is doing saves some time maybe not a lot of time,
but some.
However, there are some times when changing your limits of integration saves you a LOT
of time. Consider the integral
_
1
0
e
[x(x1)]
2
(2x 1) dx.
Make the substitution u = x(x 1) = x
2
x, so that du = (2x 1) dx. Then, the integral
becomes
_
x=1
x=0
e
u
2
du.
2.4. THE FUNDAMENTAL THEOREM OF CALCULUS 153
We cannot produce a nice anti-derivative of e
u
2
; this function has no elementary anti-
derivative. See Remark 1.1.23. Nonetheless, had we changed our limits of integration to describe
what u does, we would have found that, when x = 0, u = 0, and when x = 1, u = 0; thus, our
integral becomes
_
0
0
e
u
2
du = 0.
Therefore, in this example, switching to the u limits of integration enables us to calculate an
integral that we would not be able to calculate otherwise.
2.4.1 Exercises
In each of Exercises 1 through 15, calculate the denite integrals using the Funda-
mental Theorem of Calculus.
1.
_
5
3
x
3
+x + 2 dx.
2.
_
6
4
y
y
2
3
dy.
3.
_
2

2
sin 3t dt.
4.
_
200
2
5
z
dz.
5.
_
6
5
_
100 u
2
du.
6.
_
10
5
dw

w
2
+ 25
.
7.
_
3
3
dv
(9 +v
2
)
2
dv.
8.
_
4
0
cosh(5y) dy.
9.
_
3
1
8u 35
u
2
7u
du.
10.
_

0
sin t cos t dt.
154 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
11.
_
15
9
dx

x
2
49
.
12.
_

0
cos
2
d.
13.
_
1
0
(2x + 6)e
(x+3)
2
dx.
14.
_
4
0
_
16 +y
2
dy.
15.
_
2
0
z
2
+z + 3
z + 3
dz.
16. Suppose we want to calculate
_
1
1
dx
x
. Since ln [x[ is an anti-derivative of the integrand,
the Fundamental Theorem of Calculus tells us that
_
1
1
dx
x
= ln [x[

1
1
= ln 1 ln 1 = 0.
What is wrong with this argument?
17. Suppose that f is continuous on [a, b] and that G(x) =
_
x
2
a
f(t) dt. What is dG/dx? Hint:
Use the Chain Rule.
In each of Exercises 18 through 22, nd the average value of the function on the
closed interval if it is dened. If it is undened, explain why.
18. k(x) =

9 x
2
, [2, 5].
19. j(y) =
_
1 +y
2
, [0, 1].
20. h(z) =
1
z
2
, [2, 2].
21. g(u) = cosh u, [k, k], k > 0.
22. f(v) = sec v, [0, ].
23. Find the area under the graph of y = 3e
x
+2 and above the interval [0, ln 7] on the x-axis.
Note that 3e
x
+ 2 0.
2.4. THE FUNDAMENTAL THEOREM OF CALCULUS 155
24. Find the area trapped between the graph y = y(x) = 2 +
3
1 +x
2
and the interval [1, 1]
on the x-axis. Note that, for x in [1, 1], y(x) is sometimes positive and sometimes
negative. You need to calculate the total area, both above and below the given interval.
Given a probability density function, f(x), for a continuous random variable, X, the
probability that a X b is
_
b
a
f(x) dx.
25. A continuous random variable is uniformly distributed if f(x) = C for some (necessarily
positive) constant. This means that any all possible outcomes are equally likely to appear.
Suppose the life-span of a certain organism is uniformly distributed between 4 and 9 years
and the probability density function is f(x) = 1/5. What is the probability a given
organism will live for 6 to 8 years?
26. If I = [a, b] is the set of all possible outcomes of a random variable with density function
f(x), then it must be true that
_
b
a
f(x) dx = 1. This means the probability the random
variable falls within the set of all its possible outcomes is 100%. Prove that a uniformly
distributed random variable on the interval [a, b] has density function f(x) = 1/(b a).
27. A random variable has a range of possible outcomes between 1 and 4.
a. The density function is f(x) = cx
2
. What is c?
b. What is the probability of an outcome between 2 and 3?
The expected value or expectation of random variable X is a weighted average of
the possible values that the random variable can achieve and is denoted E(X). For
a discrete variable, E(X) =
n

i=1
x
i
p(x
i
) where each x
i
is a possible outcome that will
occur with probability p(x
i
). For a continuous random variable that can achieve a
value on the interval [a, b], E(X) =
_
b
a
xf(x) dx where f(x) is the density function.
28. Suppose two standard six-sided dice are rolled and were interested in the average, or
expected sum on the two dice. Thus, we let X can be any integer between two and 12.
a. Complete the table below by determining the probability of each possible outcome
and entering it in the second column. Note that there are 36 possible outcomes
(1, 1), (1, 2), (1, 3), .. etc. For example, there are four ways to achieve a sum of ve:
(1, 4), (4, 1), (2, 3), (3, 2). The probability of a sum of ve is therefore 4/36 = 1/9.
156 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
x
i
p(x
i
) x
i
(x
i
)
1
2
3
4
5 4/36 20/36
6
7
8
9
10
11
12
b. What is the most likely sum? That is, what sum has the highest probability of
occurring?
c. Fill in the third column by multiplying the rst two columns together. In row 5, for
example, x
i
p(x
i
) = 20/36 = 5/9.
d. What is the sum of all the entries in the third column? This is the expected, or
average sum.
29. What is the expected value of a continuous random variable uniformly distributed over
the interval [a, b]?
30. Let f(x) =
3
4
(1 x
2
).
a. Verify that this is a legitimate density function of a random variable with possible
outcomes on the interval [1, 1] by verifying that
_
1
1
f(x) dx = 1.
b. What is E(X)?
More generally, we can dene the expectation of a function of the random variable.
For example, in the dice example above, we may have been interested in the square
of the sum of the dice, rather than just the sum. The expectation of the square of
the sum is

(x
i
)
2
p(x
i
). In general, if g(X) is a function of a discrete variable, we
write: E(g(X)) =

g(x
i
) p(x
i
). If g(X) is a function of a continuous variable, we
write: E(g(X)) =
_
b
a
g(x) p(x) dx.
31. What is E(X
2
) in the dice example?
32. What is E(X
2
) for a uniformly distributed random variable on the interval [a, b]?
33. What is E(X
2
) for a random variable with domain [1, 1] and density function f(x) =
3
4
(1 x
2
)?
2.4. THE FUNDAMENTAL THEOREM OF CALCULUS 157
Variance measures the dispersion of a variable, or how spread out it is. Let = E(X).
Then, the variance of a discrete variable is Var(X) =

(x )
2
f(x). The variance
of a continuous random variable is Var(X) =
_
b
a
(x )
2
f(x) dx.
34. Show that for a discrete variable, Var(X) = E(X
2
) E(X)
2
.
35. Show that for a continuous variable, Var(X) = E(X
2
) E(X)
2
. Hint: the proof is nearly
identical to the previous problem. This shows that the rules governing nite sums are in
some sense not that dierent from the rules governing denite integrals.
36. What is the variance of the sum of two dice?
37. What is the variance of a uniformly distributed random variable on the interval [a, b]?
38. What is the variance of a random variable with density function f(x) =
3
4
(1 x
2
)?
39. Let a and b be positive integers. Prove that
_

sin axcos bxdx = 0. Hint: prove that

sin axcos bx =
1
2
(sin(ax +bx) + sin(ax bx)) and use the fact that cos x is an even func-
tion.
40. Let a and b be positive integers. Evaluate
_

sin axsin bxdx, when

a. a ,= b;
b. a = b.
Use the Fundamental Theorem of Calculus and the Chain Rule to solve Exercises
41 through 44.
41. What is
d
dx
_
3x
0
3t dt?
42. What is
d
dx
_
5x
2x
t
2
dt?
43. What is
d
dx
_
x
2
x
1 du?
44. What is
d
dx
_
x
2
0
xt dt?
45. Let g(x) =
_
x
2
0
e
x+t
dt. What is g

(2)? Hint: move e

x
outside the integral.
46. Let a and b be positive integers. Evaluate
_

cos(ax) cos(bx) dx when

158 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
a. a ,= b;
b. a = b.
47. Let A =
_
/2
0
sin
k
x
sin
k
x + cos
k
x
dx.
a. Show that A =
_
/2
0
cos
k
_

2
x
_
cos
k
_

2
x
_
+ sin
k
_

2
x
_ dx.
b. Make the substitution u =

2
x and show that A =
_
/2
0
cos
k
u
sin
k
u + cos
k
u
du. Let
the right-hand side of this equation be B.
c. Prove that A+B = /2 and conclude that A = /4.
48. Consider the expression
_
4
2
_
5
1
x
2
y+2y dxdy. We evaluate by rst viewing y as a constant
and integrating with respect to x, and then integrating with respect to y.
a. Evaluate
_
5
1
x
2
y + 2y dx. Assume y is a constant.
b. Take the denite integral of your answer to part (a) with respect to y.
c. Do you get the same answer if you switch the order of integration? That is, is your
answer to part (b) the same as
_
5
1
_
4
2
x
2
y + 2y dy dx?
Use the method in the previous problem to evaluate the integrals in the next two
exercises.
49. a.
_
2
0
_
1
0
r dr d.
b.
_
1
0
_
2
0
r d dr.
50. a.
_

0
_

0
sin xcos y dxdy.
b.
_

0
_

0
sin xcos y dy dx.
Its not always true that switching the order of integration results in the same value. However,
if a function is reasonably well-behaved, then switching the order of integration has no eect on
the nal outcome. This is part of the content of Fubinis Theorem.
2.5. IMPROPER INTEGRALS 159
2.5 Improper Integrals
In this section, we will dene the integral of a function over a set of real numbers, where the set
of real numbers need not be a closed, bounded, interval; we also dene integrals when the value
of the integrand is unbounded, even if the interval of integration is itself bounded. This really
requires a new denition, for Theorem 2.3.6 tells us that the Riemann integral of an unbounded
function does not exist.
The actual calculation of our new type of integral will involve calculating our usual integrals
on intervals of the form [a, b], where now either a or b varies, and we will then take a limit
as a or b approaches some problematic value. Of course, the calculation of the integrals on
the intervals [a, b] can/will still use the Fundamental Theorem of Calculus, and, hence, we may
apply all of our techniques from earlier sections to nd anti-derivatives.
As we wish to be able to discuss integrals over intervals [a, b] and over intervals (a, b] (and
over other sets), the notation
_
b
a
does not suce to distinguish between the types of intervals
that we care about. Thus, we adopt some new notation; if E is a subset of the real numbers,
we will write
_
E
f(x) dx
for the integral of f over the set E. Of course, right now, this has no meaning for us, unless E
is a closed interval [a, b], in which case
_
b
a
f(x) dx =
_
[a,b]
f(x) dx. Our goal in this section is to
dene
_
E
f(x) dx for sets E that need not be closed intervals.
Before we really start looking at new types of integrals, it will be helpful to have a new piece
of terminology.
Denition 2.5.1. If f is a real function, with domain D, an extension of f is a function

f whose domain is larger (or equal to) D, and such that f and

f agree at all points of D,
i.e., for all x in D,

f(x) = f(x).
In other words, an extension

f, of f, is a function whose domain includes the domain,
D, of f, and such that the restriction of

f to D is equal to f.
160 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
Example 2.5.2. For example, the function

f, with domain [0, 1] given by

f =
_
_
_
sin x
x
if 0 < x 1;
1 if x = 0
is an extension of the function f(x) = (sin x)/x, with domain (0, 1]. In fact,

f is a continuous
extension of f to [0, 1], since
lim
x0
+
sin x
x
= 1.
Now, lets start our discussion of more general notions of integration by looking at the easiest
case, one where we already know the answer: the case of a function which is Riemann integrable.
Example 2.5.3. Consider f(x) = x
2
on the interval [0, 1]. Certainly, f is Riemann integrable on
[0, 1], since f is continuous on [0, 1]. Thus, we know what
_
[0,1]
x
2
dx means; it means
_
1
0
x
2
dx,
which, by the Fundamental Theorem of Calculus, is equal to (x
3
/3)

1
0
= 1/3.
What if we omit the 0 from the interval of integration, i.e., what should
_
(0,1]
x
2
dx mean?
Theorem 2.3.14 tells us that, for Riemann integrals on closed intervals, altering the function at
a nite number of points does not change the integrability of the function or the value of the
integral. Thus, intuitively, it seems reasonable that omitting a single point of integration, like 0,
should not aect the integral. Therefore, in this case, it seems reasonable to make the denition
that
_
(0,1]
x
2
dx =
_
[0,1]
x
2
dx =
1
3
.
While this seems reasonable, it should, in a way, seem like cheating; we wanted to integrate
f(x) = x
2
on the interval (0, 1], and yet we used information about f on the larger interval [0, 1],
i.e., we used that there was a continuous extension of f from the interval (0, 1] to the interval
[0, 1]. Could we have dened
_
(0,1]
x
2
dx without using the extension to [0, 1]? Yes.
For all a such that 0 < a 1, f(x) = x
2
is Riemann integrable on the interval [a, 1], and
_
1
a
x
2
dx =
x
3
3

1
a
=
1
3

a
3
3
,
and this integral uses only that f(x) = x
2
is dened on (0, 1].
2.5. IMPROPER INTEGRALS 161
Now, we can take the limit as a approached 0 from the right to obtain our previous answer.
That is, we could have dened
_
(0,1]
x
2
dx by
_
(0,1]
x
2
dx = lim
a0
+
_
1
a
x
2
dx = lim
a0
+
_
1
3

a
3
3
_
=
1
3
.
The point is that, in this example, we obtain the same value for
_
(0,1]
x
2
dx, regardless of
whether we use the fact that x
2
extends to [0, 1] or whether we instead use the limits of integrals,
as our lower-limit of integration approaches 0.
Example 2.5.4. Lets consider a more-complicated example. Let f be the function, with
domain (0, 1], given by
f(x) =
sin x
x
.
How many choices do we have for reasonable ways to dene the integral
_
(0,1]
f(x) dx? At least
two.
First, we can dene an extension

f to the closed interval [0, 1], and then dene
_
(0,1]
sin x
x
dx =
_
[0,1]

f(x) dx.
Of course, we either need to pick a particular extension, or show that the value of
_
[0,1]

f(x) dx
is always the same, regardless of what extension we select. We could, in fact, take

f to be the
continuous extension of f given in Example 2.5.2; then,

f would certainly be Riemann integrable
by Theorem 2.3.8.
However, we dont have to use continuous extension of f. Since (sin x)/x does possess
a continuous extension to the interval [0, 1], it follows that (sin x)/x is bounded on (0, 1]. Thus,
Theorem 2.3.8 and Theorem 2.3.14 imply that it doesnt matter how we dene

f at x = 0; no
matter what value we chose for

f(0),

f will be bounded and, at least, piecewise-continuous
hence, Riemann integrable and changing the value at one point will not aect the value of the
integral. Therefore, we could dene
_
(0,1]
f(x) dx =
_
[0,1]

f dx,
162 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
for any extension of f to a function

f on [0, 1].
However, we could take the second approach, as we did in Example 2.5.3; without using an
extension at all, we could dene
_
(0,1]
sin x
x
dx = lim
a0
+
_
1
a
sin x
x
dx.
This leads to a new question or two. If 0 < a 1, we know that (sin x)/x is continuous on [a, 1]
and, hence,
_
1
a
sin x
x
dx exists. But, how do we know that the limit as a 0
+
exists and, even
if the limit does exist, how do we know that that limit is equal to what wed get by extending
(sin x)/x to [0, 1] and then taking the Riemann integral of our extension over [0, 1]?
Actually, the answers to these questions are easy, given our earlier results. Suppose that

f
is an extension of f to [0, 1]. Then, as we discussed above,

f is Riemann integrable on [0, 1]. By
Remark 2.4.4, the function on [0, 1] that sends a to
_
1
a

f(x) dx is continuous; in particular,

_
1
0

f(x) dx = lim
a0
+
_
1
a

f(x) dx = lim
a0
+
_
1
a
f(x) dx,
where the second equality above follows from the fact that, if 0 < x 1, then

f(x) = f(x),
since

f is an extension of f.
Thus, lim
a0
+
_
1
a
f(x) dx exists and equals what we would obtain by calculating the Riemann
integral of any extension of f to [0, 1]. Therefore, it seems reasonable to dene the integral of f
over the half-open interval (0, 1] by
_
(0,1]
sin x
x
dx = lim
a0
+
_
1
a
sin x
x
dx.
What we have seen in the previous two examples is that, in those cases, it made sense to
dene
_
(0,1]
f(x) dx by
_
(0,1]
f(x) dx = lim
a0
+
_
1
a
f(x) dx,
where each
_
1
a
f(x) dx is a Riemann integral. However, in those cases, we could also have dened
_
(0,1]
f(x) dx by extending f to the closed interval [0, 1] and then using the Riemann integral of
2.5. IMPROPER INTEGRALS 163
the extended function. Now we will look at an example where the approach via extensions does
not work, but the limit idea still yields a meaningful result.
Example 2.5.5. Consider the function
f(x) =
1

x
= x
1/2
on the half-open interval (0, 1]. As x approaches 0 from the right, 1/

x approaches , and
so f is unbounded on (0, 1]. Hence, any extension of f to the closed interval [0, 1] will also be
unbounded and, therefore, will not be Riemann integrable (Theorem 2.3.6).
On the other hand, using the Fundamental Theorem, and the Power Rule for Integration,
we nd
lim
a0
+
_
1
a
x
1/2
dx = lim
a0
+
_
x
1/2
1/2

1
a
_
= lim
a0
+
(2 2

a) = 2.
This means that the area under the graph of y = 1/

x and over the interval [a, 1] approaches

2 as a 0
+
. We say, simply, that the area under the graph of y = 1/

x and over the interval

(0, 1] equals 2.
-0.5 0 0.5 1 1.5
a
Figure 2.33: Area under the graph of y = 1/

x over [a, 1].

What we see in this example is that dening the integral over a half-open interval (a, b]
in terms of limits of integrals over closed intervals gives us a well-dened number, while the
Riemann integral of an extension to the closed interval [a, b] does not exist.
164 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
Before making a denition, we wish to look at one more example.
Example 2.5.6. Suppose that f(x) = e
x
and we wish to integrate f over the interval [0, ).
Note that e
x
is an anti-derivative of e
x
.
In this example, the function f itself is bounded on the given interval; e
x
is between 0 and
1 for x in [0, ). On the other hand, the interval that we want to integrate over is unbounded,
since it goes out to innity. How should we dene
_
[0,)
e
x
dx?
Here, there is no closed interval to which we can possibly extend f. However, the limit
approach still yields an answer:
lim
b
_
b
0
e
x
dx = lim
b
_
e
x

b
0
_
= lim
b
_
e
b
(e
0
)
_
= 0 + 1 = 1.
This means that the area under the graph of y = e
x
and over the interval [0, b] approaches 1 as
b . We say, simply, that the area under the graph of y = e
x
and over the interval [0, )
equals 1.
0 5
1
b
Figure 2.34: Area under the graph of y = e
x
over [0, b].
2.5. IMPROPER INTEGRALS 165
In light of the above discussion and examples, we make the following denition:
Denition 2.5.7. Suppose that f is dened on the half-open interval [a, b), where b may be
, and suppose that, for all c such that a c < b, f is Riemann integrable on the closed
interval [a, c].
Then, we let
_
b
a
f(x) dx =
_
[a,b)
f(x) dx = lim
cb

_
c
a
f(x) dx,
provided that the limit exists, in which case we say that f is integrable on [a, b), or that
the integral
_
b
a
f(x) dx converges. Otherwise, we say that
_
b
a
f(x) dx diverges.
Similarly, suppose that f is dened on the half-open interval (a, b], where a may be , and
suppose that, for all c such that a < c b, f is Riemann integrable on the closed interval
[c, b].
Then, we let
_
b
a
f(x) dx =
_
(a,b]
f(x) dx = lim
ca
+
_
b
c
f(x) dx,
provided that the limit exists, in which case we say that f is integrable on (a, b], or that
the integral
_
b
a
f(x) dx converges. Otherwise, we say that
_
b
a
f(x) dx diverges.
Naturally, if
_
b
a
f(x) dx converges, we dene
_
a
b
f(x) dx =
_
b
a
f(x) dx.
Note that we do not have a notational conict; if f is, in fact, Riemann integrable on [a, b],
then, by Theorem 2.4.3 and Remark 2.4.4, the Riemann integral
_
b
a
f(x) dx equals both of the
one-sided limits of integrals given above.
The way that we usually conclude that f is Riemann integrable on all of the closed intervals
[a, c] (or [c, b]) contained in [a, b) (or (a, b]) is that f is continuous on the entire half-open interval.
Assuming this is the case, the only way that the integral
_
b
a
f(x) dx can possibly fail to converge
is for either the interval of integration to be unbounded, or for the function f to be unbounded
on the interval of integration. We give these two types of integrals which involve unbounded
activity a name:
Denition 2.5.8. An integral
_
b
a
f(x) dx, in which a or b is , or such that f is un-
bounded on the interval (a, b) is called an improper integral.
166 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
The integrals in Example 2.5.5 and Example 2.5.6 are improper integrals, and yet the inte-
grals converge. The importance of improper integrals is that, for continuous functions, theyre
the only types of integrals which might diverge.
While the integrals in Denition 2.5.7 are the basic new types of integrals that we are dening
in this section, we are also interested in more-complicated integrals, ones which break up into a
nite number of pieces which are of the types found in Denition 2.5.7 .
Example 2.5.9. Consider the integral
_
8
1
1
x
2/3
dx.
The point x = 0 is in the interval [1, 8], the interval over which were supposed to integrate.
However, as x approaches 0 from the left or right, the integrand goes to . This is a type of
improper integral. The question is: how should we dene what such an integral means?
-2 -1 0 1 2 3 4 5 6 7 8 9
Figure 2.35: The y = x
2/3
becomes unbounded from either side of x = 0.
The answer is: we want Theorem 2.3.16, on splitting up integrals, to remain true. This
means that we want it to be true that
_
8
1
1
x
2/3
dx =
_
0
1
1
x
2/3
dx +
_
8
0
1
x
2/3
dx,
2.5. IMPROPER INTEGRALS 167
where each of the summands on the right is an integral of the type we dened in Denition 2.5.7.
We now calculate by taking limits:
lim
b0

_
b
1
x
2/3
dx + lim
a0
+
_
8
a
x
2/3
dx = lim
b0

_
3x
1/3

b
1
_
+ lim
a0
+
_
3x
1/3

8
a
_
=
lim
b0

_
3b
1/3
3(1)
1/3
_
+ lim
a0
+
_
3(8)
1/3
3a
1/3
_
= 3 + 6 = 9.
There can also be multiple problem points.
Example 2.5.10. Consider the integral
_

1
1
(x 2)(x 4)
dx.
After possibly removing a nite number of points (problem points, where unboundedness
comes into play), we want to split the interval [1, ) into a nite number of closed or half-open
intervals on which the integrand 1/[(x 2)(x 4)] is continuous; in this splitting, we allow a
pair of closed or half-open intervals to intersect each other in, at most, one point. We then add
together the resulting integrals, provided all of them exist; otherwise, we say that the original
integral diverges.
Thus, we start with the interval [1, ). We remove the two points x = 2 and x = 4, where
1/[(x 2)(x 4)] is undened. For now, this gives us intervals [1, 2), (2, 4), and (4, ). The
intervals (2, 4) and (4, ) are not closed or half-open. Why do we care? We do not want to
have to deal with some sort of simultaneous limits at the two endpoints of the intervals.
Therefore, we further split the intervals (2, 4), and (4, ). Where do we split them? At any
point in-between the endpoints. It is a theorem, which we incorporated into the statement of
Denition 2.5.11, that is doesnt matter where the splitting occurs. So, we split the interval
(2, 4) into (2, 2.5] and [2.5, 4) (pairs of half-open intervals may intersect at a point), and we split
(4, ) into (4, 7] and [7, ).
We end up with ve half-open intervals: I
1
= [1, 2), I
2
= (2, 2.5], I
3
= [2.5, 4), I
4
= (4, 7],
and I
5
= [7, ), whose union is equal to the original interval [1, ), minus a nite number of
points, and pairs of the half-open intervals intersect each other in, at most, one point.
168 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
Now, we dene
_

1
1
(x 2)(x 4)
dx
to equal the sum
_
I1
1
(x 2)(x 4)
dx +
_
I2
1
(x 2)(x 4)
dx +
_
I3
1
(x 2)(x 4)
dx +
_
I4
1
(x 2)(x 4)
dx +
_
I5
1
(x 2)(x 4)
dx,
provided that all of these integrals exist, in which case we say that
_

1
1
(x2)(x4)
dx converges.
If one (or more) of the ve separate integrals, appearing in the summation, diverges, then we
say that
_

1
1
(x2)(x4)
dx diverges.
In the remainder of this example, we will show that
_

1
1
(x 2)(x 4)
dx
diverges, by showing that the rst improper integral in the summation above diverges, i.e., we
will show that
_
2
1
1
(x 2)(x 4)
dx = lim
b2

_
b
1
1
(x 2)(x 4)
dx
diverges to .
We nd an anti-derivative of
1
(x2)(x4)
via partial fractions, as in Section 1.3. So, we rst
determine constants A and B such that
1
(x 2)(x 4)
=
A
x 2
+
B
x 4
,
for all x, other than x = 2 and x = 4. Clearing the denominators, by multiplying each side of
the equality by the big denominator on the left, i.e., by (x 2)(x 4), we obtain
1 = A(x 4) + B(x 2),
which needs to hold for all x. Plugging in x = 2, we nd that 1 = A (2) + B 0, so that
2.5. IMPROPER INTEGRALS 169
A = 1/2. Plugging in x = 4, we nd that 1 = A 0 +B 2, and so B = 1/2. Thus,
1
(x 2)(x 4)
=
1/2
x 2
+
1/2
x 4
,
and we want to calculate
lim
b2

_
b
1
1
(x 2)(x 4)
dx = lim
b2

_
b
1
_
1/2
x 2
+
1/2
x 4
_
dx, (2.3)
or show that it doesnt exist. We need to nd an anti-derivative of 1/(x a), where a is 2 or 4.
We accomplish this by making the substitution u = x a, so that du = dx, and
_
1
x a
dx =
_
1
u
du = ln [u[ + C = ln [x a[ + C.
Thus, Formula 2.3 becomes
lim
b2

_
b
1
1
(x 2)(x 4)
dx = lim
b2

_
1
2
ln [x 2[ +
1
2
ln [x 4[
_

b
1
.
As b is approaching 2 from the left, we know that all of the xs that we are considering are less
than 2 and, hence, [x2[ = 2x and [x4[ = 4x. Now, evaluating at x = b and subtracting
the value at x = 1, we nd
lim
b2

_
b
1
1
(x 2)(x 4)
dx = lim
b2

_
1
2
ln(2 b) +
1
2
ln(4 b)
1
2
ln 3
_
,
where we used that ln 1 = 0. As b approaches 2 from the left,
1
2
ln(4 b)
1
2
ln 3 approaches
1
2
ln 2
1
2
ln 3, but
1
2
ln(2 b) approaches (1/2)() = .
Therefore,
_
2
1
1
(x2)(x4)
dx diverges, and so does the integral over the bigger interval [1, ).
We should mention that, just because
_
2
1
1
(x2)(x4)
dx diverges to , that does not imply that
_

1
1
(x2)(x4)
dx also diverges to ; other parts of the summand that we split the integral into
may (and do) diverge to . Thus, we simply say that
_

1
1
(x2)(x4)
dx diverges, without
trying to specify the manner in which it diverges.
170 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
We summarize our discussion and example above into a denition. In this denition, we will
split a set into a union of closed or half-open intervals (as we did above); the point is that each
of the intervals that we split things into has a problem, an unboundedness issue, at, at most,
one endpoint. It is important that the denition gives you the same result, regardless of how
you pick the half-open intervals. The proof of this is in Theorem 2.A.11.
Denition 2.5.11. Let E be a subset of the real numbers which is the union of a nite
number of intervals. Let f be a real function, which is dened and continuous on the set E
except, perhaps, at a nite set of points P. Let E P denote the set of points in E which
are not in the set P.
Then, EP is a union of a nite number of closed or half-open intervals I
1
, I
2
, . . . , I
n
on
which f is dened and continuous. Given any such decomposition of E P into intervals,
we dene the integral of f on E by
_
E
f(x) dx =
_
I1
f(x) dx +
_
I2
f(x) dx + . . . +
_
In
f(x) dx,
provided that each integral in the summation on the right converges; in this case, we say
that
_
E
f(x) dx converges or that f is integrable on E. If any one of the integrals in the
summation above diverges, then we say that
_
E
f(x) dx diverges.
These denitions of converges, diverges, and the value of the integral are independent of the
choice of the intervals I
1
, I
2
, . . . , I
n
, as long as the given conditions are satised.
If E is an interval [a, b], [a, b), (a, b], or (a, b), then we may also use other notation; we set
_
b
a
f(x) dx =
_
E
f(x) dx, and
_
a
b
f(x) dx =
_
E
f(x) dx.
Note that, with our terminology, we may say that a function f is integrable on a set E,
even if f is not dened at a nite number of the points in E. For instance, in
Example 2.5.9, we would say that 1/x
2/3
is integrable on (1, 8), even though 1/x
2/3
is not
dened at x = 0. Be aware that other books might use dierent terminology, and might not
say that f is integrable on a set on which f is not dened.
2.5. IMPROPER INTEGRALS 171
Just as we have linearity for Riemann integrals, Theorem 2.3.19, we also have linearity for
our more general integrals, provided the individual integrals converge. The proof is essentially
identical, except that, to deal with the improper integrals, you must use that limits are linear,
i.e., you can pull constants out of limits and split up sums.
Theorem 2.5.12. (Linearity of Improper Integrals) Let E be a subset of the real
numbers which is the union of a nite number of intervals. Suppose that f and g are
integrable on E, and that a and b are any real numbers. Then, af + bg is integrable on E
and _
E
af(x) +bg(x) dx = a
_
E
f(x) dx + b
_
E
g(x) dx.
Example 2.5.13. As we saw in Example 2.5.5 and Example 2.5.9 (with dierent upper-limits
of integration), the integrals
_
1
0
1
x
1/2
dx and
_
1
0
1
x
2/3
dx
both converge; we leave it as an exercise for you to show that they converge to 2 and 3, respec-
tively.
Therefore,
_
1
0
_
5
x
1/2

7
x
2/3
_
dx
converges, and equals 5 2 7 3 = 11.
It is sometimes possible to tell that an improper integral converges, without being able to
determine what it converges to. This seemingly bizarre fact stems from a dening property
of the real numbers: every non-empty set of real numbers, which has an upper bound, has a
LEAST upper bound. See Denition 5.1.15 and Theorem 5.1.18. This least upper bound property
is the main reason that we can sometimes tell that limits, as in improper integrals, exist without
knowing the value.
We rst give a theorem which says that theres only one way for some types of improper
integrals to diverge.
172 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
Theorem 2.5.14. Let I be an interval of the form [a, b) or (a, b], where we allow the
intervals [a, ) or (, b]. Suppose that, for all x in I, f(x) 0, and that, for all closed
intervals [c, d] contained in I, f is Riemann integrable on [c, d].
Then, if there exists a real number M (an upper bound) such that, for all [c, d] contained
in I,
_
d
c
f(x) dx M, then
_
b
a
f(x) dx converges, and what it converges to is less than, or
equal to, M; in other words, if there is an upper bound M on all of the
_
d
c
f(x) dx, then
_
b
a
f(x) dx converges to the least such upper bound.
In particular, if
_
b
a
f(x) dx diverges, what it diverges to is .
Proof. See Theorem 2.A.12.
Remark 2.5.15. The analogous statement for f(x) 0 is true, and is obtained by applying
Theorem 2.5.14 to f(x), which would be non-negative.
What you nd for non-positive f is: if there exists a real number M (a lower-bound) such
that, for all [c, d] contained in I, M
_
d
c
f(x) dx, then
_
b
a
f(x) dx converges to the greatest such
lower bound. In particular, if
_
b
a
f(x) dx diverges, what it diverges to is .
Corollary 2.5.16. Let I be an interval of the form [a, b) or (a, b], where we allow the
intervals [a, ) or (, b]. Suppose that, for all x in I, 0 f(x) g(x), and that, for all
closed intervals [c, d] contained in I, f and g are Riemann integrable on [c, d].
Then, if
_
b
a
g(x) dx converges, then so does
_
b
a
f(x) dx, and what it converges to is some-
thing less than, or equal to,
_
b
a
g(x) dx. This implies that, if
_
b
a
f(x) dx diverges, then so
does
_
b
a
g(x) dx.
Proof. This is easy now. Since f g on I, Theorem 2.3.20 tells us that, for all [c, d] contained
in I,
_
d
c
f(x) dx
_
d
c
g(x) dx.
If
_
b
a
g(x) dx converges, then Theorem 2.5.14 tells us that it converges to the least upper bound
M on the integrals
_
d
c
g(x) dx, but, by the inequality above, this M is an upper bound on all of
the integrals
_
d
c
f(x) dx. The corollary now follows by applying Theorem 2.5.14 again.
2.5. IMPROPER INTEGRALS 173
Example 2.5.17. Consider the integrals
_
1
0
1 + sin x

x
dx and
_

1
2 + sin x
x
dx.
Producing manageable anti-derivatives of these integrands is problematic/impossible. Nonethe-
less, we can use Corollary 2.5.16 to determine quickly whether they converge or not.
First, note that, for all x in (0, 1],
0
1 + sin x

x

2

x
,
and that, for all x in [1, ),
0
1
x

2 + sin x
x
.
Now, we nd
_
1
0
2

x
dx = lim
a0
+
_
1
a
2

x
dx = lim
a0
+
4x
1/2

1
a
= 4 lim
a0
+
(1 a
1/2
) = 4,
and
_

1
1
x
dx = lim
b
(ln b ln 1) = .
Combining all of the above with Corollary 2.5.16, we conclude that
_
1
0
1 + sin x

x
dx converges
(to something less than or equal to 4, but we dont know what), while
_

1
2 + sin x
x
dx diverges
to .
2.5.1 Exercises
In Exercises 1 through 17, evaluate the given integral if it converges; otherwise,
show that the integral diverges.
174 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
1.
_

0
dx
(2x + 5)
2
.
2.
_

1
dt
(4 + 5t)
3/2
.
3.
_

0
dx
7x 5
.
4.
_

1
dy
(4y 3)
1/2
.
5.
_

0
sin z dz.
6.
_

0
du
(u + 3)(u + 5)
.
7.
_

4
du
(u + 3)(u + 5)
.
8.
_

0
dv
1 cos v
. Hint: use a half-angle identity to evaluate the integral.
9.
_
1
0
dt

1 t
2
.
10.
_

0
dw
1 +w
2
.
11.
_
1
0
x
1/2
ln xdx.
12.
_
1
0
x
2
ln xdx.
13.
_
k
k
dt
3

t
dt, k > 0.
14.
_

0
tan
1
z
1 +z
2
dz.
15. Let n > 1. Does
_
1
0
dx
x
n
converge or diverge? If it converges, what does it converge to?
16. Let n = 1. Does
_
1
0
dx
x
n
converge or diverge? If it converges, what does it converge to?
17. Let n < 1. Does
_
1
0
dx
x
n
converge or diverge? If it converges, what does it converge to?
2.5. IMPROPER INTEGRALS 175
Use Corollary 2.5.16 to determine whether the integrals in Exercises 18 through 24
converge or diverge. If they converge, you need not calculate the integrals.
18.
_

1
1 +x
6
dx.
19.
_

2
[ sin x[
1 +x
2
dx.
20.
_

1
e
t
t
dt.
21.
_
1
0
e
y

y
dy.
22.
_

1
e
y

y
dy.
23.
_

0
tan
1
x
x
3
+ 5x + 1
dx. Hint: See Exercise 14 above.
24.
_

1
s
e
s
1
ds. Hint: Justify and use the fact that e
s
> 1 +
s
3
6
, for all s > 0.
25. Prove that the elliptic integral
_
1
0
dx
_
(1 x
2
)(1 k
2
x
2
)
converges. Assume k
2
< 1.
26. Formulate a similar comparison theorem for functions that grow without bound at some
real number a. You may limit your statement to either a right or left-hand limit.
27. Find all real numbers p, if any exist, such that
_

1
x
p
dx converges.
28. Find all real numbers p, if any exist, such that
_

1
1 +x
p
dx converges.
Let (n) =
_

0
e
x
x
n1
dx. This function is called the gamma function.
29. Calculate (1), and (2).
30. Prove inductively that (n) = (n 1)! for n = 1, 2, 3, .....
A random variable is said to be an exponentially distributed if its density function
is
f(x) =
_
e
x
, if x 0;
0, if x < 0
for some > 0.
176 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
31. What is E(X), the expected value of an exponential random variable?
32. What is E(X
2
), if X is an exponential random variable?
33. What is the variance, V ar(X), of an exponential random variable?
A data set (in statistics) or a random variable (in probability) is said to be normally
distributed if its density function is
1

2
e
(x)
2
/2
2
.
The graph of this density function is the familiar bell curve.
34. Prove that the expected value of a normally distributed random variable is .
35. What is E(X
2
) for a normally distributed random variable?
36. Prove that the variance of a normally distributed random variable is
2
.
37. Recall that the cumulative distribution function of a continuous random variable is F(a) =
_
a

f(x) dx. F(a) is the probability that the random variable X is less than or equal to
a. What is F(a) for an exponentially distributed random variable?
38. Using the notation of the previous problem, prove that lim
a
F(a) = 1 if F is the cumulative
distribution function of an exponentially distributed random variable.
39. The survival function of a a random variable is the complement of the cumulative distribu-
tion function. That is, if F(x) is the cumulative distribution function, then S(x) = 1F(x)
is the survival function. The name of this function can be understood from an actuarial
perspective where the random variable is a life. Then S(x) is the probability that the life
will survive longer than x years. What is S(x) for an exponential function?
40. The moment generating function of a distribution provides an alternative method of cal-
culating the mean and variance of the distribution, and is dened as follows:
(t) = E(e
tX
) =
_

e
tx
f(x) dx
where f(x) is the density function. Prove that if X is exponentially distributed with
parameter , then (t) =

t
.
2.5. IMPROPER INTEGRALS 177
41. Show that

(0) = E(X) if X is exponentially distributed.

42. Show that

(0) = E(X
2
) if X is exponentially distributed. These two results are not
specic to the exponential distribution; they hold for arbitrary continuous distributions
(Bonus exercise: try to prove this!)
43. The amount of time it takes to wait in line to register for classes is exponentially distributed
with mean 12 minutes (so = 1/12).
a. Whats the probability a student will have to wait for less than 12 minutes?
b. Whats the probability a students waiting time will be between 10 and 20 minutes?
c. Whats the probability a students waiting time exceeds 25 minutes?
44. Recall that the mass m(t) of a decaying compound is given by m(t) = m
0
e
t
where
t > 0. However, this apparently exact formula is merely a model for a complicated
physical process. Radioactive decay can be be viewed in probabilistic terms using an
exponential distribution. First, note that e
t
is the survival function of an exponential
random variable. Assuming the life of the atom is exponentially distributed, it makes sense
to dene the mean life of an atom as ML =
_

0
te
t
dt. Calculate the mean lifetime of
the following atoms ( is given so that t is in years).
a. uranium-239, = 1.55 10
10
.
b. carbon-14, = 1.15 10
5
.
c. tritium, = 0.056.
45. Suppose we have 100 tritium atoms. How many atoms are anticipated to decay between
8 and 12 years based on the model in the previous problem? Note that this is the same
as asking for the probability that a single tritium atom will decay between 8 and 12 years
from now.
46. Is
_

0
xsin(x
2
) dx convergent? Justify your answer.
If f(t) is a continuous function on [0, ), then the Laplace transform of f is the
function F(s) =
_

0
f(t)e
st
dt. Laplace transforms provide a powerful technique for
solving dierential equations. Calculate the Laplace transforms of the functions
in Exercises 47 through 52. Note that s values for which the Laplace transform
integral does not exist are not in the domain of the transformed function F(s).
47. f(t) = 1.
48. f(t) = kt, where k is a constant.
178 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
49. f(t) = e
at
, where a is a constant.
50. f(t) = sin(ct), where c is a constant.
51. f(t) = cosh t.
52. f(t) = sinh t.
53. Suppose p > 1. Recall the denition of the gamma function from above. Show that the
Laplace transform of t
p
is F(s) = (p + 1)/s
p+1
. In particular, if p is a positive integer,
then F(s) = p!/s
p+1
.
2.6. NUMERICAL TECHNIQUES 179
2.6 Numerical Techniques
for Approximating Integrals
In this section, it may seem like were backing up. In Section 2.2, we looked at Riemann
sums, then, in Section 2.3, we took limits of Riemann sums to dene the denite integral, the
continuous sum of innitesimal contributions. Then, after seeing how tedious it is to calculate
integrals as limits of Riemann sums, in Section 2.4, we presented the Fundamental Theorem of
Calculus, which tells us that, if we have an anti-derivative F of a continuous function f, then
its easy to calculate the values of denite integrals of f, in terms of F. Great. So, whats our
problem?
Our problem is that there are continuous functions f for which we cannot produce man-
ageable anti-derivatives, and so we cannot use the Fundamental Theorem to calculate denite
integrals for such f. The classic example is f(x) = e
x
2
. Denite integrals of this function
are of fundamental importance in probability and statistics, and yet, as we mentioned in Re-
mark 1.1.23, e
x
2
has no elementary anti-derivative.
Thus, while the rst part of the Fundamental Theorem, Theorem 2.4.7, tells us that denite
integrals of continuous functions on closed intervals exist, our question, for functions f without
nice anti-derivatives is: can we approximate
_
b
a
f(x) dx in a better way than by using simply
Riemann sums?
The answer to this question is: yes. We will look at two rules for approximating the
values of denite integrals: the Trapezoidal (or Trapezoid) Rule and Simpsons Rule. Both of
these rules use summations that look similar to Riemann sums, but the summations are, in fact,
not Riemann sums.
The Trapezoidal Rule is very easy to derive. We approximate the graph of f, the function
we want to integrate, by using line segments between points on the graph, and (assuming for the
convenience of this discussion that f 0) we then nd the area of the region, a trapezoid, below
each line segment and above the relevant interval. Adding these areas gives an approximation
of the corresponding denite integral.
It turns out that, for many functions, the approximation of the denite integral using the
Trapezoidal Rule is worse than the approximation using midpoint Riemann sums (Riemann
sums in which all subintervals have the same length, and the sample points are the midpoints
of the subintervals; see Example 2.2.8). Hence, in some sense, the Trapezoidal Rule is useless.
However, Simpsons Rule is just as easy to use as the Trapezoidal Rule, and yet, approxi-
mating an integral using Simpsons Rule is usually stunningly more accurate than using the
Trapezoidal Rule or using midpoint Riemann sums. Simpsons Rule is based on approximating
180 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
the graph of f with a collection of parabola segments, instead of with the line segments used
in the Trapezoidal Rule. Since graphs of typical functions curve, and parabolas curve, it should
seem reasonable that Simpsons Rule is typically more accurate than the Trapezoidal Rule.
Below, we rst adopt some notation and remind you what a midpoint Riemann sum is. After
that, we discuss and derive the Trapezoidal Rule, and then discuss and state Simpsons Rule,
leaving the actual derivation of Simpsons Rule for the Technical Matters section, Section 2.A.
We also give statements regarding bounds on the errors when you use the three approximations;
the proofs of these bounds are also in Section 2.A. We rst give examples in which we do not
consider the error bounds. Finally, we give an extended example where we discuss the error,
and bounds on the error, in more detail.
Suppose that we have a function f on a closed interval [a, b].
We subdivide the interval [a, b] into n subintervals of equal length x = (b a)/n. For 0
k n, the corresponding partition of [a, b], i.e., the collection of endpoints of the subintervals,
is given by x
k
= a +kx, that is,
x
0
= a, x
1
= a + x, x
2
= a + 2x, . . . , x
n1
= a + (n 1)x, x
n
= a +nx = b.
For 1 k n, the k-th subinterval is the closed interval [x
k1
, x
k
]; its midpoint is s
k
=
(x
k1
+x
k
)/2.
If A is an approximation of
_
b
a
f(x) dx, then we refer to the dierence E =
_
b
a
f(x) dx A
as the error in the approximation.. As we usually care about whether A is within plus or minus
some amount of the actual value of the integral, it is usually the absolute value of E that we
are interested in. We refer to the absolute value of the error as the absolute error.
We should remark that we will give all our calculations below to 12 decimal places. This may
seem like ridiculous precision; however, when youre trying to compare various approximations to
each other, you want to go out to enough decimal places to actually see where the approximations
dier.
We rst use midpoint Riemann sums to approximate the denite integral. Graphically, we
view the midpoint Riemann sum (for f 0) as shown in Figure 2.36.
2.6. NUMERICAL TECHNIQUES 181
x
0
x
1
x
2
x
3
x
4
x
0
x
1
x
2
x
3
x
4
Figure 2.36: A midpoint Riemann sum.
Denition 2.6.1. (Midpoint Approximation for Integrals) Using the notation above,
the midpoint approximation of
_
b
a
f(x) dx, using n subintervals, is
_
b
a
f(x) dx x
_
f
_
x
0
+x
1
2
_
+ f
_
x
1
+x
2
2
_
+ + f
_
x
n1
+x
n
2
__
.
Example 2.6.2. Consider the integral
_
1
0
e
x
2
dx. A calculator will tell you, correctly, that, to
12 decimal places, the value of this denite integral is 0.746824132812. What approximations
do we obtain from the Midpoint Approximation using n = 2 and n = 4?
When n = 2, we have x = (1 0)/2 = 1/2, x
0
= 0, x
1
= 1/2, and x
2
= 1 and, of course,
f(x) = e
x
2
.
The Midpoint Approximation is
_
1
0
e
x
2
dx
1
2
_
e
(1/4)
2
+e
(3/4)
2
_
,
which, to 12 decimal places is 0.754597943772. Thus, the absolute error here is

0.746824132812 0.754597943772

= 0.00777381096;
182 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
not great, but pretty good, considering that we used only 2 subintervals.
Lets try n = 4. Now, x = (1 0)/4 = 1/4, x
0
= 0, x
1
= 1/4, x
2
= 1/2, x
3
= 3/4, and
x
4
= 1.
Our new Midpoint Approximation is
_
1
0
e
x
2
dx
1
4
_
e
(1/8)
2
+e
(3/8)
2
+e
(5/8)
2
+e
(7/8)
2
_
,
which, to 12 decimal places is 0.748747131891. Thus, the absolute error here is

0.746824132812 0.748747131891

= 0.001922999079;
this is roughly 1/4 of the absolute error that we had when we used n = 2.
Now lets look at the Trapezoidal Rule. As we stated at the beginning of the section,
the Trapezoidal Rule does not involve Riemann sums, but rather something that looks similar
to Riemann sums. The idea of the Trapezoidal Rule is simple; over each of our subintervals
[x
k1
, x
k
], you approximate the function f by the unique linear function L
k
(x) = m
k
x + b
k
whose graph passes through the points (x
k1
, f(x
k1
)) and (x
k
, f(x
k
)), i.e., on the graph of
f, you connect the dots (with line segments) between the points of the graph of f which
correspond to the ends of the subinterval.
x
0
x
1
x
2
x
3
x
4
x
0
x
1
x
2
x
3
x
4
Figure 2.37: Typical areas involved in the Trapezoidal Rule.
At this point, instead of integrating f, you calculate each of the integrals I
k
of the linear
2.6. NUMERICAL TECHNIQUES 183
functions L
k
(x) = m
k
x + b
k
over the interval [x
k1
, x
k
], and then you add these integrals
together. What you obtain for I
k
is
I
k
=
x
2
_
f(x
k1
) +f(x
k
)
_
.
If f is positive, you should recognize this as the area of a trapezoid: one half the height (here,
its the width) times the sum of the lengths of the bases. Adding these together, we nd the
approximation
_
b
a
f(x) dx I
1
+ I
2
+ I
3
+ + I
n1
+ I
n
=
x
2
_
f(x
0
) +f(x
1
)
_
+
x
2
_
f(x
1
) +f(x
2
)
_
+
x
2
_
f(x
2
) +f(x
3
)
_
+ +
x
2
_
f(x
n2
) +f(x
n1
)
_
+
x
2
_
f(x
n1
) +f(x
n
)
_
.
Factoring out the
x
2
, and combining the pairs of overlapping terms everywhere, except for at
the f(x
0
) and f(x
n
) terms, we obtain the following approximation.
Denition 2.6.3. (Trapezoidal Rule) Using the notation above, the Trapezoidal Rule
Approximation of
_
b
a
f(x) dx, using n subintervals, is
_
b
a
f(x) dx
x
2
_
f(x
0
) + 2f(x
1
) + 2f(x
2
) + + 2f(x
n1
) +f(x
n
)
_
.
Note that the pattern of the coecients for f(x
k
) in the Trapezoidal Rule is that the rst
and last coecients are 1s and, aside from that, the coecients are always 2s.
Example 2.6.4. As in Example 2.6.2, consider the integral
_
1
0
e
x
2
dx. Recall that, to 12
decimal places, the value of this denite integral is 0.746824132812. What approximations do
we obtain from the Trapezoidal Rule using n = 2 and n = 4?
When n = 2, as before, we have x = (1 0)/2 = 1/2, x
0
= 0, x
1
= 1/2, and x
2
= 1.
184 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
The Trapezoidal Rule Approximation is
_
1
0
e
x
2
dx
1
4
_
e
(0)
2
+ 2e
(1/2)
2
+e
(1)
2
_
,
which, to 12 decimal places is 0.731370251829. Thus, the absolute error here is

0.746824132812 0.731370251829

= 0.015453880983.
Note that this is roughly double the absolute error that we had when we used the midpoint
Riemann sum with n = 2.
What about when n = 4? As before, x = (1 0)/4 = 1/4, x
0
= 0, x
1
= 1/4, x
2
= 1/2,
x
3
= 3/4, and x
4
= 1.
Our new Trapezoidal Rule Approximation is
_
1
0
e
x
2
dx
1
8
_
e
(0)
2
+ 2e
(1/4)
2
+ 2e
(1/2)
2
+ 2e
(3/4)
2
+e
(1)
2
_
,
which, to 12 decimal places is 0.742984097800. Thus, the absolute error here is

0.746824132812 0.742984097800

= 0.003840035012.
As with the Midpoint Approximation, this new approximation, using the Trapezoidal Rule with
n = 4 has roughly 1/4 of the absolute error that we had when we used the Trapezoidal Rule
with n = 2. However, you should also notice that it is still true that the absolute error, with
n = 4, using the Trapezoidal Rule, is roughly double the error, with n = 4, using the midpoint
Riemann sum.
What is Simpsons Rule? In a sense, Simpsons Rule is the next step after the Trapezoidal
Rule. The Trapezoidal Rule takes pairs of successive points on the graph of f, as determined by
the endpoints of the subintervals. Two points determine a line, and so determine a linear function
L = mx +b. On the k-th subinterval, we approximate f by the linear function L
k
= m
k
x +b
k
2.6. NUMERICAL TECHNIQUES 185
and integrate this linear function over the integral [x
k1
, x
k
], instead of integrating f. Then, we
add together the approximations over all of the subintervals.
Simpsons Rule uses the fact that three points determine a parabola (or a line, if the points
are collinear), and so determine a unique polynomial q = q(x) = ax
2
+bx+c of degree less than
or equal to 2. Simpsons Rule takes three successive points on the graph of f, as determined
by the endpoints of the subintervals, and approximates f by the corresponding polynomial
q = q(x) = ax
2
+ bx + c over the two subintervals whose endpoints correspond to the three
points we took on the graph. For instance, over the two successive subintervals [x
0
, x
1
] and
[x
1
, x
2
], i.e., over the interval [x
0
, x
2
], Simpsons Rule approximates f by the unique function
q(x) = ax
2
+ bx + c whose graph passes through the three points (x
0
, f(x
0
)), (x
1
, f(x
1
)), and
(x
2
, f(x
2
)).
When you determine the function q(x) and integrate it over [x
0
, x
2
], what you nd is relatively
simple. You get
x
3
_
f(x
0
) +4f(x
1
) +f(x
2
)
_
. (See Proposition 2.A.14.) We can do this for pairs
of subintervals, and approximate the entire integral
_
b
a
f(x) dx, provided that n, the number
of subintervals, is even.
x
0
x
1
x
2
x
3
x
4
Figure 2.38: Typical areas involved in Simpsons Rule.
Note that we had to make Figure 2.38 unusually large for you to have any hope of seeing
that the dotted parabolas do not exactly t the graph.
Assuming that n is even, if we add together the contributions over pairs of intervals, the
186 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
approximation that we obtain is
_
b
a
f(x) dx
x
3
_
f(x
0
) + 4f(x
1
) +f(x
2
)
_
+
x
3
_
f(x
2
) + 4f(x
3
) +f(x
4
)
_
+ +
x
3
_
f(x
n4
) + 4f(x
n3
) +f(x
n2
)
_
+
x
3
_
f(x
n2
) + 4f(x
n1
) +f(x
n
)
_
.
Factoring out the
x
3
, and combining the pairs of overlapping terms everywhere, except for at
the f(x
0
) and f(x
n
) terms, we obtain the following approximation.
Denition 2.6.5. (Simpsons Rule) Using the notation above, the Simpsons Rule
approximation of
_
b
a
f(x) dx, using n subintervals, where n is even, is
_
b
a
f(x) dx
x
3
_
f(x
0
) + 4f(x
1
) + 2f(x
2
) + 4f(x
3
) + 2f(x
4
) + + 4f(x
n1
) +f(x
n
)
_
.
Note that the pattern of the coecients for f(x
k
) in Simpsons Rule is that the rst and last
coecients are 1s and, aside from that, the coecients alternate 4, 2, 4, 2, etc. The next-to-last
coecient will always be a 4, due to the fact that n is even.
Example 2.6.6. As in our previous two examples, consider the integral
_
1
0
e
x
2
dx. Recall that,
to 12 decimal places, the value of this denite integral is 0.746824132812. What approximations
do we obtain from Simpsons Rule using n = 2 and n = 4?
When n = 2, we still have x = (1 0)/2 = 1/2, x
0
= 0, x
1
= 1/2, and x
2
= 1.
The Simpsons Rule Approximation is
_
1
0
e
x
2
dx
1
6
_
e
(0)
2
+ 4e
(1/2)
2
+e
(1)
2
_
,
which, to 12 decimal places is 0.747180428910. Thus, the absolute error here is

0.746824132812 0.747180428910

= 0.000356296098.
2.6. NUMERICAL TECHNIQUES 187
Wow! This absolute error is roughly 1/20 of the absolute error using n = 2 for the midpoint
Riemann sum, and is even roughly 1/5 of the absolute error using n = 4 for the midpoint
Riemann sum (and, remember, the midpoint Riemann sums gave better approximations than
the Trapezoidal Rule).
Thus, what we see is that the approximation via Simpsons Rule, even with a smaller n
value, is signicantly more accurate than those obtained from midpoint Riemann sums or the
Trapezoidal Rule. This is typical.
Lets see what happens when n = 4. As before, x = (1 0)/4 = 1/4, x
0
= 0, x
1
= 1/4,
x
2
= 1/2, x
3
= 3/4, and x
4
= 1.
Our new Simpsons Rule Approximation is
_
1
0
e
x
2
dx
1
12
_
e
(0)
2
+ 4e
(1/4)
2
+ 2e
(1/2)
2
+ 4e
(3/4)
2
+e
(1)
2
_
,
which, to 12 decimal places is 0.746855379791. Thus, the absolute error here is

0.746824132812 0.746855379791

= 0.000031246979.
This new approximation, using Simpsons Rule with n = 4 has roughly 1/10 of the absolute
error that we had when we used Simpsons Rule with n = 2.
It is reasonable to ask if it is possible to make precise in what sense midpoint Riemann sums,
the Trapezoidal Rule, and Simpsons Rule are reasonable approximation methods. Basically,
the question is: what can an instructor say to a student who wants credit for claiming
_
1
0
e
x
2
dx
is approximately 1, 000, 000, its just that the error is really large?
A possible answer is provided by the following three theorems, which we prove in the Techni-
cal Matters section, Section 2.A. These theorems give upper bounds on the absolute error when
approximating an integral by using midpoint Riemann sums, the Trapezoidal Rule, and/or
Simpsons Rule, provided that the integrand possesses enough derivatives. In particular, these
theorems imply that, for all three approximation techniques, the error approaches zero as n
approaches .
188 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
Theorem 2.6.7. Suppose that a < b and the midpoint Riemann sum with n subintervals of
equal length x is used to approximate
_
b
a
f(x) dx.
If f

(x) exists for all x in some open interval containing the closed interval [a, b] and,
if there exists a number M 0 such that, for all x in [a, b], [f

(x)[ M, then the absolute

value of the error, E
mdpt
, satises the inequality
[E
mdpt
[
M(x)
3
n
24
=
M(b a)
3
24n
2
.
Proof. See Theorem 2.A.15.
If, in fact, f

above is continuous on the interval [a, b], then the best, i.e., smallest, value
that you can use for M above is the maximum value of f

on [a, b].
Theorem 2.6.8. Suppose that a < b and the Trapezoidal Rule with n subintervals of equal
length x is used to approximate
_
b
a
f(x) dx.
If f

(x) exists for all x in some open interval containing the closed interval [a, b] and,
if there exists a number M 0 such that, for all x in [a, b], [f

(x)[ M, then the absolute

value of the error, E
trap
, satises the inequality
[E
trap
[
M(x)
3
n
12
=
M(b a)
3
12n
2
Proof. See Theorem 2.A.16.
Remark 2.6.9. Note that our upper bound for the absolute value of the error in approximating
with the Trapezoidal Rule is precisely twice the upper bound that we had for approximating
with midpoint Riemann sums. This is why references frequently say that midpoint Riemann
sums generally provide a better approximation to denite integrals than does the Trapezoidal
Rule. Understand, however, that the general upper bounds on the absolute values of the errors
that appear in the theorems dont have to be very close to the true absolute values of the errors;
they are simply gross upper bounds.
2.6. NUMERICAL TECHNIQUES 189
In particular, the absolute value of the error using the Trapezoidal Rule can be strictly
smaller than what you obtain from midpoint Riemann sums. Thats just not what we expect,
in general.
Theorem 2.6.10. Suppose that a < b, n is even, and Simpsons Rule with n subintervals
of equal length x is used to approximate
_
b
a
f(x) dx.
If f
(4)
(x) exists for all x in some open interval containing the closed interval [a, b] and,
if there exists a number M 0 such that, for all x in [a, b], [f
(4)
(x)[ M, then the absolute
value of the error, E
Simp
, satises the inequality
[E
Simp
[
M(x)
5
n
180
=
M(b a)
5
180n
4
.
Proof. See Theorem 2.A.17.
Remark 2.6.11. If f(x) is a quadratic polynomial, then it should not be surprising that
Simpsons Rule would give the exact value of the integral; after all, Simpsons Rule is obtained
by approximating the given function with quadratic functions.
What is somewhat surprising is that, even if f(x) is a cubic polynomial, the error bound
above tells us that Simpsons Rule still yields the exact value of the integral, for the fourth
derivative of a cubic polynomial is 0, and so the M in the error statement could be taken to be
0.
Lets look at an example for which we can produce an easy anti-derivative, and can easily
nd bounds on the second and fourth derivatives, so that we can see how two dierent approx-
imations compare with the exact answer. Understand, however, that the true value of these
approximations is in cases like
_
1
0
e
x
2
dx, where we cant produce a convenient anti-derivative.
190 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
Example 2.6.12. Consider
_
2
1
1
x
2
dx =
_
2
1
x
2
dx =
x
1
1

2
1
=
1
2
+ 1 =
1
2
.
Lets approximate
_
2
1
1
x
2
dx using midpoint Riemann sums, the Trapezoidal Rule, and Simp-
sons Rule with both n = 2 and n = 4. First, though, lets go ahead and calculate the theoretical
error bounds, so that we can see how our actual errors compare with the bounds. We will also
calculate how big we would need to pick n to have an error bound in each of the three approxi-
mations that would guarantee an error of less than 0.00001.
So, we have f(x) = 1/x
2
= x
2
, a = 1, and b = 2. We calculate the derivatives that we need
for the error bounds: f

(x) = 2x
3
, f

(x) = 6x
4
, f

(x) = 24x
5
, and f
(4)
(x) = 120x
6
.
Thus, the maximum value of [f

(x)[ on the interval [1, 2] is 6, and the maximum value of [f

(4)
(x)[
on [1, 2] is 120.
Therefore, when n = 2, we know that
[E
mdpt
[
6(2 1)
3
24 2
2
=
1
16
= 0.0625;
[E
trap
[
6(2 1)
3
12 2
2
=
1
8
= 0.125;
and
[E
Simp
[
120(2 1)
5
180 2
4
=
1
24
= 0.0416.
When n = 4, we nd
[E
mdpt
[
6(2 1)
3
24 4
2
=
1
64
= 0.015625;
[E
trap
[
6(2 1)
3
12 4
2
=
1
32
= 0.03125;
and
[E
Simp
[
120(2 1)
5
180 4
4
=
1
384
= 0.00260416.
2.6. NUMERICAL TECHNIQUES 191
You need to realize that these numbers are bounds on the possible amounts of error, not
the actual errors. In fact, theres a very important point here; if we know precisely what
the error in using an approximation for an integral is, then we know the actual value of the
integral; take the approximation and add/subtract the error. Of course, in this example,
we do know the actual value of the integral, but well write it again the value of these
approximation techniques is that we can use them in cases where we do not have a way to
calculate the exact integral. Thus, the most that we can hope for, typically, is to calculate
a nice bound on the error, not to calculate the error itself.
How many subintervals would we need to use with each of the approximations to produce
bounds on the absolute values of the errors that would guarantee our approximation is within
0.00001?
With midpoint Riemann sums, we would need
[E
mdpt
[
6(2 1)
3
24 n
2

1
100, 000
,
that is, we must require 4n
2
100, 000. Thus, we would need n

25, 000 158.11, i.e., the
smallest n that makes the upper bound on the absolute value of the error less than or equal to
0.00001 is n = 159.
It is possible that a smaller n actually yields an approximation within the desired amount,
we just cant determine such an n from our theoretical bound.
With the Trapezoidal Rule, we would need
[E
trap
[
6(2 1)
3
12 n
2

1
100, 000
,
that is, we must require 2n
2
100, 000. Thus, we would need n

50, 000 223.61, i.e., the
smallest n that makes the upper bound on the absolute value of the error less than or equal to
0.00001 is n = 224.
Again, it is possible that a smaller n actually yields an approximation within the desired
amount, we just cant determine such an n from our theoretical bound.
With Simpsons Rule, we would need
[E
Simp
[
120(2 1)
5
180 n
4

1
100, 000
,
192 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
that is, we must require 3n
4
/2 100, 000. Thus, we would need n 16.07, i.e., the smallest
n that makes the upper bound on the absolute value of the error less than or equal to 0.00001
would be n = 17 if we didnt need an even n for Simpsons Rule, but we do. So, the smallest n
that we can actually use for Simpsons Rule is n = 18.
Well write it for a nal time: it is possible that a smaller n actually yields an approximation
within the desired amount, we just cant determine such an n from our theoretical bound.
You can see from the numbers above why Simpsons Rule is the preferred approximation
method, and why, in general, we prefer midpoint Riemann sums to the Trapezoidal Rule. All
three approximations are roughly equally as complicated to calculate for a xed n, and yet, we
can know that Simpsons Rule with n = 18 gives us a bound on the error that requires n = 159
for midpoint Riemann sums, and n = 224 for the Trapezoidal Rule. This is typical.
Okay - its time to actually calculate the approximations of
_
2
1
1
x
2
dx when n = 2 and n = 4.
n = 2:
x = (b a)/2 = 1/2, x
0
= 1, x
1
= 3/2, x
2
= 2.
midpoint Riemann sum:
1
2
_
1
(5/4)
2
+
1
(7/4)
2
_
=
1
2
_
16
25
+
16
49
_
0.483265306122.
Trapezoidal Rule:
1
4
_
1
1
2
+ 2
1
(3/2)
2
+
1
2
2
_
= 0.53472.
Simpsons Rule:
1
6
_
1
1
2
+ 4
1
(3/2)
2
+
1
2
2
_
= 0.504629.
As the actual value of the integral is 0.5, we see that the absolute values of the errors
in the three cases are (to within 12 decimal places), respectively, 0.016734693878, 0.03472, and
0.004629. Recall that the upper bounds on these, which we calculated earlier, were, respectively,
0.0625, 0.125, and 0.0416. What we see is that the upper bounds are certainly upper bounds,
but that these upper bounds are not particularly close to the actual absolute values of the errors.
This, too, is typical.
2.6. NUMERICAL TECHNIQUES 193
n = 4:
x = (b a)/4 = 1/4, x
0
= 1, x
1
= 5/4, x
2
= 3/2, x
3
= 7/4, x
4
= 2.
midpoint Riemann sum:
1
4
_
1
(9/8)
2
+
1
(11/8)
2
+
1
(13/8)
2
+
1
(15/8)
2
_
0.495547936480.
Trapezoidal Rule:
1
8
_
1
1
2
+ 2
1
(5/4)
2
+ 2
1
(3/2)
2
+ 2
1
(7/4)
2
+
1
2
2
_
= 0.508993764172.
Simpsons Rule:
1
12
_
1
1
2
+ 4
1
(5/4)
2
+ 2
1
(3/2)
2
+ 4
1
(7/4)
2
+
1
2
2
_
= 0.500417611489.
Here, we see that the absolute values of the errors in the three cases are (to within 12 decimal
places), respectively, 0.00445206352, 0.0089937641720, and 0.000417611489. The upper bounds
on these, which we calculated earlier, were, respectively, 0.015625, 0.03125, and 0.00260416.
Again, our upper bounds are clearly upper bounds, but they are not very close to the actual
absolute values of the errors.
Remark 2.6.13. Before leaving this section, we should make a nal remark. These days, all
scientic calculators can calculate denite integrals. How do they do this? They use a form of
Simpsons Rule, with many, very small, subintervals, so that the result is typically accurate to
within the number of signicant digits that the display can handle.
2.6.1 Exercises
For the denite integrals in Exercises 1 - 15, calculate (a) the exact value of the
denite integral, (b) a midpoint approximation, (c) a trapezoidal approximation,
194 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
and (d) an approximation using Simpsons Rule. For each of the approximations,
use n = 4. View (a) as an opportunity to practice integration techniques from the
previous chapters.
1.
_
4
0
5y
y 7
dy.
2.
_
4
0
_
x
2
+ 16 dx.
3.
_
10
6
9x 10
x
2
3x 10
dx.
4.
_
4
4
cosh(4s) ds.
5.
_
6
2
4t
2
+ 6t 2
t
2
1
dt.
6.
_
12
0
3

x
2
+ 4
dx.
7.
_
/6
0
tan(2) d.
8.
_
5
1
ln hdh.
9.
_
6
2
3z
3
+ 5z
2
+ 5z 5
z
3
z
2
dz.
10.
_
6
0
_
36 t
2
dt.
11.
_
16
12
24

m
2
9
dm.
12.
_
1
0
2
du
(144 +u
2
)
2
du.
13.
_
7
3
2p
2
+ 1
p
3
+p
dp.
14.
_
2
0
sin
2
d.
15.
_
0
6
_
x
2
+ 12x + 45 dx.
2.6. NUMERICAL TECHNIQUES 195
16. Prove that the midpoint approximation is in fact equal to
_
b
a
mxdx for arbitrary n. Hint:
show
_
xi+1
xi
mx +b dx is equal to the area of the rectangle approximating it.
Determine n, the number of intervals necessary to give an approximation of the
denite integral using the specied method with absolute error less than that given.
17.
_
/4
0
sin(4) d, Midpoint, Error< 0.001.
18.
_
/4
0
sin(4) d, Trapezoid, Error< 0.001.
19.
_
14238
5134
97x
3
56x
2
+ 42ex 500 dx, Simpsons Rule, Error< 0.000001.
Estimate each of the following unpleasant denite integrals using (a) the Midpoint
Rule, (b) the Trapezoidal Rule and (c) Simpsons Rule.
20.
_
5
1
e
x
x
dx, n = 4.
21.
_

0
sin(sin t) dt. n = 6.
22.
_
/2
0
sin(y
2
) dy. n = 8. Note: this is one of the Fresnel integrals, introduced in the
exercises of the previous chapter.
23.
_
1
0
dz

z
4
+ 1
. n = 10.
24.
_
1/2
0
ln(1 +x)
1 +x
dx. n = 6.
25. Use the fact that ln x =
_
x
1
dt
t
and the Trapezoidal Rule with n = 10 to approximate ln 2.
26. Redo the previous problem using n = 10 and Simpsons rule.
27. Use the formula

4
=
_
1
0
dy
1 +y
2
to estimate . Use the trapezoid formula with n = 10.
28. Redo the previous problem using n = 10 and Simpsons rule.
In Exercises 29 - 32, calculate an upper bound for the absolute error of the denite
integral in terms of n when approximated with (a) the Midpoint Method and (b)
the Trapezoidal Rule.
196 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
29.
_
9
1
dy
y
.
30.
_
2
0
sin
2
t dt.
31.
_
12
0
e
z
dz.
32.
_
1
1
e
x
2
dx. Hint: To nd the maximum absolute value of the derivatives, note that all
derivatives are of the form p(x)e
x
2
where p(x) is a polynomial. It takes some fortitude to
factor these polynomials and correctly apply the rst or second derivative tests, but some
of the roots are obvious.
33. What is the upper bound on the absolute error of
_
3
0
e
x
2
dx when estimated with Simp-
sons Rule? Leave your answer in terms of n. Hint: even though Simpsons Rule requires
more derivative taking than the Midpoint Method, this problem is easier than the pre-
vious problem since the polynomial appearing in the 4th derivative is easier to analyze.
34. Generalize the previous problem. Let E(n, b) be the upper bound of Simpsons approx-
imation of
_
b
0
e
x
2
dx when [0, b] is subdivided into n equally spaced intervals. What is
E(n, b)?
35. Prove that if n = 4, the Simpsons approximation of
_
b
a
x
2
dx is exact.
A random variable is said to follow the standard normal distribution if it is normally
distributed with mean zero and standard deviation one ( = 0, =
2
= 1). Use this
information to solve Exercises 36 - 40.
36. Show that the density function of a standard normal distribution is
1

2
e
x
2
/2
.
37. a. Suppose a data set follows a standard normal distribution. Use the Midpoint Rule
with n = 4 to approximate the proportion of data between 0 and 1.
b. Based on your result in (a), approximately how much of the data falls between 1
and 1? Hint: the density function is symmetric about the y-axis.
38. a. Suppose a data set follows a standard normal distribution. Use the Trapezoidal Rule
with n = 4 to approximate the proportion of data between 0 and 2.
b. Based on your result in (a), approximately how much of the data falls between 2
and 2?
2.6. NUMERICAL TECHNIQUES 197
39. Suppose that the mature height of a certain tree species is normally distributed with
mean 16 feet and standard deviation of 2 feet. What proportion of trees are between 14
and 18 feet tall? Hint: given a normal distribution with parameters and , make the
substitution z =
x

before evaluating. You should be able to use one of the prior two
problems to help answer this question.
40. Suppose a test is administered to gauge the eect of alcohol on driving ability. A large
sample of individuals is given one alcoholic beverage, and their response time to a certain
stimulus is measured. Suppose the response time is normally distributed with mean three
seconds and standard deviation 0.6 seconds. What proportion of the individuals response
times falls between 1.8 and 4.2 seconds?
Up to this point, weve used the various methods in this chapter to approximate
denite integrals of a given continuous, or at least integrable, function. Since these
approximations depend only on knowledge of the function at nitely many points,
they may be adapted to applications where a denite integral is required, but the
integrand is known only at nitely many points.
41. Recall that if the force exerted on an object x meters from the origin is given by f(x), then
the work done in moving the object from a to b meters from the origin is W =
_
b
a
f(x) dx.
Suppose a 3 kilogram mass is being moved from 9 to 17 meters from the origin and the
acceleration of the particle is measured every 2 meters in meters per second per second as
shown in the table below. Use the Midpoint Method to approximate the total work done.
x Acceleration
10 2.5
12 5
14 2
16 -1
42. Suppose that a device is placed in a race car that measure the instantaneous velocity at
two second intervals. The data is shown below. Use Simpsons Rule to estimate the total
distance traveled by the car between t = 0 and t = 14 seconds.
198 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
Time (t) Velocity (miles/hour)
0 0
2 15
4 40
6 66
8 82
10 92
12 108
14 116
43. Redo the previous problem using the Trapezoidal Rule.
Estimate each of the following denite integrals using the specied method and
number of partitions. Then make a conjecture about the value of the (possibly
improper) integral. As a side note, each of these integrals have the property that
they can be calculated directly using a powerful tool from complex analysis called
residues.
44.
_
100
0
1
1 +x
4
dx, Midpoint Method, n = 10. Make a conjecture regarding the value of
_

0
1
1 +x
4
dx.
45.
_
10
10
cos z
1 +z
2
dz, Trapezoidal Rule, n = 10. Make a conjecture regarding the value of
_

cos z
1 +z
2
dz.
46.
_

dt
5 + 3 cos t
, Simpsons Rule, n = 8. Make a conjecture regarding the exact value of
_

dt
5 + 3 cos t
.
An ordinary or simple pendulum is one in which the path of the pendulum is
assumed to trace out a portion of a circle. This contrasts with a cycloidal pendulum
where the path is a cycloid. Assuming no friction, the period T of an ordinary
pendulum can be stated in terms of the elliptic integral
T = 2

L
g

_
1
1
dw
_
(1 w
2
)(1 w
2
sin
2
(A/2))
where L is the radius of the path, g is the acceleration due to gravity, and A is the
amplitude of oscillation. That is A measures the angular position of the pendulum
when it is rst dropped.
2.6. NUMERICAL TECHNIQUES 199
47. Suppose A = /2, g = 9.8 m/sec
2
and L = 1 meter. Use Simpsons Rule with n = 4
to estimate the period. As the integral is improper, make your estimation of the interval
[.95, .95].
48. Redo the previous problem with A = /4.
49. Argue that if the period is small, the period can be approximated by
2

L
g

_
1
1
dw

1 w
2
.
This expression is therefore independent of the amplitude. Evaluate this integral directly
using L = 1 meter.
50. Is your answer to the previous problem closer to the Simpsons approximation with A =
/2 or A = /4? Why does this make sense?
200 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
Appendix 2.A Technical Matters
In this section, we want to describe the technical framework for proving many of the theorems
on denite integrals. Still, there are many details beyond the scope of what we want to present
even in this technical appendix; our primary reference for these details is [4].
Theorem 2.A.1. The limit of Riemann sums, the denite integral, given in Denition 2.3.1 is
unique, if it exists.
In addition, if f is Riemann integrable on the interval [a, b],
_
b
a
f(x) dx = L, and
(T
n
, o
n
) is a sequence of sampled partitions of [a, b] such that lim
n
[[ T
n
[[ = 0, then
lim
n
!
Sn
Pn
(f) = L.
Proof. Clearly, proving the sequence statement proves the entire theorem.
Suppose that
_
b
a
f(x) dx exists and equals L, and suppose that (T
n
, o
n
) is a sequence of
sampled partitions of [a, b] such that lim
n
[[ T
n
[[ = 0.
Let > 0. Then, by denition of the Riemann integral, there exists > 0 such that, for all
partitions T = x
0
, . . . , x
n
of [a, b], such that [[T[[ < , for all sample sets s
1
, . . . , s
n
for T,

i=1
f(s
i
)x
i
L

< .
As lim
n
[[ T
n
[[ = 0, there exists a positive integer N such that, for all n N, [[ T
n
[[ < .
Hence, for all n N, if T
n
= x
0
, . . . , x
n
and o
n
= s
1
, . . . , s
n
,

i=1
f(s
i
)x
i
L

< ,
that is,
lim
n
!
Sn
Pn
(f) = L.
Before stating any further results, we remark that Theorem 2.3.6 tells us that unbounded
functions (see Denition 2.3.4) are not Riemann integrable. Consequently, below, we will be
2.A. TECHNICAL MATTERS 201
concerned with the case where f is a bounded function, until we get to the results on improper
integrals.
It is a dening property of the real numbers that every non-empty set of real numbers, which
is bounded above, has a least upper bound or supremum. It follows, by negating, that every non-
empty set of real numbers, which is bounded below, has a greatest lower bound or inmum. For a
bounded set V of real numbers, you should think of the supremum and inmum as the numbers
that want to be the maximum and minimum values, respectively, in V , though neither the
supremum or inmum need actually be contained in V .
Now suppose that f is a bounded function (see Denition 2.3.4) on a set E of real numbers.
This means that the set V = f(E) = f(x) [ x E is bounded, and so possesses a supremum
and inmum; we denote these, respectively, by M
f
(E) and m
f
(E).
The Extreme Value Theorem, (see [2] or [4]), tells us that a continuous function f on a closed
interval [a, b] is bounded and that, in fact, M
f
([a, b]) and m
f
([a, b]) are in the set f([a, b]), so
that they are, respectively, the maximum and minimum values of f on [a, b].
Now, suppose that f is bounded on the closed interval [a, b], and that T = (x
0
, . . . , x
n
)
is a partition of [a, b]. We would like to pick sample points which maximize or minimize the
Riemann sums, but such sample points need not exist (though they would if f were continuous).
However, we can fake it by using the supremum and inmum of f on each subinterval of the
partition. Thus, we dene the upper and lower sums U
f
(T) and L
f
(T), respectively, of f with
respect to the partition T to be
U
f
(T) =
n

i=1
M
f
([x
i1
, x
i
]) x
i
,
and
L
f
(T) =
n

i=1
m
f
([x
i1
, x
i
]) x
i
.
It follows immediately that U
f
(T) is the least upper bound of all of the Riemann sums of f,
using the partition T, and that L
f
(T) is the greatest lower bound of all of the Riemann sums
of f, using the partition T.
Note that it is immediate that, if a partition T

is a renement of a partition T, then

U
f
(T

) L
f
(T

) U
f
(T) L
f
(T).
202 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
The fundamental integrability result that we need, for bounded functions, is the following,
which we state without proof; see Theorem 3.2.7 of [4].
Theorem 2.A.2. Suppose that f is bounded on the interval [a, b]. Then, f is Riemann integrable
on [a, b] if and only if, for all > 0, there exists a partition T of [a, b] such that
U
f
(T) L
f
(T) < .
Given Theorem 2.A.2, the integrability of continuous functions follows almost immediately
from a basic fact: continuous functions on closed bounded intervals [a, b] are uniformly contin-
uous. That is:
Theorem 2.A.3. If f is continuous on the closed interval [a, b], then, for all > 0, there exists
> 0 such that, if x
1
and x
2
are in [a, b] and [x
1
x
2
[ < , then f(x
1
) f(x
2
) < .
Proof. See Theorem 2.2.12 of [4].
The uniform in uniformly continuous refers to the fact that, for a given > 0, there is a
uniform > 0 that works everywhere in the interval; normal continuity allows to vary as you
change one of the points.
It is now easy to prove:
Theorem 2.A.4. Suppose that f is continuous on the interval [a, b], where a < b. Then, f is
Riemann integrable on [a, b].
Proof. Let > 0. Then, /(b a) > 0, and so, by Theorem 2.A.3, there exists > 0 such that,
if x
1
and x
2
are in [a, b] and [x
1
x
2
[ < , then [f(x
1
) f(x
2
)[ < /(b a).
Let T = (x
0
, . . . , x
n
) be a partition of [a, b] with mesh < . We claim that U
f
(T)L
f
(T) < ,
which would prove that f is Riemann integrable by Theorem 2.A.2.
As [[ T [[ < , for any two points a
i
and b
i
in the subinterval [x
i1
, x
i
], [a
i
b
i
[ < . By
considering the points in the subinterval where f attains its maximum and minimum values, we
conclude that M
f
([x
i1
, x
i
]) m
f
([x
i1
, x
i
]) < /(b a) and, hence, that
n

i=1
_
M
f
([x
i1
, x
i
]) m
f
([x
i1
, x
i
])
_
x
i
<
n

i=1

b a
x
i
=

b a
n

i=1
x
i
= ,
2.A. TECHNICAL MATTERS 203
i.e., that
U
f
(T) L
f
(T) < .
We can now prove Theorem 2.3.8.
Theorem 2.A.5. Bounded, piecewise-continuous functions on closed intervals are Riemann
integrable.
Proof. By dividing the interval [a, b] into subintervals such that each subinterval has a discon-
tinuity at, at most, one endpoint. By Theorem 2.3.16, it is enough for us to prove that f is
Riemann integrable on such subintervals. Thus, we will assume that f is bounded on [a, b], with
[f(x)[ B, for x in [a, b], and has a single discontinuity at a. Note that, if B = 0, f is identically
0 on [a, b], and were nished; so assume that B > 0.
Now, let > 0. Consider rst the partition of [a, b] given by T = (a, a + /(4B), b). As f
is continuous on [a + /(4B), b], f is integrable on this subinterval, and so, by Theorem 2.A.2,
there exists a partition

T = (x
0
, . . . , x
n
) of [a +/(4B), b] such that U
f
(

T) L
f
(

T) < /2. Let

T

be the partition of [a, b] given by (a, x

0
, . . . , x
n
).
Then, from the denitions, it follows at once that
U
f
(T

) = M
f
([a, x
0
]) /(4B) + U
f
(

T)
and
L
f
(T

) = m
f
([a, x
0
]) /(4B) + L
f
(

T).
Also, note that M
f
([a, x
0
]) B, while m
f
([a, x
0
]) B, so that m
f
([a, x
0
]) B, and
M
f
([a, x
0
]) m
f
([a, x
0
]) 2B.
Therefore,
U
f
(T

) L
f
(T

) = M
f
([a, x
0
]) /(4B) + U
f
(

T) m
f
([a, x
0
]) /(4B) L
f
(

T) =
_
M
f
([a, x
0
]) m
f
([a, x
0
])
_

4B
+
_
U
f
(

T) L
f
(

T)

<

2
+

2
= ,
and we are nished by Theorem 2.A.2.
Theorem 2.A.6. Suppose that a < b.
204 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
1. If f is Riemann integrable on [a, b] and a c < d b, then f is Riemann integrable
on [c, d], i.e., if f is Riemann integrable on a given closed interval, then f is Riemann
integrable on any closed subinterval of the given interval.
2. Suppose that a < c < b. Then, f is Riemann integrable on [a, b] if and only if f is Riemann
integrable on [a, c] and [c, b] and, when these equivalent conditions hold,
_
b
a
f(x) dx =
_
c
a
f(x) dx +
_
b
c
f(x) dx.
Proof. Suppose that f is Riemann integrable on [a, b] and a c < d b. By Theorem 2.3.6, f
must be bounded on [a, b] and, hence, is bounded on [c, d].
Let > 0. By Theorem 2.A.2, there exists a partition T of [a, b] such that U
f
(T)L
f
(T) < .
Now, let T

be the partition of [c, d] formed from the points in [c, d] T, together with c and d.
Then,
U
f
(T

) L
f
(T

) U
f
(T) L
f
(T) < ,
and so, by Theorem 2.A.2, f is integrable on [c, d].
Suppose now that a < c < b.
If f is Riemann integrable on [a, b], then part (1) tells us that f is Riemann integrable on
[a, c] and [c, b].
Suppose that f is Riemann integrable on [a, c] and [c, b]. Then, f must be bounded on [a, c]
and [c, b] and, hence, on [a, b]. Let > 0. Then, /2 > 0. By Theorem 2.A.2, there exist
partitions T
1
and T
2
of [a, c] and [c, b], respectively, such that U
f
(T
1
) L
f
(T
1
) < /2 and
U
f
(T
2
) L
f
(T
2
) < /2. Then, the partition T

= T
1
T
2
of [a, b] is such that
U
f
(T

) L
f
(T

) <

2
+

2
= .
Therefore, by Theorem 2.A.2, f is Riemann integrable on [a, b].
Now, suppose that f is Riemann integrable on [a, c] and [c, b] and, hence, by the above,
Riemann integrable on [a, b]. Let L
1
=
_
c
a
f(x) dx, L
2
=
_
b
c
f(x) dx, and L =
_
b
a
f(x) dx. For
each integer n 1, let T
1
n
and T
2
n
be partitions of [a, c] and [c, b], respectively, with mesh < 1/n.
Let o
1
n
and o
2
n
be sample sets for the partitions T
1
n
and T
2
n
, respectively. Let T
n
= T
1
n
T
2
n
and o
n
= o
1
n
o
2
n
. Then, T
n
is a partition of [a, b] of mesh < 1/n, and o
n
is a sample set for
T
n
(possibly using the sample point c as a sample point for two dierent subintervals).
2.A. TECHNICAL MATTERS 205
Using Theorem 2.3.3 three times, we nd
L
1
+ L
2
= lim
n
!
S
1
n
P
1
n
(f) + lim
n
!
S
2
n
P
2
n
(f) = lim
n
!
Sn
Pn
(f) = L.
Theorem 2.A.7. (Linearity of Integration) Denite integration over a closed interval is a
linear operation, i.e., if f and g are Riemann integrable on [a, b], then, for all constants r and
s, the function rf +sg is Riemann integrable on [a, b], and
_
b
a
_
rf(x) +sg(x)
_
dx = r
_
b
a
f(x) dx + s
_
b
a
g(x) dx.
Proof. Suppose that f and g are Riemann integrable on [a, b]. We will prove that f + g is
Riemann integrable on [a, b] and that
_
b
a
_
f(x) +g(x)
_
dx =
_
b
a
f(x) dx +
_
b
a
g(x) dx.
That constants can be pulled out of integrals is left to you in Exercise 52.
Let [c, d] be any closed subinterval of [a, b]. It is trivial to see that
M
f+g
([c, d]) M
f
([c, d]) +M
g
([c, d])
and
m
f
([c, d]) +m
g
([c, d]) m
f+g
([c, d]).
Therefore, for any partition T of [a, b], U
f+g
(T) U
f
(T) + U
g
(T) and L
f
(T) + L
g
(T)
L
f+g
(T); hence,
U
f+g
(T) L
f+g
(T) [U
f
(T) L
f
(T)] + [U
g
(T) L
g
(T)] .
It follows immediately from Theorem 2.A.2 that f +g is integrable on [a, b].
Now, let T
n
be a partition of [a, b] of mesh < 1/n, and o
n
be a sample set for T
n
. Using
206 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
Theorem 2.3.3 three times, we nd
_
b
a
_
f(x) +g(x)
_
dx = lim
n
!
Sn
Pn
(f +g) = lim
n
!
Sn
Pn
(f) + lim
n
!
Sn
Pn
(g) =
_
b
a
f(x) dx +
_
b
a
g(x) dx.
Using Theorem 2.3.19 (see Theorem 3.3.1 of [4]), it is easy to conclude Theorem 2.3.14.
Theorem 2.A.8. Suppose that f and g are dened on a closed interval [a, b], and that, except
possibly for a nite set points in [a, b], f and g are equal at each point in [a, b].
Then, f is Riemann integrable on [a, b] if and only if g is, and when f and g are Riemann
integrable,
_
b
a
f(x) dx =
_
b
a
g(x) dx.
Proof. As f and g dier on at most a nite number of points, f is bounded if and only if g is
bounded. If both functions are unbounded, then Theorem 2.3.6 tells us that both functions are
not Riemann integrable.
Now, suppose that f and g are both bounded, and that f is Riemann integrable. Then, g f
is bounded, and is equal to 0 except at, possibly, a nite number of points, r
1
, . . . , r
p
. We claim
that
_
b
a
(g f)(x) dx = 0; if we show this, we are nished by Theorem 2.3.19, since we would
then have
_
b
a
f(x) dx =
_
b
a
f(x) dx +
_
b
a
(g f)(x) dx =
_
b
a
g(x) dx.
Let M be the maximum of [(g f)(r
i
)[ for 1 i p. Let T = (x
0
, . . . , x
n
) be a partition
of the interval [a, b], and let o = (s
1
, . . . , s
n
) be a set of sample points for T.
Then,

i=1
(g f)(s
i
)x
i

i=1
[(g f)(s
i
)[ [x
i
[ M [[ T [[,
and, hence,
lim
|| P ||0
!
S
P
(g f) = 0.
Theorem 2.A.9.
2.A. TECHNICAL MATTERS 207
Theorem 2.A.10. (Fundamental Theorem of Calculus, Part 1) Suppose that f is Rie-
mann integrable on [a, b] and is continuous at a point x
0
in (a, b). Then, the integral function
I
[a,b]
f
of f on [a, b] is dierentiable at x
0
and
_
I
[a,b]
f
_

(x
0
) = f(x
0
).
Thus, if f is continuous on [a, b], then I
[a,b]
f
is an anti-derivative of f on [a, b].
Proof. From the denition of the derivative, we have
_
I
[a,b]
f
_

(x
0
) = lim
xx0
I
[a,b]
f
(x) I
[a,b]
f
(x
0
)
x x
0
= lim
xx0
_
x
a
f(t) dt
_
x0
a
f(t) dt
x x
0
=
lim
xx0
_
x
x0
f(t) dt
x x
0
.
We must show that this limit equals f(x
0
). We shall show that the limit from the right equals
f(x
0
), and leave the nearly identical argument from the left as an exercise.
Let > 0. By the denition of f being continuous at x
0
, there exists > 0 such that, for all
x such that [x x
0
[ < , [f(x) f(x
0
)[ < . We claim that, if 0 < x x
0
< , then

_
x
x0
f(t) dt
x x
0
f(x
0
)

< , (2.4)
which is what it means that lim
xx
+
0

x
x
0
f(t) dt
xx0
= f(x
0
).
To prove the claim in Formula 2.4, x an x such that 0 < x x
0
< . Then, for all t in
[x
0
, x], we have 0 t x
0
x x
0
< , and so, [t x
0
[ < . Using the dening property of
again, we conclude that, for all t in [x
0
, x], [f(t) f(x
0
)[ < .
Therefore, for all t in [x
0
, x],
f(x
0
) f(t) f(x
0
) +.
208 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
By the monotonicity of integration, Theorem 2.3.20, and Proposition 2.3.12, we conclude that
(f(x
0
) )(x x
0
) =
_
x
x0
(f(x
0
) ) dt
_
x
x0
f(t) dt
_
x
x0
(f(x
0
) +) dt = (f(x
0
) +)(x x
0
).
Hence,
f(x
0
)
_
x
x0
f(t) dt
x x
0
f(x
0
) +,
i.e.,

_
x
x0
f(t) dt
x x
0
f(x
0
)

< ,
as we claimed in Formula 2.4.
The nal conclusion of the theorem follows from the rst one, together with the continuity
of I
[a,b]
f
, which follows from Theorem 2.4.3.
Theorem 2.A.11. The integral dened in Denition 2.5.11 is independent of the splitting into
subintervals (of the type described in the denition).
Proof. We will prove a lemma, from which the theorem follows immediately.
Suppose that f is continuous on the interval I = [a, b). Let c be such that a < c < b. We
claim that
_
b
a
f(x) dx exists if and only if
_
b
c
f(x) dx exists and, in this case,
_
b
a
f(x) dx =
_
c
a
f(x) dx +
_
b
c
f(x) dx.
This is easy:
_
b
a
f(x) dx = lim
db

_
d
a
f(x) dx = lim
db

_
_
c
a
f(x) dx +
_
d
c
f(x) dx
_
=
_
c
a
f(x) dx + lim
db

_
_
d
c
f(x) dx
_
=
_
c
a
f(x) dx +
_
b
c
f(x) dx.
2.A. TECHNICAL MATTERS 209
Theorem 2.A.12. Let I be an interval of the form [a, b) or (a, b], where we allow the intervals
[a, ) or (, b]. Suppose that, for all x in I, f(x) 0, and that, for all closed intervals [c, d]
contained in I, f is Riemann integrable on [c, d].
Then, if there exists a real number M (an upper bound) such that, for all [c, d] contained
in I,
_
d
c
f(x) dx M, then
_
b
a
f(x) dx converges, and what it converges to is less than, or equal
to, M; in other words, if there is an upper bound M on all of the
_
d
c
f(x) dx, then
_
b
a
f(x) dx
converges to the least such upper bound.
In particular, if
_
b
a
f(x) dx diverges, what it diverges to is .
Proof. We will deal with the case where I = [a, b); the other cases are entirely analogous.
For a d < b, dene the function F(d) =
_
d
a
f(x) dx, which exists by assumption. As f 0,
F is an increasing function on the interval I.
If F gets unboundedly large, then, since F is increasing, lim
db
F(d) = . If F is bounded
above by M, then, as F is increasing, F converges to the least upper bound of the set F(I),
which is M.
Lemma 2.A.13. Let q(x) = c
0
+c
1
x +c
2
x
2
+c
3
x
3
, where c
i
denotes a constant. Then,
_
h
h
q(x) dx =
h
3
_
6c
0
+ 2c
2
h
2
_
=
h
3
_
q(h) + 4q(0) +q(h)
_
.
Proof. This is a trivial calculation, which we leave to the reader.
Proposition 2.A.14. Let q(x) = ax
2
+bx+c. Fix x
0
. Let h > 0. Let x
1
= x
0
+h, x
2
= x
1
+h,
and y
i
= q(x
i
), for i = 1, 2, 3. Then,
_
x2
x0
q(x) dx =
h
3
_
y
0
+ 4y
1
+y
2
_
.
Proof. One can make the substitution u = x (x
0
+ h) into the integral and reduce oneself to
the case in Lemma 2.A.13, or do a messy, but simple, algebra problem. One has to show that
_
a(x
0
+ 2h)
3
3
+
b(x
0
+ 2h)
2
2
+c(x
0
+ 2h)
_

_
ax
3
0
3
+
bx
2
0
2
+cx
0
_
=
210 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
h
3
_
_
ax
2
0
+bx
0
+c

+ 4
_
a(x
0
+h)
2
+b(x
0
+h) +c

+
_
a(x
0
+ 2h)
2
+b(x
0
+ 2h) +c

_
.
We leave either verication as an exercise.
In the proofs below, for notational convenience, we apply the inequalities above in the case
where a = 0; the general case follows at once by making the substitution u = x a.
Theorem 2.A.15. Suppose that a < b and the midpoint Riemann sum with n subintervals of
equal length x is used to approximate
_
b
a
f(x) dx.
If f

(x) exists for all x in some open interval containing the closed interval [a, b] and, if
there exists a number M 0 such that, for all x in [a, b], [f

(x)[ M, then the absolute value

of the error, E
mdpt
, satises the inequality
[E
mdpt
[
M(x)
3
n
24
=
M(b a)
3
24n
2
Proof. We shall prove this for one subinterval, centered at 0. The general result follows by
re-centering the intervals, via substitution, and then adding, which leads to the multiplication
by n in the inequality. Let h = x/2.
We will show that

_
h
h
f(x) dx 2hf(0)

Mh
3
3
.
We must look ahead to the Taylor-Lagrange Theorem, Theorem 4.3.3,, which immediately
implies that: if f is twice-dierentiable on an open interval I around a point a, and M > 0 is a
constant such that, for all x in I, [f(x)[ M, then, for all x in I,

f(x)
_
f(a) +f

(a)(x a)
_

M
2
(x a)
2
; (2.5)
Now the proof is easy.

_
h
h
f(x) dx 2hf(0)

_
h
h
f(x) dx
_
h
h
_
f(0) +f

(0)x
_
dx

_
h
h

f(x) f(0) f

(0)x

dx
_
h
h
M
2
x
2
dx =
Mh
3
3
.
2.A. TECHNICAL MATTERS 211
Theorem 2.A.16. Suppose that a < b and the Trapezoidal Rule with n subintervals of equal
length x is used to approximate
_
b
a
f(x) dx.
If f

(x) exists for all x in some open interval containing the closed interval [a, b] and, if
there exists a number M 0 such that, for all x in [a, b], [f

(x)[ M, then the absolute value

of the error, E
trap
, satises the inequality
[E
trap
[
M(x)
3
n
12
=
M(b a)
3
12n
2
Proof. We shall prove this for one subinterval, centered at 0. The general result follows by
re-centering the intervals, via substitution, and then adding, which leads to the multiplication
by n in the inequality. Let h = x/2.
We will show that

_
h
h
f(x) dx 2h
f(h) +f(h)
2

2Mh
3
3
.
We integrate by parts twice.
_
h
h
f(x) dx = xf(x)

h
h

_
h
h
xf

(x) dx = hf(h) +hf(h)

_
h
h
xf

(x) dx =
h
_
f(h) +f(h)
_

_
1
2
x
2
f

(x)

h
h

1
2
_
h
h
x
2
f

(x) dx
_
.
Therefore,

_
h
h
f(x) dx 2h
f(h) +f(h)
2

=
1
2

h
2
f

(h) h
2
f

(h)
_
h
h
x
2
f

(x) dx

=
1
2

_
h
h
(h
2
x
2
)f

(x) dx

1
2
_
h
h
(h
2
x
2
)M dx =
M
2

4h
3
3
=
2Mh
3
3
.
212 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
Theorem 2.A.17. Suppose that a < b, n is even, and Simpsons Rule with n subintervals of
equal length x is used to approximate
_
b
a
f(x) dx.
If f
(4)
(x) exists for all x in some open interval containing the closed interval [a, b] and,
if there exists a number M 0 such that, for all x in [a, b], [f
(4)
(x)[ M, then the absolute
value of the error, E
Simp
, satises the inequality
[E
Simp
[
M(x)
5
n
180
=
M(b a)
5
180n
4
.
Proof. We shall prove this for two subintervals, centered at 0. The general result follows by
re-centering the intervals, via substitution, and then adding, which leads to the multiplication
by n/2 in the inequality. Let h = x.
Let
(t) =
_
t
t
f(x) dx
t
3
_
f(t) + 4f(0) + f(t)
_
.
We will show that there exists a c such that h < c < h and
(h) =
h
5
f
(4)
(c)
90
.
Let
(t) = (t)
_
t
h
_
5
(t).
One easily calculates that (0) =

(0) =

(0) = 0 and, for t ,= 0,

(t) =
2t
2
3
_
f

(t) f

(t)
2t
+
90
h
5
(h)
_
.
Now, (h) = 0, and so, by Rolles Theorem, there exists c
1
such that 0 < c
1
< h and

(c
1
) = 0. Applying Rolles Theorem again, we nd that there exists c
2
such that 0 < c
2
<
c
1
< h and

(c
2
) = 0. Applying Rolles Theorem yet again, we nd that there exists c
3
such
that 0 < c
3
< c
2
< c
1
< h and

(c
3
) = 0.
2.A. TECHNICAL MATTERS 213
Applying the Mean Value Theorem, we conclude that there exists c such that c
3
< c < c
3
and
f

(c
3
) f

(c
3
) = 2c
3
f
(4)
(c).
Combining the previous 3 displayed formulas, we conclude that
f
(4)
(c) +
90
h
5
(h) = 0,
i.e.,
(h) =
h
5
f
(4)
(c)
90
,
which is what we wanted to show.
214 CHAPTER 2. CONTINUOUS SUMS: THE DEFINITE INTEGRAL
Chapter 3
Applications of Integration
In this chapter, we will apply our results on anti-dierentiation, denite integrals, and the
Fundamental Theorem of Calculus to a wide variety of problems involving displacement, distance
traveled, area in the plane, volume, surface areas, mass, centers of mass, rotational inertia, work,
and hydrostatic pressure.
Throughout this chapter, we state many of our general results as Propositions, though, in
many cases, these Propositions could be used as Denitions. For instance, we will see the denite
integral of speed, with respect to time, yields the total distance traveled. We assume that you
have a preconceived notion of the distance traveled, and then conclude that that distance can
be calculated by integrating speed. However, we could, instead, dene the distance traveled by
the integral of the speed. Mathematically, this latter approach is more rigorous; it is, however,
somewhat intuitively unsatisfying.
Thus, throughout this chapter, we shall usually assume that we have predened physical
terms, such as distance traveled, volume, mass, etc., and state our integration formulas for these
quantities as Propositions.
We should remark that many of the applications in this chapter are actually pre-applications,
or what some people refer to as toy problems. A pre-application or toy problem is a problem
that is stripped of many physical complications or is not inherently of interest in and of itself,
but rather is designed mainly to provide some basic example of a fundamental idea that will
serve as a building block to tackling more-dicult actual applications.
215
216 CHAPTER 3. APPLICATIONS OF INTEGRATION
3.1 Displacement and Distance Traveled
in a Straight Line
In this section, we will apply the denite integral and the Fundamental Theorem of Calculus
to problems involving the net change in position, the displacement, of an object, and the total
distance traveled by the object. If the object changes its direction, these two quantities will not
be the same; the total distance traveled would be greater, while the displacement could, in fact,
turn out to be zero if the object ends up back where it started.
The second part of the Fundamental Theorem of Calculus, Theorem 2.4.10, tells us that, if
F is an anti-derivative of a continuous function f on the interval [a, b], then
_
b
a
f(t) dt = F(t)

b
a
= F(b) F(a).
This means that, if we start with a function g(t), which is dierentiable on an open interval
containing [a, b], and g

(t) is continuous, then

_
b
a
g

(t) dt =
_
b
a
dg
dt
dt = g(b) g(a).
Suppose now that we have an object moving in a straight line, on which weve chosen positive
and negative directions, i.e., suppose that we have an object moving along a coordinate. Let
p(t) denote the position (i.e., the coordinate value) of the object at time t.
Assuming that p is continuously dierentiable, we know that the velocity v = v(t) of the
object is the rate of change of the position, with respect to time, dp/dt, and the Fundamental
Theorem tells us that
_
b
a
v(t) dt = p(b) p(a).
The quantity on the right above is the change in the position of the object between times
t = a and t = b; recall that this is called the displacement of the object between times a and b.
Of course, when we refer to the velocity function v = v(t), we are implicitly assuming that
the position function is dierentiable, for, otherwise, the velocity has no good meaning. Thus,
3.1. DISPLACEMENT AND DISTANCE TRAVELED 217
we can summarize our discussion above by:
Proposition 3.1.1. If the velocity v = v(t), as a function of time t, of an object on a
coordinate axis is continuous on the interval [a, b], then the displacement of the object
between times t = a and t = b is given by
_
b
a
v(t) dt.
Example 3.1.2. A particle is moving in a straight line in such a way that its velocity v = v(t),
in m/s, at time t seconds, is given by
v = 3t
2
12t + 8.
What is the displacement of the particle between times 1 and 3 seconds? Between times 2
and 4 seconds? What is the displacement between times 1 and t seconds, for arbitrary t?
Solution:
To answer the rst two questions, we need to calculate
_
3
1
(3t
2
12t + 8) dt and
_
4
2
(3t
2
12t + 8) dt.
We want to apply the Fundamental Theorem, and so we need to nd the (or, an) anti-derivative
_
(3t
2
12t + 8) dt = 3
t
3
3
12
t
2
2
+ 8t +C = t
3
6t
2
+ 8t +C = t(t 2)(t 4) +C.
Therefore, the displacement between times 1 and 3 seconds is
_
3
1
(3t
2
12t + 8) dt = t(t 2)(t 4)

3
1
= 3 3 = 6 meters;
218 CHAPTER 3. APPLICATIONS OF INTEGRATION
this means that the particle ended up, at time 3 seconds, 6 meters in the negative direction from
where it started at time 1 second.
Note that we do not know where the particle is at times 1 and 3 seconds, i.e., we dont know
the position function p(t). However, except for the arbitrary constant +C above, we know
p(t), and thats enough to determine the change in the position.
The displacement of the particle between times 2 and 4 seconds is
_
4
2
(3t
2
12t + 8) dt = t(t 2)(t 4)

4
2
= 0 0 = 0 meters.
This does not mean that the particle did not move between times 2 and 4 seconds; it merely
means that the particle, at time 4 seconds, ended up at the same place where the particle was
at time 2 seconds. Of course, this means that the particle had to turn around at some point.
Do not confuse displacement, the net change in position, with the total distance traveled.
The displacement here, between times 2 and 4 seconds is 0. We shall calculate the total distance
traveled by this particle, between times 2 and 4 seconds, in the next example.
Finally, in this example, we were asked to nd the displacement of the particle between times
1 and t. This is easy, except that it would now be bad form to use t for the integration variable.
You should use essentially any other variable for the dummy variable of integration; the only
variables you should avoid are t and v, and maybe d would look confusing too. Well use z for
the integration variable. Then, we quickly nd that the displacement between times 1 and t
seconds is
_
t
1
(3z
2
12z +8) dz = z(z 2)(z 4)

t
1
= t(t2)(t4)(1)(1)(3) = t
3
6t
2
+8t3 meters.
In the example above, we ran into the issue of the distinction between displacement and total
distance traveled. This distinction arises because, if the velocity v of an object is continuous and
always positive on an interval of time [a, b], then the distance traveled by the object between
times a and b is the same as the displacement, namely
_
b
a
v dt. On the other hand, if the
velocity v of an object is continuous and always negative on an interval of time [a, b], then the
displacement will be negative; it will be the negation of the distance traveled by the object
between times a and b, i.e., the distance traveled will be
_
b
a
v dt =
_
b
a
v dt.
Thus, to nd the distance traveled, we want to integrate v on intervals where v is positive,
3.1. DISPLACEMENT AND DISTANCE TRAVELED 219
and we want to integrate v on intervals where v is negative. The way to write this in one
formula is to say that we want to integrate the absolute value [v[ in all cases.
Proposition 3.1.3. If the velocity v = v(t), as a function of time t, of an object on a
coordinate axis is continuous on the interval [a, b], then the (total) distance traveled by
the object between times t = a and t = b is given by
_
b
a
[v(t)[ dt.
Remark 3.1.4. Proposition 3.1.3 tells us, in a new way, something that we already knew. It
tells us that the total distance traveled, r = r(t), by an object, between some initial time t
0
and
some arbitrary time t, is given by
r(t) =
_
t
t0
[v(z)[ dz,
where we once again have used z as a new dummy variable, and were assuming that v is
continuous.
Now, the rst part of the Fundamental Theorem, Theorem 2.4.7, tells us that
dr
dt
= [v(t)[,
i.e., that the (instantaneous) speed of an object can either be dened as the absolute value of the
velocity or as the instantaneous rate of change of the distance traveled, with respect to time.
220 CHAPTER 3. APPLICATIONS OF INTEGRATION
However, it is important to remember that average speed need not be the absolute value
of the average velocity. For instance, an object that returns to where it started, after some
amount of time, will have zero displacement and, hence, zero average velocity, but not zero
average speed. Average speed is the average rate of change of the distance traveled, with
respect to time. Thus, it is probably best, in the average and/or instantaneous setting,
to dene speed as the rate of change of the distance traveled, with respect to time, and
then take it as a theorem that the instantaneous speed is equal to the absolute value of the
instantaneous velocity.
Lets look at an example of calculating the (total) distance traveled.
Example 3.1.5. In Example 3.1.2, we had a particle moving in a straight line in such a way
that its velocity v = v(t), in m/s, at time t seconds, was given by
v = 3t
2
12t + 8.
We found that the displacement of the particle between times 2 and 4 seconds is
_
4
2
(3t
2
12t + 8) dt = t(t 2)(t 4)

4
2
= 0 0 = 0 meters.
Now well ask a dierent question.
What was the distance traveled by the particle between 2 and 4 seconds?
Solution:
Its certainly easy to write the appropriate integral. The distance traveled by the particle
between 2 and 4 seconds was
_
4
2

3t
2
12t + 8

dt meters,
but how do we calculate the integral of the absolute value?
One thing that you denitely dont do is nd an anti-derivative of 3t
2
12t + 8, and then
take its absolute value, in hope of producing an anti-derivative of

3t
2
12t +8

. This would
be completely wrong.
3.1. DISPLACEMENT AND DISTANCE TRAVELED 221
Theres really only one thing to do; split up the integral into integrals over subintervals on
which 3t
2
12t + 8 is always 0 or is always 0. Then use that, if 3t
2
12t + 8 0, then

3t
2
12t +8

= 3t
2
12t +8 and, if 3t
2
12t +8 0, then

3t
2
12t +8

= (3t
2
12t +8).
As v(t) = 3t
2
12t + 8 is a continuous function, it can switch signs only at points where
it hits 0. We nd these points, then check the sign of v(t) in-between the zeroes. Setting
3t
2
12t + 8 = 0, and using the quadratic formula, we nd that the zeroes of v(t) occur where
t =
12

144 96
6
= 2
2

3
3
.
Obviously, 2 2

3/3 < 2. However, 2 + 2

3/3 is in the interval [2, 4]. Thus, the sign of v(t)

does not switch from positive to negative, or vice-versa, on the interval [2, 2 + 2

3/3] or on
the interval [2 + 2

3/3, 4]. As v(2) = 4 and v(4) = 8, we nd that v(t) 0 on the interval

[2, 2 + 2

3/3], and that v(t) 0 on the interval [2 + 2

3/3, 4].
Therefore, the distance traveled, in meters, by the particle between times 2 and 4 seconds
was
_
4
2

3t
2
12t + 8

dt =
_
2+2

3/3
2

3t
2
12t + 8

dt +
_
4
2+2

3/3

3t
2
12t + 8

dt =
_
2+2

3/3
2
(3t
2
12t + 8) dt +
_
4
2+2

3/3
(3t
2
12t + 8) dt.
Using plus or minus our anti-derivative from Example 3.1.2, we nd that the sum above
equals
t(t 2)(t 4)

2+2

3/3
2
+ t(t 2)(t 4)

4
2+2

3/3
=
_

_
2 + 2

3/3
__
2

3/3
__
2 + 2

3/3
_
0
_
+
_
0
_
2 + 2

3/3
__
2

3/3
__
2 + 2

3/3
_
_
=
16

3
9
+
16

3
9
=
32

3
9
6.1584.
It is somewhat interesting to note that, since the displacement between times 2 and 4 seconds
was 0, the average velocity between times 2 and 4 seconds was 0, but the average speed was
222 CHAPTER 3. APPLICATIONS OF INTEGRATION
approximately 6.1584/(4 2) = 3.0792 meters per second.
As you can see, the calculation of the distance traveled and/or average speed can be, and
usually is, signicantly more dicult than the calculation of the displacement and/or average
velocity.
We should make a nal remark in this section.
Remark 3.1.6. It may seem as though we have competing notions of what average velocity
and average speed mean.
Suppose that, at time t, we let p(t) denote the position of an object which is moving along
line, let r(t) denote the total distance traveled by the object, and let v(t) denote the velocity,
which we will assume is a continuous function of t.
Then, the denition of the average velocity and average speed of the object, between times
t = a and t = b (or, on the interval [a, b]), where a ,= b, are
average velocity =
p(b) p(a)
b a
and average speed =
r(b) r(a)
b a
.
However, in Denition 2.3.30, the average value of any Riemann integrable function f(t) on
the interval [a, b] was dened to be
1
b a
_
b
a
f(t) dt,
and we could consider the average value of the (instantaneous) velocity and speed functions,
v(t) and [v(t)[, respectively.
Do we need to worry that average velocity and average speed might refer to the average
values of the velocity and speed functions, instead of referring to our original denitions? No.
As we discussed in this section,
_
b
a
v(t) dt = p(b) p(a) and
_
b
a
[v(t)[ dt = r(b) r(a). Thus,
the average value of the velocity function v(t) equals the average velocity from our original
denition, and the average value of the speed function [v(t)[ equals the average speed from our
original denition.
3.1. DISPLACEMENT AND DISTANCE TRAVELED 223
3.1.1 Exercises
Throughout the exercises, assume that units of length are meters and that units of
time are given in seconds and, hence, that velocities are given in m/s and acceler-
ations in m/s
2
.
In Exercises 1 - 10, the velocity function of particle is given. Find the total dis-
placement of the particle between the times t
0
and t
1
.
1. v(t) = t
3
3t
2
+ 1, t
0
= 0, t
1
= 10.
2. v(t) = 1/(t 3) + cos 4t, t
0
= 2, t
1
= 4.
3. v(t) = cosh 3t, t
0
= 5, t
1
= 5.
4. v(t) = sinh 3t, t
0
= 5, t
1
= 5.
5. v(t) =
1

4t
2
64
, t
0
= 5, t
1
= 8.
6. v(t) = t
2
/(t
2
1), t
0
= 3, t
1
= 6.
7. v(t) = tan
2
t, t
0
= 0, t
1
= /4.
8. v(t) =
2t

12 t
2
, t
0
= 0, t
1
= 3.
9. v(t) =
4
25 + 36t
2
, t
0
= 0, t
1
= 1.
10. v(t) = (cos t)e
sin t
, t
0
= 0, t
1
= .
Calculate the total distance traveled by the particle over the given time interval
with the given velocity function.
11. v(t) = t
2
5t + 6, [0, 5].
12. v(t) = 2t
2
+ 12t + 1, [4, 4].
13. v(t) = sin 2t, [0, /2].
14. v(t) = e
t
1, [0, 2].
15. v(t) = e
3t
+t
2
+ cos
2
t + cosh t + 1, [4, 4].
224 CHAPTER 3. APPLICATIONS OF INTEGRATION
16. v(t) =
t
2
9t + 18
t
2
+ 3t + 2
, [0, 8].
17. v(t) =
6t

24 3t
2
, [2, 2].
18. v(t) =
_
4t
2
+ 4t + 10, [3, 5].
19. v(t) = ln(t/2), [1, 4].
20. v(t) = sin t + cos t, [0, 2].
In Exercises 21 - 25, calculate (a) the average velocity and (b) the average speed of
the particle traveling with velocity function v(t) over the given time interval.
21. v(t) = sin 3t, [/6, /3].
22. v(t) = t
2
+ 5t 14, [0, 12].
23. v(t) =
6t + 8
t
2
+t 6
, [2, 1].
24. v(t) = sec
3
4t, [/10, /12].
25. v(t) = t cosh t
2
, [2, 4].
In Exercises 26 - 30, a particle is moving with velocity v(t). Calculate the position
function in terms of the time t from the given initial time t
0
where the particles
position is p(t
0
). Assume that t > t
0
.
26. v(t) = t
2
+ 5t 24, p(0) = 12.
27. v(t) = t ln t, p(1) = 0.
28. v(t) = cot
2
t, p(/4) = 3. Assume t < .
29. v(t) =
t 1
t
2
+ 1
, p(1) = 5.
30. v(t) = sec t tan t, p(0) = 3. Assume t < /2.
In Exercises 31 - 34, a particle is moving with velocity v(t). Calculate the total
distance D(t) traveled by the particle between the specied starting time t
0
and an
arbitrary time t, where t t
0
.
31. v(t) = e
t
+ cos
2
t +t
4
, t
0
= 0.
32. v(t) = t
2
+ 7t 18, t
0
= 3.
3.1. DISPLACEMENT AND DISTANCE TRAVELED 225
33. v(t) = t
2
+ 7t 18, t
0
= 1.
34. v(t) = sinh t, t
0
= 0.
35. Suppose v(t) = (t +2)(t 3) and were interested in calculating the total distance traveled
by the particle between t = 0 and t = t
1
where t
1
> 0. Then the integral D(t
1
) =
_
t1
0
[(t +2)(t 3)[ dt will need to be evaluated in two pieces because of the sign change at
t = 3. Is the function D(t
1
) continuous at t
1
= 3? Is it dierentiable?
36. Redo the previous problem if v(t) = (t 3)
2
. Why is this function easier to work with
when calculating the distance function than the one in the previous problem?
37. Suppose a paratrooper jumps from an airplane with initial velocity v(0) = 0 and accelerates
downward at g = 9.8 m/s
2
for three seconds. She opens her parachute and then accelerates
upward at a rate of 2 m/s
2
for 5 more seconds.
a. Write an expression for v(t), where up is used as the positive direction.
b. What is the paratroopers displacement over the eight seconds?
c. What is the paratroopers average velocity over the eight seconds?
38. Suppose the A train is traveling at 60 mph. The conductor sees the B train in front of
him on the same track moving in the same direction with speed 30 mph. The conductor
puts on the breaks causing constant deceleration of a mph when the trains are exactly 1/4
of a mile apart.
a. How far will the A train travel, in terms of a, before it comes to rest?
b. How far will the B train travel, in terms of a, before the A train comes to rest?
c. What is the minimum a must be to prevent a collision?
39. Suppose v(t) is continuous. Prove, or provide a counterexample to, the statement
_
b
a
[v(t)[ dt =

_
b
a
v(t) dt

.
40. Consider the position function
p(t) =
_
t
2
sin(1/t) if t ,= 0;
0 if t = 0.
226 CHAPTER 3. APPLICATIONS OF INTEGRATION
a. Prove that p is dierentiable for all t.
b. Prove that p

(t) is not continuous at t = 0.

c. What are the ramications of (a) and (b) on the application of the fundamental
theorem of calculus? Specically, is it true that
_
b
0
p

(t) dt = p(b) p(0)?

In each of Exercises 41 through 43, you are given the acceleration function a(t) for a
particle and the velocity of the particle at time t = 0. Calculate the average velocity
of the particle between times t = 0 and t = t
1
.
41. a(t) = 2 sin 3t 3 cos 2t. v(0) = 4, t
1
= /6.
42. a(t) =
t
2
t + 1
+ 4. v(0) = 6, t
1
= 9.
43. a(t) =
3
25 t
2
. v(0) = 4, t
1
= 3.
44. Calculate the average acceleration of the functions in the three previous problems.
In some applications, it is reasonable to consider the limit of the displacement or
speed of a moving particle as t . These limits may or may not exist and should
be evaluated using improper integrals.
45. Suppose a force is acting on a particle in such a way that the velocity of the particle is
v(t) = e
t
. What is the limit as t
1
of the distance the particle has traveled between
times t = 0 and t = t
1
?
46. Using the velocity function in the previous problem,
a. What is the average velocity between times t = 0 and t = t
1
?
b. What is the limit of the average velocity between times t = 0 and t = t
1
as t
1
?
47. Consider a particle traveling in such a way that its velocity function is modeled by a
damped oscillation. Specically, assume v(t) = e
t
sin t.
a. Show that [v(t)[ 0 as t .
b. Show that
_
k
(k1)
[v(t)[ dt =
1
2
e
k
(1 +e

) for k = 1, 2, 3, ....
c. What is
_
n
0
[v(t)[ dt, where n is a positive integer? In other words, whats the total
distance traveled by the particle during this time interval?
d. What is the limit, if it exists, of
_
n
0
[v(t)[ dt as n ?
3.1. DISPLACEMENT AND DISTANCE TRAVELED 227
As we shall discuss at length in Section 3.3, the formula for the distance traveled
by an object traveling along a curved path in two or three-dimensional Euclidean
space is a natural generalization of our current formula for distance traveled along
the x-axis. Suppose that p is a dierentiable curve which gives the position of an
object in three-space. Let p(t) = (x(t), y(t), z(t)), where t [a, b]. Then the velocity
vector and instantaneous speed of the particle at t = t
0
are, respectively:
v(t
0
) = p

(t) = (x

(t
0
), y

(t
0
), z

(t
0
) and [v(t
0
)[ =
_
[x

(t
0
)]
2
+ [y

(t
0
)]
2
+ [z

(t
0
)]
2
.
This notation is convenient because the formula for the total distance traveled
between times t
0
and t is exactly the same as the one-dimensional formula:
r(t) =
_
t
t0
[v(z)[ dz.
48. Suppose an object is traveling in a circular orbit with radius R. The position function is
p(t) = (Rcos t, Rsin t). What is the total distance traveled by the particle between times
t
1
and t
2
? Show that this is equal to the arc length of a sector of a circle with central
angle t
2
t
1
.
49. The previous problem conrms our intuition that calculating the distance traveled by a
dynamic particle is equivalent to measuring the length of a path. A DNA molecule is
shaped like a helix- a spiral around a cylinder. The approximate radius of the molecule is
10 angstroms, and the helix rises by about 34 angstroms per revolution.
a. Show that this path can be parameterized as p(t) = (10 cos t, 10 sin t, 17t/).
b. Calculate p

(t). If we think of p as the position of a particle traversing the helix, then

p

(t) is the velocity vector.

c. Calculate [ p

(t)[. One can interpret this as either the instantaneous speed of a particle
traversing the helix, or as the length of the tangent vector.
d. What is the length, or distance traveled by a particle traversing the helix, in one
revolution?
e. A DNA molecule has approximately 285 million turns. What is the approximate
length of a DNA molecule?
228 CHAPTER 3. APPLICATIONS OF INTEGRATION
3.2 Area in the Plane
Given a continuous or, at least, integrable (Riemann integrable, or in the extended manner
dened for improper integrals), we have looked at examples of calculating the area under the
graph and above intervals on the x-axis or, for negative functions, areas above the graph and
under intervals on the x-axis.
In this section, we will look at the more general problem of calculating the area trapped
between the graphs of two dierent functions. This is a relatively easy application of integration.
However, as we shall see, the main new diculty is reminiscent of the diculty we saw in the
last section, where we wanted to calculate distance traveled, instead of displacement: we need
to integrate the absolute value of a function.
Back in Proposition 2.3.11, we saw that if f 0 and integrable on an interval [a, b], then
_
b
a
f(x) dx is equal to the area under the graph and above the interval [a, b]. We also saw that
if f 0 and integrable on an interval [a, b], then
_
b
a
f(x) dx =
_
b
a
[f(x)] dx is equal to the
area above the graph and under the interval [a, b]. One formula that unites these two results is
given in:
Proposition 3.2.1. Suppose that we have a function y = f(x) dened and integrable on an
interval, I, from a to b. Then, the area between the graph of y = f(x) and the x-axis,
for x in the interval I (or, on I), is
_
b
a
[f(x)[ dx.
Of course, the absolute value signs in Proposition 3.2.1 give a very succinct way of expressing
the result, but they hide the diculty: to actually evaluate
_
b
a
[f(x)[ dx, we have to do what
we did for distance traveled problems in the previous section, namely, split the integral up into
pieces over various subintervals on which f(x) is 0 and on which f(x) is 0. We then use
that, if f(x) 0, then [f(x)[ = f(x) and, if f(x) 0, then [f(x)[ = f(x).
3.2. AREA IN THE PLANE 229
Example 3.2.2. Calculate the area between the x-axis and the graph of y = ln x, for
1
2
x 2.
See Figure 3.1.
-1 0 1 2 3
-1
-0.5
0.5
1
Figure 3.1: Area between the x-axis and y = ln x,
1
2
x 2.
Solution:
The area of the region is given by
_
2
1/2
[ ln x[ dx.
But how do we calculate this?
We use that ln x 0 if 0 < x 1, ln x 0 if x 1 and, as we saw in Example 1.1.21,
integration by parts tells us that
_
ln xdx = xln xx+C. (Dont get confused: the derivative
of ln x is 1/x, but were integrating here.)
Therefore,
_
2
1/2
[ ln x[ dx =
_
1
1/2
[ ln x[ dx +
_
2
1
[ ln x[ dx =
_
1
1/2
ln xdx +
_
2
1
ln xdx =
(xln x x)

1
1/2
+ (xln x x)

2
1
=
230 CHAPTER 3. APPLICATIONS OF INTEGRATION
_
(0 1) +
_
1
2
ln
_
1
2
_

1
2
__
+ [(2 ln 2 2) (0 1)] =
1
2

1
2
ln 2 + 2 ln 2 1 =
3 ln 2 1
2
.
If we think of the x-axis as the graph of y = g(x) = 0, then nding the area between the
graph of y = f(x) and the x-axis is the same as nding the area between the graphs of y = f(x)
and y = g(x). Our question now is: can we use integration to nd the area trapped between
two graphs in the more general case in which we dont assume that one graph is the x-axis?
Lets think about this. Consider the functions y = f(x) = x
2
and y = g(x) = 4. Suppose we
want to nd the area between the graphs of f and g for 1 x 3.
-1 0 1 2 3 4 5
5
10
y=4
y=x
2
Figure 3.2: The area between the graphs of y = x
2
and y = 4.
What we do is split the problem into two pieces, pieces which correspond to which function
is bigger, f or g. Hence, we nd the area between the graphs for 1 x 2 and add to that the
area between the graphs for 2 x 3.
How do we nd the area between the graphs for 1 x 2? What we could do is nd
the area under the graph of y = 4, namely
_
2
1
4 dx (which, even without integrating, we know
3.2. AREA IN THE PLANE 231
is 4 1, since the region is a rectangle of height 4 and width 1), and then subtract the missing
area,
_
2
1
x
2
dx, under the graph of y = x
2
. We would nd that the area between the graphs for
1 x 2 is
_
2
1
4 dx
_
2
1
x
2
dx =
_
2
1
(4 x
2
) dx =
_
2
1
(g(x) f(x)) dx.
Similarly, we could calculate the area between the graphs of f and g for 2 x 3 by taking
the area,
_
3
2
x
2
dx, under the graph of y = x
2
, and above the interval [2, 3] and then subtracting
the missing area
_
3
2
4 dx. We would nd that the area between the graphs for 1 x 2 is
_
3
2
x
2
dx
_
3
2
4 dx =
_
3
2
(x
2
4) dx =
_
3
2
(f(x) g(x)) dx.
However, theres one psychological reason and one practical reason why, on the dierent
subintervals, you dont want to think of taking the entire area under one graph and subtracting
the entire missing area under the other graph; you want to think of integrating the dierences,
i.e., you want to think of the problem in terms of
_
2
1
(g(x) f(x)) dx and
_
3
2
(f(x) g(x)) dx in
the rst place.
Psychologically, you shouldnt think of involving any regions other than those that are
trapped between the two graphs. Instead of subtracting areas of entire regions, you should
think of taking the continuous sum of areas of innitesimal rectangles that lie between the two
graphs, rectangles of innitesimal width dx, and height (g(x) f(x)), if 1 x 2, or height
(f(x) g(x)), if 2 x 3. See Figure 3.3. Thus, we have rectangles of innitesimal area
dA = (g(x) f(x)) dx or dA = (f(x) g(x)) dx, and we should add up these innitesimal areas,
by taking the integral, to nd the total area.
In a practical sense, for other functions and intervals, it could be signicantly easier to
calculate
_
b
a
(f(x) g(x)) dx than to separately calculate
_
b
a
f(x) dx and
_
b
a
g(x) dx and then
subtract. How is this possible? Nasty terms might cancel out in f(x) g(x). For instance, if
f(x) = x
2
+ e
x
2
and g(x) = 4 + e
x
2
, then
_
3
2
(f(x) g(x)) dx =
_
3
2
(x
2
4) dx, which is easy
to calculate, but you would not succeed in calculating either
_
3
2
f(x) dx or
_
3
2
g(x) dx.
So, what we have seen is that, over subintervals [a, b] where g(x) f(x), i.e., where g(x)
f(x) 0, we nd the area between the graphs of y = f(x) and y = g(x) by calculating
_
b
a
(g(x) f(x)) dx. Over subintervals [a, b] where f(x) g(x), i.e., where f(x) g(x) 0, we
nd the area between the graphs of y = f(x) and y = g(x) by calculating
_
b
a
(f(x) g(x)) dx.
Therefore, regardless of which function is greater, our integrand is always [f(x) g(x)[.
232 CHAPTER 3. APPLICATIONS OF INTEGRATION
-1 0 1 2 3 4 5
5
10
y=4
y=x
2
height = g(x)-f(x)
height = f(x)-g(x)
Figure 3.3: Innitesimally wide rectangles between the graphs of y = x
2
and y = 4.
Proposition 3.2.3. Suppose that we have functions y = f(x) and y = g(x) dened on an
interval, I, from a to b, and that f(x) g(x) is integrable on I. Then, the area between
the graphs of y = f(x) and y = g(x), for x in the interval I (or, on I), is
_
b
a
[f(x) g(x)[ dx.
Note that, if g(x) = 0, so that the graph of y = g(x) is just the x-axis, then Proposition 3.2.3
reduces to Proposition 3.2.1.
Example 3.2.4. Lets nish our problem from the discussion above. We have the functions
y = f(x) = x
2
and y = g(x) = 4, and we want to nd the area between the graphs of f and g
for 1 x 3.
Solution:
We need to calculate
_
3
1

x
2
4

dx.
3.2. AREA IN THE PLANE 233
In order to deal with the absolute value, we need to know where x
2
4 0 and where
x
2
4 0. As x
2
4 is continuous, we rst nd where x
2
4 = 0. This is easy; it happens
when x = 2. Deleting these two zeroes divides the real line into three subintervals (, 2),
(2, 2), and (2, ), and the sign, , of x
2
4 cannot change on a given one of these subintervals,
for the Intermediate Value Theorem tells us that continuous functions must pass through zero
to switch signs.
Thus, to nd the sign of x
2
4 on each of these subintervals, we may simply pick any x
value in the subinterval, evaluate x
2
4 there, and see whether its positive or negative. In fact,
as were integrating from 1 to 3, we dont actually care about what happens on the interval
(, 2). From the interval (2, 2), well pick x = 0, and nd then that x
2
4 = 4 < 0; thus,
x
2
4 < 0 for all x in the interval (2, 2), and so x
2
4 0 for all x in [2, 2]. In particular,
x
2
4 0 for each x in the interval [1, 2]. When x = 3, x
2
4 = 5 > 0, and so x
2
4 0 for
all x in the interval [2, ). Of course, it follows that x
2
4 0 for all x in the interval [2, 3].
Hence, we nd
_
3
1

x
2
4

dx =
_
2
1

_
x
2
4
_
dx +
_
3
2
_
x
2
4
_
dx =
_

x
3
3
+ 4x
_

2
1
+
_
x
3
3
4x
_

3
2
=
__

8
3
+ 8
_

1
3
+ 4
__
+
_
(9 12)
_
8
3
8
__
= 4.
Example 3.2.5. Of course, not all problems about area between graphs involve a change in
signs of f(x) g(x). Consider the problem of nding the area between the graphs of y = sin x
and y = cos x, for 0 x /6.
All that you need to know is that, for 0 x /6, cos x sin x, so that cos x sin x 0,
and, hence, [ cos x sin x[ = cos x sin x. Thus, the area between the graphs is
Area =
_
/6
0
(cos x sin x) dx = (sin x + cos x)

/6
0
=
1
2
+

3
2
0 1 =

3 1
2
.
234 CHAPTER 3. APPLICATIONS OF INTEGRATION
-0.5
0
0.5
1
1.5
!/6
y=cos(x)
y=sin(x)
Figure 3.4: Area between y = sin x and y = cos x, 0 x /6.
Really, in a problem where one function is always bigger than (or equal to) the other function,
you shouldnt think in terms of absolute values at all; just take the bigger function, subtract the
smaller function, and integrate.
There are times when you are not explicitly given the limits of integration. Lets look at a
typical such problem.
Example 3.2.6. Find the area of the bounded region between the graphs of y = f(x) = x
3
and y = g(x) = x
2
.
Solution: If you look at the graphs in Figure 3.5, you can see the region between the graphs
of y = x
3
and y = x
2
naturally breaks up into three pieces: one piece where x 0, one piece
where 0 x 1, and one piece where x 1. But the pieces where x 0 and x 1 are
unbounded, i.e., go out innitely far. The bounded region between the graphs is the portion
where 0 x 1; the part thats really trapped, or completely bordered, by the graphs.
After we know this, the problem is easy, except for one small issue. You probably normally
think that x
3
is bigger than x
2
. Right? Doesnt cubing a number give you something bigger than
squaring it? Not for numbers between 0 and 1! (And certainly not for negative numbers,
either.) If 0 x 1, then x
2
x
3
. Therefore, for 0 x 1, x
2
x
3
0, and we dont need
to use absolute values to nd the area. The area of the bounded region is simply
Area =
_
1
0
(x
2
x
3
) dx =
x
3
3

x
4
4

1
0
=
1
12
.
3.2. AREA IN THE PLANE 235
-0.5 0 0.5 1 1.5
-0.5
0.5
1
1.5
y=x
3
y=x
2
Figure 3.5: Area of bounded region between y = x
3
and y = x
2
.
In the following example, we look at a problem in which nasty pieces of the functions involved
cancel out.
Example 3.2.7. Let f(x) = e
x
+ e
x
2
and g(x) = e
3x+4
+ e
x
2
. Find the area between the
graphs of f and g for 0 x 3.
Solution:
You might be inclined to graph f and g by hand, or by using a calculator or computer. The
problem is that the e
x
2
terms get so big so fast that you will have diculty seeing the relatively
small dierence between the values of the functions. The good news is that we dont need to
see a picture in order to solve the problem.
We need to calculate
_
3
0
[f(x) g(x)[ dx =
_
3
0

e
x
e
3x+4

dx.
To deal with the absolute value, we need to determine where e
x
e
3x+4
is positive and
where its negative. So, we nd where e
x
e
3x+4
= 0, i.e., where e
x
= e
3x+4
. Taking
natural logs of both sides of the equation (or using that the exponential function is one-to-one),
236 CHAPTER 3. APPLICATIONS OF INTEGRATION
we nd that x = 3x + 4, and so x = 2.
Therefore, since e
x
e
3x+4
is continuous and is zero only at x = 2, we must have that
e
x
e
3x+4
is either always positive or always negative on each of the intervals (, 2) and
(2, ). We simply check the sign of the function at some point in each interval. When x = 0,
e
x
e
3x+4
= e
0
e
4
< 0; hence, e
x
e
3x+4
< 0 for all x in (, 2), and so e
x
e
3x+4
0
for all x in (, 2]. Similarly, when x = 3, e
x
e
3x+4
= e
3
e
5
> 0, and we conclude
that e
x
e
3x+4
0 for all x in [2, ).
Thus, we nd
_
3
0

e
x
e
3x+4

dx =
_
2
0

e
x
e
3x+4

dx +
_
3
2

e
x
e
3x+4

dx =
_
2
0

_
e
x
e
3x+4
_
dx +
_
3
2
_
e
x
e
3x+4
_
dx.
Lets produce an anti-derivative for all functions of the form e
ax+b
, so that we can use this
result to integrate each piece in the integrals above. Consider
_
e
ax+b
dx, where a and b are
constants, and a ,= 0. Make the substitution u = ax + b, so that du = a dx, i.e., dx = du/a.
Then, we nd
_
e
ax+b
dx =
_
e
u

du
a
=
1
a
_
e
u
du =
1
a
e
u
+C =
e
ax+b
a
+C.
Now, we apply this formula to e
x
and e
3x+4
, and obtain
_
2
0

_
e
x
e
3x+4
_
dx = e
x

e
3x+4
3

2
0
= e
2

e
2
3

_
1
e
4
3
_
=
2e
2
3
1 +
e
4
3
.
We also nd that
_
3
2
_
e
x
e
3x+4
_
dx = e
x
+
e
3x+4
3

3
2
=
3.2. AREA IN THE PLANE 237
e
3
+
e
5
3

_
e
2
+
e
2
3
_
= e
3
+
e
5
3
+
2e
2
3
.
Putting together all of our above work, we nd:
Area =
_
3
0

e
x
e
3x+4

dx =
2e
2
3
1 +
e
4
3
e
3
+
e
5
3
+
2e
2
3
=
e
4
+ 4e
2
+e
5
3
e
3
1.
In the next example, we consider the same two functions that we did above, but now we look
at the area of an unbounded region, a region that extends out innitely far. This, of course,
leads to an improper integral.
Example 3.2.8. Let f(x) = e
x
+ e
x
2
and g(x) = e
3x+4
+ e
x
2
. Find the area between the
graphs of f and g for 0 x < .
Solution:
These are the same functions that we used in the previous example, and so we can use much
of our work from that problem.
We need to calculate
_

0

e
x
e
3x+4

dx.
We could do this either one of two ways.
First, we could proceed exactly as we did in the previous example, and write
_

0

e
x
e
3x+4

dx =
_
2
0

e
x
e
3x+4

dx +
_

2

e
x
e
3x+4

dx =
_
2
0

_
e
x
e
3x+4
_
dx +
_

2
_
e
x
e
3x+4
_
dx.
We already calculated, in the previous example, the integral on the left; we would still need to
calculate the improper integral on the right.
238 CHAPTER 3. APPLICATIONS OF INTEGRATION
However, given that we already calculated
_
3
0

e
x
e
3x+4

dx is the previous example, it

would also be reasonable to split the integral up in the following manner:
_

0

e
x
e
3x+4

dx =
_
3
0

e
x
e
3x+4

dx +
_

3

e
x
e
3x+4

dx =
e
4
+ 4e
2
+e
5
3
e
3
1 +
_

3
_
e
x
e
3x+4
_
dx,
where, again, we still need to calculate the improper integral on the right.
Well use the second splitting of the integral, though it makes little dierence. We nd
_

3
_
e
x
e
3x+4
_
dx = lim
p
_
e
x
+
e
3x+4
3

p
3
_
.
lim
p
_
e
p
+
e
3p+4
3
+e
3

e
5
3
_
= e
3

e
5
3
.
Therefore, our nal answer is
Area =
_

0

e
x
e
3x+4

dx =
e
4
+ 4e
2
+e
5
3
e
3
1 + e
3

e
5
3
=
e
4
+ 4e
2
3
1.
What if you want to nd the area between curves that arent given to you by having y in
terms of x, but instead are given to you by x being specied in terms of y? For instance, how
do you nd the area between the curves/lines x = 4 y
2
and x = 2 y for 0 y 2?
3.2. AREA IN THE PLANE 239
-1 0 1 2 3 4
-1
1
2
3
Figure 3.6: Area between x = 4 y
2
and x = 2 y, 0 y 2.
There are several ways to approach this problem, some good, some not so good. Lets start
with the not so good.
Probably the worst way to approach this problem is to think I know how to do area problems
only when I have y in terms of x, so Ill rewrite both equations, solving for y. You get
y =

4 x (theres no plus or minus sign because were looking where y 0) and y = 2 x.

Understand that in other problems, this step, of solving algebraically for y in terms of x, might
be impossible (in any reasonable way). Thats a big downside to this method in general, but we
can try to proceed anyway with our specic problem.
So, now you need to nd the area between the two graphs of y =

4 x and y = 2 x, for
0 y 2. If you look at Figure 3.6, its easy to see another diculty: in terms of x, the region
that were trying to nd the area of is between y = 2 x and y =

4 x when 0 x 2,
but is between y = 0 and y =

4 x when 2 x 4. This means that we need to split the

problem up as the sum of two integrals, i.e., we need to calculate
_
2
0
_
4 x (2 x)
_
dx +
_
4
2
_
4 x 0
_
dx.
Yuck. We would do this if we had to, and nd the answer of 10/3, but surely theres an easier
way.
Even if you want to nd areas by always integrating with respect to x, you dont have to
do what we did in the previous method. You could just tell yourself that ipping the graph,
interchanging the x- and y-axes, doesnt change the area of the region. In fact, instead of
actually ipping the graph to save time, eort, and space you can simply relabel the old
240 CHAPTER 3. APPLICATIONS OF INTEGRATION
x-axis as now being the y-axis, and the old y-axis as now being the x-axis, and swap all of the x
and ys in the equations and in the bounds. This way, we are now looking for the area between
the curves/lines y = 4 x
2
and y = 2 x for 0 x 2, but the region looks the same as it did
before. Note the new labels on the axes in Figure 3.7.
-1 0 1 2 3 4
-1
1
2
3
x
y
Figure 3.7: Make the x-axis the vertical axis, and the y-axis the horizontal.
Now we know what to do. To calculate the area, we evaluate
_
2
0
_
(4 x
2
) (2 x)
_
dx =
_
2
0
_
2 +x x
2
_
dx = 2x +
x
2
2

x
3
3

2
0
=
10
3
. (3.1)
But the best method for nding the area of the given region is to realize that, while you could
switch the xs and ys, as we did above, we dont need to; just dont panic about using y for your
integration variable in the rst place. Had you not exchanged the xs and ys in the rst place,
but instead used y for the integration variable, you would have had ys throughout Formula 3.1,
in place of xs, but of course youd get the same answer:
_
2
0
_
(4 y
2
) (2 y)
_
dy =
_
2
0
_
2 +y y
2
_
dy = 2y +
y
2
2

y
3
3

2
0
=
10
3
.
Geometrically, with the x-axis and y-axis in their usual positions, this means we are taking
the continuous sum of areas of innitesimally high rectangles, of height dy, whose length is given
3.2. AREA IN THE PLANE 241
by the dierence of the x-coordinates on the two curves, as in Figure 3.8
-1 0 1 2 3 4
-1
1
2
3
} dy
Figure 3.8: An innitesimal rectangle between x = 4 y
2
and x = 2 y.
Even though its really nothing new, for the sake of completeness, and for reference, we give
the general proposition which relates to the discussion above.
Proposition 3.2.9. Suppose that we have functions x = f(y) and x = g(y) dened on an
interval, I, from a to b, and that [f(y) g(y)[ is integrable on I. Then, the area between
the graphs of x = f(y) and x = g(y), for y in the interval I, is
_
b
a
[f(y) g(y)[ dy.
A particular case of Proposition 3.2.9 is when one of the curves is the graph of x = 0, i.e.,
the y-axis. Then, the integral for the area collapses to just the integral of the absolute value of
the other function.
242 CHAPTER 3. APPLICATIONS OF INTEGRATION
Example 3.2.10. Suppose, for instance, that you want to nd the area between the graph of
x = sin y and the y-axis, for

4
x

2
.
-1 0 1
-!/4
!/2
Figure 3.9: Area between x = sin y and the y-axis, /4 y /2.
We need to calculate
_
/2
/4
[ sin y[ dy. We nd
_
/2
/4
[ sin y[ dy =
_
0
/4
[ sin y[ dy +
_
/2
0
[ sin y[ dy =
_
0
/4
sin y dy +
_
/2
0
sin y dy = cos y

0
/4
+ (cos y)

/2
0
=
1
1

2
+ 0 (1) = 2
1

2
=
4

2
2
.
3.2.1 Exercises
In each of Exercises 1 through 5, nd the total area between the x-axis and the
graph of y = f(x), for the indicated values of x. Also, possibly using a calculator or
graphing software, sketch the region whose area you are calculating, and include in
your sketch some typical innitesimal rectangles with innitesimal width dx.
1. y = sin x, 0 x 2
3.2. AREA IN THE PLANE 243
2. y = cos x, 0 x 2
3. y = 9 x
2
, 4 x 2
4. y = x
3
+x
2
2x, 3 x 2
5. y = xe
(x
2
)
, 1 x 1
In each of Exercises 6 through 10, nd the area between the y-axis and the graph of
x = f(y), for the indicated values of y. Also, possibly using a calculator or graphing
software, sketch the region whose area you are calculating, and include in your
sketch some typical innitesimal rectangles with innitesimal height dy.
6. x = sin y, 0 y 2
7. x = cos y, 0 y 2
8. x = 9 y
2
, 4 y 2
9. x = y
3
+y
2
2y, 3 y 2
10. x = ye
(y
2
)
, 1 y 1
In each of Exercises 11 through 20, nd the area of the bounded region between
the graphs of the two given functions. Use integration with respect to whichever
variable seems most convenient.
11. y = 3 x
2
, y = 6
12. y = 3 x
2
, y = 3x 1
13. x = 3 y
2
, x = 3y 1
14. x = y
2
, y = x
2
15. y = x(e
x
e), y = 0
16. x = y(e
y
e), x = 0
17. x =
4
1 +y
2
, x = 2
18. y =
10
1 +x
2
, y = 2
19. y =
10
1 +x
2
, y = 1
7
1 +x
2
244 CHAPTER 3. APPLICATIONS OF INTEGRATION
20. x =
10
1 +y
2
, x = 1
7
1 +y
2
In each of Exercises 21 through 36, calculate the total area bounded by the graphs
of y = f(x) and y = g(x), or x = f(y) and x = g(y), for values of the independent
variable in the given interval.
21. f(x) = x
2
16, g(x) = 9, [10, 10].
22. f(x) = sin x, g(x) = cos x, [0, 2].
23. f(y) = sin y, g(y) = sin 2y, [0, 2].
24. f(y) = e
y
2
+e
3y+1
, g(y) = e
2y+2
+e
y
2
, [4, 3].
25. f(x) = x
4
+x
3
3x
2
+ 5x + 5, g(x) = x
4
+ 2x
3
3x
2
+ 5x + 6, [4, 6].
26. f(x) = ln(x
x
tan x) ln(sin x), g(x) = ln(e
x
sec x), [1, e/2].
27. f(y) = 3/(y + 2), g(y) = 2/(y 4), [1, 3].
28. f(y) =
tan y
1 tan y tan 1
, g(y) =
tan 1
1 tan y tan 1
, [0, /4].
29. f(x) = x
4
+x
2
+ 2, g(x) = x
3
+ 3x, [3, 3].
30. f(x) = tan
1
x +x
2
+ 3, g(x) = 2x
2
+x + tan
1
x 4, [3, 5].
31. f(y) = e
2y
+ 4, g(y) = e
2y
+ 2, [5, 5].
32. f(y) =
y
4
+y
2
_
y
6
+y
4
, g(y) =
y
2
_
y
2
+ 1
, [1, 4].
33. f(x) = sinh
1
x, g(x) = cosh
1
x, [1, 10].
34. f(x) =
x
2
(x 2)(x + 2)
, g(x) =
4
(x 2)(x + 2)
, [1, 1].
35. f(y) = [y[, g(y) = [y[ + 1, [1/2, 1/2].
36. f(y) = 2
y
, g(y) = 3
y
, [1, 2].
37. Prove that if f, g and h are continuous functions on [a, b] then the area between the graphs
of f and g is equal to the area between the graphs of f +h and g +h.
38. Suppose the area enclosed by two continuous functions f and g on the interval [a, b] is
A =
_
b
a
[f(x) g(x)[ dx. Suppose c is an arbitrary real number. Is it true that the area
between the graphs of cf(x) and cg(x) is cA? If not, how can you correct this statement?
3.2. AREA IN THE PLANE 245
39. Calculate the limiting area between the curves e
at
and e
bt
between t = 0 and t = U
where U . Assume a and b are positive distinct real numbers.
40. Calculate the area between the curves h(y) = ln(ay) and g(y) = ln(by) on the interval
[c, d] where a, b and c are positive and a ,= b.
41. Calculate the area between the curves a(t) = 1/t and b(t) = 1/t
2
on the interval [1, U].
Does the limiting area converge as U ?
42. Redo the previous problem with a(t) = 1/t
2
and b(t) = 1/t
3
.
43. What is the area enclosed by the curves f(x) = sin nx and g(x) = cos nx on the interval
[0, 2] where n is a positive integer?
In each of Exercises 44 through 47, use the Midpoint Rule to approximate the areas
bound by the two curves on the interval [0, 1]. Use a partition of n = 4 evenly spaced
points.
44. f(x) = e
x
2
, g(x) = e
x
.
45. f(x) = e
x
k
, g(x) = e
x
, k > 0.
46. f(x) = e
x
k
, g(x) = e
x
m
, k, m > 0.
47. f(x) =

1 +x
3
, g(x) = x + 1.
In Exercises 48 through 51, you are given the area A between the graphs of f(x)
and g(x) over an interval I. You are also given the two functions with one unknown
parameter c. Solve for the parameter.
48. A = 24, f(x) = 2x
2
3x + 4, g(x) = x
2
+cx 6, I = [0, 6].
49. A = 2 + 2

3 /6, f(x) = sin x, g(x) = c, I = [0, ], c > 0.

50. A = 16, f(x) = c [x[, g(x) = [x[, I = [c, c].
51. A = 9 18, f(x) =

c
2
x
2
, g(x) = c x, I = [0, c].
52. Consider the region R in the rst quadrant, under the graph of y = 1x
2
. Find the value
of b so that the horizontal line given by y = b splits R into two sections with equal areas.
53. Suppose that f is continuous on [a, b].
a. Prove that [f[ is integrable on [a, b].
246 CHAPTER 3. APPLICATIONS OF INTEGRATION
b. Argue that one can choose k = 1 such that k
_
b
a
f(x) dx 0.
c. Conclude that

_
b
a
f(x) dx

_
b
a
[f(x)[ dx.
d. Use part (c) to show that a lower bound for the area enclosed between the graphs of
f(x) and g(x) on the interval [a, b] is

_
b
a
f(x) dx
_
b
a
g(x) dx

.
54. Argue that if f and h are continuous on [a, b], then the area enclosed by f and the x-axis
is less than or equal to the sum of the area enclosed by f and h, and the area enclosed by
h and the x-axis.
55. Suppose f and g are continuous on [a, b]. Let h(x) = [f(x)[ and j(x) = [g(x)[. Prove that
the area enclosed by j and h on the interval [a, b] is less than or equal to the area enclosed
by f and g on the interval [a, b].
Its often more convenient to describe a function f using polar coordinates. If we
assume that the distance from the origin to a point on the graph, r, depends only
on the angle between the x-axis and ray connecting the origin to the point, then
we can write r = f(). Under reasonable assumptions, the area of the polar region
dened for [a, b] is A =
1
2
_
b
a
f
2
() d. Here, we assume that b a 2 to prevent
double counting. Calculate the areas of the polar regions in Exercises 36 - 40.
56. r = , [0, ].
57. r = sin 3, [0, /3].
58. 4 = sin n, [0, /n], n a positive integer.
59. r = r
0
, [0, 2]. What familiar shape is this?
60. r = 1 + sin , [0, 2].
The area between two polar regions dened by the functions r
1
() and r
2
() is given
by A =
1
2
_
b
a
[r
1
() r
2
()[ d. Calculate the areas between the polar curves over the
given angular interval in Exercises 41 - 45
61. r
1
() = 1/2, r
2
() = cos 3, I = [/9, /9].
62. r
1
() = 3 cos , r
2
() = 1 + cos , [/3, /3].
63. r
1
= sin , r
2
= cos , I = [0, /4].
64. Recalculate the previous problem by writing the equations of the graphs of the two func-
tions in cartesian coordinates.
3.2. AREA IN THE PLANE 247
65. Use the polar area formula to nd the area of the annular region enclosed by circles of
radii R
1
and R
2
, 0 < R
1
< R
2
.
Recall that the area below the graph of a probability density function (pdf ) is
interpreted as a probability. Specically, if f(x) is a pdf of a random variable
X, then the probability that X is between a and b is
_
b
a
f(x) dx. In nance and
economics, the area between two curves can be used to assess the materiality in
choosing one model over another. We explore this idea in Exercises 46 - 48.
66. Suppose that the XY Z corporation is considering a $200,000 investment. It is known
with certainty that the most XY Z can lose is its initial investment of $200,000 and the
most it can prot is $700,000. Analyst A believes the prot is uniformly distributed
on the interval [200000, 700000]. Analyst B believes the prot has density function
f
2
(x) =
x
2
117, 000, 000
.
a. What is f
1
(x), the pdf based on Analyst As assumptions?
b. What is the area between f
1
(x) and f
2
(x) on the interval [500, 700]? What is the
probabilistic interpretation of this area?
67. Under which analysts assumptions is it more likely for XY Z to experience a positive
prot from this investment?
68. Calculate the area between the graphs of the functions xf
1
(x) and xf
2
(x) on the interval
[200, 700]. What does this area represent?
69. Suppose f(x) is a strictly increasing continuous function dened on all reals with the
property that f(0) = 0. Then f possesses an inverse function g. Just as
_
a
0
f(x) dx is the
area below the graph of f on the interval [0, a],
_
b
0
g(y) dy is the area to the left of the
graph of g along the (vertical) interval [0, b].
a. Assume a and b are positive. Draw a picture to prove that
ab
_
a
0
f(x) dx +
_
b
0
g(y) dy.
This is Youngs Inequality.
b. Based on the picture, formulate a conjecture about when the inequality is an equality.
248 CHAPTER 3. APPLICATIONS OF INTEGRATION
70. Use the previous problem and the function f(x) = x
k
, k > 0 to prove that
ab
a
p
p
+
b
q
q
if a, b 0, 1 < p and
1
p
+
1
q
= 1.
71. a. Suppose f and g are two continuous functions, both of which are positive for all real
x. Suppose p > 1 is a real number, and q is the unique real such that
1
p
+
1
q
= 1.
Suppose f and g have the additional properties:
_
_
b
a
[f(x)]
p
dx
_
1/p
= 1 and
_
_
b
a
[g(x)]
q
dx
_
1/q
= 1.
Integrate both sides of the inequality in the previous problem to prove that
_
b
a
f(x)g(x) dx 1.
b. Now suppose f and g satisfy all the conditions in part (a) except the last two. Prove
that
_
b
a
f(x)g(x) dx
_
_
b
a
[f(x)]
p
dx
_
1/p

_
_
b
a
[g(x)]
q
dx
_
1/q
.
This result is known as H olders Inequality.
3.3. DISTANCE TRAVELED IN SPACE AND ARC LENGTH 249
3.3 Distance Traveled in Space
and Arc Length
In Section 3.1, we looked at an object which was moving in a straight line, and found that its
displacement and distance traveled could be calculated by integrating its velocity v(t) and the
speed [v(t)[, respectively.
Suppose, however, that an object is moving along some curved path in a plane or in space.
How do we use the velocity and/or speed of the object to calculate the displacement and the
distance traveled? Amazingly, the answer is that, aside from writing things in terms of vectors,
the formulas look exactly the same.
Its also true that the total distance traveled by an object is equal to the length of the curve
(or line) that it travels along, provided that the object doesnt move back over points that its
already hit. This means that we can use the same techniques to calculate the arc length of a
curve that we use in calculating the distance traveled.
In this section, we will use a small amount of material on vectors and vector-valued functions,
such as the material in Appendix A. While this discussion could be put o until you take multi-
variable Calculus, it is really not much more dicult than analyzing motion in a straight line.
Displacement and distance traveled:
Suppose that an object is moving in xyz-space, R
3
, in such a way that, at time t, its x-, y-,
and z-coordinates are given by continuously dierentiable functions x(t), y(t), z(t). Then, the
position function or position vector or, simply, position of the object at time t is
p(t) = (x(t), y(t), z(t)).
When we say that p : [a, b] R
3
is continuously dierentiable, you may wonder what is
meant at the endpoints a and b of the closed interval, since you cant take two-sided limits
there. We shall always mean that p is dened and dierentiable on an open interval which
contains [a, b], and that p

is continuous when restricted to the closed interval [a, b].

For a given t value, we sometimes think of p(t) as simply the point (x(t), y(t), z(t)), and, at
other times, think of p(t) as the vector which is represented by the arrow from the origin to the
point p(t); the context should always make it clear how we are thinking of p(t).
250 CHAPTER 3. APPLICATIONS OF INTEGRATION
If the object is moving in the xy-plane, R
2
, we can still think of it as moving in xyz-space,
but always with z-coordinate equal to 0. Thus, we will discuss motion in xyz-space, and that
will also tell you what happens for objects moving in a plane; you take our discussion for motion
in space and either set z(t) = 0 for all t, or just omit the z-coordinate. On the other hand,
diagrams are much easier to produce and understand in the xy-plane; so our diagrams will
typically show the 2-dimensional situation, and we leave to you to picture the 3-dimensional
case.
The velocity (function or vector) of the object is
v(t) = p

(t) = (x

(t), y

(t), z

(t)).
While we shall not derive it here, we will mention that, if the velocity vector v(t) is drawn
as initiating from the point p(t), and v(t) ,=

0, then v(t) will be tangent to the curve that the
object is moving along. See Figure 3.10.
p(t)
p(t)
v(t)
v(t)
Figure 3.10: Velocity vectors are tangent to the curve dened by p(t).
As the velocity is the derivative of the position, if we are given v(t), we can anti-dierentiate
each component function, and obtain
_
v(t) dt = (x(t) +C
x
, y(t) +C
y
, z(t) +C
z
) =
(x(t), y(t), z(t)) + (C
x
, C
y
, C
z
) = p(t) +

C,
3.3. DISTANCE TRAVELED IN SPACE AND ARC LENGTH 251
where C
x
, C
y
, and C
z
are constants.
Now, suppose that we have two times a and b, where a b. Then, by denition, the denite
integral from a to b of a vector-valued function is obtained by integrating each component
separately; so,
_
b
a
v(t) dt =
_
_
b
a
x

(t) dt,
_
b
a
y

(t) dt,
_
b
a
z

(t) dt
_
.
By applying the Fundamental Theorem of Calculus to each component, we nd that this last
vector of integrals equals
_
x(b) x(a), y(b) y(a), z(b) z(a)
_
= p(b) p(a).
The vector p(b) p(a) is the change in the position of the object between times t = a and t = b;
as in the case of motion in a line, this change in position is the displacement of the object. The
magnitude of the displacement is the (straight line) distance between the points p(a) and p(b).
p(a)
p(b)
p(b)-p(a)
Figure 3.11: A typical displacement vector.
From our discussion above, we arrive at the more general vector version of Proposition 3.1.1:
Proposition 3.3.1. If the velocity v = v(t), as a function of time t, of an object is con-
tinuous on the interval [a, b], then the displacement of the object between times t = a and
t = b is given by
_
b
a
v(t) dt.
252 CHAPTER 3. APPLICATIONS OF INTEGRATION
Example 3.3.2. A particles velocity is given by v(t) = (e
t
,

2, e
t
) meters per second, where
t is in seconds. What is the displacement of the particle between times t = 0 and t = 1 seconds?
What is the magnitude of the displacement?
Solution:
This is easy. We nd the displacement is
p(1) p(0) =
_
1
0
v(t) dt =
_
1
0
(e
t
,

2, e
t
) dt = (e
t
,

2 t, e
t
)

1
0
=
(e,

2, e
1
) (1, 0, 1) = (e 1,

2, e
1
1) meters.
The magnitude [ p(1) p(0)[ of the displacement vector is
[ p(1) p(0)[ =

(e,

2, e
1
) (1, 0, 1)

=
_
(e 1)
2
+ (

2)
2
+ (e
1
1)
2
=
_
e
2
2e + 1 + 2 +e
2
2e
1
+ 1 =
_
(e e
1
)
2
2(e +e
1
3) meters,
where we have left the answer in this form for a reason that will be apparent in the next example.
If we look at the example above, and anti-dierentiate v(t), we nd that the position p(t) is
given by
p(t) = (e
t
,

2 t, e
t
) +

C,
where

C is a constant vector. This means that, between times 0 and 1 second, the particle
moves from the point (1, 0, 1) +

C to the point (e,

2, e
1
) +

C, not along a straight line, but
rather along the curved path p(t) = (e
t
,

2 t, e
t
) +

C.
The (straight line) distance between the initial and nal points of the particle is the distance
between (e,

2, e
1
) +

C and (1, 0, 1) +

C; this is equal to the magnitude of the displacement
[ p(1) p(0)[.
But, what if we want to know the total distance traveled by the particle along the curved
path given by p(t) = (e
t
,

2 t, e
t
) +

C? This distance traveled along the curved path should
(and will) turn out to be bigger than
_
(e +e
1
)
2
2(e +e
1
3).
3.3. DISTANCE TRAVELED IN SPACE AND ARC LENGTH 253
To determine how to calculate the distance traveled by an object, rst think about approx-
imating the distance traveled during a small interval of time, and then we add together the
resulting small distances, and take the limit as the size of the time interval approaches zero.
Not surprisingly, this results in an integral.
Suppose that we have two times t
0
< t
1
, and lets think about what happens when t = t
1
t
0
is small (close to 0). If t is close to 0, then the distance s that the object travels (along its
possibly curved path) should approximately equal the distance along a straight line between the
starting position p(t
0
) and the ending position p(t
1
); this straight line distance is [ p(t
1
) p(t
0
)[.
Thus, we have
s [ p(t
1
) p(t
0
)[.
We can rewrite this distance as
s

p(t
0
+ t) p(t
0
)
t

t,
where we used that t is positive. As t 0,
_
p(t
0
+t) p(t
0
)
_
/t approaches v(t) = p

(t
0
).
Using dierential notation and innitesimal terminology, this means that the innitesimal
distance ds traveled by the object in the innitesimal time interval dt is
ds = [v(t)[ dt.
The innitesimal value ds is frequently referred to an element of arc length. Of course, to obtain
the total distance traveled, we take a continuous sum of the innitesimal distances traveled, i.e.,
we take an integral.
Proposition 3.3.3. If the velocity v = v(t), as a function of time t, of an object is contin-
uous on the interval [a, b], then the distance traveled by the object between times t = a
and t = b is given by
_
t=b
t=a
ds =
_
b
a
[v(t)[ dt,
where ds = [v(t)[ dt.
254 CHAPTER 3. APPLICATIONS OF INTEGRATION
Example 3.3.4. Lets return to the situation in Example 3.3.2. A particles velocity is given
by v(t) = (e
t
,

2, e
t
) meters per second, where t is in seconds. What is the distance traveled
by the particle between times t = 0 and t = 1 seconds?
Solution:
We calculate that the distance traveled:
_
1
0
[v(t)[ dt =
_
1
0
_
(e
t
)
2
+ (

2)
2
+ (e
t
)
2
dt =
_
1
0
_
e
2t
+ 2 +e
2t
dt =
_
1
0
_
(e
t
+e
t
)
2
dt =
_
1
0
(e
t
+e
t
) dt =
_
e
t
e
t
_

1
0
= (e e
1
) (1 1) =
e e
1
meters.
Note that, since e + e
1
3 > 0, the distance traveled is greater than the magnitude of the
displacement,
_
(e e
1
)
2
2(e +e
1
3) meters,
which we found in Example 3.3.2. In other words, the distance traveled along a curved path
between two points is greater than the straight line distance between the points. Good!
If the velocity v(t) of an object is continuous on the interval [a, b], then we may dene the
distance traveled function between times a and t, where a t b, to be
s(t) =
_
t
a
[v(z)[ dz,
where z is just a dummy variable, which we introduce since t is now a limit of integration.
Then, the rst part of the Fundamental Theorem of Calculus, Theorem 2.4.7, tells us that
ds
dt
= [v(t)[.
Denition 3.3.5. The speed of an object, whose velocity function is continuous, is the
magnitude of the velocity or, equivalently, the rate of change of the distance traveled, with
respect to time.
3.3. DISTANCE TRAVELED IN SPACE AND ARC LENGTH 255
Remark 3.3.6. It is possible to weaken the assumptions that we made about our position and
velocity functions, and still calculate the distance traveled via integration.
Assuming that you dont believe that objects can teleport, then any moving object should
have a continuous position function. If p(t) is continuous, the distance traveled by an object
during times t such that a < t < b is the same as the distance traveled during times t such that
a t b.
Thus, to calculate the distance traveled, it is enough for p(t) to be continuous on the closed
interval [a, b], continuously dierentiable on the open interval (a, b), and for the integral of
[v(t)[ = [ p

(t)[ on the open interval (a, b), as dened in Denition 2.5.11, to exist. This distance
traveled is thus, again, simply
_
b
a
[v(t)[ dt,
provided that this integral exists.
We should mention that, while we assume, for physical reasons (non-teleportation), that
the position function is continuous, it is fairly common in physics and engineering to allow the
velocity of an object to be discontinuous.
For instance, a pitched baseball may be moving at 90 mph in one direction and then, when
its hit by a bat, moves in the other direction at around 110 mph. The change in velocity of the
baseball is 200 mph, and for most practical purposes, that changes occurs instantly at the time
when the ball is struck by the bat. Thus, we think of the velocity function as being discontinuous
at the time at which the bat strikes the ball.
Of course, what really happens is that the bat strikes the ball, and the ball deforms while
making contact with the bat for a very small amount of time. Hence, the velocity function of the
ball is not truly discontinuous. However, it is so dicult to analyze exactly what happens to the
balls velocity (or, more precisely, the velocity of the center of mass of the ball; see Section 3.8)
during the tiny time interval of contact with the bat that the situation is usually idealized and
described as an instantaneous change in velocity of 200 mph.
This discussion does not merely apply to baseballs; it applies often when one object strikes
another. If you jump into the air 1 foot, and land on the ground, then, just before you strike
the ground, your speed downward is approximately 8 ft/s. After you hit the ground, your speed
is zero. Once again, for most practical purposes, it is reasonable to say that you stop instantly.
If you think about it, if an objects velocity changes by some nite amount in zero time, then
the acceleration of the object must be innite, and so, by Newtons 2nd Law of Motion, the
force acting on the object must be innite. Thus, we typically say things like the bat strikes
256 CHAPTER 3. APPLICATIONS OF INTEGRATION
the baseball with an instantaneous innite force or the ground exerts an instantaneous innite
force on you when you land on it after jumping. Such instantaneous innite forces are usually
described via the Dirac delta function.
Whats really happening in these cases is that an extremely large force is acting in a com-
plicated manner over an extremely small period of time. We just idealize the situation to the
instantaneous, discontinuous, setting.
Example 3.3.7. Suppose that a particle moves in the xy-plane in such a way that its position
p(t), in feet, at time t seconds, where 1 t 1, is given by
p(t) =
_
t,
_
1 t
2
_
=
_
t, (1 t
2
)
1/2
_
.
Describe geometrically the path that the particle takes, and nd the distance that the particle
travels between t = 1 and t = 1 seconds.
Solution:
At time t seconds, x-coordinate of the particle is x = t, and the y-coordinate is y =

1 t
2
.
Thus, (x, y) is a point on the path that the particle moves along if and only if 1 x 1, y 0,
and y =

1 x
2
. Squaring both sides of this last equation yields y
2
= 1 x
2
, or x
2
+ y
2
= 1.
Therefore, the path of the particle lies on the circle of radius 1 foot, centered at the origin,
but only the part where y 0. Hence, the particle moves along the top half of the unit circle,
centered at the origin.
We nd the velocity
v(t) = p

(t) =
_
1,
1
2
_
1 t
2
_
1/2
(2t)
_
=
_
1,
t

1 t
2
_
ft/s,
for 1 < t < 1. Note that the velocity vector does not exist when t = 1.
In Figure 3.12, we have drawn the semicircle and indicated, in red, some velocity vectors.
Note that the lengths of the velocity vectors do not match the speed; we had to scale the
magnitude in order to t things in a diagram of reasonable size. What is important for you to
get from the diagram is the direction of the velocity, and that the magnitude of the velocity, the
speed, gets larger as the particle gets closer to the points (1, 0) and (1, 0), becoming innite
exactly at these endpoints of the semicircle.
3.3. DISTANCE TRAVELED IN SPACE AND ARC LENGTH 257
-1.5 -1 -0.5 0 0.5 1 1.5
-1
1
t=-1
t=0
t=1
t=0
Figure 3.12: Velocity vectors have innite magnitude when t = x = 1.
We know from high school geometry how far the particle travels; the length/circumference
of the top half of the unit circle is one half of circumference of the entire circle: (2 1)/2 =
feet. Lets make certain that the integral
_
1
1
[v(t)[ dt gives us the same answer.
First, we nd the speed, in ft/s:
[v(t)[ =

(1)
2
+
_
t

1 t
2
_
2
=
_
1 t
2
1 t
2
+
t
2
1 t
2
=
1

1 t
2
.
Thus, the distance traveled by the particle, between times t = 1 and t = 1 second, is given
by
_
1
1
[v(t)[ dt =
_
1
1
1

1 t
2
dt.
This is an improper integral (see Section 2.5) with two problem points: t = 1 and t = 1.
Hence, we split the integral, and calculate
_
1
1
1

1 t
2
dt =
_
0
1
1

1 t
2
dt +
_
1
0
1

1 t
2
dt =
lim
a1
+
_
0
a
1

1 t
2
dt + lim
b1

_
b
0
1

1 t
2
dt =
258 CHAPTER 3. APPLICATIONS OF INTEGRATION
lim
a1
+
_
sin
1
t

0
a
_
+ lim
b1

_
sin
1
t

b
0
_
=
_
0 lim
a1
+
sin
1
a
_
+
_
lim
b1

sin
1
b 0
_
=

2
_
+

2
= feet.
Whew! It sure took a while to verify that the length of a semicircle of radius 1 is .
Arc length:
The term length in this context is usually augmented and is referred to as arc length, in
order to emphasize that we mean the length along something which curves (i.e., along something
which is composed of arcs).
Example 3.3.8. If all we want is to use integration to verify that the arc length of a semicircle
of radius 1 is equal to , perhaps it would be better to think of a particle which moves around
the semicircle at a dierent speed from that of the particle in Example 3.3.7.
Suppose that a particle moves in the xy-plane in such a way that its position p(t), in feet,
at time t seconds, where 0 t , is given by
p(t) = (cos t, sin t).
Then, we see that the x- and y-coordinates of the particle are given by x = cos t and y = sin t,
so that x
2
+y
2
= (cos t)
2
+sin
2
t = 1. Hence, once again, the particle is always on the circle of
radius 1, centered at the origin. We also see that y = sin t 0 for 0 t , that p(0) = (1, 0),
and that p() = (1, 0). Finally, it is important to note that p(t) is one-to-one, i.e., the particle
is never at the same point at two (or more) dierent times.
3.3. DISTANCE TRAVELED IN SPACE AND ARC LENGTH 259
-1 0 1
-1
1
t=0
t=!/2
t=!
Figure 3.13: Moving around the semicircle with constant speed.
Thus, as in Example 3.3.7, the particle starts at (1, 0) and moves clockwise around the
semicircle of radius 1, centered at the origin, ending at (1, 0). However, the speed at which this
particle is moving is very dierent from the speed of the particle in Example 3.3.7.
We nd v(t) = p

(t) = (sin t, cos t), and so the speed is

[v(t)[ =
_
sin
2
t + cos
2
t = 1 ft/s.
Hence, we nd that our current particle moves with a constant speed of 1 ft/s.
Our integral for the distance that the particle travels, i.e., for the arc length of a semicircle
of radius 1 is now:
_

0
[v(t)[ dt =
_

0
1 dt = t

0
= feet.
Comparing this example with Example 3.3.7, we see that selecting/imagining a particle thats
moving in the right way can make the calculation of the arc length of a curve much easier.
One of the important things that you should have gotten out of the last two examples is
that, if you want the arc length of a curve, you can think of an object moving along the curve
and calculate the distance traveled by the object, provided that the object does not travel along
260 CHAPTER 3. APPLICATIONS OF INTEGRATION
any portions of the curve more than once; this means that the position function should be one-
to-one. We also saw that the calculation of the distance traveled can be made much easier by
making a nice choice of a position function.
In fact, it doesnt matter whether or not we actually think of an object moving along a given
curve. Whats important is that we have a one-to-one, continuously dierentiable function p(t)
from a closed interval [a, b], where a < b, into the xy-plane or into xyz-space, that plays the
role of a position function, in that the points on the curve that were interested in are precisely
the points that you get from p(t), i.e., the curve whose arc length we want is the range of the
function p. The term simple is frequently used in this context to indicate that p is one-to-one.
We also require, for now, a condition, regularity, which, in terms of motion, would say that
the object never stops at times in the open interval (a, b); this technical condition is very useful
mathematically.
Denition 3.3.9. A simple regular parameterization of a curve in R
n
(e.g., in R
2
,
the xy-plane, or in R
3
, xyz-space) is a one-to-one, continuously dierentiable function p
from a closed interval [a, b], where a < b, into R
n
, such that, for all t in [a, b], p

(t) ,=

0.
The range of a simple regular parameterization, that is, the set of points that p(t) passes
through, is called a simple regular curve.
We say that a simple regular parameterization p, with domain [a, b], parameterizes the
simple regular curve which is its range, and that the parameterization starts at the point
p(a) and ends at the point p(b).
Remark 3.3.10. We have dened a simple regular curve C as a set of points in R
2
or R
3
,
and you should have an intuitive idea of what a curve looks like, but how can we rigorously,
mathematically, say when a set of points matches your intuition for what a curve is?
What we have said in Denition 3.3.9 is that a rigorous denition of a simple regular curve is
that it is the range of a simple regular parametrization; but keep in mind that, as we saw in the
previous two examples, a simple regular curve, like a semicircle, can have more than one simple
regular parameterization. In fact, it is not dicult to show that any simple regular curve has
an innite number of dierent simple regular parameterizations. Nonetheless, as we saw in the
previous two examples, the curve itself has a length which is independent of the simple regular
parameterization, i.e.,
_
b
a
[ p

(t)[ dt will give the same number, regardless of what simple regular
parameterization you use for a given simple regular curve.
3.3. DISTANCE TRAVELED IN SPACE AND ARC LENGTH 261
Finally, we should mention that, in many references, a curve is dened to be the param-
eterization p, not the range of this function. This is mathematically convenient, but does
not agree with most peoples intuition for what a curve is, and so we will not use the
terminology.
What we have seen in our discussion and examples is that an innitesimal piece of arc length,
ds is equal to [ p

(t)[ dt, and so we have:

Proposition 3.3.11. Suppose that C is a simple regular curve in R
2
or in R
3
. Then, for
any simple regular parameterization p(t), with domain [a, b], of C, the arc length of C is
_
t=b
t=a
ds =
_
b
a
[ p

(t)[ dt.
Curves in the plane:
In our next examples of arc length calculations, we are going to look at simple regular curves
in R
2
, the xy-plane. In the plane, there are three common ways to specify simple regular
parameterizations.
One way is to specify appropriate coordinate (or component) functions x = x(t) and y = y(t),
and then the parameterization is, of course, p(t) = (x(t), y(t)). Then, Proposition 3.3.11 tells
us:
Proposition 3.3.12. If p(t) = (x(t), y(t)), for a t b, is a simple regular parameteriza-
tion of a curve C in R
2
, then
ds =

_
dx
dt
_
2
+
_
dy
dt
_
2
dt,
and
arc length of C =
_
b
a

_
dx
dt
_
2
+
_
dy
dt
_
2
dt.
262 CHAPTER 3. APPLICATIONS OF INTEGRATION
However, you are probably most familiar with curves described as the graph of y = f(x),
where a x b, and where f is continuously dierentiable. There is an obvious parame-
terization here: x = t and y = f(t), i.e., p(t) = (t, f(t)). This p is clearly one-to-one, since
(t
1
, f(t
1
)) = (t
2
, f(t
2
)) immediately implies that t
1
= t
2
. In addition, using that t = x, we have

_
dx
dt
_
2
+
_
dy
dt
_
2
=

_
dx
dx
_
2
+
_
df
dx
_
2
=

1 +
_
df
dx
_
2
> 0.
Therefore, if f is continuously dierentiable on an open interval containing [a, b], then p(t) =
(t, f(t)) is a simple regular parameterization.
Proposition 3.3.13. If y = f(x) is continuously dierentiable on an open interval con-
taining [a, b], then
ds =

1 +
_
dy
dx
_
2
dx
and the arc length of the graph of f, for a x b, is
_
b
a

1 +
_
dy
dx
_
2
dx.
While its less common, a curve may also be described by specifying x as a function of y,
instead of giving y as a function of x. Of course, all this does is interchange the roles of x and
y.
Proposition 3.3.14. If x = f(y) is continuously dierentiable on an open interval con-
taining [a, b], then
ds =

1 +
_
dx
dy
_
2
dy,
and the arc length of the graph of f, for a y b, is
_
b
a

1 +
_
dx
dy
_
2
dy.
3.3. DISTANCE TRAVELED IN SPACE AND ARC LENGTH 263
dx
dy
ds
Figure 3.14: Innitesimally, arc length and straight line distance are equal.
All three of the formulas for arc length above are frequently thought of as stemming from
an innitesimal version of the Pythagorean Theorem.
That is, if we perform algebra with the dierentials, assuming that dt, dx, and dy are positive,
then what we are integrating in Proposition 3.3.12, Proposition 3.3.13, and Proposition 3.3.14
is always
ds =
_
(dx)
2
+ (dy)
2
,
which is frequently written as
(ds)
2
= (dx)
2
+ (dy)
2
.
Before we give a couple of examples, we should mention that the anti-derivatives that arise
in calculating arc length are usually very dicult, or impossible, to nd (as elementary func-
tions). Such denite integrals can be approximated using methods such as Simpsons Rule
(Denition 2.6.5). The examples below are very special.
Example 3.3.15. Find the arc length of the graph of y =
2
3
x
3/2
, for 0 x 8.
Solution:
We use Proposition 3.3.13 and nd
arc length =
_
8
0

1 +
_
dy
dx
_
2
dx =
_
8
0
_
1 + (x
1/2
)
2
dx =
_
8
0

1 +x dx.
264 CHAPTER 3. APPLICATIONS OF INTEGRATION
-1 0 1 2 3 4 5 6 7 8
5
10
15
Figure 3.15: The graph of y = 2x
3/2
/3.
We make the substitution u = 1 +x, which tells us that du = dx, and that, as x goes from 0 to
8, u goes 1 + 0 = 1 to 1 + 8 = 9. Thus, we have that
arc length =
_
9
1
u
1/2
du =
u
3/2
3/2

9
1
=
2
3
_
9
3/2
1
3/2
_
=
52
3
.
Example 3.3.16. Find the arc length of the graph of x =
e
y
+e
y
2
= cosh(y), for 0 y ln 3.
-0.5 0 0.5 1 1.5 2
-0.5
0.5
1
Figure 3.16: The graph of x = (e
y
+e
y
)/2.
Solution:
3.3. DISTANCE TRAVELED IN SPACE AND ARC LENGTH 265
We use Proposition 3.3.14 and nd
arc length =
_
ln 3
0

1 +
_
dx
dy
_
2
dy =
_
ln 3
0

1 +
_
e
y
e
y
2
_
2
dy =
_
ln 3
0

4
4
+
(e
y
)
2
2 + (e
y
)
2
4
dy =
_
ln 3
0

(e
y
)
2
+ 2 + (e
y
)
2
4
dy =
_
ln 3
0

_
e
y
+e
y
2
_
2
dy =
_
ln 3
0
e
y
+e
y
2
dy =
1
2
_
e
y
e
y
_

ln 3
0
=
1
2
_
(3 3
1
) (1 1)

=
13
3
.
If some nite collection of simple regular curves intersect each other in a nite number of
points, then the arc length of all of the curves combined is simply the sum of the arc lengths of
the individual curves; this is because the overlap at a nite number of points does not have any
eect on the total arc length. We call the union of a nite number of simple regular curves, which
may intersect each other in a nite number of points, a piecewise-simple regular curve.
We can nd arc lengths of some piecewise-simple regular curves by using a parameterization
p(t) that is not one-to-one, but for which a nite number of points might be repeated for a nite
number of t values.
Example 3.3.17. Verify, using integration, that the arc length of a circle of radius r > 0 is
2r.
Solution:
Consider the parameterization of the circle of radius r, centered at the origin, given by
p(t) = (r cos t, r sin t) = r(cos t, sin t), for 0 t 2.
This is not a simple regular parameterization, since the function is not one-to-one: p(0) =
p(2) = (1, 0). However, this overlap at (1, 0) is the only problem; p(t) is continuously dier-
266 CHAPTER 3. APPLICATIONS OF INTEGRATION
entiable, its range is the circle in question, the only repeated point occurs when t = 0 and
t = 2, and
[ p

(t)[ = [r(sin t, cos t)[ = r [(sin t, cos t)[ = r

_
(sin t)
2
+ (cos t)
2
= r ,= 0.
Thus, we claim that the arc length is given by
_
2
0
[ p

(t)[ dt =
_
2
0
r dt = rt

2
0
= 2r,
even though the parameterization is not one-to-one.
Why is this calculation correct? Because we can look at the circle C in question as the union
of the top semicircle C
1
(including the endpoints at (r, 0) and (r, 0)) and the bottom semicircle
C
2
(including the endpoints). Both C
1
and C
2
are simple regular curves, parameterized by the
restrictions of p to the intervals [0, ] and [, 2], respectively. Thus, C is piecewise-simple
regular curve and its arc length is the sum of the arc lengths of C
1
and C
2
, i.e.,
_

0
[ p

(t)[ dt +
_
2

[ p

(t)[ dt =
_
2
0
[ p

(t)[ dt,
which is what we found to equal 2r.
Parameterizing by arc length:
Suppose that p : [a, b] R
n
is a simple regular parameterization of a curve C, which has
length L. Just as we dened the distance traveled function (immediately before Denition 3.3.5),
we can dene the arc length function l : [a, b] [0, L] by
s = l(t) =
_
t
a
[ p

(u)[ du;
thus, s = l(t) gives the arc length along C from the point p(a) to the point p(t).
As before, the rst part of the Fundamental Theorem of Calculus, Theorem 2.4.7, tells us
3.3. DISTANCE TRAVELED IN SPACE AND ARC LENGTH 267
that
l

(t) =
ds
dt
= [ p

(t)[.
As p is a regular parameterization, [ p

(t)[ > 0, and so the Inverse Function Theorem implies that

l has a monotonically increasing, dierentiable inverse function, i.e., there exists a dierentiable
function t = m(s) such that m(0) = a, m(L) = b, and l(m(s)) = s, for all s in the interval
[0, L], and m(l(t)) = t, for all t in the interval [a, b]. From the Chain Rule (or the formula for
the derivative of an inverse function), we have that
m

(s) =
1
l

(m(s))
=
1
[ p

(m(s))[
It is easy to check that the composed function q = p m : [0, L] R
n
is again a simple
regular parameterization of C, and has the property that q(s) is the unique point on the curve
C which is distance (along C) s away from q(0) = p(a).
Denition 3.3.18. A parameterization such as q, where the parameter measures arc length
along C, is said to be a parameterization by arc length.
Note that if q : [0, L] R
n
is a simple regular parameterization, parameterized by arc length,
then, applying the Chain Rule (to each component of the vector), we nd
[q

(s)[ = [( p m)

(s)[ = [m

(s) p

(m(s))[ = [m

(s)[ [ p

(m(s))[ =
1
[ p

(m(s))[
[ p

(m(s))[ = 1.
This, of course, gives us what it had better give us: if we use q to dene the arc length
function, we get
_
s
0
[q

(u)[ du =
_
s
0
1 du = u

s
0
= s,
which, in words, says that if you calculate the arc length between the starting point of a (pa-
rameterized) simple regular curve and the point thats the distance s away (along C) from the
starting point, then what you get is s. If that sounds stupidly obvious, GOOD.
One nal remark on parameterizing by arc length: if we return to the situation of an object
moving along a curve, with a simple regular position function p = p(t), where t is time, then
the parameter t is also the arc length (so that p(t) is a parameterization by arc length) if and
268 CHAPTER 3. APPLICATIONS OF INTEGRATION
only if, for all t, [ p

(t)[ = 1. Thus, a parameterization by arc length is also referred to as a

parameterization with unit speed, or speed 1.
More general parameterized curves:
If we combine parameterizations like those in Example 3.3.7 and in Example 3.3.17, we nd
that we can deal with curves that are the ranges of parameterizations which may fail to satisfy
various conditions at a nite number of points. We dont need for our parameterizations to have
closed intervals as their domains. We also dont require that p is continuously dierentiable
everywhere, and so p

need not be dened at some points, and even the improper integral (recall
Section 2.5) of [ p

(t)[ is not automatically guaranteed to exist.

All of these things make the denition below very technical. You need to recall our most
general notion of integrating over fairly general sets of points from Denition 2.5.11. The term
rectiable, used below, is a technical term which implies that a reasonable notion of arc length
exists for the curve in question.
Denition 3.3.19. A rectiable parameterization of a curve in R
n
is a continuous
function p from an interval I, whose interior is non-empty, into R
n
, such that there exists a
nite set of points F (which could be empty) in I such that, the restriction of p to the points
of I which are not in F (i.e., to the set I F) is one-to-one, continuously dierentiable, has
an everywhere non-zero derivative, and such that the (possibly) improper integral
_
I
[ p

(t)[ dt
exists.
The range of a rectiable parameterization is a rectiable, piecewise-regular, curve.
Of course, we have:
Proposition 3.3.20. Suppose that C is a rectiable, piecewise-regular, curve. Then, for
any rectiable parameterization p(t), with domain I, of C, the arc length of C is
_
I
[ p

(t)[ dt.
3.3. DISTANCE TRAVELED IN SPACE AND ARC LENGTH 269
3.3.1 Exercises
You are given the velocity function v(t) of a object in Exercises 1 - 6. Calculate the
displacement of the object over the given time interval.
1. v(t) = (3 sin 2t, 3 cos 2t, 4t), [0, 2].
2. v(t) = (t
2
+t, 4t
3
, 0), [0, 3].
3. v(t) = (3 + 4t, 5 7t, 11t), [3, 7].
4. v(t) = (15, t + 4, sinh t), [5, 8].
5. v(t) = ([t[, [t[, [t[), [3, 3].
6. v(t) = (t ln t, 1/t, e
t
), [1, 5].
Say whether the given map is a simple regular parameterization of a curve. If its
not, say why it is not.
7.

(t) = (2t
3
, 16, cos t), t [10, 10].
8. (t) = (cos t, [t + 2[, sin t), t [, 0].
9. u(t) = (cos t, [t + 2[, sin t), t [0, ].
10. w(t) = (1, t, t
2
), t [12, 20].
11.

h(t) = (t
3
, t
3
, t
3
), t [5, 5].
In each of Exercises 12 through 14, give an arc length parameterization of the given
curve.
12. (t) = (cos 5t, sin 5t), t [0, 2/5].
13. w(t) = (3 + 6t, 2 5t, 7 +t), t [0, 4].
14. z(t) =
_
t
3
/3, 0, 3 + (t
2
/2)
_
, t [1, 3].
15. Suppose : [a, b] R
n
is a regular curve parameterized by arc length. Show that the
length of the curve is b a.
16. A regular curve : [a, b] R
n
is closed if (a) = (b). Note that must still be
one-to-one everywhere else.
270 CHAPTER 3. APPLICATIONS OF INTEGRATION
a. If is a closed regular position function, show that the total displacement over the
interval is

0.
b. Give a simple example that shows that while the displacement is zero, the total
distance traveled may be non-zero.
You are given the position function of a particle. Calculate the total distance
traveled by the particle over the given time interval.
17. p(t) = (t, t, 3), t [0, /4].
18. p(t) = (t, 12, 0.5e
2t
), t [0, 1].
19. p(t) = (sin 3t, cos 3t, 2t
3/2
), t [0, ].
Calculate the arc length of the graph of the function over the interval I.
20. f(x) = ax +b, a ,= 0, I = [c, d].
21. g(x) = ln(sin x), I = [/4, /2].
22. f(y) = e
ay
, a > 0, I = [0, 1].
23. h(y) = 5 ln y, I = [1, 5].
24. f(x) =
x
3
12
+
1
x
. I = [1, 2].
25. Generalize the previous problem. Let a > 0 and f(x) =
x
3
6

a
+

a
2x
, I = [1, 2] and calculate
the arc length of the graph.
26. Suppose p(t) = (a cos t, a sin t, b cos t, b sin t), t [0, 2] is the position function of a particle
traveling in R
4
. What is the total distance traveled?
27. Parameterize the curve in the previous example by arc length.
28. What is the range of the curve (t
3
, t
3
), t R? Is this curve regular?
29. Is the path (t, [t[) regular? Is it rectiable on the in the interval [1, 1]?
30. The intersection of a sphere centered at the origin with radius R and the plane z =
z
0
, R < z
0
< R is a circle. Give a parameterization of this circle and calculate its
circumference in terms of z
0
and R.
31. Suppose you make a journey (on the Earth) consisting of three legs. Assume the Earth is
a sphere centered at the origin with radius R. You start at the North Pole and travel to a
point on the equator along a line of longitude. You then walk along the equator through
90 degrees of longitude. Finally, you travel back along the equator.
3.3. DISTANCE TRAVELED IN SPACE AND ARC LENGTH 271
a. Give a piecewise dierentiable parameterization for your journey.
b. What is the total distance traveled?
32. In Dierential Calculus, we dened the curvature, , function. If is a regular parame-
terization, then we set (t) =

(t)

(t)

where T(t) is the unit tangent vector. Prove that if

is parameterized by an arc length parameter s, then (s) =

dT
ds

.
33. Suppose C is the range of some simple regular curve : [a, b] R
3
. Suppose : [c, d] R
3
is another simple regular parameterization of C. Wed like to make sure that the arc length
of C is the same whether we use or .
a. Assume without loss of generality that (a) = (c) and (b) = (d). Let f : [a, b]
[c, d] be the function f =
1
. Let u = f(t) and show that
d
dt
=
d
du
du
dt
.
b. Carefully justify the equality:
_
b
a
[

(t)[ dt =
_
d
c
[

(u)[ du.
34. Assume the bottom of a tire with radius r is initially positioned at the origin. The bottom
of the tire is marked with a red piece of tape. As the tire rolls clockwise, the tape traces
out a cycloid. The gure below shows a cycloid with r = 2.
a. Let be the angle between the angle the tape makes with the vertical and let a()
be the vector from the center of the tire to the piece of tape. What is a()?
b. Let

b() be the vector from the origin to the center of the tire. What is

b()?
c. Using vector addition, determine the position function, p() of the piece of tape.
d. What is the distance traveled by the piece of tape from 0 2?
e. Show that the velocity vector vanishes when = 2k, k an integer.

Cycloid
35. What is the arc length of the cycloid between two consecutive cusps?
272 CHAPTER 3. APPLICATIONS OF INTEGRATION
36. Suppose that an outer circle orbits clockwise along the outside of an inner circle. This is
a common gear conguration. As in the previous problem, suppose a point on the outer
circle is marked in red. The path traced about by the mark is called an epicycloid. Suppose
the inner and outer circles have radii r
1
and r
2
, respectively and that the outer circle is
initially positioned on top of the inner circle with the marker touching the inner circe.
a. Let be the angle through which the outer circle rotates, and the angle between the
y-axis and the vector connecting the centers of the two circles, measured clockwise.
Show that the position vector of the center of the outer circle is given by a() =
((r
1
+r
2
) sin , (r
1
+r
2
) cos )).
b. Show that the vector from the center of the outer circle to the red marker is

b() =
(r
2
sin
(r
1
+r
2
)
r
1
, r
2
cos
(r
1
+r
2
)
r
1
).
c. Justify the equation r
1
= r
2
.
d. Conclude that the position function of the red mark is given by
p() =
_
(r
1
+r
2
) sin(
r
2

r
1
) r
2
sin
(r
1
+r
2
)
r
1
, (r
1
+r
2
) cos(
r
2

r
1
) r
2
cos
(r
1
+r
2
)
r
1
_
.

Epicycloid
37. Calculate the velocity vector of the epicycloid in the previous problem.
38. A curve c(t) with domain (, ) is periodic if there exists a constant t
0
such that
c(t) = c(t t
0
). An epicycloid may fail to be periodic, as suggested by the picture below.
3.3. DISTANCE TRAVELED IN SPACE AND ARC LENGTH 273
Determine a necessary and sucient condition for an epicycloid to be periodic. Hint:
consider the relationship between r
1
and r
2
.
Epicycloid II
39. Suppose c() is a parameterization of an epicycloid with innitely many cusps and that
r
1
= 1. Prove that given > 0, there exists
0
such that c(
0
) is a cusp and the arc length
between the points p = (1, 0) and c(
0
) is less than . Note: theres nothing special about
the point (1, 0). Given an arbitrary point p on the inner circle, the above argument holds.
We say the cusps are dense on the inner circle.
40. Let c(t) = (t, t sin(1/t)) for t ,= 0 and let c(0) = 0.
a. Show that c is everywhere continuous.
b. Show that c is not a rectiable parameterization.
41. If a small circle of radius r
2
revolves around the inside of a larger circle with radius r
1
,
the curve dened by the path of a x point on the inner cycloid is a hypocycloid. A
parameterization of a hypocycloid is given below.
x() = (r
1
r
2
) cos +r
2
cos
_
(r
1
r
2
)
r
2
_
y() = (r
1
r
2
) sin r
2
sin
_
(r
1
r
2
)
r
2
_
Calculate x

() and y

().
274 CHAPTER 3. APPLICATIONS OF INTEGRATION
42. Show that in a hypocycloid where the ratio of r
1
to r
2
is 4 to 1, the coordinates satisfy
x
2/3
+y
2/3
= C, where C is some constant.
43. Students not familiar with the dot product should skip this problem. Suppose
p and q are two distinct points in R
3
. Wed like to conrm our intuition that the straight
line segment connecting these two points is the shortest curve among those with endpoints
p and q. More precisely, well show that any curve between these two points must be at
least as long as the straight line segment between them.
a. Let c : [a, b] R
3
be any simple regular curve such that c(a) = p and c(b) = q. Let
v be any unit vector (so, [v[ = 1). Prove that (q p) v =
_
b
a
c

(t) v dt.
b. Prove that
_
b
a
c

(t) v dt
_
b
a
[c /(t)[ dt. Hint: one way to do this is to use the fact
that x y = [x[[y[ cos .
c. Let v =
p q
[p q[
and conclude that
[p q[
_
b
a
[c

(t)[ dt.
In other words, the length of an arbitrary curve from p to q is at least as long as the
straight line segment between p and q.
44. A curve may be parametrized for unbounded time, and yet still have nite length. Let
p(t) = (ae
bt
cos t, ae
bt
sin t) where a, b > 0 and t R. The range of this parameterization
is called the logarithmic spiral.
a. Calculate [ p

(t)[.
b. What is lim
t
_
t
0
[ p

(u)[ du?
3.3. DISTANCE TRAVELED IN SPACE AND ARC LENGTH 275
x
Logarithmic Spiral
45. Its often more convenient to parameterize a curve in polar coordinates rather than Carte-
sian coordinates. Recall that the relationship between the two coordinate systems is given
by the equations:
x = r cos
y = r sin .
a. Suppose c() = (x(), y()) = (r() cos , r() sin ). This means the distance from a
point on the curve to the origin, r, is a function of . Calculate dx/d and dy/d via
the chain and product rules.